diff --git a/CMakeLists.txt b/CMakeLists.txt index ceaac9a78..694c628b7 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -83,6 +83,7 @@ CMAKE_DEPENDENT_OPTION(ENABLE_SOFTWARE_RENDERER "Enables the software renderer" CMAKE_DEPENDENT_OPTION(ENABLE_OPENGL "Enables the OpenGL renderer" ON "NOT APPLE" OFF) option(ENABLE_VULKAN "Enables the Vulkan renderer" ON) +option(ENABLE_PROFILING "Enables integration with the Tracy profiler" ON) option(USE_DISCORD_PRESENCE "Enables Discord Rich Presence" OFF) CMAKE_DEPENDENT_OPTION(CITRA_ENABLE_BUNDLE_TARGET "Enable the distribution bundling target." ON "NOT ANDROID AND NOT IOS" OFF) diff --git a/externals/CMakeLists.txt b/externals/CMakeLists.txt index 878528e68..635d8f97e 100644 --- a/externals/CMakeLists.txt +++ b/externals/CMakeLists.txt @@ -129,10 +129,6 @@ else() add_subdirectory(inih) endif() -# MicroProfile -add_library(microprofile INTERFACE) -target_include_directories(microprofile SYSTEM INTERFACE ./microprofile) - # Nihstro add_library(nihstro-headers INTERFACE) target_include_directories(nihstro-headers SYSTEM INTERFACE ./nihstro/include) @@ -401,3 +397,8 @@ if (ENABLE_VULKAN) add_subdirectory(libadrenotools) endif() endif() + +# Tracy +if (ENABLE_PROFILING) + add_subdirectory(tracy) +endif() diff --git a/externals/microprofile/README.md b/externals/microprofile/README.md deleted file mode 100644 index 0a58d1c5a..000000000 --- a/externals/microprofile/README.md +++ /dev/null @@ -1,7 +0,0 @@ -# microprofile - -MicroProfile is a embeddable profiler in a single file, written in C++ - -It can display profile information in the application, or by generating captures via a minimal built in webserver. - -For more information see the project webpage at https://bitbucket.org/jonasmeyer/microprofile diff --git a/externals/microprofile/microprofile.h b/externals/microprofile/microprofile.h deleted file mode 100644 index 3db1e15f7..000000000 --- a/externals/microprofile/microprofile.h +++ /dev/null @@ -1,3597 +0,0 @@ -#pragma once -// This is free and unencumbered software released into the public domain. -// Anyone is free to copy, modify, publish, use, compile, sell, or -// distribute this software, either in source code form or as a compiled -// binary, for any purpose, commercial or non-commercial, and by any -// means. -// In jurisdictions that recognize copyright laws, the author or authors -// of this software dedicate any and all copyright interest in the -// software to the public domain. We make this dedication for the benefit -// of the public at large and to the detriment of our heirs and -// successors. We intend this dedication to be an overt act of -// relinquishment in perpetuity of all present and future rights to this -// software under copyright law. -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -// IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR -// OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, -// ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR -// OTHER DEALINGS IN THE SOFTWARE. -// For more information, please refer to -// -// *********************************************************************** -// -// -// -// -// Howto: -// Call these functions from your code: -// MicroProfileOnThreadCreate -// MicroProfileMouseButton -// MicroProfileMousePosition -// MicroProfileModKey -// MicroProfileFlip <-- Call this once per frame -// MicroProfileDraw <-- Call this once per frame -// MicroProfileToggleDisplayMode <-- Bind to a key to toggle profiling -// MicroProfileTogglePause <-- Bind to a key to toggle pause -// -// Use these macros in your code in blocks you want to time: -// -// MICROPROFILE_DECLARE -// MICROPROFILE_DEFINE -// MICROPROFILE_DECLARE_GPU -// MICROPROFILE_DEFINE_GPU -// MICROPROFILE_SCOPE -// MICROPROFILE_SCOPEI -// MICROPROFILE_SCOPEGPU -// MICROPROFILE_SCOPEGPUI -// MICROPROFILE_META -// -// -// Usage: -// -// { -// MICROPROFILE_SCOPEI("GroupName", "TimerName", nColorRgb): -// ..Code to be timed.. -// } -// -// MICROPROFILE_DECLARE / MICROPROFILE_DEFINE allows defining groups in a shared place, to ensure sorting of the timers -// -// (in global scope) -// MICROPROFILE_DEFINE(g_ProfileFisk, "Fisk", "Skalle", nSomeColorRgb); -// -// (in some other file) -// MICROPROFILE_DECLARE(g_ProfileFisk); -// -// void foo(){ -// MICROPROFILE_SCOPE(g_ProfileFisk); -// } -// -// Once code is instrumented the gui is activeted by calling MicroProfileToggleDisplayMode or by clicking in the upper left corner of -// the screen -// -// The following functions must be implemented before the profiler is usable -// debug render: -// void MicroProfileDrawText(int nX, int nY, uint32_t nColor, const char* pText, uint32_t nNumCharacters); -// void MicroProfileDrawBox(int nX, int nY, int nX1, int nY1, uint32_t nColor, MicroProfileBoxType = MicroProfileBoxTypeFlat); -// void MicroProfileDrawLine2D(uint32_t nVertices, float* pVertices, uint32_t nColor); -// Gpu time stamps: (See below for d3d/opengl helper) -// uint32_t MicroProfileGpuInsertTimeStamp(); -// uint64_t MicroProfileGpuGetTimeStamp(uint32_t nKey); -// uint64_t MicroProfileTicksPerSecondGpu(); -// threading: -// const char* MicroProfileGetThreadName(); Threadnames in detailed view -// -// Default implementations of Gpu timestamp functions: -// Opengl: -// in .c file where MICROPROFILE_IMPL is defined: -// #define MICROPROFILE_GPU_TIMERS_GL -// call MicroProfileGpuInitGL() on startup -// D3D11: -// in .c file where MICROPROFILE_IMPL is defined: -// #define MICROPROFILE_GPU_TIMERS_D3D11 -// call MICROPROFILE_GPU_TIMERS_D3D11(). Pass Device & ImmediateContext -// -// Limitations: -// GPU timestamps can only be inserted from one thread. - - - -#ifndef MICROPROFILE_ENABLED -#define MICROPROFILE_ENABLED 1 -#endif - -#include -typedef uint64_t MicroProfileToken; -typedef uint16_t MicroProfileGroupId; - -#if 0 == MICROPROFILE_ENABLED - -#define MICROPROFILE_DECLARE(var) -#define MICROPROFILE_DEFINE(var, group, name, color) -#define MICROPROFILE_REGISTER_GROUP(group, color, category) -#define MICROPROFILE_DECLARE_GPU(var) -#define MICROPROFILE_DEFINE_GPU(var, name, color) -#define MICROPROFILE_SCOPE(var) do{}while(0) -#define MICROPROFILE_SCOPEI(group, name, color) do{}while(0) -#define MICROPROFILE_SCOPEGPU(var) do{}while(0) -#define MICROPROFILE_SCOPEGPUI( name, color) do{}while(0) -#define MICROPROFILE_META_CPU(name, count) -#define MICROPROFILE_META_GPU(name, count) -#define MICROPROFILE_FORCEENABLECPUGROUP(s) do{} while(0) -#define MICROPROFILE_FORCEDISABLECPUGROUP(s) do{} while(0) -#define MICROPROFILE_FORCEENABLEGPUGROUP(s) do{} while(0) -#define MICROPROFILE_FORCEDISABLEGPUGROUP(s) do{} while(0) -#define MICROPROFILE_SCOPE_TOKEN(token) - -#define MicroProfileGetTime(group, name) 0.f -#define MicroProfileOnThreadCreate(foo) do{}while(0) -#define MicroProfileFlip() do{}while(0) -#define MicroProfileSetAggregateFrames(a) do{}while(0) -#define MicroProfileGetAggregateFrames() 0 -#define MicroProfileGetCurrentAggregateFrames() 0 -#define MicroProfileTogglePause() do{}while(0) -#define MicroProfileToggleAllGroups() do{} while(0) -#define MicroProfileDumpTimers() do{}while(0) -#define MicroProfileShutdown() do{}while(0) -#define MicroProfileSetForceEnable(a) do{} while(0) -#define MicroProfileGetForceEnable() false -#define MicroProfileSetEnableAllGroups(a) do{} while(0) -#define MicroProfileEnableCategory(a) do{} while(0) -#define MicroProfileDisableCategory(a) do{} while(0) -#define MicroProfileGetEnableAllGroups() false -#define MicroProfileSetForceMetaCounters(a) -#define MicroProfileGetForceMetaCounters() 0 -#define MicroProfileEnableMetaCounter(c) do{}while(0) -#define MicroProfileDisableMetaCounter(c) do{}while(0) -#define MicroProfileDumpFile(html,csv) do{} while(0) -#define MicroProfileWebServerPort() ((uint32_t)-1) - -#else - -#include -#include -#include -#include -#include -#include -#include - -#ifndef MICROPROFILE_API -#define MICROPROFILE_API -#endif - -MICROPROFILE_API int64_t MicroProfileTicksPerSecondCpu(); - - -#if defined(__APPLE__) -#include -#include -#include -#include -#include -#if TARGET_OS_IPHONE -#define MICROPROFILE_IOS -#endif - -#define MP_TICK() mach_absolute_time() -inline int64_t MicroProfileTicksPerSecondCpu() -{ - static int64_t nTicksPerSecond = 0; - if(nTicksPerSecond == 0) - { - mach_timebase_info_data_t sTimebaseInfo; - mach_timebase_info(&sTimebaseInfo); - nTicksPerSecond = 1000000000ll * sTimebaseInfo.denom / sTimebaseInfo.numer; - } - return nTicksPerSecond; -} -inline uint64_t MicroProfileGetCurrentThreadId() -{ - uint64_t tid; - pthread_threadid_np(pthread_self(), &tid); - return tid; -} - -#define MP_BREAK() __builtin_trap() -#define MP_THREAD_LOCAL __thread -#define MP_STRCASECMP strcasecmp -#define MP_GETCURRENTTHREADID() MicroProfileGetCurrentThreadId() -typedef uint64_t ThreadIdType; -#elif defined(_WIN32) -int64_t MicroProfileGetTick(); -#define MP_TICK() MicroProfileGetTick() -#define MP_BREAK() __debugbreak() -#define MP_THREAD_LOCAL thread_local -#define MP_STRCASECMP _stricmp -#define MP_GETCURRENTTHREADID() GetCurrentThreadId() -typedef uint32_t ThreadIdType; - -#elif !defined(_WIN32) -#include -#include -inline int64_t MicroProfileTicksPerSecondCpu() -{ - return 1000000000ll; -} - -inline int64_t MicroProfileGetTick() -{ - timespec ts; - clock_gettime(CLOCK_REALTIME, &ts); - return 1000000000ll * ts.tv_sec + ts.tv_nsec; -} -#define MP_TICK() MicroProfileGetTick() -#define MP_BREAK() __builtin_trap() -#define MP_THREAD_LOCAL __thread -#define MP_STRCASECMP strcasecmp -#define MP_GETCURRENTTHREADID() (uint64_t)pthread_self() -typedef uint64_t ThreadIdType; -#endif - - -#ifndef MP_GETCURRENTTHREADID -#define MP_GETCURRENTTHREADID() 0 -typedef uint32_t ThreadIdType; -#endif - - -#define MP_ASSERT(a) do{if(!(a)){MP_BREAK();} }while(0) -#define MICROPROFILE_DECLARE(var) extern MicroProfileToken g_mp_##var -#define MICROPROFILE_DEFINE(var, group, name, color) MicroProfileToken g_mp_##var = MicroProfileGetToken(group, name, color, MicroProfileTokenTypeCpu) -#define MICROPROFILE_REGISTER_GROUP(group, category, color) MicroProfileRegisterGroup(group, category, color) -#define MICROPROFILE_DECLARE_GPU(var) extern MicroProfileToken g_mp_##var -#define MICROPROFILE_DEFINE_GPU(var, name, color) MicroProfileToken g_mp_##var = MicroProfileGetToken("GPU", name, color, MicroProfileTokenTypeGpu) -#define MICROPROFILE_TOKEN_PASTE0(a, b) a ## b -#define MICROPROFILE_TOKEN_PASTE(a, b) MICROPROFILE_TOKEN_PASTE0(a,b) -#define MICROPROFILE_TOKEN(var) g_mp_##var -#define MICROPROFILE_SCOPE(var) MicroProfileScopeHandler MICROPROFILE_TOKEN_PASTE(foo, __LINE__)(g_mp_##var) -#define MICROPROFILE_SCOPE_TOKEN(token) MicroProfileScopeHandler MICROPROFILE_TOKEN_PASTE(foo, __LINE__)(token) -#define MICROPROFILE_SCOPEI(group, name, color) static MicroProfileToken MICROPROFILE_TOKEN_PASTE(g_mp,__LINE__) = MicroProfileGetToken(group, name, color, MicroProfileTokenTypeCpu); MicroProfileScopeHandler MICROPROFILE_TOKEN_PASTE(foo,__LINE__)( MICROPROFILE_TOKEN_PASTE(g_mp,__LINE__)) -#define MICROPROFILE_SCOPEGPU(var) MicroProfileScopeGpuHandler MICROPROFILE_TOKEN_PASTE(foo, __LINE__)(g_mp_##var) -#define MICROPROFILE_SCOPEGPUI(name, color) static MicroProfileToken MICROPROFILE_TOKEN_PASTE(g_mp,__LINE__) = MicroProfileGetToken("GPU", name, color, MicroProfileTokenTypeGpu); MicroProfileScopeGpuHandler MICROPROFILE_TOKEN_PASTE(foo,__LINE__)( MICROPROFILE_TOKEN_PASTE(g_mp,__LINE__)) -#define MICROPROFILE_META_CPU(name, count) static MicroProfileToken MICROPROFILE_TOKEN_PASTE(g_mp_meta,__LINE__) = MicroProfileGetMetaToken(name); MicroProfileMetaUpdate(MICROPROFILE_TOKEN_PASTE(g_mp_meta,__LINE__), count, MicroProfileTokenTypeCpu) -#define MICROPROFILE_META_GPU(name, count) static MicroProfileToken MICROPROFILE_TOKEN_PASTE(g_mp_meta,__LINE__) = MicroProfileGetMetaToken(name); MicroProfileMetaUpdate(MICROPROFILE_TOKEN_PASTE(g_mp_meta,__LINE__), count, MicroProfileTokenTypeGpu) - - -#ifndef MICROPROFILE_USE_THREAD_NAME_CALLBACK -#define MICROPROFILE_USE_THREAD_NAME_CALLBACK 0 -#endif - -#ifndef MICROPROFILE_PER_THREAD_BUFFER_SIZE -#define MICROPROFILE_PER_THREAD_BUFFER_SIZE (2048<<10) -#endif - -#ifndef MICROPROFILE_MAX_FRAME_HISTORY -#define MICROPROFILE_MAX_FRAME_HISTORY 512 -#endif - -#ifndef MICROPROFILE_PRINTF -#define MICROPROFILE_PRINTF printf -#endif - -#ifndef MICROPROFILE_META_MAX -#define MICROPROFILE_META_MAX 8 -#endif - -#ifndef MICROPROFILE_WEBSERVER_PORT -#define MICROPROFILE_WEBSERVER_PORT 1338 -#endif - -#ifndef MICROPROFILE_WEBSERVER -#define MICROPROFILE_WEBSERVER 1 -#endif - -#ifndef MICROPROFILE_WEBSERVER_MAXFRAMES -#define MICROPROFILE_WEBSERVER_MAXFRAMES 30 -#endif - -#ifndef MICROPROFILE_WEBSERVER_SOCKET_BUFFER_SIZE -#define MICROPROFILE_WEBSERVER_SOCKET_BUFFER_SIZE (16<<10) -#endif - -#ifndef MICROPROFILE_GPU_TIMERS -#define MICROPROFILE_GPU_TIMERS 1 -#endif - -#ifndef MICROPROFILE_GPU_FRAME_DELAY -#define MICROPROFILE_GPU_FRAME_DELAY 3 //must be > 0 -#endif - - -#ifndef MICROPROFILE_NAME_MAX_LEN -#define MICROPROFILE_NAME_MAX_LEN 64 -#endif - -#define MICROPROFILE_FORCEENABLECPUGROUP(s) MicroProfileForceEnableGroup(s, MicroProfileTokenTypeCpu) -#define MICROPROFILE_FORCEDISABLECPUGROUP(s) MicroProfileForceDisableGroup(s, MicroProfileTokenTypeCpu) -#define MICROPROFILE_FORCEENABLEGPUGROUP(s) MicroProfileForceEnableGroup(s, MicroProfileTokenTypeGpu) -#define MICROPROFILE_FORCEDISABLEGPUGROUP(s) MicroProfileForceDisableGroup(s, MicroProfileTokenTypeGpu) - -#define MICROPROFILE_INVALID_TICK ((uint64_t)-1) -#define MICROPROFILE_GROUP_MASK_ALL 0xffffffffffff - - -#define MICROPROFILE_INVALID_TOKEN (uint64_t)-1 - -enum MicroProfileTokenType -{ - MicroProfileTokenTypeCpu, - MicroProfileTokenTypeGpu, -}; - -enum MicroProfileBoxType -{ - MicroProfileBoxTypeBar, - MicroProfileBoxTypeFlat, -}; - - - -struct MicroProfile; - -MICROPROFILE_API void MicroProfileInit(); -MICROPROFILE_API void MicroProfileShutdown(); -MICROPROFILE_API MicroProfileToken MicroProfileFindToken(const char* sGroup, const char* sName); -MICROPROFILE_API MicroProfileToken MicroProfileGetToken(const char* sGroup, const char* sName, uint32_t nColor, MicroProfileTokenType Token = MicroProfileTokenTypeCpu); -MICROPROFILE_API MicroProfileToken MicroProfileGetMetaToken(const char* pName); -MICROPROFILE_API void MicroProfileMetaUpdate(MicroProfileToken, int nCount, MicroProfileTokenType eTokenType); -MICROPROFILE_API uint64_t MicroProfileEnter(MicroProfileToken nToken); -MICROPROFILE_API void MicroProfileLeave(MicroProfileToken nToken, uint64_t nTick); -MICROPROFILE_API uint64_t MicroProfileGpuEnter(MicroProfileToken nToken); -MICROPROFILE_API void MicroProfileGpuLeave(MicroProfileToken nToken, uint64_t nTick); -inline uint16_t MicroProfileGetTimerIndex(MicroProfileToken t){ return (t&0xffff); } -inline uint64_t MicroProfileGetGroupMask(MicroProfileToken t){ return ((t>>16)&MICROPROFILE_GROUP_MASK_ALL);} -inline MicroProfileToken MicroProfileMakeToken(uint64_t nGroupMask, uint16_t nTimer){ return (nGroupMask<<16) | nTimer;} - -MICROPROFILE_API void MicroProfileFlip(); //! call once per frame. -MICROPROFILE_API void MicroProfileTogglePause(); -MICROPROFILE_API void MicroProfileForceEnableGroup(const char* pGroup, MicroProfileTokenType Type); -MICROPROFILE_API void MicroProfileForceDisableGroup(const char* pGroup, MicroProfileTokenType Type); -MICROPROFILE_API float MicroProfileGetTime(const char* pGroup, const char* pName); -MICROPROFILE_API void MicroProfileContextSwitchSearch(uint32_t* pContextSwitchStart, uint32_t* pContextSwitchEnd, uint64_t nBaseTicksCpu, uint64_t nBaseTicksEndCpu); -MICROPROFILE_API void MicroProfileOnThreadCreate(const char* pThreadName); //should be called from newly created threads -MICROPROFILE_API void MicroProfileOnThreadExit(); //call on exit to reuse log -MICROPROFILE_API void MicroProfileInitThreadLog(); -MICROPROFILE_API void MicroProfileSetForceEnable(bool bForceEnable); -MICROPROFILE_API bool MicroProfileGetForceEnable(); -MICROPROFILE_API void MicroProfileSetEnableAllGroups(bool bEnable); -MICROPROFILE_API void MicroProfileEnableCategory(const char* pCategory); -MICROPROFILE_API void MicroProfileDisableCategory(const char* pCategory); -MICROPROFILE_API bool MicroProfileGetEnableAllGroups(); -MICROPROFILE_API void MicroProfileSetForceMetaCounters(bool bEnable); -MICROPROFILE_API bool MicroProfileGetForceMetaCounters(); -MICROPROFILE_API void MicroProfileEnableMetaCounter(const char* pMet); -MICROPROFILE_API void MicroProfileDisableMetaCounter(const char* pMet); -MICROPROFILE_API void MicroProfileSetAggregateFrames(int frames); -MICROPROFILE_API int MicroProfileGetAggregateFrames(); -MICROPROFILE_API int MicroProfileGetCurrentAggregateFrames(); -MICROPROFILE_API MicroProfile* MicroProfileGet(); -MICROPROFILE_API void MicroProfileGetRange(uint32_t nPut, uint32_t nGet, uint32_t nRange[2][2]); -MICROPROFILE_API std::recursive_mutex& MicroProfileGetMutex(); -MICROPROFILE_API void MicroProfileStartContextSwitchTrace(); -MICROPROFILE_API void MicroProfileStopContextSwitchTrace(); -MICROPROFILE_API bool MicroProfileIsLocalThread(uint32_t nThreadId); - - -#if MICROPROFILE_WEBSERVER -MICROPROFILE_API void MicroProfileDumpFile(const char* pHtml, const char* pCsv); -MICROPROFILE_API uint32_t MicroProfileWebServerPort(); -#else -#define MicroProfileDumpFile(c) do{} while(0) -#define MicroProfileWebServerPort() ((uint32_t)-1) -#endif - - - - -#if MICROPROFILE_GPU_TIMERS -MICROPROFILE_API uint32_t MicroProfileGpuInsertTimeStamp(); -MICROPROFILE_API uint64_t MicroProfileGpuGetTimeStamp(uint32_t nKey); -MICROPROFILE_API uint64_t MicroProfileTicksPerSecondGpu(); -MICROPROFILE_API int MicroProfileGetGpuTickReference(int64_t* pOutCPU, int64_t* pOutGpu); -#else -#define MicroProfileGpuInsertTimeStamp() 1 -#define MicroProfileGpuGetTimeStamp(a) 0 -#define MicroProfileTicksPerSecondGpu() 1 -#define MicroProfileGetGpuTickReference(a,b) 0 -#endif - -#if MICROPROFILE_GPU_TIMERS_D3D11 -#define MICROPROFILE_D3D_MAX_QUERIES (8<<10) -MICROPROFILE_API void MicroProfileGpuInitD3D11(void* pDevice, void* pDeviceContext); -#endif - -#if MICROPROFILE_GPU_TIMERS_GL -#define MICROPROFILE_GL_MAX_QUERIES (8<<10) -MICROPROFILE_API void MicroProfileGpuInitGL(); -#endif - - - -#if MICROPROFILE_USE_THREAD_NAME_CALLBACK -MICROPROFILE_API const char* MicroProfileGetThreadName(); -#else -#define MicroProfileGetThreadName() "" -#endif - -#if !defined(MICROPROFILE_THREAD_NAME_FROM_ID) -#define MICROPROFILE_THREAD_NAME_FROM_ID(a) "" -#endif - - -struct MicroProfileScopeHandler -{ - MicroProfileToken nToken; - uint64_t nTick; - MicroProfileScopeHandler(MicroProfileToken Token):nToken(Token) - { - nTick = MicroProfileEnter(nToken); - } - ~MicroProfileScopeHandler() - { - MicroProfileLeave(nToken, nTick); - } -}; - -struct MicroProfileScopeGpuHandler -{ - MicroProfileToken nToken; - uint64_t nTick; - MicroProfileScopeGpuHandler(MicroProfileToken Token):nToken(Token) - { - nTick = MicroProfileGpuEnter(nToken); - } - ~MicroProfileScopeGpuHandler() - { - MicroProfileGpuLeave(nToken, nTick); - } -}; - - - -#define MICROPROFILE_MAX_TIMERS 1024 -#define MICROPROFILE_MAX_GROUPS 48 //dont bump! no. of bits used it bitmask -#define MICROPROFILE_MAX_CATEGORIES 16 -#define MICROPROFILE_MAX_GRAPHS 5 -#define MICROPROFILE_GRAPH_HISTORY 128 -#define MICROPROFILE_BUFFER_SIZE ((MICROPROFILE_PER_THREAD_BUFFER_SIZE)/sizeof(MicroProfileLogEntry)) -#define MICROPROFILE_MAX_CONTEXT_SWITCH_THREADS 256 -#define MICROPROFILE_STACK_MAX 32 -//#define MICROPROFILE_MAX_PRESETS 5 -#define MICROPROFILE_ANIM_DELAY_PRC 0.5f -#define MICROPROFILE_GAP_TIME 50 //extra ms to fetch to close timers from earlier frames - - -#ifndef MICROPROFILE_MAX_THREADS -#define MICROPROFILE_MAX_THREADS 32 -#endif - -#ifndef MICROPROFILE_UNPACK_RED -#define MICROPROFILE_UNPACK_RED(c) ((c)>>16) -#endif - -#ifndef MICROPROFILE_UNPACK_GREEN -#define MICROPROFILE_UNPACK_GREEN(c) ((c)>>8) -#endif - -#ifndef MICROPROFILE_UNPACK_BLUE -#define MICROPROFILE_UNPACK_BLUE(c) ((c)) -#endif - -#ifndef MICROPROFILE_DEFAULT_PRESET -#define MICROPROFILE_DEFAULT_PRESET "Default" -#endif - - -#ifndef MICROPROFILE_CONTEXT_SWITCH_TRACE -#if defined(_WIN32) -#define MICROPROFILE_CONTEXT_SWITCH_TRACE 1 -#elif defined(__APPLE__) -#define MICROPROFILE_CONTEXT_SWITCH_TRACE 0 //disabled until dtrace script is working. -#else -#define MICROPROFILE_CONTEXT_SWITCH_TRACE 0 -#endif -#endif - -#if MICROPROFILE_CONTEXT_SWITCH_TRACE -#define MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE (128*1024) //2mb with 16 byte entry size -#else -#define MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE (1) -#endif - -#ifndef MICROPROFILE_MINIZ -#define MICROPROFILE_MINIZ 0 -#endif - -#ifdef _WIN32 -#include -typedef UINT_PTR MpSocket; -#else -typedef int MpSocket; -#endif - - -#ifndef _WIN32 -typedef pthread_t MicroProfileThread; -#elif defined(_MSC_VER) -typedef HANDLE MicroProfileThread; -#else -typedef std::thread* MicroProfileThread; -#endif - - - -enum MicroProfileDrawMask -{ - MP_DRAW_OFF = 0x0, - MP_DRAW_BARS = 0x1, - MP_DRAW_DETAILED = 0x2, - MP_DRAW_HIDDEN = 0x3, -}; - -enum MicroProfileDrawBarsMask -{ - MP_DRAW_TIMERS = 0x1, - MP_DRAW_AVERAGE = 0x2, - MP_DRAW_MAX = 0x4, - MP_DRAW_CALL_COUNT = 0x8, - MP_DRAW_TIMERS_EXCLUSIVE = 0x10, - MP_DRAW_AVERAGE_EXCLUSIVE = 0x20, - MP_DRAW_MAX_EXCLUSIVE = 0x40, - MP_DRAW_META_FIRST = 0x80, - MP_DRAW_ALL = 0xffffffff, - -}; - -typedef uint64_t MicroProfileLogEntry; - -struct MicroProfileTimer -{ - uint64_t nTicks; - uint32_t nCount; -}; - -struct MicroProfileCategory -{ - char pName[MICROPROFILE_NAME_MAX_LEN]; - uint64_t nGroupMask; -}; - -struct MicroProfileGroupInfo -{ - char pName[MICROPROFILE_NAME_MAX_LEN]; - uint32_t nNameLen; - uint32_t nGroupIndex; - uint32_t nNumTimers; - uint32_t nMaxTimerNameLen; - uint32_t nColor; - uint32_t nCategory; - MicroProfileTokenType Type; -}; - -struct MicroProfileTimerInfo -{ - MicroProfileToken nToken; - uint32_t nTimerIndex; - uint32_t nGroupIndex; - char pName[MICROPROFILE_NAME_MAX_LEN]; - uint32_t nNameLen; - uint32_t nColor; - bool bGraph; -}; - -struct MicroProfileGraphState -{ - int64_t nHistory[MICROPROFILE_GRAPH_HISTORY]; - MicroProfileToken nToken; - int32_t nKey; -}; - -struct MicroProfileContextSwitch -{ - ThreadIdType nThreadOut; - ThreadIdType nThreadIn; - int64_t nCpu : 8; - int64_t nTicks : 56; -}; - - -struct MicroProfileFrameState -{ - int64_t nFrameStartCpu; - int64_t nFrameStartGpu; - uint32_t nLogStart[MICROPROFILE_MAX_THREADS]; -}; - -struct MicroProfileThreadLog -{ - std::array Log{}; - - std::atomic nPut{0}; - std::atomic nGet{0}; - uint32_t nActive = 0; - uint32_t nGpu = 0; - ThreadIdType nThreadId{}; - - std::array nStack{}; - std::array nChildTickStack{}; - uint32_t nStackPos = 0; - - - std::array nGroupStackPos{}; - std::array nGroupTicks{}; - std::array nAggregateGroupTicks{}; - enum - { - THREAD_MAX_LEN = 64, - }; - char ThreadName[64]{}; - int nFreeListNext = 0; - - void Reset() { - Log.fill({}); - nPut = 0; - nGet = 0; - nActive = 0; - nGpu = 0; - nThreadId = {}; - nStack.fill(0); - nChildTickStack.fill(0); - nStackPos = 0; - nGroupStackPos.fill(0); - nGroupTicks.fill(0); - nAggregateGroupTicks.fill(0); - std::fill(std::begin(ThreadName), std::end(ThreadName), '\0'); - nFreeListNext = 0; - } -}; - -#if MICROPROFILE_GPU_TIMERS_D3D11 -struct MicroProfileD3D11Frame -{ - uint32_t m_nQueryStart; - uint32_t m_nQueryCount; - uint32_t m_nRateQueryStarted; - void* m_pRateQuery; -}; - -struct MicroProfileGpuTimerState -{ - uint32_t bInitialized; - void* m_pDevice; - void* m_pDeviceContext; - void* m_pQueries[MICROPROFILE_D3D_MAX_QUERIES]; - int64_t m_nQueryResults[MICROPROFILE_D3D_MAX_QUERIES]; - uint32_t m_nQueryPut; - uint32_t m_nQueryGet; - uint32_t m_nQueryFrame; - int64_t m_nQueryFrequency; - MicroProfileD3D11Frame m_QueryFrames[MICROPROFILE_GPU_FRAME_DELAY]; -}; -#elif MICROPROFILE_GPU_TIMERS_GL -struct MicroProfileGpuTimerState -{ - uint32_t GLTimers[MICROPROFILE_GL_MAX_QUERIES]; - uint32_t GLTimerPos; -}; -#else -struct MicroProfileGpuTimerState{}; -#endif - -struct MicroProfile -{ - uint32_t nTotalTimers; - uint32_t nGroupCount; - uint32_t nCategoryCount; - uint32_t nAggregateClear; - uint32_t nAggregateFlip; - uint32_t nAggregateFlipCount; - uint32_t nAggregateFrames; - - uint64_t nAggregateFlipTick; - - uint32_t nDisplay; - uint32_t nBars; - uint64_t nActiveGroup; - uint32_t nActiveBars; - - uint64_t nForceGroup; - uint32_t nForceEnable; - uint32_t nForceMetaCounters; - - uint64_t nForceGroupUI; - uint64_t nActiveGroupWanted; - uint32_t nAllGroupsWanted; - uint32_t nAllThreadsWanted; - - uint32_t nOverflow; - - uint64_t nGroupMask; - uint32_t nRunning; - uint32_t nToggleRunning; - uint32_t nMaxGroupSize; - uint32_t nDumpFileNextFrame; - uint32_t nAutoClearFrames; - char HtmlDumpPath[512]; - char CsvDumpPath[512]; - - int64_t nPauseTicks; - - float fReferenceTime; - float fRcpReferenceTime; - - MicroProfileCategory CategoryInfo[MICROPROFILE_MAX_CATEGORIES]; - MicroProfileGroupInfo GroupInfo[MICROPROFILE_MAX_GROUPS]; - MicroProfileTimerInfo TimerInfo[MICROPROFILE_MAX_TIMERS]; - uint8_t TimerToGroup[MICROPROFILE_MAX_TIMERS]; - - MicroProfileTimer AccumTimers[MICROPROFILE_MAX_TIMERS]; - uint64_t AccumMaxTimers[MICROPROFILE_MAX_TIMERS]; - uint64_t AccumTimersExclusive[MICROPROFILE_MAX_TIMERS]; - uint64_t AccumMaxTimersExclusive[MICROPROFILE_MAX_TIMERS]; - - MicroProfileTimer Frame[MICROPROFILE_MAX_TIMERS]; - uint64_t FrameExclusive[MICROPROFILE_MAX_TIMERS]; - - MicroProfileTimer Aggregate[MICROPROFILE_MAX_TIMERS]; - uint64_t AggregateMax[MICROPROFILE_MAX_TIMERS]; - uint64_t AggregateExclusive[MICROPROFILE_MAX_TIMERS]; - uint64_t AggregateMaxExclusive[MICROPROFILE_MAX_TIMERS]; - - - uint64_t FrameGroup[MICROPROFILE_MAX_GROUPS]; - uint64_t AccumGroup[MICROPROFILE_MAX_GROUPS]; - uint64_t AccumGroupMax[MICROPROFILE_MAX_GROUPS]; - - uint64_t AggregateGroup[MICROPROFILE_MAX_GROUPS]; - uint64_t AggregateGroupMax[MICROPROFILE_MAX_GROUPS]; - - - struct - { - uint64_t nCounters[MICROPROFILE_MAX_TIMERS]; - - uint64_t nAccum[MICROPROFILE_MAX_TIMERS]; - uint64_t nAccumMax[MICROPROFILE_MAX_TIMERS]; - - uint64_t nAggregate[MICROPROFILE_MAX_TIMERS]; - uint64_t nAggregateMax[MICROPROFILE_MAX_TIMERS]; - - uint64_t nSum; - uint64_t nSumAccum; - uint64_t nSumAccumMax; - uint64_t nSumAggregate; - uint64_t nSumAggregateMax; - - const char* pName; - } MetaCounters[MICROPROFILE_META_MAX]; - - MicroProfileGraphState Graph[MICROPROFILE_MAX_GRAPHS]; - uint32_t nGraphPut; - - uint32_t nThreadActive[MICROPROFILE_MAX_THREADS]; - MicroProfileThreadLog* Pool[MICROPROFILE_MAX_THREADS]; - uint32_t nNumLogs; - uint32_t nMemUsage; - int nFreeListHead; - - uint32_t nFrameCurrent; - uint32_t nFrameCurrentIndex; - uint32_t nFramePut; - uint64_t nFramePutIndex; - - MicroProfileFrameState Frames[MICROPROFILE_MAX_FRAME_HISTORY]; - - uint64_t nFlipTicks; - uint64_t nFlipAggregate; - uint64_t nFlipMax; - uint64_t nFlipAggregateDisplay; - uint64_t nFlipMaxDisplay; - - MicroProfileThread ContextSwitchThread; - bool bContextSwitchRunning; - bool bContextSwitchStop; - bool bContextSwitchAllThreads; - bool bContextSwitchNoBars; - uint32_t nContextSwitchUsage; - uint32_t nContextSwitchLastPut; - - int64_t nContextSwitchHoverTickIn; - int64_t nContextSwitchHoverTickOut; - uint32_t nContextSwitchHoverThread; - uint32_t nContextSwitchHoverThreadBefore; - uint32_t nContextSwitchHoverThreadAfter; - uint8_t nContextSwitchHoverCpu; - uint8_t nContextSwitchHoverCpuNext; - - uint32_t nContextSwitchPut; - MicroProfileContextSwitch ContextSwitch[MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE]; - - - MpSocket ListenerSocket; - uint32_t nWebServerPort; - - char WebServerBuffer[MICROPROFILE_WEBSERVER_SOCKET_BUFFER_SIZE]; - uint32_t WebServerPut; - - uint64_t nWebServerDataSent; - - MicroProfileGpuTimerState GPU; - - -}; - -#define MP_LOG_TICK_MASK 0x0000ffffffffffff -#define MP_LOG_INDEX_MASK 0x3fff000000000000 -#define MP_LOG_BEGIN_MASK 0xc000000000000000 -#define MP_LOG_GPU_EXTRA 0x3 -#define MP_LOG_META 0x2 -#define MP_LOG_ENTER 0x1 -#define MP_LOG_LEAVE 0x0 - - -inline int MicroProfileLogType(MicroProfileLogEntry Index) -{ - return (int)(((MP_LOG_BEGIN_MASK & Index)>>62) & 0x3ULL); -} - -inline uint64_t MicroProfileLogTimerIndex(MicroProfileLogEntry Index) -{ - return (0x3fff&(Index>>48)); -} - -inline MicroProfileLogEntry MicroProfileMakeLogIndex(uint64_t nBegin, MicroProfileToken nToken, int64_t nTick) -{ - MicroProfileLogEntry Entry = (nBegin<<62) | ((0x3fff&nToken)<<48) | (MP_LOG_TICK_MASK&nTick); - int t = MicroProfileLogType(Entry); - uint64_t nTimerIndex = MicroProfileLogTimerIndex(Entry); - MP_ASSERT(static_cast(t) == nBegin); - MP_ASSERT(nTimerIndex == (nToken&0x3fff)); - return Entry; - -} - -inline int64_t MicroProfileLogTickDifference(MicroProfileLogEntry Start, MicroProfileLogEntry End) -{ - uint64_t nStart = Start; - uint64_t nEnd = End; - int64_t nDifference = ((nEnd<<16) - (nStart<<16)); - return nDifference >> 16; -} - -inline int64_t MicroProfileLogGetTick(MicroProfileLogEntry e) -{ - return MP_LOG_TICK_MASK & e; -} - -inline int64_t MicroProfileLogSetTick(MicroProfileLogEntry e, int64_t nTick) -{ - return (MP_LOG_TICK_MASK & nTick) | (e & ~MP_LOG_TICK_MASK); -} - -template -T MicroProfileMin(T a, T b) -{ return a < b ? a : b; } - -template -T MicroProfileMax(T a, T b) -{ return a > b ? a : b; } - -inline int64_t MicroProfileMsToTick(float fMs, int64_t nTicksPerSecond) -{ - return (int64_t)(fMs*0.001f*(float)nTicksPerSecond); -} - -inline float MicroProfileTickToMsMultiplier(int64_t nTicksPerSecond) -{ - return 1000.f / (float)nTicksPerSecond; -} - -inline uint16_t MicroProfileGetGroupIndex(MicroProfileToken t) -{ - return (uint16_t)MicroProfileGet()->TimerToGroup[MicroProfileGetTimerIndex(t)]; -} - - - -#ifdef MICROPROFILE_IMPL - -#ifdef _WIN32 -#include -#define snprintf _snprintf - -#ifdef _MSC_VER -#pragma warning(push) -#pragma warning(disable: 4244) -#endif -int64_t MicroProfileTicksPerSecondCpu() -{ - static int64_t nTicksPerSecond = 0; - if(nTicksPerSecond == 0) - { - QueryPerformanceFrequency((LARGE_INTEGER*)&nTicksPerSecond); - } - return nTicksPerSecond; -} -int64_t MicroProfileGetTick() -{ - int64_t ticks; - QueryPerformanceCounter((LARGE_INTEGER*)&ticks); - return ticks; -} - -#endif - -#if defined(MICROPROFILE_WEBSERVER) || defined(MICROPROFILE_CONTEXT_SWITCH_TRACE) - - -typedef void* (*MicroProfileThreadFunc)(void*); - -#ifndef _WIN32 -typedef pthread_t MicroProfileThread; -inline void MicroProfileThreadStart(MicroProfileThread* pThread, MicroProfileThreadFunc Func) -{ - pthread_attr_t Attr; - int r = pthread_attr_init(&Attr); - MP_ASSERT(r == 0); - pthread_create(pThread, &Attr, Func, 0); -} -inline void MicroProfileThreadJoin(MicroProfileThread* pThread) -{ - int r = pthread_join(*pThread, 0); - MP_ASSERT(r == 0); -} -#elif defined(_MSC_VER) -typedef HANDLE MicroProfileThread; -DWORD _stdcall ThreadTrampoline(void* pFunc) -{ - MicroProfileThreadFunc F = (MicroProfileThreadFunc)pFunc; - - // The return value of F will always return a void*, however, this is for - // compatibility with pthreads. The underlying "address" of the pointer - // is always a 32-bit value, so this cast is safe to perform. - return static_cast(reinterpret_cast(F(0))); -} - -inline void MicroProfileThreadStart(MicroProfileThread* pThread, MicroProfileThreadFunc Func) -{ - *pThread = CreateThread(0, 0, ThreadTrampoline, Func, 0, 0); -} -inline void MicroProfileThreadJoin(MicroProfileThread* pThread) -{ - WaitForSingleObject(*pThread, INFINITE); - CloseHandle(*pThread); -} -#else -#include -typedef std::thread* MicroProfileThread; -inline void MicroProfileThreadStart(MicroProfileThread* pThread, MicroProfileThreadFunc Func) -{ - *pThread = new std::thread(Func, nullptr); -} -inline void MicroProfileThreadJoin(MicroProfileThread* pThread) -{ - (*pThread)->join(); - delete *pThread; -} -#endif -#endif - -#if MICROPROFILE_WEBSERVER - -#ifdef _WIN32 -#define MP_INVALID_SOCKET(f) (f == INVALID_SOCKET) -#endif - -#ifndef _WIN32 -#include -#include -#include -#define MP_INVALID_SOCKET(f) (f < 0) -#endif - - -void MicroProfileWebServerStart(); -void MicroProfileWebServerStop(); -bool MicroProfileWebServerUpdate(); -void MicroProfileDumpToFile(); - -#else - -#define MicroProfileWebServerStart() do{}while(0) -#define MicroProfileWebServerStop() do{}while(0) -#define MicroProfileWebServerUpdate() false -#define MicroProfileDumpToFile() do{} while(0) -#endif - - -#if MICROPROFILE_GPU_TIMERS_D3D11 -void MicroProfileGpuFlip(); -void MicroProfileGpuShutdown(); -#else -#define MicroProfileGpuFlip() do{}while(0) -#define MicroProfileGpuShutdown() do{}while(0) -#endif - - - -#include -#include -#include -#include - - -#ifndef MICROPROFILE_DEBUG -#define MICROPROFILE_DEBUG 0 -#endif - - -#define S g_MicroProfile - -MicroProfile g_MicroProfile; -MicroProfileThreadLog* g_MicroProfileGpuLog = 0; -#ifdef MICROPROFILE_IOS -// iOS doesn't support __thread -static pthread_key_t g_MicroProfileThreadLogKey; -static pthread_once_t g_MicroProfileThreadLogKeyOnce = PTHREAD_ONCE_INIT; -static void MicroProfileCreateThreadLogKey() -{ - pthread_key_create(&g_MicroProfileThreadLogKey, NULL); -} -#else -MP_THREAD_LOCAL MicroProfileThreadLog* g_MicroProfileThreadLog = 0; -#endif -static std::atomic g_bUseLock{false}; /// This is used because windows does not support using mutexes under dll init(which is where global initialization is handled) - - -MICROPROFILE_DEFINE(g_MicroProfileFlip, "MicroProfile", "MicroProfileFlip", 0x3355ee); -MICROPROFILE_DEFINE(g_MicroProfileThreadLoop, "MicroProfile", "ThreadLoop", 0x3355ee); -MICROPROFILE_DEFINE(g_MicroProfileClear, "MicroProfile", "Clear", 0x3355ee); -MICROPROFILE_DEFINE(g_MicroProfileAccumulate, "MicroProfile", "Accumulate", 0x3355ee); -MICROPROFILE_DEFINE(g_MicroProfileContextSwitchSearch,"MicroProfile", "ContextSwitchSearch", 0xDD7300); - -inline std::recursive_mutex& MicroProfileMutex() -{ - static std::recursive_mutex Mutex; - return Mutex; -} -std::recursive_mutex& MicroProfileGetMutex() -{ - return MicroProfileMutex(); -} - -MICROPROFILE_API MicroProfile* MicroProfileGet() -{ - return &g_MicroProfile; -} - - -MicroProfileThreadLog* MicroProfileCreateThreadLog(const char* pName); - - -void MicroProfileInit() -{ - std::recursive_mutex& mutex = MicroProfileMutex(); - bool bUseLock = g_bUseLock; - if(bUseLock) - mutex.lock(); - static bool bOnce = true; - if(bOnce) - { - S.nMemUsage += sizeof(S); - bOnce = false; - memset(&S, 0, sizeof(S)); - for(int i = 0; i < MICROPROFILE_MAX_GROUPS; ++i) - { - S.GroupInfo[i].pName[0] = '\0'; - } - for(int i = 0; i < MICROPROFILE_MAX_CATEGORIES; ++i) - { - S.CategoryInfo[i].pName[0] = '\0'; - S.CategoryInfo[i].nGroupMask = 0; - } - strcpy(&S.CategoryInfo[0].pName[0], "default"); - S.nCategoryCount = 1; - for(int i = 0; i < MICROPROFILE_MAX_TIMERS; ++i) - { - S.TimerInfo[i].pName[0] = '\0'; - } - S.nGroupCount = 0; - S.nAggregateFlipTick = MP_TICK(); - S.nActiveGroup = 0; - S.nActiveBars = 0; - S.nForceGroup = 0; - S.nAllGroupsWanted = 0; - S.nActiveGroupWanted = 0; - S.nAllThreadsWanted = 1; - S.nAggregateFlip = 0; - S.nTotalTimers = 0; - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - { - S.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN; - } - S.nRunning = 1; - S.fReferenceTime = 33.33f; - S.fRcpReferenceTime = 1.f / S.fReferenceTime; - S.nFreeListHead = -1; - int64_t nTick = MP_TICK(); - for(int i = 0; i < MICROPROFILE_MAX_FRAME_HISTORY; ++i) - { - S.Frames[i].nFrameStartCpu = nTick; - S.Frames[i].nFrameStartGpu = -1; - } - - MicroProfileThreadLog* pGpu = MicroProfileCreateThreadLog("GPU"); - g_MicroProfileGpuLog = pGpu; - MP_ASSERT(S.Pool[0] == pGpu); - pGpu->nGpu = 1; - pGpu->nThreadId = 0; - - S.nWebServerDataSent = (uint64_t)-1; - } - if(bUseLock) - mutex.unlock(); -} - -void MicroProfileShutdown() -{ - std::lock_guard Lock(MicroProfileMutex()); - MicroProfileWebServerStop(); - MicroProfileStopContextSwitchTrace(); - MicroProfileGpuShutdown(); -} - -#ifdef MICROPROFILE_IOS -inline MicroProfileThreadLog* MicroProfileGetThreadLog() -{ - pthread_once(&g_MicroProfileThreadLogKeyOnce, MicroProfileCreateThreadLogKey); - return (MicroProfileThreadLog*)pthread_getspecific(g_MicroProfileThreadLogKey); -} - -inline void MicroProfileSetThreadLog(MicroProfileThreadLog* pLog) -{ - pthread_once(&g_MicroProfileThreadLogKeyOnce, MicroProfileCreateThreadLogKey); - pthread_setspecific(g_MicroProfileThreadLogKey, pLog); -} -#else -inline MicroProfileThreadLog* MicroProfileGetThreadLog() -{ - return g_MicroProfileThreadLog; -} -inline void MicroProfileSetThreadLog(MicroProfileThreadLog* pLog) -{ - g_MicroProfileThreadLog = pLog; -} -#endif - - -MicroProfileThreadLog* MicroProfileCreateThreadLog(const char* pName) -{ - MicroProfileThreadLog* pLog = 0; - if(S.nFreeListHead != -1) - { - pLog = S.Pool[S.nFreeListHead]; - MP_ASSERT(pLog->nPut.load() == 0); - MP_ASSERT(pLog->nGet.load() == 0); - S.nFreeListHead = S.Pool[S.nFreeListHead]->nFreeListNext; - pLog->Reset(); - } - else - { - pLog = new MicroProfileThreadLog; - S.nMemUsage += sizeof(MicroProfileThreadLog); - S.Pool[S.nNumLogs++] = pLog; - } - int len = (int)strlen(pName); - int maxlen = sizeof(pLog->ThreadName)-1; - len = len < maxlen ? len : maxlen; - memcpy(&pLog->ThreadName[0], pName, len); - pLog->ThreadName[len] = '\0'; - pLog->nThreadId = MP_GETCURRENTTHREADID(); - pLog->nFreeListNext = -1; - pLog->nActive = 1; - return pLog; -} - -void MicroProfileOnThreadCreate(const char* pThreadName) -{ - g_bUseLock = true; - MicroProfileInit(); - std::lock_guard Lock(MicroProfileMutex()); - MP_ASSERT(MicroProfileGetThreadLog() == 0); - MicroProfileThreadLog* pLog = MicroProfileCreateThreadLog(pThreadName ? pThreadName : MicroProfileGetThreadName()); - MP_ASSERT(pLog); - MicroProfileSetThreadLog(pLog); -} - -void MicroProfileOnThreadExit() -{ - std::lock_guard Lock(MicroProfileMutex()); - MicroProfileThreadLog* pLog = MicroProfileGetThreadLog(); - if(pLog) - { - int32_t nLogIndex = -1; - for(int i = 0; i < MICROPROFILE_MAX_THREADS; ++i) - { - if(pLog == S.Pool[i]) - { - nLogIndex = i; - break; - } - } - MP_ASSERT(nLogIndex < MICROPROFILE_MAX_THREADS && nLogIndex > 0); - pLog->nFreeListNext = S.nFreeListHead; - pLog->nActive = 0; - pLog->nPut.store(0); - pLog->nGet.store(0); - S.nFreeListHead = nLogIndex; - for(int i = 0; i < MICROPROFILE_MAX_FRAME_HISTORY; ++i) - { - S.Frames[i].nLogStart[nLogIndex] = 0; - } - pLog->nGroupStackPos.fill(0); - pLog->nGroupTicks.fill(0); - } -} - -void MicroProfileInitThreadLog() -{ - MicroProfileOnThreadCreate(nullptr); -} - - -struct MicroProfileScopeLock -{ - bool bUseLock; - std::recursive_mutex& m; - MicroProfileScopeLock(std::recursive_mutex& m) : bUseLock(g_bUseLock), m(m) - { - if(bUseLock) - m.lock(); - } - ~MicroProfileScopeLock() - { - if(bUseLock) - m.unlock(); - } -}; - -MicroProfileToken MicroProfileFindToken(const char* pGroup, const char* pName) -{ - MicroProfileInit(); - MicroProfileScopeLock L(MicroProfileMutex()); - for(uint32_t i = 0; i < S.nTotalTimers; ++i) - { - if(!MP_STRCASECMP(pName, S.TimerInfo[i].pName) && !MP_STRCASECMP(pGroup, S.GroupInfo[S.TimerToGroup[i]].pName)) - { - return S.TimerInfo[i].nToken; - } - } - return MICROPROFILE_INVALID_TOKEN; -} - -inline uint16_t MicroProfileGetGroup(const char* pGroup, MicroProfileTokenType Type) -{ - for(uint32_t i = 0; i < S.nGroupCount; ++i) - { - if(!MP_STRCASECMP(pGroup, S.GroupInfo[i].pName)) - { - return i; - } - } - uint16_t nGroupIndex = 0xffff; - uint32_t nLen = (uint32_t)strlen(pGroup); - if(nLen > MICROPROFILE_NAME_MAX_LEN-1) - nLen = MICROPROFILE_NAME_MAX_LEN-1; - memcpy(&S.GroupInfo[S.nGroupCount].pName[0], pGroup, nLen); - S.GroupInfo[S.nGroupCount].pName[nLen] = '\0'; - S.GroupInfo[S.nGroupCount].nNameLen = nLen; - S.GroupInfo[S.nGroupCount].nNumTimers = 0; - S.GroupInfo[S.nGroupCount].nGroupIndex = S.nGroupCount; - S.GroupInfo[S.nGroupCount].Type = Type; - S.GroupInfo[S.nGroupCount].nMaxTimerNameLen = 0; - S.GroupInfo[S.nGroupCount].nColor = 0x88888888; - S.GroupInfo[S.nGroupCount].nCategory = 0; - S.CategoryInfo[0].nGroupMask |= (1ll << (uint64_t)S.nGroupCount); - nGroupIndex = S.nGroupCount++; - S.nGroupMask = (S.nGroupMask<<1)|1; - MP_ASSERT(nGroupIndex < MICROPROFILE_MAX_GROUPS); - return nGroupIndex; -} - -inline void MicroProfileRegisterGroup(const char* pGroup, const char* pCategory, uint32_t nColor) -{ - int nCategoryIndex = -1; - for(uint32_t i = 0; i < S.nCategoryCount; ++i) - { - if(!MP_STRCASECMP(pCategory, S.CategoryInfo[i].pName)) - { - nCategoryIndex = (int)i; - break; - } - } - if(-1 == nCategoryIndex && S.nCategoryCount < MICROPROFILE_MAX_CATEGORIES) - { - MP_ASSERT(S.CategoryInfo[S.nCategoryCount].pName[0] == '\0'); - nCategoryIndex = (int)S.nCategoryCount++; - uint32_t nLen = (uint32_t)strlen(pCategory); - if(nLen > MICROPROFILE_NAME_MAX_LEN-1) - nLen = MICROPROFILE_NAME_MAX_LEN-1; - memcpy(&S.CategoryInfo[nCategoryIndex].pName[0], pCategory, nLen); - S.CategoryInfo[nCategoryIndex].pName[nLen] = '\0'; - } - uint16_t nGroup = MicroProfileGetGroup(pGroup, 0 != MP_STRCASECMP(pGroup, "gpu")?MicroProfileTokenTypeCpu : MicroProfileTokenTypeGpu); - S.GroupInfo[nGroup].nColor = nColor; - if(nCategoryIndex >= 0) - { - uint64_t nBit = 1ll << nGroup; - uint32_t nOldCategory = S.GroupInfo[nGroup].nCategory; - S.CategoryInfo[nOldCategory].nGroupMask &= ~nBit; - S.CategoryInfo[nCategoryIndex].nGroupMask |= nBit; - S.GroupInfo[nGroup].nCategory = nCategoryIndex; - } -} - -MicroProfileToken MicroProfileGetToken(const char* pGroup, const char* pName, uint32_t nColor, MicroProfileTokenType Type) -{ - MicroProfileInit(); - MicroProfileScopeLock L(MicroProfileMutex()); - MicroProfileToken ret = MicroProfileFindToken(pGroup, pName); - if(ret != MICROPROFILE_INVALID_TOKEN) - return ret; - uint16_t nGroupIndex = MicroProfileGetGroup(pGroup, Type); - uint16_t nTimerIndex = (uint16_t)(S.nTotalTimers++); - uint64_t nGroupMask = 1ll << nGroupIndex; - MicroProfileToken nToken = MicroProfileMakeToken(nGroupMask, nTimerIndex); - S.GroupInfo[nGroupIndex].nNumTimers++; - S.GroupInfo[nGroupIndex].nMaxTimerNameLen = MicroProfileMax(S.GroupInfo[nGroupIndex].nMaxTimerNameLen, (uint32_t)strlen(pName)); - MP_ASSERT(S.GroupInfo[nGroupIndex].Type == Type); //dont mix cpu & gpu timers in the same group - S.nMaxGroupSize = MicroProfileMax(S.nMaxGroupSize, S.GroupInfo[nGroupIndex].nNumTimers); - S.TimerInfo[nTimerIndex].nToken = nToken; - uint32_t nLen = (uint32_t)strlen(pName); - if(nLen > MICROPROFILE_NAME_MAX_LEN-1) - nLen = MICROPROFILE_NAME_MAX_LEN-1; - memcpy(&S.TimerInfo[nTimerIndex].pName, pName, nLen); - S.TimerInfo[nTimerIndex].pName[nLen] = '\0'; - S.TimerInfo[nTimerIndex].nNameLen = nLen; - S.TimerInfo[nTimerIndex].nColor = nColor&0xffffff; - S.TimerInfo[nTimerIndex].nGroupIndex = nGroupIndex; - S.TimerInfo[nTimerIndex].nTimerIndex = nTimerIndex; - S.TimerToGroup[nTimerIndex] = nGroupIndex; - return nToken; -} - -MicroProfileToken MicroProfileGetMetaToken(const char* pName) -{ - MicroProfileInit(); - MicroProfileScopeLock L(MicroProfileMutex()); - for(uint32_t i = 0; i < MICROPROFILE_META_MAX; ++i) - { - if(!S.MetaCounters[i].pName) - { - S.MetaCounters[i].pName = pName; - return i; - } - else if(!MP_STRCASECMP(pName, S.MetaCounters[i].pName)) - { - return i; - } - } - MP_ASSERT(0);//out of slots, increase MICROPROFILE_META_MAX - return (MicroProfileToken)-1; -} - - -inline void MicroProfileLogPut(MicroProfileToken nToken_, uint64_t nTick, uint64_t nBegin, MicroProfileThreadLog* pLog) -{ - MP_ASSERT(pLog != 0); //this assert is hit if MicroProfileOnCreateThread is not called - MP_ASSERT(pLog->nActive); - uint32_t nPos = pLog->nPut.load(std::memory_order_relaxed); - uint32_t nNextPos = (nPos+1) % MICROPROFILE_BUFFER_SIZE; - if(nNextPos == pLog->nGet.load(std::memory_order_relaxed)) - { - S.nOverflow = 100; - } - else - { - pLog->Log[nPos] = MicroProfileMakeLogIndex(nBegin, nToken_, nTick); - pLog->nPut.store(nNextPos, std::memory_order_release); - } -} - -uint64_t MicroProfileEnter(MicroProfileToken nToken_) -{ - if(MicroProfileGetGroupMask(nToken_) & S.nActiveGroup) - { - if(!MicroProfileGetThreadLog()) - { - MicroProfileInitThreadLog(); - } - uint64_t nTick = MP_TICK(); - MicroProfileLogPut(nToken_, nTick, MP_LOG_ENTER, MicroProfileGetThreadLog()); - return nTick; - } - return MICROPROFILE_INVALID_TICK; -} - -void MicroProfileMetaUpdate(MicroProfileToken nToken, int nCount, MicroProfileTokenType eTokenType) -{ - if((MP_DRAW_META_FIRST<nTicks <= S.nPauseTicks) - { - uint32_t nPut = S.nContextSwitchPut; - S.ContextSwitch[nPut] = *pContextSwitch; - S.nContextSwitchPut = (S.nContextSwitchPut+1) % MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE; - } -} - - -void MicroProfileGetRange(uint32_t nPut, uint32_t nGet, uint32_t nRange[2][2]) -{ - if(nPut > nGet) - { - nRange[0][0] = nGet; - nRange[0][1] = nPut; - nRange[1][0] = nRange[1][1] = 0; - } - else if(nPut != nGet) - { - MP_ASSERT(nGet != MICROPROFILE_BUFFER_SIZE); - uint32_t nCountEnd = MICROPROFILE_BUFFER_SIZE - nGet; - nRange[0][0] = nGet; - nRange[0][1] = nGet + nCountEnd; - nRange[1][0] = 0; - nRange[1][1] = nPut; - } -} - -void MicroProfileFlip() -{ - #if 0 - //verify LogEntry wraps correctly - MicroProfileLogEntry c = MP_LOG_TICK_MASK-5000; - for(int i = 0; i < 10000; ++i, c += 1) - { - MicroProfileLogEntry l2 = (c+2500) & MP_LOG_TICK_MASK; - MP_ASSERT(2500 == MicroProfileLogTickDifference(c, l2)); - } - #endif - MICROPROFILE_SCOPE(g_MicroProfileFlip); - std::lock_guard Lock(MicroProfileMutex()); - - - MicroProfileGpuFlip(); - - if(S.nToggleRunning) - { - S.nRunning = !S.nRunning; - if(!S.nRunning) - S.nPauseTicks = MP_TICK(); - S.nToggleRunning = 0; - for(uint32_t i = 0; i < MICROPROFILE_MAX_THREADS; ++i) - { - MicroProfileThreadLog* pLog = S.Pool[i]; - if(pLog) - { - pLog->nStackPos = 0; - } - } - } - uint32_t nAggregateClear = S.nAggregateClear || S.nAutoClearFrames, nAggregateFlip = 0; - if(S.nDumpFileNextFrame) - { - MicroProfileDumpToFile(); - S.nDumpFileNextFrame = 0; - S.nAutoClearFrames = MICROPROFILE_GPU_FRAME_DELAY + 3; //hide spike from dumping webpage - } - if(S.nWebServerDataSent == (uint64_t)-1) - { - MicroProfileWebServerStart(); - S.nWebServerDataSent = 0; - } - - if(MicroProfileWebServerUpdate()) - { - S.nAutoClearFrames = MICROPROFILE_GPU_FRAME_DELAY + 3; //hide spike from dumping webpage - } - - if(S.nAutoClearFrames) - { - nAggregateClear = 1; - nAggregateFlip = 1; - S.nAutoClearFrames -= 1; - } - - - if(S.nRunning || S.nForceEnable) - { - S.nFramePutIndex++; - S.nFramePut = (S.nFramePut+1) % MICROPROFILE_MAX_FRAME_HISTORY; - MP_ASSERT((S.nFramePutIndex % MICROPROFILE_MAX_FRAME_HISTORY) == S.nFramePut); - S.nFrameCurrent = (S.nFramePut + MICROPROFILE_MAX_FRAME_HISTORY - MICROPROFILE_GPU_FRAME_DELAY - 1) % MICROPROFILE_MAX_FRAME_HISTORY; - S.nFrameCurrentIndex++; - uint32_t nFrameNext = (S.nFrameCurrent+1) % MICROPROFILE_MAX_FRAME_HISTORY; - - uint32_t nContextSwitchPut = S.nContextSwitchPut; - if(S.nContextSwitchLastPut < nContextSwitchPut) - { - S.nContextSwitchUsage = (nContextSwitchPut - S.nContextSwitchLastPut); - } - else - { - S.nContextSwitchUsage = MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE - S.nContextSwitchLastPut + nContextSwitchPut; - } - S.nContextSwitchLastPut = nContextSwitchPut; - - MicroProfileFrameState* pFramePut = &S.Frames[S.nFramePut]; - MicroProfileFrameState* pFrameCurrent = &S.Frames[S.nFrameCurrent]; - MicroProfileFrameState* pFrameNext = &S.Frames[nFrameNext]; - - pFramePut->nFrameStartCpu = MP_TICK(); - pFramePut->nFrameStartGpu = (uint32_t)MicroProfileGpuInsertTimeStamp(); - if(static_cast(pFrameNext->nFrameStartGpu) != (uint64_t)-1) - pFrameNext->nFrameStartGpu = MicroProfileGpuGetTimeStamp((uint32_t)pFrameNext->nFrameStartGpu); - - if(static_cast(pFrameCurrent->nFrameStartGpu) == (uint64_t)-1) - pFrameCurrent->nFrameStartGpu = pFrameNext->nFrameStartGpu + 1; - - uint64_t nFrameStartCpu = pFrameCurrent->nFrameStartCpu; - uint64_t nFrameEndCpu = pFrameNext->nFrameStartCpu; - - { - uint64_t nTick = nFrameEndCpu - nFrameStartCpu; - S.nFlipTicks = nTick; - S.nFlipAggregate += nTick; - S.nFlipMax = MicroProfileMax(S.nFlipMax, nTick); - } - - uint8_t* pTimerToGroup = &S.TimerToGroup[0]; - for(uint32_t i = 0; i < MICROPROFILE_MAX_THREADS; ++i) - { - MicroProfileThreadLog* pLog = S.Pool[i]; - if(!pLog) - { - pFramePut->nLogStart[i] = 0; - } - else - { - uint32_t nPut = pLog->nPut.load(std::memory_order_acquire); - pFramePut->nLogStart[i] = nPut; - MP_ASSERT(nPut< MICROPROFILE_BUFFER_SIZE); - //need to keep last frame around to close timers. timers more than 1 frame old is ditched. - pLog->nGet.store(nPut, std::memory_order_relaxed); - } - } - - if(S.nRunning) - { - uint64_t* pFrameGroup = &S.FrameGroup[0]; - { - MICROPROFILE_SCOPE(g_MicroProfileClear); - for(uint32_t i = 0; i < S.nTotalTimers; ++i) - { - S.Frame[i].nTicks = 0; - S.Frame[i].nCount = 0; - S.FrameExclusive[i] = 0; - } - for(uint32_t i = 0; i < MICROPROFILE_MAX_GROUPS; ++i) - { - pFrameGroup[i] = 0; - } - for(uint32_t j = 0; j < MICROPROFILE_META_MAX; ++j) - { - if(S.MetaCounters[j].pName && 0 != (S.nActiveBars & (MP_DRAW_META_FIRST<nGroupStackPos[0]; - int64_t nGroupTicks[MICROPROFILE_MAX_GROUPS] = {0}; - - - uint32_t nPut = pFrameNext->nLogStart[i]; - uint32_t nGet = pFrameCurrent->nLogStart[i]; - uint32_t nRange[2][2] = { {0, 0}, {0, 0}, }; - MicroProfileGetRange(nPut, nGet, nRange); - - - //fetch gpu results. - if(pLog->nGpu) - { - for(uint32_t j = 0; j < 2; ++j) - { - uint32_t nStart = nRange[j][0]; - uint32_t nEnd = nRange[j][1]; - for(uint32_t k = nStart; k < nEnd; ++k) - { - MicroProfileLogEntry L = pLog->Log[k]; - if(MicroProfileLogType(L) < MP_LOG_META) - { - pLog->Log[k] = MicroProfileLogSetTick(L, MicroProfileGpuGetTimeStamp((uint32_t)MicroProfileLogGetTick(L))); - } - } - } - } - - - uint32_t* pStack = &pLog->nStack[0]; - int64_t* pChildTickStack = &pLog->nChildTickStack[0]; - uint32_t nStackPos = pLog->nStackPos; - - for(uint32_t j = 0; j < 2; ++j) - { - uint32_t nStart = nRange[j][0]; - uint32_t nEnd = nRange[j][1]; - for(uint32_t k = nStart; k < nEnd; ++k) - { - MicroProfileLogEntry LE = pLog->Log[k]; - int nType = MicroProfileLogType(LE); - - if(MP_LOG_ENTER == nType) - { - int nTimer = MicroProfileLogTimerIndex(LE); - uint8_t nGroup = pTimerToGroup[nTimer]; - MP_ASSERT(nStackPos < MICROPROFILE_STACK_MAX); - MP_ASSERT(nGroup < MICROPROFILE_MAX_GROUPS); - pGroupStackPos[nGroup]++; - pStack[nStackPos++] = k; - pChildTickStack[nStackPos] = 0; - - } - else if(MP_LOG_META == nType) - { - if(nStackPos) - { - int64_t nMetaIndex = MicroProfileLogTimerIndex(LE); - int64_t nMetaCount = MicroProfileLogGetTick(LE); - MP_ASSERT(nMetaIndex < MICROPROFILE_META_MAX); - int64_t nCounter = MicroProfileLogTimerIndex(pLog->Log[pStack[nStackPos-1]]); - S.MetaCounters[nMetaIndex].nCounters[nCounter] += nMetaCount; - } - } - else if(MP_LOG_LEAVE == nType) - { - int nTimer = MicroProfileLogTimerIndex(LE); - uint8_t nGroup = pTimerToGroup[nTimer]; - MP_ASSERT(nGroup < MICROPROFILE_MAX_GROUPS); - if(nStackPos) - { - int64_t nTickStart = pLog->Log[pStack[nStackPos-1]]; - int64_t nTicks = MicroProfileLogTickDifference(nTickStart, LE); - int64_t nChildTicks = pChildTickStack[nStackPos]; - nStackPos--; - pChildTickStack[nStackPos] += nTicks; - - uint32_t nTimerIndex = MicroProfileLogTimerIndex(LE); - S.Frame[nTimerIndex].nTicks += nTicks; - S.FrameExclusive[nTimerIndex] += (nTicks-nChildTicks); - S.Frame[nTimerIndex].nCount += 1; - - MP_ASSERT(nGroup < MICROPROFILE_MAX_GROUPS); - uint8_t nGroupStackPos = pGroupStackPos[nGroup]; - if(nGroupStackPos) - { - nGroupStackPos--; - if(0 == nGroupStackPos) - { - nGroupTicks[nGroup] += nTicks; - } - pGroupStackPos[nGroup] = nGroupStackPos; - } - } - } - } - } - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - pLog->nGroupTicks[j] += nGroupTicks[j]; - pFrameGroup[j] += nGroupTicks[j]; - } - pLog->nStackPos = nStackPos; - } - } - { - MICROPROFILE_SCOPE(g_MicroProfileAccumulate); - for(uint32_t i = 0; i < S.nTotalTimers; ++i) - { - S.AccumTimers[i].nTicks += S.Frame[i].nTicks; - S.AccumTimers[i].nCount += S.Frame[i].nCount; - S.AccumMaxTimers[i] = MicroProfileMax(S.AccumMaxTimers[i], S.Frame[i].nTicks); - S.AccumTimersExclusive[i] += S.FrameExclusive[i]; - S.AccumMaxTimersExclusive[i] = MicroProfileMax(S.AccumMaxTimersExclusive[i], S.FrameExclusive[i]); - } - - for(uint32_t i = 0; i < MICROPROFILE_MAX_GROUPS; ++i) - { - S.AccumGroup[i] += pFrameGroup[i]; - S.AccumGroupMax[i] = MicroProfileMax(S.AccumGroupMax[i], pFrameGroup[i]); - } - - for(uint32_t j = 0; j < MICROPROFILE_META_MAX; ++j) - { - if(S.MetaCounters[j].pName && 0 != (S.nActiveBars & (MP_DRAW_META_FIRST<nAggregateGroupTicks[0], &pLog->nGroupTicks[0], sizeof(pLog->nAggregateGroupTicks)); - - if(nAggregateClear) - { - memset(&pLog->nGroupTicks[0], 0, sizeof(pLog->nGroupTicks)); - } - } - - for(uint32_t j = 0; j < MICROPROFILE_META_MAX; ++j) - { - if(S.MetaCounters[j].pName && 0 != (S.nActiveBars & (MP_DRAW_META_FIRST<= 0) - { - if(bEnabled) - { - S.nActiveGroupWanted |= S.CategoryInfo[nCategoryIndex].nGroupMask; - } - else - { - S.nActiveGroupWanted &= ~S.CategoryInfo[nCategoryIndex].nGroupMask; - } - } -} - - -void MicroProfileEnableCategory(const char* pCategory) -{ - MicroProfileEnableCategory(pCategory, true); -} -void MicroProfileDisableCategory(const char* pCategory) -{ - MicroProfileEnableCategory(pCategory, false); -} - -bool MicroProfileGetEnableAllGroups() -{ - return 0 != S.nAllGroupsWanted; -} - -void MicroProfileSetForceMetaCounters(bool bForce) -{ - S.nForceMetaCounters = bForce ? 1 : 0; -} - -bool MicroProfileGetForceMetaCounters() -{ - return 0 != S.nForceMetaCounters; -} - -void MicroProfileEnableMetaCounter(const char* pMeta) -{ - for(uint32_t i = 0; i < MICROPROFILE_META_MAX; ++i) - { - if(S.MetaCounters[i].pName && 0 == MP_STRCASECMP(S.MetaCounters[i].pName, pMeta)) - { - S.nBars |= (MP_DRAW_META_FIRST< Lock(MicroProfileMutex()); - uint16_t nGroup = MicroProfileGetGroup(pGroup, Type); - S.nForceGroup |= (1ll << nGroup); -} - -void MicroProfileForceDisableGroup(const char* pGroup, MicroProfileTokenType Type) -{ - MicroProfileInit(); - std::lock_guard Lock(MicroProfileMutex()); - uint16_t nGroup = MicroProfileGetGroup(pGroup, Type); - S.nForceGroup &= ~(1ll << nGroup); -} - - -inline void MicroProfileCalcAllTimers(float* pTimers, float* pAverage, float* pMax, float* pCallAverage, float* pExclusive, float* pAverageExclusive, float* pMaxExclusive, float* pTotal, uint32_t nSize) -{ - for(uint32_t i = 0; i < S.nTotalTimers && i < nSize; ++i) - { - const uint32_t nGroupId = S.TimerInfo[i].nGroupIndex; - const float fToMs = MicroProfileTickToMsMultiplier(S.GroupInfo[nGroupId].Type == MicroProfileTokenTypeGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu()); - uint32_t nTimer = i; - uint32_t nIdx = i * 2; - uint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1; - uint32_t nAggregateCount = S.Aggregate[nTimer].nCount ? S.Aggregate[nTimer].nCount : 1; - float fToPrc = S.fRcpReferenceTime; - float fMs = fToMs * (S.Frame[nTimer].nTicks); - float fPrc = MicroProfileMin(fMs * fToPrc, 1.f); - float fAverageMs = fToMs * (S.Aggregate[nTimer].nTicks / nAggregateFrames); - float fAveragePrc = MicroProfileMin(fAverageMs * fToPrc, 1.f); - float fMaxMs = fToMs * (S.AggregateMax[nTimer]); - float fMaxPrc = MicroProfileMin(fMaxMs * fToPrc, 1.f); - float fCallAverageMs = fToMs * (S.Aggregate[nTimer].nTicks / nAggregateCount); - float fCallAveragePrc = MicroProfileMin(fCallAverageMs * fToPrc, 1.f); - float fMsExclusive = fToMs * (S.FrameExclusive[nTimer]); - float fPrcExclusive = MicroProfileMin(fMsExclusive * fToPrc, 1.f); - float fAverageMsExclusive = fToMs * (S.AggregateExclusive[nTimer] / nAggregateFrames); - float fAveragePrcExclusive = MicroProfileMin(fAverageMsExclusive * fToPrc, 1.f); - float fMaxMsExclusive = fToMs * (S.AggregateMaxExclusive[nTimer]); - float fMaxPrcExclusive = MicroProfileMin(fMaxMsExclusive * fToPrc, 1.f); - float fTotalMs = fToMs * S.Aggregate[nTimer].nTicks; - pTimers[nIdx] = fMs; - pTimers[nIdx+1] = fPrc; - pAverage[nIdx] = fAverageMs; - pAverage[nIdx+1] = fAveragePrc; - pMax[nIdx] = fMaxMs; - pMax[nIdx+1] = fMaxPrc; - pCallAverage[nIdx] = fCallAverageMs; - pCallAverage[nIdx+1] = fCallAveragePrc; - pExclusive[nIdx] = fMsExclusive; - pExclusive[nIdx+1] = fPrcExclusive; - pAverageExclusive[nIdx] = fAverageMsExclusive; - pAverageExclusive[nIdx+1] = fAveragePrcExclusive; - pMaxExclusive[nIdx] = fMaxMsExclusive; - pMaxExclusive[nIdx+1] = fMaxPrcExclusive; - pTotal[nIdx] = fTotalMs; - pTotal[nIdx+1] = 0.f; - } -} - -void MicroProfileTogglePause() -{ - S.nToggleRunning = 1; -} - -float MicroProfileGetTime(const char* pGroup, const char* pName) -{ - MicroProfileToken nToken = MicroProfileFindToken(pGroup, pName); - if(nToken == MICROPROFILE_INVALID_TOKEN) - { - return 0.f; - } - uint32_t nTimerIndex = MicroProfileGetTimerIndex(nToken); - uint32_t nGroupIndex = MicroProfileGetGroupIndex(nToken); - float fToMs = MicroProfileTickToMsMultiplier(S.GroupInfo[nGroupIndex].Type == MicroProfileTokenTypeGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu()); - return S.Frame[nTimerIndex].nTicks * fToMs; -} - - -void MicroProfileContextSwitchSearch(uint32_t* pContextSwitchStart, uint32_t* pContextSwitchEnd, uint64_t nBaseTicksCpu, uint64_t nBaseTicksEndCpu) -{ - MICROPROFILE_SCOPE(g_MicroProfileContextSwitchSearch); - uint32_t nContextSwitchPut = S.nContextSwitchPut; - uint64_t nContextSwitchStart, nContextSwitchEnd; - nContextSwitchStart = nContextSwitchEnd = (nContextSwitchPut + MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE - 1) % MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE; - int64_t nSearchEnd = nBaseTicksEndCpu + MicroProfileMsToTick(30.f, MicroProfileTicksPerSecondCpu()); - int64_t nSearchBegin = nBaseTicksCpu - MicroProfileMsToTick(30.f, MicroProfileTicksPerSecondCpu()); - for(uint32_t i = 0; i < MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE; ++i) - { - uint32_t nIndex = (nContextSwitchPut + MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE - (i+1)) % MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE; - MicroProfileContextSwitch& CS = S.ContextSwitch[nIndex]; - if(CS.nTicks > nSearchEnd) - { - nContextSwitchEnd = nIndex; - } - if(CS.nTicks > nSearchBegin) - { - nContextSwitchStart = nIndex; - } - } - *pContextSwitchStart = nContextSwitchStart; - *pContextSwitchEnd = nContextSwitchEnd; -} - - - -#if MICROPROFILE_WEBSERVER - -#define MICROPROFILE_EMBED_HTML - -extern const char* g_MicroProfileHtml_begin[]; -extern size_t g_MicroProfileHtml_begin_sizes[]; -extern size_t g_MicroProfileHtml_begin_count; -extern const char* g_MicroProfileHtml_end[]; -extern size_t g_MicroProfileHtml_end_sizes[]; -extern size_t g_MicroProfileHtml_end_count; - -typedef void MicroProfileWriteCallback(void* Handle, size_t size, const char* pData); - -uint32_t MicroProfileWebServerPort() -{ - return S.nWebServerPort; -} - -void MicroProfileDumpFile(const char* pHtml, const char* pCsv) -{ - S.nDumpFileNextFrame = 0; - if(pHtml) - { - uint32_t nLen = strlen(pHtml); - if(nLen > sizeof(S.HtmlDumpPath)-1) - { - return; - } - memcpy(S.HtmlDumpPath, pHtml, nLen+1); - S.nDumpFileNextFrame |= 1; - } - if(pCsv) - { - uint32_t nLen = strlen(pCsv); - if(nLen > sizeof(S.CsvDumpPath)-1) - { - return; - } - memcpy(S.CsvDumpPath, pCsv, nLen+1); - S.nDumpFileNextFrame |= 2; - } -} - -void MicroProfilePrintf(MicroProfileWriteCallback CB, void* Handle, const char* pFmt, ...) -{ - char buffer[32*1024]; - va_list args; - va_start (args, pFmt); -#ifdef _WIN32 - size_t size = vsprintf_s(buffer, pFmt, args); -#else - size_t size = vsnprintf(buffer, sizeof(buffer)-1, pFmt, args); -#endif - CB(Handle, size, &buffer[0]); - va_end (args); -} - -#define printf(...) MicroProfilePrintf(CB, Handle, __VA_ARGS__) -void MicroProfileDumpCsv(MicroProfileWriteCallback CB, void* Handle, int nMaxFrames) -{ - uint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1; - float fToMsCPU = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()); - float fToMsGPU = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondGpu()); - - printf("frames,%d\n", nAggregateFrames); - printf("group,name,average,max,callaverage\n"); - - uint32_t nNumTimers = S.nTotalTimers; - uint32_t nBlockSize = 2 * nNumTimers; - float* pTimers = (float*)alloca(nBlockSize * 8 * sizeof(float)); - float* pAverage = pTimers + nBlockSize; - float* pMax = pTimers + 2 * nBlockSize; - float* pCallAverage = pTimers + 3 * nBlockSize; - float* pTimersExclusive = pTimers + 4 * nBlockSize; - float* pAverageExclusive = pTimers + 5 * nBlockSize; - float* pMaxExclusive = pTimers + 6 * nBlockSize; - float* pTotal = pTimers + 7 * nBlockSize; - - MicroProfileCalcAllTimers(pTimers, pAverage, pMax, pCallAverage, pTimersExclusive, pAverageExclusive, pMaxExclusive, pTotal, nNumTimers); - - for(uint32_t i = 0; i < S.nTotalTimers; ++i) - { - uint32_t nIdx = i * 2; - printf("\"%s\",\"%s\",%f,%f,%f\n", S.TimerInfo[i].pName, S.GroupInfo[S.TimerInfo[i].nGroupIndex].pName, pAverage[nIdx], pMax[nIdx], pCallAverage[nIdx]); - } - - printf("\n\n"); - - printf("group,average,max,total\n"); - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - const char* pGroupName = S.GroupInfo[j].pName; - float fToMs = S.GroupInfo[j].Type == MicroProfileTokenTypeGpu ? fToMsGPU : fToMsCPU; - if(pGroupName[0] != '\0') - { - printf("\"%s\",%.3f,%.3f,%.3f\n", pGroupName, fToMs * S.AggregateGroup[j] / nAggregateFrames, fToMs * S.AggregateGroup[j] / nAggregateFrames, fToMs * S.AggregateGroup[j]); - } - } - - printf("\n\n"); - printf("group,thread,average,total\n"); - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - for(uint32_t i = 0; i < S.nNumLogs; ++i) - { - if(S.Pool[i]) - { - const char* pThreadName = &S.Pool[i]->ThreadName[0]; - // MicroProfilePrintf(CB, Handle, "var ThreadGroupTime%d = [", i); - float fToMs = S.Pool[i]->nGpu ? fToMsGPU : fToMsCPU; - { - uint64_t nTicks = S.Pool[i]->nAggregateGroupTicks[j]; - float fTime = nTicks / nAggregateFrames * fToMs; - float fTimeTotal = nTicks * fToMs; - if(fTimeTotal > 0.01f) - { - const char* pGroupName = S.GroupInfo[j].pName; - printf("\"%s\",\"%s\",%.3f,%.3f\n", pGroupName, pThreadName, fTime, fTimeTotal); - } - } - } - } - } - - printf("\n\n"); - printf("frametimecpu\n"); - - const uint32_t nCount = MICROPROFILE_MAX_FRAME_HISTORY - MICROPROFILE_GPU_FRAME_DELAY - 3; - const uint32_t nStart = S.nFrameCurrent; - for(uint32_t i = nCount; i > 0; i--) - { - uint32_t nFrame = (nStart + MICROPROFILE_MAX_FRAME_HISTORY - i) % MICROPROFILE_MAX_FRAME_HISTORY; - uint32_t nFrameNext = (nStart + MICROPROFILE_MAX_FRAME_HISTORY - i + 1) % MICROPROFILE_MAX_FRAME_HISTORY; - uint64_t nTicks = S.Frames[nFrameNext].nFrameStartCpu - S.Frames[nFrame].nFrameStartCpu; - printf("%f,", nTicks * fToMsCPU); - } - printf("\n"); - - printf("\n\n"); - printf("frametimegpu\n"); - - for(uint32_t i = nCount; i > 0; i--) - { - uint32_t nFrame = (nStart + MICROPROFILE_MAX_FRAME_HISTORY - i) % MICROPROFILE_MAX_FRAME_HISTORY; - uint32_t nFrameNext = (nStart + MICROPROFILE_MAX_FRAME_HISTORY - i + 1) % MICROPROFILE_MAX_FRAME_HISTORY; - uint64_t nTicks = S.Frames[nFrameNext].nFrameStartGpu - S.Frames[nFrame].nFrameStartGpu; - printf("%f,", nTicks * fToMsGPU); - } - printf("\n\n"); - printf("Meta\n");//only single frame snapshot - printf("name,average,max,total\n"); - for(int j = 0; j < MICROPROFILE_META_MAX; ++j) - { - if(S.MetaCounters[j].pName) - { - printf("\"%s\",%f,%lld,%lld\n",S.MetaCounters[j].pName, S.MetaCounters[j].nSumAggregate / (float)nAggregateFrames, S.MetaCounters[j].nSumAggregateMax,S.MetaCounters[j].nSumAggregate); - } - } -} -#undef printf - -void MicroProfileDumpHtml(MicroProfileWriteCallback CB, void* Handle, int nMaxFrames, const char* pHost) -{ - uint32_t nRunning = S.nRunning; - S.nRunning = 0; - //stall pushing of timers - uint64_t nActiveGroup = S.nActiveGroup; - S.nActiveGroup = 0; - S.nPauseTicks = MP_TICK(); - - - for(size_t i = 0; i < g_MicroProfileHtml_begin_count; ++i) - { - CB(Handle, g_MicroProfileHtml_begin_sizes[i]-1, g_MicroProfileHtml_begin[i]); - } - //dump info - uint64_t nTicks = MP_TICK(); - - float fToMsCPU = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()); - float fToMsGPU = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondGpu()); - float fAggregateMs = fToMsCPU * (nTicks - S.nAggregateFlipTick); - MicroProfilePrintf(CB, Handle, "var DumpHost = '%s';\n", pHost ? pHost : ""); - time_t CaptureTime; - time(&CaptureTime); - MicroProfilePrintf(CB, Handle, "var DumpUtcCaptureTime = %ld;\n", CaptureTime); - MicroProfilePrintf(CB, Handle, "var AggregateInfo = {'Frames':%d, 'Time':%f};\n", S.nAggregateFrames, fAggregateMs); - - //categories - MicroProfilePrintf(CB, Handle, "var CategoryInfo = Array(%d);\n",S.nCategoryCount); - for(uint32_t i = 0; i < S.nCategoryCount; ++i) - { - MicroProfilePrintf(CB, Handle, "CategoryInfo[%d] = \"%s\";\n", i, S.CategoryInfo[i].pName); - } - - //groups - MicroProfilePrintf(CB, Handle, "var GroupInfo = Array(%d);\n\n",S.nGroupCount); - uint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1; - float fRcpAggregateFrames = 1.f / nAggregateFrames; - for(uint32_t i = 0; i < S.nGroupCount; ++i) - { - MP_ASSERT(i == S.GroupInfo[i].nGroupIndex); - float fToMs = S.GroupInfo[i].Type == MicroProfileTokenTypeCpu ? fToMsCPU : fToMsGPU; - MicroProfilePrintf(CB, Handle, "GroupInfo[%d] = MakeGroup(%d, \"%s\", %d, %d, %d, %f, %f, %f, '#%02x%02x%02x');\n", - S.GroupInfo[i].nGroupIndex, - S.GroupInfo[i].nGroupIndex, - S.GroupInfo[i].pName, - S.GroupInfo[i].nCategory, - S.GroupInfo[i].nNumTimers, - S.GroupInfo[i].Type == MicroProfileTokenTypeGpu?1:0, - fToMs * S.AggregateGroup[i], - fToMs * S.AggregateGroup[i] / nAggregateFrames, - fToMs * S.AggregateGroupMax[i], - MICROPROFILE_UNPACK_RED(S.GroupInfo[i].nColor) & 0xff, - MICROPROFILE_UNPACK_GREEN(S.GroupInfo[i].nColor) & 0xff, - MICROPROFILE_UNPACK_BLUE(S.GroupInfo[i].nColor) & 0xff); - } - //timers - - uint32_t nNumTimers = S.nTotalTimers; - uint32_t nBlockSize = 2 * nNumTimers; - float* pTimers = (float*)alloca(nBlockSize * 8 * sizeof(float)); - float* pAverage = pTimers + nBlockSize; - float* pMax = pTimers + 2 * nBlockSize; - float* pCallAverage = pTimers + 3 * nBlockSize; - float* pTimersExclusive = pTimers + 4 * nBlockSize; - float* pAverageExclusive = pTimers + 5 * nBlockSize; - float* pMaxExclusive = pTimers + 6 * nBlockSize; - float* pTotal = pTimers + 7 * nBlockSize; - - MicroProfileCalcAllTimers(pTimers, pAverage, pMax, pCallAverage, pTimersExclusive, pAverageExclusive, pMaxExclusive, pTotal, nNumTimers); - - MicroProfilePrintf(CB, Handle, "\nvar TimerInfo = Array(%d);\n\n", S.nTotalTimers); - for(uint32_t i = 0; i < S.nTotalTimers; ++i) - { - uint32_t nIdx = i * 2; - MP_ASSERT(i == S.TimerInfo[i].nTimerIndex); - MicroProfilePrintf(CB, Handle, "var Meta%d = [", i); - bool bOnce = true; - for(int j = 0; j < MICROPROFILE_META_MAX; ++j) - { - if(S.MetaCounters[j].pName) - { - uint32_t lala = S.MetaCounters[j].nCounters[i]; - MicroProfilePrintf(CB, Handle, bOnce ? "%d" : ",%d", lala); - bOnce = false; - } - } - MicroProfilePrintf(CB, Handle, "];\n"); - MicroProfilePrintf(CB, Handle, "var MetaAvg%d = [", i); - bOnce = true; - for(int j = 0; j < MICROPROFILE_META_MAX; ++j) - { - if(S.MetaCounters[j].pName) - { - MicroProfilePrintf(CB, Handle, bOnce ? "%f" : ",%f", fRcpAggregateFrames * S.MetaCounters[j].nAggregate[i]); - bOnce = false; - } - } - MicroProfilePrintf(CB, Handle, "];\n"); - MicroProfilePrintf(CB, Handle, "var MetaMax%d = [", i); - bOnce = true; - for(int j = 0; j < MICROPROFILE_META_MAX; ++j) - { - if(S.MetaCounters[j].pName) - { - MicroProfilePrintf(CB, Handle, bOnce ? "%d" : ",%d", S.MetaCounters[j].nAggregateMax[i]); - bOnce = false; - } - } - MicroProfilePrintf(CB, Handle, "];\n"); - - - uint32_t nColor = S.TimerInfo[i].nColor; - uint32_t nColorDark = (nColor >> 1) & ~0x80808080; - MicroProfilePrintf(CB, Handle, "TimerInfo[%d] = MakeTimer(%d, \"%s\", %d, '#%02x%02x%02x','#%02x%02x%02x', %f, %f, %f, %f, %f, %d, %f, Meta%d, MetaAvg%d, MetaMax%d);\n", S.TimerInfo[i].nTimerIndex, S.TimerInfo[i].nTimerIndex, S.TimerInfo[i].pName, S.TimerInfo[i].nGroupIndex, - MICROPROFILE_UNPACK_RED(nColor) & 0xff, - MICROPROFILE_UNPACK_GREEN(nColor) & 0xff, - MICROPROFILE_UNPACK_BLUE(nColor) & 0xff, - MICROPROFILE_UNPACK_RED(nColorDark) & 0xff, - MICROPROFILE_UNPACK_GREEN(nColorDark) & 0xff, - MICROPROFILE_UNPACK_BLUE(nColorDark) & 0xff, - pAverage[nIdx], - pMax[nIdx], - pAverageExclusive[nIdx], - pMaxExclusive[nIdx], - pCallAverage[nIdx], - S.Aggregate[i].nCount, - pTotal[nIdx], - i,i,i); - - } - - MicroProfilePrintf(CB, Handle, "\nvar ThreadNames = ["); - for(uint32_t i = 0; i < S.nNumLogs; ++i) - { - if(S.Pool[i]) - { - MicroProfilePrintf(CB, Handle, "'%s',", S.Pool[i]->ThreadName); - } - else - { - MicroProfilePrintf(CB, Handle, "'Thread %d',", i); - } - } - MicroProfilePrintf(CB, Handle, "];\n\n"); - - - for(uint32_t i = 0; i < S.nNumLogs; ++i) - { - if(S.Pool[i]) - { - MicroProfilePrintf(CB, Handle, "var ThreadGroupTime%d = [", i); - float fToMs = S.Pool[i]->nGpu ? fToMsGPU : fToMsCPU; - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - MicroProfilePrintf(CB, Handle, "%f,", S.Pool[i]->nAggregateGroupTicks[j]/nAggregateFrames * fToMs); - } - MicroProfilePrintf(CB, Handle, "];\n"); - } - } - MicroProfilePrintf(CB, Handle, "\nvar ThreadGroupTimeArray = ["); - for(uint32_t i = 0; i < S.nNumLogs; ++i) - { - if(S.Pool[i]) - { - MicroProfilePrintf(CB, Handle, "ThreadGroupTime%d,", i); - } - } - MicroProfilePrintf(CB, Handle, "];\n"); - - - for(uint32_t i = 0; i < S.nNumLogs; ++i) - { - if(S.Pool[i]) - { - MicroProfilePrintf(CB, Handle, "var ThreadGroupTimeTotal%d = [", i); - float fToMs = S.Pool[i]->nGpu ? fToMsGPU : fToMsCPU; - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - MicroProfilePrintf(CB, Handle, "%f,", S.Pool[i]->nAggregateGroupTicks[j] * fToMs); - } - MicroProfilePrintf(CB, Handle, "];\n"); - } - } - MicroProfilePrintf(CB, Handle, "\nvar ThreadGroupTimeTotalArray = ["); - for(uint32_t i = 0; i < S.nNumLogs; ++i) - { - if(S.Pool[i]) - { - MicroProfilePrintf(CB, Handle, "ThreadGroupTimeTotal%d,", i); - } - } - MicroProfilePrintf(CB, Handle, "];"); - - - - - MicroProfilePrintf(CB, Handle, "\nvar ThreadIds = ["); - for(uint32_t i = 0; i < S.nNumLogs; ++i) - { - if(S.Pool[i]) - { - ThreadIdType ThreadId = S.Pool[i]->nThreadId; - if(!ThreadId) - { - ThreadId = (ThreadIdType)-1; - } - MicroProfilePrintf(CB, Handle, "%d,", ThreadId); - } - else - { - MicroProfilePrintf(CB, Handle, "-1,", i); - } - } - MicroProfilePrintf(CB, Handle, "];\n\n"); - - MicroProfilePrintf(CB, Handle, "\nvar MetaNames = ["); - for(int i = 0; i < MICROPROFILE_META_MAX; ++i) - { - if(S.MetaCounters[i].pName) - { - MicroProfilePrintf(CB, Handle, "'%s',", S.MetaCounters[i].pName); - } - } - - - MicroProfilePrintf(CB, Handle, "];\n\n"); - - - - uint32_t nNumFrames = (MICROPROFILE_MAX_FRAME_HISTORY - MICROPROFILE_GPU_FRAME_DELAY - 3); //leave a few to not overwrite - nNumFrames = MicroProfileMin(nNumFrames, (uint32_t)nMaxFrames); - - - uint32_t nFirstFrame = (S.nFrameCurrent + MICROPROFILE_MAX_FRAME_HISTORY - nNumFrames) % MICROPROFILE_MAX_FRAME_HISTORY; - uint32_t nLastFrame = (nFirstFrame + nNumFrames) % MICROPROFILE_MAX_FRAME_HISTORY; - MP_ASSERT(nLastFrame == (S.nFrameCurrent % MICROPROFILE_MAX_FRAME_HISTORY)); - MP_ASSERT(nFirstFrame < MICROPROFILE_MAX_FRAME_HISTORY); - MP_ASSERT(nLastFrame < MICROPROFILE_MAX_FRAME_HISTORY); - const int64_t nTickStart = S.Frames[nFirstFrame].nFrameStartCpu; - const int64_t nTickEnd = S.Frames[nLastFrame].nFrameStartCpu; - int64_t nTickStartGpu = S.Frames[nFirstFrame].nFrameStartGpu; - - int64_t nTickReferenceCpu, nTickReferenceGpu; - int64_t nTicksPerSecondCpu = MicroProfileTicksPerSecondCpu(); - int64_t nTicksPerSecondGpu = MicroProfileTicksPerSecondGpu(); - int nTickReference = 0; - if(MicroProfileGetGpuTickReference(&nTickReferenceCpu, &nTickReferenceGpu)) - { - nTickStartGpu = (nTickStart - nTickReferenceCpu) * nTicksPerSecondGpu / nTicksPerSecondCpu + nTickReferenceGpu; - nTickReference = 1; - } - - -#if MICROPROFILE_DEBUG - printf("dumping %d frames\n", nNumFrames); - printf("dumping frame %d to %d\n", nFirstFrame, nLastFrame); -#endif - - - uint32_t* nTimerCounter = (uint32_t*)alloca(sizeof(uint32_t)* S.nTotalTimers); - memset(nTimerCounter, 0, sizeof(uint32_t) * S.nTotalTimers); - - MicroProfilePrintf(CB, Handle, "var Frames = Array(%d);\n", nNumFrames); - for(uint32_t i = 0; i < nNumFrames; ++i) - { - uint32_t nFrameIndex = (nFirstFrame + i) % MICROPROFILE_MAX_FRAME_HISTORY; - uint32_t nFrameIndexNext = (nFrameIndex + 1) % MICROPROFILE_MAX_FRAME_HISTORY; - - for(uint32_t j = 0; j < S.nNumLogs; ++j) - { - MicroProfileThreadLog* pLog = S.Pool[j]; - int64_t nStartTickBase = pLog->nGpu ? nTickStartGpu : nTickStart; - uint32_t nLogStart = S.Frames[nFrameIndex].nLogStart[j]; - uint32_t nLogEnd = S.Frames[nFrameIndexNext].nLogStart[j]; - - float fToMsCpu = MicroProfileTickToMsMultiplier(nTicksPerSecondCpu); - float fToMsBase = MicroProfileTickToMsMultiplier(pLog->nGpu ? nTicksPerSecondGpu : nTicksPerSecondCpu); - MicroProfilePrintf(CB, Handle, "var ts_%d_%d = [", i, j); - if(nLogStart != nLogEnd) - { - uint32_t k = nLogStart; - uint32_t nLogType = MicroProfileLogType(pLog->Log[k]); - float fToMs = nLogType == MP_LOG_GPU_EXTRA ? fToMsCpu : fToMsBase; - int64_t nStartTick = nLogType == MP_LOG_GPU_EXTRA ? nTickStart : nStartTickBase; - float fTime = nLogType == MP_LOG_META ? 0.f : MicroProfileLogTickDifference(nStartTick, pLog->Log[k]) * fToMs; - MicroProfilePrintf(CB, Handle, "%f", fTime); - for(k = (k+1) % MICROPROFILE_BUFFER_SIZE; k != nLogEnd; k = (k+1) % MICROPROFILE_BUFFER_SIZE) - { - uint32_t nLogType = MicroProfileLogType(pLog->Log[k]); - float fToMs = nLogType == MP_LOG_GPU_EXTRA ? fToMsCpu : fToMsBase; - nStartTick = nLogType == MP_LOG_GPU_EXTRA ? nTickStart : nStartTickBase; - float fTime = nLogType == MP_LOG_META ? 0.f : MicroProfileLogTickDifference(nStartTick, pLog->Log[k]) * fToMs; - MicroProfilePrintf(CB, Handle, ",%f", fTime); - } - } - MicroProfilePrintf(CB, Handle, "];\n"); - MicroProfilePrintf(CB, Handle, "var tt_%d_%d = [", i, j); - if(nLogStart != nLogEnd) - { - uint32_t k = nLogStart; - MicroProfilePrintf(CB, Handle, "%d", MicroProfileLogType(pLog->Log[k])); - for(k = (k+1) % MICROPROFILE_BUFFER_SIZE; k != nLogEnd; k = (k+1) % MICROPROFILE_BUFFER_SIZE) - { - uint32_t nLogType = MicroProfileLogType(pLog->Log[k]); - if(nLogType == MP_LOG_META) - { - //for meta, store the count + 3, which is the tick part - nLogType = 3 + MicroProfileLogGetTick(pLog->Log[k]); - } - MicroProfilePrintf(CB, Handle, ",%d", nLogType); - } - } - MicroProfilePrintf(CB, Handle, "];\n"); - - MicroProfilePrintf(CB, Handle, "var ti_%d_%d = [", i, j); - if(nLogStart != nLogEnd) - { - uint32_t k = nLogStart; - MicroProfilePrintf(CB, Handle, "%d", (uint32_t)MicroProfileLogTimerIndex(pLog->Log[k])); - for(k = (k+1) % MICROPROFILE_BUFFER_SIZE; k != nLogEnd; k = (k+1) % MICROPROFILE_BUFFER_SIZE) - { - uint32_t nTimerIndex = (uint32_t)MicroProfileLogTimerIndex(pLog->Log[k]); - MicroProfilePrintf(CB, Handle, ",%d", nTimerIndex); - nTimerCounter[nTimerIndex]++; - } - } - MicroProfilePrintf(CB, Handle, "];\n"); - - } - - MicroProfilePrintf(CB, Handle, "var ts%d = [", i); - for(uint32_t j = 0; j < S.nNumLogs; ++j) - { - MicroProfilePrintf(CB, Handle, "ts_%d_%d,", i, j); - } - MicroProfilePrintf(CB, Handle, "];\n"); - MicroProfilePrintf(CB, Handle, "var tt%d = [", i); - for(uint32_t j = 0; j < S.nNumLogs; ++j) - { - MicroProfilePrintf(CB, Handle, "tt_%d_%d,", i, j); - } - MicroProfilePrintf(CB, Handle, "];\n"); - - MicroProfilePrintf(CB, Handle, "var ti%d = [", i); - for(uint32_t j = 0; j < S.nNumLogs; ++j) - { - MicroProfilePrintf(CB, Handle, "ti_%d_%d,", i, j); - } - MicroProfilePrintf(CB, Handle, "];\n"); - - - int64_t nFrameStart = S.Frames[nFrameIndex].nFrameStartCpu; - int64_t nFrameEnd = S.Frames[nFrameIndexNext].nFrameStartCpu; - - float fToMs = MicroProfileTickToMsMultiplier(nTicksPerSecondCpu); - float fFrameMs = MicroProfileLogTickDifference(nTickStart, nFrameStart) * fToMs; - float fFrameEndMs = MicroProfileLogTickDifference(nTickStart, nFrameEnd) * fToMs; - float fFrameGpuMs = 0; - float fFrameGpuEndMs = 0; - if(nTickReference) - { - fFrameGpuMs = MicroProfileLogTickDifference(nTickStartGpu, S.Frames[nFrameIndex].nFrameStartGpu) * fToMsGPU; - fFrameGpuEndMs = MicroProfileLogTickDifference(nTickStartGpu, S.Frames[nFrameIndexNext].nFrameStartGpu) * fToMsGPU; - } - MicroProfilePrintf(CB, Handle, "Frames[%d] = MakeFrame(%d, %f, %f, %f, %f, ts%d, tt%d, ti%d);\n", i, 0, fFrameMs, fFrameEndMs, fFrameGpuMs, fFrameGpuEndMs, i, i, i); - } - - uint32_t nContextSwitchStart = 0; - uint32_t nContextSwitchEnd = 0; - MicroProfileContextSwitchSearch(&nContextSwitchStart, &nContextSwitchEnd, nTickStart, nTickEnd); - - uint32_t nWrittenBefore = S.nWebServerDataSent; - MicroProfilePrintf(CB, Handle, "var CSwitchThreadInOutCpu = ["); - for(uint32_t j = nContextSwitchStart; j != nContextSwitchEnd; j = (j+1) % MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE) - { - MicroProfileContextSwitch CS = S.ContextSwitch[j]; - int nCpu = CS.nCpu; - MicroProfilePrintf(CB, Handle, "%d,%d,%d,", CS.nThreadIn, CS.nThreadOut, nCpu); - } - MicroProfilePrintf(CB, Handle, "];\n"); - MicroProfilePrintf(CB, Handle, "var CSwitchTime = ["); - float fToMsCpu = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()); - for(uint32_t j = nContextSwitchStart; j != nContextSwitchEnd; j = (j+1) % MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE) - { - MicroProfileContextSwitch CS = S.ContextSwitch[j]; - float fTime = MicroProfileLogTickDifference(nTickStart, CS.nTicks) * fToMsCpu; - MicroProfilePrintf(CB, Handle, "%f,", fTime); - } - MicroProfilePrintf(CB, Handle, "];\n"); - uint32_t nWrittenAfter = S.nWebServerDataSent; - MicroProfilePrintf(CB, Handle, "//CSwitch Size %d\n", nWrittenAfter - nWrittenBefore); - - - for(size_t i = 0; i < g_MicroProfileHtml_end_count; ++i) - { - CB(Handle, g_MicroProfileHtml_end_sizes[i]-1, g_MicroProfileHtml_end[i]); - } - - uint32_t* nGroupCounter = (uint32_t*)alloca(sizeof(uint32_t)* S.nGroupCount); - - memset(nGroupCounter, 0, sizeof(uint32_t) * S.nGroupCount); - for(uint32_t i = 0; i < S.nTotalTimers; ++i) - { - uint32_t nGroupIndex = S.TimerInfo[i].nGroupIndex; - nGroupCounter[nGroupIndex] += nTimerCounter[i]; - } - - uint32_t* nGroupCounterSort = (uint32_t*)alloca(sizeof(uint32_t)* S.nGroupCount); - uint32_t* nTimerCounterSort = (uint32_t*)alloca(sizeof(uint32_t)* S.nTotalTimers); - for(uint32_t i = 0; i < S.nGroupCount; ++i) - { - nGroupCounterSort[i] = i; - } - for(uint32_t i = 0; i < S.nTotalTimers; ++i) - { - nTimerCounterSort[i] = i; - } - std::sort(nGroupCounterSort, nGroupCounterSort + S.nGroupCount, - [nGroupCounter](const uint32_t l, const uint32_t r) - { - return nGroupCounter[l] > nGroupCounter[r]; - } - ); - - std::sort(nTimerCounterSort, nTimerCounterSort + S.nTotalTimers, - [nTimerCounter](const uint32_t l, const uint32_t r) - { - return nTimerCounter[l] > nTimerCounter[r]; - } - ); - - MicroProfilePrintf(CB, Handle, "\n\n"); - - S.nActiveGroup = nActiveGroup; - S.nRunning = nRunning; - -#if MICROPROFILE_DEBUG - int64_t nTicksEnd = MP_TICK(); - float fMs = fToMsCpu * (nTicksEnd - S.nPauseTicks); - printf("html dump took %6.2fms\n", fMs); -#endif - - -} - -void MicroProfileWriteFile(void* Handle, size_t nSize, const char* pData) -{ - fwrite(pData, nSize, 1, (FILE*)Handle); -} - -void MicroProfileDumpToFile() -{ - std::lock_guard Lock(MicroProfileMutex()); - if(S.nDumpFileNextFrame&1) - { - FILE* F = fopen(S.HtmlDumpPath, "w"); - if(F) - { - MicroProfileDumpHtml(MicroProfileWriteFile, F, MICROPROFILE_WEBSERVER_MAXFRAMES, S.HtmlDumpPath); - fclose(F); - } - } - if(S.nDumpFileNextFrame&2) - { - FILE* F = fopen(S.CsvDumpPath, "w"); - if(F) - { - MicroProfileDumpCsv(MicroProfileWriteFile, F, MICROPROFILE_WEBSERVER_MAXFRAMES); - fclose(F); - } - } -} - -void MicroProfileFlushSocket(MpSocket Socket) -{ - send(Socket, &S.WebServerBuffer[0], S.WebServerPut, 0); - S.WebServerPut = 0; - -} - -void MicroProfileWriteSocket(void* Handle, size_t nSize, const char* pData) -{ - S.nWebServerDataSent += nSize; - MpSocket Socket = *(MpSocket*)Handle; - if(nSize > MICROPROFILE_WEBSERVER_SOCKET_BUFFER_SIZE / 2) - { - MicroProfileFlushSocket(Socket); - send(Socket, pData, nSize, 0); - - } - else - { - memcpy(&S.WebServerBuffer[S.WebServerPut], pData, nSize); - S.WebServerPut += nSize; - if(S.WebServerPut > MICROPROFILE_WEBSERVER_SOCKET_BUFFER_SIZE/2) - { - MicroProfileFlushSocket(Socket); - } - } -} - -#if MICROPROFILE_MINIZ -#ifndef MICROPROFILE_COMPRESS_BUFFER_SIZE -#define MICROPROFILE_COMPRESS_BUFFER_SIZE (256<<10) -#endif - -#define MICROPROFILE_COMPRESS_CHUNK (MICROPROFILE_COMPRESS_BUFFER_SIZE/2) -struct MicroProfileCompressedSocketState -{ - unsigned char DeflateOut[MICROPROFILE_COMPRESS_CHUNK]; - unsigned char DeflateIn[MICROPROFILE_COMPRESS_CHUNK]; - mz_stream Stream; - MpSocket Socket; - uint32_t nSize; - uint32_t nCompressedSize; - uint32_t nFlushes; - uint32_t nMemmoveBytes; -}; - -void MicroProfileCompressedSocketFlush(MicroProfileCompressedSocketState* pState) -{ - mz_stream& Stream = pState->Stream; - unsigned char* pSendStart = &pState->DeflateOut[0]; - unsigned char* pSendEnd = &pState->DeflateOut[MICROPROFILE_COMPRESS_CHUNK - Stream.avail_out]; - if(pSendStart != pSendEnd) - { - send(pState->Socket, (const char*)pSendStart, pSendEnd - pSendStart, 0); - pState->nCompressedSize += pSendEnd - pSendStart; - } - Stream.next_out = &pState->DeflateOut[0]; - Stream.avail_out = MICROPROFILE_COMPRESS_CHUNK; - -} -void MicroProfileCompressedSocketStart(MicroProfileCompressedSocketState* pState, MpSocket Socket) -{ - mz_stream& Stream = pState->Stream; - memset(&Stream, 0, sizeof(Stream)); - Stream.next_out = &pState->DeflateOut[0]; - Stream.avail_out = MICROPROFILE_COMPRESS_CHUNK; - Stream.next_in = &pState->DeflateIn[0]; - Stream.avail_in = 0; - mz_deflateInit(&Stream, Z_DEFAULT_COMPRESSION); - pState->Socket = Socket; - pState->nSize = 0; - pState->nCompressedSize = 0; - pState->nFlushes = 0; - pState->nMemmoveBytes = 0; - -} -void MicroProfileCompressedSocketFinish(MicroProfileCompressedSocketState* pState) -{ - mz_stream& Stream = pState->Stream; - MicroProfileCompressedSocketFlush(pState); - int r = mz_deflate(&Stream, MZ_FINISH); - MP_ASSERT(r == MZ_STREAM_END); - MicroProfileCompressedSocketFlush(pState); - r = mz_deflateEnd(&Stream); - MP_ASSERT(r == MZ_OK); -} - -void MicroProfileCompressedWriteSocket(void* Handle, size_t nSize, const char* pData) -{ - MicroProfileCompressedSocketState* pState = (MicroProfileCompressedSocketState*)Handle; - mz_stream& Stream = pState->Stream; - const unsigned char* pDeflateInEnd = Stream.next_in + Stream.avail_in; - const unsigned char* pDeflateInStart = &pState->DeflateIn[0]; - const unsigned char* pDeflateInRealEnd = &pState->DeflateIn[MICROPROFILE_COMPRESS_CHUNK]; - pState->nSize += nSize; - if(nSize <= pDeflateInRealEnd - pDeflateInEnd) - { - memcpy((void*)pDeflateInEnd, pData, nSize); - Stream.avail_in += nSize; - MP_ASSERT(Stream.next_in + Stream.avail_in <= pDeflateInRealEnd); - return; - } - int Flush = 0; - while(nSize) - { - pDeflateInEnd = Stream.next_in + Stream.avail_in; - if(Flush) - { - pState->nFlushes++; - MicroProfileCompressedSocketFlush(pState); - pDeflateInRealEnd = &pState->DeflateIn[MICROPROFILE_COMPRESS_CHUNK]; - if(pDeflateInEnd == pDeflateInRealEnd) - { - if(Stream.avail_in) - { - MP_ASSERT(pDeflateInStart != Stream.next_in); - memmove((void*)pDeflateInStart, Stream.next_in, Stream.avail_in); - pState->nMemmoveBytes += Stream.avail_in; - } - Stream.next_in = pDeflateInStart; - pDeflateInEnd = Stream.next_in + Stream.avail_in; - } - } - size_t nSpace = pDeflateInRealEnd - pDeflateInEnd; - size_t nBytes = MicroProfileMin(nSpace, nSize); - MP_ASSERT(nBytes + pDeflateInEnd <= pDeflateInRealEnd); - memcpy((void*)pDeflateInEnd, pData, nBytes); - Stream.avail_in += nBytes; - nSize -= nBytes; - pData += nBytes; - int r = mz_deflate(&Stream, MZ_NO_FLUSH); - Flush = r == MZ_BUF_ERROR || nBytes == 0 || Stream.avail_out == 0 ? 1 : 0; - MP_ASSERT(r == MZ_BUF_ERROR || r == MZ_OK); - if(r == MZ_BUF_ERROR) - { - r = mz_deflate(&Stream, MZ_SYNC_FLUSH); - } - } -} -#endif - - -#ifndef MicroProfileSetNonBlocking //fcntl doesnt work on a some unix like platforms.. -void MicroProfileSetNonBlocking(MpSocket Socket, int NonBlocking) -{ -#ifdef _WIN32 - u_long nonBlocking = NonBlocking ? 1 : 0; - ioctlsocket(Socket, FIONBIO, &nonBlocking); -#else - int Options = fcntl(Socket, F_GETFL); - if(NonBlocking) - { - fcntl(Socket, F_SETFL, Options|O_NONBLOCK); - } - else - { - fcntl(Socket, F_SETFL, Options&(~O_NONBLOCK)); - } -#endif -} -#endif - -void MicroProfileWebServerStart() -{ -#ifdef _WIN32 - WSADATA wsa; - if(WSAStartup(MAKEWORD(2, 2), &wsa)) - { - S.ListenerSocket = -1; - return; - } -#endif - - S.ListenerSocket = socket(PF_INET, SOCK_STREAM, 6); - MP_ASSERT(!MP_INVALID_SOCKET(S.ListenerSocket)); - MicroProfileSetNonBlocking(S.ListenerSocket, 1); - - S.nWebServerPort = (uint32_t)-1; - struct sockaddr_in Addr; - Addr.sin_family = AF_INET; - Addr.sin_addr.s_addr = INADDR_ANY; - for(int i = 0; i < 20; ++i) - { - Addr.sin_port = htons(MICROPROFILE_WEBSERVER_PORT+i); - if(0 == bind(S.ListenerSocket, (sockaddr*)&Addr, sizeof(Addr))) - { - S.nWebServerPort = MICROPROFILE_WEBSERVER_PORT+i; - break; - } - } - listen(S.ListenerSocket, 8); -} - -void MicroProfileWebServerStop() -{ -#ifdef _WIN32 - closesocket(S.ListenerSocket); - WSACleanup(); -#else - close(S.ListenerSocket); -#endif -} - -int MicroProfileParseGet(const char* pGet) -{ - const char* pStart = pGet; - while(*pGet != '\0') - { - if(*pGet < '0' || *pGet > '9') - return 0; - pGet++; - } - int nFrames = atoi(pStart); - if(nFrames) - { - return nFrames; - } - else - { - return MICROPROFILE_WEBSERVER_MAXFRAMES; - } -} -bool MicroProfileWebServerUpdate() -{ - MICROPROFILE_SCOPEI("MicroProfile", "Webserver-update", -1); - MpSocket Connection = accept(S.ListenerSocket, 0, 0); - bool bServed = false; - if(!MP_INVALID_SOCKET(Connection)) - { - std::lock_guard Lock(MicroProfileMutex()); - char Req[8192]; - MicroProfileSetNonBlocking(Connection, 0); - int nReceived = recv(Connection, Req, sizeof(Req)-1, 0); - if(nReceived > 0) - { - Req[nReceived] = '\0'; -#if MICROPROFILE_MINIZ -#define MICROPROFILE_HTML_HEADER "HTTP/1.0 200 OK\r\nContent-Type: text/html\r\nContent-Encoding: deflate\r\nExpires: Tue, 01 Jan 2199 16:00:00 GMT\r\n\r\n" -#else -#define MICROPROFILE_HTML_HEADER "HTTP/1.0 200 OK\r\nContent-Type: text/html\r\nExpires: Tue, 01 Jan 2199 16:00:00 GMT\r\n\r\n" -#endif - char* pHttp = strstr(Req, "HTTP/"); - char* pGet = strstr(Req, "GET /"); - char* pHost = strstr(Req, "Host: "); - auto Terminate = [](char* pString) - { - char* pEnd = pString; - while(*pEnd != '\0') - { - if(*pEnd == '\r' || *pEnd == '\n' || *pEnd == ' ') - { - *pEnd = '\0'; - return; - } - pEnd++; - } - }; - if(pHost) - { - pHost += sizeof("Host: ")-1; - Terminate(pHost); - } - - if(pHttp && pGet) - { - *pHttp = '\0'; - pGet += sizeof("GET /")-1; - Terminate(pGet); - int nFrames = MicroProfileParseGet(pGet); - if(nFrames) - { - uint64_t nTickStart = MP_TICK(); - send(Connection, MICROPROFILE_HTML_HEADER, sizeof(MICROPROFILE_HTML_HEADER)-1, 0); - uint64_t nDataStart = S.nWebServerDataSent; - S.WebServerPut = 0; - #if 0 == MICROPROFILE_MINIZ - MicroProfileDumpHtml(MicroProfileWriteSocket, &Connection, nFrames, pHost); - uint64_t nDataEnd = S.nWebServerDataSent; - uint64_t nTickEnd = MP_TICK(); - uint64_t nDiff = (nTickEnd - nTickStart); - float fMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()) * nDiff; - int nKb = ((nDataEnd-nDataStart)>>10) + 1; - int nCompressedKb = nKb; - MicroProfilePrintf(MicroProfileWriteSocket, &Connection, "\n\n\n",nKb, fMs); - MicroProfileFlushSocket(Connection); - #else - MicroProfileCompressedSocketState CompressState; - MicroProfileCompressedSocketStart(&CompressState, Connection); - MicroProfileDumpHtml(MicroProfileCompressedWriteSocket, &CompressState, nFrames, pHost); - S.nWebServerDataSent += CompressState.nSize; - uint64_t nDataEnd = S.nWebServerDataSent; - uint64_t nTickEnd = MP_TICK(); - uint64_t nDiff = (nTickEnd - nTickStart); - float fMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()) * nDiff; - int nKb = ((nDataEnd-nDataStart)>>10) + 1; - int nCompressedKb = ((CompressState.nCompressedSize)>>10) + 1; - MicroProfilePrintf(MicroProfileCompressedWriteSocket, &CompressState, "\n\n\n", nKb, nCompressedKb, fMs); - MicroProfileCompressedSocketFinish(&CompressState); - MicroProfileFlushSocket(Connection); - #endif - - #if MICROPROFILE_DEBUG - printf("\n\n\n", nKb, nCompressedKb, fMs); - #endif - } - } - } -#ifdef _WIN32 - closesocket(Connection); -#else - close(Connection); -#endif - } - return bServed; -} -#endif - - - - -#if MICROPROFILE_CONTEXT_SWITCH_TRACE -//functions that need to be implemented per platform. -void* MicroProfileTraceThread(void* unused); -bool MicroProfileIsLocalThread(uint32_t nThreadId); - - -void MicroProfileStartContextSwitchTrace() -{ - if(!S.bContextSwitchRunning) - { - S.bContextSwitchRunning = true; - S.bContextSwitchStop = false; - MicroProfileThreadStart(&S.ContextSwitchThread, MicroProfileTraceThread); - } -} - -void MicroProfileStopContextSwitchTrace() -{ - if(S.bContextSwitchRunning) - { - S.bContextSwitchStop = true; - MicroProfileThreadJoin(&S.ContextSwitchThread); - } -} - - -#ifdef _WIN32 -#define INITGUID -#include -#include -#include - - -static GUID g_MicroProfileThreadClassGuid = { 0x3d6fa8d1, 0xfe05, 0x11d0, 0x9d, 0xda, 0x00, 0xc0, 0x4f, 0xd7, 0xba, 0x7c }; - -struct MicroProfileSCSwitch -{ - uint32_t NewThreadId; - uint32_t OldThreadId; - int8_t NewThreadPriority; - int8_t OldThreadPriority; - uint8_t PreviousCState; - int8_t SpareByte; - int8_t OldThreadWaitReason; - int8_t OldThreadWaitMode; - int8_t OldThreadState; - int8_t OldThreadWaitIdealProcessor; - uint32_t NewThreadWaitTime; - uint32_t Reserved; -}; - - -VOID WINAPI MicroProfileContextSwitchCallback(PEVENT_TRACE pEvent) -{ - if (pEvent->Header.Guid == g_MicroProfileThreadClassGuid) - { - if (pEvent->Header.Class.Type == 36) - { - MicroProfileSCSwitch* pCSwitch = (MicroProfileSCSwitch*) pEvent->MofData; - if ((pCSwitch->NewThreadId != 0) || (pCSwitch->OldThreadId != 0)) - { - MicroProfileContextSwitch Switch; - Switch.nThreadOut = pCSwitch->OldThreadId; - Switch.nThreadIn = pCSwitch->NewThreadId; - Switch.nCpu = pEvent->BufferContext.ProcessorNumber; - Switch.nTicks = pEvent->Header.TimeStamp.QuadPart; - MicroProfileContextSwitchPut(&Switch); - } - } - } -} - -ULONG WINAPI MicroProfileBufferCallback(PEVENT_TRACE_LOGFILE Buffer) -{ - return (S.bContextSwitchStop || !S.bContextSwitchRunning) ? FALSE : TRUE; -} - - -struct MicroProfileKernelTraceProperties : public EVENT_TRACE_PROPERTIES -{ - char dummy[sizeof(KERNEL_LOGGER_NAME)]; -}; - -void MicroProfileContextSwitchShutdownTrace() -{ - TRACEHANDLE SessionHandle = 0; - MicroProfileKernelTraceProperties sessionProperties; - - ZeroMemory(&sessionProperties, sizeof(sessionProperties)); - sessionProperties.Wnode.BufferSize = sizeof(sessionProperties); - sessionProperties.Wnode.Flags = WNODE_FLAG_TRACED_GUID; - sessionProperties.Wnode.ClientContext = 1; //QPC clock resolution - sessionProperties.Wnode.Guid = SystemTraceControlGuid; - sessionProperties.BufferSize = 1; - sessionProperties.NumberOfBuffers = 128; - sessionProperties.EnableFlags = EVENT_TRACE_FLAG_CSWITCH; - sessionProperties.LogFileMode = EVENT_TRACE_REAL_TIME_MODE; - sessionProperties.MaximumFileSize = 0; - sessionProperties.LoggerNameOffset = sizeof(EVENT_TRACE_PROPERTIES); - sessionProperties.LogFileNameOffset = 0; - - EVENT_TRACE_LOGFILE log; - ZeroMemory(&log, sizeof(log)); - log.LoggerName = KERNEL_LOGGER_NAME; - log.ProcessTraceMode = 0; - TRACEHANDLE hLog = OpenTrace(&log); - if (hLog) - { - ControlTrace(SessionHandle, KERNEL_LOGGER_NAME, &sessionProperties, EVENT_TRACE_CONTROL_STOP); - } - CloseTrace(hLog); - - -} - -void* MicroProfileTraceThread(void* unused) -{ - - MicroProfileContextSwitchShutdownTrace(); - ULONG status = ERROR_SUCCESS; - TRACEHANDLE SessionHandle = 0; - MicroProfileKernelTraceProperties sessionProperties; - - ZeroMemory(&sessionProperties, sizeof(sessionProperties)); - sessionProperties.Wnode.BufferSize = sizeof(sessionProperties); - sessionProperties.Wnode.Flags = WNODE_FLAG_TRACED_GUID; - sessionProperties.Wnode.ClientContext = 1; //QPC clock resolution - sessionProperties.Wnode.Guid = SystemTraceControlGuid; - sessionProperties.BufferSize = 1; - sessionProperties.NumberOfBuffers = 128; - sessionProperties.EnableFlags = EVENT_TRACE_FLAG_CSWITCH|EVENT_TRACE_FLAG_PROCESS; - sessionProperties.LogFileMode = EVENT_TRACE_REAL_TIME_MODE; - sessionProperties.MaximumFileSize = 0; - sessionProperties.LoggerNameOffset = sizeof(EVENT_TRACE_PROPERTIES); - sessionProperties.LogFileNameOffset = 0; - - - status = StartTrace((PTRACEHANDLE) &SessionHandle, KERNEL_LOGGER_NAME, &sessionProperties); - - if (ERROR_SUCCESS != status) - { - S.bContextSwitchRunning = false; - return 0; - } - - EVENT_TRACE_LOGFILE log; - ZeroMemory(&log, sizeof(log)); - - log.LoggerName = KERNEL_LOGGER_NAME; - log.ProcessTraceMode = PROCESS_TRACE_MODE_REAL_TIME | PROCESS_TRACE_MODE_RAW_TIMESTAMP; - log.EventCallback = MicroProfileContextSwitchCallback; - log.BufferCallback = MicroProfileBufferCallback; - - TRACEHANDLE hLog = OpenTrace(&log); - ProcessTrace(&hLog, 1, 0, 0); - CloseTrace(hLog); - MicroProfileContextSwitchShutdownTrace(); - - S.bContextSwitchRunning = false; - return 0; -} - -bool MicroProfileIsLocalThread(uint32_t nThreadId) -{ - HANDLE h = OpenThread(THREAD_QUERY_LIMITED_INFORMATION, FALSE, nThreadId); - if(h == NULL) - return false; - DWORD hProcess = GetProcessIdOfThread(h); - CloseHandle(h); - return GetCurrentProcessId() == hProcess; -} - -#elif defined(__APPLE__) -#include -void* MicroProfileTraceThread(void* unused) -{ - FILE* pFile = fopen("mypipe", "r"); - if(!pFile) - { - printf("CONTEXT SWITCH FAILED TO OPEN FILE: make sure to run dtrace script\n"); - S.bContextSwitchRunning = false; - return 0; - } - printf("STARTING TRACE THREAD\n"); - char* pLine = 0; - size_t cap = 0; - size_t len = 0; - struct timeval tv; - - gettimeofday(&tv, NULL); - - uint64_t nsSinceEpoch = ((uint64_t)(tv.tv_sec) * 1000000 + (uint64_t)(tv.tv_usec)) * 1000; - uint64_t nTickEpoch = MP_TICK(); - uint32_t nLastThread[MICROPROFILE_MAX_CONTEXT_SWITCH_THREADS] = {0}; - mach_timebase_info_data_t sTimebaseInfo; - mach_timebase_info(&sTimebaseInfo); - S.bContextSwitchRunning = true; - - uint64_t nProcessed = 0; - uint64_t nProcessedLast = 0; - while((len = getline(&pLine, &cap, pFile))>0 && !S.bContextSwitchStop) - { - nProcessed += len; - if(nProcessed - nProcessedLast > 10<<10) - { - nProcessedLast = nProcessed; - printf("processed %llukb %llukb\n", (nProcessed-nProcessedLast)>>10,nProcessed >>10); - } - - char* pX = strchr(pLine, 'X'); - if(pX) - { - int cpu = atoi(pX+1); - char* pX2 = strchr(pX + 1, 'X'); - char* pX3 = strchr(pX2 + 1, 'X'); - int thread = atoi(pX2+1); - char* lala; - int64_t timestamp = strtoll(pX3 + 1, &lala, 10); - MicroProfileContextSwitch Switch; - - //convert to ticks. - uint64_t nDeltaNsSinceEpoch = timestamp - nsSinceEpoch; - uint64_t nDeltaTickSinceEpoch = sTimebaseInfo.numer * nDeltaNsSinceEpoch / sTimebaseInfo.denom; - uint64_t nTicks = nDeltaTickSinceEpoch + nTickEpoch; - if(cpu < MICROPROFILE_MAX_CONTEXT_SWITCH_THREADS) - { - Switch.nThreadOut = nLastThread[cpu]; - Switch.nThreadIn = thread; - nLastThread[cpu] = thread; - Switch.nCpu = cpu; - Switch.nTicks = nTicks; - MicroProfileContextSwitchPut(&Switch); - } - } - } - printf("EXITING TRACE THREAD\n"); - S.bContextSwitchRunning = false; - return 0; -} - -bool MicroProfileIsLocalThread(uint32_t nThreadId) -{ - return false; -} - -#endif -#else - -bool MicroProfileIsLocalThread([[maybe_unused]] uint32_t nThreadId) { return false; } -void MicroProfileStopContextSwitchTrace(){} -void MicroProfileStartContextSwitchTrace(){} - -#endif - - - - -#if MICROPROFILE_GPU_TIMERS_D3D11 -uint32_t MicroProfileGpuInsertTimeStamp() -{ - MicroProfileD3D11Frame& Frame = S.GPU.m_QueryFrames[S.GPU.m_nQueryFrame]; - if(Frame.m_nRateQueryStarted) - { - uint32_t nCurrent = (Frame.m_nQueryStart + Frame.m_nQueryCount) % MICROPROFILE_D3D_MAX_QUERIES; - uint32_t nNext = (nCurrent + 1) % MICROPROFILE_D3D_MAX_QUERIES; - if(nNext != S.GPU.m_nQueryGet) - { - Frame.m_nQueryCount++; - ID3D11Query* pQuery = (ID3D11Query*)S.GPU.m_pQueries[nCurrent]; - ID3D11DeviceContext* pContext = (ID3D11DeviceContext*)S.GPU.m_pDeviceContext; - pContext->End(pQuery); - S.GPU.m_nQueryPut = nNext; - return nCurrent; - } - } - return (uint32_t)-1; -} - -uint64_t MicroProfileGpuGetTimeStamp(uint32_t nIndex) -{ - if(nIndex == (uint32_t)-1) - { - return (uint64_t)-1; - } - int64_t nResult = S.GPU.m_nQueryResults[nIndex]; - MP_ASSERT(nResult != -1); - return nResult; -} - -bool MicroProfileGpuGetData(void* pQuery, void* pData, uint32_t nDataSize) -{ - HRESULT hr; - do - { - hr = ((ID3D11DeviceContext*)S.GPU.m_pDeviceContext)->GetData((ID3D11Query*)pQuery, pData, nDataSize, 0); - }while(hr == S_FALSE); - switch(hr) - { - case DXGI_ERROR_DEVICE_REMOVED: - case DXGI_ERROR_INVALID_CALL: - case E_INVALIDARG: - MP_BREAK(); - return false; - - } - return true; -} - -uint64_t MicroProfileTicksPerSecondGpu() -{ - return S.GPU.m_nQueryFrequency; -} - -void MicroProfileGpuFlip() -{ - MicroProfileD3D11Frame& CurrentFrame = S.GPU.m_QueryFrames[S.GPU.m_nQueryFrame]; - ID3D11DeviceContext* pContext = (ID3D11DeviceContext*)S.GPU.m_pDeviceContext; - if(CurrentFrame.m_nRateQueryStarted) - { - pContext->End((ID3D11Query*)CurrentFrame.m_pRateQuery); - } - uint32_t nNextFrame = (S.GPU.m_nQueryFrame + 1) % MICROPROFILE_GPU_FRAME_DELAY; - MicroProfileD3D11Frame& OldFrame = S.GPU.m_QueryFrames[nNextFrame]; - if(OldFrame.m_nRateQueryStarted) - { - struct RateQueryResult - { - uint64_t nFrequency; - BOOL bDisjoint; - }; - RateQueryResult Result; - if(MicroProfileGpuGetData(OldFrame.m_pRateQuery, &Result, sizeof(Result))) - { - if(S.GPU.m_nQueryFrequency != (int64_t)Result.nFrequency) - { - if(S.GPU.m_nQueryFrequency) - { - OutputDebugString("Query freq changing"); - } - S.GPU.m_nQueryFrequency = Result.nFrequency; - } - uint32_t nStart = OldFrame.m_nQueryStart; - uint32_t nCount = OldFrame.m_nQueryCount; - for(uint32_t i = 0; i < nCount; ++i) - { - uint32_t nIndex = (i + nStart) % MICROPROFILE_D3D_MAX_QUERIES; - - - - if(!MicroProfileGpuGetData(S.GPU.m_pQueries[nIndex], &S.GPU.m_nQueryResults[nIndex], sizeof(uint64_t))) - { - S.GPU.m_nQueryResults[nIndex] = -1; - } - } - } - else - { - uint32_t nStart = OldFrame.m_nQueryStart; - uint32_t nCount = OldFrame.m_nQueryCount; - for(uint32_t i = 0; i < nCount; ++i) - { - uint32_t nIndex = (i + nStart) % MICROPROFILE_D3D_MAX_QUERIES; - S.GPU.m_nQueryResults[nIndex] = -1; - } - } - S.GPU.m_nQueryGet = (OldFrame.m_nQueryStart + OldFrame.m_nQueryCount) % MICROPROFILE_D3D_MAX_QUERIES; - } - - S.GPU.m_nQueryFrame = nNextFrame; - MicroProfileD3D11Frame& NextFrame = S.GPU.m_QueryFrames[nNextFrame]; - pContext->Begin((ID3D11Query*)NextFrame.m_pRateQuery); - NextFrame.m_nQueryStart = S.GPU.m_nQueryPut; - NextFrame.m_nQueryCount = 0; - - NextFrame.m_nRateQueryStarted = 1; -} - -void MicroProfileGpuInitD3D11(void* pDevice_, void* pDeviceContext_) -{ - ID3D11Device* pDevice = (ID3D11Device*)pDevice_; - ID3D11DeviceContext* pDeviceContext = (ID3D11DeviceContext*)pDeviceContext_; - S.GPU.m_pDeviceContext = pDeviceContext_; - - D3D11_QUERY_DESC Desc; - Desc.MiscFlags = 0; - Desc.Query = D3D11_QUERY_TIMESTAMP; - for(uint32_t i = 0; i < MICROPROFILE_D3D_MAX_QUERIES; ++i) - { - HRESULT hr = pDevice->CreateQuery(&Desc, (ID3D11Query**)&S.GPU.m_pQueries[i]); - MP_ASSERT(hr == S_OK); - S.GPU.m_nQueryResults[i] = -1; - } - S.GPU.m_nQueryPut = 0; - S.GPU.m_nQueryGet = 0; - S.GPU.m_nQueryFrame = 0; - S.GPU.m_nQueryFrequency = 0; - Desc.Query = D3D11_QUERY_TIMESTAMP_DISJOINT; - for(uint32_t i = 0; i < MICROPROFILE_GPU_FRAME_DELAY; ++i) - { - S.GPU.m_QueryFrames[i].m_nQueryStart = 0; - S.GPU.m_QueryFrames[i].m_nQueryCount = 0; - S.GPU.m_QueryFrames[i].m_nRateQueryStarted = 0; - HRESULT hr = pDevice->CreateQuery(&Desc, (ID3D11Query**)&S.GPU.m_QueryFrames[i].m_pRateQuery); - MP_ASSERT(hr == S_OK); - } -} - - -void MicroProfileGpuShutdown() -{ - for(uint32_t i = 0; i < MICROPROFILE_D3D_MAX_QUERIES; ++i) - { - ((ID3D11Query*)&S.GPU.m_pQueries[i])->Release(); - S.GPU.m_pQueries[i] = 0; - } - for(uint32_t i = 0; i < MICROPROFILE_GPU_FRAME_DELAY; ++i) - { - ((ID3D11Query*)S.GPU.m_QueryFrames[i].m_pRateQuery)->Release(); - S.GPU.m_QueryFrames[i].m_pRateQuery = 0; - } -} - -int MicroProfileGetGpuTickReference(int64_t* pOutCPU, int64_t* pOutGpu) -{ - return 0; -} - - -#elif MICROPROFILE_GPU_TIMERS_GL -void MicroProfileGpuInitGL() -{ - S.GPU.GLTimerPos = 0; - glGenQueries(MICROPROFILE_GL_MAX_QUERIES, &S.GPU.GLTimers[0]); -} - -uint32_t MicroProfileGpuInsertTimeStamp() -{ - uint32_t nIndex = (S.GPU.GLTimerPos+1)%MICROPROFILE_GL_MAX_QUERIES; - glQueryCounter(S.GPU.GLTimers[nIndex], GL_TIMESTAMP); - S.GPU.GLTimerPos = nIndex; - return nIndex; -} -uint64_t MicroProfileGpuGetTimeStamp(uint32_t nKey) -{ - uint64_t result; - glGetQueryObjectui64v(S.GPU.GLTimers[nKey], GL_QUERY_RESULT, &result); - return result; -} - -uint64_t MicroProfileTicksPerSecondGpu() -{ - return 1000000000ll; -} - -int MicroProfileGetGpuTickReference(int64_t* pOutCpu, int64_t* pOutGpu) -{ - int64_t nGpuTimeStamp; - glGetInteger64v(GL_TIMESTAMP, &nGpuTimeStamp); - if(nGpuTimeStamp) - { - *pOutCpu = MP_TICK(); - *pOutGpu = nGpuTimeStamp; - #if 0 //debug test if timestamp diverges - static int64_t nTicksPerSecondCpu = MicroProfileTicksPerSecondCpu(); - static int64_t nTicksPerSecondGpu = MicroProfileTicksPerSecondGpu(); - static int64_t nGpuStart = 0; - static int64_t nCpuStart = 0; - if(!nCpuStart) - { - nCpuStart = *pOutCpu; - nGpuStart = *pOutGpu; - } - static int nCountDown = 100; - if(0 == nCountDown--) - { - int64_t nCurCpu = *pOutCpu; - int64_t nCurGpu = *pOutGpu; - double fDistanceCpu = (nCurCpu - nCpuStart) / (double)nTicksPerSecondCpu; - double fDistanceGpu = (nCurGpu - nGpuStart) / (double)nTicksPerSecondGpu; - - char buf[254]; - snprintf(buf, sizeof(buf)-1,"Distance %f %f diff %f\n", fDistanceCpu, fDistanceGpu, fDistanceCpu-fDistanceGpu); - OutputDebugString(buf); - nCountDown = 100; - } - #endif - return 1; - } - return 0; -} - - -#endif - -#undef S - -#ifdef _MSC_VER -#pragma warning(pop) -#endif - - - - - -#endif -#endif -#ifdef MICROPROFILE_EMBED_HTML -#include "microprofile_html.h" -#endif diff --git a/externals/microprofile/microprofile_html.h b/externals/microprofile/microprofile_html.h deleted file mode 100644 index 01b624b60..000000000 --- a/externals/microprofile/microprofile_html.h +++ /dev/null @@ -1,3868 +0,0 @@ -///start file generated from microprofile.html -#ifdef MICROPROFILE_EMBED_HTML -const char g_MicroProfileHtml_begin_0[] = -"\n" -"\n" -"\n" -"MicroProfile Capture\n" -"\n" -"\n" -"\n" -"\n" -"
\n" -"\n" -"
\n" -"\n" -"\n" -" "; - -const size_t g_MicroProfileHtml_end_2_size = sizeof(g_MicroProfileHtml_end_2); -const char* g_MicroProfileHtml_end[] = { -&g_MicroProfileHtml_end_0[0], -&g_MicroProfileHtml_end_1[0], -&g_MicroProfileHtml_end_2[0], -}; -size_t g_MicroProfileHtml_end_sizes[] = { -sizeof(g_MicroProfileHtml_end_0), -sizeof(g_MicroProfileHtml_end_1), -sizeof(g_MicroProfileHtml_end_2), -}; -size_t g_MicroProfileHtml_end_count = 3; -#endif //MICROPROFILE_EMBED_HTML - -///end file generated from microprofile.html diff --git a/externals/microprofile/microprofileui.h b/externals/microprofile/microprofileui.h deleted file mode 100644 index 0553734b1..000000000 --- a/externals/microprofile/microprofileui.h +++ /dev/null @@ -1,3354 +0,0 @@ -#pragma once -// This is free and unencumbered software released into the public domain. -// Anyone is free to copy, modify, publish, use, compile, sell, or -// distribute this software, either in source code form or as a compiled -// binary, for any purpose, commercial or non-commercial, and by any -// means. -// In jurisdictions that recognize copyright laws, the author or authors -// of this software dedicate any and all copyright interest in the -// software to the public domain. We make this dedication for the benefit -// of the public at large and to the detriment of our heirs and -// successors. We intend this dedication to be an overt act of -// relinquishment in perpetuity of all present and future rights to this -// software under copyright law. -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -// IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR -// OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, -// ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR -// OTHER DEALINGS IN THE SOFTWARE. -// For more information, please refer to -// -// *********************************************************************** -// -// -// - - -#ifndef MICROPROFILE_ENABLED -#error "microprofile.h must be included before including microprofileui.h" -#endif - -#ifndef MICROPROFILEUI_ENABLED -#define MICROPROFILEUI_ENABLED MICROPROFILE_ENABLED -#endif - -#ifndef MICROPROFILEUI_API -#define MICROPROFILEUI_API -#endif - - -#if 0 == MICROPROFILEUI_ENABLED -#define MicroProfileMouseButton(foo, bar) do{}while(0) -#define MicroProfileMousePosition(foo, bar, z) do{}while(0) -#define MicroProfileModKey(key) do{}while(0) -#define MicroProfileDraw(foo, bar) do{}while(0) -#define MicroProfileIsDrawing() 0 -#define MicroProfileToggleDisplayMode() do{}while(0) -#define MicroProfileSetDisplayMode(f) do{}while(0) -#else - -#ifndef MICROPROFILE_DRAWCURSOR -#define MICROPROFILE_DRAWCURSOR 0 -#endif - -#ifndef MICROPROFILE_DETAILED_BAR_NAMES -#define MICROPROFILE_DETAILED_BAR_NAMES 1 -#endif - -#ifndef MICROPROFILE_TEXT_WIDTH -#define MICROPROFILE_TEXT_WIDTH 5 -#endif - -#ifndef MICROPROFILE_TEXT_HEIGHT -#define MICROPROFILE_TEXT_HEIGHT 8 -#endif - -#ifndef MICROPROFILE_DETAILED_BAR_HEIGHT -#define MICROPROFILE_DETAILED_BAR_HEIGHT 12 -#endif - -#ifndef MICROPROFILE_DETAILED_CONTEXT_SWITCH_HEIGHT -#define MICROPROFILE_DETAILED_CONTEXT_SWITCH_HEIGHT 7 -#endif - -#ifndef MICROPROFILE_GRAPH_WIDTH -#define MICROPROFILE_GRAPH_WIDTH 256 -#endif - -#ifndef MICROPROFILE_GRAPH_HEIGHT -#define MICROPROFILE_GRAPH_HEIGHT 256 -#endif - -#ifndef MICROPROFILE_BORDER_SIZE -#define MICROPROFILE_BORDER_SIZE 1 -#endif - -#ifndef MICROPROFILE_HELP_LEFT -#define MICROPROFILE_HELP_LEFT "Left-Click" -#endif - -#ifndef MICROPROFILE_HELP_ALT -#define MICROPROFILE_HELP_ALT "Alt-Click" -#endif - -#ifndef MICROPROFILE_HELP_MOD -#define MICROPROFILE_HELP_MOD "Mod" -#endif - -#ifndef MICROPROFILE_BAR_WIDTH -#define MICROPROFILE_BAR_WIDTH 100 -#endif - -#ifndef MICROPROFILE_CUSTOM_MAX -#define MICROPROFILE_CUSTOM_MAX 8 -#endif - -#ifndef MICROPROFILE_CUSTOM_MAX_TIMERS -#define MICROPROFILE_CUSTOM_MAX_TIMERS 64 -#endif - -#ifndef MICROPROFILE_CUSTOM_PADDING -#define MICROPROFILE_CUSTOM_PADDING 12 -#endif - - -#define MICROPROFILE_FRAME_HISTORY_HEIGHT 50 -#define MICROPROFILE_FRAME_HISTORY_WIDTH 7 -#define MICROPROFILE_FRAME_HISTORY_COLOR_CPU 0xffff7f27 //255 127 39 -#define MICROPROFILE_FRAME_HISTORY_COLOR_GPU 0xff37a0ee //55 160 238 -#define MICROPROFILE_FRAME_HISTORY_COLOR_HIGHTLIGHT 0x7733bb44 -#define MICROPROFILE_FRAME_COLOR_HIGHTLIGHT 0x20009900 -#define MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU 0x20996600 -#define MICROPROFILE_NUM_FRAMES (MICROPROFILE_MAX_FRAME_HISTORY - (MICROPROFILE_GPU_FRAME_DELAY+1)) - -#define MICROPROFILE_TOOLTIP_MAX_STRINGS (32 + MICROPROFILE_MAX_GROUPS*2) -#define MICROPROFILE_TOOLTIP_STRING_BUFFER_SIZE (4*1024) -#define MICROPROFILE_TOOLTIP_MAX_LOCKED 3 - - -enum -{ - MICROPROFILE_CUSTOM_BARS = 0x1, - MICROPROFILE_CUSTOM_BAR_SOURCE_MAX = 0x2, - MICROPROFILE_CUSTOM_BAR_SOURCE_AVG = 0, - MICROPROFILE_CUSTOM_STACK = 0x4, - MICROPROFILE_CUSTOM_STACK_SOURCE_MAX = 0x8, - MICROPROFILE_CUSTOM_STACK_SOURCE_AVG = 0, -}; - - -MICROPROFILEUI_API void MicroProfileDraw(uint32_t nWidth, uint32_t nHeight); //! call if drawing microprofilers -MICROPROFILEUI_API bool MicroProfileIsDrawing(); -MICROPROFILEUI_API void MicroProfileToggleGraph(MicroProfileToken nToken); -MICROPROFILEUI_API bool MicroProfileDrawGraph(uint32_t nScreenWidth, uint32_t nScreenHeight); -MICROPROFILEUI_API void MicroProfileToggleDisplayMode(); //switch between off, bars, detailed -MICROPROFILEUI_API void MicroProfileSetDisplayMode(int); //switch between off, bars, detailed -MICROPROFILEUI_API void MicroProfileClearGraph(); -MICROPROFILEUI_API void MicroProfileMousePosition(uint32_t nX, uint32_t nY, int nWheelDelta); -MICROPROFILEUI_API void MicroProfileModKey(uint32_t nKeyState); -MICROPROFILEUI_API void MicroProfileMouseButton(uint32_t nLeft, uint32_t nRight); -MICROPROFILEUI_API void MicroProfileDrawLineVertical(int nX, int nTop, int nBottom, uint32_t nColor); -MICROPROFILEUI_API void MicroProfileDrawLineHorizontal(int nLeft, int nRight, int nY, uint32_t nColor); -MICROPROFILEUI_API void MicroProfileLoadPreset(const char* pSuffix); -MICROPROFILEUI_API void MicroProfileSavePreset(const char* pSuffix); - -MICROPROFILEUI_API void MicroProfileDrawText(int nX, int nY, uint32_t nColor, const char* pText, uint32_t nNumCharacters); -MICROPROFILEUI_API void MicroProfileDrawBox(int nX, int nY, int nX1, int nY1, uint32_t nColor, MicroProfileBoxType = MicroProfileBoxTypeFlat); -MICROPROFILEUI_API void MicroProfileDrawLine2D(uint32_t nVertices, float* pVertices, uint32_t nColor); -MICROPROFILEUI_API void MicroProfileDumpTimers(); - -MICROPROFILEUI_API void MicroProfileInitUI(); - -MICROPROFILEUI_API void MicroProfileCustomGroupToggle(const char* pCustomName); -MICROPROFILEUI_API void MicroProfileCustomGroupEnable(const char* pCustomName); -MICROPROFILEUI_API void MicroProfileCustomGroupEnable(uint32_t nIndex); -MICROPROFILEUI_API void MicroProfileCustomGroupDisable(); -MICROPROFILEUI_API void MicroProfileCustomGroup(const char* pCustomName, uint32_t nMaxTimers, uint32_t nAggregateFlip, float fReferenceTime, uint32_t nFlags); -MICROPROFILEUI_API void MicroProfileCustomGroupAddTimer(const char* pCustomName, const char* pGroup, const char* pTimer); - -#ifdef MICROPROFILEUI_IMPL -#include -#include -#include -#include -#include -#include -#include - -MICROPROFILE_DEFINE(g_MicroProfileDetailed, "MicroProfile", "Detailed View", 0x8888000); -MICROPROFILE_DEFINE(g_MicroProfileDrawGraph, "MicroProfile", "Draw Graph", 0xff44ee00); -MICROPROFILE_DEFINE(g_MicroProfileDrawBarView, "MicroProfile", "DrawBarView", 0x00dd77); -MICROPROFILE_DEFINE(g_MicroProfileDraw,"MicroProfile", "Draw", 0x737373); - - -struct MicroProfileStringArray -{ - const char* ppStrings[MICROPROFILE_TOOLTIP_MAX_STRINGS]; - char Buffer[MICROPROFILE_TOOLTIP_STRING_BUFFER_SIZE]; - char* pBufferPos; - uint32_t nNumStrings; -}; - -struct MicroProfileGroupMenuItem -{ - uint32_t nIsCategory; - uint32_t nCategoryIndex; - uint32_t nIndex; - const char* pName; -}; - -struct MicroProfileCustom -{ - char pName[MICROPROFILE_NAME_MAX_LEN]; - uint32_t nFlags; - uint32_t nAggregateFlip; - uint32_t nNumTimers; - uint32_t nMaxTimers; - uint64_t nGroupMask; - float fReference; - uint64_t* pTimers; -}; - -struct SOptionDesc -{ - SOptionDesc(){} - SOptionDesc(uint8_t nSubType, uint8_t nIndex, const char* fmt, ...):nSubType(nSubType), nIndex(nIndex) - { - va_list args; - va_start (args, fmt); - vsprintf(Text, fmt, args); - va_end(args); - } - char Text[32]; - uint8_t nSubType; - uint8_t nIndex; - bool bSelected; -}; -static const std::array g_MicroProfileAggregatePresets{0, 10, 20, 30, 60, 120}; -static const std::array g_MicroProfileReferenceTimePresets{5.f, 10.f, 15.f,20.f, 33.33f, 66.66f, 100.f, 250.f, 500.f, 1000.f}; -static const std::array g_MicroProfileOpacityPresets{0x40, 0x80, 0xc0, 0xff}; -static const std::array g_MicroProfilePresetNames -{ - MICROPROFILE_DEFAULT_PRESET, - "Render", - "GPU", - "Lighting", - "AI", - "Visibility", - "Sound", -}; - -enum -{ - MICROPROFILE_NUM_REFERENCE_PRESETS = g_MicroProfileReferenceTimePresets.size(), - MICROPROFILE_NUM_OPACITY_PRESETS = g_MicroProfileOpacityPresets.size(), -#if MICROPROFILE_CONTEXT_SWITCH_TRACE - MICROPROFILE_OPTION_SIZE = MICROPROFILE_NUM_REFERENCE_PRESETS + MICROPROFILE_NUM_OPACITY_PRESETS * 2 + 2 + 7, -#else - MICROPROFILE_OPTION_SIZE = MICROPROFILE_NUM_REFERENCE_PRESETS + MICROPROFILE_NUM_OPACITY_PRESETS * 2 + 2 + 3, -#endif -}; - -struct MicroProfileUI -{ - //menu/mouse over stuff - uint64_t nHoverToken; - int64_t nHoverTime; - int nHoverFrame; -#if MICROPROFILE_DEBUG - uint64_t nHoverAddressEnter; - uint64_t nHoverAddressLeave; -#endif - - uint32_t nWidth; - uint32_t nHeight; - - - int nOffsetX; - int nOffsetY; - float fDetailedOffset; //display offset relative to start of latest displayable frame. - float fDetailedRange; //no. of ms to display - float fDetailedOffsetTarget; - float fDetailedRangeTarget; - uint32_t nOpacityBackground; - uint32_t nOpacityForeground; - bool bShowSpikes; - - - - uint32_t nMouseX; - uint32_t nMouseY; - uint32_t nMouseDownX; - uint32_t nMouseDownY; - int nMouseWheelDelta; - uint32_t nMouseDownLeft; - uint32_t nMouseDownRight; - uint32_t nMouseLeft; - uint32_t nMouseRight; - uint32_t nMouseLeftMod; - uint32_t nMouseRightMod; - uint32_t nModDown; - uint32_t nActiveMenu; - - MicroProfileLogEntry* pDisplayMouseOver; - - int64_t nRangeBegin; - int64_t nRangeEnd; - int64_t nRangeBeginGpu; - int64_t nRangeEndGpu; - uint32_t nRangeBeginIndex; - uint32_t nRangeEndIndex; - MicroProfileThreadLog* pRangeLog; - uint32_t nHoverColor; - uint32_t nHoverColorShared; - - MicroProfileStringArray LockedToolTips[MICROPROFILE_TOOLTIP_MAX_LOCKED]; - uint32_t nLockedToolTipColor[MICROPROFILE_TOOLTIP_MAX_LOCKED]; - int LockedToolTipFront; - - MicroProfileGroupMenuItem GroupMenu[MICROPROFILE_MAX_GROUPS + MICROPROFILE_MAX_CATEGORIES]; - uint32_t GroupMenuCount; - - - uint32_t nCustomActive; - uint32_t nCustomTimerCount; - uint32_t nCustomCount; - MicroProfileCustom Custom[MICROPROFILE_CUSTOM_MAX]; - uint64_t CustomTimer[MICROPROFILE_CUSTOM_MAX_TIMERS]; - - SOptionDesc Options[MICROPROFILE_OPTION_SIZE]; - - -}; - -MicroProfileUI g_MicroProfileUI; -#define UI g_MicroProfileUI -static const std::array g_nMicroProfileBackColors{ 0x474747, 0x313131 }; -#define MICROPROFILE_NUM_CONTEXT_SWITCH_COLORS 16 -static const std::array g_nMicroProfileContextSwitchThreadColors //palette generated by http://tools.medialab.sciences-po.fr/iwanthue/index.php -{ - 0x63607B, - 0x755E2B, - 0x326A55, - 0x523135, - 0x904F42, - 0x87536B, - 0x346875, - 0x5E6046, - 0x35404C, - 0x224038, - 0x413D1E, - 0x5E3A26, - 0x5D6161, - 0x4C6234, - 0x7D564F, - 0x5C4352, -}; - - -void MicroProfileInitUI() -{ - static bool bInitialized = false; - if(!bInitialized) - { - bInitialized = true; - g_MicroProfileUI = {}; - UI.nActiveMenu = UINT32_MAX; - UI.fDetailedOffsetTarget = UI.fDetailedOffset = 0.f; - UI.fDetailedRangeTarget = UI.fDetailedRange = 50.f; - - UI.nOpacityBackground = 0xff<<24; - UI.nOpacityForeground = 0xff<<24; - - UI.bShowSpikes = false; - - UI.nWidth = 100; - UI.nHeight = 100; - - UI.nCustomActive = UINT32_MAX; - UI.nCustomTimerCount = 0; - UI.nCustomCount = 0; - - int nIndex = 0; - UI.Options[nIndex++] = SOptionDesc(0xff, 0, "%s", "Reference"); - for(int i = 0; i < MICROPROFILE_NUM_REFERENCE_PRESETS; ++i) - { - UI.Options[nIndex++] = SOptionDesc(0, i, " %6.2fms", g_MicroProfileReferenceTimePresets[i]); - } - UI.Options[nIndex++] = SOptionDesc(0xff, 0, "%s", "BG Opacity"); - for(int i = 0; i < MICROPROFILE_NUM_OPACITY_PRESETS; ++i) - { - UI.Options[nIndex++] = SOptionDesc(1, i, " %7d%%", (i+1)*25); - } - UI.Options[nIndex++] = SOptionDesc(0xff, 0, "%s", "FG Opacity"); - for(int i = 0; i < MICROPROFILE_NUM_OPACITY_PRESETS; ++i) - { - UI.Options[nIndex++] = SOptionDesc(2, i, " %7d%%", (i+1)*25); - } - UI.Options[nIndex++] = SOptionDesc(0xff, 0, "%s", "Spike Display"); - UI.Options[nIndex++] = SOptionDesc(3, 0, "%s", " Enable"); - -#if MICROPROFILE_CONTEXT_SWITCH_TRACE - UI.Options[nIndex++] = SOptionDesc(0xff, 0, "%s", "CSwitch Trace"); - UI.Options[nIndex++] = SOptionDesc(4, 0, "%s", " Enable"); - UI.Options[nIndex++] = SOptionDesc(4, 1, "%s", " All Threads"); - UI.Options[nIndex++] = SOptionDesc(4, 2, "%s", " No Bars"); -#endif - MP_ASSERT(nIndex == MICROPROFILE_OPTION_SIZE); - } -} - -void MicroProfileSetDisplayMode(int nValue) -{ - MicroProfile& S = *MicroProfileGet(); - nValue = nValue >= 0 && nValue < 4 ? nValue : S.nDisplay; - S.nDisplay = nValue; - UI.nOffsetY = 0; -} - -void MicroProfileToggleDisplayMode() -{ - MicroProfile& S = *MicroProfileGet(); - S.nDisplay = (S.nDisplay + 1) % 4; - UI.nOffsetY = 0; -} - - -void MicroProfileStringArrayClear(MicroProfileStringArray* pArray) -{ - pArray->nNumStrings = 0; - pArray->pBufferPos = &pArray->Buffer[0]; -} - -void MicroProfileStringArrayAddLiteral(MicroProfileStringArray* pArray, const char* pLiteral) -{ - MP_ASSERT(pArray->nNumStrings < MICROPROFILE_TOOLTIP_MAX_STRINGS); - pArray->ppStrings[pArray->nNumStrings++] = pLiteral; -} - -void MicroProfileStringArrayFormat(MicroProfileStringArray* pArray, const char* fmt, ...) -{ - MP_ASSERT(pArray->nNumStrings < MICROPROFILE_TOOLTIP_MAX_STRINGS); - pArray->ppStrings[pArray->nNumStrings++] = pArray->pBufferPos; - va_list args; - va_start (args, fmt); - pArray->pBufferPos += 1 + vsprintf(pArray->pBufferPos, fmt, args); - va_end(args); - MP_ASSERT(pArray->pBufferPos < pArray->Buffer + MICROPROFILE_TOOLTIP_STRING_BUFFER_SIZE); -} -void MicroProfileStringArrayCopy(MicroProfileStringArray* pDest, MicroProfileStringArray* pSrc) -{ - memcpy(&pDest->ppStrings[0], &pSrc->ppStrings[0], sizeof(pDest->ppStrings)); - memcpy(&pDest->Buffer[0], &pSrc->Buffer[0], sizeof(pDest->Buffer)); - for(uint32_t i = 0; i < MICROPROFILE_TOOLTIP_MAX_STRINGS; ++i) - { - if(i < pSrc->nNumStrings) - { - if(pSrc->ppStrings[i] >= &pSrc->Buffer[0] && pSrc->ppStrings[i] < &pSrc->Buffer[0] + MICROPROFILE_TOOLTIP_STRING_BUFFER_SIZE) - { - pDest->ppStrings[i] += &pDest->Buffer[0] - &pSrc->Buffer[0]; - } - } - } - pDest->nNumStrings = pSrc->nNumStrings; -} - -void MicroProfileFloatWindowSize(const char** ppStrings, uint32_t nNumStrings, uint32_t* pColors, uint32_t& nWidth, uint32_t& nHeight, uint32_t* pStringLengths = 0) -{ - uint32_t* nStringLengths = pStringLengths ? pStringLengths : (uint32_t*)alloca(nNumStrings * sizeof(uint32_t)); - uint32_t nTextCount = nNumStrings/2; - for(uint32_t i = 0; i < nTextCount; ++i) - { - uint32_t i0 = i * 2; - uint32_t s0, s1; - nStringLengths[i0] = s0 = (uint32_t)strlen(ppStrings[i0]); - nStringLengths[i0+1] = s1 = (uint32_t)strlen(ppStrings[i0+1]); - nWidth = MicroProfileMax(s0+s1, nWidth); - } - nWidth = (MICROPROFILE_TEXT_WIDTH+1) * (2+nWidth) + 2 * MICROPROFILE_BORDER_SIZE; - if(pColors) - nWidth += MICROPROFILE_TEXT_WIDTH + 1; - nHeight = (MICROPROFILE_TEXT_HEIGHT+1) * nTextCount + 2 * MICROPROFILE_BORDER_SIZE; -} - -void MicroProfileDrawFloatWindow(uint32_t nX, uint32_t nY, const char** ppStrings, uint32_t nNumStrings, uint32_t nColor, uint32_t* pColors = 0) -{ - uint32_t nWidth = 0, nHeight = 0; - uint32_t* nStringLengths = (uint32_t*)alloca(nNumStrings * sizeof(uint32_t)); - MicroProfileFloatWindowSize(ppStrings, nNumStrings, pColors, nWidth, nHeight, nStringLengths); - uint32_t nTextCount = nNumStrings/2; - if(nX + nWidth > UI.nWidth) - nX = UI.nWidth - nWidth; - if(nY + nHeight > UI.nHeight) - nY = UI.nHeight - nHeight; - MicroProfileDrawBox(nX-1, nY-1, nX + nWidth+1, nY + nHeight+1, 0xff000000|nColor); - MicroProfileDrawBox(nX, nY, nX + nWidth, nY + nHeight, 0xff000000); - if(pColors) - { - nX += MICROPROFILE_TEXT_WIDTH+1; - nWidth -= MICROPROFILE_TEXT_WIDTH+1; - } - for(uint32_t i = 0; i < nTextCount; ++i) - { - int i0 = i * 2; - if(pColors) - { - MicroProfileDrawBox(nX-MICROPROFILE_TEXT_WIDTH, nY, nX, nY + MICROPROFILE_TEXT_WIDTH, pColors[i]|0xff000000); - } - MicroProfileDrawText(nX + 1, nY + 1, UINT32_MAX, ppStrings[i0], (uint32_t)strlen(ppStrings[i0])); - MicroProfileDrawText(nX + nWidth - nStringLengths[i0+1] * (MICROPROFILE_TEXT_WIDTH+1), nY + 1, UINT32_MAX, ppStrings[i0+1], (uint32_t)strlen(ppStrings[i0+1])); - nY += (MICROPROFILE_TEXT_HEIGHT+1); - } -} -void MicroProfileDrawTextBox(uint32_t nX, uint32_t nY, const char** ppStrings, uint32_t nNumStrings, uint32_t nColor, uint32_t* pColors = 0) -{ - uint32_t nWidth = 0, nHeight = 0; - uint32_t* nStringLengths = (uint32_t*)alloca(nNumStrings * sizeof(uint32_t)); - for(uint32_t i = 0; i < nNumStrings; ++i) - { - nStringLengths[i] = (uint32_t)strlen(ppStrings[i]); - nWidth = MicroProfileMax(nWidth, nStringLengths[i]); - nHeight++; - } - nWidth = (MICROPROFILE_TEXT_WIDTH+1) * (2+nWidth) + 2 * MICROPROFILE_BORDER_SIZE; - nHeight = (MICROPROFILE_TEXT_HEIGHT+1) * nHeight + 2 * MICROPROFILE_BORDER_SIZE; - if(nX + nWidth > UI.nWidth) - nX = UI.nWidth - nWidth; - if(nY + nHeight > UI.nHeight) - nY = UI.nHeight - nHeight; - MicroProfileDrawBox(nX, nY, nX + nWidth, nY + nHeight, 0xff000000); - for(uint32_t i = 0; i < nNumStrings; ++i) - { - MicroProfileDrawText(nX + 1, nY + 1, UINT32_MAX, ppStrings[i], (uint32_t)strlen(ppStrings[i])); - nY += (MICROPROFILE_TEXT_HEIGHT+1); - } -} - - - -void MicroProfileToolTipMeta(MicroProfileStringArray* pToolTip) -{ - MicroProfile& S = *MicroProfileGet(); - if(UI.nRangeBeginIndex != UI.nRangeEndIndex && UI.pRangeLog) - { - uint64_t nMetaSum[MICROPROFILE_META_MAX] = {0}; - uint64_t nMetaSumInclusive[MICROPROFILE_META_MAX] = {0}; - int nStackDepth = 0; - uint32_t nRange[2][2]; - MicroProfileThreadLog* pLog = UI.pRangeLog; - - - MicroProfileGetRange(UI.nRangeEndIndex, UI.nRangeBeginIndex, nRange); - for(uint32_t i = 0; i < 2; ++i) - { - uint32_t nStart = nRange[i][0]; - uint32_t nEnd = nRange[i][1]; - for(uint32_t j = nStart; j < nEnd; ++j) - { - MicroProfileLogEntry LE = pLog->Log[j]; - int nType = MicroProfileLogType(LE); - switch(nType) - { - case MP_LOG_META: - { - int64_t nMetaIndex = MicroProfileLogTimerIndex(LE); - int64_t nMetaCount = MicroProfileLogGetTick(LE); - MP_ASSERT(nMetaIndex < MICROPROFILE_META_MAX); - if(nStackDepth>1) - { - nMetaSumInclusive[nMetaIndex] += nMetaCount; - } - else - { - nMetaSum[nMetaIndex] += nMetaCount; - } - } - break; - case MP_LOG_LEAVE: - if(nStackDepth) - { - nStackDepth--; - } - else - { - for(int i = 0; i < MICROPROFILE_META_MAX; ++i) - { - nMetaSumInclusive[i] += nMetaSum[i]; - nMetaSum[i] = 0; - } - } - break; - case MP_LOG_ENTER: - nStackDepth++; - break; - } - - } - } - bool bSpaced = false; - for(int i = 0; i < MICROPROFILE_META_MAX; ++i) - { - if(S.MetaCounters[i].pName && (nMetaSum[i]||nMetaSumInclusive[i])) - { - if(!bSpaced) - { - bSpaced = true; - MicroProfileStringArrayAddLiteral(pToolTip, ""); - MicroProfileStringArrayAddLiteral(pToolTip, ""); - } - MicroProfileStringArrayFormat(pToolTip, "%s excl", S.MetaCounters[i].pName); - MicroProfileStringArrayFormat(pToolTip, "%5d", nMetaSum[i]); - MicroProfileStringArrayFormat(pToolTip, "%s incl", S.MetaCounters[i].pName); - MicroProfileStringArrayFormat(pToolTip, "%5d", nMetaSum[i] + nMetaSumInclusive[i]); - } - } - } -} - -void MicroProfileDrawFloatTooltip(uint32_t nX, uint32_t nY, uint32_t nToken, uint64_t nTime) -{ - MicroProfile& S = *MicroProfileGet(); - - uint32_t nIndex = MicroProfileGetTimerIndex(nToken); - uint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1; - uint32_t nAggregateCount = S.Aggregate[nIndex].nCount ? S.Aggregate[nIndex].nCount : 1; - - uint32_t nGroupId = MicroProfileGetGroupIndex(nToken); - uint32_t nTimerId = MicroProfileGetTimerIndex(nToken); - bool bGpu = S.GroupInfo[nGroupId].Type == MicroProfileTokenTypeGpu; - - float fToMs = MicroProfileTickToMsMultiplier(bGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu()); - - float fMs = fToMs * (nTime); - float fFrameMs = fToMs * (S.Frame[nIndex].nTicks); - float fAverage = fToMs * (S.Aggregate[nIndex].nTicks/nAggregateFrames); - float fCallAverage = fToMs * (S.Aggregate[nIndex].nTicks / nAggregateCount); - float fMax = fToMs * (S.AggregateMax[nIndex]); - - float fFrameMsExclusive = fToMs * (S.FrameExclusive[nIndex]); - float fAverageExclusive = fToMs * (S.AggregateExclusive[nIndex]/nAggregateFrames); - float fMaxExclusive = fToMs * (S.AggregateMaxExclusive[nIndex]); - - float fGroupAverage = fToMs * (S.AggregateGroup[nGroupId] / nAggregateFrames); - float fGroupMax = fToMs * (S.AggregateGroupMax[nGroupId]); - float fGroup = fToMs * (S.FrameGroup[nGroupId]); - - - MicroProfileStringArray ToolTip; - MicroProfileStringArrayClear(&ToolTip); - const char* pGroupName = S.GroupInfo[nGroupId].pName; - const char* pTimerName = S.TimerInfo[nTimerId].pName; - MicroProfileStringArrayAddLiteral(&ToolTip, "Timer:"); - MicroProfileStringArrayFormat(&ToolTip, "%s", pTimerName); - -#if MICROPROFILE_DEBUG - MicroProfileStringArrayFormat(&ToolTip,"0x%p", UI.nHoverAddressEnter); - MicroProfileStringArrayFormat(&ToolTip,"0x%p", UI.nHoverAddressLeave); -#endif - - if(nTime != (uint64_t)0) - { - MicroProfileStringArrayAddLiteral(&ToolTip, "Time:"); - MicroProfileStringArrayFormat(&ToolTip,"%6.3fms", fMs); - MicroProfileStringArrayAddLiteral(&ToolTip, ""); - MicroProfileStringArrayAddLiteral(&ToolTip, ""); - } - - MicroProfileStringArrayAddLiteral(&ToolTip, "Frame Time:"); - MicroProfileStringArrayFormat(&ToolTip,"%6.3fms", fFrameMs); - - MicroProfileStringArrayAddLiteral(&ToolTip, "Average:"); - MicroProfileStringArrayFormat(&ToolTip,"%6.3fms", fAverage); - - MicroProfileStringArrayAddLiteral(&ToolTip, "Max:"); - MicroProfileStringArrayFormat(&ToolTip,"%6.3fms", fMax); - - MicroProfileStringArrayAddLiteral(&ToolTip, ""); - MicroProfileStringArrayAddLiteral(&ToolTip, ""); - - MicroProfileStringArrayAddLiteral(&ToolTip, "Frame Call Average:"); - MicroProfileStringArrayFormat(&ToolTip,"%6.3fms", fCallAverage); - - MicroProfileStringArrayAddLiteral(&ToolTip, "Frame Call Count:"); - MicroProfileStringArrayFormat(&ToolTip, "%6d", nAggregateCount / nAggregateFrames); - - MicroProfileStringArrayAddLiteral(&ToolTip, ""); - MicroProfileStringArrayAddLiteral(&ToolTip, ""); - - MicroProfileStringArrayAddLiteral(&ToolTip, "Exclusive Frame Time:"); - MicroProfileStringArrayFormat(&ToolTip, "%6.3fms", fFrameMsExclusive); - - MicroProfileStringArrayAddLiteral(&ToolTip, "Exclusive Average:"); - MicroProfileStringArrayFormat(&ToolTip, "%6.3fms", fAverageExclusive); - - MicroProfileStringArrayAddLiteral(&ToolTip, "Exclusive Max:"); - MicroProfileStringArrayFormat(&ToolTip, "%6.3fms", fMaxExclusive); - - MicroProfileStringArrayAddLiteral(&ToolTip, ""); - MicroProfileStringArrayAddLiteral(&ToolTip, ""); - - MicroProfileStringArrayAddLiteral(&ToolTip, "Group:"); - MicroProfileStringArrayFormat(&ToolTip, "%s", pGroupName); - MicroProfileStringArrayAddLiteral(&ToolTip, "Frame Time:"); - MicroProfileStringArrayFormat(&ToolTip, "%6.3f", fGroup); - MicroProfileStringArrayAddLiteral(&ToolTip, "Frame Average:"); - MicroProfileStringArrayFormat(&ToolTip, "%6.3f", fGroupAverage); - MicroProfileStringArrayAddLiteral(&ToolTip, "Frame Max:"); - MicroProfileStringArrayFormat(&ToolTip, "%6.3f", fGroupMax); - - - - - MicroProfileToolTipMeta(&ToolTip); - - - MicroProfileDrawFloatWindow(nX, nY+20, &ToolTip.ppStrings[0], ToolTip.nNumStrings, S.TimerInfo[nTimerId].nColor); - - if(UI.nMouseLeftMod) - { - int nIndex = (g_MicroProfileUI.LockedToolTipFront + MICROPROFILE_TOOLTIP_MAX_LOCKED - 1) % MICROPROFILE_TOOLTIP_MAX_LOCKED; - g_MicroProfileUI.nLockedToolTipColor[nIndex] = S.TimerInfo[nTimerId].nColor; - MicroProfileStringArrayCopy(&g_MicroProfileUI.LockedToolTips[nIndex], &ToolTip); - g_MicroProfileUI.LockedToolTipFront = nIndex; - - } -} - - -void MicroProfileZoomTo(int64_t nTickStart, int64_t nTickEnd) -{ - MicroProfile& S = *MicroProfileGet(); - - int64_t nStart = S.Frames[S.nFrameCurrent].nFrameStartCpu; - float fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()); - UI.fDetailedOffsetTarget = MicroProfileLogTickDifference(nStart, nTickStart) * fToMs; - UI.fDetailedRangeTarget = MicroProfileLogTickDifference(nTickStart, nTickEnd) * fToMs; -} - -void MicroProfileCenter(int64_t nTickCenter) -{ - MicroProfile& S = *MicroProfileGet(); - int64_t nStart = S.Frames[S.nFrameCurrent].nFrameStartCpu; - float fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()); - float fCenter = MicroProfileLogTickDifference(nStart, nTickCenter) * fToMs; - UI.fDetailedOffsetTarget = UI.fDetailedOffset = fCenter - 0.5f * UI.fDetailedRange; -} -#ifdef MICROPROFILE_DEBUG -uint64_t* g_pMicroProfileDumpStart = 0; -uint64_t* g_pMicroProfileDumpEnd = 0; -void MicroProfileDebugDumpRange() -{ - MicroProfile& S = *MicroProfileGet(); - if(g_pMicroProfileDumpStart != g_pMicroProfileDumpEnd) - { - uint64_t* pStart = g_pMicroProfileDumpStart; - uint64_t* pEnd = g_pMicroProfileDumpEnd; - while(pStart != pEnd) - { - uint64_t nTick = MicroProfileLogGetTick(*pStart); - uint64_t nToken = MicroProfileLogTimerIndex(*pStart); - uint32_t nTimerId = MicroProfileGetTimerIndex(nToken); - - const char* pTimerName = S.TimerInfo[nTimerId].pName; - char buffer[256]; - int type = MicroProfileLogType(*pStart); - - const char* pBegin = type == MP_LOG_LEAVE ? "END" : - (type == MP_LOG_ENTER ? "BEGIN" : "META"); - snprintf(buffer, 255, "DUMP 0x%p: %s :: %llx: %s\n", pStart, pBegin, nTick, pTimerName); -#ifdef _WIN32 - OutputDebugString(buffer); -#else - printf("%s", buffer); -#endif - pStart++; - } - - g_pMicroProfileDumpStart = g_pMicroProfileDumpEnd; - } -} -#define MP_DEBUG_DUMP_RANGE() MicroProfileDebugDumpRange(); -#else -#define MP_DEBUG_DUMP_RANGE() do{} while(0) -#endif - -#define MICROPROFILE_HOVER_DIST 0.5f - -void MicroProfileDrawDetailedContextSwitchBars(uint32_t nY, uint32_t nThreadId, uint32_t nContextSwitchStart, uint32_t nContextSwitchEnd, int64_t nBaseTicks, uint32_t nBaseY) -{ - MicroProfile& S = *MicroProfileGet(); - int64_t nTickIn = -1; - uint32_t nThreadBefore = UINT32_MAX; - float fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()); - float fMsToScreen = UI.nWidth / UI.fDetailedRange; - float fMouseX = (float)UI.nMouseX; - float fMouseY = (float)UI.nMouseY; - - - for(uint32_t j = nContextSwitchStart; j != nContextSwitchEnd; j = (j+1) % MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE) - { - MP_ASSERT(j < MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE); - MicroProfileContextSwitch CS = S.ContextSwitch[j]; - - if(nTickIn == -1) - { - if(CS.nThreadIn == nThreadId) - { - nTickIn = CS.nTicks; - nThreadBefore = CS.nThreadOut; - } - } - else - { - if(CS.nThreadOut == nThreadId) - { - int64_t nTickOut = CS.nTicks; - float fMsStart = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickIn); - float fMsEnd = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickOut); - if(fMsStart <= fMsEnd) - { - float fXStart = fMsStart * fMsToScreen; - float fXEnd = fMsEnd * fMsToScreen; - float fYStart = (float)nY; - float fYEnd = fYStart + (MICROPROFILE_DETAILED_CONTEXT_SWITCH_HEIGHT); - uint32_t nColor = g_nMicroProfileContextSwitchThreadColors[CS.nCpu%MICROPROFILE_NUM_CONTEXT_SWITCH_COLORS]; - float fXDist = MicroProfileMax(fXStart - fMouseX, fMouseX - fXEnd); - bool bHover = fXDist < MICROPROFILE_HOVER_DIST && fYStart <= fMouseY && fMouseY <= fYEnd && nBaseY < fMouseY; - if(bHover) - { - UI.nRangeBegin = nTickIn; - UI.nRangeEnd = nTickOut; - S.nContextSwitchHoverTickIn = nTickIn; - S.nContextSwitchHoverTickOut = nTickOut; - S.nContextSwitchHoverThread = CS.nThreadOut; - S.nContextSwitchHoverThreadBefore = nThreadBefore; - S.nContextSwitchHoverThreadAfter = CS.nThreadIn; - S.nContextSwitchHoverCpuNext = CS.nCpu; - nColor = UI.nHoverColor; - } - if(CS.nCpu == S.nContextSwitchHoverCpu) - { - nColor = UI.nHoverColorShared; - } - MicroProfileDrawBox(fXStart, fYStart, fXEnd, fYEnd, nColor|UI.nOpacityForeground, MicroProfileBoxTypeFlat); - } - nTickIn = -1; - } - } - } -} - -void MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int nBaseY, int nSelectedFrame) -{ - MicroProfile& S = *MicroProfileGet(); - MP_DEBUG_DUMP_RANGE(); - int nY = nBaseY - UI.nOffsetY; - [[maybe_unused]] int64_t nNumBoxes = 0; - [[maybe_unused]] int64_t nNumLines = 0; - - uint32_t nFrameNext = (S.nFrameCurrent+1) % MICROPROFILE_MAX_FRAME_HISTORY; - MicroProfileFrameState* pFrameCurrent = &S.Frames[S.nFrameCurrent]; - MicroProfileFrameState* pFrameNext = &S.Frames[nFrameNext]; - - UI.nRangeBegin = 0; - UI.nRangeEnd = 0; - UI.nRangeBeginGpu = 0; - UI.nRangeEndGpu = 0; - UI.nRangeBeginIndex = UI.nRangeEndIndex = 0; - UI.pRangeLog = 0; - int64_t nFrameStartCpu = pFrameCurrent->nFrameStartCpu; - int64_t nFrameStartGpu = pFrameCurrent->nFrameStartGpu; - int64_t nTicksPerSecondCpu = MicroProfileTicksPerSecondCpu(); - int64_t nTicksPerSecondGpu = MicroProfileTicksPerSecondGpu(); - float fToMsCpu = MicroProfileTickToMsMultiplier(nTicksPerSecondCpu); - float fToMsGpu = MicroProfileTickToMsMultiplier(nTicksPerSecondGpu); - - float fDetailedOffset = UI.fDetailedOffset; - float fDetailedRange = UI.fDetailedRange; - - - int64_t nDetailedOffsetTicksCpu = MicroProfileMsToTick(fDetailedOffset, MicroProfileTicksPerSecondCpu()); - int64_t nDetailedOffsetTicksGpu = MicroProfileMsToTick(fDetailedOffset, MicroProfileTicksPerSecondGpu()); - int64_t nBaseTicksCpu = nDetailedOffsetTicksCpu + nFrameStartCpu; - int64_t nBaseTicksGpu = nDetailedOffsetTicksGpu + nFrameStartGpu; - int64_t nBaseTicksEndCpu = nBaseTicksCpu + MicroProfileMsToTick(fDetailedRange, MicroProfileTicksPerSecondCpu()); - - int64_t nTickReferenceCpu = 0, nTickReferenceGpu = 0; - static int64_t nRefCpu = 0, nRefGpu = 0; - if(MicroProfileGetGpuTickReference(&nTickReferenceCpu, &nTickReferenceGpu)) - { - if(0 == nRefCpu || std::abs(nRefCpu-nBaseTicksCpu) > std::abs(nTickReferenceCpu-nBaseTicksCpu)) - { - nRefCpu = nTickReferenceCpu; - nRefGpu = nTickReferenceGpu; - } - else - { - nTickReferenceCpu = nRefCpu; - nTickReferenceGpu = nRefGpu; - } - nBaseTicksGpu = (nBaseTicksCpu - nTickReferenceCpu) * nTicksPerSecondGpu / nTicksPerSecondCpu + nTickReferenceGpu; - } - int64_t nBaseTicksEndGpu = nBaseTicksCpu + MicroProfileMsToTick(fDetailedRange, MicroProfileTicksPerSecondCpu()); - - MicroProfileFrameState* pFrameFirst = pFrameCurrent; - int64_t nGapTime = MicroProfileTicksPerSecondCpu() * MICROPROFILE_GAP_TIME / 1000; - for(uint32_t i = 0; i < MICROPROFILE_MAX_FRAME_HISTORY - MICROPROFILE_GPU_FRAME_DELAY; ++i) - { - uint32_t nNextIndex = (S.nFrameCurrent + MICROPROFILE_MAX_FRAME_HISTORY - i) % MICROPROFILE_MAX_FRAME_HISTORY; - pFrameFirst = &S.Frames[nNextIndex]; - if(pFrameFirst->nFrameStartCpu <= nBaseTicksCpu-nGapTime) - break; - } - - float fMsBase = fToMsCpu * nDetailedOffsetTicksCpu; - float fMs = fDetailedRange; - float fMsEnd = fMs + fMsBase; - float fWidth = (float)nWidth; - float fMsToScreen = fWidth / fMs; - - { - float fRate = floor(2*(log10(fMs)-1))/2; - float fStep = powf(10.f, fRate); - float fRcpStep = 1.f / fStep; - int nColorIndex = (int)(floor(fMsBase*fRcpStep)); - float fStart = floor(fMsBase*fRcpStep) * fStep; - for(float f = fStart; f < fMsEnd; ) - { - float fStart = f; - float fNext = f + fStep; - MicroProfileDrawBox(((fStart-fMsBase) * fMsToScreen), nBaseY, (fNext-fMsBase) * fMsToScreen+1, nBaseY + nHeight, UI.nOpacityBackground | g_nMicroProfileBackColors[nColorIndex++ & 1]); - f = fNext; - } - } - - nY += MICROPROFILE_TEXT_HEIGHT+1; - MicroProfileLogEntry* pMouseOver = UI.pDisplayMouseOver; - MicroProfileLogEntry* pMouseOverNext = 0; - uint64_t nMouseOverToken = pMouseOver ? MicroProfileLogTimerIndex(*pMouseOver) : MICROPROFILE_INVALID_TOKEN; - float fMouseX = (float)UI.nMouseX; - float fMouseY = (float)UI.nMouseY; - uint64_t nHoverToken = MICROPROFILE_INVALID_TOKEN; - int64_t nHoverTime = 0; - - static int nHoverCounter = 155; - static int nHoverCounterDelta = 10; - nHoverCounter += nHoverCounterDelta; - if(nHoverCounter >= 245) - nHoverCounterDelta = -10; - else if(nHoverCounter < 100) - nHoverCounterDelta = 10; - UI.nHoverColor = (nHoverCounter<<24)|(nHoverCounter<<16)|(nHoverCounter<<8)|nHoverCounter; - uint32_t nHoverCounterShared = nHoverCounter>>2; - UI.nHoverColorShared = (nHoverCounterShared<<24)|(nHoverCounterShared<<16)|(nHoverCounterShared<<8)|nHoverCounterShared; - - uint32_t nLinesDrawn[MICROPROFILE_STACK_MAX]={0}; - - uint32_t nContextSwitchHoverThreadAfter = S.nContextSwitchHoverThreadAfter; - uint32_t nContextSwitchHoverThreadBefore = S.nContextSwitchHoverThreadBefore; - S.nContextSwitchHoverThread = S.nContextSwitchHoverThreadAfter = S.nContextSwitchHoverThreadBefore = UINT32_MAX; - - uint32_t nContextSwitchStart = UINT32_MAX; - uint32_t nContextSwitchEnd = UINT32_MAX; - S.nContextSwitchHoverCpuNext = 0xff; - S.nContextSwitchHoverTickIn = -1; - S.nContextSwitchHoverTickOut = -1; - if(S.bContextSwitchRunning) - { - MicroProfileContextSwitchSearch(&nContextSwitchStart, &nContextSwitchEnd, nBaseTicksCpu, nBaseTicksEndCpu); - } - - bool bSkipBarView = S.bContextSwitchRunning && S.bContextSwitchNoBars; - - if(!bSkipBarView) - { - for(uint32_t i = 0; i < MICROPROFILE_MAX_THREADS; ++i) - { - MicroProfileThreadLog* pLog = S.Pool[i]; - if(!pLog) - continue; - - uint32_t nPut = pFrameNext->nLogStart[i]; - ///note: this may display new samples as old data, but this will only happen when - // unpaused, where the detailed view is hardly perceptible - uint32_t nFront = S.Pool[i]->nPut.load(std::memory_order_relaxed); - MicroProfileFrameState* pFrameLogFirst = pFrameCurrent; - MicroProfileFrameState* pFrameLogLast = pFrameNext; - uint32_t nGet = pFrameLogFirst->nLogStart[i]; - do - { - MP_ASSERT(pFrameLogFirst >= &S.Frames[0] && pFrameLogFirst < &S.Frames[MICROPROFILE_MAX_FRAME_HISTORY]); - uint32_t nNewGet = pFrameLogFirst->nLogStart[i]; - bool bIsValid = false; - if(nPut < nFront) - { - bIsValid = nNewGet <= nPut || nNewGet >= nFront; - } - else - { - bIsValid = nNewGet <= nPut && nNewGet >= nFront; - } - if(bIsValid) - { - nGet = nNewGet; - pFrameLogFirst--; - if(pFrameLogFirst < &S.Frames[0]) - pFrameLogFirst = &S.Frames[MICROPROFILE_MAX_FRAME_HISTORY-1]; - } - else - { - break; - } - }while(pFrameLogFirst != pFrameFirst); - - - if (nGet == UINT32_MAX) { - continue; - } - MP_ASSERT(nGet != UINT32_MAX); - - nPut = pFrameLogLast->nLogStart[i]; - - uint32_t nRange[2][2] = { {0, 0}, {0, 0}, }; - - MicroProfileGetRange(nPut, nGet, nRange); - if(nPut == nGet) - continue; - uint32_t nMaxStackDepth = 0; - - bool bGpu = pLog->nGpu != 0; - float fToMs = bGpu ? fToMsGpu : fToMsCpu; - int64_t nBaseTicks = bGpu ? nBaseTicksGpu : nBaseTicksCpu; - char ThreadName[MicroProfileThreadLog::THREAD_MAX_LEN + 16]; - uint64_t nThreadId = pLog->nThreadId; - snprintf(ThreadName, sizeof(ThreadName)-1, "%04" PRIx64 ": %s", nThreadId, &pLog->ThreadName[0] ); - nY += 3; - uint32_t nThreadColor = UINT32_MAX; - if(pLog->nThreadId == nContextSwitchHoverThreadAfter || pLog->nThreadId == nContextSwitchHoverThreadBefore) - nThreadColor = UI.nHoverColorShared|0x906060; - MicroProfileDrawText(0, nY, nThreadColor, &ThreadName[0], (uint32_t)strlen(&ThreadName[0])); - nY += 3; - nY += MICROPROFILE_TEXT_HEIGHT + 1; - - if(S.bContextSwitchRunning) - { - MicroProfileDrawDetailedContextSwitchBars(nY, pLog->nThreadId, nContextSwitchStart, nContextSwitchEnd, nBaseTicks, nBaseY); - nY -= MICROPROFILE_DETAILED_BAR_HEIGHT; - nY += MICROPROFILE_DETAILED_CONTEXT_SWITCH_HEIGHT+1; - } - - uint32_t nYDelta = MICROPROFILE_DETAILED_BAR_HEIGHT; - uint32_t nStack[MICROPROFILE_STACK_MAX]; - uint32_t nStackPos = 0; - for(uint32_t j = 0; j < 2; ++j) - { - uint32_t nStart = nRange[j][0]; - uint32_t nEnd = nRange[j][1]; - for(uint32_t k = nStart; k < nEnd; ++k) - { - MicroProfileLogEntry* pEntry = &pLog->Log[k]; - int nType = MicroProfileLogType(*pEntry); - if(MP_LOG_ENTER == nType) - { - MP_ASSERT(nStackPos < MICROPROFILE_STACK_MAX); - nStack[nStackPos++] = k; - } - else if(MP_LOG_META == nType) - { - - } - else if(MP_LOG_LEAVE == nType) - { - if(0 == nStackPos) - { - continue; - } - - MicroProfileLogEntry* pEntryEnter = &pLog->Log[nStack[nStackPos-1]]; - if(MicroProfileLogTimerIndex(*pEntryEnter) != MicroProfileLogTimerIndex(*pEntry)) - { - //uprintf("mismatch %llx %llx\n", pEntryEnter->nToken, pEntry->nToken); - continue; - } - int64_t nTickStart = MicroProfileLogGetTick(*pEntryEnter); - int64_t nTickEnd = MicroProfileLogGetTick(*pEntry); - uint64_t nTimerIndex = MicroProfileLogTimerIndex(*pEntry); - uint32_t nColor = S.TimerInfo[nTimerIndex].nColor; - if(nMouseOverToken == nTimerIndex) - { - if(pEntry == pMouseOver) - { - nColor = UI.nHoverColor; - if(bGpu) - { - UI.nRangeBeginGpu = *pEntryEnter; - UI.nRangeEndGpu = *pEntry; - uint32_t nCpuBegin = (nStack[nStackPos-1] + 1) % MICROPROFILE_BUFFER_SIZE; - uint32_t nCpuEnd = (k + 1) % MICROPROFILE_BUFFER_SIZE; - MicroProfileLogEntry LogCpuBegin = pLog->Log[nCpuBegin]; - MicroProfileLogEntry LogCpuEnd = pLog->Log[nCpuEnd]; - if(MicroProfileLogType(LogCpuBegin)==3 && MicroProfileLogType(LogCpuEnd) == 3) - { - UI.nRangeBegin = LogCpuBegin; - UI.nRangeEnd = LogCpuEnd; - } - UI.nRangeBeginIndex = nStack[nStackPos-1]; - UI.nRangeEndIndex = k; - UI.pRangeLog = pLog; - } - else - { - UI.nRangeBegin = *pEntryEnter; - UI.nRangeEnd = *pEntry; - UI.nRangeBeginIndex = nStack[nStackPos-1]; - UI.nRangeEndIndex = k; - UI.pRangeLog = pLog; - - } - } - else - { - nColor = UI.nHoverColorShared; - } - } - - nMaxStackDepth = MicroProfileMax(nMaxStackDepth, nStackPos); - float fMsStart = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickStart); - float fMsEnd = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickEnd); - float fXStart = fMsStart * fMsToScreen; - float fXEnd = fMsEnd * fMsToScreen; - float fYStart = (float)(nY + nStackPos * nYDelta); - float fYEnd = fYStart + (MICROPROFILE_DETAILED_BAR_HEIGHT); - float fXDist = MicroProfileMax(fXStart - fMouseX, fMouseX - fXEnd); - bool bHover = fXDist < MICROPROFILE_HOVER_DIST && fYStart <= fMouseY && fMouseY <= fYEnd && nBaseY < fMouseY; - uint32_t nIntegerWidth = (uint32_t)(fXEnd - fXStart); - if(nIntegerWidth) - { - if(bHover && UI.nActiveMenu == UINT32_MAX) - { - nHoverToken = MicroProfileLogTimerIndex(*pEntry); - #if MICROPROFILE_DEBUG - UI.nHoverAddressEnter = (uint64_t)pEntryEnter; - UI.nHoverAddressLeave = (uint64_t)pEntry; - #endif - nHoverTime = MicroProfileLogTickDifference(nTickStart, nTickEnd); - pMouseOverNext = pEntry; - } - - MicroProfileDrawBox(fXStart, fYStart, fXEnd, fYEnd, nColor|UI.nOpacityForeground, MicroProfileBoxTypeBar); -#if MICROPROFILE_DETAILED_BAR_NAMES - if(nIntegerWidth>3*MICROPROFILE_TEXT_WIDTH) - { - float fXStartText = MicroProfileMax(fXStart, 0.f); - int nTextWidth = (int)(fXEnd - fXStartText); - int nCharacters = (nTextWidth - 2*MICROPROFILE_TEXT_WIDTH) / MICROPROFILE_TEXT_WIDTH; - if(nCharacters>0) - { - MicroProfileDrawText(fXStartText + 1, fYStart + 1, UINT32_MAX, S.TimerInfo[nTimerIndex].pName, MicroProfileMin(S.TimerInfo[nTimerIndex].nNameLen, nCharacters)); - } - } -#endif - ++nNumBoxes; - } - else - { - float fXAvg = 0.5f * (fXStart + fXEnd); - int nLineX = (int)floor(fXAvg+0.5f); - if(nLineX != (int)nLinesDrawn[nStackPos]) - { - if(bHover && UI.nActiveMenu == UINT32_MAX) - { - nHoverToken = (uint32_t)MicroProfileLogTimerIndex(*pEntry); - nHoverTime = MicroProfileLogTickDifference(nTickStart, nTickEnd); - pMouseOverNext = pEntry; - } - nLinesDrawn[nStackPos] = nLineX; - MicroProfileDrawLineVertical(nLineX, fYStart + 0.5f, fYEnd + 0.5f, nColor|UI.nOpacityForeground); - ++nNumLines; - } - } - nStackPos--; - if(0 == nStackPos) - { - if(bGpu ? (nTickStart > nBaseTicksEndGpu) : (nTickStart > nBaseTicksEndCpu)) - { - break; - } - } - } - } - } - nY += nMaxStackDepth * nYDelta + MICROPROFILE_DETAILED_BAR_HEIGHT+1; - } - } - if(S.bContextSwitchRunning && (S.bContextSwitchAllThreads||S.bContextSwitchNoBars)) - { - uint32_t nNumThreads = 0; - uint32_t nThreads[MICROPROFILE_MAX_CONTEXT_SWITCH_THREADS]; - for(uint32_t i = 0; i < MICROPROFILE_MAX_THREADS && S.Pool[i]; ++i) - nThreads[nNumThreads++] = S.Pool[i]->nThreadId; - uint32_t nNumThreadsBase = nNumThreads; - if(S.bContextSwitchAllThreads) - { - for(uint32_t i = nContextSwitchStart; i != nContextSwitchEnd; i = (i+1) % MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE) - { - MicroProfileContextSwitch CS = S.ContextSwitch[i]; - ThreadIdType nThreadId = CS.nThreadIn; - if(nThreadId) - { - bool bSeen = false; - for(uint32_t j = 0; j < nNumThreads; ++j) - { - if(nThreads[j] == nThreadId) - { - bSeen = true; - break; - } - } - if(!bSeen) - { - nThreads[nNumThreads++] = nThreadId; - } - } - if(nNumThreads == MICROPROFILE_MAX_CONTEXT_SWITCH_THREADS) - { - S.nOverflow = 10; - break; - } - } - std::sort(&nThreads[nNumThreadsBase], &nThreads[nNumThreads]); - } - uint32_t nStart = nNumThreadsBase; - if(S.bContextSwitchNoBars) - nStart = 0; - for(uint32_t i = nStart; i < nNumThreads; ++i) - { - ThreadIdType nThreadId = nThreads[i]; - if(nThreadId) - { - char ThreadName[MicroProfileThreadLog::THREAD_MAX_LEN + 16]; - const char* cLocal = MicroProfileIsLocalThread(nThreadId) ? "*": " "; - -#if defined(_WIN32) - // nThreadId is 32-bit on Windows - int nStrLen = snprintf(ThreadName, sizeof(ThreadName)-1, "%04x: %s%s", nThreadId, cLocal, i < nNumThreadsBase ? &S.Pool[i]->ThreadName[0] : MICROPROFILE_THREAD_NAME_FROM_ID(nThreadId) ); -#else - int nStrLen = snprintf(ThreadName, sizeof(ThreadName)-1, "%04" PRIx64 ": %s%s", nThreadId, cLocal, i < nNumThreadsBase ? &S.Pool[i]->ThreadName[0] : MICROPROFILE_THREAD_NAME_FROM_ID(nThreadId) ); -#endif - uint32_t nThreadColor = UINT32_MAX; - if(nThreadId == nContextSwitchHoverThreadAfter || nThreadId == nContextSwitchHoverThreadBefore) - nThreadColor = UI.nHoverColorShared|0x906060; - MicroProfileDrawDetailedContextSwitchBars(nY+2, nThreadId, nContextSwitchStart, nContextSwitchEnd, nBaseTicksCpu, nBaseY); - MicroProfileDrawText(0, nY, nThreadColor, &ThreadName[0], nStrLen); - nY += MICROPROFILE_TEXT_HEIGHT+1; - } - } - } - - S.nContextSwitchHoverCpu = S.nContextSwitchHoverCpuNext; - - UI.pDisplayMouseOver = pMouseOverNext; - - if(!S.nRunning) - { - if(nHoverToken != MICROPROFILE_INVALID_TOKEN && nHoverTime) - { - UI.nHoverToken = nHoverToken; - UI.nHoverTime = nHoverTime; - } - - if(nSelectedFrame != -1) - { - UI.nRangeBegin = S.Frames[nSelectedFrame].nFrameStartCpu; - UI.nRangeEnd = S.Frames[(nSelectedFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY].nFrameStartCpu; - UI.nRangeBeginGpu = S.Frames[nSelectedFrame].nFrameStartGpu; - UI.nRangeEndGpu = S.Frames[(nSelectedFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY].nFrameStartGpu; - } - if(UI.nRangeBegin != UI.nRangeEnd) - { - float fMsStart = fToMsCpu * MicroProfileLogTickDifference(nBaseTicksCpu, UI.nRangeBegin); - float fMsEnd = fToMsCpu * MicroProfileLogTickDifference(nBaseTicksCpu, UI.nRangeEnd); - float fXStart = fMsStart * fMsToScreen; - float fXEnd = fMsEnd * fMsToScreen; - MicroProfileDrawBox(fXStart, nBaseY, fXEnd, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT, MicroProfileBoxTypeFlat); - MicroProfileDrawLineVertical(fXStart, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT | 0x44000000); - MicroProfileDrawLineVertical(fXEnd, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT | 0x44000000); - - fMsStart += fDetailedOffset; - fMsEnd += fDetailedOffset; - char sBuffer[32]; - uint32_t nLenStart = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsStart); - float fStartTextWidth = (float)((1+MICROPROFILE_TEXT_WIDTH) * nLenStart); - float fStartTextX = fXStart - fStartTextWidth - 2; - MicroProfileDrawBox(fStartTextX, nBaseY, fStartTextX + fStartTextWidth + 2, MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat); - MicroProfileDrawText(fStartTextX+1, nBaseY, UINT32_MAX, sBuffer, nLenStart); - uint32_t nLenEnd = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsEnd); - MicroProfileDrawBox(fXEnd+1, nBaseY, fXEnd+1+(1+MICROPROFILE_TEXT_WIDTH) * nLenEnd + 3, MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat); - MicroProfileDrawText(fXEnd+2, nBaseY+1, UINT32_MAX, sBuffer, nLenEnd); - - if(UI.nMouseRight) - { - MicroProfileZoomTo(UI.nRangeBegin, UI.nRangeEnd); - } - } - - if(UI.nRangeBeginGpu != UI.nRangeEndGpu) - { - float fMsStart = fToMsGpu * MicroProfileLogTickDifference(nBaseTicksGpu, UI.nRangeBeginGpu); - float fMsEnd = fToMsGpu * MicroProfileLogTickDifference(nBaseTicksGpu, UI.nRangeEndGpu); - float fXStart = fMsStart * fMsToScreen; - float fXEnd = fMsEnd * fMsToScreen; - MicroProfileDrawBox(fXStart, nBaseY, fXEnd, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU, MicroProfileBoxTypeFlat); - MicroProfileDrawLineVertical(fXStart, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU | 0x44000000); - MicroProfileDrawLineVertical(fXEnd, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU | 0x44000000); - - nBaseY += MICROPROFILE_TEXT_HEIGHT+1; - - fMsStart += fDetailedOffset; - fMsEnd += fDetailedOffset; - char sBuffer[32]; - uint32_t nLenStart = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsStart); - float fStartTextWidth = (float)((1+MICROPROFILE_TEXT_WIDTH) * nLenStart); - float fStartTextX = fXStart - fStartTextWidth - 2; - MicroProfileDrawBox(fStartTextX, nBaseY, fStartTextX + fStartTextWidth + 2, MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat); - MicroProfileDrawText(fStartTextX+1, nBaseY, UINT32_MAX, sBuffer, nLenStart); - uint32_t nLenEnd = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsEnd); - MicroProfileDrawBox(fXEnd+1, nBaseY, fXEnd+1+(1+MICROPROFILE_TEXT_WIDTH) * nLenEnd + 3, MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat); - MicroProfileDrawText(fXEnd+2, nBaseY+1, UINT32_MAX, sBuffer, nLenEnd); - } - } -} - - -void MicroProfileDrawDetailedFrameHistory(uint32_t nWidth, uint32_t nHeight, uint32_t nBaseY, uint32_t nSelectedFrame) -{ - MicroProfile& S = *MicroProfileGet(); - - const uint32_t nBarHeight = MICROPROFILE_FRAME_HISTORY_HEIGHT; - float fBaseX = (float)nWidth; - float fDx = fBaseX / MICROPROFILE_NUM_FRAMES; - - uint32_t nLastIndex = (S.nFrameCurrent+1) % MICROPROFILE_MAX_FRAME_HISTORY; - MicroProfileDrawBox(0, nBaseY, nWidth, nBaseY+MICROPROFILE_FRAME_HISTORY_HEIGHT, 0xff000000 | g_nMicroProfileBackColors[0], MicroProfileBoxTypeFlat); - float fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()) * S.fRcpReferenceTime; - float fToMsGpu = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondGpu()) * S.fRcpReferenceTime; - - - MicroProfileFrameState* pFrameCurrent = &S.Frames[S.nFrameCurrent]; - uint64_t nFrameStartCpu = pFrameCurrent->nFrameStartCpu; - int64_t nDetailedOffsetTicksCpu = MicroProfileMsToTick(UI.fDetailedOffset, MicroProfileTicksPerSecondCpu()); - int64_t nCpuStart = nDetailedOffsetTicksCpu + nFrameStartCpu; - int64_t nCpuEnd = nCpuStart + MicroProfileMsToTick(UI.fDetailedRange, MicroProfileTicksPerSecondCpu());; - - - float fSelectionStart = (float)nWidth; - float fSelectionEnd = 0.f; - for(uint32_t i = 0; i < MICROPROFILE_NUM_FRAMES; ++i) - { - uint32_t nIndex = (S.nFrameCurrent + MICROPROFILE_MAX_FRAME_HISTORY - i) % MICROPROFILE_MAX_FRAME_HISTORY; - MicroProfileFrameState* pCurrent = &S.Frames[nIndex]; - MicroProfileFrameState* pNext = &S.Frames[nLastIndex]; - - int64_t nTicks = pNext->nFrameStartCpu - pCurrent->nFrameStartCpu; - int64_t nTicksGpu = pNext->nFrameStartGpu - pCurrent->nFrameStartGpu; - float fScale = fToMs * nTicks; - float fScaleGpu = fToMsGpu * nTicksGpu; - fScale = fScale > 1.f ? 0.f : 1.f - fScale; - fScaleGpu = fScaleGpu > 1.f ? 0.f : 1.f - fScaleGpu; - float fXEnd = fBaseX; - float fXStart = fBaseX - fDx; - fBaseX = fXStart; - uint32_t nColor = MICROPROFILE_FRAME_HISTORY_COLOR_CPU; - if(nIndex == nSelectedFrame) - nColor = UINT32_MAX; - MicroProfileDrawBox(fXStart, nBaseY + fScale * nBarHeight, fXEnd, nBaseY+MICROPROFILE_FRAME_HISTORY_HEIGHT, nColor, MicroProfileBoxTypeBar); - if(pNext->nFrameStartCpu > nCpuStart) - { - fSelectionStart = fXStart; - } - if(pCurrent->nFrameStartCpu < nCpuEnd && fSelectionEnd == 0.f) - { - fSelectionEnd = fXEnd; - } - nLastIndex = nIndex; - } - MicroProfileDrawBox(fSelectionStart, nBaseY, fSelectionEnd, nBaseY+MICROPROFILE_FRAME_HISTORY_HEIGHT, MICROPROFILE_FRAME_HISTORY_COLOR_HIGHTLIGHT, MicroProfileBoxTypeFlat); -} -void MicroProfileDrawDetailedView(uint32_t nWidth, uint32_t nHeight) -{ - MicroProfile& S = *MicroProfileGet(); - - MICROPROFILE_SCOPE(g_MicroProfileDetailed); - uint32_t nBaseY = MICROPROFILE_TEXT_HEIGHT + 1; - - int nSelectedFrame = -1; - if(UI.nMouseY > nBaseY && UI.nMouseY <= nBaseY + MICROPROFILE_FRAME_HISTORY_HEIGHT && UI.nActiveMenu == UINT32_MAX) - { - - nSelectedFrame = ((MICROPROFILE_NUM_FRAMES) * (UI.nWidth-UI.nMouseX) / UI.nWidth); - nSelectedFrame = (S.nFrameCurrent + MICROPROFILE_MAX_FRAME_HISTORY - nSelectedFrame) % MICROPROFILE_MAX_FRAME_HISTORY; - UI.nHoverFrame = nSelectedFrame; - if(UI.nMouseRight) - { - int64_t nRangeBegin = S.Frames[nSelectedFrame].nFrameStartCpu; - int64_t nRangeEnd = S.Frames[(nSelectedFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY].nFrameStartCpu; - MicroProfileZoomTo(nRangeBegin, nRangeEnd); - } - if(UI.nMouseDownLeft) - { - uint64_t nFrac = (1024 * (MICROPROFILE_NUM_FRAMES) * (UI.nMouseX) / UI.nWidth) % 1024; - int64_t nRangeBegin = S.Frames[nSelectedFrame].nFrameStartCpu; - int64_t nRangeEnd = S.Frames[(nSelectedFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY].nFrameStartCpu; - MicroProfileCenter(nRangeBegin + (nRangeEnd-nRangeBegin) * nFrac / 1024); - } - } - else - { - UI.nHoverFrame = -1; - } - - MicroProfileDrawDetailedBars(nWidth, nHeight, nBaseY + MICROPROFILE_FRAME_HISTORY_HEIGHT, nSelectedFrame); - MicroProfileDrawDetailedFrameHistory(nWidth, nHeight, nBaseY, nSelectedFrame); -} - -void MicroProfileDrawTextRight(uint32_t nX, uint32_t nY, uint32_t nColor, const char* pStr, uint32_t nStrLen) -{ - MicroProfileDrawText(nX - nStrLen * (MICROPROFILE_TEXT_WIDTH+1), nY, nColor, pStr, nStrLen); -} -void MicroProfileDrawHeader(int32_t nX, uint32_t nWidth, const char* pName) -{ - if(pName) - { - MicroProfileDrawBox(nX-8, MICROPROFILE_TEXT_HEIGHT + 2, nX + nWidth+5, MICROPROFILE_TEXT_HEIGHT + 2 + (MICROPROFILE_TEXT_HEIGHT+1), 0xff000000|g_nMicroProfileBackColors[1]); - MicroProfileDrawText(nX, MICROPROFILE_TEXT_HEIGHT + 2, UINT32_MAX, pName, (uint32_t)strlen(pName)); - } -} - - -typedef void (*MicroProfileLoopGroupCallback)(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pData); - -void MicroProfileLoopActiveGroupsDraw(int32_t nX, int32_t nY, const char* pName, MicroProfileLoopGroupCallback CB, void* pData) -{ - MicroProfile& S = *MicroProfileGet(); - nY += MICROPROFILE_TEXT_HEIGHT + 2; - uint64_t nGroup = S.nAllGroupsWanted ? S.nGroupMask : S.nActiveGroupWanted; - uint32_t nCount = 0; - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - uint64_t nMask = 1ULL << j; - if(nMask & nGroup) - { - nY += MICROPROFILE_TEXT_HEIGHT + 1; - for(uint32_t i = 0; i < S.nTotalTimers;++i) - { - uint64_t nTokenMask = MicroProfileGetGroupMask(S.TimerInfo[i].nToken); - if(nTokenMask & nMask) - { - if(nY >= 0) - CB(i, nCount, nMask, nX, nY, pData); - - nCount += 2; - nY += MICROPROFILE_TEXT_HEIGHT + 1; - - if(nY > (int)UI.nHeight) - return; - } - } - - } - } -} - - -void MicroProfileCalcTimers(float* pTimers, float* pAverage, float* pMax, float* pCallAverage, float* pExclusive, float* pAverageExclusive, float* pMaxExclusive, uint64_t nGroup, uint32_t nSize) -{ - MicroProfile& S = *MicroProfileGet(); - - uint32_t nCount = 0; - uint64_t nMask = 1; - - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - if(nMask & nGroup) - { - const float fToMs = MicroProfileTickToMsMultiplier(S.GroupInfo[j].Type == MicroProfileTokenTypeGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu()); - for(uint32_t i = 0; i < S.nTotalTimers;++i) - { - uint64_t nTokenMask = MicroProfileGetGroupMask(S.TimerInfo[i].nToken); - if(nTokenMask & nMask) - { - { - uint32_t nTimer = i; - uint32_t nIdx = nCount; - uint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1; - uint32_t nAggregateCount = S.Aggregate[nTimer].nCount ? S.Aggregate[nTimer].nCount : 1; - float fToPrc = S.fRcpReferenceTime; - float fMs = fToMs * (S.Frame[nTimer].nTicks); - float fPrc = MicroProfileMin(fMs * fToPrc, 1.f); - float fAverageMs = fToMs * (S.Aggregate[nTimer].nTicks / nAggregateFrames); - float fAveragePrc = MicroProfileMin(fAverageMs * fToPrc, 1.f); - float fMaxMs = fToMs * (S.AggregateMax[nTimer]); - float fMaxPrc = MicroProfileMin(fMaxMs * fToPrc, 1.f); - float fCallAverageMs = fToMs * (S.Aggregate[nTimer].nTicks / nAggregateCount); - float fCallAveragePrc = MicroProfileMin(fCallAverageMs * fToPrc, 1.f); - float fMsExclusive = fToMs * (S.FrameExclusive[nTimer]); - float fPrcExclusive = MicroProfileMin(fMsExclusive * fToPrc, 1.f); - float fAverageMsExclusive = fToMs * (S.AggregateExclusive[nTimer] / nAggregateFrames); - float fAveragePrcExclusive = MicroProfileMin(fAverageMsExclusive * fToPrc, 1.f); - float fMaxMsExclusive = fToMs * (S.AggregateMaxExclusive[nTimer]); - float fMaxPrcExclusive = MicroProfileMin(fMaxMsExclusive * fToPrc, 1.f); - pTimers[nIdx] = fMs; - pTimers[nIdx+1] = fPrc; - pAverage[nIdx] = fAverageMs; - pAverage[nIdx+1] = fAveragePrc; - pMax[nIdx] = fMaxMs; - pMax[nIdx+1] = fMaxPrc; - pCallAverage[nIdx] = fCallAverageMs; - pCallAverage[nIdx+1] = fCallAveragePrc; - pExclusive[nIdx] = fMsExclusive; - pExclusive[nIdx+1] = fPrcExclusive; - pAverageExclusive[nIdx] = fAverageMsExclusive; - pAverageExclusive[nIdx+1] = fAveragePrcExclusive; - pMaxExclusive[nIdx] = fMaxMsExclusive; - pMaxExclusive[nIdx+1] = fMaxPrcExclusive; - } - nCount += 2; - } - } - } - nMask <<= 1; - } -} - -#define SBUF_MAX 32 - -void MicroProfileDrawBarArrayCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra) -{ - const uint32_t nHeight = MICROPROFILE_TEXT_HEIGHT; - const uint32_t nTextWidth = 6 * (1+MICROPROFILE_TEXT_WIDTH); - const float fWidth = (float)MICROPROFILE_BAR_WIDTH; - - float* pTimers = ((float**)pExtra)[0]; - float* pTimers2 = ((float**)pExtra)[1]; - MicroProfile& S = *MicroProfileGet(); - char sBuffer[SBUF_MAX]; - if (pTimers2 && pTimers2[nIdx] > 0.1f) - snprintf(sBuffer, SBUF_MAX-1, "%5.2f %3.1fx", pTimers[nIdx], pTimers[nIdx] / pTimers2[nIdx]); - else - snprintf(sBuffer, SBUF_MAX-1, "%5.2f", pTimers[nIdx]); - if (!pTimers2) - MicroProfileDrawBox(nX + nTextWidth, nY, nX + nTextWidth + fWidth * pTimers[nIdx+1], nY + nHeight, UI.nOpacityForeground|S.TimerInfo[nTimer].nColor, MicroProfileBoxTypeBar); - MicroProfileDrawText(nX, nY, UINT32_MAX, sBuffer, (uint32_t)strlen(sBuffer)); -} - - -uint32_t MicroProfileDrawBarArray(int32_t nX, int32_t nY, float* pTimers, const char* pName, uint32_t nTotalHeight, float* pTimers2 = NULL) -{ - const uint32_t nTextWidth = 6 * (1+MICROPROFILE_TEXT_WIDTH); - const uint32_t nWidth = MICROPROFILE_BAR_WIDTH; - - MicroProfileDrawLineVertical(nX-5, 0, nTotalHeight+nY, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]); - float* pTimersArray[2] = {pTimers, pTimers2}; - MicroProfileLoopActiveGroupsDraw(nX, nY, pName, MicroProfileDrawBarArrayCallback, pTimersArray); - MicroProfileDrawHeader(nX, nTextWidth + nWidth, pName); - return nWidth + 5 + nTextWidth; - -} -void MicroProfileDrawBarCallCountCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra) -{ - MicroProfile& S = *MicroProfileGet(); - char sBuffer[SBUF_MAX]; - int nLen = snprintf(sBuffer, SBUF_MAX-1, "%5d", S.Frame[nTimer].nCount);//fix - MicroProfileDrawText(nX, nY, UINT32_MAX, sBuffer, nLen); -} - -uint32_t MicroProfileDrawBarCallCount(int32_t nX, int32_t nY, const char* pName) -{ - MicroProfileLoopActiveGroupsDraw(nX, nY, pName, MicroProfileDrawBarCallCountCallback, 0); - const uint32_t nTextWidth = 6 * MICROPROFILE_TEXT_WIDTH; - MicroProfileDrawHeader(nX, 5 + nTextWidth, pName); - return 5 + nTextWidth; -} - -struct MicroProfileMetaAverageArgs -{ - uint64_t* pCounters; - float fRcpFrames; -}; - -void MicroProfileDrawBarMetaAverageCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra) -{ - MicroProfileMetaAverageArgs* pArgs = (MicroProfileMetaAverageArgs*)pExtra; - uint64_t* pCounters = pArgs->pCounters; - float fRcpFrames = pArgs->fRcpFrames; - char sBuffer[SBUF_MAX]; - int nLen = snprintf(sBuffer, SBUF_MAX-1, "%5.2f", pCounters[nTimer] * fRcpFrames); - MicroProfileDrawText(nX - nLen * (MICROPROFILE_TEXT_WIDTH+1), nY, UINT32_MAX, sBuffer, nLen); -} - -uint32_t MicroProfileDrawBarMetaAverage(int32_t nX, int32_t nY, uint64_t* pCounters, const char* pName, uint32_t nTotalHeight) -{ - if(!pName) - return 0; - MicroProfileDrawLineVertical(nX-5, 0, nTotalHeight+nY, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]); - uint32_t nTextWidth = (1+MICROPROFILE_TEXT_WIDTH) * MicroProfileMax(6, (uint32_t)strlen(pName)); - float fRcpFrames = 1.f / (MicroProfileGet()->nAggregateFrames ? MicroProfileGet()->nAggregateFrames : 1); - MicroProfileMetaAverageArgs Args = {pCounters, fRcpFrames}; - MicroProfileLoopActiveGroupsDraw(nX + nTextWidth, nY, pName, MicroProfileDrawBarMetaAverageCallback, &Args); - MicroProfileDrawHeader(nX, 5 + nTextWidth, pName); - return 5 + nTextWidth; -} - - -void MicroProfileDrawBarMetaCountCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra) -{ - uint64_t* pCounters = (uint64_t*)pExtra; - char sBuffer[SBUF_MAX]; - int nLen = snprintf(sBuffer, SBUF_MAX-1, "%5" PRIu64, pCounters[nTimer]); - MicroProfileDrawText(nX - nLen * (MICROPROFILE_TEXT_WIDTH+1), nY, UINT32_MAX, sBuffer, nLen); -} - -uint32_t MicroProfileDrawBarMetaCount(int32_t nX, int32_t nY, uint64_t* pCounters, const char* pName, uint32_t nTotalHeight) -{ - if(!pName) - return 0; - - MicroProfileDrawLineVertical(nX-5, 0, nTotalHeight+nY, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]); - uint32_t nTextWidth = (1+MICROPROFILE_TEXT_WIDTH) * MicroProfileMax(6, (uint32_t)strlen(pName)); - MicroProfileLoopActiveGroupsDraw(nX + nTextWidth, nY, pName, MicroProfileDrawBarMetaCountCallback, pCounters); - MicroProfileDrawHeader(nX, 5 + nTextWidth, pName); - return 5 + nTextWidth; -} - -void MicroProfileDrawBarLegendCallback(uint32_t nTimer, uint32_t nIdx, uint64_t nGroupMask, uint32_t nX, uint32_t nY, void* pExtra) -{ - MicroProfile& S = *MicroProfileGet(); - if (S.TimerInfo[nTimer].bGraph) - { - MicroProfileDrawText(nX, nY, S.TimerInfo[nTimer].nColor, ">", 1); - } - MicroProfileDrawTextRight(nX, nY, S.TimerInfo[nTimer].nColor, S.TimerInfo[nTimer].pName, (uint32_t)strlen(S.TimerInfo[nTimer].pName)); - if(UI.nMouseY >= nY && UI.nMouseY < nY + MICROPROFILE_TEXT_HEIGHT+1) - { - UI.nHoverToken = nTimer; - UI.nHoverTime = 0; - } -} - -uint32_t MicroProfileDrawBarLegend(int32_t nX, int32_t nY, uint32_t nTotalHeight, uint32_t nMaxWidth) -{ - MicroProfileDrawLineVertical(nX-5, nY, nTotalHeight, UI.nOpacityBackground | g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]); - MicroProfileLoopActiveGroupsDraw(nMaxWidth, nY, 0, MicroProfileDrawBarLegendCallback, 0); - return nX; -} - -bool MicroProfileDrawGraph(uint32_t nScreenWidth, uint32_t nScreenHeight) -{ - MicroProfile& S = *MicroProfileGet(); - - MICROPROFILE_SCOPE(g_MicroProfileDrawGraph); - bool bEnabled = false; - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - if(S.Graph[i].nToken != MICROPROFILE_INVALID_TOKEN) - bEnabled = true; - if(!bEnabled) - return false; - - uint32_t nX = nScreenWidth - MICROPROFILE_GRAPH_WIDTH; - uint32_t nY = nScreenHeight - MICROPROFILE_GRAPH_HEIGHT; - MicroProfileDrawBox(nX, nY, nX + MICROPROFILE_GRAPH_WIDTH, nY + MICROPROFILE_GRAPH_HEIGHT, 0x88000000 | g_nMicroProfileBackColors[0]); - bool bMouseOver = UI.nMouseX >= nX && UI.nMouseY >= nY; - float fMouseXPrc =(float(UI.nMouseX - nX)) / MICROPROFILE_GRAPH_WIDTH; - if(bMouseOver) - { - float fXAvg = fMouseXPrc * MICROPROFILE_GRAPH_WIDTH + nX; - MicroProfileDrawLineVertical(fXAvg, nY, nY + MICROPROFILE_GRAPH_HEIGHT, UINT32_MAX); - } - - - float fY = (float)nScreenHeight; - float fDX = MICROPROFILE_GRAPH_WIDTH * 1.f / MICROPROFILE_GRAPH_HISTORY; - float fDY = MICROPROFILE_GRAPH_HEIGHT; - uint32_t nPut = S.nGraphPut; - float* pGraphData = (float*)alloca(sizeof(float)* MICROPROFILE_GRAPH_HISTORY*2); - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - { - if(S.Graph[i].nToken != MICROPROFILE_INVALID_TOKEN) - { - uint32_t nGroupId = MicroProfileGetGroupIndex(S.Graph[i].nToken); - bool bGpu = S.GroupInfo[nGroupId].Type == MicroProfileTokenTypeGpu; - float fToMs = MicroProfileTickToMsMultiplier(bGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu()); - float fToPrc = fToMs * S.fRcpReferenceTime * 3 / 4; - - float fX = (float)nX; - for(uint32_t j = 0; j < MICROPROFILE_GRAPH_HISTORY; ++j) - { - float fWeigth = MicroProfileMin(fToPrc * (S.Graph[i].nHistory[(j+nPut)%MICROPROFILE_GRAPH_HISTORY]), 1.f); - pGraphData[(j*2)] = fX; - pGraphData[(j*2)+1] = fY - fDY * fWeigth; - fX += fDX; - } - MicroProfileDrawLine2D(MICROPROFILE_GRAPH_HISTORY, pGraphData, S.TimerInfo[MicroProfileGetTimerIndex(S.Graph[i].nToken)].nColor); - } - } - { - float fY1 = 0.25f * MICROPROFILE_GRAPH_HEIGHT + nY; - float fY2 = 0.50f * MICROPROFILE_GRAPH_HEIGHT + nY; - float fY3 = 0.75f * MICROPROFILE_GRAPH_HEIGHT + nY; - MicroProfileDrawLineHorizontal(nX, nX + MICROPROFILE_GRAPH_WIDTH, fY1, 0xffdd4444); - MicroProfileDrawLineHorizontal(nX, nX + MICROPROFILE_GRAPH_WIDTH, fY2, 0xff000000| g_nMicroProfileBackColors[0]); - MicroProfileDrawLineHorizontal(nX, nX + MICROPROFILE_GRAPH_WIDTH, fY3, 0xff000000|g_nMicroProfileBackColors[0]); - - char buf[32]; - int nLen = snprintf(buf, sizeof(buf)-1, "%5.2fms", S.fReferenceTime); - MicroProfileDrawText(nX+1, fY1 - (2+MICROPROFILE_TEXT_HEIGHT), UINT32_MAX, buf, nLen); - } - - - - if(bMouseOver) - { - uint32_t pColors[MICROPROFILE_MAX_GRAPHS]; - MicroProfileStringArray Strings; - MicroProfileStringArrayClear(&Strings); - uint32_t nTextCount = 0; - uint32_t nGraphIndex = (S.nGraphPut + MICROPROFILE_GRAPH_HISTORY - int(MICROPROFILE_GRAPH_HISTORY*(1.f - fMouseXPrc))) % MICROPROFILE_GRAPH_HISTORY; - - uint32_t nX = UI.nMouseX; - uint32_t nY = UI.nMouseY + 20; - - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - { - if(S.Graph[i].nToken != MICROPROFILE_INVALID_TOKEN) - { - uint32_t nGroupId = MicroProfileGetGroupIndex(S.Graph[i].nToken); - bool bGpu = S.GroupInfo[nGroupId].Type == MicroProfileTokenTypeGpu; - float fToMs = MicroProfileTickToMsMultiplier(bGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu()); - uint32_t nIndex = MicroProfileGetTimerIndex(S.Graph[i].nToken); - uint32_t nColor = S.TimerInfo[nIndex].nColor; - const char* pName = S.TimerInfo[nIndex].pName; - pColors[nTextCount++] = nColor; - MicroProfileStringArrayAddLiteral(&Strings, pName); - MicroProfileStringArrayFormat(&Strings, "%5.2fms", fToMs * (S.Graph[i].nHistory[nGraphIndex])); - } - } - if(nTextCount) - { - MicroProfileDrawFloatWindow(nX, nY, Strings.ppStrings, Strings.nNumStrings, 0, pColors); - } - - if(UI.nMouseRight) - { - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - { - S.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN; - } - } - } - - return bMouseOver; -} - -void MicroProfileDumpTimers() -{ - MicroProfile& S = *MicroProfileGet(); - - uint64_t nActiveGroup = S.nGroupMask; - - uint32_t nNumTimers = S.nTotalTimers; - uint32_t nBlockSize = 2 * nNumTimers; - float* pTimers = (float*)alloca(nBlockSize * 7 * sizeof(float)); - float* pAverage = pTimers + nBlockSize; - float* pMax = pTimers + 2 * nBlockSize; - float* pCallAverage = pTimers + 3 * nBlockSize; - float* pTimersExclusive = pTimers + 4 * nBlockSize; - float* pAverageExclusive = pTimers + 5 * nBlockSize; - float* pMaxExclusive = pTimers + 6 * nBlockSize; - MicroProfileCalcTimers(pTimers, pAverage, pMax, pCallAverage, pTimersExclusive, pAverageExclusive, pMaxExclusive, nActiveGroup, nNumTimers); - - MICROPROFILE_PRINTF("%11s, ", "Time"); - MICROPROFILE_PRINTF("%11s, ", "Average"); - MICROPROFILE_PRINTF("%11s, ", "Max"); - MICROPROFILE_PRINTF("%11s, ", "Call Avg"); - MICROPROFILE_PRINTF("%9s, ", "Count"); - MICROPROFILE_PRINTF("%11s, ", "Excl"); - MICROPROFILE_PRINTF("%11s, ", "Avg Excl"); - MICROPROFILE_PRINTF("%11s, \n", "Max Excl"); - - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - uint64_t nMask = 1ULL << j; - if(nMask & nActiveGroup) - { - MICROPROFILE_PRINTF("%s\n", S.GroupInfo[j].pName); - for(uint32_t i = 0; i < S.nTotalTimers;++i) - { - uint64_t nTokenMask = MicroProfileGetGroupMask(S.TimerInfo[i].nToken); - if(nTokenMask & nMask) - { - uint32_t nIdx = i * 2; - MICROPROFILE_PRINTF("%9.2fms, ", pTimers[nIdx]); - MICROPROFILE_PRINTF("%9.2fms, ", pAverage[nIdx]); - MICROPROFILE_PRINTF("%9.2fms, ", pMax[nIdx]); - MICROPROFILE_PRINTF("%9.2fms, ", pCallAverage[nIdx]); - MICROPROFILE_PRINTF("%9d, ", S.Frame[i].nCount); - MICROPROFILE_PRINTF("%9.2fms, ", pTimersExclusive[nIdx]); - MICROPROFILE_PRINTF("%9.2fms, ", pAverageExclusive[nIdx]); - MICROPROFILE_PRINTF("%9.2fms, ", pMaxExclusive[nIdx]); - MICROPROFILE_PRINTF("%s\n", S.TimerInfo[i].pName); - } - } - } - } -} - -void MicroProfileDrawBarView(uint32_t nScreenWidth, uint32_t nScreenHeight) -{ - MicroProfile& S = *MicroProfileGet(); - - uint64_t nActiveGroup = S.nAllGroupsWanted ? S.nGroupMask : S.nActiveGroupWanted; - if(!nActiveGroup) - return; - MICROPROFILE_SCOPE(g_MicroProfileDrawBarView); - - const uint32_t nHeight = MICROPROFILE_TEXT_HEIGHT; - int nColorIndex = 0; - uint32_t nMaxTimerNameLen = 1; - uint32_t nNumTimers = 0; - uint32_t nNumGroups = 0; - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - if(nActiveGroup & (1ULL << j)) - { - nNumTimers += S.GroupInfo[j].nNumTimers; - nNumGroups += 1; - nMaxTimerNameLen = MicroProfileMax(nMaxTimerNameLen, S.GroupInfo[j].nMaxTimerNameLen); - } - } - uint32_t nTimerWidth = 2+(4+nMaxTimerNameLen) * (MICROPROFILE_TEXT_WIDTH+1); - uint32_t nX = nTimerWidth + UI.nOffsetX; - uint32_t nY = nHeight + 3 - UI.nOffsetY; - uint32_t nBlockSize = 2 * nNumTimers; - float* pTimers = (float*)alloca(nBlockSize * 7 * sizeof(float)); - float* pAverage = pTimers + nBlockSize; - float* pMax = pTimers + 2 * nBlockSize; - float* pCallAverage = pTimers + 3 * nBlockSize; - float* pTimersExclusive = pTimers + 4 * nBlockSize; - float* pAverageExclusive = pTimers + 5 * nBlockSize; - float* pMaxExclusive = pTimers + 6 * nBlockSize; - MicroProfileCalcTimers(pTimers, pAverage, pMax, pCallAverage, pTimersExclusive, pAverageExclusive, pMaxExclusive, nActiveGroup, nNumTimers); - uint32_t nWidth = 0; - { - uint32_t nMetaIndex = 0; - for(uint32_t i = 1; i ; i <<= 1) - { - if(S.nBars & i) - { - if(i >= MP_DRAW_META_FIRST) - { - if(nMetaIndex < MICROPROFILE_META_MAX && S.MetaCounters[nMetaIndex].pName) - { - uint32_t nStrWidth = strlen(S.MetaCounters[nMetaIndex].pName); - if(S.nBars & MP_DRAW_TIMERS) - nWidth += 6 + (1+MICROPROFILE_TEXT_WIDTH) * (nStrWidth); - if(S.nBars & MP_DRAW_AVERAGE) - nWidth += 6 + (1+MICROPROFILE_TEXT_WIDTH) * (nStrWidth + 4); - if(S.nBars & MP_DRAW_MAX) - nWidth += 6 + (1+MICROPROFILE_TEXT_WIDTH) * (nStrWidth + 4); - } - } - else - { - nWidth += MICROPROFILE_BAR_WIDTH + 6 + 6 * (1+MICROPROFILE_TEXT_WIDTH); - if(i & MP_DRAW_CALL_COUNT) - nWidth += 6 + 6 * MICROPROFILE_TEXT_WIDTH; - } - } - if(i >= MP_DRAW_META_FIRST) - { - ++nMetaIndex; - } - } - nWidth += (1+nMaxTimerNameLen) * (MICROPROFILE_TEXT_WIDTH+1); - for(uint32_t i = 0; i < nNumTimers+nNumGroups+1; ++i) - { - uint32_t nY0 = nY + i * (nHeight + 1); - bool bInside = (UI.nActiveMenu == UINT32_MAX) && ((UI.nMouseY >= nY0) && (UI.nMouseY < (nY0 + nHeight + 1))); - MicroProfileDrawBox(nX, nY0, nWidth+nX, nY0 + (nHeight+1)+1, UI.nOpacityBackground | (g_nMicroProfileBackColors[nColorIndex++ & 1] + ((bInside) ? 0x002c2c2c : 0))); - } - nX += 10; - } - int nTotalHeight = (nNumTimers+nNumGroups+1) * (nHeight+1); - uint32_t nLegendOffset = 1; - if(S.nBars & MP_DRAW_TIMERS) - nX += MicroProfileDrawBarArray(nX, nY, pTimers, "Time", nTotalHeight) + 1; - if(S.nBars & MP_DRAW_AVERAGE) - nX += MicroProfileDrawBarArray(nX, nY, pAverage, "Average", nTotalHeight) + 1; - if(S.nBars & MP_DRAW_MAX) - nX += MicroProfileDrawBarArray(nX, nY, pMax, (!UI.bShowSpikes) ? "Max Time" : "Max Time, Spike", nTotalHeight, UI.bShowSpikes ? pAverage : NULL) + 1; - if(S.nBars & MP_DRAW_CALL_COUNT) - { - nX += MicroProfileDrawBarArray(nX, nY, pCallAverage, "Call Average", nTotalHeight) + 1; - nX += MicroProfileDrawBarCallCount(nX, nY, "Count") + 1; - } - if(S.nBars & MP_DRAW_TIMERS_EXCLUSIVE) - nX += MicroProfileDrawBarArray(nX, nY, pTimersExclusive, "Exclusive Time", nTotalHeight) + 1; - if(S.nBars & MP_DRAW_AVERAGE_EXCLUSIVE) - nX += MicroProfileDrawBarArray(nX, nY, pAverageExclusive, "Exclusive Average", nTotalHeight) + 1; - if(S.nBars & MP_DRAW_MAX_EXCLUSIVE) - nX += MicroProfileDrawBarArray(nX, nY, pMaxExclusive, (!UI.bShowSpikes) ? "Exclusive Max Time" :"Excl Max Time, Spike", nTotalHeight, UI.bShowSpikes ? pAverageExclusive : NULL) + 1; - - for(int i = 0; i < MICROPROFILE_META_MAX; ++i) - { - if(0 != (S.nBars & (MP_DRAW_META_FIRST<= nY0) && (UI.nMouseY < (nY0 + nHeight + 1))); - MicroProfileDrawBox(nX, nY0, nTimerWidth, nY0 + (nHeight+1)+1, 0xff0000000 | (g_nMicroProfileBackColors[nColorIndex++ & 1] + ((bInside) ? 0x002c2c2c : 0))); - } - nX += MicroProfileDrawBarLegend(nX, nY, nTotalHeight, nTimerWidth-5) + 1; - - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - if(nActiveGroup & (1ULL << j)) - { - MicroProfileDrawText(nX, nY + (1+nHeight) * nLegendOffset, UINT32_MAX, S.GroupInfo[j].pName, S.GroupInfo[j].nNameLen); - nLegendOffset += S.GroupInfo[j].nNumTimers+1; - } - } - MicroProfileDrawHeader(nX, nTimerWidth-5, "Group"); - MicroProfileDrawTextRight(nTimerWidth-3, MICROPROFILE_TEXT_HEIGHT + 2, UINT32_MAX, "Timer", 5); - MicroProfileDrawLineVertical(nTimerWidth, 0, nTotalHeight+nY, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]); - MicroProfileDrawLineHorizontal(0, nWidth, 2*MICROPROFILE_TEXT_HEIGHT + 3, UI.nOpacityBackground|g_nMicroProfileBackColors[0]|g_nMicroProfileBackColors[1]); -} - -typedef const char* (*MicroProfileSubmenuCallback)(int, bool* bSelected); -typedef void (*MicroProfileClickCallback)(int); - - -const char* MicroProfileUIMenuMode(int nIndex, bool* bSelected) -{ - MicroProfile& S = *MicroProfileGet(); - switch(nIndex) - { - case 0: - *bSelected = S.nDisplay == MP_DRAW_DETAILED; - return "Detailed"; - case 1: - *bSelected = S.nDisplay == MP_DRAW_BARS; - return "Timers"; - case 2: - *bSelected = S.nDisplay == MP_DRAW_HIDDEN; - return "Hidden"; - case 3: - *bSelected = true; - return "Off"; - case 4: - *bSelected = true; - return "------"; - case 5: - *bSelected = S.nForceEnable != 0; - return "Force Enable"; - - default: return 0; - } -} - -const char* MicroProfileUIMenuGroups(int nIndex, bool* bSelected) -{ - MicroProfile& S = *MicroProfileGet(); - *bSelected = false; - if(nIndex == 0) - { - *bSelected = S.nAllGroupsWanted != 0; - return "[ALL]"; - } - else - { - nIndex = nIndex-1; - if(static_cast(nIndex) < UI.GroupMenuCount) - { - MicroProfileGroupMenuItem& Item = UI.GroupMenu[nIndex]; - static char buffer[MICROPROFILE_NAME_MAX_LEN+32]; - if(Item.nIsCategory) - { - uint64_t nGroupMask = S.CategoryInfo[Item.nIndex].nGroupMask; - *bSelected = nGroupMask == (nGroupMask & S.nActiveGroupWanted); - snprintf(buffer, sizeof(buffer)-1, "[%s]", Item.pName); - } - else - { - *bSelected = 0 != (S.nActiveGroupWanted & (1ULL << Item.nIndex)); - snprintf(buffer, sizeof(buffer)-1, " %s", Item.pName); - } - return buffer; - } - return 0; - } -} - -const char* MicroProfileUIMenuAggregate(int nIndex, bool* bSelected) -{ - MicroProfile& S = *MicroProfileGet(); - if(static_cast(nIndex) < g_MicroProfileAggregatePresets.size()) - { - uint32_t val = g_MicroProfileAggregatePresets[nIndex]; - *bSelected = S.nAggregateFlip == val; - if (0 == val) - { - return "Infinite"; - } - else - { - static char buf[128]; - snprintf(buf, sizeof(buf)-1, "%7u", val); - return buf; - } - } - return 0; - -} - -const char* MicroProfileUIMenuTimers(int nIndex, bool* bSelected) -{ - MicroProfile& S = *MicroProfileGet(); - *bSelected = 0 != (S.nBars & (1 << nIndex)); - switch(nIndex) - { - case 0: return "Time"; - case 1: return "Average"; - case 2: return "Max"; - case 3: return "Call Count"; - case 4: return "Exclusive Timers"; - case 5: return "Exclusive Average"; - case 6: return "Exclusive Max"; - } - int nMetaIndex = nIndex - 7; - if(nMetaIndex < MICROPROFILE_META_MAX) - { - return S.MetaCounters[nMetaIndex].pName; - } - return 0; -} - -const char* MicroProfileUIMenuOptions(int nIndex, bool* bSelected) -{ - MicroProfile& S = *MicroProfileGet(); - if(nIndex >= MICROPROFILE_OPTION_SIZE) return 0; - switch(UI.Options[nIndex].nSubType) - { - case 0: - *bSelected = S.fReferenceTime == g_MicroProfileReferenceTimePresets[UI.Options[nIndex].nIndex]; - break; - case 1: - *bSelected = UI.nOpacityBackground>>24 == g_MicroProfileOpacityPresets[UI.Options[nIndex].nIndex]; - break; - case 2: - *bSelected = UI.nOpacityForeground>>24 == g_MicroProfileOpacityPresets[UI.Options[nIndex].nIndex]; - break; - case 3: - *bSelected = UI.bShowSpikes; - break; -#if MICROPROFILE_CONTEXT_SWITCH_TRACE - case 4: - { - switch(UI.Options[nIndex].nIndex) - { - case 0: - *bSelected = S.bContextSwitchRunning; - break; - case 1: - *bSelected = S.bContextSwitchAllThreads; - break; - case 2: - *bSelected = S.bContextSwitchNoBars; - break; - } - } - break; -#endif - } - return UI.Options[nIndex].Text; -} - -const char* MicroProfileUIMenuPreset(int nIndex, bool* bSelected) -{ - static char buf[128]; - *bSelected = false; - int nNumPresets = static_cast(g_MicroProfilePresetNames.size()); - int nIndexSave = nIndex - nNumPresets - 1; - if (nIndex == nNumPresets) - { - return "--"; - } - else if(nIndexSave >=0 && nIndexSave < nNumPresets) - { - snprintf(buf, sizeof(buf)-1, "Save '%s'", g_MicroProfilePresetNames[nIndexSave]); - return buf; - } - else if(nIndex < nNumPresets) - { - snprintf(buf, sizeof(buf)-1, "Load '%s'", g_MicroProfilePresetNames[nIndex]); - return buf; - } - else - { - return 0; - } -} - -const char* MicroProfileUIMenuCustom(int nIndex, bool* bSelected) -{ - if(UINT32_MAX == UI.nCustomActive) - { - *bSelected = nIndex == 0; - } - else - { - *bSelected = nIndex-2 == static_cast(UI.nCustomActive); - } - switch(nIndex) - { - case 0: return "Disable"; - case 1: return "--"; - default: - nIndex -= 2; - if(static_cast(nIndex) < UI.nCustomCount) - { - return UI.Custom[nIndex].pName; - } - else - { - return 0; - } - } -} - -const char* MicroProfileUIMenuEmpty(int nIndex, bool* bSelected) -{ - return 0; -} - - -void MicroProfileUIClickMode(int nIndex) -{ - MicroProfile& S = *MicroProfileGet(); - switch(nIndex) - { - case 0: - S.nDisplay = MP_DRAW_DETAILED; - break; - case 1: - S.nDisplay = MP_DRAW_BARS; - break; - case 2: - S.nDisplay = MP_DRAW_HIDDEN; - break; - case 3: - S.nDisplay = 0; - break; - case 4: - break; - case 5: - S.nForceEnable = !S.nForceEnable; - break; - } -} - -void MicroProfileUIClickGroups(int nIndex) -{ - MicroProfile& S = *MicroProfileGet(); - if(nIndex == 0) - S.nAllGroupsWanted = 1-S.nAllGroupsWanted; - else - { - nIndex -= 1; - if(static_cast(nIndex) < UI.GroupMenuCount) - { - MicroProfileGroupMenuItem& Item = UI.GroupMenu[nIndex]; - if(Item.nIsCategory) - { - uint64_t nGroupMask = S.CategoryInfo[Item.nIndex].nGroupMask; - if(nGroupMask != (nGroupMask & S.nActiveGroupWanted)) - { - S.nActiveGroupWanted |= nGroupMask; - } - else - { - S.nActiveGroupWanted &= ~nGroupMask; - } - } - else - { - MP_ASSERT(Item.nIndex < S.nGroupCount); - S.nActiveGroupWanted ^= (1ULL << Item.nIndex); - } - } - } -} - -void MicroProfileUIClickAggregate(int nIndex) -{ - MicroProfile& S = *MicroProfileGet(); - S.nAggregateFlip = g_MicroProfileAggregatePresets[nIndex]; - if(0 == S.nAggregateFlip) - { - S.nAggregateClear = 1; - } -} - -void MicroProfileUIClickTimers(int nIndex) -{ - MicroProfile& S = *MicroProfileGet(); - S.nBars ^= (1 << nIndex); -} - -void MicroProfileUIClickOptions(int nIndex) -{ - MicroProfile& S = *MicroProfileGet(); - switch(UI.Options[nIndex].nSubType) - { - case 0: - S.fReferenceTime = g_MicroProfileReferenceTimePresets[UI.Options[nIndex].nIndex]; - S.fRcpReferenceTime = 1.f / S.fReferenceTime; - break; - case 1: - UI.nOpacityBackground = g_MicroProfileOpacityPresets[UI.Options[nIndex].nIndex]<<24; - break; - case 2: - UI.nOpacityForeground = g_MicroProfileOpacityPresets[UI.Options[nIndex].nIndex]<<24; - break; - case 3: - UI.bShowSpikes = !UI.bShowSpikes; - break; -#if MICROPROFILE_CONTEXT_SWITCH_TRACE - case 4: - { - switch(UI.Options[nIndex].nIndex) - { - case 0: - if(S.bContextSwitchRunning) - { - MicroProfileStopContextSwitchTrace(); - } - else - { - MicroProfileStartContextSwitchTrace(); - } - break; - case 1: - S.bContextSwitchAllThreads = !S.bContextSwitchAllThreads; - break; - case 2: - S.bContextSwitchNoBars= !S.bContextSwitchNoBars; - break; - - } - } - break; -#endif - } -} - -void MicroProfileUIClickPreset(int nIndex) -{ - int nNumPresets = static_cast(g_MicroProfilePresetNames.size()); - int nIndexSave = nIndex - nNumPresets - 1; - if(nIndexSave >= 0 && nIndexSave < nNumPresets) - { - MicroProfileSavePreset(g_MicroProfilePresetNames[nIndexSave]); - } - else if(nIndex >= 0 && nIndex < nNumPresets) - { - MicroProfileLoadPreset(g_MicroProfilePresetNames[nIndex]); - } -} - -void MicroProfileUIClickCustom(int nIndex) -{ - if(nIndex == 0) - { - MicroProfileCustomGroupDisable(); - } - else - { - MicroProfileCustomGroupEnable(nIndex-2); - } - -} - -void MicroProfileUIClickEmpty(int nIndex) -{ - -} - - -void MicroProfileDrawMenu(uint32_t nWidth, uint32_t nHeight) -{ - MicroProfile& S = *MicroProfileGet(); - - uint32_t nX = 0; - uint32_t nY = 0; - -#define SBUF_SIZE 256 - char buffer[256]; - MicroProfileDrawBox(nX, nY, nX + nWidth, nY + (MICROPROFILE_TEXT_HEIGHT+1)+1, 0xff000000|g_nMicroProfileBackColors[1]); - -#define MICROPROFILE_MENU_MAX 16 - const char* pMenuText[MICROPROFILE_MENU_MAX] = {0}; - uint32_t nMenuX[MICROPROFILE_MENU_MAX] = {0}; - uint32_t nNumMenuItems = 0; - - int nLen = snprintf(buffer, 127, "MicroProfile"); - MicroProfileDrawText(nX, nY, UINT32_MAX, buffer, nLen); - nX += (sizeof("MicroProfile")+2) * (MICROPROFILE_TEXT_WIDTH+1); - pMenuText[nNumMenuItems++] = "Mode"; - pMenuText[nNumMenuItems++] = "Groups"; - char AggregateText[64]; - snprintf(AggregateText, sizeof(AggregateText)-1, "Aggregate[%d]", S.nAggregateFlip ? S.nAggregateFlip : S.nAggregateFlipCount); - pMenuText[nNumMenuItems++] = &AggregateText[0]; - pMenuText[nNumMenuItems++] = "Timers"; - pMenuText[nNumMenuItems++] = "Options"; - pMenuText[nNumMenuItems++] = "Preset"; - pMenuText[nNumMenuItems++] = "Custom"; - const int nPauseIndex = nNumMenuItems; - pMenuText[nNumMenuItems++] = S.nRunning ? "Pause" : "Unpause"; - pMenuText[nNumMenuItems++] = "Help"; - - if(S.nOverflow) - { - pMenuText[nNumMenuItems++] = "!BUFFERSFULL!"; - } - - - if(UI.GroupMenuCount != S.nGroupCount + S.nCategoryCount) - { - UI.GroupMenuCount = S.nGroupCount + S.nCategoryCount; - for(uint32_t i = 0; i < S.nCategoryCount; ++i) - { - UI.GroupMenu[i].nIsCategory = 1; - UI.GroupMenu[i].nCategoryIndex = i; - UI.GroupMenu[i].nIndex = i; - UI.GroupMenu[i].pName = S.CategoryInfo[i].pName; - } - for(uint32_t i = 0; i < S.nGroupCount; ++i) - { - uint32_t idx = i + S.nCategoryCount; - UI.GroupMenu[idx].nIsCategory = 0; - UI.GroupMenu[idx].nCategoryIndex = S.GroupInfo[i].nCategory; - UI.GroupMenu[idx].nIndex = i; - UI.GroupMenu[idx].pName = S.GroupInfo[i].pName; - } - std::sort(&UI.GroupMenu[0], &UI.GroupMenu[UI.GroupMenuCount], - [] (const MicroProfileGroupMenuItem& l, const MicroProfileGroupMenuItem& r) -> bool - { - if(l.nCategoryIndex < r.nCategoryIndex) - { - return true; - } - else if(r.nCategoryIndex < l.nCategoryIndex) - { - return false; - } - if(r.nIsCategory || l.nIsCategory) - { - return l.nIsCategory > r.nIsCategory; - } - return MP_STRCASECMP(l.pName, r.pName)<0; - } - ); - } - - MicroProfileSubmenuCallback GroupCallback[MICROPROFILE_MENU_MAX] = - { - MicroProfileUIMenuMode, - MicroProfileUIMenuGroups, - MicroProfileUIMenuAggregate, - MicroProfileUIMenuTimers, - MicroProfileUIMenuOptions, - MicroProfileUIMenuPreset, - MicroProfileUIMenuCustom, - MicroProfileUIMenuEmpty, - MicroProfileUIMenuEmpty, - MicroProfileUIMenuEmpty, - }; - - MicroProfileClickCallback CBClick[MICROPROFILE_MENU_MAX] = - { - MicroProfileUIClickMode, - MicroProfileUIClickGroups, - MicroProfileUIClickAggregate, - MicroProfileUIClickTimers, - MicroProfileUIClickOptions, - MicroProfileUIClickPreset, - MicroProfileUIClickCustom, - MicroProfileUIClickEmpty, - MicroProfileUIClickEmpty, - MicroProfileUIClickEmpty, - }; - - - uint32_t nSelectMenu = UINT32_MAX; - for(uint32_t i = 0; i < nNumMenuItems; ++i) - { - nMenuX[i] = nX; - uint32_t nLen = (uint32_t)strlen(pMenuText[i]); - uint32_t nEnd = nX + nLen * (MICROPROFILE_TEXT_WIDTH+1); - if(UI.nMouseY <= MICROPROFILE_TEXT_HEIGHT && UI.nMouseX <= nEnd && UI.nMouseX >= nX) - { - MicroProfileDrawBox(nX-1, nY, nX + nLen * (MICROPROFILE_TEXT_WIDTH+1), nY +(MICROPROFILE_TEXT_HEIGHT+1)+1, 0xff888888); - nSelectMenu = i; - if((UI.nMouseLeft || UI.nMouseRight) && i == (uint32_t)nPauseIndex) - { - S.nToggleRunning = 1; - } - } - MicroProfileDrawText(nX, nY, UINT32_MAX, pMenuText[i], (uint32_t)strlen(pMenuText[i])); - nX += (nLen+1) * (MICROPROFILE_TEXT_WIDTH+1); - } - uint32_t nMenu = nSelectMenu != UINT32_MAX ? nSelectMenu : UI.nActiveMenu; - UI.nActiveMenu = nMenu; - if(UINT32_MAX != nMenu) - { - nX = nMenuX[nMenu]; - nY += MICROPROFILE_TEXT_HEIGHT+1; - MicroProfileSubmenuCallback CB = GroupCallback[nMenu]; - int nNumLines = 0; - bool bSelected = false; - const char* pString = CB(nNumLines, &bSelected); - uint32_t nWidth = 0, nHeight = 0; - while(pString) - { - nWidth = MicroProfileMax(nWidth, (int)strlen(pString)); - nNumLines++; - pString = CB(nNumLines, &bSelected); - } - nWidth = (2+nWidth) * (MICROPROFILE_TEXT_WIDTH+1); - nHeight = nNumLines * (MICROPROFILE_TEXT_HEIGHT+1); - if(UI.nMouseY <= nY + nHeight+0 && UI.nMouseY >= nY-0 && UI.nMouseX <= nX + nWidth + 0 && UI.nMouseX >= nX - 0) - { - UI.nActiveMenu = nMenu; - } - else if(nSelectMenu == UINT32_MAX) - { - UI.nActiveMenu = UINT32_MAX; - } - MicroProfileDrawBox(nX, nY, nX + nWidth, nY + nHeight, 0xff000000|g_nMicroProfileBackColors[1]); - for(int i = 0; i < nNumLines; ++i) - { - bool bSelected = false; - const char* pString = CB(i, &bSelected); - if(UI.nMouseY >= nY && UI.nMouseY < nY + MICROPROFILE_TEXT_HEIGHT + 1) - { - if(UI.nMouseLeft || UI.nMouseRight) - { - CBClick[nMenu](i); - } - MicroProfileDrawBox(nX, nY, nX + nWidth, nY + MICROPROFILE_TEXT_HEIGHT + 1, 0xff888888); - } - int nLen = snprintf(buffer, SBUF_SIZE-1, "%c %s", bSelected ? '*' : ' ' ,pString); - MicroProfileDrawText(nX, nY, UINT32_MAX, buffer, nLen); - nY += MICROPROFILE_TEXT_HEIGHT+1; - } - } - - - { - static char FrameTimeMessage[64]; - float fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()); - uint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1; - float fMs = fToMs * (S.nFlipTicks); - float fAverageMs = fToMs * (S.nFlipAggregateDisplay / nAggregateFrames); - float fMaxMs = fToMs * S.nFlipMaxDisplay; - int nLen = snprintf(FrameTimeMessage, sizeof(FrameTimeMessage)-1, "Time[%6.2f] Avg[%6.2f] Max[%6.2f]", fMs, fAverageMs, fMaxMs); - pMenuText[nNumMenuItems++] = &FrameTimeMessage[0]; - MicroProfileDrawText(nWidth - nLen * (MICROPROFILE_TEXT_WIDTH+1), 0, UINT32_MAX, FrameTimeMessage, nLen); - } -} - - -void MicroProfileMoveGraph() -{ - - int nZoom = UI.nMouseWheelDelta; - int nPanX = 0; - int nPanY = 0; - static int X = 0, Y = 0; - if(UI.nMouseDownLeft && !UI.nModDown) - { - nPanX = UI.nMouseX - X; - nPanY = UI.nMouseY - Y; - } - X = UI.nMouseX; - Y = UI.nMouseY; - - if(nZoom) - { - float fOldRange = UI.fDetailedRange; - if(nZoom>0) - { - UI.fDetailedRangeTarget = UI.fDetailedRange *= UI.nModDown ? 1.40f : 1.05f; - } - else - { - float fNewDetailedRange = UI.fDetailedRange / (UI.nModDown ? 1.40f : 1.05f); - if(fNewDetailedRange < 1e-4f) //100ns - fNewDetailedRange = 1e-4f; - UI.fDetailedRangeTarget = UI.fDetailedRange = fNewDetailedRange; - } - - float fDiff = fOldRange - UI.fDetailedRange; - float fMousePrc = MicroProfileMax((float)UI.nMouseX / UI.nWidth ,0.f); - UI.fDetailedOffsetTarget = UI.fDetailedOffset += fDiff * fMousePrc; - - } - if(nPanX) - { - UI.fDetailedOffsetTarget = UI.fDetailedOffset += -nPanX * UI.fDetailedRange / UI.nWidth; - } - UI.nOffsetY -= nPanY; - UI.nOffsetX += nPanX; - if(UI.nOffsetX > 0) - UI.nOffsetX = 0; - if(UI.nOffsetY<0) - UI.nOffsetY = 0; -} - -void MicroProfileDrawCustom(uint32_t nWidth, uint32_t nHeight) -{ - if(UINT32_MAX != UI.nCustomActive) - { - MicroProfile& S = *MicroProfileGet(); - MP_ASSERT(UI.nCustomActive < MICROPROFILE_CUSTOM_MAX); - MicroProfileCustom* pCustom = &UI.Custom[UI.nCustomActive]; - uint32_t nCount = pCustom->nNumTimers; - uint32_t nAggregateFrames = S.nAggregateFrames ? S.nAggregateFrames : 1; - uint32_t nExtraOffset = 1 + ((pCustom->nFlags & MICROPROFILE_CUSTOM_STACK) != 0 ? 3 : 0); - uint32_t nOffsetYBase = nHeight - (nExtraOffset+nCount)* (1+MICROPROFILE_TEXT_HEIGHT) - MICROPROFILE_CUSTOM_PADDING; - uint32_t nOffsetY = nOffsetYBase; - float fReference = pCustom->fReference; - float fRcpReference = 1.f / fReference; - uint32_t nReducedWidth = UI.nWidth - 2*MICROPROFILE_CUSTOM_PADDING - MICROPROFILE_GRAPH_WIDTH; - - char Buffer[MICROPROFILE_NAME_MAX_LEN*2+1]; - float* pTime = (float*)alloca(sizeof(float)*nCount); - float* pTimeAvg = (float*)alloca(sizeof(float)*nCount); - float* pTimeMax = (float*)alloca(sizeof(float)*nCount); - uint32_t* pColors = (uint32_t*)alloca(sizeof(uint32_t)*nCount); - uint32_t nMaxOffsetX = 0; - MicroProfileDrawBox(MICROPROFILE_CUSTOM_PADDING-1, nOffsetY-1, MICROPROFILE_CUSTOM_PADDING+nReducedWidth+1, UI.nHeight - MICROPROFILE_CUSTOM_PADDING+1, 0x88000000|g_nMicroProfileBackColors[0]); - - for(uint32_t i = 0; i < nCount; ++i) - { - uint16_t nTimerIndex = MicroProfileGetTimerIndex(pCustom->pTimers[i]); - uint16_t nGroupIndex = MicroProfileGetGroupIndex(pCustom->pTimers[i]); - float fToMs = MicroProfileTickToMsMultiplier(S.GroupInfo[nGroupIndex].Type == MicroProfileTokenTypeGpu ? MicroProfileTicksPerSecondGpu() : MicroProfileTicksPerSecondCpu()); - pTime[i] = S.Frame[nTimerIndex].nTicks * fToMs; - pTimeAvg[i] = fToMs * (S.Aggregate[nTimerIndex].nTicks / nAggregateFrames); - pTimeMax[i] = fToMs * (S.AggregateMax[nTimerIndex]); - pColors[i] = S.TimerInfo[nTimerIndex].nColor; - } - - MicroProfileDrawText(MICROPROFILE_CUSTOM_PADDING + 3*MICROPROFILE_TEXT_WIDTH, nOffsetY, UINT32_MAX, "Avg", sizeof("Avg")-1); - MicroProfileDrawText(MICROPROFILE_CUSTOM_PADDING + 13*MICROPROFILE_TEXT_WIDTH, nOffsetY, UINT32_MAX, "Max", sizeof("Max")-1); - for(uint32_t i = 0; i < nCount; ++i) - { - nOffsetY += (1+MICROPROFILE_TEXT_HEIGHT); - uint16_t nTimerIndex = MicroProfileGetTimerIndex(pCustom->pTimers[i]); - uint16_t nGroupIndex = MicroProfileGetGroupIndex(pCustom->pTimers[i]); - MicroProfileTimerInfo* pTimerInfo = &S.TimerInfo[nTimerIndex]; - int nSize; - uint32_t nOffsetX = MICROPROFILE_CUSTOM_PADDING; - nSize = snprintf(Buffer, sizeof(Buffer)-1, "%6.2f", pTimeAvg[i]); - MicroProfileDrawText(nOffsetX, nOffsetY, UINT32_MAX, Buffer, nSize); - nOffsetX += (nSize+2) * (MICROPROFILE_TEXT_WIDTH+1); - nSize = snprintf(Buffer, sizeof(Buffer)-1, "%6.2f", pTimeMax[i]); - MicroProfileDrawText(nOffsetX, nOffsetY, UINT32_MAX, Buffer, nSize); - nOffsetX += (nSize+2) * (MICROPROFILE_TEXT_WIDTH+1); - nSize = snprintf(Buffer, sizeof(Buffer)-1, "%s:%s", S.GroupInfo[nGroupIndex].pName, pTimerInfo->pName); - MicroProfileDrawText(nOffsetX, nOffsetY, pTimerInfo->nColor, Buffer, nSize); - nOffsetX += (nSize+2) * (MICROPROFILE_TEXT_WIDTH+1); - nMaxOffsetX = MicroProfileMax(nMaxOffsetX, nOffsetX); - } - uint32_t nMaxWidth = nReducedWidth- nMaxOffsetX; - - if(pCustom->nFlags & MICROPROFILE_CUSTOM_BARS) - { - nOffsetY = nOffsetYBase; - float* pMs = pCustom->nFlags & MICROPROFILE_CUSTOM_BAR_SOURCE_MAX ? pTimeMax : pTimeAvg; - const char* pString = pCustom->nFlags & MICROPROFILE_CUSTOM_BAR_SOURCE_MAX ? "Max" : "Avg"; - MicroProfileDrawText(nMaxOffsetX, nOffsetY, UINT32_MAX, pString, static_cast(strlen(pString))); - int nSize = snprintf(Buffer, sizeof(Buffer)-1, "%6.2fms", fReference); - MicroProfileDrawText(nReducedWidth - (1+nSize) * (MICROPROFILE_TEXT_WIDTH+1), nOffsetY, UINT32_MAX, Buffer, nSize); - for(uint32_t i = 0; i < nCount; ++i) - { - nOffsetY += (1+MICROPROFILE_TEXT_HEIGHT); - uint32_t nWidth = MicroProfileMin(nMaxWidth, (uint32_t)(nMaxWidth * pMs[i] * fRcpReference)); - MicroProfileDrawBox(nMaxOffsetX, nOffsetY, nMaxOffsetX+nWidth, nOffsetY+MICROPROFILE_TEXT_HEIGHT, pColors[i]|0xff000000); - } - } - if(pCustom->nFlags & MICROPROFILE_CUSTOM_STACK) - { - nOffsetY += 2*(1+MICROPROFILE_TEXT_HEIGHT); - const char* pString = pCustom->nFlags & MICROPROFILE_CUSTOM_STACK_SOURCE_MAX ? "Max" : "Avg"; - MicroProfileDrawText(MICROPROFILE_CUSTOM_PADDING, nOffsetY, UINT32_MAX, pString, static_cast(strlen(pString))); - int nSize = snprintf(Buffer, sizeof(Buffer)-1, "%6.2fms", fReference); - MicroProfileDrawText(nReducedWidth - (1+nSize) * (MICROPROFILE_TEXT_WIDTH+1), nOffsetY, UINT32_MAX, Buffer, nSize); - nOffsetY += (1+MICROPROFILE_TEXT_HEIGHT); - float fPosX = MICROPROFILE_CUSTOM_PADDING; - float* pMs = pCustom->nFlags & MICROPROFILE_CUSTOM_STACK_SOURCE_MAX ? pTimeMax : pTimeAvg; - for(uint32_t i = 0; i < nCount; ++i) - { - float fWidth = pMs[i] * fRcpReference * nReducedWidth; - uint32_t nX = fPosX; - fPosX += fWidth; - uint32_t nXEnd = fPosX; - if(nX < nXEnd) - { - MicroProfileDrawBox(nX, nOffsetY, nXEnd, nOffsetY+MICROPROFILE_TEXT_HEIGHT, pColors[i]|0xff000000); - } - } - } - } -} -void MicroProfileDraw(uint32_t nWidth, uint32_t nHeight) -{ - MICROPROFILE_SCOPE(g_MicroProfileDraw); - MicroProfile& S = *MicroProfileGet(); - - { - static int once = 0; - if(0 == once) - { - std::recursive_mutex& m = MicroProfileGetMutex(); - m.lock(); - MicroProfileInitUI(); - - - - uint32_t nDisplay = S.nDisplay; - MicroProfileLoadPreset(MICROPROFILE_DEFAULT_PRESET); - once++; - S.nDisplay = nDisplay;// dont load display, just state - m.unlock(); - - } - } - - - if(S.nDisplay) - { - std::recursive_mutex& m = MicroProfileGetMutex(); - m.lock(); - UI.nWidth = nWidth; - UI.nHeight = nHeight; - UI.nHoverToken = MICROPROFILE_INVALID_TOKEN; - UI.nHoverTime = 0; - UI.nHoverFrame = -1; - if(S.nDisplay != MP_DRAW_DETAILED) - S.nContextSwitchHoverThread = S.nContextSwitchHoverThreadAfter = S.nContextSwitchHoverThreadBefore = UINT32_MAX; - MicroProfileMoveGraph(); - - - if(S.nDisplay == MP_DRAW_DETAILED) - { - MicroProfileDrawDetailedView(nWidth, nHeight); - } - else if(S.nDisplay == MP_DRAW_BARS && S.nBars) - { - MicroProfileDrawBarView(nWidth, nHeight); - } - - MicroProfileDrawMenu(nWidth, nHeight); - bool bMouseOverGraph = MicroProfileDrawGraph(nWidth, nHeight); - MicroProfileDrawCustom(nWidth, nHeight); - bool bHidden = S.nDisplay == MP_DRAW_HIDDEN; - if(!bHidden) - { - uint32_t nLockedToolTipX = 3; - bool bDeleted = false; - for(int i = 0; i < MICROPROFILE_TOOLTIP_MAX_LOCKED; ++i) - { - int nIndex = (g_MicroProfileUI.LockedToolTipFront + i) % MICROPROFILE_TOOLTIP_MAX_LOCKED; - if(g_MicroProfileUI.LockedToolTips[nIndex].ppStrings[0]) - { - uint32_t nToolTipWidth = 0, nToolTipHeight = 0; - MicroProfileFloatWindowSize(g_MicroProfileUI.LockedToolTips[nIndex].ppStrings, g_MicroProfileUI.LockedToolTips[nIndex].nNumStrings, 0, nToolTipWidth, nToolTipHeight, 0); - uint32_t nStartY = nHeight - nToolTipHeight - 2; - if(!bDeleted && UI.nMouseY > nStartY && UI.nMouseX > nLockedToolTipX && UI.nMouseX <= nLockedToolTipX + nToolTipWidth && (UI.nMouseLeft || UI.nMouseRight) ) - { - bDeleted = true; - int j = i; - for(; j < MICROPROFILE_TOOLTIP_MAX_LOCKED-1; ++j) - { - int nIndex0 = (g_MicroProfileUI.LockedToolTipFront + j) % MICROPROFILE_TOOLTIP_MAX_LOCKED; - int nIndex1 = (g_MicroProfileUI.LockedToolTipFront + j+1) % MICROPROFILE_TOOLTIP_MAX_LOCKED; - MicroProfileStringArrayCopy(&g_MicroProfileUI.LockedToolTips[nIndex0], &g_MicroProfileUI.LockedToolTips[nIndex1]); - } - MicroProfileStringArrayClear(&g_MicroProfileUI.LockedToolTips[(g_MicroProfileUI.LockedToolTipFront + j) % MICROPROFILE_TOOLTIP_MAX_LOCKED]); - } - else - { - MicroProfileDrawFloatWindow(nLockedToolTipX, nHeight-nToolTipHeight-2, &g_MicroProfileUI.LockedToolTips[nIndex].ppStrings[0], g_MicroProfileUI.LockedToolTips[nIndex].nNumStrings, g_MicroProfileUI.nLockedToolTipColor[nIndex]); - nLockedToolTipX += nToolTipWidth + 4; - } - } - } - - if(UI.nActiveMenu == 8) - { - if(S.nDisplay & MP_DRAW_DETAILED) - { - MicroProfileStringArray DetailedHelp; - MicroProfileStringArrayClear(&DetailedHelp); - MicroProfileStringArrayFormat(&DetailedHelp, "%s", MICROPROFILE_HELP_LEFT); - MicroProfileStringArrayAddLiteral(&DetailedHelp, "Toggle Graph"); - MicroProfileStringArrayFormat(&DetailedHelp, "%s", MICROPROFILE_HELP_ALT); - MicroProfileStringArrayAddLiteral(&DetailedHelp, "Zoom"); - MicroProfileStringArrayFormat(&DetailedHelp, "%s + %s", MICROPROFILE_HELP_MOD, MICROPROFILE_HELP_LEFT); - MicroProfileStringArrayAddLiteral(&DetailedHelp, "Lock Tooltip"); - MicroProfileStringArrayAddLiteral(&DetailedHelp, "Drag"); - MicroProfileStringArrayAddLiteral(&DetailedHelp, "Pan View"); - MicroProfileStringArrayAddLiteral(&DetailedHelp, "Mouse Wheel"); - MicroProfileStringArrayAddLiteral(&DetailedHelp, "Zoom"); - MicroProfileDrawFloatWindow(nWidth, MICROPROFILE_FRAME_HISTORY_HEIGHT+20, DetailedHelp.ppStrings, DetailedHelp.nNumStrings, 0xff777777); - - MicroProfileStringArray DetailedHistoryHelp; - MicroProfileStringArrayClear(&DetailedHistoryHelp); - MicroProfileStringArrayFormat(&DetailedHistoryHelp, "%s", MICROPROFILE_HELP_LEFT); - MicroProfileStringArrayAddLiteral(&DetailedHistoryHelp, "Center View"); - MicroProfileStringArrayFormat(&DetailedHistoryHelp, "%s", MICROPROFILE_HELP_ALT); - MicroProfileStringArrayAddLiteral(&DetailedHistoryHelp, "Zoom to frame"); - MicroProfileDrawFloatWindow(nWidth, 20, DetailedHistoryHelp.ppStrings, DetailedHistoryHelp.nNumStrings, 0xff777777); - - - - } - else if(0 != (S.nDisplay & MP_DRAW_BARS) && S.nBars) - { - MicroProfileStringArray BarHelp; - MicroProfileStringArrayClear(&BarHelp); - MicroProfileStringArrayFormat(&BarHelp, "%s", MICROPROFILE_HELP_LEFT); - MicroProfileStringArrayAddLiteral(&BarHelp, "Toggle Graph"); - MicroProfileStringArrayFormat(&BarHelp, "%s + %s", MICROPROFILE_HELP_MOD, MICROPROFILE_HELP_LEFT); - MicroProfileStringArrayAddLiteral(&BarHelp, "Lock Tooltip"); - MicroProfileStringArrayAddLiteral(&BarHelp, "Drag"); - MicroProfileStringArrayAddLiteral(&BarHelp, "Pan View"); - MicroProfileDrawFloatWindow(nWidth, MICROPROFILE_FRAME_HISTORY_HEIGHT+20, BarHelp.ppStrings, BarHelp.nNumStrings, 0xff777777); - - } - MicroProfileStringArray Debug; - MicroProfileStringArrayClear(&Debug); - MicroProfileStringArrayAddLiteral(&Debug, "Memory Usage"); - MicroProfileStringArrayFormat(&Debug, "%4.2fmb", S.nMemUsage / (1024.f * 1024.f)); - MicroProfileStringArrayAddLiteral(&Debug, "Web Server Port"); - MicroProfileStringArrayFormat(&Debug, "%d", MicroProfileWebServerPort()); - uint32_t nFrameNext = (S.nFrameCurrent+1) % MICROPROFILE_MAX_FRAME_HISTORY; - MicroProfileFrameState* pFrameCurrent = &S.Frames[S.nFrameCurrent]; - MicroProfileFrameState* pFrameNext = &S.Frames[nFrameNext]; - - - MicroProfileStringArrayAddLiteral(&Debug, ""); - MicroProfileStringArrayAddLiteral(&Debug, ""); - MicroProfileStringArrayAddLiteral(&Debug, "Usage"); - MicroProfileStringArrayAddLiteral(&Debug, "markers [frames] "); - -#if MICROPROFILE_CONTEXT_SWITCH_TRACE - MicroProfileStringArrayAddLiteral(&Debug, "Context Switch"); - MicroProfileStringArrayFormat(&Debug, "%9d [%7d]", S.nContextSwitchUsage, MICROPROFILE_CONTEXT_SWITCH_BUFFER_SIZE / S.nContextSwitchUsage ); -#endif - - for(int i = 0; i < MICROPROFILE_MAX_THREADS; ++i) - { - if(pFrameCurrent->nLogStart[i] && S.Pool[i]) - { - uint32_t nEnd = pFrameNext->nLogStart[i]; - uint32_t nStart = pFrameCurrent->nLogStart[i]; - uint32_t nUsage = nStart < nEnd ? (nEnd - nStart) : (nEnd + MICROPROFILE_BUFFER_SIZE - nStart); - uint32_t nFrameSupport = MICROPROFILE_BUFFER_SIZE / nUsage; - MicroProfileStringArrayFormat(&Debug, "%s", &S.Pool[i]->ThreadName[0]); - MicroProfileStringArrayFormat(&Debug, "%9d [%7d]", nUsage, nFrameSupport); - } - } - - MicroProfileDrawFloatWindow(0, nHeight-10, Debug.ppStrings, Debug.nNumStrings, 0xff777777); - } - - - - if(UI.nActiveMenu == UINT32_MAX && !bMouseOverGraph) - { - if(UI.nHoverToken != MICROPROFILE_INVALID_TOKEN) - { - MicroProfileDrawFloatTooltip(UI.nMouseX, UI.nMouseY, UI.nHoverToken, UI.nHoverTime); - } - else if(S.nContextSwitchHoverThreadAfter != UINT32_MAX && S.nContextSwitchHoverThreadBefore != UINT32_MAX) - { - float fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()); - MicroProfileStringArray ToolTip; - MicroProfileStringArrayClear(&ToolTip); - MicroProfileStringArrayAddLiteral(&ToolTip, "Context Switch"); - MicroProfileStringArrayFormat(&ToolTip, "%04x", S.nContextSwitchHoverThread); - MicroProfileStringArrayAddLiteral(&ToolTip, "Before"); - MicroProfileStringArrayFormat(&ToolTip, "%04x", S.nContextSwitchHoverThreadBefore); - MicroProfileStringArrayAddLiteral(&ToolTip, "After"); - MicroProfileStringArrayFormat(&ToolTip, "%04x", S.nContextSwitchHoverThreadAfter); - MicroProfileStringArrayAddLiteral(&ToolTip, "Duration"); - int64_t nDifference = MicroProfileLogTickDifference(S.nContextSwitchHoverTickIn, S.nContextSwitchHoverTickOut); - MicroProfileStringArrayFormat(&ToolTip, "%6.2fms", fToMs * nDifference ); - MicroProfileStringArrayAddLiteral(&ToolTip, "CPU"); - MicroProfileStringArrayFormat(&ToolTip, "%d", S.nContextSwitchHoverCpu); - MicroProfileDrawFloatWindow(UI.nMouseX, UI.nMouseY+20, &ToolTip.ppStrings[0], ToolTip.nNumStrings, UINT32_MAX); - - - } - else if(UI.nHoverFrame != -1) - { - uint32_t nNextFrame = (UI.nHoverFrame+1)%MICROPROFILE_MAX_FRAME_HISTORY; - int64_t nTick = S.Frames[UI.nHoverFrame].nFrameStartCpu; - int64_t nTickNext = S.Frames[nNextFrame].nFrameStartCpu; - int64_t nTickGpu = S.Frames[UI.nHoverFrame].nFrameStartGpu; - int64_t nTickNextGpu = S.Frames[nNextFrame].nFrameStartGpu; - - float fToMs = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondCpu()); - float fToMsGpu = MicroProfileTickToMsMultiplier(MicroProfileTicksPerSecondGpu()); - float fMs = fToMs * (nTickNext - nTick); - float fMsGpu = fToMsGpu * (nTickNextGpu - nTickGpu); - MicroProfileStringArray ToolTip; - MicroProfileStringArrayClear(&ToolTip); - MicroProfileStringArrayFormat(&ToolTip, "Frame %d", UI.nHoverFrame); - #if MICROPROFILE_DEBUG - MicroProfileStringArrayFormat(&ToolTip, "%p", &S.Frames[UI.nHoverFrame]); - #else - MicroProfileStringArrayAddLiteral(&ToolTip, ""); - #endif - MicroProfileStringArrayAddLiteral(&ToolTip, "CPU Time"); - MicroProfileStringArrayFormat(&ToolTip, "%6.2fms", fMs); - MicroProfileStringArrayAddLiteral(&ToolTip, "GPU Time"); - MicroProfileStringArrayFormat(&ToolTip, "%6.2fms", fMsGpu); - #if MICROPROFILE_DEBUG - for(int i = 0; i < MICROPROFILE_MAX_THREADS; ++i) - { - if(S.Frames[UI.nHoverFrame].nLogStart[i]) - { - MicroProfileStringArrayFormat(&ToolTip, "%d", i); - MicroProfileStringArrayFormat(&ToolTip, "%d", S.Frames[UI.nHoverFrame].nLogStart[i]); - } - } - #endif - MicroProfileDrawFloatWindow(UI.nMouseX, UI.nMouseY+20, &ToolTip.ppStrings[0], ToolTip.nNumStrings, UINT32_MAX); - } - if(UI.nMouseLeft) - { - if(UI.nHoverToken != MICROPROFILE_INVALID_TOKEN) - MicroProfileToggleGraph(UI.nHoverToken); - } - } - } - -#if MICROPROFILE_DRAWCURSOR - { - float fCursor[8] = - { - MicroProfileMax(0, (int)UI.nMouseX-3), UI.nMouseY, - MicroProfileMin(nWidth, UI.nMouseX+3), UI.nMouseY, - UI.nMouseX, MicroProfileMax((int)UI.nMouseY-3, 0), - UI.nMouseX, MicroProfileMin(nHeight, UI.nMouseY+3), - }; - MicroProfileDrawLine2D(2, &fCursor[0], 0xff00ff00); - MicroProfileDrawLine2D(2, &fCursor[4], 0xff00ff00); - } -#endif - m.unlock(); - } - else if(UI.nCustomActive != UINT32_MAX) - { - std::recursive_mutex& m = MicroProfileGetMutex(); - m.lock(); - MicroProfileDrawGraph(nWidth, nHeight); - MicroProfileDrawCustom(nWidth, nHeight); - m.unlock(); - - } - UI.nMouseLeft = UI.nMouseRight = 0; - UI.nMouseLeftMod = UI.nMouseRightMod = 0; - UI.nMouseWheelDelta = 0; - if(S.nOverflow) - S.nOverflow--; - - UI.fDetailedOffset = UI.fDetailedOffset + (UI.fDetailedOffsetTarget - UI.fDetailedOffset) * MICROPROFILE_ANIM_DELAY_PRC; - UI.fDetailedRange = UI.fDetailedRange + (UI.fDetailedRangeTarget - UI.fDetailedRange) * MICROPROFILE_ANIM_DELAY_PRC; - - -} - -bool MicroProfileIsDrawing() -{ - MicroProfile& S = *MicroProfileGet(); - return S.nDisplay != 0; -} - -void MicroProfileToggleGraph(MicroProfileToken nToken) -{ - MicroProfile& S = *MicroProfileGet(); - uint32_t nTimerId = MicroProfileGetTimerIndex(nToken); - nToken &= 0xffff; - int32_t nMinSort = 0x7fffffff; - int32_t nFreeIndex = -1; - int32_t nMinIndex = 0; - int32_t nMaxSort = 0x80000000; - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - { - if(S.Graph[i].nToken == MICROPROFILE_INVALID_TOKEN) - nFreeIndex = i; - if(S.Graph[i].nToken == nToken) - { - S.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN; - S.TimerInfo[nTimerId].bGraph = false; - return; - } - if(S.Graph[i].nKey < nMinSort) - { - nMinSort = S.Graph[i].nKey; - nMinIndex = i; - } - if(S.Graph[i].nKey > nMaxSort) - { - nMaxSort = S.Graph[i].nKey; - } - } - int nIndex = nFreeIndex > -1 ? nFreeIndex : nMinIndex; - if (nFreeIndex == -1) - { - uint32_t idx = MicroProfileGetTimerIndex(S.Graph[nIndex].nToken); - S.TimerInfo[idx].bGraph = false; - } - S.Graph[nIndex].nToken = nToken; - S.Graph[nIndex].nKey = nMaxSort+1; - memset(&S.Graph[nIndex].nHistory[0], 0, sizeof(S.Graph[nIndex].nHistory)); - S.TimerInfo[nTimerId].bGraph = true; -} - - -void MicroProfileMousePosition(uint32_t nX, uint32_t nY, int nWheelDelta) -{ - UI.nMouseX = nX; - UI.nMouseY = nY; - UI.nMouseWheelDelta = nWheelDelta; -} - -void MicroProfileModKey(uint32_t nKeyState) -{ - UI.nModDown = nKeyState ? 1 : 0; -} - -void MicroProfileClearGraph() -{ - MicroProfile& S = *MicroProfileGet(); - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - { - if(S.Graph[i].nToken != 0) - { - S.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN; - } - } -} - -void MicroProfileMouseButton(uint32_t nLeft, uint32_t nRight) -{ - bool bCanRelease = abs((int)(UI.nMouseDownX - UI.nMouseX)) + abs((int)(UI.nMouseDownY - UI.nMouseY)) < 3; - - if(0 == nLeft && UI.nMouseDownLeft && bCanRelease) - { - if(UI.nModDown) - UI.nMouseLeftMod = 1; - else - UI.nMouseLeft = 1; - } - - if(0 == nRight && UI.nMouseDownRight && bCanRelease) - { - if(UI.nModDown) - UI.nMouseRightMod = 1; - else - UI.nMouseRight = 1; - } - if((nLeft || nRight) && !(UI.nMouseDownLeft || UI.nMouseDownRight)) - { - UI.nMouseDownX = UI.nMouseX; - UI.nMouseDownY = UI.nMouseY; - } - - UI.nMouseDownLeft = nLeft; - UI.nMouseDownRight = nRight; - -} - -void MicroProfileDrawLineVertical(int nX, int nTop, int nBottom, uint32_t nColor) -{ - MicroProfileDrawBox(nX, nTop, nX + 1, nBottom, nColor); -} - -void MicroProfileDrawLineHorizontal(int nLeft, int nRight, int nY, uint32_t nColor) -{ - MicroProfileDrawBox(nLeft, nY, nRight, nY + 1, nColor); -} - - - -#include - -#define MICROPROFILE_PRESET_HEADER_MAGIC 0x28586813 -#define MICROPROFILE_PRESET_HEADER_VERSION 0x00000102 -struct MicroProfilePresetHeader -{ - uint32_t nMagic; - uint32_t nVersion; - //groups, threads, aggregate, reference frame, graphs timers - uint32_t nGroups[MICROPROFILE_MAX_GROUPS]; - uint32_t nThreads[MICROPROFILE_MAX_THREADS]; - uint32_t nGraphName[MICROPROFILE_MAX_GRAPHS]; - uint32_t nGraphGroupName[MICROPROFILE_MAX_GRAPHS]; - uint32_t nAllGroupsWanted; - uint32_t nAllThreadsWanted; - uint32_t nAggregateFlip; - float fReferenceTime; - uint32_t nBars; - uint32_t nDisplay; - uint32_t nOpacityBackground; - uint32_t nOpacityForeground; - uint32_t nShowSpikes; -}; - -#ifndef MICROPROFILE_PRESET_FILENAME_FUNC -#define MICROPROFILE_PRESET_FILENAME_FUNC MicroProfilePresetFilename -static const char* MicroProfilePresetFilename(const char* pSuffix) -{ - static char filename[512]; - snprintf(filename, sizeof(filename)-1, ".microprofilepreset.%s", pSuffix); - return filename; -} -#endif - -void MicroProfileSavePreset(const char* pPresetName) -{ - std::lock_guard Lock(MicroProfileGetMutex()); - FILE* F = fopen(MICROPROFILE_PRESET_FILENAME_FUNC(pPresetName), "wb"); - if(!F) return; - - MicroProfile& S = *MicroProfileGet(); - - MicroProfilePresetHeader Header; - memset(&Header, 0, sizeof(Header)); - Header.nAggregateFlip = S.nAggregateFlip; - Header.nBars = S.nBars; - Header.fReferenceTime = S.fReferenceTime; - Header.nAllGroupsWanted = S.nAllGroupsWanted; - Header.nAllThreadsWanted = S.nAllThreadsWanted; - Header.nMagic = MICROPROFILE_PRESET_HEADER_MAGIC; - Header.nVersion = MICROPROFILE_PRESET_HEADER_VERSION; - Header.nDisplay = S.nDisplay; - Header.nOpacityBackground = UI.nOpacityBackground; - Header.nOpacityForeground = UI.nOpacityForeground; - Header.nShowSpikes = UI.bShowSpikes ? 1 : 0; - fwrite(&Header, sizeof(Header), 1, F); - uint64_t nMask = 1; - for(uint32_t i = 0; i < MICROPROFILE_MAX_GROUPS; ++i) - { - if(S.nActiveGroupWanted & nMask) - { - uint32_t offset = ftell(F); - const char* pName = S.GroupInfo[i].pName; - int nLen = (int)strlen(pName)+1; - fwrite(pName, nLen, 1, F); - Header.nGroups[i] = offset; - } - nMask <<= 1; - } - for(uint32_t i = 0; i < MICROPROFILE_MAX_THREADS; ++i) - { - MicroProfileThreadLog* pLog = S.Pool[i]; - if(pLog && S.nThreadActive[i]) - { - uint32_t nOffset = ftell(F); - const char* pName = &pLog->ThreadName[0]; - int nLen = (int)strlen(pName)+1; - fwrite(pName, nLen, 1, F); - Header.nThreads[i] = nOffset; - } - } - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - { - MicroProfileToken nToken = S.Graph[i].nToken; - if(nToken != MICROPROFILE_INVALID_TOKEN) - { - uint32_t nGroupIndex = MicroProfileGetGroupIndex(nToken); - uint32_t nTimerIndex = MicroProfileGetTimerIndex(nToken); - const char* pGroupName = S.GroupInfo[nGroupIndex].pName; - const char* pTimerName = S.TimerInfo[nTimerIndex].pName; - MP_ASSERT(pGroupName); - MP_ASSERT(pTimerName); - int nGroupLen = (int)strlen(pGroupName)+1; - int nTimerLen = (int)strlen(pTimerName)+1; - - uint32_t nOffsetGroup = ftell(F); - fwrite(pGroupName, nGroupLen, 1, F); - uint32_t nOffsetTimer = ftell(F); - fwrite(pTimerName, nTimerLen, 1, F); - Header.nGraphName[i] = nOffsetTimer; - Header.nGraphGroupName[i] = nOffsetGroup; - } - } - fseek(F, 0, SEEK_SET); - fwrite(&Header, sizeof(Header), 1, F); - - fclose(F); - -} - - - -void MicroProfileLoadPreset(const char* pSuffix) -{ - std::lock_guard Lock(MicroProfileGetMutex()); - FILE* F = fopen(MICROPROFILE_PRESET_FILENAME_FUNC(pSuffix), "rb"); - if(!F) - { - return; - } - fseek(F, 0, SEEK_END); - int nSize = ftell(F); - char* const pBuffer = (char*)alloca(nSize); - fseek(F, 0, SEEK_SET); - int nRead = (int)fread(pBuffer, nSize, 1, F); - fclose(F); - if(1 != nRead) - return; - - MicroProfile& S = *MicroProfileGet(); - - MicroProfilePresetHeader& Header = *(MicroProfilePresetHeader*)pBuffer; - - if(Header.nMagic != MICROPROFILE_PRESET_HEADER_MAGIC || Header.nVersion != MICROPROFILE_PRESET_HEADER_VERSION) - { - return; - } - - S.nAggregateFlip = Header.nAggregateFlip; - S.nBars = Header.nBars; - S.fReferenceTime = Header.fReferenceTime; - S.fRcpReferenceTime = 1.f / Header.fReferenceTime; - S.nAllGroupsWanted = Header.nAllGroupsWanted; - S.nAllThreadsWanted = Header.nAllThreadsWanted; - S.nDisplay = Header.nDisplay; - S.nActiveGroupWanted = 0; - UI.nOpacityBackground = Header.nOpacityBackground; - UI.nOpacityForeground = Header.nOpacityForeground; - UI.bShowSpikes = Header.nShowSpikes == 1; - - memset(&S.nThreadActive[0], 0, sizeof(S.nThreadActive)); - - for(uint32_t i = 0; i < MICROPROFILE_MAX_GROUPS; ++i) - { - if(Header.nGroups[i]) - { - const char* pGroupName = pBuffer + Header.nGroups[i]; - for(uint32_t j = 0; j < MICROPROFILE_MAX_GROUPS; ++j) - { - if(0 == MP_STRCASECMP(pGroupName, S.GroupInfo[j].pName)) - { - S.nActiveGroupWanted |= (1ULL << j); - } - } - } - } - for(uint32_t i = 0; i < MICROPROFILE_MAX_THREADS; ++i) - { - if(Header.nThreads[i]) - { - const char* pThreadName = pBuffer + Header.nThreads[i]; - for(uint32_t j = 0; j < MICROPROFILE_MAX_THREADS; ++j) - { - MicroProfileThreadLog* pLog = S.Pool[j]; - if(pLog && 0 == MP_STRCASECMP(pThreadName, &pLog->ThreadName[0])) - { - S.nThreadActive[j] = 1; - } - } - } - } - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - { - MicroProfileToken nPrevToken = S.Graph[i].nToken; - S.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN; - if(Header.nGraphName[i] && Header.nGraphGroupName[i]) - { - const char* pGraphName = pBuffer + Header.nGraphName[i]; - const char* pGraphGroupName = pBuffer + Header.nGraphGroupName[i]; - for(uint32_t j = 0; j < S.nTotalTimers; ++j) - { - uint64_t nGroupIndex = S.TimerInfo[j].nGroupIndex; - if(0 == MP_STRCASECMP(pGraphName, S.TimerInfo[j].pName) && 0 == MP_STRCASECMP(pGraphGroupName, S.GroupInfo[nGroupIndex].pName)) - { - MicroProfileToken nToken = MicroProfileMakeToken(1ULL << nGroupIndex, (uint16_t)j); - S.Graph[i].nToken = nToken; // note: group index is stored here but is checked without in MicroProfileToggleGraph()! - S.TimerInfo[j].bGraph = true; - if(nToken != nPrevToken) - { - memset(&S.Graph[i].nHistory, 0, sizeof(S.Graph[i].nHistory)); - } - break; - } - } - } - } -} - -uint32_t MicroProfileCustomGroupFind(const char* pCustomName) -{ - for(uint32_t i = 0; i < UI.nCustomCount; ++i) - { - if(!MP_STRCASECMP(pCustomName, UI.Custom[i].pName)) - { - return i; - } - } - return UINT32_MAX; -} - -uint32_t MicroProfileCustomGroup(const char* pCustomName) -{ - for(uint32_t i = 0; i < UI.nCustomCount; ++i) - { - if(!MP_STRCASECMP(pCustomName, UI.Custom[i].pName)) - { - return i; - } - } - MP_ASSERT(UI.nCustomCount < MICROPROFILE_CUSTOM_MAX); - uint32_t nIndex = UI.nCustomCount; - UI.nCustomCount++; - memset(&UI.Custom[nIndex], 0, sizeof(UI.Custom[nIndex])); - size_t nLen = strlen(pCustomName); - if(nLen > MICROPROFILE_NAME_MAX_LEN-1) - nLen = MICROPROFILE_NAME_MAX_LEN-1; - memcpy(&UI.Custom[nIndex].pName[0], pCustomName, nLen); - UI.Custom[nIndex].pName[nLen] = '\0'; - return nIndex; -} -void MicroProfileCustomGroup(const char* pCustomName, uint32_t nMaxTimers, uint32_t nAggregateFlip, float fReferenceTime, uint32_t nFlags) -{ - uint32_t nIndex = MicroProfileCustomGroup(pCustomName); - MP_ASSERT(UI.Custom[nIndex].pTimers == 0);//only call once! - UI.Custom[nIndex].pTimers = &UI.CustomTimer[UI.nCustomTimerCount]; - UI.Custom[nIndex].nMaxTimers = nMaxTimers; - UI.Custom[nIndex].fReference = fReferenceTime; - UI.nCustomTimerCount += nMaxTimers; - MP_ASSERT(UI.nCustomTimerCount <= MICROPROFILE_CUSTOM_MAX_TIMERS); //bump MICROPROFILE_CUSTOM_MAX_TIMERS - UI.Custom[nIndex].nFlags = nFlags; - UI.Custom[nIndex].nAggregateFlip = nAggregateFlip; -} - -void MicroProfileCustomGroupEnable(uint32_t nIndex) -{ - if(nIndex < UI.nCustomCount) - { - MicroProfile& S = *MicroProfileGet(); - S.nForceGroupUI = UI.Custom[nIndex].nGroupMask; - MicroProfileSetAggregateFrames(UI.Custom[nIndex].nAggregateFlip); - S.fReferenceTime = UI.Custom[nIndex].fReference; - S.fRcpReferenceTime = 1.f / UI.Custom[nIndex].fReference; - UI.nCustomActive = nIndex; - - for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i) - { - if(S.Graph[i].nToken != MICROPROFILE_INVALID_TOKEN) - { - uint32_t nTimerId = MicroProfileGetTimerIndex(S.Graph[i].nToken); - S.TimerInfo[nTimerId].bGraph = false; - S.Graph[i].nToken = MICROPROFILE_INVALID_TOKEN; - } - } - - for(uint32_t i = 0; i < UI.Custom[nIndex].nNumTimers; ++i) - { - if(i == MICROPROFILE_MAX_GRAPHS) - { - break; - } - S.Graph[i].nToken = UI.Custom[nIndex].pTimers[i]; - S.Graph[i].nKey = i; - uint32_t nTimerId = MicroProfileGetTimerIndex(S.Graph[i].nToken); - S.TimerInfo[nTimerId].bGraph = true; - } - } -} - -void MicroProfileCustomGroupToggle(const char* pCustomName) -{ - uint32_t nIndex = MicroProfileCustomGroupFind(pCustomName); - if(nIndex == UINT32_MAX || nIndex == UI.nCustomActive) - { - MicroProfileCustomGroupDisable(); - } - else - { - MicroProfileCustomGroupEnable(nIndex); - } -} - -void MicroProfileCustomGroupEnable(const char* pCustomName) -{ - uint32_t nIndex = MicroProfileCustomGroupFind(pCustomName); - MicroProfileCustomGroupEnable(nIndex); -} -void MicroProfileCustomGroupDisable() -{ - MicroProfile& S = *MicroProfileGet(); - S.nForceGroupUI = 0; - UI.nCustomActive = UINT32_MAX; -} - -void MicroProfileCustomGroupAddTimer(const char* pCustomName, const char* pGroup, const char* pTimer) -{ - uint32_t nIndex = MicroProfileCustomGroupFind(pCustomName); - if(UINT32_MAX == nIndex) - { - return; - } - uint32_t nTimerIndex = UI.Custom[nIndex].nNumTimers; - MP_ASSERT(nTimerIndex < UI.Custom[nIndex].nMaxTimers); - uint64_t nToken = MicroProfileFindToken(pGroup, pTimer); - MP_ASSERT(nToken != MICROPROFILE_INVALID_TOKEN); //Timer must be registered first. - UI.Custom[nIndex].pTimers[nTimerIndex] = nToken; - uint16_t nGroup = MicroProfileGetGroupIndex(nToken); - UI.Custom[nIndex].nGroupMask |= (1ULL << nGroup); - UI.Custom[nIndex].nNumTimers++; -} - -#undef UI - -#endif -#endif diff --git a/externals/tracy/CMakeLists.txt b/externals/tracy/CMakeLists.txt new file mode 100644 index 000000000..5b00b6e66 --- /dev/null +++ b/externals/tracy/CMakeLists.txt @@ -0,0 +1,3 @@ +add_library(tracy STATIC public/TracyClient.cpp) +target_include_directories(tracy PUBLIC public) +target_compile_definitions(tracy PRIVATE TRACY_ENABLE TRACY_ON_DEMAND) diff --git a/externals/tracy/public/TracyClient.cpp b/externals/tracy/public/TracyClient.cpp new file mode 100644 index 000000000..26387b762 --- /dev/null +++ b/externals/tracy/public/TracyClient.cpp @@ -0,0 +1,58 @@ +// +// Tracy profiler +// ---------------- +// +// For fast integration, compile and +// link with this source file (and none +// other) in your executable (or in the +// main DLL / shared object on multi-DLL +// projects). +// + +// Define TRACY_ENABLE to enable profiler. + +#include "common/TracySystem.cpp" + +#ifdef TRACY_ENABLE + +#ifdef _MSC_VER +# pragma warning(push, 0) +#endif + +#include "common/tracy_lz4.cpp" +#include "client/TracyProfiler.cpp" +#include "client/TracyCallstack.cpp" +#include "client/TracySysPower.cpp" +#include "client/TracySysTime.cpp" +#include "client/TracySysTrace.cpp" +#include "common/TracySocket.cpp" +#include "client/tracy_rpmalloc.cpp" +#include "client/TracyDxt1.cpp" +#include "client/TracyAlloc.cpp" +#include "client/TracyOverride.cpp" + +#if TRACY_HAS_CALLSTACK == 2 || TRACY_HAS_CALLSTACK == 3 || TRACY_HAS_CALLSTACK == 4 || TRACY_HAS_CALLSTACK == 6 +# include "libbacktrace/alloc.cpp" +# include "libbacktrace/dwarf.cpp" +# include "libbacktrace/fileline.cpp" +# include "libbacktrace/mmapio.cpp" +# include "libbacktrace/posix.cpp" +# include "libbacktrace/sort.cpp" +# include "libbacktrace/state.cpp" +# if TRACY_HAS_CALLSTACK == 4 +# include "libbacktrace/macho.cpp" +# else +# include "libbacktrace/elf.cpp" +# endif +# include "common/TracyStackFrames.cpp" +#endif + +#ifdef _MSC_VER +# pragma comment(lib, "ws2_32.lib") +# pragma comment(lib, "dbghelp.lib") +# pragma comment(lib, "advapi32.lib") +# pragma comment(lib, "user32.lib") +# pragma warning(pop) +#endif + +#endif diff --git a/externals/tracy/public/client/TracyAlloc.cpp b/externals/tracy/public/client/TracyAlloc.cpp new file mode 100644 index 000000000..c675b6d3f --- /dev/null +++ b/externals/tracy/public/client/TracyAlloc.cpp @@ -0,0 +1,43 @@ +#include "../common/TracyAlloc.hpp" + +#ifdef TRACY_USE_RPMALLOC + +#include + +#include "../common/TracyForceInline.hpp" +#include "../common/TracyYield.hpp" + +namespace tracy +{ + +extern thread_local bool RpThreadInitDone; +extern std::atomic RpInitDone; +extern std::atomic RpInitLock; + +tracy_no_inline static void InitRpmallocPlumbing() +{ + const auto done = RpInitDone.load( std::memory_order_acquire ); + if( !done ) + { + int expected = 0; + while( !RpInitLock.compare_exchange_weak( expected, 1, std::memory_order_release, std::memory_order_relaxed ) ) { expected = 0; YieldThread(); } + const auto done = RpInitDone.load( std::memory_order_acquire ); + if( !done ) + { + rpmalloc_initialize(); + RpInitDone.store( 1, std::memory_order_release ); + } + RpInitLock.store( 0, std::memory_order_release ); + } + rpmalloc_thread_initialize(); + RpThreadInitDone = true; +} + +TRACY_API void InitRpmalloc() +{ + if( !RpThreadInitDone ) InitRpmallocPlumbing(); +} + +} + +#endif diff --git a/externals/tracy/public/client/TracyArmCpuTable.hpp b/externals/tracy/public/client/TracyArmCpuTable.hpp new file mode 100644 index 000000000..2b47c3a60 --- /dev/null +++ b/externals/tracy/public/client/TracyArmCpuTable.hpp @@ -0,0 +1,419 @@ +namespace tracy +{ + +#if defined __linux__ && defined __ARM_ARCH + +static const char* DecodeArmImplementer( uint32_t v ) +{ + static char buf[16]; + switch( v ) + { + case 0x41: return "ARM"; + case 0x42: return "Broadcom"; + case 0x43: return "Cavium"; + case 0x44: return "DEC"; + case 0x46: return "Fujitsu"; + case 0x48: return "HiSilicon"; + case 0x49: return "Infineon"; + case 0x4d: return "Motorola"; + case 0x4e: return "Nvidia"; + case 0x50: return "Applied Micro"; + case 0x51: return "Qualcomm"; + case 0x53: return "Samsung"; + case 0x54: return "Texas Instruments"; + case 0x56: return "Marvell"; + case 0x61: return "Apple"; + case 0x66: return "Faraday"; + case 0x68: return "HXT"; + case 0x69: return "Intel"; + case 0xc0: return "Ampere Computing"; + default: break; + } + sprintf( buf, "0x%x", v ); + return buf; +} + +static const char* DecodeArmPart( uint32_t impl, uint32_t part ) +{ + static char buf[16]; + switch( impl ) + { + case 0x41: // ARM + switch( part ) + { + case 0x810: return "810"; + case 0x920: return "920"; + case 0x922: return "922"; + case 0x926: return "926"; + case 0x940: return "940"; + case 0x946: return "946"; + case 0x966: return "966"; + case 0xa20: return "1020"; + case 0xa22: return "1022"; + case 0xa26: return "1026"; + case 0xb02: return "11 MPCore"; + case 0xb36: return "1136"; + case 0xb56: return "1156"; + case 0xb76: return "1176"; + case 0xc05: return " Cortex-A5"; + case 0xc07: return " Cortex-A7"; + case 0xc08: return " Cortex-A8"; + case 0xc09: return " Cortex-A9"; + case 0xc0c: return " Cortex-A12"; + case 0xc0d: return " Rockchip RK3288"; + case 0xc0e: return " Cortex-A17"; + case 0xc0f: return " Cortex-A15"; + case 0xc14: return " Cortex-R4"; + case 0xc15: return " Cortex-R5"; + case 0xc17: return " Cortex-R7"; + case 0xc18: return " Cortex-R8"; + case 0xc20: return " Cortex-M0"; + case 0xc21: return " Cortex-M1"; + case 0xc23: return " Cortex-M3"; + case 0xc24: return " Cortex-M4"; + case 0xc27: return " Cortex-M7"; + case 0xc60: return " Cortex-M0+"; + case 0xd00: return " AArch64 simulator"; + case 0xd01: return " Cortex-A32"; + case 0xd02: return " Cortex-A34"; + case 0xd03: return " Cortex-A53"; + case 0xd04: return " Cortex-A35"; + case 0xd05: return " Cortex-A55"; + case 0xd06: return " Cortex-A65"; + case 0xd07: return " Cortex-A57"; + case 0xd08: return " Cortex-A72"; + case 0xd09: return " Cortex-A73"; + case 0xd0a: return " Cortex-A75"; + case 0xd0b: return " Cortex-A76"; + case 0xd0c: return " Neoverse N1"; + case 0xd0d: return " Cortex-A77"; + case 0xd0e: return " Cortex-A76AE"; + case 0xd0f: return " AEMv8"; + case 0xd13: return " Cortex-R52"; + case 0xd20: return " Cortex-M23"; + case 0xd21: return " Cortex-M33"; + case 0xd22: return " Cortex-M55"; + case 0xd40: return " Neoverse V1"; + case 0xd41: return " Cortex-A78"; + case 0xd42: return " Cortex-A78AE"; + case 0xd43: return " Cortex-A65AE"; + case 0xd44: return " Cortex-X1"; + case 0xd47: return " Cortex-A710"; + case 0xd48: return " Cortex-X2"; + case 0xd49: return " Neoverse N2"; + case 0xd4a: return " Neoverse E1"; + case 0xd4b: return " Cortex-A78C"; + case 0xd4c: return " Cortex-X1C"; + default: break; + } + case 0x42: // Broadcom + switch( part ) + { + case 0xf: return " Brahma B15"; + case 0x100: return " Brahma B53"; + case 0x516: return " ThunderX2"; + default: break; + } + case 0x43: // Cavium + switch( part ) + { + case 0xa0: return " ThunderX"; + case 0xa1: return " ThunderX 88XX"; + case 0xa2: return " ThunderX 81XX"; + case 0xa3: return " ThunderX 83XX"; + case 0xaf: return " ThunderX2 99xx"; + case 0xb0: return " OcteonTX2"; + case 0xb1: return " OcteonTX2 T98"; + case 0xb2: return " OcteonTX2 T96"; + case 0xb3: return " OcteonTX2 F95"; + case 0xb4: return " OcteonTX2 F95N"; + case 0xb5: return " OcteonTX2 F95MM"; + case 0xb6: return " OcteonTX2 F95O"; + case 0xb8: return " ThunderX3 T110"; + default: break; + } + case 0x44: // DEC + switch( part ) + { + case 0xa10: return " SA110"; + case 0xa11: return " SA1100"; + default: break; + } + case 0x46: // Fujitsu + switch( part ) + { + case 0x1: return " A64FX"; + default: break; + } + case 0x48: // HiSilicon + switch( part ) + { + case 0xd01: return " TSV100"; + case 0xd40: return " Kirin 980"; + default: break; + } + case 0x4e: // Nvidia + switch( part ) + { + case 0x0: return " Denver"; + case 0x3: return " Denver 2"; + case 0x4: return " Carmel"; + default: break; + } + case 0x50: // Applied Micro + switch( part ) + { + case 0x0: return " X-Gene"; + default: break; + } + case 0x51: // Qualcomm + switch( part ) + { + case 0xf: return " Scorpion"; + case 0x2d: return " Scorpion"; + case 0x4d: return " Krait"; + case 0x6f: return " Krait"; + case 0x200: return " Kryo"; + case 0x201: return " Kryo Silver (Snapdragon 821)"; + case 0x205: return " Kryo Gold"; + case 0x211: return " Kryo Silver (Snapdragon 820)"; + case 0x800: return " Kryo 260 / 280 Gold"; + case 0x801: return " Kryo 260 / 280 Silver"; + case 0x802: return " Kryo 385 Gold"; + case 0x803: return " Kryo 385 Silver"; + case 0x804: return " Kryo 485 Gold"; + case 0x805: return " Kryo 4xx/5xx Silver"; + case 0xc00: return " Falkor"; + case 0xc01: return " Saphira"; + default: break; + } + case 0x53: // Samsung + switch( part ) + { + case 0x1: return " Exynos M1/M2"; + case 0x2: return " Exynos M3"; + case 0x3: return " Exynos M4"; + case 0x4: return " Exynos M5"; + default: break; + } + case 0x54: // Texas Instruments + switch( part ) + { + case 0x925: return " TI925"; + default: break; + } + case 0x56: // Marvell + switch( part ) + { + case 0x131: return " Feroceon 88FR131"; + case 0x581: return " PJ4 / PJ4B"; + case 0x584: return " PJ4B-MP / PJ4C"; + default: break; + } + case 0x61: // Apple + switch( part ) + { + case 0x1: return " Cyclone"; + case 0x2: return " Typhoon"; + case 0x3: return " Typhoon/Capri"; + case 0x4: return " Twister"; + case 0x5: return " Twister/Elba/Malta"; + case 0x6: return " Hurricane"; + case 0x7: return " Hurricane/Myst"; + case 0x22: return " M1 Icestorm"; + case 0x23: return " M1 Firestorm"; + case 0x24: return " M1 Icestorm Pro"; + case 0x25: return " M1 Firestorm Pro"; + case 0x28: return " M1 Icestorm Max"; + case 0x29: return " M1 Firestorm Max"; + default: break; + } + case 0x66: // Faraday + switch( part ) + { + case 0x526: return " FA526"; + case 0x626: return " FA626"; + default: break; + } + case 0x68: // HXT + switch( part ) + { + case 0x0: return " Phecda"; + default: break; + } + case 0xc0: // Ampere Computing + switch( part ) + { + case 0xac3: return " Ampere1"; + default: break; + } + default: break; + } + sprintf( buf, " 0x%x", part ); + return buf; +} + +#elif defined __APPLE__ && TARGET_OS_IPHONE == 1 + +static const char* DecodeIosDevice( const char* id ) +{ + static const char* DeviceTable[] = { + "i386", "32-bit simulator", + "x86_64", "64-bit simulator", + "iPhone1,1", "iPhone", + "iPhone1,2", "iPhone 3G", + "iPhone2,1", "iPhone 3GS", + "iPhone3,1", "iPhone 4 (GSM)", + "iPhone3,2", "iPhone 4 (GSM)", + "iPhone3,3", "iPhone 4 (CDMA)", + "iPhone4,1", "iPhone 4S", + "iPhone5,1", "iPhone 5 (A1428)", + "iPhone5,2", "iPhone 5 (A1429)", + "iPhone5,3", "iPhone 5c (A1456/A1532)", + "iPhone5,4", "iPhone 5c (A1507/A1516/1526/A1529)", + "iPhone6,1", "iPhone 5s (A1433/A1533)", + "iPhone6,2", "iPhone 5s (A1457/A1518/A1528/A1530)", + "iPhone7,1", "iPhone 6 Plus", + "iPhone7,2", "iPhone 6", + "iPhone8,1", "iPhone 6S", + "iPhone8,2", "iPhone 6S Plus", + "iPhone8,4", "iPhone SE", + "iPhone9,1", "iPhone 7 (CDMA)", + "iPhone9,2", "iPhone 7 Plus (CDMA)", + "iPhone9,3", "iPhone 7 (GSM)", + "iPhone9,4", "iPhone 7 Plus (GSM)", + "iPhone10,1", "iPhone 8 (CDMA)", + "iPhone10,2", "iPhone 8 Plus (CDMA)", + "iPhone10,3", "iPhone X (CDMA)", + "iPhone10,4", "iPhone 8 (GSM)", + "iPhone10,5", "iPhone 8 Plus (GSM)", + "iPhone10,6", "iPhone X (GSM)", + "iPhone11,2", "iPhone XS", + "iPhone11,4", "iPhone XS Max", + "iPhone11,6", "iPhone XS Max China", + "iPhone11,8", "iPhone XR", + "iPhone12,1", "iPhone 11", + "iPhone12,3", "iPhone 11 Pro", + "iPhone12,5", "iPhone 11 Pro Max", + "iPhone12,8", "iPhone SE 2nd Gen", + "iPhone13,1", "iPhone 12 Mini", + "iPhone13,2", "iPhone 12", + "iPhone13,3", "iPhone 12 Pro", + "iPhone13,4", "iPhone 12 Pro Max", + "iPhone14,2", "iPhone 13 Pro", + "iPhone14,3", "iPhone 13 Pro Max", + "iPhone14,4", "iPhone 13 Mini", + "iPhone14,5", "iPhone 13", + "iPhone14,6", "iPhone SE 3rd Gen", + "iPhone14,7", "iPhone 14", + "iPhone14,8", "iPhone 14 Plus", + "iPhone15,2", "iPhone 14 Pro", + "iPhone15,3", "iPhone 14 Pro Max", + "iPhone15,4", "iPhone 15", + "iPhone15,5", "iPhone 15 Plus", + "iPhone16,1", "iPhone 15 Pro", + "iPhone16,2", "iPhone 15 Pro Max", + "iPad1,1", "iPad (A1219/A1337)", + "iPad2,1", "iPad 2 (A1395)", + "iPad2,2", "iPad 2 (A1396)", + "iPad2,3", "iPad 2 (A1397)", + "iPad2,4", "iPad 2 (A1395)", + "iPad2,5", "iPad Mini (A1432)", + "iPad2,6", "iPad Mini (A1454)", + "iPad2,7", "iPad Mini (A1455)", + "iPad3,1", "iPad 3 (A1416)", + "iPad3,2", "iPad 3 (A1403)", + "iPad3,3", "iPad 3 (A1430)", + "iPad3,4", "iPad 4 (A1458)", + "iPad3,5", "iPad 4 (A1459)", + "iPad3,6", "iPad 4 (A1460)", + "iPad4,1", "iPad Air (A1474)", + "iPad4,2", "iPad Air (A1475)", + "iPad4,3", "iPad Air (A1476)", + "iPad4,4", "iPad Mini 2 (A1489)", + "iPad4,5", "iPad Mini 2 (A1490)", + "iPad4,6", "iPad Mini 2 (A1491)", + "iPad4,7", "iPad Mini 3 (A1599)", + "iPad4,8", "iPad Mini 3 (A1600)", + "iPad4,9", "iPad Mini 3 (A1601)", + "iPad5,1", "iPad Mini 4 (A1538)", + "iPad5,2", "iPad Mini 4 (A1550)", + "iPad5,3", "iPad Air 2 (A1566)", + "iPad5,4", "iPad Air 2 (A1567)", + "iPad6,3", "iPad Pro 9.7\" (A1673)", + "iPad6,4", "iPad Pro 9.7\" (A1674)", + "iPad6,5", "iPad Pro 9.7\" (A1675)", + "iPad6,7", "iPad Pro 12.9\" (A1584)", + "iPad6,8", "iPad Pro 12.9\" (A1652)", + "iPad6,11", "iPad 5th gen (A1822)", + "iPad6,12", "iPad 5th gen (A1823)", + "iPad7,1", "iPad Pro 12.9\" 2nd gen (A1670)", + "iPad7,2", "iPad Pro 12.9\" 2nd gen (A1671/A1821)", + "iPad7,3", "iPad Pro 10.5\" (A1701)", + "iPad7,4", "iPad Pro 10.5\" (A1709)", + "iPad7,5", "iPad 6th gen (A1893)", + "iPad7,6", "iPad 6th gen (A1954)", + "iPad7,11", "iPad 7th gen 10.2\" (Wifi)", + "iPad7,12", "iPad 7th gen 10.2\" (Wifi+Cellular)", + "iPad8,1", "iPad Pro 11\" (A1980)", + "iPad8,2", "iPad Pro 11\" (A1980)", + "iPad8,3", "iPad Pro 11\" (A1934/A1979/A2013)", + "iPad8,4", "iPad Pro 11\" (A1934/A1979/A2013)", + "iPad8,5", "iPad Pro 12.9\" 3rd gen (A1876)", + "iPad8,6", "iPad Pro 12.9\" 3rd gen (A1876)", + "iPad8,7", "iPad Pro 12.9\" 3rd gen (A1895/A1983/A2014)", + "iPad8,8", "iPad Pro 12.9\" 3rd gen (A1895/A1983/A2014)", + "iPad8,9", "iPad Pro 11\" 2nd gen (Wifi)", + "iPad8,10", "iPad Pro 11\" 2nd gen (Wifi+Cellular)", + "iPad8,11", "iPad Pro 12.9\" 4th gen (Wifi)", + "iPad8,12", "iPad Pro 12.9\" 4th gen (Wifi+Cellular)", + "iPad11,1", "iPad Mini 5th gen (A2133)", + "iPad11,2", "iPad Mini 5th gen (A2124/A2125/A2126)", + "iPad11,3", "iPad Air 3rd gen (A2152)", + "iPad11,4", "iPad Air 3rd gen (A2123/A2153/A2154)", + "iPad11,6", "iPad 8th gen (WiFi)", + "iPad11,7", "iPad 8th gen (WiFi+Cellular)", + "iPad12,1", "iPad 9th Gen (WiFi)", + "iPad12,2", "iPad 9th Gen (WiFi+Cellular)", + "iPad13,1", "iPad Air 4th gen (WiFi)", + "iPad13,2", "iPad Air 4th gen (WiFi+Cellular)", + "iPad13,4", "iPad Pro 11\" 3rd gen", + "iPad13,5", "iPad Pro 11\" 3rd gen", + "iPad13,6", "iPad Pro 11\" 3rd gen", + "iPad13,7", "iPad Pro 11\" 3rd gen", + "iPad13,8", "iPad Pro 12.9\" 5th gen", + "iPad13,9", "iPad Pro 12.9\" 5th gen", + "iPad13,10", "iPad Pro 12.9\" 5th gen", + "iPad13,11", "iPad Pro 12.9\" 5th gen", + "iPad13,16", "iPad Air 5th Gen (WiFi)", + "iPad13,17", "iPad Air 5th Gen (WiFi+Cellular)", + "iPad13,18", "iPad 10th Gen", + "iPad13,19", "iPad 10th Gen", + "iPad14,1", "iPad mini 6th Gen (WiFi)", + "iPad14,2", "iPad mini 6th Gen (WiFi+Cellular)", + "iPad14,3", "iPad Pro 11\" 4th Gen", + "iPad14,4", "iPad Pro 11\" 4th Gen", + "iPad14,5", "iPad Pro 12.9\" 6th Gen", + "iPad14,6", "iPad Pro 12.9\" 6th Gen", + "iPod1,1", "iPod Touch", + "iPod2,1", "iPod Touch 2nd gen", + "iPod3,1", "iPod Touch 3rd gen", + "iPod4,1", "iPod Touch 4th gen", + "iPod5,1", "iPod Touch 5th gen", + "iPod7,1", "iPod Touch 6th gen", + "iPod9,1", "iPod Touch 7th gen", + nullptr + }; + + auto ptr = DeviceTable; + while( *ptr ) + { + if( strcmp( ptr[0], id ) == 0 ) return ptr[1]; + ptr += 2; + } + return id; +} + +#endif + +} diff --git a/externals/tracy/public/client/TracyCallstack.cpp b/externals/tracy/public/client/TracyCallstack.cpp new file mode 100644 index 000000000..e8c20bf1b --- /dev/null +++ b/externals/tracy/public/client/TracyCallstack.cpp @@ -0,0 +1,1369 @@ +#include +#include +#include +#include +#include "TracyCallstack.hpp" +#include "TracyDebug.hpp" +#include "TracyFastVector.hpp" +#include "TracyStringHelpers.hpp" +#include "../common/TracyAlloc.hpp" +#include "../common/TracySystem.hpp" + + +#ifdef TRACY_HAS_CALLSTACK + +#if TRACY_HAS_CALLSTACK == 1 +# ifndef NOMINMAX +# define NOMINMAX +# endif +# include +# include +# include +# ifdef _MSC_VER +# pragma warning( push ) +# pragma warning( disable : 4091 ) +# endif +# include +# ifdef _MSC_VER +# pragma warning( pop ) +# endif +#elif TRACY_HAS_CALLSTACK == 2 || TRACY_HAS_CALLSTACK == 3 || TRACY_HAS_CALLSTACK == 4 || TRACY_HAS_CALLSTACK == 6 +# include "../libbacktrace/backtrace.hpp" +# include +# include +# include +# include +#elif TRACY_HAS_CALLSTACK == 5 +# include +# include +#endif + +#ifdef TRACY_DBGHELP_LOCK +# include "TracyProfiler.hpp" + +# define DBGHELP_INIT TracyConcat( TRACY_DBGHELP_LOCK, Init() ) +# define DBGHELP_LOCK TracyConcat( TRACY_DBGHELP_LOCK, Lock() ); +# define DBGHELP_UNLOCK TracyConcat( TRACY_DBGHELP_LOCK, Unlock() ); + +extern "C" +{ + void DBGHELP_INIT; + void DBGHELP_LOCK; + void DBGHELP_UNLOCK; +}; +#endif + +#if TRACY_HAS_CALLSTACK == 2 || TRACY_HAS_CALLSTACK == 3 || TRACY_HAS_CALLSTACK == 4 || TRACY_HAS_CALLSTACK == 5 || TRACY_HAS_CALLSTACK == 6 +// If you want to use your own demangling functionality (e.g. for another language), +// define TRACY_DEMANGLE and provide your own implementation of the __tracy_demangle +// function. The input parameter is a function name. The demangle function must +// identify whether this name is mangled, and fail if it is not. Failure is indicated +// by returning nullptr. If demangling succeeds, a pointer to the C string containing +// demangled function must be returned. The demangling function is responsible for +// managing memory for this string. It is expected that it will be internally reused. +// When a call to ___tracy_demangle is made, previous contents of the string memory +// do not need to be preserved. Function may return string of any length, but the +// profiler can choose to truncate it. +extern "C" const char* ___tracy_demangle( const char* mangled ); + +#ifndef TRACY_DEMANGLE +constexpr size_t ___tracy_demangle_buffer_len = 1024*1024; +char* ___tracy_demangle_buffer; + +void ___tracy_init_demangle_buffer() +{ + ___tracy_demangle_buffer = (char*)tracy::tracy_malloc( ___tracy_demangle_buffer_len ); +} + +void ___tracy_free_demangle_buffer() +{ + tracy::tracy_free( ___tracy_demangle_buffer ); +} + +extern "C" const char* ___tracy_demangle( const char* mangled ) +{ + if( !mangled || mangled[0] != '_' ) return nullptr; + if( strlen( mangled ) > ___tracy_demangle_buffer_len ) return nullptr; + int status; + size_t len = ___tracy_demangle_buffer_len; + return abi::__cxa_demangle( mangled, ___tracy_demangle_buffer, &len, &status ); +} +#endif +#endif + +#if TRACY_HAS_CALLSTACK == 3 +# define TRACY_USE_IMAGE_CACHE +# include +#endif + +namespace tracy +{ + +#ifdef TRACY_USE_IMAGE_CACHE +// when we have access to dl_iterate_phdr(), we can build a cache of address ranges to image paths +// so we can quickly determine which image an address falls into. +// We refresh this cache only when we hit an address that doesn't fall into any known range. +class ImageCache +{ +public: + struct ImageEntry + { + void* m_startAddress = nullptr; + void* m_endAddress = nullptr; + char* m_name = nullptr; + }; + + ImageCache() + : m_images( 512 ) + { + Refresh(); + } + + ~ImageCache() + { + Clear(); + } + + const ImageEntry* GetImageForAddress( void* address ) + { + const ImageEntry* entry = GetImageForAddressImpl( address ); + if( !entry ) + { + Refresh(); + return GetImageForAddressImpl( address ); + } + return entry; + } + +private: + tracy::FastVector m_images; + bool m_updated; + + static int Callback( struct dl_phdr_info* info, size_t size, void* data ) + { + ImageCache* cache = reinterpret_cast( data ); + + const auto startAddress = reinterpret_cast( info->dlpi_addr ); + if( cache->Contains( startAddress ) ) return 0; + + [[maybe_unused]] const uint32_t headerCount = info->dlpi_phnum; + assert( headerCount > 0); + const auto endAddress = reinterpret_cast( info->dlpi_addr + + info->dlpi_phdr[info->dlpi_phnum - 1].p_vaddr + info->dlpi_phdr[info->dlpi_phnum - 1].p_memsz); + + ImageEntry* image = cache->m_images.push_next(); + image->m_startAddress = startAddress; + image->m_endAddress = endAddress; + + const char* imageName = nullptr; + // the base executable name isn't provided when iterating with dl_iterate_phdr, get it with dladdr() + if( info->dlpi_name && info->dlpi_name[0] != '\0' ) + { + imageName = info->dlpi_name; + } + else + { + Dl_info dlInfo; + if( dladdr( (void *)info->dlpi_addr, &dlInfo ) ) + { + imageName = dlInfo.dli_fname; + } + } + + if( imageName ) + { + size_t sz = strlen( imageName ) + 1; + image->m_name = (char*)tracy_malloc( sz ); + memcpy( image->m_name, imageName, sz ); + } + else + { + image->m_name = nullptr; + } + + cache->m_updated = true; + + return 0; + } + + bool Contains( void* startAddress ) const + { + return std::any_of( m_images.begin(), m_images.end(), [startAddress]( const ImageEntry& entry ) { return startAddress == entry.m_startAddress; } ); + } + + void Refresh() + { + m_updated = false; + dl_iterate_phdr( Callback, this ); + + if( m_updated ) + { + std::sort( m_images.begin(), m_images.end(), + []( const ImageEntry& lhs, const ImageEntry& rhs ) { return lhs.m_startAddress > rhs.m_startAddress; } ); + } + } + + const ImageEntry* GetImageForAddressImpl( void* address ) const + { + auto it = std::lower_bound( m_images.begin(), m_images.end(), address, + []( const ImageEntry& lhs, const void* rhs ) { return lhs.m_startAddress > rhs; } ); + + if( it != m_images.end() && address < it->m_endAddress ) + { + return it; + } + return nullptr; + } + + void Clear() + { + for( ImageEntry& entry : m_images ) + { + tracy_free( entry.m_name ); + } + + m_images.clear(); + } +}; +#endif //#ifdef TRACY_USE_IMAGE_CACHE + +// when "TRACY_SYMBOL_OFFLINE_RESOLVE" is set, instead of fully resolving symbols at runtime, +// simply resolve the offset and image name (which will be enough the resolving to be done offline) +#ifdef TRACY_SYMBOL_OFFLINE_RESOLVE +constexpr bool s_shouldResolveSymbolsOffline = true; +#else +static bool s_shouldResolveSymbolsOffline = false; +bool ShouldResolveSymbolsOffline() +{ + const char* symbolOfflineResolve = GetEnvVar( "TRACY_SYMBOL_OFFLINE_RESOLVE" ); + return (symbolOfflineResolve && symbolOfflineResolve[0] == '1'); +} +#endif // #ifdef TRACY_SYMBOL_OFFLINE_RESOLVE + +#if TRACY_HAS_CALLSTACK == 1 + +enum { MaxCbTrace = 64 }; +enum { MaxNameSize = 8*1024 }; + +int cb_num; +CallstackEntry cb_data[MaxCbTrace]; + +extern "C" +{ + typedef DWORD (__stdcall *t_SymAddrIncludeInlineTrace)( HANDLE hProcess, DWORD64 Address ); + typedef BOOL (__stdcall *t_SymQueryInlineTrace)( HANDLE hProcess, DWORD64 StartAddress, DWORD StartContext, DWORD64 StartRetAddress, DWORD64 CurAddress, LPDWORD CurContext, LPDWORD CurFrameIndex ); + typedef BOOL (__stdcall *t_SymFromInlineContext)( HANDLE hProcess, DWORD64 Address, ULONG InlineContext, PDWORD64 Displacement, PSYMBOL_INFO Symbol ); + typedef BOOL (__stdcall *t_SymGetLineFromInlineContext)( HANDLE hProcess, DWORD64 qwAddr, ULONG InlineContext, DWORD64 qwModuleBaseAddress, PDWORD pdwDisplacement, PIMAGEHLP_LINE64 Line64 ); + + t_SymAddrIncludeInlineTrace _SymAddrIncludeInlineTrace = 0; + t_SymQueryInlineTrace _SymQueryInlineTrace = 0; + t_SymFromInlineContext _SymFromInlineContext = 0; + t_SymGetLineFromInlineContext _SymGetLineFromInlineContext = 0; + + TRACY_API ___tracy_t_RtlWalkFrameChain ___tracy_RtlWalkFrameChain = 0; +} + +struct ModuleCache +{ + uint64_t start; + uint64_t end; + char* name; +}; + +static FastVector* s_modCache; + + +struct KernelDriver +{ + uint64_t addr; + const char* mod; + const char* path; +}; + +KernelDriver* s_krnlCache = nullptr; +size_t s_krnlCacheCnt; + +void InitCallstackCritical() +{ + ___tracy_RtlWalkFrameChain = (___tracy_t_RtlWalkFrameChain)GetProcAddress( GetModuleHandleA( "ntdll.dll" ), "RtlWalkFrameChain" ); +} + +void DbgHelpInit() +{ + if( s_shouldResolveSymbolsOffline ) return; + + _SymAddrIncludeInlineTrace = (t_SymAddrIncludeInlineTrace)GetProcAddress(GetModuleHandleA("dbghelp.dll"), "SymAddrIncludeInlineTrace"); + _SymQueryInlineTrace = (t_SymQueryInlineTrace)GetProcAddress(GetModuleHandleA("dbghelp.dll"), "SymQueryInlineTrace"); + _SymFromInlineContext = (t_SymFromInlineContext)GetProcAddress(GetModuleHandleA("dbghelp.dll"), "SymFromInlineContext"); + _SymGetLineFromInlineContext = (t_SymGetLineFromInlineContext)GetProcAddress(GetModuleHandleA("dbghelp.dll"), "SymGetLineFromInlineContext"); + +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_INIT; + DBGHELP_LOCK; +#endif + + SymInitialize( GetCurrentProcess(), nullptr, true ); + SymSetOptions( SYMOPT_LOAD_LINES ); + +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_UNLOCK; +#endif +} + +DWORD64 DbgHelpLoadSymbolsForModule( const char* imageName, uint64_t baseOfDll, uint32_t bllSize ) +{ + if( s_shouldResolveSymbolsOffline ) return 0; + return SymLoadModuleEx( GetCurrentProcess(), nullptr, imageName, nullptr, baseOfDll, bllSize, nullptr, 0 ); +} + +ModuleCache* LoadSymbolsForModuleAndCache( const char* imageName, uint32_t imageNameLength, uint64_t baseOfDll, uint32_t dllSize ) +{ + DbgHelpLoadSymbolsForModule( imageName, baseOfDll, dllSize ); + + ModuleCache* cachedModule = s_modCache->push_next(); + cachedModule->start = baseOfDll; + cachedModule->end = baseOfDll + dllSize; + + // when doing offline symbol resolution, we must store the full path of the dll for the resolving to work + if( s_shouldResolveSymbolsOffline ) + { + cachedModule->name = (char*)tracy_malloc_fast(imageNameLength + 1); + memcpy(cachedModule->name, imageName, imageNameLength); + cachedModule->name[imageNameLength] = '\0'; + } + else + { + auto ptr = imageName + imageNameLength; + while (ptr > imageName && *ptr != '\\' && *ptr != '/') ptr--; + if (ptr > imageName) ptr++; + const auto namelen = imageName + imageNameLength - ptr; + cachedModule->name = (char*)tracy_malloc_fast(namelen + 3); + cachedModule->name[0] = '['; + memcpy(cachedModule->name + 1, ptr, namelen); + cachedModule->name[namelen + 1] = ']'; + cachedModule->name[namelen + 2] = '\0'; + } + + return cachedModule; +} + +void InitCallstack() +{ +#ifndef TRACY_SYMBOL_OFFLINE_RESOLVE + s_shouldResolveSymbolsOffline = ShouldResolveSymbolsOffline(); +#endif //#ifndef TRACY_SYMBOL_OFFLINE_RESOLVE + if( s_shouldResolveSymbolsOffline ) + { + TracyDebug("TRACY: enabling offline symbol resolving!\n"); + } + + DbgHelpInit(); + +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_LOCK; +#endif + + // use TRACY_NO_DBGHELP_INIT_LOAD=1 to disable preloading of driver + // and process module symbol loading at startup time - they will be loaded on demand later + // Sometimes this process can take a very long time and prevent resolving callstack frames + // symbols during that time. + const char* noInitLoadEnv = GetEnvVar( "TRACY_NO_DBGHELP_INIT_LOAD" ); + const bool initTimeModuleLoad = !( noInitLoadEnv && noInitLoadEnv[0] == '1' ); + if ( !initTimeModuleLoad ) + { + TracyDebug("TRACY: skipping init time dbghelper module load\n"); + } + + DWORD needed; + LPVOID dev[4096]; + if( initTimeModuleLoad && EnumDeviceDrivers( dev, sizeof(dev), &needed ) != 0 ) + { + char windir[MAX_PATH]; + if( !GetWindowsDirectoryA( windir, sizeof( windir ) ) ) memcpy( windir, "c:\\windows", 11 ); + const auto windirlen = strlen( windir ); + + const auto sz = needed / sizeof( LPVOID ); + s_krnlCache = (KernelDriver*)tracy_malloc( sizeof(KernelDriver) * sz ); + int cnt = 0; + for( size_t i=0; i", 2 ); + s_krnlCache[cnt] = KernelDriver { (uint64_t)dev[i], buf }; + + const auto len = GetDeviceDriverFileNameA( dev[i], fn, sizeof( fn ) ); + if( len != 0 ) + { + char full[MAX_PATH]; + char* path = fn; + + if( memcmp( fn, "\\SystemRoot\\", 12 ) == 0 ) + { + memcpy( full, windir, windirlen ); + strcpy( full + windirlen, fn + 11 ); + path = full; + } + + DbgHelpLoadSymbolsForModule( path, (DWORD64)dev[i], 0 ); + + const auto psz = strlen( path ); + auto pptr = (char*)tracy_malloc_fast( psz+1 ); + memcpy( pptr, path, psz ); + pptr[psz] = '\0'; + s_krnlCache[cnt].path = pptr; + } + + cnt++; + } + } + s_krnlCacheCnt = cnt; + std::sort( s_krnlCache, s_krnlCache + s_krnlCacheCnt, []( const KernelDriver& lhs, const KernelDriver& rhs ) { return lhs.addr > rhs.addr; } ); + } + + s_modCache = (FastVector*)tracy_malloc( sizeof( FastVector ) ); + new(s_modCache) FastVector( 512 ); + + HANDLE proc = GetCurrentProcess(); + HMODULE mod[1024]; + if( initTimeModuleLoad && EnumProcessModules( proc, mod, sizeof( mod ), &needed ) != 0 ) + { + const auto sz = needed / sizeof( HMODULE ); + for( size_t i=0; i 0 ) + { + // This may be a new module loaded since our call to SymInitialize. + // Just in case, force DbgHelp to load its pdb ! + LoadSymbolsForModuleAndCache( name, nameLength, (DWORD64)info.lpBaseOfDll, info.SizeOfImage ); + } + } + } + } + +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_UNLOCK; +#endif +} + +void EndCallstack() +{ +} + +const char* DecodeCallstackPtrFast( uint64_t ptr ) +{ + if( s_shouldResolveSymbolsOffline ) return "[unresolved]"; + + static char ret[MaxNameSize]; + const auto proc = GetCurrentProcess(); + + char buf[sizeof( SYMBOL_INFO ) + MaxNameSize]; + auto si = (SYMBOL_INFO*)buf; + si->SizeOfStruct = sizeof( SYMBOL_INFO ); + si->MaxNameLen = MaxNameSize; + +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_LOCK; +#endif + if( SymFromAddr( proc, ptr, nullptr, si ) == 0 ) + { + *ret = '\0'; + } + else + { + memcpy( ret, si->Name, si->NameLen ); + ret[si->NameLen] = '\0'; + } +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_UNLOCK; +#endif + return ret; +} + +const char* GetKernelModulePath( uint64_t addr ) +{ + assert( addr >> 63 != 0 ); + if( !s_krnlCache ) return nullptr; + auto it = std::lower_bound( s_krnlCache, s_krnlCache + s_krnlCacheCnt, addr, []( const KernelDriver& lhs, const uint64_t& rhs ) { return lhs.addr > rhs; } ); + if( it == s_krnlCache + s_krnlCacheCnt ) return nullptr; + return it->path; +} + +struct ModuleNameAndBaseAddress +{ + const char* name; + uint64_t baseAddr; +}; + +ModuleNameAndBaseAddress GetModuleNameAndPrepareSymbols( uint64_t addr ) +{ + if( ( addr >> 63 ) != 0 ) + { + if( s_krnlCache ) + { + auto it = std::lower_bound( s_krnlCache, s_krnlCache + s_krnlCacheCnt, addr, []( const KernelDriver& lhs, const uint64_t& rhs ) { return lhs.addr > rhs; } ); + if( it != s_krnlCache + s_krnlCacheCnt ) + { + return ModuleNameAndBaseAddress{ it->mod, it->addr }; + } + } + return ModuleNameAndBaseAddress{ "", addr }; + } + + for( auto& v : *s_modCache ) + { + if( addr >= v.start && addr < v.end ) + { + return ModuleNameAndBaseAddress{ v.name, v.start }; + } + } + + HMODULE mod[1024]; + DWORD needed; + HANDLE proc = GetCurrentProcess(); + + InitRpmalloc(); + if( EnumProcessModules( proc, mod, sizeof( mod ), &needed ) != 0 ) + { + const auto sz = needed / sizeof( HMODULE ); + for( size_t i=0; i= base && addr < base + info.SizeOfImage ) + { + char name[1024]; + const auto nameLength = GetModuleFileNameA( mod[i], name, 1021 ); + if( nameLength > 0 ) + { + // since this is the first time we encounter this module, load its symbols (needed for modules loaded after SymInitialize) + ModuleCache* cachedModule = LoadSymbolsForModuleAndCache( name, nameLength, (DWORD64)info.lpBaseOfDll, info.SizeOfImage ); + return ModuleNameAndBaseAddress{ cachedModule->name, cachedModule->start }; + } + } + } + } + } + + return ModuleNameAndBaseAddress{ "[unknown]", 0x0 }; +} + +CallstackSymbolData DecodeSymbolAddress( uint64_t ptr ) +{ + CallstackSymbolData sym; + + if( s_shouldResolveSymbolsOffline ) + { + sym.file = "[unknown]"; + sym.line = 0; + sym.needFree = false; + return sym; + } + + IMAGEHLP_LINE64 line; + DWORD displacement = 0; + line.SizeOfStruct = sizeof(IMAGEHLP_LINE64); +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_LOCK; +#endif + const auto res = SymGetLineFromAddr64( GetCurrentProcess(), ptr, &displacement, &line ); + if( res == 0 || line.LineNumber >= 0xF00000 ) + { + sym.file = "[unknown]"; + sym.line = 0; + sym.needFree = false; + } + else + { + sym.file = CopyString( line.FileName ); + sym.line = line.LineNumber; + sym.needFree = true; + } +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_UNLOCK; +#endif + return sym; +} + +CallstackEntryData DecodeCallstackPtr( uint64_t ptr ) +{ +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_LOCK; +#endif + + InitRpmalloc(); + + const ModuleNameAndBaseAddress moduleNameAndAddress = GetModuleNameAndPrepareSymbols( ptr ); + + if( s_shouldResolveSymbolsOffline ) + { +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_UNLOCK; +#endif + + cb_data[0].symAddr = ptr - moduleNameAndAddress.baseAddr; + cb_data[0].symLen = 0; + + cb_data[0].name = CopyStringFast("[unresolved]"); + cb_data[0].file = CopyStringFast("[unknown]"); + cb_data[0].line = 0; + + return { cb_data, 1, moduleNameAndAddress.name }; + } + + int write; + const auto proc = GetCurrentProcess(); + +#if !defined TRACY_NO_CALLSTACK_INLINES + BOOL doInline = FALSE; + DWORD ctx = 0; + DWORD inlineNum = 0; + if( _SymAddrIncludeInlineTrace ) + { + inlineNum = _SymAddrIncludeInlineTrace( proc, ptr ); + if( inlineNum > MaxCbTrace - 1 ) inlineNum = MaxCbTrace - 1; + DWORD idx; + if( inlineNum != 0 ) doInline = _SymQueryInlineTrace( proc, ptr, 0, ptr, ptr, &ctx, &idx ); + } + if( doInline ) + { + write = inlineNum; + cb_num = 1 + inlineNum; + } + else +#endif + { + write = 0; + cb_num = 1; + } + + char buf[sizeof( SYMBOL_INFO ) + MaxNameSize]; + auto si = (SYMBOL_INFO*)buf; + si->SizeOfStruct = sizeof( SYMBOL_INFO ); + si->MaxNameLen = MaxNameSize; + + const auto symValid = SymFromAddr( proc, ptr, nullptr, si ) != 0; + + IMAGEHLP_LINE64 line; + DWORD displacement = 0; + line.SizeOfStruct = sizeof(IMAGEHLP_LINE64); + + { + const char* filename; + const auto res = SymGetLineFromAddr64( proc, ptr, &displacement, &line ); + if( res == 0 || line.LineNumber >= 0xF00000 ) + { + filename = "[unknown]"; + cb_data[write].line = 0; + } + else + { + filename = line.FileName; + cb_data[write].line = line.LineNumber; + } + + cb_data[write].name = symValid ? CopyStringFast( si->Name, si->NameLen ) : CopyStringFast( moduleNameAndAddress.name ); + cb_data[write].file = CopyStringFast( filename ); + if( symValid ) + { + cb_data[write].symLen = si->Size; + cb_data[write].symAddr = si->Address; + } + else + { + cb_data[write].symLen = 0; + cb_data[write].symAddr = 0; + } + } + +#if !defined TRACY_NO_CALLSTACK_INLINES + if( doInline ) + { + for( DWORD i=0; iName, si->NameLen ) : CopyStringFast( moduleNameAndAddress.name ); + cb.file = CopyStringFast( filename ); + if( symInlineValid ) + { + cb.symLen = si->Size; + cb.symAddr = si->Address; + } + else + { + cb.symLen = 0; + cb.symAddr = 0; + } + + ctx++; + } + } +#endif +#ifdef TRACY_DBGHELP_LOCK + DBGHELP_UNLOCK; +#endif + + return { cb_data, uint8_t( cb_num ), moduleNameAndAddress.name }; +} + +#elif TRACY_HAS_CALLSTACK == 2 || TRACY_HAS_CALLSTACK == 3 || TRACY_HAS_CALLSTACK == 4 || TRACY_HAS_CALLSTACK == 6 + +enum { MaxCbTrace = 64 }; + +struct backtrace_state* cb_bts = nullptr; + +int cb_num; +CallstackEntry cb_data[MaxCbTrace]; +int cb_fixup; +#ifdef TRACY_USE_IMAGE_CACHE +static ImageCache* s_imageCache = nullptr; +#endif //#ifdef TRACY_USE_IMAGE_CACHE + +#ifdef TRACY_DEBUGINFOD +debuginfod_client* s_debuginfod; + +struct DebugInfo +{ + uint8_t* buildid; + size_t buildid_size; + char* filename; + int fd; +}; + +FastVector s_di_known( 16 ); +#endif + +#ifdef __linux +struct KernelSymbol +{ + uint64_t addr; + const char* name; + const char* mod; +}; + +KernelSymbol* s_kernelSym = nullptr; +size_t s_kernelSymCnt; + +static void InitKernelSymbols() +{ + FILE* f = fopen( "/proc/kallsyms", "rb" ); + if( !f ) return; + tracy::FastVector tmpSym( 1024 ); + size_t linelen = 16 * 1024; // linelen must be big enough to prevent reallocs in getline() + auto linebuf = (char*)tracy_malloc( linelen ); + ssize_t sz; + while( ( sz = getline( &linebuf, &linelen, f ) ) != -1 ) + { + auto ptr = linebuf; + uint64_t addr = 0; + while( *ptr != ' ' ) + { + auto v = *ptr; + if( v >= '0' && v <= '9' ) + { + v -= '0'; + } + else if( v >= 'a' && v <= 'f' ) + { + v -= 'a'; + v += 10; + } + else if( v >= 'A' && v <= 'F' ) + { + v -= 'A'; + v += 10; + } + else + { + assert( false ); + } + assert( ( v & ~0xF ) == 0 ); + addr <<= 4; + addr |= v; + ptr++; + } + if( addr == 0 ) continue; + ptr++; + if( *ptr != 'T' && *ptr != 't' ) continue; + ptr += 2; + const auto namestart = ptr; + while( *ptr != '\t' && *ptr != '\n' ) ptr++; + const auto nameend = ptr; + const char* modstart = nullptr; + const char* modend; + if( *ptr == '\t' ) + { + ptr += 2; + modstart = ptr; + while( *ptr != ']' ) ptr++; + modend = ptr; + } + + auto strname = (char*)tracy_malloc_fast( nameend - namestart + 1 ); + memcpy( strname, namestart, nameend - namestart ); + strname[nameend-namestart] = '\0'; + + char* strmod = nullptr; + if( modstart ) + { + strmod = (char*)tracy_malloc_fast( modend - modstart + 1 ); + memcpy( strmod, modstart, modend - modstart ); + strmod[modend-modstart] = '\0'; + } + + auto sym = tmpSym.push_next(); + sym->addr = addr; + sym->name = strname; + sym->mod = strmod; + } + tracy_free_fast( linebuf ); + fclose( f ); + if( tmpSym.empty() ) return; + + std::sort( tmpSym.begin(), tmpSym.end(), []( const KernelSymbol& lhs, const KernelSymbol& rhs ) { return lhs.addr > rhs.addr; } ); + s_kernelSymCnt = tmpSym.size(); + s_kernelSym = (KernelSymbol*)tracy_malloc_fast( sizeof( KernelSymbol ) * s_kernelSymCnt ); + memcpy( s_kernelSym, tmpSym.data(), sizeof( KernelSymbol ) * s_kernelSymCnt ); + TracyDebug( "Loaded %zu kernel symbols\n", s_kernelSymCnt ); +} +#endif + +char* NormalizePath( const char* path ) +{ + if( path[0] != '/' ) return nullptr; + + const char* ptr = path; + const char* end = path; + while( *end ) end++; + + char* res = (char*)tracy_malloc( end - ptr + 1 ); + size_t rsz = 0; + + while( ptr < end ) + { + const char* next = ptr; + while( next < end && *next != '/' ) next++; + size_t lsz = next - ptr; + switch( lsz ) + { + case 2: + if( memcmp( ptr, "..", 2 ) == 0 ) + { + const char* back = res + rsz - 1; + while( back > res && *back != '/' ) back--; + rsz = back - res; + ptr = next + 1; + continue; + } + break; + case 1: + if( *ptr == '.' ) + { + ptr = next + 1; + continue; + } + break; + case 0: + ptr = next + 1; + continue; + } + if( rsz != 1 ) res[rsz++] = '/'; + memcpy( res+rsz, ptr, lsz ); + rsz += lsz; + ptr = next + 1; + } + + if( rsz == 0 ) + { + memcpy( res, "/", 2 ); + } + else + { + res[rsz] = '\0'; + } + return res; +} + +void InitCallstackCritical() +{ +} + +void InitCallstack() +{ + InitRpmalloc(); + +#ifdef TRACY_USE_IMAGE_CACHE + s_imageCache = (ImageCache*)tracy_malloc( sizeof( ImageCache ) ); + new(s_imageCache) ImageCache(); +#endif //#ifdef TRACY_USE_IMAGE_CACHE + +#ifndef TRACY_SYMBOL_OFFLINE_RESOLVE + s_shouldResolveSymbolsOffline = ShouldResolveSymbolsOffline(); +#endif //#ifndef TRACY_SYMBOL_OFFLINE_RESOLVE + if( s_shouldResolveSymbolsOffline ) + { + cb_bts = nullptr; // disable use of libbacktrace calls + TracyDebug("TRACY: enabling offline symbol resolving!\n"); + } + else + { + cb_bts = backtrace_create_state( nullptr, 0, nullptr, nullptr ); + } + +#ifndef TRACY_DEMANGLE + ___tracy_init_demangle_buffer(); +#endif + +#ifdef __linux + InitKernelSymbols(); +#endif +#ifdef TRACY_DEBUGINFOD + s_debuginfod = debuginfod_begin(); +#endif +} + +#ifdef TRACY_DEBUGINFOD +void ClearDebugInfoVector( FastVector& vec ) +{ + for( auto& v : vec ) + { + tracy_free( v.buildid ); + tracy_free( v.filename ); + if( v.fd >= 0 ) close( v.fd ); + } + vec.clear(); +} + +DebugInfo* FindDebugInfo( FastVector& vec, const uint8_t* buildid_data, size_t buildid_size ) +{ + for( auto& v : vec ) + { + if( v.buildid_size == buildid_size && memcmp( v.buildid, buildid_data, buildid_size ) == 0 ) + { + return &v; + } + } + return nullptr; +} + +int GetDebugInfoDescriptor( const char* buildid_data, size_t buildid_size, const char* filename ) +{ + auto buildid = (uint8_t*)buildid_data; + auto it = FindDebugInfo( s_di_known, buildid, buildid_size ); + if( it ) return it->fd >= 0 ? dup( it->fd ) : -1; + + int fd = debuginfod_find_debuginfo( s_debuginfod, buildid, buildid_size, nullptr ); + it = s_di_known.push_next(); + it->buildid_size = buildid_size; + it->buildid = (uint8_t*)tracy_malloc( buildid_size ); + memcpy( it->buildid, buildid, buildid_size ); + const auto fnsz = strlen( filename ) + 1; + it->filename = (char*)tracy_malloc( fnsz ); + memcpy( it->filename, filename, fnsz ); + it->fd = fd >= 0 ? fd : -1; + TracyDebug( "DebugInfo descriptor query: %i, fn: %s\n", fd, filename ); + return it->fd; +} + +const uint8_t* GetBuildIdForImage( const char* image, size_t& size ) +{ + assert( image ); + for( auto& v : s_di_known ) + { + if( strcmp( image, v.filename ) == 0 ) + { + size = v.buildid_size; + return v.buildid; + } + } + return nullptr; +} + +debuginfod_client* GetDebuginfodClient() +{ + return s_debuginfod; +} +#endif + +void EndCallstack() +{ +#ifdef TRACY_USE_IMAGE_CACHE + if( s_imageCache ) + { + s_imageCache->~ImageCache(); + tracy_free( s_imageCache ); + } +#endif //#ifdef TRACY_USE_IMAGE_CACHE +#ifndef TRACY_DEMANGLE + ___tracy_free_demangle_buffer(); +#endif +#ifdef TRACY_DEBUGINFOD + ClearDebugInfoVector( s_di_known ); + debuginfod_end( s_debuginfod ); +#endif +} + +const char* DecodeCallstackPtrFast( uint64_t ptr ) +{ + static char ret[1024]; + auto vptr = (void*)ptr; + const char* symname = nullptr; + Dl_info dlinfo; + if( dladdr( vptr, &dlinfo ) && dlinfo.dli_sname ) + { + symname = dlinfo.dli_sname; + } + if( symname ) + { + strcpy( ret, symname ); + } + else + { + *ret = '\0'; + } + return ret; +} + +static int SymbolAddressDataCb( void* data, uintptr_t pc, uintptr_t lowaddr, const char* fn, int lineno, const char* function ) +{ + auto& sym = *(CallstackSymbolData*)data; + if( !fn ) + { + sym.file = "[unknown]"; + sym.line = 0; + sym.needFree = false; + } + else + { + sym.file = NormalizePath( fn ); + if( !sym.file ) sym.file = CopyString( fn ); + sym.line = lineno; + sym.needFree = true; + } + + return 1; +} + +static void SymbolAddressErrorCb( void* data, const char* /*msg*/, int /*errnum*/ ) +{ + auto& sym = *(CallstackSymbolData*)data; + sym.file = "[unknown]"; + sym.line = 0; + sym.needFree = false; +} + +CallstackSymbolData DecodeSymbolAddress( uint64_t ptr ) +{ + CallstackSymbolData sym; + if( cb_bts ) + { + backtrace_pcinfo( cb_bts, ptr, SymbolAddressDataCb, SymbolAddressErrorCb, &sym ); + } + else + { + SymbolAddressErrorCb(&sym, nullptr, 0); + } + + return sym; +} + +static int CallstackDataCb( void* /*data*/, uintptr_t pc, uintptr_t lowaddr, const char* fn, int lineno, const char* function ) +{ + cb_data[cb_num].symLen = 0; + cb_data[cb_num].symAddr = (uint64_t)lowaddr; + + if( !fn && !function ) + { + const char* symname = nullptr; + auto vptr = (void*)pc; + ptrdiff_t symoff = 0; + + Dl_info dlinfo; + if( dladdr( vptr, &dlinfo ) ) + { + symname = dlinfo.dli_sname; + symoff = (char*)pc - (char*)dlinfo.dli_saddr; + const char* demangled = ___tracy_demangle( symname ); + if( demangled ) symname = demangled; + } + + if( !symname ) symname = "[unknown]"; + + if( symoff == 0 ) + { + const auto len = std::min( strlen( symname ), std::numeric_limits::max() ); + cb_data[cb_num].name = CopyStringFast( symname, len ); + } + else + { + char buf[32]; + const auto offlen = sprintf( buf, " + %td", symoff ); + const auto namelen = std::min( strlen( symname ), std::numeric_limits::max() - offlen ); + auto name = (char*)tracy_malloc_fast( namelen + offlen + 1 ); + memcpy( name, symname, namelen ); + memcpy( name + namelen, buf, offlen ); + name[namelen + offlen] = '\0'; + cb_data[cb_num].name = name; + } + + cb_data[cb_num].file = CopyStringFast( "[unknown]" ); + cb_data[cb_num].line = 0; + } + else + { + if( !fn ) fn = "[unknown]"; + if( !function ) + { + function = "[unknown]"; + } + else + { + const char* demangled = ___tracy_demangle( function ); + if( demangled ) function = demangled; + } + + const auto len = std::min( strlen( function ), std::numeric_limits::max() ); + cb_data[cb_num].name = CopyStringFast( function, len ); + cb_data[cb_num].file = NormalizePath( fn ); + if( !cb_data[cb_num].file ) cb_data[cb_num].file = CopyStringFast( fn ); + cb_data[cb_num].line = lineno; + } + + if( ++cb_num >= MaxCbTrace ) + { + return 1; + } + else + { + return 0; + } +} + +static void CallstackErrorCb( void* /*data*/, const char* /*msg*/, int /*errnum*/ ) +{ + for( int i=0; i> 63 == 0 ) + { + const char* imageName = nullptr; + uint64_t imageBaseAddress = 0x0; + +#ifdef TRACY_USE_IMAGE_CACHE + const auto* image = s_imageCache->GetImageForAddress((void*)ptr); + if( image ) + { + imageName = image->m_name; + imageBaseAddress = uint64_t(image->m_startAddress); + } +#else + Dl_info dlinfo; + if( dladdr( (void*)ptr, &dlinfo ) ) + { + imageName = dlinfo.dli_fname; + imageBaseAddress = uint64_t( dlinfo.dli_fbase ); + } +#endif + + if( s_shouldResolveSymbolsOffline ) + { + cb_num = 1; + GetSymbolForOfflineResolve( (void*)ptr, imageBaseAddress, cb_data[0] ); + } + else + { + cb_num = 0; + backtrace_pcinfo( cb_bts, ptr, CallstackDataCb, CallstackErrorCb, nullptr ); + assert( cb_num > 0 ); + + backtrace_syminfo( cb_bts, ptr, SymInfoCallback, SymInfoError, nullptr ); + } + + return { cb_data, uint8_t( cb_num ), imageName ? imageName : "[unknown]" }; + } +#ifdef __linux + else if( s_kernelSym ) + { + auto it = std::lower_bound( s_kernelSym, s_kernelSym + s_kernelSymCnt, ptr, []( const KernelSymbol& lhs, const uint64_t& rhs ) { return lhs.addr > rhs; } ); + if( it != s_kernelSym + s_kernelSymCnt ) + { + cb_data[0].name = CopyStringFast( it->name ); + cb_data[0].file = CopyStringFast( "" ); + cb_data[0].line = 0; + cb_data[0].symLen = 0; + cb_data[0].symAddr = it->addr; + return { cb_data, 1, it->mod ? it->mod : "" }; + } + } +#endif + + cb_data[0].name = CopyStringFast( "[unknown]" ); + cb_data[0].file = CopyStringFast( "" ); + cb_data[0].line = 0; + cb_data[0].symLen = 0; + cb_data[0].symAddr = 0; + return { cb_data, 1, "" }; +} + +#elif TRACY_HAS_CALLSTACK == 5 + +void InitCallstackCritical() +{ +} + +void InitCallstack() +{ + ___tracy_init_demangle_buffer(); +} + +void EndCallstack() +{ + ___tracy_free_demangle_buffer(); +} + +const char* DecodeCallstackPtrFast( uint64_t ptr ) +{ + static char ret[1024]; + auto vptr = (void*)ptr; + const char* symname = nullptr; + Dl_info dlinfo; + if( dladdr( vptr, &dlinfo ) && dlinfo.dli_sname ) + { + symname = dlinfo.dli_sname; + } + if( symname ) + { + strcpy( ret, symname ); + } + else + { + *ret = '\0'; + } + return ret; +} + +CallstackSymbolData DecodeSymbolAddress( uint64_t ptr ) +{ + const char* symloc = nullptr; + Dl_info dlinfo; + if( dladdr( (void*)ptr, &dlinfo ) ) symloc = dlinfo.dli_fname; + if( !symloc ) symloc = "[unknown]"; + return CallstackSymbolData { symloc, 0, false, 0 }; +} + +CallstackEntryData DecodeCallstackPtr( uint64_t ptr ) +{ + static CallstackEntry cb; + cb.line = 0; + + const char* symname = nullptr; + const char* symloc = nullptr; + auto vptr = (void*)ptr; + ptrdiff_t symoff = 0; + void* symaddr = nullptr; + + Dl_info dlinfo; + if( dladdr( vptr, &dlinfo ) ) + { + symloc = dlinfo.dli_fname; + symname = dlinfo.dli_sname; + symoff = (char*)ptr - (char*)dlinfo.dli_saddr; + symaddr = dlinfo.dli_saddr; + const char* demangled = ___tracy_demangle( symname ); + if( demangled ) symname = demangled; + } + + if( !symname ) symname = "[unknown]"; + if( !symloc ) symloc = "[unknown]"; + + if( symoff == 0 ) + { + const auto len = std::min( strlen( symname ), std::numeric_limits::max() ); + cb.name = CopyString( symname, len ); + } + else + { + char buf[32]; + const auto offlen = sprintf( buf, " + %td", symoff ); + const auto namelen = std::min( strlen( symname ), std::numeric_limits::max() - offlen ); + auto name = (char*)tracy_malloc( namelen + offlen + 1 ); + memcpy( name, symname, namelen ); + memcpy( name + namelen, buf, offlen ); + name[namelen + offlen] = '\0'; + cb.name = name; + } + + cb.file = CopyString( "[unknown]" ); + cb.symLen = 0; + cb.symAddr = (uint64_t)symaddr; + + return { &cb, 1, symloc }; +} + +#endif + +} + +#endif diff --git a/externals/tracy/public/client/TracyCallstack.h b/externals/tracy/public/client/TracyCallstack.h new file mode 100644 index 000000000..2c7ecad9f --- /dev/null +++ b/externals/tracy/public/client/TracyCallstack.h @@ -0,0 +1,35 @@ +#ifndef __TRACYCALLSTACK_H__ +#define __TRACYCALLSTACK_H__ + +#ifndef TRACY_NO_CALLSTACK + +# if !defined _WIN32 +# include +# endif + +# if defined _WIN32 +# include "../common/TracyUwp.hpp" +# ifndef TRACY_UWP +# define TRACY_HAS_CALLSTACK 1 +# endif +# elif defined __ANDROID__ +# if !defined __arm__ || __ANDROID_API__ >= 21 +# define TRACY_HAS_CALLSTACK 2 +# else +# define TRACY_HAS_CALLSTACK 5 +# endif +# elif defined __linux +# if defined _GNU_SOURCE && defined __GLIBC__ +# define TRACY_HAS_CALLSTACK 3 +# else +# define TRACY_HAS_CALLSTACK 2 +# endif +# elif defined __APPLE__ +# define TRACY_HAS_CALLSTACK 4 +# elif defined BSD +# define TRACY_HAS_CALLSTACK 6 +# endif + +#endif + +#endif diff --git a/externals/tracy/public/client/TracyCallstack.hpp b/externals/tracy/public/client/TracyCallstack.hpp new file mode 100644 index 000000000..e1de3821a --- /dev/null +++ b/externals/tracy/public/client/TracyCallstack.hpp @@ -0,0 +1,153 @@ +#ifndef __TRACYCALLSTACK_HPP__ +#define __TRACYCALLSTACK_HPP__ + +#include "../common/TracyApi.h" +#include "../common/TracyForceInline.hpp" +#include "TracyCallstack.h" + +#if TRACY_HAS_CALLSTACK == 2 || TRACY_HAS_CALLSTACK == 5 +# include +#elif TRACY_HAS_CALLSTACK >= 3 +# ifdef TRACY_LIBUNWIND_BACKTRACE + // libunwind is, in general, significantly faster than execinfo based backtraces +# define UNW_LOCAL_ONLY +# include +# else +# include +# endif +#endif + +#ifndef TRACY_HAS_CALLSTACK + +namespace tracy +{ +static tracy_force_inline void* Callstack( int depth ) { return nullptr; } +} + +#else + +#ifdef TRACY_DEBUGINFOD +# include +#endif + +#include +#include + +#include "../common/TracyAlloc.hpp" + +namespace tracy +{ + +struct CallstackSymbolData +{ + const char* file; + uint32_t line; + bool needFree; + uint64_t symAddr; +}; + +struct CallstackEntry +{ + const char* name; + const char* file; + uint32_t line; + uint32_t symLen; + uint64_t symAddr; +}; + +struct CallstackEntryData +{ + const CallstackEntry* data; + uint8_t size; + const char* imageName; +}; + +CallstackSymbolData DecodeSymbolAddress( uint64_t ptr ); +const char* DecodeCallstackPtrFast( uint64_t ptr ); +CallstackEntryData DecodeCallstackPtr( uint64_t ptr ); +void InitCallstack(); +void InitCallstackCritical(); +void EndCallstack(); +const char* GetKernelModulePath( uint64_t addr ); + +#ifdef TRACY_DEBUGINFOD +const uint8_t* GetBuildIdForImage( const char* image, size_t& size ); +debuginfod_client* GetDebuginfodClient(); +#endif + +#if TRACY_HAS_CALLSTACK == 1 + +extern "C" +{ + typedef unsigned long (__stdcall *___tracy_t_RtlWalkFrameChain)( void**, unsigned long, unsigned long ); + TRACY_API extern ___tracy_t_RtlWalkFrameChain ___tracy_RtlWalkFrameChain; +} + +static tracy_force_inline void* Callstack( int depth ) +{ + assert( depth >= 1 && depth < 63 ); + auto trace = (uintptr_t*)tracy_malloc( ( 1 + depth ) * sizeof( uintptr_t ) ); + const auto num = ___tracy_RtlWalkFrameChain( (void**)( trace + 1 ), depth, 0 ); + *trace = num; + return trace; +} + +#elif TRACY_HAS_CALLSTACK == 2 || TRACY_HAS_CALLSTACK == 5 + +struct BacktraceState +{ + void** current; + void** end; +}; + +static _Unwind_Reason_Code tracy_unwind_callback( struct _Unwind_Context* ctx, void* arg ) +{ + auto state = (BacktraceState*)arg; + uintptr_t pc = _Unwind_GetIP( ctx ); + if( pc ) + { + if( state->current == state->end ) return _URC_END_OF_STACK; + *state->current++ = (void*)pc; + } + return _URC_NO_REASON; +} + +static tracy_force_inline void* Callstack( int depth ) +{ + assert( depth >= 1 && depth < 63 ); + + auto trace = (uintptr_t*)tracy_malloc( ( 1 + depth ) * sizeof( uintptr_t ) ); + BacktraceState state = { (void**)(trace+1), (void**)(trace+1+depth) }; + _Unwind_Backtrace( tracy_unwind_callback, &state ); + + *trace = (uintptr_t*)state.current - trace + 1; + + return trace; +} + +#elif TRACY_HAS_CALLSTACK == 3 || TRACY_HAS_CALLSTACK == 4 || TRACY_HAS_CALLSTACK == 6 + +static tracy_force_inline void* Callstack( int depth ) +{ + assert( depth >= 1 ); + + auto trace = (uintptr_t*)tracy_malloc( ( 1 + (size_t)depth ) * sizeof( uintptr_t ) ); + +#ifdef TRACY_LIBUNWIND_BACKTRACE + size_t num = unw_backtrace( (void**)(trace+1), depth ); +#else + const auto num = (size_t)backtrace( (void**)(trace+1), depth ); +#endif + + *trace = num; + + return trace; +} + +#endif + +} + +#endif + +#endif diff --git a/externals/tracy/public/client/TracyCpuid.hpp b/externals/tracy/public/client/TracyCpuid.hpp new file mode 100644 index 000000000..9820be00b --- /dev/null +++ b/externals/tracy/public/client/TracyCpuid.hpp @@ -0,0 +1,12 @@ +#ifndef __TRACYCPUID_HPP__ +#define __TRACYCPUID_HPP__ + +// Prior to GCC 11 the cpuid.h header did not have any include guards and thus +// including it more than once would cause a compiler error due to symbol +// redefinitions. In order to support older GCC versions, we have to wrap this +// include between custom include guards to prevent this issue. +// See also https://github.com/wolfpld/tracy/issues/452 + +#include + +#endif diff --git a/externals/tracy/public/client/TracyDebug.hpp b/externals/tracy/public/client/TracyDebug.hpp new file mode 100644 index 000000000..8723356f4 --- /dev/null +++ b/externals/tracy/public/client/TracyDebug.hpp @@ -0,0 +1,11 @@ +#ifndef __TRACYPRINT_HPP__ +#define __TRACYPRINT_HPP__ + +#ifdef TRACY_VERBOSE +# include +# define TracyDebug(...) fprintf( stderr, __VA_ARGS__ ); +#else +# define TracyDebug(...) +#endif + +#endif diff --git a/externals/tracy/public/client/TracyDxt1.cpp b/externals/tracy/public/client/TracyDxt1.cpp new file mode 100644 index 000000000..930d09820 --- /dev/null +++ b/externals/tracy/public/client/TracyDxt1.cpp @@ -0,0 +1,644 @@ +#include "TracyDxt1.hpp" +#include "../common/TracyForceInline.hpp" + +#include +#include +#include + +#ifdef __ARM_NEON +# include +#endif + +#if defined __AVX__ && !defined __SSE4_1__ +# define __SSE4_1__ +#endif + +#if defined __SSE4_1__ || defined __AVX2__ +# ifdef _MSC_VER +# include +# else +# include +# ifndef _mm256_cvtsi256_si32 +# define _mm256_cvtsi256_si32( v ) ( _mm_cvtsi128_si32( _mm256_castsi256_si128( v ) ) ) +# endif +# endif +#endif + +namespace tracy +{ + +static inline uint16_t to565( uint8_t r, uint8_t g, uint8_t b ) +{ + return ( ( r & 0xF8 ) << 8 ) | ( ( g & 0xFC ) << 3 ) | ( b >> 3 ); +} + +static inline uint16_t to565( uint32_t c ) +{ + return + ( ( c & 0xF80000 ) >> 19 ) | + ( ( c & 0x00FC00 ) >> 5 ) | + ( ( c & 0x0000F8 ) << 8 ); +} + +static const uint16_t DivTable[255*3+1] = { + 0xffff, 0xffff, 0xffff, 0xffff, 0xcccc, 0xaaaa, 0x9249, 0x8000, 0x71c7, 0x6666, 0x5d17, 0x5555, 0x4ec4, 0x4924, 0x4444, 0x4000, + 0x3c3c, 0x38e3, 0x35e5, 0x3333, 0x30c3, 0x2e8b, 0x2c85, 0x2aaa, 0x28f5, 0x2762, 0x25ed, 0x2492, 0x234f, 0x2222, 0x2108, 0x2000, + 0x1f07, 0x1e1e, 0x1d41, 0x1c71, 0x1bac, 0x1af2, 0x1a41, 0x1999, 0x18f9, 0x1861, 0x17d0, 0x1745, 0x16c1, 0x1642, 0x15c9, 0x1555, + 0x14e5, 0x147a, 0x1414, 0x13b1, 0x1352, 0x12f6, 0x129e, 0x1249, 0x11f7, 0x11a7, 0x115b, 0x1111, 0x10c9, 0x1084, 0x1041, 0x1000, + 0x0fc0, 0x0f83, 0x0f48, 0x0f0f, 0x0ed7, 0x0ea0, 0x0e6c, 0x0e38, 0x0e07, 0x0dd6, 0x0da7, 0x0d79, 0x0d4c, 0x0d20, 0x0cf6, 0x0ccc, + 0x0ca4, 0x0c7c, 0x0c56, 0x0c30, 0x0c0c, 0x0be8, 0x0bc5, 0x0ba2, 0x0b81, 0x0b60, 0x0b40, 0x0b21, 0x0b02, 0x0ae4, 0x0ac7, 0x0aaa, + 0x0a8e, 0x0a72, 0x0a57, 0x0a3d, 0x0a23, 0x0a0a, 0x09f1, 0x09d8, 0x09c0, 0x09a9, 0x0991, 0x097b, 0x0964, 0x094f, 0x0939, 0x0924, + 0x090f, 0x08fb, 0x08e7, 0x08d3, 0x08c0, 0x08ad, 0x089a, 0x0888, 0x0876, 0x0864, 0x0853, 0x0842, 0x0831, 0x0820, 0x0810, 0x0800, + 0x07f0, 0x07e0, 0x07d1, 0x07c1, 0x07b3, 0x07a4, 0x0795, 0x0787, 0x0779, 0x076b, 0x075d, 0x0750, 0x0743, 0x0736, 0x0729, 0x071c, + 0x070f, 0x0703, 0x06f7, 0x06eb, 0x06df, 0x06d3, 0x06c8, 0x06bc, 0x06b1, 0x06a6, 0x069b, 0x0690, 0x0685, 0x067b, 0x0670, 0x0666, + 0x065c, 0x0652, 0x0648, 0x063e, 0x0634, 0x062b, 0x0621, 0x0618, 0x060f, 0x0606, 0x05fd, 0x05f4, 0x05eb, 0x05e2, 0x05d9, 0x05d1, + 0x05c9, 0x05c0, 0x05b8, 0x05b0, 0x05a8, 0x05a0, 0x0598, 0x0590, 0x0588, 0x0581, 0x0579, 0x0572, 0x056b, 0x0563, 0x055c, 0x0555, + 0x054e, 0x0547, 0x0540, 0x0539, 0x0532, 0x052b, 0x0525, 0x051e, 0x0518, 0x0511, 0x050b, 0x0505, 0x04fe, 0x04f8, 0x04f2, 0x04ec, + 0x04e6, 0x04e0, 0x04da, 0x04d4, 0x04ce, 0x04c8, 0x04c3, 0x04bd, 0x04b8, 0x04b2, 0x04ad, 0x04a7, 0x04a2, 0x049c, 0x0497, 0x0492, + 0x048d, 0x0487, 0x0482, 0x047d, 0x0478, 0x0473, 0x046e, 0x0469, 0x0465, 0x0460, 0x045b, 0x0456, 0x0452, 0x044d, 0x0448, 0x0444, + 0x043f, 0x043b, 0x0436, 0x0432, 0x042d, 0x0429, 0x0425, 0x0421, 0x041c, 0x0418, 0x0414, 0x0410, 0x040c, 0x0408, 0x0404, 0x0400, + 0x03fc, 0x03f8, 0x03f4, 0x03f0, 0x03ec, 0x03e8, 0x03e4, 0x03e0, 0x03dd, 0x03d9, 0x03d5, 0x03d2, 0x03ce, 0x03ca, 0x03c7, 0x03c3, + 0x03c0, 0x03bc, 0x03b9, 0x03b5, 0x03b2, 0x03ae, 0x03ab, 0x03a8, 0x03a4, 0x03a1, 0x039e, 0x039b, 0x0397, 0x0394, 0x0391, 0x038e, + 0x038b, 0x0387, 0x0384, 0x0381, 0x037e, 0x037b, 0x0378, 0x0375, 0x0372, 0x036f, 0x036c, 0x0369, 0x0366, 0x0364, 0x0361, 0x035e, + 0x035b, 0x0358, 0x0355, 0x0353, 0x0350, 0x034d, 0x034a, 0x0348, 0x0345, 0x0342, 0x0340, 0x033d, 0x033a, 0x0338, 0x0335, 0x0333, + 0x0330, 0x032e, 0x032b, 0x0329, 0x0326, 0x0324, 0x0321, 0x031f, 0x031c, 0x031a, 0x0317, 0x0315, 0x0313, 0x0310, 0x030e, 0x030c, + 0x0309, 0x0307, 0x0305, 0x0303, 0x0300, 0x02fe, 0x02fc, 0x02fa, 0x02f7, 0x02f5, 0x02f3, 0x02f1, 0x02ef, 0x02ec, 0x02ea, 0x02e8, + 0x02e6, 0x02e4, 0x02e2, 0x02e0, 0x02de, 0x02dc, 0x02da, 0x02d8, 0x02d6, 0x02d4, 0x02d2, 0x02d0, 0x02ce, 0x02cc, 0x02ca, 0x02c8, + 0x02c6, 0x02c4, 0x02c2, 0x02c0, 0x02be, 0x02bc, 0x02bb, 0x02b9, 0x02b7, 0x02b5, 0x02b3, 0x02b1, 0x02b0, 0x02ae, 0x02ac, 0x02aa, + 0x02a8, 0x02a7, 0x02a5, 0x02a3, 0x02a1, 0x02a0, 0x029e, 0x029c, 0x029b, 0x0299, 0x0297, 0x0295, 0x0294, 0x0292, 0x0291, 0x028f, + 0x028d, 0x028c, 0x028a, 0x0288, 0x0287, 0x0285, 0x0284, 0x0282, 0x0280, 0x027f, 0x027d, 0x027c, 0x027a, 0x0279, 0x0277, 0x0276, + 0x0274, 0x0273, 0x0271, 0x0270, 0x026e, 0x026d, 0x026b, 0x026a, 0x0268, 0x0267, 0x0265, 0x0264, 0x0263, 0x0261, 0x0260, 0x025e, + 0x025d, 0x025c, 0x025a, 0x0259, 0x0257, 0x0256, 0x0255, 0x0253, 0x0252, 0x0251, 0x024f, 0x024e, 0x024d, 0x024b, 0x024a, 0x0249, + 0x0247, 0x0246, 0x0245, 0x0243, 0x0242, 0x0241, 0x0240, 0x023e, 0x023d, 0x023c, 0x023b, 0x0239, 0x0238, 0x0237, 0x0236, 0x0234, + 0x0233, 0x0232, 0x0231, 0x0230, 0x022e, 0x022d, 0x022c, 0x022b, 0x022a, 0x0229, 0x0227, 0x0226, 0x0225, 0x0224, 0x0223, 0x0222, + 0x0220, 0x021f, 0x021e, 0x021d, 0x021c, 0x021b, 0x021a, 0x0219, 0x0218, 0x0216, 0x0215, 0x0214, 0x0213, 0x0212, 0x0211, 0x0210, + 0x020f, 0x020e, 0x020d, 0x020c, 0x020b, 0x020a, 0x0209, 0x0208, 0x0207, 0x0206, 0x0205, 0x0204, 0x0203, 0x0202, 0x0201, 0x0200, + 0x01ff, 0x01fe, 0x01fd, 0x01fc, 0x01fb, 0x01fa, 0x01f9, 0x01f8, 0x01f7, 0x01f6, 0x01f5, 0x01f4, 0x01f3, 0x01f2, 0x01f1, 0x01f0, + 0x01ef, 0x01ee, 0x01ed, 0x01ec, 0x01eb, 0x01ea, 0x01e9, 0x01e9, 0x01e8, 0x01e7, 0x01e6, 0x01e5, 0x01e4, 0x01e3, 0x01e2, 0x01e1, + 0x01e0, 0x01e0, 0x01df, 0x01de, 0x01dd, 0x01dc, 0x01db, 0x01da, 0x01da, 0x01d9, 0x01d8, 0x01d7, 0x01d6, 0x01d5, 0x01d4, 0x01d4, + 0x01d3, 0x01d2, 0x01d1, 0x01d0, 0x01cf, 0x01cf, 0x01ce, 0x01cd, 0x01cc, 0x01cb, 0x01cb, 0x01ca, 0x01c9, 0x01c8, 0x01c7, 0x01c7, + 0x01c6, 0x01c5, 0x01c4, 0x01c3, 0x01c3, 0x01c2, 0x01c1, 0x01c0, 0x01c0, 0x01bf, 0x01be, 0x01bd, 0x01bd, 0x01bc, 0x01bb, 0x01ba, + 0x01ba, 0x01b9, 0x01b8, 0x01b7, 0x01b7, 0x01b6, 0x01b5, 0x01b4, 0x01b4, 0x01b3, 0x01b2, 0x01b2, 0x01b1, 0x01b0, 0x01af, 0x01af, + 0x01ae, 0x01ad, 0x01ad, 0x01ac, 0x01ab, 0x01aa, 0x01aa, 0x01a9, 0x01a8, 0x01a8, 0x01a7, 0x01a6, 0x01a6, 0x01a5, 0x01a4, 0x01a4, + 0x01a3, 0x01a2, 0x01a2, 0x01a1, 0x01a0, 0x01a0, 0x019f, 0x019e, 0x019e, 0x019d, 0x019c, 0x019c, 0x019b, 0x019a, 0x019a, 0x0199, + 0x0198, 0x0198, 0x0197, 0x0197, 0x0196, 0x0195, 0x0195, 0x0194, 0x0193, 0x0193, 0x0192, 0x0192, 0x0191, 0x0190, 0x0190, 0x018f, + 0x018f, 0x018e, 0x018d, 0x018d, 0x018c, 0x018b, 0x018b, 0x018a, 0x018a, 0x0189, 0x0189, 0x0188, 0x0187, 0x0187, 0x0186, 0x0186, + 0x0185, 0x0184, 0x0184, 0x0183, 0x0183, 0x0182, 0x0182, 0x0181, 0x0180, 0x0180, 0x017f, 0x017f, 0x017e, 0x017e, 0x017d, 0x017d, + 0x017c, 0x017b, 0x017b, 0x017a, 0x017a, 0x0179, 0x0179, 0x0178, 0x0178, 0x0177, 0x0177, 0x0176, 0x0175, 0x0175, 0x0174, 0x0174, + 0x0173, 0x0173, 0x0172, 0x0172, 0x0171, 0x0171, 0x0170, 0x0170, 0x016f, 0x016f, 0x016e, 0x016e, 0x016d, 0x016d, 0x016c, 0x016c, + 0x016b, 0x016b, 0x016a, 0x016a, 0x0169, 0x0169, 0x0168, 0x0168, 0x0167, 0x0167, 0x0166, 0x0166, 0x0165, 0x0165, 0x0164, 0x0164, + 0x0163, 0x0163, 0x0162, 0x0162, 0x0161, 0x0161, 0x0160, 0x0160, 0x015f, 0x015f, 0x015e, 0x015e, 0x015d, 0x015d, 0x015d, 0x015c, + 0x015c, 0x015b, 0x015b, 0x015a, 0x015a, 0x0159, 0x0159, 0x0158, 0x0158, 0x0158, 0x0157, 0x0157, 0x0156, 0x0156 +}; + +#if defined __ARM_NEON && defined __aarch64__ +static const uint16_t DivTableNEON[255*3+1] = { + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x1c71, 0x1af2, 0x1999, 0x1861, 0x1745, 0x1642, 0x1555, 0x147a, 0x13b1, 0x12f6, 0x1249, 0x11a7, 0x1111, 0x1084, 0x1000, + 0x0f83, 0x0f0f, 0x0ea0, 0x0e38, 0x0dd6, 0x0d79, 0x0d20, 0x0ccc, 0x0c7c, 0x0c30, 0x0be8, 0x0ba2, 0x0b60, 0x0b21, 0x0ae4, 0x0aaa, + 0x0a72, 0x0a3d, 0x0a0a, 0x09d8, 0x09a9, 0x097b, 0x094f, 0x0924, 0x08fb, 0x08d3, 0x08ad, 0x0888, 0x0864, 0x0842, 0x0820, 0x0800, + 0x07e0, 0x07c1, 0x07a4, 0x0787, 0x076b, 0x0750, 0x0736, 0x071c, 0x0703, 0x06eb, 0x06d3, 0x06bc, 0x06a6, 0x0690, 0x067b, 0x0666, + 0x0652, 0x063e, 0x062b, 0x0618, 0x0606, 0x05f4, 0x05e2, 0x05d1, 0x05c0, 0x05b0, 0x05a0, 0x0590, 0x0581, 0x0572, 0x0563, 0x0555, + 0x0547, 0x0539, 0x052b, 0x051e, 0x0511, 0x0505, 0x04f8, 0x04ec, 0x04e0, 0x04d4, 0x04c8, 0x04bd, 0x04b2, 0x04a7, 0x049c, 0x0492, + 0x0487, 0x047d, 0x0473, 0x0469, 0x0460, 0x0456, 0x044d, 0x0444, 0x043b, 0x0432, 0x0429, 0x0421, 0x0418, 0x0410, 0x0408, 0x0400, + 0x03f8, 0x03f0, 0x03e8, 0x03e0, 0x03d9, 0x03d2, 0x03ca, 0x03c3, 0x03bc, 0x03b5, 0x03ae, 0x03a8, 0x03a1, 0x039b, 0x0394, 0x038e, + 0x0387, 0x0381, 0x037b, 0x0375, 0x036f, 0x0369, 0x0364, 0x035e, 0x0358, 0x0353, 0x034d, 0x0348, 0x0342, 0x033d, 0x0338, 0x0333, + 0x032e, 0x0329, 0x0324, 0x031f, 0x031a, 0x0315, 0x0310, 0x030c, 0x0307, 0x0303, 0x02fe, 0x02fa, 0x02f5, 0x02f1, 0x02ec, 0x02e8, + 0x02e4, 0x02e0, 0x02dc, 0x02d8, 0x02d4, 0x02d0, 0x02cc, 0x02c8, 0x02c4, 0x02c0, 0x02bc, 0x02b9, 0x02b5, 0x02b1, 0x02ae, 0x02aa, + 0x02a7, 0x02a3, 0x02a0, 0x029c, 0x0299, 0x0295, 0x0292, 0x028f, 0x028c, 0x0288, 0x0285, 0x0282, 0x027f, 0x027c, 0x0279, 0x0276, + 0x0273, 0x0270, 0x026d, 0x026a, 0x0267, 0x0264, 0x0261, 0x025e, 0x025c, 0x0259, 0x0256, 0x0253, 0x0251, 0x024e, 0x024b, 0x0249, + 0x0246, 0x0243, 0x0241, 0x023e, 0x023c, 0x0239, 0x0237, 0x0234, 0x0232, 0x0230, 0x022d, 0x022b, 0x0229, 0x0226, 0x0224, 0x0222, + 0x021f, 0x021d, 0x021b, 0x0219, 0x0216, 0x0214, 0x0212, 0x0210, 0x020e, 0x020c, 0x020a, 0x0208, 0x0206, 0x0204, 0x0202, 0x0200, + 0x01fe, 0x01fc, 0x01fa, 0x01f8, 0x01f6, 0x01f4, 0x01f2, 0x01f0, 0x01ee, 0x01ec, 0x01ea, 0x01e9, 0x01e7, 0x01e5, 0x01e3, 0x01e1, + 0x01e0, 0x01de, 0x01dc, 0x01da, 0x01d9, 0x01d7, 0x01d5, 0x01d4, 0x01d2, 0x01d0, 0x01cf, 0x01cd, 0x01cb, 0x01ca, 0x01c8, 0x01c7, + 0x01c5, 0x01c3, 0x01c2, 0x01c0, 0x01bf, 0x01bd, 0x01bc, 0x01ba, 0x01b9, 0x01b7, 0x01b6, 0x01b4, 0x01b3, 0x01b2, 0x01b0, 0x01af, + 0x01ad, 0x01ac, 0x01aa, 0x01a9, 0x01a8, 0x01a6, 0x01a5, 0x01a4, 0x01a2, 0x01a1, 0x01a0, 0x019e, 0x019d, 0x019c, 0x019a, 0x0199, + 0x0198, 0x0197, 0x0195, 0x0194, 0x0193, 0x0192, 0x0190, 0x018f, 0x018e, 0x018d, 0x018b, 0x018a, 0x0189, 0x0188, 0x0187, 0x0186, + 0x0184, 0x0183, 0x0182, 0x0181, 0x0180, 0x017f, 0x017e, 0x017d, 0x017b, 0x017a, 0x0179, 0x0178, 0x0177, 0x0176, 0x0175, 0x0174, + 0x0173, 0x0172, 0x0171, 0x0170, 0x016f, 0x016e, 0x016d, 0x016c, 0x016b, 0x016a, 0x0169, 0x0168, 0x0167, 0x0166, 0x0165, 0x0164, + 0x0163, 0x0162, 0x0161, 0x0160, 0x015f, 0x015e, 0x015d, 0x015c, 0x015b, 0x015a, 0x0159, 0x0158, 0x0158, 0x0157, 0x0156, 0x0155, + 0x0154, 0x0153, 0x0152, 0x0151, 0x0150, 0x0150, 0x014f, 0x014e, 0x014d, 0x014c, 0x014b, 0x014a, 0x014a, 0x0149, 0x0148, 0x0147, + 0x0146, 0x0146, 0x0145, 0x0144, 0x0143, 0x0142, 0x0142, 0x0141, 0x0140, 0x013f, 0x013e, 0x013e, 0x013d, 0x013c, 0x013b, 0x013b, + 0x013a, 0x0139, 0x0138, 0x0138, 0x0137, 0x0136, 0x0135, 0x0135, 0x0134, 0x0133, 0x0132, 0x0132, 0x0131, 0x0130, 0x0130, 0x012f, + 0x012e, 0x012e, 0x012d, 0x012c, 0x012b, 0x012b, 0x012a, 0x0129, 0x0129, 0x0128, 0x0127, 0x0127, 0x0126, 0x0125, 0x0125, 0x0124, + 0x0123, 0x0123, 0x0122, 0x0121, 0x0121, 0x0120, 0x0120, 0x011f, 0x011e, 0x011e, 0x011d, 0x011c, 0x011c, 0x011b, 0x011b, 0x011a, + 0x0119, 0x0119, 0x0118, 0x0118, 0x0117, 0x0116, 0x0116, 0x0115, 0x0115, 0x0114, 0x0113, 0x0113, 0x0112, 0x0112, 0x0111, 0x0111, + 0x0110, 0x010f, 0x010f, 0x010e, 0x010e, 0x010d, 0x010d, 0x010c, 0x010c, 0x010b, 0x010a, 0x010a, 0x0109, 0x0109, 0x0108, 0x0108, + 0x0107, 0x0107, 0x0106, 0x0106, 0x0105, 0x0105, 0x0104, 0x0104, 0x0103, 0x0103, 0x0102, 0x0102, 0x0101, 0x0101, 0x0100, 0x0100, + 0x00ff, 0x00ff, 0x00fe, 0x00fe, 0x00fd, 0x00fd, 0x00fc, 0x00fc, 0x00fb, 0x00fb, 0x00fa, 0x00fa, 0x00f9, 0x00f9, 0x00f8, 0x00f8, + 0x00f7, 0x00f7, 0x00f6, 0x00f6, 0x00f5, 0x00f5, 0x00f4, 0x00f4, 0x00f4, 0x00f3, 0x00f3, 0x00f2, 0x00f2, 0x00f1, 0x00f1, 0x00f0, + 0x00f0, 0x00f0, 0x00ef, 0x00ef, 0x00ee, 0x00ee, 0x00ed, 0x00ed, 0x00ed, 0x00ec, 0x00ec, 0x00eb, 0x00eb, 0x00ea, 0x00ea, 0x00ea, + 0x00e9, 0x00e9, 0x00e8, 0x00e8, 0x00e7, 0x00e7, 0x00e7, 0x00e6, 0x00e6, 0x00e5, 0x00e5, 0x00e5, 0x00e4, 0x00e4, 0x00e3, 0x00e3, + 0x00e3, 0x00e2, 0x00e2, 0x00e1, 0x00e1, 0x00e1, 0x00e0, 0x00e0, 0x00e0, 0x00df, 0x00df, 0x00de, 0x00de, 0x00de, 0x00dd, 0x00dd, + 0x00dd, 0x00dc, 0x00dc, 0x00db, 0x00db, 0x00db, 0x00da, 0x00da, 0x00da, 0x00d9, 0x00d9, 0x00d9, 0x00d8, 0x00d8, 0x00d7, 0x00d7, + 0x00d7, 0x00d6, 0x00d6, 0x00d6, 0x00d5, 0x00d5, 0x00d5, 0x00d4, 0x00d4, 0x00d4, 0x00d3, 0x00d3, 0x00d3, 0x00d2, 0x00d2, 0x00d2, + 0x00d1, 0x00d1, 0x00d1, 0x00d0, 0x00d0, 0x00d0, 0x00cf, 0x00cf, 0x00cf, 0x00ce, 0x00ce, 0x00ce, 0x00cd, 0x00cd, 0x00cd, 0x00cc, + 0x00cc, 0x00cc, 0x00cb, 0x00cb, 0x00cb, 0x00ca, 0x00ca, 0x00ca, 0x00c9, 0x00c9, 0x00c9, 0x00c9, 0x00c8, 0x00c8, 0x00c8, 0x00c7, + 0x00c7, 0x00c7, 0x00c6, 0x00c6, 0x00c6, 0x00c5, 0x00c5, 0x00c5, 0x00c5, 0x00c4, 0x00c4, 0x00c4, 0x00c3, 0x00c3, 0x00c3, 0x00c3, + 0x00c2, 0x00c2, 0x00c2, 0x00c1, 0x00c1, 0x00c1, 0x00c1, 0x00c0, 0x00c0, 0x00c0, 0x00bf, 0x00bf, 0x00bf, 0x00bf, 0x00be, 0x00be, + 0x00be, 0x00bd, 0x00bd, 0x00bd, 0x00bd, 0x00bc, 0x00bc, 0x00bc, 0x00bc, 0x00bb, 0x00bb, 0x00bb, 0x00ba, 0x00ba, 0x00ba, 0x00ba, + 0x00b9, 0x00b9, 0x00b9, 0x00b9, 0x00b8, 0x00b8, 0x00b8, 0x00b8, 0x00b7, 0x00b7, 0x00b7, 0x00b7, 0x00b6, 0x00b6, 0x00b6, 0x00b6, + 0x00b5, 0x00b5, 0x00b5, 0x00b5, 0x00b4, 0x00b4, 0x00b4, 0x00b4, 0x00b3, 0x00b3, 0x00b3, 0x00b3, 0x00b2, 0x00b2, 0x00b2, 0x00b2, + 0x00b1, 0x00b1, 0x00b1, 0x00b1, 0x00b0, 0x00b0, 0x00b0, 0x00b0, 0x00af, 0x00af, 0x00af, 0x00af, 0x00ae, 0x00ae, 0x00ae, 0x00ae, + 0x00ae, 0x00ad, 0x00ad, 0x00ad, 0x00ad, 0x00ac, 0x00ac, 0x00ac, 0x00ac, 0x00ac, 0x00ab, 0x00ab, 0x00ab, 0x00ab, +}; +#endif + + +static tracy_force_inline uint64_t ProcessRGB( const uint8_t* src ) +{ +#ifdef __SSE4_1__ + __m128i px0 = _mm_loadu_si128(((__m128i*)src) + 0); + __m128i px1 = _mm_loadu_si128(((__m128i*)src) + 1); + __m128i px2 = _mm_loadu_si128(((__m128i*)src) + 2); + __m128i px3 = _mm_loadu_si128(((__m128i*)src) + 3); + + __m128i smask = _mm_set1_epi32( 0xF8FCF8 ); + __m128i sd0 = _mm_and_si128( px0, smask ); + __m128i sd1 = _mm_and_si128( px1, smask ); + __m128i sd2 = _mm_and_si128( px2, smask ); + __m128i sd3 = _mm_and_si128( px3, smask ); + + __m128i sc = _mm_shuffle_epi32(sd0, _MM_SHUFFLE(0, 0, 0, 0)); + + __m128i sc0 = _mm_cmpeq_epi8(sd0, sc); + __m128i sc1 = _mm_cmpeq_epi8(sd1, sc); + __m128i sc2 = _mm_cmpeq_epi8(sd2, sc); + __m128i sc3 = _mm_cmpeq_epi8(sd3, sc); + + __m128i sm0 = _mm_and_si128(sc0, sc1); + __m128i sm1 = _mm_and_si128(sc2, sc3); + __m128i sm = _mm_and_si128(sm0, sm1); + + if( _mm_testc_si128(sm, _mm_set1_epi32(-1)) ) + { + return uint64_t( to565( src[0], src[1], src[2] ) ) << 16; + } + + __m128i amask = _mm_set1_epi32( 0xFFFFFF ); + px0 = _mm_and_si128( px0, amask ); + px1 = _mm_and_si128( px1, amask ); + px2 = _mm_and_si128( px2, amask ); + px3 = _mm_and_si128( px3, amask ); + + __m128i min0 = _mm_min_epu8( px0, px1 ); + __m128i min1 = _mm_min_epu8( px2, px3 ); + __m128i min2 = _mm_min_epu8( min0, min1 ); + + __m128i max0 = _mm_max_epu8( px0, px1 ); + __m128i max1 = _mm_max_epu8( px2, px3 ); + __m128i max2 = _mm_max_epu8( max0, max1 ); + + __m128i min3 = _mm_shuffle_epi32( min2, _MM_SHUFFLE( 2, 3, 0, 1 ) ); + __m128i max3 = _mm_shuffle_epi32( max2, _MM_SHUFFLE( 2, 3, 0, 1 ) ); + __m128i min4 = _mm_min_epu8( min2, min3 ); + __m128i max4 = _mm_max_epu8( max2, max3 ); + + __m128i min5 = _mm_shuffle_epi32( min4, _MM_SHUFFLE( 0, 0, 2, 2 ) ); + __m128i max5 = _mm_shuffle_epi32( max4, _MM_SHUFFLE( 0, 0, 2, 2 ) ); + __m128i rmin = _mm_min_epu8( min4, min5 ); + __m128i rmax = _mm_max_epu8( max4, max5 ); + + __m128i range1 = _mm_subs_epu8( rmax, rmin ); + __m128i range2 = _mm_sad_epu8( rmax, rmin ); + + uint32_t vrange = _mm_cvtsi128_si32( range2 ) >> 1; + __m128i range = _mm_set1_epi16( DivTable[vrange] ); + + __m128i inset1 = _mm_srli_epi16( range1, 4 ); + __m128i inset = _mm_and_si128( inset1, _mm_set1_epi8( 0xF ) ); + __m128i min = _mm_adds_epu8( rmin, inset ); + __m128i max = _mm_subs_epu8( rmax, inset ); + + __m128i c0 = _mm_subs_epu8( px0, rmin ); + __m128i c1 = _mm_subs_epu8( px1, rmin ); + __m128i c2 = _mm_subs_epu8( px2, rmin ); + __m128i c3 = _mm_subs_epu8( px3, rmin ); + + __m128i is0 = _mm_maddubs_epi16( c0, _mm_set1_epi8( 1 ) ); + __m128i is1 = _mm_maddubs_epi16( c1, _mm_set1_epi8( 1 ) ); + __m128i is2 = _mm_maddubs_epi16( c2, _mm_set1_epi8( 1 ) ); + __m128i is3 = _mm_maddubs_epi16( c3, _mm_set1_epi8( 1 ) ); + + __m128i s0 = _mm_hadd_epi16( is0, is1 ); + __m128i s1 = _mm_hadd_epi16( is2, is3 ); + + __m128i m0 = _mm_mulhi_epu16( s0, range ); + __m128i m1 = _mm_mulhi_epu16( s1, range ); + + __m128i p0 = _mm_packus_epi16( m0, m1 ); + + __m128i p1 = _mm_or_si128( _mm_srai_epi32( p0, 6 ), _mm_srai_epi32( p0, 12 ) ); + __m128i p2 = _mm_or_si128( _mm_srai_epi32( p0, 18 ), p0 ); + __m128i p3 = _mm_or_si128( p1, p2 ); + __m128i p =_mm_shuffle_epi8( p3, _mm_set1_epi32( 0x0C080400 ) ); + + uint32_t vmin = _mm_cvtsi128_si32( min ); + uint32_t vmax = _mm_cvtsi128_si32( max ); + uint32_t vp = _mm_cvtsi128_si32( p ); + + return uint64_t( ( uint64_t( to565( vmin ) ) << 16 ) | to565( vmax ) | ( uint64_t( vp ) << 32 ) ); +#elif defined __ARM_NEON +# ifdef __aarch64__ + uint8x16x4_t px = vld4q_u8( src ); + + uint8x16_t lr = px.val[0]; + uint8x16_t lg = px.val[1]; + uint8x16_t lb = px.val[2]; + + uint8_t rmaxr = vmaxvq_u8( lr ); + uint8_t rmaxg = vmaxvq_u8( lg ); + uint8_t rmaxb = vmaxvq_u8( lb ); + + uint8_t rminr = vminvq_u8( lr ); + uint8_t rming = vminvq_u8( lg ); + uint8_t rminb = vminvq_u8( lb ); + + int rr = rmaxr - rminr; + int rg = rmaxg - rming; + int rb = rmaxb - rminb; + + int vrange1 = rr + rg + rb; + uint16_t vrange2 = DivTableNEON[vrange1]; + + uint8_t insetr = rr >> 4; + uint8_t insetg = rg >> 4; + uint8_t insetb = rb >> 4; + + uint8_t minr = rminr + insetr; + uint8_t ming = rming + insetg; + uint8_t minb = rminb + insetb; + + uint8_t maxr = rmaxr - insetr; + uint8_t maxg = rmaxg - insetg; + uint8_t maxb = rmaxb - insetb; + + uint8x16_t cr = vsubq_u8( lr, vdupq_n_u8( rminr ) ); + uint8x16_t cg = vsubq_u8( lg, vdupq_n_u8( rming ) ); + uint8x16_t cb = vsubq_u8( lb, vdupq_n_u8( rminb ) ); + + uint16x8_t is0l = vaddl_u8( vget_low_u8( cr ), vget_low_u8( cg ) ); + uint16x8_t is0h = vaddl_u8( vget_high_u8( cr ), vget_high_u8( cg ) ); + uint16x8_t is1l = vaddw_u8( is0l, vget_low_u8( cb ) ); + uint16x8_t is1h = vaddw_u8( is0h, vget_high_u8( cb ) ); + + int16x8_t range = vdupq_n_s16( vrange2 ); + uint16x8_t m0 = vreinterpretq_u16_s16( vqdmulhq_s16( vreinterpretq_s16_u16( is1l ), range ) ); + uint16x8_t m1 = vreinterpretq_u16_s16( vqdmulhq_s16( vreinterpretq_s16_u16( is1h ), range ) ); + + uint8x8_t p00 = vmovn_u16( m0 ); + uint8x8_t p01 = vmovn_u16( m1 ); + uint8x16_t p0 = vcombine_u8( p00, p01 ); + + uint32x4_t p1 = vaddq_u32( vshrq_n_u32( vreinterpretq_u32_u8( p0 ), 6 ), vshrq_n_u32( vreinterpretq_u32_u8( p0 ), 12 ) ); + uint32x4_t p2 = vaddq_u32( vshrq_n_u32( vreinterpretq_u32_u8( p0 ), 18 ), vreinterpretq_u32_u8( p0 ) ); + uint32x4_t p3 = vaddq_u32( p1, p2 ); + + uint16x4x2_t p4 = vuzp_u16( vget_low_u16( vreinterpretq_u16_u32( p3 ) ), vget_high_u16( vreinterpretq_u16_u32( p3 ) ) ); + uint8x8x2_t p = vuzp_u8( vreinterpret_u8_u16( p4.val[0] ), vreinterpret_u8_u16( p4.val[0] ) ); + + uint32_t vp; + vst1_lane_u32( &vp, vreinterpret_u32_u8( p.val[0] ), 0 ); + + return uint64_t( ( uint64_t( to565( minr, ming, minb ) ) << 16 ) | to565( maxr, maxg, maxb ) | ( uint64_t( vp ) << 32 ) ); +# else + uint32x4_t px0 = vld1q_u32( (uint32_t*)src ); + uint32x4_t px1 = vld1q_u32( (uint32_t*)src + 4 ); + uint32x4_t px2 = vld1q_u32( (uint32_t*)src + 8 ); + uint32x4_t px3 = vld1q_u32( (uint32_t*)src + 12 ); + + uint32x4_t smask = vdupq_n_u32( 0xF8FCF8 ); + uint32x4_t sd0 = vandq_u32( smask, px0 ); + uint32x4_t sd1 = vandq_u32( smask, px1 ); + uint32x4_t sd2 = vandq_u32( smask, px2 ); + uint32x4_t sd3 = vandq_u32( smask, px3 ); + + uint32x4_t sc = vdupq_n_u32( sd0[0] ); + + uint32x4_t sc0 = vceqq_u32( sd0, sc ); + uint32x4_t sc1 = vceqq_u32( sd1, sc ); + uint32x4_t sc2 = vceqq_u32( sd2, sc ); + uint32x4_t sc3 = vceqq_u32( sd3, sc ); + + uint32x4_t sm0 = vandq_u32( sc0, sc1 ); + uint32x4_t sm1 = vandq_u32( sc2, sc3 ); + int64x2_t sm = vreinterpretq_s64_u32( vandq_u32( sm0, sm1 ) ); + + if( sm[0] == -1 && sm[1] == -1 ) + { + return uint64_t( to565( src[0], src[1], src[2] ) ) << 16; + } + + uint32x4_t mask = vdupq_n_u32( 0xFFFFFF ); + uint8x16_t l0 = vreinterpretq_u8_u32( vandq_u32( mask, px0 ) ); + uint8x16_t l1 = vreinterpretq_u8_u32( vandq_u32( mask, px1 ) ); + uint8x16_t l2 = vreinterpretq_u8_u32( vandq_u32( mask, px2 ) ); + uint8x16_t l3 = vreinterpretq_u8_u32( vandq_u32( mask, px3 ) ); + + uint8x16_t min0 = vminq_u8( l0, l1 ); + uint8x16_t min1 = vminq_u8( l2, l3 ); + uint8x16_t min2 = vminq_u8( min0, min1 ); + + uint8x16_t max0 = vmaxq_u8( l0, l1 ); + uint8x16_t max1 = vmaxq_u8( l2, l3 ); + uint8x16_t max2 = vmaxq_u8( max0, max1 ); + + uint8x16_t min3 = vreinterpretq_u8_u32( vrev64q_u32( vreinterpretq_u32_u8( min2 ) ) ); + uint8x16_t max3 = vreinterpretq_u8_u32( vrev64q_u32( vreinterpretq_u32_u8( max2 ) ) ); + + uint8x16_t min4 = vminq_u8( min2, min3 ); + uint8x16_t max4 = vmaxq_u8( max2, max3 ); + + uint8x16_t min5 = vcombine_u8( vget_high_u8( min4 ), vget_low_u8( min4 ) ); + uint8x16_t max5 = vcombine_u8( vget_high_u8( max4 ), vget_low_u8( max4 ) ); + + uint8x16_t rmin = vminq_u8( min4, min5 ); + uint8x16_t rmax = vmaxq_u8( max4, max5 ); + + uint8x16_t range1 = vsubq_u8( rmax, rmin ); + uint8x8_t range2 = vget_low_u8( range1 ); + uint8x8x2_t range3 = vzip_u8( range2, vdup_n_u8( 0 ) ); + uint16x4_t range4 = vreinterpret_u16_u8( range3.val[0] ); + + uint16_t vrange1; + uint16x4_t range5 = vpadd_u16( range4, range4 ); + uint16x4_t range6 = vpadd_u16( range5, range5 ); + vst1_lane_u16( &vrange1, range6, 0 ); + + uint32_t vrange2 = ( 2 << 16 ) / uint32_t( vrange1 + 1 ); + uint16x8_t range = vdupq_n_u16( vrange2 ); + + uint8x16_t inset = vshrq_n_u8( range1, 4 ); + uint8x16_t min = vaddq_u8( rmin, inset ); + uint8x16_t max = vsubq_u8( rmax, inset ); + + uint8x16_t c0 = vsubq_u8( l0, rmin ); + uint8x16_t c1 = vsubq_u8( l1, rmin ); + uint8x16_t c2 = vsubq_u8( l2, rmin ); + uint8x16_t c3 = vsubq_u8( l3, rmin ); + + uint16x8_t is0 = vpaddlq_u8( c0 ); + uint16x8_t is1 = vpaddlq_u8( c1 ); + uint16x8_t is2 = vpaddlq_u8( c2 ); + uint16x8_t is3 = vpaddlq_u8( c3 ); + + uint16x4_t is4 = vpadd_u16( vget_low_u16( is0 ), vget_high_u16( is0 ) ); + uint16x4_t is5 = vpadd_u16( vget_low_u16( is1 ), vget_high_u16( is1 ) ); + uint16x4_t is6 = vpadd_u16( vget_low_u16( is2 ), vget_high_u16( is2 ) ); + uint16x4_t is7 = vpadd_u16( vget_low_u16( is3 ), vget_high_u16( is3 ) ); + + uint16x8_t s0 = vcombine_u16( is4, is5 ); + uint16x8_t s1 = vcombine_u16( is6, is7 ); + + uint16x8_t m0 = vreinterpretq_u16_s16( vqdmulhq_s16( vreinterpretq_s16_u16( s0 ), vreinterpretq_s16_u16( range ) ) ); + uint16x8_t m1 = vreinterpretq_u16_s16( vqdmulhq_s16( vreinterpretq_s16_u16( s1 ), vreinterpretq_s16_u16( range ) ) ); + + uint8x8_t p00 = vmovn_u16( m0 ); + uint8x8_t p01 = vmovn_u16( m1 ); + uint8x16_t p0 = vcombine_u8( p00, p01 ); + + uint32x4_t p1 = vaddq_u32( vshrq_n_u32( vreinterpretq_u32_u8( p0 ), 6 ), vshrq_n_u32( vreinterpretq_u32_u8( p0 ), 12 ) ); + uint32x4_t p2 = vaddq_u32( vshrq_n_u32( vreinterpretq_u32_u8( p0 ), 18 ), vreinterpretq_u32_u8( p0 ) ); + uint32x4_t p3 = vaddq_u32( p1, p2 ); + + uint16x4x2_t p4 = vuzp_u16( vget_low_u16( vreinterpretq_u16_u32( p3 ) ), vget_high_u16( vreinterpretq_u16_u32( p3 ) ) ); + uint8x8x2_t p = vuzp_u8( vreinterpret_u8_u16( p4.val[0] ), vreinterpret_u8_u16( p4.val[0] ) ); + + uint32_t vmin, vmax, vp; + vst1q_lane_u32( &vmin, vreinterpretq_u32_u8( min ), 0 ); + vst1q_lane_u32( &vmax, vreinterpretq_u32_u8( max ), 0 ); + vst1_lane_u32( &vp, vreinterpret_u32_u8( p.val[0] ), 0 ); + + return uint64_t( ( uint64_t( to565( vmin ) ) << 16 ) | to565( vmax ) | ( uint64_t( vp ) << 32 ) ); +# endif +#else + uint32_t ref; + memcpy( &ref, src, 4 ); + uint32_t refMask = ref & 0xF8FCF8; + auto stmp = src + 4; + for( int i=1; i<16; i++ ) + { + uint32_t px; + memcpy( &px, stmp, 4 ); + if( ( px & 0xF8FCF8 ) != refMask ) break; + stmp += 4; + } + if( stmp == src + 64 ) + { + return uint64_t( to565( ref ) ) << 16; + } + + uint8_t min[3] = { src[0], src[1], src[2] }; + uint8_t max[3] = { src[0], src[1], src[2] }; + auto tmp = src + 4; + for( int i=1; i<16; i++ ) + { + for( int j=0; j<3; j++ ) + { + if( tmp[j] < min[j] ) min[j] = tmp[j]; + else if( tmp[j] > max[j] ) max[j] = tmp[j]; + } + tmp += 4; + } + + const uint32_t range = DivTable[max[0] - min[0] + max[1] - min[1] + max[2] - min[2]]; + const uint32_t rmin = min[0] + min[1] + min[2]; + for( int i=0; i<3; i++ ) + { + const uint8_t inset = ( max[i] - min[i] ) >> 4; + min[i] += inset; + max[i] -= inset; + } + + uint32_t data = 0; + for( int i=0; i<16; i++ ) + { + const uint32_t c = src[0] + src[1] + src[2] - rmin; + const uint8_t idx = ( c * range ) >> 16; + data |= idx << (i*2); + src += 4; + } + + return uint64_t( ( uint64_t( to565( min[0], min[1], min[2] ) ) << 16 ) | to565( max[0], max[1], max[2] ) | ( uint64_t( data ) << 32 ) ); +#endif +} + +#ifdef __AVX2__ +static tracy_force_inline void ProcessRGB_AVX( const uint8_t* src, char*& dst ) +{ + __m256i px0 = _mm256_loadu_si256(((__m256i*)src) + 0); + __m256i px1 = _mm256_loadu_si256(((__m256i*)src) + 1); + __m256i px2 = _mm256_loadu_si256(((__m256i*)src) + 2); + __m256i px3 = _mm256_loadu_si256(((__m256i*)src) + 3); + + __m256i smask = _mm256_set1_epi32( 0xF8FCF8 ); + __m256i sd0 = _mm256_and_si256( px0, smask ); + __m256i sd1 = _mm256_and_si256( px1, smask ); + __m256i sd2 = _mm256_and_si256( px2, smask ); + __m256i sd3 = _mm256_and_si256( px3, smask ); + + __m256i sc = _mm256_shuffle_epi32(sd0, _MM_SHUFFLE(0, 0, 0, 0)); + + __m256i sc0 = _mm256_cmpeq_epi8( sd0, sc ); + __m256i sc1 = _mm256_cmpeq_epi8( sd1, sc ); + __m256i sc2 = _mm256_cmpeq_epi8( sd2, sc ); + __m256i sc3 = _mm256_cmpeq_epi8( sd3, sc ); + + __m256i sm0 = _mm256_and_si256( sc0, sc1 ); + __m256i sm1 = _mm256_and_si256( sc2, sc3 ); + __m256i sm = _mm256_and_si256( sm0, sm1 ); + + const int64_t solid0 = 1 - _mm_testc_si128( _mm256_castsi256_si128( sm ), _mm_set1_epi32( -1 ) ); + const int64_t solid1 = 1 - _mm_testc_si128( _mm256_extracti128_si256( sm, 1 ), _mm_set1_epi32( -1 ) ); + + if( solid0 + solid1 == 0 ) + { + const auto c0 = uint64_t( to565( src[0], src[1], src[2] ) ) << 16; + const auto c1 = uint64_t( to565( src[16], src[17], src[18] ) ) << 16; + memcpy( dst, &c0, 8 ); + memcpy( dst+8, &c1, 8 ); + dst += 16; + return; + } + + __m256i amask = _mm256_set1_epi32( 0xFFFFFF ); + px0 = _mm256_and_si256( px0, amask ); + px1 = _mm256_and_si256( px1, amask ); + px2 = _mm256_and_si256( px2, amask ); + px3 = _mm256_and_si256( px3, amask ); + + __m256i min0 = _mm256_min_epu8( px0, px1 ); + __m256i min1 = _mm256_min_epu8( px2, px3 ); + __m256i min2 = _mm256_min_epu8( min0, min1 ); + + __m256i max0 = _mm256_max_epu8( px0, px1 ); + __m256i max1 = _mm256_max_epu8( px2, px3 ); + __m256i max2 = _mm256_max_epu8( max0, max1 ); + + __m256i min3 = _mm256_shuffle_epi32( min2, _MM_SHUFFLE( 2, 3, 0, 1 ) ); + __m256i max3 = _mm256_shuffle_epi32( max2, _MM_SHUFFLE( 2, 3, 0, 1 ) ); + __m256i min4 = _mm256_min_epu8( min2, min3 ); + __m256i max4 = _mm256_max_epu8( max2, max3 ); + + __m256i min5 = _mm256_shuffle_epi32( min4, _MM_SHUFFLE( 0, 0, 2, 2 ) ); + __m256i max5 = _mm256_shuffle_epi32( max4, _MM_SHUFFLE( 0, 0, 2, 2 ) ); + __m256i rmin = _mm256_min_epu8( min4, min5 ); + __m256i rmax = _mm256_max_epu8( max4, max5 ); + + __m256i range1 = _mm256_subs_epu8( rmax, rmin ); + __m256i range2 = _mm256_sad_epu8( rmax, rmin ); + + uint16_t vrange0 = DivTable[_mm256_cvtsi256_si32( range2 ) >> 1]; + uint16_t vrange1 = DivTable[_mm256_extract_epi16( range2, 8 ) >> 1]; + __m256i range00 = _mm256_set1_epi16( vrange0 ); + __m256i range = _mm256_inserti128_si256( range00, _mm_set1_epi16( vrange1 ), 1 ); + + __m256i inset1 = _mm256_srli_epi16( range1, 4 ); + __m256i inset = _mm256_and_si256( inset1, _mm256_set1_epi8( 0xF ) ); + __m256i min = _mm256_adds_epu8( rmin, inset ); + __m256i max = _mm256_subs_epu8( rmax, inset ); + + __m256i c0 = _mm256_subs_epu8( px0, rmin ); + __m256i c1 = _mm256_subs_epu8( px1, rmin ); + __m256i c2 = _mm256_subs_epu8( px2, rmin ); + __m256i c3 = _mm256_subs_epu8( px3, rmin ); + + __m256i is0 = _mm256_maddubs_epi16( c0, _mm256_set1_epi8( 1 ) ); + __m256i is1 = _mm256_maddubs_epi16( c1, _mm256_set1_epi8( 1 ) ); + __m256i is2 = _mm256_maddubs_epi16( c2, _mm256_set1_epi8( 1 ) ); + __m256i is3 = _mm256_maddubs_epi16( c3, _mm256_set1_epi8( 1 ) ); + + __m256i s0 = _mm256_hadd_epi16( is0, is1 ); + __m256i s1 = _mm256_hadd_epi16( is2, is3 ); + + __m256i m0 = _mm256_mulhi_epu16( s0, range ); + __m256i m1 = _mm256_mulhi_epu16( s1, range ); + + __m256i p0 = _mm256_packus_epi16( m0, m1 ); + + __m256i p1 = _mm256_or_si256( _mm256_srai_epi32( p0, 6 ), _mm256_srai_epi32( p0, 12 ) ); + __m256i p2 = _mm256_or_si256( _mm256_srai_epi32( p0, 18 ), p0 ); + __m256i p3 = _mm256_or_si256( p1, p2 ); + __m256i p =_mm256_shuffle_epi8( p3, _mm256_set1_epi32( 0x0C080400 ) ); + + __m256i mm0 = _mm256_unpacklo_epi8( _mm256_setzero_si256(), min ); + __m256i mm1 = _mm256_unpacklo_epi8( _mm256_setzero_si256(), max ); + __m256i mm2 = _mm256_unpacklo_epi64( mm1, mm0 ); + __m256i mmr = _mm256_slli_epi64( _mm256_srli_epi64( mm2, 11 ), 11 ); + __m256i mmg = _mm256_slli_epi64( _mm256_srli_epi64( mm2, 26 ), 5 ); + __m256i mmb = _mm256_srli_epi64( _mm256_slli_epi64( mm2, 16 ), 59 ); + __m256i mm3 = _mm256_or_si256( mmr, mmg ); + __m256i mm4 = _mm256_or_si256( mm3, mmb ); + __m256i mm5 = _mm256_shuffle_epi8( mm4, _mm256_set1_epi32( 0x09080100 ) ); + + __m256i d0 = _mm256_unpacklo_epi32( mm5, p ); + __m256i d1 = _mm256_permute4x64_epi64( d0, _MM_SHUFFLE( 3, 2, 2, 0 ) ); + __m128i d2 = _mm256_castsi256_si128( d1 ); + + __m128i mask = _mm_set_epi64x( 0xFFFF0000 | -solid1, 0xFFFF0000 | -solid0 ); + __m128i d3 = _mm_and_si128( d2, mask ); + _mm_storeu_si128( (__m128i*)dst, d3 ); + dst += 16; +} +#endif + +void CompressImageDxt1( const char* src, char* dst, int w, int h ) +{ + assert( (w % 4) == 0 && (h % 4) == 0 ); + +#ifdef __AVX2__ + if( w%8 == 0 ) + { + uint32_t buf[8*4]; + int i = 0; + + auto blocks = w * h / 32; + do + { + auto tmp = (char*)buf; + memcpy( tmp, src, 8*4 ); + memcpy( tmp + 8*4, src + w * 4, 8*4 ); + memcpy( tmp + 16*4, src + w * 8, 8*4 ); + memcpy( tmp + 24*4, src + w * 12, 8*4 ); + src += 8*4; + if( ++i == w/8 ) + { + src += w * 3 * 4; + i = 0; + } + + ProcessRGB_AVX( (uint8_t*)buf, dst ); + } + while( --blocks ); + } + else +#endif + { + uint32_t buf[4*4]; + int i = 0; + + auto ptr = dst; + auto blocks = w * h / 16; + do + { + auto tmp = (char*)buf; + memcpy( tmp, src, 4*4 ); + memcpy( tmp + 4*4, src + w * 4, 4*4 ); + memcpy( tmp + 8*4, src + w * 8, 4*4 ); + memcpy( tmp + 12*4, src + w * 12, 4*4 ); + src += 4*4; + if( ++i == w/4 ) + { + src += w * 3 * 4; + i = 0; + } + + const auto c = ProcessRGB( (uint8_t*)buf ); + memcpy( ptr, &c, sizeof( uint64_t ) ); + ptr += sizeof( uint64_t ); + } + while( --blocks ); + } +} + +} diff --git a/externals/tracy/public/client/TracyDxt1.hpp b/externals/tracy/public/client/TracyDxt1.hpp new file mode 100644 index 000000000..c23135427 --- /dev/null +++ b/externals/tracy/public/client/TracyDxt1.hpp @@ -0,0 +1,11 @@ +#ifndef __TRACYDXT1_HPP__ +#define __TRACYDXT1_HPP__ + +namespace tracy +{ + +void CompressImageDxt1( const char* src, char* dst, int w, int h ); + +} + +#endif diff --git a/externals/tracy/public/client/TracyFastVector.hpp b/externals/tracy/public/client/TracyFastVector.hpp new file mode 100644 index 000000000..38accc926 --- /dev/null +++ b/externals/tracy/public/client/TracyFastVector.hpp @@ -0,0 +1,118 @@ +#ifndef __TRACYFASTVECTOR_HPP__ +#define __TRACYFASTVECTOR_HPP__ + +#include +#include + +#include "../common/TracyAlloc.hpp" +#include "../common/TracyForceInline.hpp" + +namespace tracy +{ + +template +class FastVector +{ +public: + using iterator = T*; + using const_iterator = const T*; + + FastVector( size_t capacity ) + : m_ptr( (T*)tracy_malloc( sizeof( T ) * capacity ) ) + , m_write( m_ptr ) + , m_end( m_ptr + capacity ) + { + assert( capacity != 0 ); + } + + FastVector( const FastVector& ) = delete; + FastVector( FastVector&& ) = delete; + + ~FastVector() + { + tracy_free( m_ptr ); + } + + FastVector& operator=( const FastVector& ) = delete; + FastVector& operator=( FastVector&& ) = delete; + + bool empty() const { return m_ptr == m_write; } + size_t size() const { return m_write - m_ptr; } + + T* data() { return m_ptr; } + const T* data() const { return m_ptr; }; + + T* begin() { return m_ptr; } + const T* begin() const { return m_ptr; } + T* end() { return m_write; } + const T* end() const { return m_write; } + + T& front() { assert( !empty() ); return m_ptr[0]; } + const T& front() const { assert( !empty() ); return m_ptr[0]; } + + T& back() { assert( !empty() ); return m_write[-1]; } + const T& back() const { assert( !empty() ); return m_write[-1]; } + + T& operator[]( size_t idx ) { return m_ptr[idx]; } + const T& operator[]( size_t idx ) const { return m_ptr[idx]; } + + T* push_next() + { + if( m_write == m_end ) AllocMore(); + return m_write++; + } + + T* prepare_next() + { + if( m_write == m_end ) AllocMore(); + return m_write; + } + + void commit_next() + { + m_write++; + } + + void clear() + { + m_write = m_ptr; + } + + void swap( FastVector& vec ) + { + const auto ptr1 = m_ptr; + const auto ptr2 = vec.m_ptr; + const auto write1 = m_write; + const auto write2 = vec.m_write; + const auto end1 = m_end; + const auto end2 = vec.m_end; + + m_ptr = ptr2; + vec.m_ptr = ptr1; + m_write = write2; + vec.m_write = write1; + m_end = end2; + vec.m_end = end1; + } + +private: + tracy_no_inline void AllocMore() + { + const auto cap = size_t( m_end - m_ptr ) * 2; + const auto size = size_t( m_write - m_ptr ); + T* ptr = (T*)tracy_malloc( sizeof( T ) * cap ); + memcpy( ptr, m_ptr, size * sizeof( T ) ); + tracy_free_fast( m_ptr ); + m_ptr = ptr; + m_write = m_ptr + size; + m_end = m_ptr + cap; + } + + T* m_ptr; + T* m_write; + T* m_end; +}; + +} + +#endif diff --git a/externals/tracy/public/client/TracyLock.hpp b/externals/tracy/public/client/TracyLock.hpp new file mode 100644 index 000000000..d12a3c16d --- /dev/null +++ b/externals/tracy/public/client/TracyLock.hpp @@ -0,0 +1,546 @@ +#ifndef __TRACYLOCK_HPP__ +#define __TRACYLOCK_HPP__ + +#include +#include + +#include "../common/TracySystem.hpp" +#include "../common/TracyAlign.hpp" +#include "TracyProfiler.hpp" + +namespace tracy +{ + +class LockableCtx +{ +public: + tracy_force_inline LockableCtx( const SourceLocationData* srcloc ) + : m_id( GetLockCounter().fetch_add( 1, std::memory_order_relaxed ) ) +#ifdef TRACY_ON_DEMAND + , m_lockCount( 0 ) + , m_active( false ) +#endif + { + assert( m_id != (std::numeric_limits::max)() ); + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockAnnounce ); + MemWrite( &item->lockAnnounce.id, m_id ); + MemWrite( &item->lockAnnounce.time, Profiler::GetTime() ); + MemWrite( &item->lockAnnounce.lckloc, (uint64_t)srcloc ); + MemWrite( &item->lockAnnounce.type, LockType::Lockable ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + + LockableCtx( const LockableCtx& ) = delete; + LockableCtx& operator=( const LockableCtx& ) = delete; + + tracy_force_inline ~LockableCtx() + { + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockTerminate ); + MemWrite( &item->lockTerminate.id, m_id ); + MemWrite( &item->lockTerminate.time, Profiler::GetTime() ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + + tracy_force_inline bool BeforeLock() + { +#ifdef TRACY_ON_DEMAND + bool queue = false; + const auto locks = m_lockCount.fetch_add( 1, std::memory_order_relaxed ); + const auto active = m_active.load( std::memory_order_relaxed ); + if( locks == 0 || active ) + { + const bool connected = GetProfiler().IsConnected(); + if( active != connected ) m_active.store( connected, std::memory_order_relaxed ); + if( connected ) queue = true; + } + if( !queue ) return false; +#endif + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockWait ); + MemWrite( &item->lockWait.thread, GetThreadHandle() ); + MemWrite( &item->lockWait.id, m_id ); + MemWrite( &item->lockWait.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + return true; + } + + tracy_force_inline void AfterLock() + { + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockObtain ); + MemWrite( &item->lockObtain.thread, GetThreadHandle() ); + MemWrite( &item->lockObtain.id, m_id ); + MemWrite( &item->lockObtain.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline void AfterUnlock() + { +#ifdef TRACY_ON_DEMAND + m_lockCount.fetch_sub( 1, std::memory_order_relaxed ); + if( !m_active.load( std::memory_order_relaxed ) ) return; + if( !GetProfiler().IsConnected() ) + { + m_active.store( false, std::memory_order_relaxed ); + return; + } +#endif + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockRelease ); + MemWrite( &item->lockRelease.id, m_id ); + MemWrite( &item->lockRelease.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline void AfterTryLock( bool acquired ) + { +#ifdef TRACY_ON_DEMAND + if( !acquired ) return; + + bool queue = false; + const auto locks = m_lockCount.fetch_add( 1, std::memory_order_relaxed ); + const auto active = m_active.load( std::memory_order_relaxed ); + if( locks == 0 || active ) + { + const bool connected = GetProfiler().IsConnected(); + if( active != connected ) m_active.store( connected, std::memory_order_relaxed ); + if( connected ) queue = true; + } + if( !queue ) return; +#endif + + if( acquired ) + { + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockObtain ); + MemWrite( &item->lockObtain.thread, GetThreadHandle() ); + MemWrite( &item->lockObtain.id, m_id ); + MemWrite( &item->lockObtain.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + } + } + + tracy_force_inline void Mark( const SourceLocationData* srcloc ) + { +#ifdef TRACY_ON_DEMAND + const auto active = m_active.load( std::memory_order_relaxed ); + if( !active ) return; + const auto connected = GetProfiler().IsConnected(); + if( !connected ) + { + if( active ) m_active.store( false, std::memory_order_relaxed ); + return; + } +#endif + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockMark ); + MemWrite( &item->lockMark.thread, GetThreadHandle() ); + MemWrite( &item->lockMark.id, m_id ); + MemWrite( &item->lockMark.srcloc, (uint64_t)srcloc ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline void CustomName( const char* name, size_t size ) + { + assert( size < (std::numeric_limits::max)() ); + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, name, size ); + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockName ); + MemWrite( &item->lockNameFat.id, m_id ); + MemWrite( &item->lockNameFat.name, (uint64_t)ptr ); + MemWrite( &item->lockNameFat.size, (uint16_t)size ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + +private: + uint32_t m_id; + +#ifdef TRACY_ON_DEMAND + std::atomic m_lockCount; + std::atomic m_active; +#endif +}; + +template +class Lockable +{ +public: + tracy_force_inline Lockable( const SourceLocationData* srcloc ) + : m_ctx( srcloc ) + { + } + + Lockable( const Lockable& ) = delete; + Lockable& operator=( const Lockable& ) = delete; + + tracy_force_inline void lock() + { + const auto runAfter = m_ctx.BeforeLock(); + m_lockable.lock(); + if( runAfter ) m_ctx.AfterLock(); + } + + tracy_force_inline void unlock() + { + m_lockable.unlock(); + m_ctx.AfterUnlock(); + } + + tracy_force_inline bool try_lock() + { + const auto acquired = m_lockable.try_lock(); + m_ctx.AfterTryLock( acquired ); + return acquired; + } + + tracy_force_inline void Mark( const SourceLocationData* srcloc ) + { + m_ctx.Mark( srcloc ); + } + + tracy_force_inline void CustomName( const char* name, size_t size ) + { + m_ctx.CustomName( name, size ); + } + +private: + T m_lockable; + LockableCtx m_ctx; +}; + + +class SharedLockableCtx +{ +public: + tracy_force_inline SharedLockableCtx( const SourceLocationData* srcloc ) + : m_id( GetLockCounter().fetch_add( 1, std::memory_order_relaxed ) ) +#ifdef TRACY_ON_DEMAND + , m_lockCount( 0 ) + , m_active( false ) +#endif + { + assert( m_id != (std::numeric_limits::max)() ); + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockAnnounce ); + MemWrite( &item->lockAnnounce.id, m_id ); + MemWrite( &item->lockAnnounce.time, Profiler::GetTime() ); + MemWrite( &item->lockAnnounce.lckloc, (uint64_t)srcloc ); + MemWrite( &item->lockAnnounce.type, LockType::SharedLockable ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + + SharedLockableCtx( const SharedLockableCtx& ) = delete; + SharedLockableCtx& operator=( const SharedLockableCtx& ) = delete; + + tracy_force_inline ~SharedLockableCtx() + { + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockTerminate ); + MemWrite( &item->lockTerminate.id, m_id ); + MemWrite( &item->lockTerminate.time, Profiler::GetTime() ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + + tracy_force_inline bool BeforeLock() + { +#ifdef TRACY_ON_DEMAND + bool queue = false; + const auto locks = m_lockCount.fetch_add( 1, std::memory_order_relaxed ); + const auto active = m_active.load( std::memory_order_relaxed ); + if( locks == 0 || active ) + { + const bool connected = GetProfiler().IsConnected(); + if( active != connected ) m_active.store( connected, std::memory_order_relaxed ); + if( connected ) queue = true; + } + if( !queue ) return false; +#endif + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockWait ); + MemWrite( &item->lockWait.thread, GetThreadHandle() ); + MemWrite( &item->lockWait.id, m_id ); + MemWrite( &item->lockWait.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + return true; + } + + tracy_force_inline void AfterLock() + { + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockObtain ); + MemWrite( &item->lockObtain.thread, GetThreadHandle() ); + MemWrite( &item->lockObtain.id, m_id ); + MemWrite( &item->lockObtain.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline void AfterUnlock() + { +#ifdef TRACY_ON_DEMAND + m_lockCount.fetch_sub( 1, std::memory_order_relaxed ); + if( !m_active.load( std::memory_order_relaxed ) ) return; + if( !GetProfiler().IsConnected() ) + { + m_active.store( false, std::memory_order_relaxed ); + return; + } +#endif + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockRelease ); + MemWrite( &item->lockRelease.id, m_id ); + MemWrite( &item->lockRelease.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline void AfterTryLock( bool acquired ) + { +#ifdef TRACY_ON_DEMAND + if( !acquired ) return; + + bool queue = false; + const auto locks = m_lockCount.fetch_add( 1, std::memory_order_relaxed ); + const auto active = m_active.load( std::memory_order_relaxed ); + if( locks == 0 || active ) + { + const bool connected = GetProfiler().IsConnected(); + if( active != connected ) m_active.store( connected, std::memory_order_relaxed ); + if( connected ) queue = true; + } + if( !queue ) return; +#endif + + if( acquired ) + { + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockObtain ); + MemWrite( &item->lockObtain.thread, GetThreadHandle() ); + MemWrite( &item->lockObtain.id, m_id ); + MemWrite( &item->lockObtain.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + } + } + + tracy_force_inline bool BeforeLockShared() + { +#ifdef TRACY_ON_DEMAND + bool queue = false; + const auto locks = m_lockCount.fetch_add( 1, std::memory_order_relaxed ); + const auto active = m_active.load( std::memory_order_relaxed ); + if( locks == 0 || active ) + { + const bool connected = GetProfiler().IsConnected(); + if( active != connected ) m_active.store( connected, std::memory_order_relaxed ); + if( connected ) queue = true; + } + if( !queue ) return false; +#endif + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockSharedWait ); + MemWrite( &item->lockWait.thread, GetThreadHandle() ); + MemWrite( &item->lockWait.id, m_id ); + MemWrite( &item->lockWait.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + return true; + } + + tracy_force_inline void AfterLockShared() + { + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockSharedObtain ); + MemWrite( &item->lockObtain.thread, GetThreadHandle() ); + MemWrite( &item->lockObtain.id, m_id ); + MemWrite( &item->lockObtain.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline void AfterUnlockShared() + { +#ifdef TRACY_ON_DEMAND + m_lockCount.fetch_sub( 1, std::memory_order_relaxed ); + if( !m_active.load( std::memory_order_relaxed ) ) return; + if( !GetProfiler().IsConnected() ) + { + m_active.store( false, std::memory_order_relaxed ); + return; + } +#endif + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockSharedRelease ); + MemWrite( &item->lockReleaseShared.thread, GetThreadHandle() ); + MemWrite( &item->lockReleaseShared.id, m_id ); + MemWrite( &item->lockReleaseShared.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline void AfterTryLockShared( bool acquired ) + { +#ifdef TRACY_ON_DEMAND + if( !acquired ) return; + + bool queue = false; + const auto locks = m_lockCount.fetch_add( 1, std::memory_order_relaxed ); + const auto active = m_active.load( std::memory_order_relaxed ); + if( locks == 0 || active ) + { + const bool connected = GetProfiler().IsConnected(); + if( active != connected ) m_active.store( connected, std::memory_order_relaxed ); + if( connected ) queue = true; + } + if( !queue ) return; +#endif + + if( acquired ) + { + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockSharedObtain ); + MemWrite( &item->lockObtain.thread, GetThreadHandle() ); + MemWrite( &item->lockObtain.id, m_id ); + MemWrite( &item->lockObtain.time, Profiler::GetTime() ); + Profiler::QueueSerialFinish(); + } + } + + tracy_force_inline void Mark( const SourceLocationData* srcloc ) + { +#ifdef TRACY_ON_DEMAND + const auto active = m_active.load( std::memory_order_relaxed ); + if( !active ) return; + const auto connected = GetProfiler().IsConnected(); + if( !connected ) + { + if( active ) m_active.store( false, std::memory_order_relaxed ); + return; + } +#endif + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockMark ); + MemWrite( &item->lockMark.thread, GetThreadHandle() ); + MemWrite( &item->lockMark.id, m_id ); + MemWrite( &item->lockMark.srcloc, (uint64_t)srcloc ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline void CustomName( const char* name, size_t size ) + { + assert( size < (std::numeric_limits::max)() ); + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, name, size ); + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::LockName ); + MemWrite( &item->lockNameFat.id, m_id ); + MemWrite( &item->lockNameFat.name, (uint64_t)ptr ); + MemWrite( &item->lockNameFat.size, (uint16_t)size ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + +private: + uint32_t m_id; + +#ifdef TRACY_ON_DEMAND + std::atomic m_lockCount; + std::atomic m_active; +#endif +}; + +template +class SharedLockable +{ +public: + tracy_force_inline SharedLockable( const SourceLocationData* srcloc ) + : m_ctx( srcloc ) + { + } + + SharedLockable( const SharedLockable& ) = delete; + SharedLockable& operator=( const SharedLockable& ) = delete; + + tracy_force_inline void lock() + { + const auto runAfter = m_ctx.BeforeLock(); + m_lockable.lock(); + if( runAfter ) m_ctx.AfterLock(); + } + + tracy_force_inline void unlock() + { + m_lockable.unlock(); + m_ctx.AfterUnlock(); + } + + tracy_force_inline bool try_lock() + { + const auto acquired = m_lockable.try_lock(); + m_ctx.AfterTryLock( acquired ); + return acquired; + } + + tracy_force_inline void lock_shared() + { + const auto runAfter = m_ctx.BeforeLockShared(); + m_lockable.lock_shared(); + if( runAfter ) m_ctx.AfterLockShared(); + } + + tracy_force_inline void unlock_shared() + { + m_lockable.unlock_shared(); + m_ctx.AfterUnlockShared(); + } + + tracy_force_inline bool try_lock_shared() + { + const auto acquired = m_lockable.try_lock_shared(); + m_ctx.AfterTryLockShared( acquired ); + return acquired; + } + + tracy_force_inline void Mark( const SourceLocationData* srcloc ) + { + m_ctx.Mark( srcloc ); + } + + tracy_force_inline void CustomName( const char* name, size_t size ) + { + m_ctx.CustomName( name, size ); + } + +private: + T m_lockable; + SharedLockableCtx m_ctx; +}; + + +} + +#endif diff --git a/externals/tracy/public/client/TracyOverride.cpp b/externals/tracy/public/client/TracyOverride.cpp new file mode 100644 index 000000000..591508a7f --- /dev/null +++ b/externals/tracy/public/client/TracyOverride.cpp @@ -0,0 +1,26 @@ +#ifdef TRACY_ENABLE +# ifdef __linux__ +# include "TracyDebug.hpp" +# ifdef TRACY_VERBOSE +# include +# include +# endif + +extern "C" int dlclose( void* hnd ) +{ +#ifdef TRACY_VERBOSE + struct link_map* lm; + if( dlinfo( hnd, RTLD_DI_LINKMAP, &lm ) == 0 ) + { + TracyDebug( "Overriding dlclose for %s\n", lm->l_name ); + } + else + { + TracyDebug( "Overriding dlclose for unknown object (%s)\n", dlerror() ); + } +#endif + return 0; +} + +# endif +#endif diff --git a/externals/tracy/public/client/TracyProfiler.cpp b/externals/tracy/public/client/TracyProfiler.cpp new file mode 100644 index 000000000..2e278daa9 --- /dev/null +++ b/externals/tracy/public/client/TracyProfiler.cpp @@ -0,0 +1,4530 @@ +#ifdef TRACY_ENABLE + +#ifdef _WIN32 +# ifndef NOMINMAX +# define NOMINMAX +# endif +# include +# include +# include +# include +# include +# include "../common/TracyUwp.hpp" +#else +# include +# include +#endif + +#ifdef _GNU_SOURCE +# include +#endif + +#ifdef __linux__ +# include +# include +# include +# include +#endif + +#if defined __APPLE__ || defined BSD +# include +# include +#endif + +#if defined __APPLE__ +# include "TargetConditionals.h" +# include +#endif + +#ifdef __ANDROID__ +# include +# include +# include +# include +# include +# include +#endif + +#ifdef __QNX__ +# include +# include +# include +# include +# include +#endif + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "../common/TracyAlign.hpp" +#include "../common/TracyAlloc.hpp" +#include "../common/TracySocket.hpp" +#include "../common/TracySystem.hpp" +#include "../common/TracyYield.hpp" +#include "../common/tracy_lz4.hpp" +#include "tracy_rpmalloc.hpp" +#include "TracyCallstack.hpp" +#include "TracyDebug.hpp" +#include "TracyDxt1.hpp" +#include "TracyScoped.hpp" +#include "TracyProfiler.hpp" +#include "TracyThread.hpp" +#include "TracyArmCpuTable.hpp" +#include "TracySysTrace.hpp" +#include "../tracy/TracyC.h" + +#ifdef TRACY_PORT +# ifndef TRACY_DATA_PORT +# define TRACY_DATA_PORT TRACY_PORT +# endif +# ifndef TRACY_BROADCAST_PORT +# define TRACY_BROADCAST_PORT TRACY_PORT +# endif +#endif + +#ifdef __APPLE__ +# ifndef TRACY_DELAYED_INIT +# define TRACY_DELAYED_INIT +# endif +#else +# ifdef __GNUC__ +# define init_order( val ) __attribute__ ((init_priority(val))) +# else +# define init_order(x) +# endif +#endif + +#if defined _WIN32 +# include +extern "C" typedef LONG (WINAPI *t_RtlGetVersion)( PRTL_OSVERSIONINFOW ); +extern "C" typedef BOOL (WINAPI *t_GetLogicalProcessorInformationEx)( LOGICAL_PROCESSOR_RELATIONSHIP, PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX, PDWORD ); +#else +# include +# include +#endif +#if defined __linux__ +# include +# include +#endif + +#if !defined _WIN32 && ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) +# include "TracyCpuid.hpp" +#endif + +#if !( ( defined _WIN32 && _WIN32_WINNT >= _WIN32_WINNT_VISTA ) || defined __linux__ ) +# include +#endif + +#ifdef __QNX__ +extern char* __progname; +#endif + +namespace tracy +{ + +#ifdef __ANDROID__ +// Implementation helpers of EnsureReadable(address). +// This is so far only needed on Android, where it is common for libraries to be mapped +// with only executable, not readable, permissions. Typical example (line from /proc/self/maps): +/* +746b63b000-746b6dc000 --xp 00042000 07:48 35 /apex/com.android.runtime/lib64/bionic/libc.so +*/ +// See https://github.com/wolfpld/tracy/issues/125 . +// To work around this, we parse /proc/self/maps and we use mprotect to set read permissions +// on any mappings that contain symbols addresses hit by HandleSymbolCodeQuery. + +namespace { +// Holds some information about a single memory mapping. +struct MappingInfo { + // Start of address range. Inclusive. + uintptr_t start_address; + // End of address range. Exclusive, so the mapping is the half-open interval + // [start, end) and its length in bytes is `end - start`. As in /proc/self/maps. + uintptr_t end_address; + // Read/Write/Executable permissions. + bool perm_r, perm_w, perm_x; +}; +} // anonymous namespace + + // Internal implementation helper for LookUpMapping(address). + // + // Parses /proc/self/maps returning a vector. + // /proc/self/maps is assumed to be sorted by ascending address, so the resulting + // vector is sorted by ascending address too. +static std::vector ParseMappings() +{ + std::vector result; + FILE* file = fopen( "/proc/self/maps", "r" ); + if( !file ) return result; + char line[1024]; + while( fgets( line, sizeof( line ), file ) ) + { + uintptr_t start_addr; + uintptr_t end_addr; +#if defined(__LP64__) + if( sscanf( line, "%lx-%lx", &start_addr, &end_addr ) != 2 ) continue; +#else + if (sscanf( line, "%dx-%dx", &start_addr, &end_addr ) != 2 ) continue; +#endif + char* first_space = strchr( line, ' ' ); + if( !first_space ) continue; + char* perm = first_space + 1; + char* second_space = strchr( perm, ' ' ); + if( !second_space || second_space - perm != 4 ) continue; + result.emplace_back(); + auto& mapping = result.back(); + mapping.start_address = start_addr; + mapping.end_address = end_addr; + mapping.perm_r = perm[0] == 'r'; + mapping.perm_w = perm[1] == 'w'; + mapping.perm_x = perm[2] == 'x'; + } + fclose( file ); + return result; +} + +// Internal implementation helper for LookUpMapping(address). +// +// Takes as input an `address` and a known vector `mappings`, assumed to be +// sorted by increasing addresses, as /proc/self/maps seems to be. +// Returns a pointer to the MappingInfo describing the mapping that this +// address belongs to, or nullptr if the address isn't in `mappings`. +static MappingInfo* LookUpMapping(std::vector& mappings, uintptr_t address) +{ + // Comparison function for std::lower_bound. Returns true if all addresses in `m1` + // are lower than `addr`. + auto Compare = []( const MappingInfo& m1, uintptr_t addr ) { + // '<=' because the address ranges are half-open intervals, [start, end). + return m1.end_address <= addr; + }; + auto iter = std::lower_bound( mappings.begin(), mappings.end(), address, Compare ); + if( iter == mappings.end() || iter->start_address > address) { + return nullptr; + } + return &*iter; +} + +// Internal implementation helper for EnsureReadable(address). +// +// Takes as input an `address` and returns a pointer to a MappingInfo +// describing the mapping that this address belongs to, or nullptr if +// the address isn't in any known mapping. +// +// This function is stateful and not reentrant (assumes to be called from +// only one thread). It holds a vector of mappings parsed from /proc/self/maps. +// +// Attempts to react to mappings changes by re-parsing /proc/self/maps. +static MappingInfo* LookUpMapping(uintptr_t address) +{ + // Static state managed by this function. Not constant, we mutate that state as + // we turn some mappings readable. Initially parsed once here, updated as needed below. + static std::vector s_mappings = ParseMappings(); + MappingInfo* mapping = LookUpMapping( s_mappings, address ); + if( mapping ) return mapping; + + // This address isn't in any known mapping. Try parsing again, maybe + // mappings changed. + s_mappings = ParseMappings(); + return LookUpMapping( s_mappings, address ); +} + +// Internal implementation helper for EnsureReadable(address). +// +// Attempts to make the specified `mapping` readable if it isn't already. +// Returns true if and only if the mapping is readable. +static bool EnsureReadable( MappingInfo& mapping ) +{ + if( mapping.perm_r ) + { + // The mapping is already readable. + return true; + } + int prot = PROT_READ; + if( mapping.perm_w ) prot |= PROT_WRITE; + if( mapping.perm_x ) prot |= PROT_EXEC; + if( mprotect( reinterpret_cast( mapping.start_address ), + mapping.end_address - mapping.start_address, prot ) == -1 ) + { + // Failed to make the mapping readable. Shouldn't happen, hasn't + // been observed yet. If it happened in practice, we should consider + // adding a bool to MappingInfo to track this to avoid retrying mprotect + // everytime on such mappings. + return false; + } + // The mapping is now readable. Update `mapping` so the next call will be fast. + mapping.perm_r = true; + return true; +} + +// Attempts to set the read permission on the entire mapping containing the +// specified address. Returns true if and only if the mapping is now readable. +static bool EnsureReadable( uintptr_t address ) +{ + MappingInfo* mapping = LookUpMapping(address); + return mapping && EnsureReadable( *mapping ); +} + +#endif // defined __ANDROID__ + +#ifndef TRACY_DELAYED_INIT + +struct InitTimeWrapper +{ + int64_t val; +}; + +struct ProducerWrapper +{ + tracy::moodycamel::ConcurrentQueue::ExplicitProducer* ptr; +}; + +struct ThreadHandleWrapper +{ + uint32_t val; +}; +#endif + + +#if defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 +static inline void CpuId( uint32_t* regs, uint32_t leaf ) +{ + memset(regs, 0, sizeof(uint32_t) * 4); +#if defined _WIN32 + __cpuidex( (int*)regs, leaf, 0 ); +#else + __get_cpuid( leaf, regs, regs+1, regs+2, regs+3 ); +#endif +} + +static void InitFailure( const char* msg ) +{ +#if defined _WIN32 + bool hasConsole = false; + bool reopen = false; + const auto attached = AttachConsole( ATTACH_PARENT_PROCESS ); + if( attached ) + { + hasConsole = true; + reopen = true; + } + else + { + const auto err = GetLastError(); + if( err == ERROR_ACCESS_DENIED ) + { + hasConsole = true; + } + } + if( hasConsole ) + { + fprintf( stderr, "Tracy Profiler initialization failure: %s\n", msg ); + if( reopen ) + { + freopen( "CONOUT$", "w", stderr ); + fprintf( stderr, "Tracy Profiler initialization failure: %s\n", msg ); + } + } + else + { +# ifndef TRACY_UWP + MessageBoxA( nullptr, msg, "Tracy Profiler initialization failure", MB_ICONSTOP ); +# endif + } +#else + fprintf( stderr, "Tracy Profiler initialization failure: %s\n", msg ); +#endif + exit( 1 ); +} + +static bool CheckHardwareSupportsInvariantTSC() +{ + const char* noCheck = GetEnvVar( "TRACY_NO_INVARIANT_CHECK" ); + if( noCheck && noCheck[0] == '1' ) return true; + + uint32_t regs[4]; + CpuId( regs, 1 ); + if( !( regs[3] & ( 1 << 4 ) ) ) + { +#if !defined TRACY_TIMER_QPC && !defined TRACY_TIMER_FALLBACK + InitFailure( "CPU doesn't support RDTSC instruction." ); +#else + return false; +#endif + } + CpuId( regs, 0x80000007 ); + if( regs[3] & ( 1 << 8 ) ) return true; + + return false; +} + +#if defined TRACY_TIMER_FALLBACK && defined TRACY_HW_TIMER +bool HardwareSupportsInvariantTSC() +{ + static bool cachedResult = CheckHardwareSupportsInvariantTSC(); + return cachedResult; +} +#endif + +static int64_t SetupHwTimer() +{ +#if !defined TRACY_TIMER_QPC && !defined TRACY_TIMER_FALLBACK + if( !CheckHardwareSupportsInvariantTSC() ) + { +#if defined _WIN32 + InitFailure( "CPU doesn't support invariant TSC.\nDefine TRACY_NO_INVARIANT_CHECK=1 to ignore this error, *if you know what you are doing*.\nAlternatively you may rebuild the application with the TRACY_TIMER_QPC or TRACY_TIMER_FALLBACK define to use lower resolution timer." ); +#else + InitFailure( "CPU doesn't support invariant TSC.\nDefine TRACY_NO_INVARIANT_CHECK=1 to ignore this error, *if you know what you are doing*.\nAlternatively you may rebuild the application with the TRACY_TIMER_FALLBACK define to use lower resolution timer." ); +#endif + } +#endif + + return Profiler::GetTime(); +} +#else +static int64_t SetupHwTimer() +{ + return Profiler::GetTime(); +} +#endif + +static const char* GetProcessName() +{ + const char* processName = "unknown"; +#ifdef _WIN32 + static char buf[_MAX_PATH]; + GetModuleFileNameA( nullptr, buf, _MAX_PATH ); + const char* ptr = buf; + while( *ptr != '\0' ) ptr++; + while( ptr > buf && *ptr != '\\' && *ptr != '/' ) ptr--; + if( ptr > buf ) ptr++; + processName = ptr; +#elif defined __ANDROID__ +# if __ANDROID_API__ >= 21 + auto buf = getprogname(); + if( buf ) processName = buf; +# endif +#elif defined __linux__ && defined _GNU_SOURCE + if( program_invocation_short_name ) processName = program_invocation_short_name; +#elif defined __APPLE__ || defined BSD + auto buf = getprogname(); + if( buf ) processName = buf; +#elif defined __QNX__ + processName = __progname; +#endif + return processName; +} + +static const char* GetProcessExecutablePath() +{ +#ifdef _WIN32 + static char buf[_MAX_PATH]; + GetModuleFileNameA( nullptr, buf, _MAX_PATH ); + return buf; +#elif defined __ANDROID__ + return nullptr; +#elif defined __linux__ && defined _GNU_SOURCE + return program_invocation_name; +#elif defined __APPLE__ + static char buf[1024]; + uint32_t size = 1024; + _NSGetExecutablePath( buf, &size ); + return buf; +#elif defined __DragonFly__ + static char buf[1024]; + readlink( "/proc/curproc/file", buf, 1024 ); + return buf; +#elif defined __FreeBSD__ + static char buf[1024]; + int mib[4]; + mib[0] = CTL_KERN; + mib[1] = KERN_PROC; + mib[2] = KERN_PROC_PATHNAME; + mib[3] = -1; + size_t cb = 1024; + sysctl( mib, 4, buf, &cb, nullptr, 0 ); + return buf; +#elif defined __NetBSD__ + static char buf[1024]; + readlink( "/proc/curproc/exe", buf, 1024 ); + return buf; +#elif defined __QNX__ + static char buf[_PC_PATH_MAX + 1]; + _cmdname(buf); + return buf; +#else + return nullptr; +#endif +} + +#if defined __linux__ && defined __ARM_ARCH +static uint32_t GetHex( char*& ptr, int skip ) +{ + uint32_t ret; + ptr += skip; + char* end; + if( ptr[0] == '0' && ptr[1] == 'x' ) + { + ptr += 2; + ret = strtol( ptr, &end, 16 ); + } + else + { + ret = strtol( ptr, &end, 10 ); + } + ptr = end; + return ret; +} +#endif + +static const char* GetHostInfo() +{ + static char buf[1024]; + auto ptr = buf; +#if defined _WIN32 +# ifdef TRACY_UWP + auto GetVersion = &::GetVersionEx; +# else + auto GetVersion = (t_RtlGetVersion)GetProcAddress( GetModuleHandleA( "ntdll.dll" ), "RtlGetVersion" ); +# endif + if( !GetVersion ) + { +# ifdef __MINGW32__ + ptr += sprintf( ptr, "OS: Windows (MingW)\n" ); +# else + ptr += sprintf( ptr, "OS: Windows\n" ); +# endif + } + else + { + RTL_OSVERSIONINFOW ver = { sizeof( RTL_OSVERSIONINFOW ) }; + GetVersion( &ver ); + +# ifdef __MINGW32__ + ptr += sprintf( ptr, "OS: Windows %i.%i.%i (MingW)\n", (int)ver.dwMajorVersion, (int)ver.dwMinorVersion, (int)ver.dwBuildNumber ); +# else + ptr += sprintf( ptr, "OS: Windows %i.%i.%i\n", ver.dwMajorVersion, ver.dwMinorVersion, ver.dwBuildNumber ); +# endif + } +#elif defined __linux__ + struct utsname utsName; + uname( &utsName ); +# if defined __ANDROID__ + ptr += sprintf( ptr, "OS: Linux %s (Android)\n", utsName.release ); +# else + ptr += sprintf( ptr, "OS: Linux %s\n", utsName.release ); +# endif +#elif defined __APPLE__ +# if TARGET_OS_IPHONE == 1 + ptr += sprintf( ptr, "OS: Darwin (iOS)\n" ); +# elif TARGET_OS_MAC == 1 + ptr += sprintf( ptr, "OS: Darwin (OSX)\n" ); +# else + ptr += sprintf( ptr, "OS: Darwin (unknown)\n" ); +# endif +#elif defined __DragonFly__ + ptr += sprintf( ptr, "OS: BSD (DragonFly)\n" ); +#elif defined __FreeBSD__ + ptr += sprintf( ptr, "OS: BSD (FreeBSD)\n" ); +#elif defined __NetBSD__ + ptr += sprintf( ptr, "OS: BSD (NetBSD)\n" ); +#elif defined __OpenBSD__ + ptr += sprintf( ptr, "OS: BSD (OpenBSD)\n" ); +#elif defined __QNX__ + ptr += sprintf( ptr, "OS: QNX\n" ); +#else + ptr += sprintf( ptr, "OS: unknown\n" ); +#endif + +#if defined _MSC_VER +# if defined __clang__ + ptr += sprintf( ptr, "Compiler: MSVC clang-cl %i.%i.%i\n", __clang_major__, __clang_minor__, __clang_patchlevel__ ); +# else + ptr += sprintf( ptr, "Compiler: MSVC %i\n", _MSC_VER ); +# endif +#elif defined __clang__ + ptr += sprintf( ptr, "Compiler: clang %i.%i.%i\n", __clang_major__, __clang_minor__, __clang_patchlevel__ ); +#elif defined __GNUC__ + ptr += sprintf( ptr, "Compiler: gcc %i.%i.%i\n", __GNUC__, __GNUC_MINOR__, __GNUC_PATCHLEVEL__ ); +#else + ptr += sprintf( ptr, "Compiler: unknown\n" ); +#endif + +#if defined _WIN32 + InitWinSock(); + + char hostname[512]; + gethostname( hostname, 512 ); + +# ifdef TRACY_UWP + const char* user = ""; +# else + DWORD userSz = UNLEN+1; + char user[UNLEN+1]; + GetUserNameA( user, &userSz ); +# endif + + ptr += sprintf( ptr, "User: %s@%s\n", user, hostname ); +#else + char hostname[_POSIX_HOST_NAME_MAX]{}; + char user[_POSIX_LOGIN_NAME_MAX]{}; + + gethostname( hostname, _POSIX_HOST_NAME_MAX ); +# if defined __ANDROID__ + const auto login = getlogin(); + if( login ) + { + strcpy( user, login ); + } + else + { + memcpy( user, "(?)", 4 ); + } +# else + getlogin_r( user, _POSIX_LOGIN_NAME_MAX ); +# endif + + ptr += sprintf( ptr, "User: %s@%s\n", user, hostname ); +#endif + +#if defined __i386 || defined _M_IX86 + ptr += sprintf( ptr, "Arch: x86\n" ); +#elif defined __x86_64__ || defined _M_X64 + ptr += sprintf( ptr, "Arch: x64\n" ); +#elif defined __aarch64__ + ptr += sprintf( ptr, "Arch: ARM64\n" ); +#elif defined __ARM_ARCH + ptr += sprintf( ptr, "Arch: ARM\n" ); +#else + ptr += sprintf( ptr, "Arch: unknown\n" ); +#endif + +#if defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 + uint32_t regs[4]; + char cpuModel[4*4*3+1] = {}; + auto modelPtr = cpuModel; + for( uint32_t i=0x80000002; i<0x80000005; ++i ) + { + CpuId( regs, i ); + memcpy( modelPtr, regs, sizeof( regs ) ); modelPtr += sizeof( regs ); + } + + ptr += sprintf( ptr, "CPU: %s\n", cpuModel ); +#elif defined __linux__ && defined __ARM_ARCH + bool cpuFound = false; + FILE* fcpuinfo = fopen( "/proc/cpuinfo", "rb" ); + if( fcpuinfo ) + { + enum { BufSize = 4*1024 }; + char buf[BufSize]; + const auto sz = fread( buf, 1, BufSize, fcpuinfo ); + fclose( fcpuinfo ); + const auto end = buf + sz; + auto cptr = buf; + + uint32_t impl = 0; + uint32_t var = 0; + uint32_t part = 0; + uint32_t rev = 0; + + while( end - cptr > 20 ) + { + while( end - cptr > 20 && memcmp( cptr, "CPU ", 4 ) != 0 ) + { + cptr += 4; + while( end - cptr > 20 && *cptr != '\n' ) cptr++; + cptr++; + } + if( end - cptr <= 20 ) break; + cptr += 4; + if( memcmp( cptr, "implementer\t: ", 14 ) == 0 ) + { + if( impl != 0 ) break; + impl = GetHex( cptr, 14 ); + } + else if( memcmp( cptr, "variant\t: ", 10 ) == 0 ) var = GetHex( cptr, 10 ); + else if( memcmp( cptr, "part\t: ", 7 ) == 0 ) part = GetHex( cptr, 7 ); + else if( memcmp( cptr, "revision\t: ", 11 ) == 0 ) rev = GetHex( cptr, 11 ); + while( *cptr != '\n' && *cptr != '\0' ) cptr++; + cptr++; + } + + if( impl != 0 || var != 0 || part != 0 || rev != 0 ) + { + cpuFound = true; + ptr += sprintf( ptr, "CPU: %s%s r%ip%i\n", DecodeArmImplementer( impl ), DecodeArmPart( impl, part ), var, rev ); + } + } + if( !cpuFound ) + { + ptr += sprintf( ptr, "CPU: unknown\n" ); + } +#elif defined __APPLE__ && TARGET_OS_IPHONE == 1 + { + size_t sz; + sysctlbyname( "hw.machine", nullptr, &sz, nullptr, 0 ); + auto str = (char*)tracy_malloc( sz ); + sysctlbyname( "hw.machine", str, &sz, nullptr, 0 ); + ptr += sprintf( ptr, "Device: %s\n", DecodeIosDevice( str ) ); + tracy_free( str ); + } +#else + ptr += sprintf( ptr, "CPU: unknown\n" ); +#endif +#ifdef __ANDROID__ + char deviceModel[PROP_VALUE_MAX+1]; + char deviceManufacturer[PROP_VALUE_MAX+1]; + __system_property_get( "ro.product.model", deviceModel ); + __system_property_get( "ro.product.manufacturer", deviceManufacturer ); + ptr += sprintf( ptr, "Device: %s %s\n", deviceManufacturer, deviceModel ); +#endif + + ptr += sprintf( ptr, "CPU cores: %i\n", std::thread::hardware_concurrency() ); + +#if defined _WIN32 + MEMORYSTATUSEX statex; + statex.dwLength = sizeof( statex ); + GlobalMemoryStatusEx( &statex ); +# ifdef _MSC_VER + ptr += sprintf( ptr, "RAM: %I64u MB\n", statex.ullTotalPhys / 1024 / 1024 ); +# else + ptr += sprintf( ptr, "RAM: %llu MB\n", statex.ullTotalPhys / 1024 / 1024 ); +# endif +#elif defined __linux__ + struct sysinfo sysInfo; + sysinfo( &sysInfo ); + ptr += sprintf( ptr, "RAM: %lu MB\n", sysInfo.totalram / 1024 / 1024 ); +#elif defined __APPLE__ + size_t memSize; + size_t sz = sizeof( memSize ); + sysctlbyname( "hw.memsize", &memSize, &sz, nullptr, 0 ); + ptr += sprintf( ptr, "RAM: %zu MB\n", memSize / 1024 / 1024 ); +#elif defined BSD + size_t memSize; + size_t sz = sizeof( memSize ); + sysctlbyname( "hw.physmem", &memSize, &sz, nullptr, 0 ); + ptr += sprintf( ptr, "RAM: %zu MB\n", memSize / 1024 / 1024 ); +#elif defined __QNX__ + struct asinfo_entry *entries = SYSPAGE_ENTRY(asinfo); + size_t count = SYSPAGE_ENTRY_SIZE(asinfo) / sizeof(struct asinfo_entry); + char *strings = SYSPAGE_ENTRY(strings)->data; + + uint64_t memSize = 0; + size_t i; + for (i = 0; i < count; i++) { + struct asinfo_entry *entry = &entries[i]; + if (strcmp(strings + entry->name, "ram") == 0) { + memSize += entry->end - entry->start + 1; + } + } + memSize = memSize / 1024 / 1024; + ptr += sprintf( ptr, "RAM: %llu MB\n", memSize); +#else + ptr += sprintf( ptr, "RAM: unknown\n" ); +#endif + + return buf; +} + +static uint64_t GetPid() +{ +#if defined _WIN32 + return uint64_t( GetCurrentProcessId() ); +#else + return uint64_t( getpid() ); +#endif +} + +void Profiler::AckServerQuery() +{ + QueueItem item; + MemWrite( &item.hdr.type, QueueType::AckServerQueryNoop ); + NeedDataSize( QueueDataSize[(int)QueueType::AckServerQueryNoop] ); + AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::AckServerQueryNoop] ); +} + +void Profiler::AckSymbolCodeNotAvailable() +{ + QueueItem item; + MemWrite( &item.hdr.type, QueueType::AckSymbolCodeNotAvailable ); + NeedDataSize( QueueDataSize[(int)QueueType::AckSymbolCodeNotAvailable] ); + AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::AckSymbolCodeNotAvailable] ); +} + +static BroadcastMessage& GetBroadcastMessage( const char* procname, size_t pnsz, int& len, int port ) +{ + static BroadcastMessage msg; + + msg.broadcastVersion = BroadcastVersion; + msg.protocolVersion = ProtocolVersion; + msg.listenPort = port; + msg.pid = GetPid(); + + memcpy( msg.programName, procname, pnsz ); + memset( msg.programName + pnsz, 0, WelcomeMessageProgramNameSize - pnsz ); + + len = int( offsetof( BroadcastMessage, programName ) + pnsz + 1 ); + return msg; +} + +#if defined _WIN32 && !defined TRACY_UWP && !defined TRACY_NO_CRASH_HANDLER +static DWORD s_profilerThreadId = 0; +static DWORD s_symbolThreadId = 0; +static char s_crashText[1024]; + +LONG WINAPI CrashFilter( PEXCEPTION_POINTERS pExp ) +{ + if( !GetProfiler().IsConnected() ) return EXCEPTION_CONTINUE_SEARCH; + + const unsigned ec = pExp->ExceptionRecord->ExceptionCode; + auto msgPtr = s_crashText; + switch( ec ) + { + case EXCEPTION_ACCESS_VIOLATION: + msgPtr += sprintf( msgPtr, "Exception EXCEPTION_ACCESS_VIOLATION (0x%x). ", ec ); + switch( pExp->ExceptionRecord->ExceptionInformation[0] ) + { + case 0: + msgPtr += sprintf( msgPtr, "Read violation at address 0x%" PRIxPTR ".", pExp->ExceptionRecord->ExceptionInformation[1] ); + break; + case 1: + msgPtr += sprintf( msgPtr, "Write violation at address 0x%" PRIxPTR ".", pExp->ExceptionRecord->ExceptionInformation[1] ); + break; + case 8: + msgPtr += sprintf( msgPtr, "DEP violation at address 0x%" PRIxPTR ".", pExp->ExceptionRecord->ExceptionInformation[1] ); + break; + default: + break; + } + break; + case EXCEPTION_ARRAY_BOUNDS_EXCEEDED: + msgPtr += sprintf( msgPtr, "Exception EXCEPTION_ARRAY_BOUNDS_EXCEEDED (0x%x). ", ec ); + break; + case EXCEPTION_DATATYPE_MISALIGNMENT: + msgPtr += sprintf( msgPtr, "Exception EXCEPTION_DATATYPE_MISALIGNMENT (0x%x). ", ec ); + break; + case EXCEPTION_FLT_DIVIDE_BY_ZERO: + msgPtr += sprintf( msgPtr, "Exception EXCEPTION_FLT_DIVIDE_BY_ZERO (0x%x). ", ec ); + break; + case EXCEPTION_ILLEGAL_INSTRUCTION: + msgPtr += sprintf( msgPtr, "Exception EXCEPTION_ILLEGAL_INSTRUCTION (0x%x). ", ec ); + break; + case EXCEPTION_IN_PAGE_ERROR: + msgPtr += sprintf( msgPtr, "Exception EXCEPTION_IN_PAGE_ERROR (0x%x). ", ec ); + break; + case EXCEPTION_INT_DIVIDE_BY_ZERO: + msgPtr += sprintf( msgPtr, "Exception EXCEPTION_INT_DIVIDE_BY_ZERO (0x%x). ", ec ); + break; + case EXCEPTION_PRIV_INSTRUCTION: + msgPtr += sprintf( msgPtr, "Exception EXCEPTION_PRIV_INSTRUCTION (0x%x). ", ec ); + break; + case EXCEPTION_STACK_OVERFLOW: + msgPtr += sprintf( msgPtr, "Exception EXCEPTION_STACK_OVERFLOW (0x%x). ", ec ); + break; + default: + return EXCEPTION_CONTINUE_SEARCH; + } + + { + GetProfiler().SendCallstack( 60, "KiUserExceptionDispatcher" ); + + TracyQueuePrepare( QueueType::CrashReport ); + item->crashReport.time = Profiler::GetTime(); + item->crashReport.text = (uint64_t)s_crashText; + TracyQueueCommit( crashReportThread ); + } + + HANDLE h = CreateToolhelp32Snapshot( TH32CS_SNAPTHREAD, 0 ); + if( h == INVALID_HANDLE_VALUE ) return EXCEPTION_CONTINUE_SEARCH; + + THREADENTRY32 te = { sizeof( te ) }; + if( !Thread32First( h, &te ) ) + { + CloseHandle( h ); + return EXCEPTION_CONTINUE_SEARCH; + } + + const auto pid = GetCurrentProcessId(); + const auto tid = GetCurrentThreadId(); + + do + { + if( te.th32OwnerProcessID == pid && te.th32ThreadID != tid && te.th32ThreadID != s_profilerThreadId && te.th32ThreadID != s_symbolThreadId ) + { + HANDLE th = OpenThread( THREAD_SUSPEND_RESUME, FALSE, te.th32ThreadID ); + if( th != INVALID_HANDLE_VALUE ) + { + SuspendThread( th ); + CloseHandle( th ); + } + } + } + while( Thread32Next( h, &te ) ); + CloseHandle( h ); + + { + TracyLfqPrepare( QueueType::Crash ); + TracyLfqCommit; + } + + std::this_thread::sleep_for( std::chrono::milliseconds( 500 ) ); + GetProfiler().RequestShutdown(); + while( !GetProfiler().HasShutdownFinished() ) { std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) ); }; + + return EXCEPTION_CONTINUE_SEARCH; +} +#endif + +static Profiler* s_instance = nullptr; +static Thread* s_thread; +#ifndef TRACY_NO_FRAME_IMAGE +static Thread* s_compressThread; +#endif +#ifdef TRACY_HAS_CALLSTACK +static Thread* s_symbolThread; +std::atomic s_symbolThreadGone { false }; +#endif +#ifdef TRACY_HAS_SYSTEM_TRACING +static Thread* s_sysTraceThread = nullptr; +#endif + +#if defined __linux__ && !defined TRACY_NO_CRASH_HANDLER +# ifndef TRACY_CRASH_SIGNAL +# define TRACY_CRASH_SIGNAL SIGPWR +# endif + +static long s_profilerTid = 0; +static long s_symbolTid = 0; +static char s_crashText[1024]; +static std::atomic s_alreadyCrashed( false ); + +static void ThreadFreezer( int /*signal*/ ) +{ + for(;;) sleep( 1000 ); +} + +static inline void HexPrint( char*& ptr, uint64_t val ) +{ + if( val == 0 ) + { + *ptr++ = '0'; + return; + } + + static const char HexTable[16] = { '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'a', 'b', 'c', 'd', 'e', 'f' }; + char buf[16]; + auto bptr = buf; + + do + { + *bptr++ = HexTable[val%16]; + val /= 16; + } + while( val > 0 ); + + do + { + *ptr++ = *--bptr; + } + while( bptr != buf ); +} + +static void CrashHandler( int signal, siginfo_t* info, void* /*ucontext*/ ) +{ + bool expected = false; + if( !s_alreadyCrashed.compare_exchange_strong( expected, true ) ) ThreadFreezer( signal ); + + struct sigaction act = {}; + act.sa_handler = SIG_DFL; + sigaction( SIGABRT, &act, nullptr ); + + auto msgPtr = s_crashText; + switch( signal ) + { + case SIGILL: + strcpy( msgPtr, "Illegal Instruction.\n" ); + while( *msgPtr ) msgPtr++; + switch( info->si_code ) + { + case ILL_ILLOPC: + strcpy( msgPtr, "Illegal opcode.\n" ); + break; + case ILL_ILLOPN: + strcpy( msgPtr, "Illegal operand.\n" ); + break; + case ILL_ILLADR: + strcpy( msgPtr, "Illegal addressing mode.\n" ); + break; + case ILL_ILLTRP: + strcpy( msgPtr, "Illegal trap.\n" ); + break; + case ILL_PRVOPC: + strcpy( msgPtr, "Privileged opcode.\n" ); + break; + case ILL_PRVREG: + strcpy( msgPtr, "Privileged register.\n" ); + break; + case ILL_COPROC: + strcpy( msgPtr, "Coprocessor error.\n" ); + break; + case ILL_BADSTK: + strcpy( msgPtr, "Internal stack error.\n" ); + break; + default: + break; + } + break; + case SIGFPE: + strcpy( msgPtr, "Floating-point exception.\n" ); + while( *msgPtr ) msgPtr++; + switch( info->si_code ) + { + case FPE_INTDIV: + strcpy( msgPtr, "Integer divide by zero.\n" ); + break; + case FPE_INTOVF: + strcpy( msgPtr, "Integer overflow.\n" ); + break; + case FPE_FLTDIV: + strcpy( msgPtr, "Floating-point divide by zero.\n" ); + break; + case FPE_FLTOVF: + strcpy( msgPtr, "Floating-point overflow.\n" ); + break; + case FPE_FLTUND: + strcpy( msgPtr, "Floating-point underflow.\n" ); + break; + case FPE_FLTRES: + strcpy( msgPtr, "Floating-point inexact result.\n" ); + break; + case FPE_FLTINV: + strcpy( msgPtr, "Floating-point invalid operation.\n" ); + break; + case FPE_FLTSUB: + strcpy( msgPtr, "Subscript out of range.\n" ); + break; + default: + break; + } + break; + case SIGSEGV: + strcpy( msgPtr, "Invalid memory reference.\n" ); + while( *msgPtr ) msgPtr++; + switch( info->si_code ) + { + case SEGV_MAPERR: + strcpy( msgPtr, "Address not mapped to object.\n" ); + break; + case SEGV_ACCERR: + strcpy( msgPtr, "Invalid permissions for mapped object.\n" ); + break; +# ifdef SEGV_BNDERR + case SEGV_BNDERR: + strcpy( msgPtr, "Failed address bound checks.\n" ); + break; +# endif +# ifdef SEGV_PKUERR + case SEGV_PKUERR: + strcpy( msgPtr, "Access was denied by memory protection keys.\n" ); + break; +# endif + default: + break; + } + break; + case SIGPIPE: + strcpy( msgPtr, "Broken pipe.\n" ); + while( *msgPtr ) msgPtr++; + break; + case SIGBUS: + strcpy( msgPtr, "Bus error.\n" ); + while( *msgPtr ) msgPtr++; + switch( info->si_code ) + { + case BUS_ADRALN: + strcpy( msgPtr, "Invalid address alignment.\n" ); + break; + case BUS_ADRERR: + strcpy( msgPtr, "Nonexistent physical address.\n" ); + break; + case BUS_OBJERR: + strcpy( msgPtr, "Object-specific hardware error.\n" ); + break; +# ifdef BUS_MCEERR_AR + case BUS_MCEERR_AR: + strcpy( msgPtr, "Hardware memory error consumed on a machine check; action required.\n" ); + break; +# endif +# ifdef BUS_MCEERR_AO + case BUS_MCEERR_AO: + strcpy( msgPtr, "Hardware memory error detected in process but not consumed; action optional.\n" ); + break; +# endif + default: + break; + } + break; + case SIGABRT: + strcpy( msgPtr, "Abort signal from abort().\n" ); + break; + default: + abort(); + } + while( *msgPtr ) msgPtr++; + + if( signal != SIGPIPE ) + { + strcpy( msgPtr, "Fault address: 0x" ); + while( *msgPtr ) msgPtr++; + HexPrint( msgPtr, uint64_t( info->si_addr ) ); + *msgPtr++ = '\n'; + } + + { + GetProfiler().SendCallstack( 60, "__kernel_rt_sigreturn" ); + + TracyQueuePrepare( QueueType::CrashReport ); + item->crashReport.time = Profiler::GetTime(); + item->crashReport.text = (uint64_t)s_crashText; + TracyQueueCommit( crashReportThread ); + } + + DIR* dp = opendir( "/proc/self/task" ); + if( !dp ) abort(); + + const auto selfTid = syscall( SYS_gettid ); + + struct dirent* ep; + while( ( ep = readdir( dp ) ) != nullptr ) + { + if( ep->d_name[0] == '.' ) continue; + int tid = atoi( ep->d_name ); + if( tid != selfTid && tid != s_profilerTid && tid != s_symbolTid ) + { + syscall( SYS_tkill, tid, TRACY_CRASH_SIGNAL ); + } + } + closedir( dp ); + +#ifdef TRACY_HAS_CALLSTACK + if( selfTid == s_symbolTid ) s_symbolThreadGone.store( true, std::memory_order_release ); +#endif + + TracyLfqPrepare( QueueType::Crash ); + TracyLfqCommit; + + std::this_thread::sleep_for( std::chrono::milliseconds( 500 ) ); + GetProfiler().RequestShutdown(); + while( !GetProfiler().HasShutdownFinished() ) { std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) ); }; + + abort(); +} +#endif + + +enum { QueuePrealloc = 256 * 1024 }; + +TRACY_API int64_t GetFrequencyQpc() +{ +#if defined _WIN32 + LARGE_INTEGER t; + QueryPerformanceFrequency( &t ); + return t.QuadPart; +#else + return 0; +#endif +} + +#ifdef TRACY_DELAYED_INIT +struct ThreadNameData; +TRACY_API moodycamel::ConcurrentQueue& GetQueue(); + +struct ProfilerData +{ + int64_t initTime = SetupHwTimer(); + moodycamel::ConcurrentQueue queue; + Profiler profiler; + std::atomic lockCounter { 0 }; + std::atomic gpuCtxCounter { 0 }; + std::atomic threadNameData { nullptr }; +}; + +struct ProducerWrapper +{ + ProducerWrapper( ProfilerData& data ) : detail( data.queue ), ptr( data.queue.get_explicit_producer( detail ) ) {} + moodycamel::ProducerToken detail; + tracy::moodycamel::ConcurrentQueue::ExplicitProducer* ptr; +}; + +struct ProfilerThreadData +{ + ProfilerThreadData( ProfilerData& data ) : token( data ), gpuCtx( { nullptr } ) {} + ProducerWrapper token; + GpuCtxWrapper gpuCtx; +# ifdef TRACY_ON_DEMAND + LuaZoneState luaZoneState; +# endif +}; + +std::atomic RpInitDone { 0 }; +std::atomic RpInitLock { 0 }; +thread_local bool RpThreadInitDone = false; +thread_local bool RpThreadShutdown = false; + +# ifdef TRACY_MANUAL_LIFETIME +ProfilerData* s_profilerData = nullptr; +static ProfilerThreadData& GetProfilerThreadData(); +static std::atomic s_isProfilerStarted { false }; +TRACY_API void StartupProfiler() +{ + s_profilerData = (ProfilerData*)tracy_malloc( sizeof( ProfilerData ) ); + new (s_profilerData) ProfilerData(); + s_profilerData->profiler.SpawnWorkerThreads(); + GetProfilerThreadData().token = ProducerWrapper( *s_profilerData ); + s_isProfilerStarted.store( true, std::memory_order_seq_cst ); +} +static ProfilerData& GetProfilerData() +{ + assert( s_profilerData ); + return *s_profilerData; +} +TRACY_API void ShutdownProfiler() +{ + s_isProfilerStarted.store( false, std::memory_order_seq_cst ); + s_profilerData->~ProfilerData(); + tracy_free( s_profilerData ); + s_profilerData = nullptr; + rpmalloc_finalize(); + RpThreadInitDone = false; + RpInitDone.store( 0, std::memory_order_release ); +} +TRACY_API bool IsProfilerStarted() +{ + return s_isProfilerStarted.load( std::memory_order_seq_cst ); +} +# else +static std::atomic profilerDataLock { 0 }; +static std::atomic profilerData { nullptr }; + +static ProfilerData& GetProfilerData() +{ + auto ptr = profilerData.load( std::memory_order_acquire ); + if( !ptr ) + { + int expected = 0; + while( !profilerDataLock.compare_exchange_weak( expected, 1, std::memory_order_release, std::memory_order_relaxed ) ) { expected = 0; YieldThread(); } + ptr = profilerData.load( std::memory_order_acquire ); + if( !ptr ) + { + ptr = (ProfilerData*)tracy_malloc( sizeof( ProfilerData ) ); + new (ptr) ProfilerData(); + profilerData.store( ptr, std::memory_order_release ); + } + profilerDataLock.store( 0, std::memory_order_release ); + } + return *ptr; +} +# endif + +// GCC prior to 8.4 had a bug with function-inline thread_local variables. Versions of glibc beginning with +// 2.18 may attempt to work around this issue, which manifests as a crash while running static destructors +// if this function is compiled into a shared object. Unfortunately, centos7 ships with glibc 2.17. If running +// on old GCC, use the old-fashioned way as a workaround +// See: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85400 +#if !defined(__clang__) && defined(__GNUC__) && ((__GNUC__ < 8) || ((__GNUC__ == 8) && (__GNUC_MINOR__ < 4))) +struct ProfilerThreadDataKey +{ +public: + ProfilerThreadDataKey() + { + int val = pthread_key_create(&m_key, sDestructor); + static_cast(val); // unused + assert(val == 0); + } + ~ProfilerThreadDataKey() + { + int val = pthread_key_delete(m_key); + static_cast(val); // unused + assert(val == 0); + } + ProfilerThreadData& get() + { + void* p = pthread_getspecific(m_key); + if (!p) + { + p = (ProfilerThreadData*)tracy_malloc( sizeof( ProfilerThreadData ) ); + new (p) ProfilerThreadData(GetProfilerData()); + pthread_setspecific(m_key, p); + } + return *static_cast(p); + } +private: + pthread_key_t m_key; + + static void sDestructor(void* p) + { + ((ProfilerThreadData*)p)->~ProfilerThreadData(); + tracy_free(p); + } +}; + +static ProfilerThreadData& GetProfilerThreadData() +{ + static ProfilerThreadDataKey key; + return key.get(); +} +#else +static ProfilerThreadData& GetProfilerThreadData() +{ + thread_local ProfilerThreadData data( GetProfilerData() ); + return data; +} +#endif + +TRACY_API moodycamel::ConcurrentQueue::ExplicitProducer* GetToken() { return GetProfilerThreadData().token.ptr; } +TRACY_API Profiler& GetProfiler() { return GetProfilerData().profiler; } +TRACY_API moodycamel::ConcurrentQueue& GetQueue() { return GetProfilerData().queue; } +TRACY_API int64_t GetInitTime() { return GetProfilerData().initTime; } +TRACY_API std::atomic& GetLockCounter() { return GetProfilerData().lockCounter; } +TRACY_API std::atomic& GetGpuCtxCounter() { return GetProfilerData().gpuCtxCounter; } +TRACY_API GpuCtxWrapper& GetGpuCtx() { return GetProfilerThreadData().gpuCtx; } +TRACY_API uint32_t GetThreadHandle() { return detail::GetThreadHandleImpl(); } +std::atomic& GetThreadNameData() { return GetProfilerData().threadNameData; } + +# ifdef TRACY_ON_DEMAND +TRACY_API LuaZoneState& GetLuaZoneState() { return GetProfilerThreadData().luaZoneState; } +# endif + +# ifndef TRACY_MANUAL_LIFETIME +namespace +{ + const auto& __profiler_init = GetProfiler(); +} +# endif + +#else + +// MSVC static initialization order solution. gcc/clang uses init_order() to avoid all this. + +// 1a. But s_queue is needed for initialization of variables in point 2. +extern moodycamel::ConcurrentQueue s_queue; + +// 2. If these variables would be in the .CRT$XCB section, they would be initialized only in main thread. +thread_local moodycamel::ProducerToken init_order(107) s_token_detail( s_queue ); +thread_local ProducerWrapper init_order(108) s_token { s_queue.get_explicit_producer( s_token_detail ) }; +thread_local ThreadHandleWrapper init_order(104) s_threadHandle { detail::GetThreadHandleImpl() }; + +# ifdef _MSC_VER +// 1. Initialize these static variables before all other variables. +# pragma warning( disable : 4075 ) +# pragma init_seg( ".CRT$XCB" ) +# endif + +static InitTimeWrapper init_order(101) s_initTime { SetupHwTimer() }; +std::atomic init_order(102) RpInitDone( 0 ); +std::atomic init_order(102) RpInitLock( 0 ); +thread_local bool RpThreadInitDone = false; +thread_local bool RpThreadShutdown = false; +moodycamel::ConcurrentQueue init_order(103) s_queue( QueuePrealloc ); +std::atomic init_order(104) s_lockCounter( 0 ); +std::atomic init_order(104) s_gpuCtxCounter( 0 ); + +thread_local GpuCtxWrapper init_order(104) s_gpuCtx { nullptr }; + +struct ThreadNameData; +static std::atomic init_order(104) s_threadNameDataInstance( nullptr ); +std::atomic& s_threadNameData = s_threadNameDataInstance; + +# ifdef TRACY_ON_DEMAND +thread_local LuaZoneState init_order(104) s_luaZoneState { 0, false }; +# endif + +static Profiler init_order(105) s_profiler; + +TRACY_API moodycamel::ConcurrentQueue::ExplicitProducer* GetToken() { return s_token.ptr; } +TRACY_API Profiler& GetProfiler() { return s_profiler; } +TRACY_API moodycamel::ConcurrentQueue& GetQueue() { return s_queue; } +TRACY_API int64_t GetInitTime() { return s_initTime.val; } +TRACY_API std::atomic& GetLockCounter() { return s_lockCounter; } +TRACY_API std::atomic& GetGpuCtxCounter() { return s_gpuCtxCounter; } +TRACY_API GpuCtxWrapper& GetGpuCtx() { return s_gpuCtx; } +TRACY_API uint32_t GetThreadHandle() { return s_threadHandle.val; } + +std::atomic& GetThreadNameData() { return s_threadNameData; } + +# ifdef TRACY_ON_DEMAND +TRACY_API LuaZoneState& GetLuaZoneState() { return s_luaZoneState; } +# endif +#endif + +TRACY_API bool ProfilerAvailable() { return s_instance != nullptr; } +TRACY_API bool ProfilerAllocatorAvailable() { return !RpThreadShutdown; } + +Profiler::Profiler() + : m_timeBegin( 0 ) + , m_mainThread( detail::GetThreadHandleImpl() ) + , m_epoch( std::chrono::duration_cast( std::chrono::system_clock::now().time_since_epoch() ).count() ) + , m_shutdown( false ) + , m_shutdownManual( false ) + , m_shutdownFinished( false ) + , m_sock( nullptr ) + , m_broadcast( nullptr ) + , m_noExit( false ) + , m_userPort( 0 ) + , m_zoneId( 1 ) + , m_samplingPeriod( 0 ) + , m_stream( LZ4_createStream() ) + , m_buffer( (char*)tracy_malloc( TargetFrameSize*3 ) ) + , m_bufferOffset( 0 ) + , m_bufferStart( 0 ) + , m_lz4Buf( (char*)tracy_malloc( LZ4Size + sizeof( lz4sz_t ) ) ) + , m_serialQueue( 1024*1024 ) + , m_serialDequeue( 1024*1024 ) +#ifndef TRACY_NO_FRAME_IMAGE + , m_fiQueue( 16 ) + , m_fiDequeue( 16 ) +#endif + , m_symbolQueue( 8*1024 ) + , m_frameCount( 0 ) + , m_isConnected( false ) +#ifdef TRACY_ON_DEMAND + , m_connectionId( 0 ) + , m_deferredQueue( 64*1024 ) +#endif + , m_paramCallback( nullptr ) + , m_sourceCallback( nullptr ) + , m_queryImage( nullptr ) + , m_queryData( nullptr ) + , m_crashHandlerInstalled( false ) + , m_programName( nullptr ) +{ + assert( !s_instance ); + s_instance = this; + +#ifndef TRACY_DELAYED_INIT +# ifdef _MSC_VER + // 3. But these variables need to be initialized in main thread within the .CRT$XCB section. Do it here. + s_token_detail = moodycamel::ProducerToken( s_queue ); + s_token = ProducerWrapper { s_queue.get_explicit_producer( s_token_detail ) }; + s_threadHandle = ThreadHandleWrapper { m_mainThread }; +# endif +#endif + + CalibrateTimer(); + CalibrateDelay(); + ReportTopology(); + +#ifndef TRACY_NO_EXIT + const char* noExitEnv = GetEnvVar( "TRACY_NO_EXIT" ); + if( noExitEnv && noExitEnv[0] == '1' ) + { + m_noExit = true; + } +#endif + + const char* userPort = GetEnvVar( "TRACY_PORT" ); + if( userPort ) + { + m_userPort = atoi( userPort ); + } + +#if !defined(TRACY_DELAYED_INIT) || !defined(TRACY_MANUAL_LIFETIME) + SpawnWorkerThreads(); +#endif +} + +void Profiler::InstallCrashHandler() +{ + +#if defined __linux__ && !defined TRACY_NO_CRASH_HANDLER + struct sigaction threadFreezer = {}; + threadFreezer.sa_handler = ThreadFreezer; + sigaction( TRACY_CRASH_SIGNAL, &threadFreezer, &m_prevSignal.pwr ); + + struct sigaction crashHandler = {}; + crashHandler.sa_sigaction = CrashHandler; + crashHandler.sa_flags = SA_SIGINFO; + sigaction( SIGILL, &crashHandler, &m_prevSignal.ill ); + sigaction( SIGFPE, &crashHandler, &m_prevSignal.fpe ); + sigaction( SIGSEGV, &crashHandler, &m_prevSignal.segv ); + sigaction( SIGPIPE, &crashHandler, &m_prevSignal.pipe ); + sigaction( SIGBUS, &crashHandler, &m_prevSignal.bus ); + sigaction( SIGABRT, &crashHandler, &m_prevSignal.abrt ); +#endif + +#if defined _WIN32 && !defined TRACY_UWP && !defined TRACY_NO_CRASH_HANDLER + m_exceptionHandler = AddVectoredExceptionHandler( 1, CrashFilter ); +#endif + +#ifndef TRACY_NO_CRASH_HANDLER + m_crashHandlerInstalled = true; +#endif + +} + +void Profiler::RemoveCrashHandler() +{ +#if defined _WIN32 && !defined TRACY_UWP + if( m_crashHandlerInstalled ) RemoveVectoredExceptionHandler( m_exceptionHandler ); +#endif + +#if defined __linux__ && !defined TRACY_NO_CRASH_HANDLER + if( m_crashHandlerInstalled ) + { + sigaction( TRACY_CRASH_SIGNAL, &m_prevSignal.pwr, nullptr ); + sigaction( SIGILL, &m_prevSignal.ill, nullptr ); + sigaction( SIGFPE, &m_prevSignal.fpe, nullptr ); + sigaction( SIGSEGV, &m_prevSignal.segv, nullptr ); + sigaction( SIGPIPE, &m_prevSignal.pipe, nullptr ); + sigaction( SIGBUS, &m_prevSignal.bus, nullptr ); + sigaction( SIGABRT, &m_prevSignal.abrt, nullptr ); + } +#endif + m_crashHandlerInstalled = false; +} + +void Profiler::SpawnWorkerThreads() +{ +#ifdef TRACY_HAS_SYSTEM_TRACING + // use TRACY_NO_SYS_TRACE=1 to force disabling sys tracing (even if available in the underlying system) + // as it can have significant impact on the size of the traces + const char* noSysTrace = GetEnvVar( "TRACY_NO_SYS_TRACE" ); + const bool disableSystrace = (noSysTrace && noSysTrace[0] == '1'); + if( disableSystrace ) + { + TracyDebug("TRACY: Sys Trace was disabled by 'TRACY_NO_SYS_TRACE=1'\n"); + } + else if( SysTraceStart( m_samplingPeriod ) ) + { + s_sysTraceThread = (Thread*)tracy_malloc( sizeof( Thread ) ); + new(s_sysTraceThread) Thread( SysTraceWorker, nullptr ); + std::this_thread::sleep_for( std::chrono::milliseconds( 1 ) ); + } +#endif + + s_thread = (Thread*)tracy_malloc( sizeof( Thread ) ); + new(s_thread) Thread( LaunchWorker, this ); + +#ifndef TRACY_NO_FRAME_IMAGE + s_compressThread = (Thread*)tracy_malloc( sizeof( Thread ) ); + new(s_compressThread) Thread( LaunchCompressWorker, this ); +#endif + +#ifdef TRACY_HAS_CALLSTACK + s_symbolThread = (Thread*)tracy_malloc( sizeof( Thread ) ); + new(s_symbolThread) Thread( LaunchSymbolWorker, this ); +#endif + +#if defined _WIN32 && !defined TRACY_UWP && !defined TRACY_NO_CRASH_HANDLER + s_profilerThreadId = GetThreadId( s_thread->Handle() ); +# ifdef TRACY_HAS_CALLSTACK + s_symbolThreadId = GetThreadId( s_symbolThread->Handle() ); +# endif +#endif + +#ifdef TRACY_HAS_CALLSTACK + InitCallstackCritical(); +#endif + + m_timeBegin.store( GetTime(), std::memory_order_relaxed ); +} + +Profiler::~Profiler() +{ + m_shutdown.store( true, std::memory_order_relaxed ); + + RemoveCrashHandler(); + +#ifdef TRACY_HAS_SYSTEM_TRACING + if( s_sysTraceThread ) + { + SysTraceStop(); + s_sysTraceThread->~Thread(); + tracy_free( s_sysTraceThread ); + } +#endif + +#ifdef TRACY_HAS_CALLSTACK + s_symbolThread->~Thread(); + tracy_free( s_symbolThread ); +#endif + +#ifndef TRACY_NO_FRAME_IMAGE + s_compressThread->~Thread(); + tracy_free( s_compressThread ); +#endif + + s_thread->~Thread(); + tracy_free( s_thread ); + +#ifdef TRACY_HAS_CALLSTACK + EndCallstack(); +#endif + + tracy_free( m_lz4Buf ); + tracy_free( m_buffer ); + LZ4_freeStream( (LZ4_stream_t*)m_stream ); + + if( m_sock ) + { + m_sock->~Socket(); + tracy_free( m_sock ); + } + + if( m_broadcast ) + { + m_broadcast->~UdpBroadcast(); + tracy_free( m_broadcast ); + } + + assert( s_instance ); + s_instance = nullptr; +} + +bool Profiler::ShouldExit() +{ + return s_instance->m_shutdown.load( std::memory_order_relaxed ); +} + +void Profiler::Worker() +{ +#if defined __linux__ && !defined TRACY_NO_CRASH_HANDLER + s_profilerTid = syscall( SYS_gettid ); +#endif + + ThreadExitHandler threadExitHandler; + + SetThreadName( "Tracy Profiler" ); + +#ifdef TRACY_DATA_PORT + const bool dataPortSearch = false; + auto dataPort = m_userPort != 0 ? m_userPort : TRACY_DATA_PORT; +#else + const bool dataPortSearch = m_userPort == 0; + auto dataPort = m_userPort != 0 ? m_userPort : 8086; +#endif +#ifdef TRACY_BROADCAST_PORT + const auto broadcastPort = TRACY_BROADCAST_PORT; +#else + const auto broadcastPort = 8086; +#endif + + while( m_timeBegin.load( std::memory_order_relaxed ) == 0 ) std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) ); + +#ifdef TRACY_USE_RPMALLOC + rpmalloc_thread_initialize(); +#endif + + m_exectime = 0; + const auto execname = GetProcessExecutablePath(); + if( execname ) + { + struct stat st; + if( stat( execname, &st ) == 0 ) + { + m_exectime = (uint64_t)st.st_mtime; + } + } + + const auto procname = GetProcessName(); + const auto pnsz = std::min( strlen( procname ), WelcomeMessageProgramNameSize - 1 ); + + const auto hostinfo = GetHostInfo(); + const auto hisz = std::min( strlen( hostinfo ), WelcomeMessageHostInfoSize - 1 ); + + const uint64_t pid = GetPid(); + + uint8_t flags = 0; + +#ifdef TRACY_ON_DEMAND + flags |= WelcomeFlag::OnDemand; +#endif +#ifdef __APPLE__ + flags |= WelcomeFlag::IsApple; +#endif +#ifndef TRACY_NO_CODE_TRANSFER + flags |= WelcomeFlag::CodeTransfer; +#endif +#ifdef _WIN32 + flags |= WelcomeFlag::CombineSamples; +# ifndef TRACY_NO_CONTEXT_SWITCH + flags |= WelcomeFlag::IdentifySamples; +# endif +#endif + +#if defined __i386 || defined _M_IX86 + uint8_t cpuArch = CpuArchX86; +#elif defined __x86_64__ || defined _M_X64 + uint8_t cpuArch = CpuArchX64; +#elif defined __aarch64__ + uint8_t cpuArch = CpuArchArm64; +#elif defined __ARM_ARCH + uint8_t cpuArch = CpuArchArm32; +#else + uint8_t cpuArch = CpuArchUnknown; +#endif + +#if defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 + uint32_t regs[4]; + char manufacturer[12]; + CpuId( regs, 0 ); + memcpy( manufacturer, regs+1, 4 ); + memcpy( manufacturer+4, regs+3, 4 ); + memcpy( manufacturer+8, regs+2, 4 ); + + CpuId( regs, 1 ); + uint32_t cpuId = ( regs[0] & 0xFFF ) | ( ( regs[0] & 0xFFF0000 ) >> 4 ); +#else + const char manufacturer[12] = {}; + uint32_t cpuId = 0; +#endif + + WelcomeMessage welcome; + MemWrite( &welcome.timerMul, m_timerMul ); + MemWrite( &welcome.initBegin, GetInitTime() ); + MemWrite( &welcome.initEnd, m_timeBegin.load( std::memory_order_relaxed ) ); + MemWrite( &welcome.delay, m_delay ); + MemWrite( &welcome.resolution, m_resolution ); + MemWrite( &welcome.epoch, m_epoch ); + MemWrite( &welcome.exectime, m_exectime ); + MemWrite( &welcome.pid, pid ); + MemWrite( &welcome.samplingPeriod, m_samplingPeriod ); + MemWrite( &welcome.flags, flags ); + MemWrite( &welcome.cpuArch, cpuArch ); + memcpy( welcome.cpuManufacturer, manufacturer, 12 ); + MemWrite( &welcome.cpuId, cpuId ); + memcpy( welcome.programName, procname, pnsz ); + memset( welcome.programName + pnsz, 0, WelcomeMessageProgramNameSize - pnsz ); + memcpy( welcome.hostInfo, hostinfo, hisz ); + memset( welcome.hostInfo + hisz, 0, WelcomeMessageHostInfoSize - hisz ); + + moodycamel::ConsumerToken token( GetQueue() ); + + ListenSocket listen; + bool isListening = false; + if( !dataPortSearch ) + { + isListening = listen.Listen( dataPort, 4 ); + } + else + { + for( uint32_t i=0; i<20; i++ ) + { + if( listen.Listen( dataPort+i, 4 ) ) + { + dataPort += i; + isListening = true; + break; + } + } + } + if( !isListening ) + { + for(;;) + { + if( ShouldExit() ) + { + m_shutdownFinished.store( true, std::memory_order_relaxed ); + return; + } + + ClearQueues( token ); + std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) ); + } + } + +#ifndef TRACY_NO_BROADCAST + m_broadcast = (UdpBroadcast*)tracy_malloc( sizeof( UdpBroadcast ) ); + new(m_broadcast) UdpBroadcast(); +# ifdef TRACY_ONLY_LOCALHOST + const char* addr = "127.255.255.255"; +# elif defined TRACY_CLIENT_ADDRESS + const char* addr = TRACY_CLIENT_ADDRESS; +# elif defined __QNX__ + // global broadcast address of 255.255.255.255 is not well-supported by QNX, + // use the interface broadcast address instead, e.g. "const char* addr = 192.168.1.255;" +# error Need to specify TRACY_CLIENT_ADDRESS for a QNX target. +# else + const char* addr = "255.255.255.255"; +# endif + if( !m_broadcast->Open( addr, broadcastPort ) ) + { + m_broadcast->~UdpBroadcast(); + tracy_free( m_broadcast ); + m_broadcast = nullptr; + } +#endif + + int broadcastLen = 0; + auto& broadcastMsg = GetBroadcastMessage( procname, pnsz, broadcastLen, dataPort ); + uint64_t lastBroadcast = 0; + + // Connections loop. + // Each iteration of the loop handles whole connection. Multiple iterations will only + // happen in the on-demand mode or when handshake fails. + for(;;) + { + // Wait for incoming connection + for(;;) + { +#ifndef TRACY_NO_EXIT + if( !m_noExit && ShouldExit() ) + { + if( m_broadcast ) + { + broadcastMsg.activeTime = -1; + m_broadcast->Send( broadcastPort, &broadcastMsg, broadcastLen ); + } + m_shutdownFinished.store( true, std::memory_order_relaxed ); + return; + } +#endif + m_sock = listen.Accept(); + if( m_sock ) break; +#ifndef TRACY_ON_DEMAND + ProcessSysTime(); +# ifdef TRACY_HAS_SYSPOWER + m_sysPower.Tick(); +# endif +#endif + + if( m_broadcast ) + { + const auto t = std::chrono::high_resolution_clock::now().time_since_epoch().count(); + if( t - lastBroadcast > 3000000000 ) // 3s + { + m_programNameLock.lock(); + if( m_programName ) + { + broadcastMsg = GetBroadcastMessage( m_programName, strlen( m_programName ), broadcastLen, dataPort ); + m_programName = nullptr; + } + m_programNameLock.unlock(); + + lastBroadcast = t; + const auto ts = std::chrono::duration_cast( std::chrono::system_clock::now().time_since_epoch() ).count(); + broadcastMsg.activeTime = int32_t( ts - m_epoch ); + assert( broadcastMsg.activeTime >= 0 ); + m_broadcast->Send( broadcastPort, &broadcastMsg, broadcastLen ); + } + } + } + + if( m_broadcast ) + { + lastBroadcast = 0; + broadcastMsg.activeTime = -1; + m_broadcast->Send( broadcastPort, &broadcastMsg, broadcastLen ); + } + + // Handshake + { + char shibboleth[HandshakeShibbolethSize]; + auto res = m_sock->ReadRaw( shibboleth, HandshakeShibbolethSize, 2000 ); + if( !res || memcmp( shibboleth, HandshakeShibboleth, HandshakeShibbolethSize ) != 0 ) + { + m_sock->~Socket(); + tracy_free( m_sock ); + m_sock = nullptr; + continue; + } + + uint32_t protocolVersion; + res = m_sock->ReadRaw( &protocolVersion, sizeof( protocolVersion ), 2000 ); + if( !res ) + { + m_sock->~Socket(); + tracy_free( m_sock ); + m_sock = nullptr; + continue; + } + + if( protocolVersion != ProtocolVersion ) + { + HandshakeStatus status = HandshakeProtocolMismatch; + m_sock->Send( &status, sizeof( status ) ); + m_sock->~Socket(); + tracy_free( m_sock ); + m_sock = nullptr; + continue; + } + } + +#ifdef TRACY_ON_DEMAND + const auto currentTime = GetTime(); + ClearQueues( token ); + m_connectionId.fetch_add( 1, std::memory_order_release ); +#endif + m_isConnected.store( true, std::memory_order_release ); + InstallCrashHandler(); + + HandshakeStatus handshake = HandshakeWelcome; + m_sock->Send( &handshake, sizeof( handshake ) ); + + LZ4_resetStream( (LZ4_stream_t*)m_stream ); + m_sock->Send( &welcome, sizeof( welcome ) ); + + m_threadCtx = 0; + m_refTimeSerial = 0; + m_refTimeCtx = 0; + m_refTimeGpu = 0; + +#ifdef TRACY_ON_DEMAND + OnDemandPayloadMessage onDemand; + onDemand.frames = m_frameCount.load( std::memory_order_relaxed ); + onDemand.currentTime = currentTime; + + m_sock->Send( &onDemand, sizeof( onDemand ) ); + + m_deferredLock.lock(); + for( auto& item : m_deferredQueue ) + { + uint64_t ptr; + uint16_t size; + const auto idx = MemRead( &item.hdr.idx ); + switch( (QueueType)idx ) + { + case QueueType::MessageAppInfo: + ptr = MemRead( &item.messageFat.text ); + size = MemRead( &item.messageFat.size ); + SendSingleString( (const char*)ptr, size ); + break; + case QueueType::LockName: + ptr = MemRead( &item.lockNameFat.name ); + size = MemRead( &item.lockNameFat.size ); + SendSingleString( (const char*)ptr, size ); + break; + case QueueType::GpuContextName: + ptr = MemRead( &item.gpuContextNameFat.ptr ); + size = MemRead( &item.gpuContextNameFat.size ); + SendSingleString( (const char*)ptr, size ); + break; + default: + break; + } + AppendData( &item, QueueDataSize[idx] ); + } + m_deferredLock.unlock(); +#endif + + // Main communications loop + int keepAlive = 0; + for(;;) + { + ProcessSysTime(); +#ifdef TRACY_HAS_SYSPOWER + m_sysPower.Tick(); +#endif + const auto status = Dequeue( token ); + const auto serialStatus = DequeueSerial(); + if( status == DequeueStatus::ConnectionLost || serialStatus == DequeueStatus::ConnectionLost ) + { + break; + } + else if( status == DequeueStatus::QueueEmpty && serialStatus == DequeueStatus::QueueEmpty ) + { + if( ShouldExit() ) break; + if( m_bufferOffset != m_bufferStart ) + { + if( !CommitData() ) break; + } + if( keepAlive == 500 ) + { + QueueItem ka; + ka.hdr.type = QueueType::KeepAlive; + AppendData( &ka, QueueDataSize[ka.hdr.idx] ); + if( !CommitData() ) break; + + keepAlive = 0; + } + else if( !m_sock->HasData() ) + { + keepAlive++; + std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) ); + } + } + else + { + keepAlive = 0; + } + + bool connActive = true; + while( m_sock->HasData() ) + { + connActive = HandleServerQuery(); + if( !connActive ) break; + } + if( !connActive ) break; + } + if( ShouldExit() ) break; + + m_isConnected.store( false, std::memory_order_release ); + RemoveCrashHandler(); + +#ifdef TRACY_ON_DEMAND + m_bufferOffset = 0; + m_bufferStart = 0; +#endif + + m_sock->~Socket(); + tracy_free( m_sock ); + m_sock = nullptr; + +#ifndef TRACY_ON_DEMAND + // Client is no longer available here. Accept incoming connections, but reject handshake. + for(;;) + { + if( ShouldExit() ) + { + m_shutdownFinished.store( true, std::memory_order_relaxed ); + return; + } + + ClearQueues( token ); + + m_sock = listen.Accept(); + if( m_sock ) + { + char shibboleth[HandshakeShibbolethSize]; + auto res = m_sock->ReadRaw( shibboleth, HandshakeShibbolethSize, 1000 ); + if( !res || memcmp( shibboleth, HandshakeShibboleth, HandshakeShibbolethSize ) != 0 ) + { + m_sock->~Socket(); + tracy_free( m_sock ); + m_sock = nullptr; + continue; + } + + uint32_t protocolVersion; + res = m_sock->ReadRaw( &protocolVersion, sizeof( protocolVersion ), 1000 ); + if( !res ) + { + m_sock->~Socket(); + tracy_free( m_sock ); + m_sock = nullptr; + continue; + } + + HandshakeStatus status = HandshakeNotAvailable; + m_sock->Send( &status, sizeof( status ) ); + m_sock->~Socket(); + tracy_free( m_sock ); + } + } +#endif + } + // End of connections loop + + // Wait for symbols thread to terminate. Symbol resolution will continue in this thread. +#ifdef TRACY_HAS_CALLSTACK + while( s_symbolThreadGone.load() == false ) { YieldThread(); } +#endif + + // Client is exiting. Send items remaining in queues. + for(;;) + { + const auto status = Dequeue( token ); + const auto serialStatus = DequeueSerial(); + if( status == DequeueStatus::ConnectionLost || serialStatus == DequeueStatus::ConnectionLost ) + { + m_shutdownFinished.store( true, std::memory_order_relaxed ); + return; + } + else if( status == DequeueStatus::QueueEmpty && serialStatus == DequeueStatus::QueueEmpty ) + { + if( m_bufferOffset != m_bufferStart ) CommitData(); + break; + } + + while( m_sock->HasData() ) + { + if( !HandleServerQuery() ) + { + m_shutdownFinished.store( true, std::memory_order_relaxed ); + return; + } + } + +#ifdef TRACY_HAS_CALLSTACK + for(;;) + { + auto si = m_symbolQueue.front(); + if( !si ) break; + HandleSymbolQueueItem( *si ); + m_symbolQueue.pop(); + } +#endif + } + + // Send client termination notice to the server + QueueItem terminate; + MemWrite( &terminate.hdr.type, QueueType::Terminate ); + if( !SendData( (const char*)&terminate, 1 ) ) + { + m_shutdownFinished.store( true, std::memory_order_relaxed ); + return; + } + // Handle remaining server queries + for(;;) + { + while( m_sock->HasData() ) + { + if( !HandleServerQuery() ) + { + m_shutdownFinished.store( true, std::memory_order_relaxed ); + return; + } + } +#ifdef TRACY_HAS_CALLSTACK + for(;;) + { + auto si = m_symbolQueue.front(); + if( !si ) break; + HandleSymbolQueueItem( *si ); + m_symbolQueue.pop(); + } +#endif + const auto status = Dequeue( token ); + const auto serialStatus = DequeueSerial(); + if( status == DequeueStatus::ConnectionLost || serialStatus == DequeueStatus::ConnectionLost ) + { + m_shutdownFinished.store( true, std::memory_order_relaxed ); + return; + } + if( m_bufferOffset != m_bufferStart ) + { + if( !CommitData() ) + { + m_shutdownFinished.store( true, std::memory_order_relaxed ); + return; + } + } + } +} + +#ifndef TRACY_NO_FRAME_IMAGE +void Profiler::CompressWorker() +{ + ThreadExitHandler threadExitHandler; + SetThreadName( "Tracy DXT1" ); + while( m_timeBegin.load( std::memory_order_relaxed ) == 0 ) std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) ); + +#ifdef TRACY_USE_RPMALLOC + rpmalloc_thread_initialize(); +#endif + + for(;;) + { + const auto shouldExit = ShouldExit(); + + { + bool lockHeld = true; + while( !m_fiLock.try_lock() ) + { + if( m_shutdownManual.load( std::memory_order_relaxed ) ) + { + lockHeld = false; + break; + } + } + if( !m_fiQueue.empty() ) m_fiQueue.swap( m_fiDequeue ); + if( lockHeld ) + { + m_fiLock.unlock(); + } + } + + const auto sz = m_fiDequeue.size(); + if( sz > 0 ) + { + auto fi = m_fiDequeue.data(); + auto end = fi + sz; + while( fi != end ) + { + const auto w = fi->w; + const auto h = fi->h; + const auto csz = size_t( w * h / 2 ); + auto etc1buf = (char*)tracy_malloc( csz ); + CompressImageDxt1( (const char*)fi->image, etc1buf, w, h ); + tracy_free( fi->image ); + + TracyLfqPrepare( QueueType::FrameImage ); + MemWrite( &item->frameImageFat.image, (uint64_t)etc1buf ); + MemWrite( &item->frameImageFat.frame, fi->frame ); + MemWrite( &item->frameImageFat.w, w ); + MemWrite( &item->frameImageFat.h, h ); + uint8_t flip = fi->flip; + MemWrite( &item->frameImageFat.flip, flip ); + TracyLfqCommit; + + fi++; + } + m_fiDequeue.clear(); + } + else + { + std::this_thread::sleep_for( std::chrono::milliseconds( 20 ) ); + } + + if( shouldExit ) + { + return; + } + } +} +#endif + +static void FreeAssociatedMemory( const QueueItem& item ) +{ + if( item.hdr.idx >= (int)QueueType::Terminate ) return; + + uint64_t ptr; + switch( item.hdr.type ) + { + case QueueType::ZoneText: + case QueueType::ZoneName: + ptr = MemRead( &item.zoneTextFat.text ); + tracy_free( (void*)ptr ); + break; + case QueueType::MessageColor: + case QueueType::MessageColorCallstack: + ptr = MemRead( &item.messageColorFat.text ); + tracy_free( (void*)ptr ); + break; + case QueueType::Message: + case QueueType::MessageCallstack: +#ifndef TRACY_ON_DEMAND + case QueueType::MessageAppInfo: +#endif + ptr = MemRead( &item.messageFat.text ); + tracy_free( (void*)ptr ); + break; + case QueueType::ZoneBeginAllocSrcLoc: + case QueueType::ZoneBeginAllocSrcLocCallstack: + ptr = MemRead( &item.zoneBegin.srcloc ); + tracy_free( (void*)ptr ); + break; + case QueueType::GpuZoneBeginAllocSrcLoc: + case QueueType::GpuZoneBeginAllocSrcLocCallstack: + case QueueType::GpuZoneBeginAllocSrcLocSerial: + case QueueType::GpuZoneBeginAllocSrcLocCallstackSerial: + ptr = MemRead( &item.gpuZoneBegin.srcloc ); + tracy_free( (void*)ptr ); + break; + case QueueType::CallstackSerial: + case QueueType::Callstack: + ptr = MemRead( &item.callstackFat.ptr ); + tracy_free( (void*)ptr ); + break; + case QueueType::CallstackAlloc: + ptr = MemRead( &item.callstackAllocFat.nativePtr ); + tracy_free( (void*)ptr ); + ptr = MemRead( &item.callstackAllocFat.ptr ); + tracy_free( (void*)ptr ); + break; + case QueueType::CallstackSample: + case QueueType::CallstackSampleContextSwitch: + ptr = MemRead( &item.callstackSampleFat.ptr ); + tracy_free( (void*)ptr ); + break; + case QueueType::FrameImage: + ptr = MemRead( &item.frameImageFat.image ); + tracy_free( (void*)ptr ); + break; +#ifdef TRACY_HAS_CALLSTACK + case QueueType::CallstackFrameSize: + { + InitRpmalloc(); + auto size = MemRead( &item.callstackFrameSizeFat.size ); + auto data = (const CallstackEntry*)MemRead( &item.callstackFrameSizeFat.data ); + for( uint8_t i=0; i( &item.symbolInformationFat.needFree ); + if( needFree ) + { + ptr = MemRead( &item.symbolInformationFat.fileString ); + tracy_free( (void*)ptr ); + } + break; + } + case QueueType::SymbolCodeMetadata: + ptr = MemRead( &item.symbolCodeMetadata.ptr ); + tracy_free( (void*)ptr ); + break; +#endif +#ifndef TRACY_ON_DEMAND + case QueueType::LockName: + ptr = MemRead( &item.lockNameFat.name ); + tracy_free( (void*)ptr ); + break; + case QueueType::GpuContextName: + ptr = MemRead( &item.gpuContextNameFat.ptr ); + tracy_free( (void*)ptr ); + break; +#endif +#ifdef TRACY_ON_DEMAND + case QueueType::MessageAppInfo: + case QueueType::GpuContextName: + // Don't free memory associated with deferred messages. + break; +#endif +#ifdef TRACY_HAS_SYSTEM_TRACING + case QueueType::ExternalNameMetadata: + ptr = MemRead( &item.externalNameMetadata.name ); + tracy_free( (void*)ptr ); + ptr = MemRead( &item.externalNameMetadata.threadName ); + tracy_free_fast( (void*)ptr ); + break; +#endif + case QueueType::SourceCodeMetadata: + ptr = MemRead( &item.sourceCodeMetadata.ptr ); + tracy_free( (void*)ptr ); + break; + default: + break; + } +} + +void Profiler::ClearQueues( moodycamel::ConsumerToken& token ) +{ + for(;;) + { + const auto sz = GetQueue().try_dequeue_bulk_single( token, [](const uint64_t&){}, []( QueueItem* item, size_t sz ) { assert( sz > 0 ); while( sz-- > 0 ) FreeAssociatedMemory( *item++ ); } ); + if( sz == 0 ) break; + } + + ClearSerial(); +} + +void Profiler::ClearSerial() +{ + bool lockHeld = true; + while( !m_serialLock.try_lock() ) + { + if( m_shutdownManual.load( std::memory_order_relaxed ) ) + { + lockHeld = false; + break; + } + } + for( auto& v : m_serialQueue ) FreeAssociatedMemory( v ); + m_serialQueue.clear(); + if( lockHeld ) + { + m_serialLock.unlock(); + } + + for( auto& v : m_serialDequeue ) FreeAssociatedMemory( v ); + m_serialDequeue.clear(); +} + +Profiler::DequeueStatus Profiler::Dequeue( moodycamel::ConsumerToken& token ) +{ + bool connectionLost = false; + const auto sz = GetQueue().try_dequeue_bulk_single( token, + [this, &connectionLost] ( const uint32_t& threadId ) + { + if( ThreadCtxCheck( threadId ) == ThreadCtxStatus::ConnectionLost ) connectionLost = true; + }, + [this, &connectionLost] ( QueueItem* item, size_t sz ) + { + if( connectionLost ) return; + InitRpmalloc(); + assert( sz > 0 ); + int64_t refThread = m_refTimeThread; + int64_t refCtx = m_refTimeCtx; + int64_t refGpu = m_refTimeGpu; + while( sz-- > 0 ) + { + uint64_t ptr; + uint16_t size; + auto idx = MemRead( &item->hdr.idx ); + if( idx < (int)QueueType::Terminate ) + { + switch( (QueueType)idx ) + { + case QueueType::ZoneText: + case QueueType::ZoneName: + ptr = MemRead( &item->zoneTextFat.text ); + size = MemRead( &item->zoneTextFat.size ); + SendSingleString( (const char*)ptr, size ); + tracy_free_fast( (void*)ptr ); + break; + case QueueType::Message: + case QueueType::MessageCallstack: + ptr = MemRead( &item->messageFat.text ); + size = MemRead( &item->messageFat.size ); + SendSingleString( (const char*)ptr, size ); + tracy_free_fast( (void*)ptr ); + break; + case QueueType::MessageColor: + case QueueType::MessageColorCallstack: + ptr = MemRead( &item->messageColorFat.text ); + size = MemRead( &item->messageColorFat.size ); + SendSingleString( (const char*)ptr, size ); + tracy_free_fast( (void*)ptr ); + break; + case QueueType::MessageAppInfo: + ptr = MemRead( &item->messageFat.text ); + size = MemRead( &item->messageFat.size ); + SendSingleString( (const char*)ptr, size ); +#ifndef TRACY_ON_DEMAND + tracy_free_fast( (void*)ptr ); +#endif + break; + case QueueType::ZoneBeginAllocSrcLoc: + case QueueType::ZoneBeginAllocSrcLocCallstack: + { + int64_t t = MemRead( &item->zoneBegin.time ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->zoneBegin.time, dt ); + ptr = MemRead( &item->zoneBegin.srcloc ); + SendSourceLocationPayload( ptr ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::Callstack: + ptr = MemRead( &item->callstackFat.ptr ); + SendCallstackPayload( ptr ); + tracy_free_fast( (void*)ptr ); + break; + case QueueType::CallstackAlloc: + ptr = MemRead( &item->callstackAllocFat.nativePtr ); + if( ptr != 0 ) + { + CutCallstack( (void*)ptr, "lua_pcall" ); + SendCallstackPayload( ptr ); + tracy_free_fast( (void*)ptr ); + } + ptr = MemRead( &item->callstackAllocFat.ptr ); + SendCallstackAlloc( ptr ); + tracy_free_fast( (void*)ptr ); + break; + case QueueType::CallstackSample: + case QueueType::CallstackSampleContextSwitch: + { + ptr = MemRead( &item->callstackSampleFat.ptr ); + SendCallstackPayload64( ptr ); + tracy_free_fast( (void*)ptr ); + int64_t t = MemRead( &item->callstackSampleFat.time ); + int64_t dt = t - refCtx; + refCtx = t; + MemWrite( &item->callstackSampleFat.time, dt ); + break; + } + case QueueType::FrameImage: + { + ptr = MemRead( &item->frameImageFat.image ); + const auto w = MemRead( &item->frameImageFat.w ); + const auto h = MemRead( &item->frameImageFat.h ); + const auto csz = size_t( w * h / 2 ); + SendLongString( ptr, (const char*)ptr, csz, QueueType::FrameImageData ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::ZoneBegin: + case QueueType::ZoneBeginCallstack: + { + int64_t t = MemRead( &item->zoneBegin.time ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->zoneBegin.time, dt ); + break; + } + case QueueType::ZoneEnd: + { + int64_t t = MemRead( &item->zoneEnd.time ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->zoneEnd.time, dt ); + break; + } + case QueueType::GpuZoneBegin: + case QueueType::GpuZoneBeginCallstack: + { + int64_t t = MemRead( &item->gpuZoneBegin.cpuTime ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->gpuZoneBegin.cpuTime, dt ); + break; + } + case QueueType::GpuZoneBeginAllocSrcLoc: + case QueueType::GpuZoneBeginAllocSrcLocCallstack: + { + int64_t t = MemRead( &item->gpuZoneBegin.cpuTime ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->gpuZoneBegin.cpuTime, dt ); + ptr = MemRead( &item->gpuZoneBegin.srcloc ); + SendSourceLocationPayload( ptr ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::GpuZoneEnd: + { + int64_t t = MemRead( &item->gpuZoneEnd.cpuTime ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->gpuZoneEnd.cpuTime, dt ); + break; + } + case QueueType::GpuContextName: + ptr = MemRead( &item->gpuContextNameFat.ptr ); + size = MemRead( &item->gpuContextNameFat.size ); + SendSingleString( (const char*)ptr, size ); +#ifndef TRACY_ON_DEMAND + tracy_free_fast( (void*)ptr ); +#endif + break; + case QueueType::PlotDataInt: + case QueueType::PlotDataFloat: + case QueueType::PlotDataDouble: + { + int64_t t = MemRead( &item->plotDataInt.time ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->plotDataInt.time, dt ); + break; + } + case QueueType::ContextSwitch: + { + int64_t t = MemRead( &item->contextSwitch.time ); + int64_t dt = t - refCtx; + refCtx = t; + MemWrite( &item->contextSwitch.time, dt ); + break; + } + case QueueType::ThreadWakeup: + { + int64_t t = MemRead( &item->threadWakeup.time ); + int64_t dt = t - refCtx; + refCtx = t; + MemWrite( &item->threadWakeup.time, dt ); + break; + } + case QueueType::GpuTime: + { + int64_t t = MemRead( &item->gpuTime.gpuTime ); + int64_t dt = t - refGpu; + refGpu = t; + MemWrite( &item->gpuTime.gpuTime, dt ); + break; + } +#ifdef TRACY_HAS_CALLSTACK + case QueueType::CallstackFrameSize: + { + auto data = (const CallstackEntry*)MemRead( &item->callstackFrameSizeFat.data ); + auto datasz = MemRead( &item->callstackFrameSizeFat.size ); + auto imageName = (const char*)MemRead( &item->callstackFrameSizeFat.imageName ); + SendSingleString( imageName ); + AppendData( item++, QueueDataSize[idx] ); + + for( uint8_t i=0; i( &item->symbolInformationFat.fileString ); + auto needFree = MemRead( &item->symbolInformationFat.needFree ); + SendSingleString( fileString ); + if( needFree ) tracy_free_fast( (void*)fileString ); + break; + } + case QueueType::SymbolCodeMetadata: + { + auto symbol = MemRead( &item->symbolCodeMetadata.symbol ); + auto ptr = (const char*)MemRead( &item->symbolCodeMetadata.ptr ); + auto size = MemRead( &item->symbolCodeMetadata.size ); + SendLongString( symbol, ptr, size, QueueType::SymbolCode ); + tracy_free_fast( (void*)ptr ); + ++item; + continue; + } +#endif +#ifdef TRACY_HAS_SYSTEM_TRACING + case QueueType::ExternalNameMetadata: + { + auto thread = MemRead( &item->externalNameMetadata.thread ); + auto name = (const char*)MemRead( &item->externalNameMetadata.name ); + auto threadName = (const char*)MemRead( &item->externalNameMetadata.threadName ); + SendString( thread, threadName, QueueType::ExternalThreadName ); + SendString( thread, name, QueueType::ExternalName ); + tracy_free_fast( (void*)threadName ); + tracy_free_fast( (void*)name ); + ++item; + continue; + } +#endif + case QueueType::SourceCodeMetadata: + { + auto ptr = (const char*)MemRead( &item->sourceCodeMetadata.ptr ); + auto size = MemRead( &item->sourceCodeMetadata.size ); + auto id = MemRead( &item->sourceCodeMetadata.id ); + SendLongString( (uint64_t)id, ptr, size, QueueType::SourceCode ); + tracy_free_fast( (void*)ptr ); + ++item; + continue; + } + default: + assert( false ); + break; + } + } + if( !AppendData( item++, QueueDataSize[idx] ) ) + { + connectionLost = true; + m_refTimeThread = refThread; + m_refTimeCtx = refCtx; + m_refTimeGpu = refGpu; + return; + } + } + m_refTimeThread = refThread; + m_refTimeCtx = refCtx; + m_refTimeGpu = refGpu; + } + ); + if( connectionLost ) return DequeueStatus::ConnectionLost; + return sz > 0 ? DequeueStatus::DataDequeued : DequeueStatus::QueueEmpty; +} + +Profiler::DequeueStatus Profiler::DequeueContextSwitches( tracy::moodycamel::ConsumerToken& token, int64_t& timeStop ) +{ + const auto sz = GetQueue().try_dequeue_bulk_single( token, [] ( const uint64_t& ) {}, + [this, &timeStop] ( QueueItem* item, size_t sz ) + { + assert( sz > 0 ); + int64_t refCtx = m_refTimeCtx; + while( sz-- > 0 ) + { + FreeAssociatedMemory( *item ); + if( timeStop < 0 ) return; + const auto idx = MemRead( &item->hdr.idx ); + if( idx == (uint8_t)QueueType::ContextSwitch ) + { + const auto csTime = MemRead( &item->contextSwitch.time ); + if( csTime > timeStop ) + { + timeStop = -1; + m_refTimeCtx = refCtx; + return; + } + int64_t dt = csTime - refCtx; + refCtx = csTime; + MemWrite( &item->contextSwitch.time, dt ); + if( !AppendData( item, QueueDataSize[(int)QueueType::ContextSwitch] ) ) + { + timeStop = -2; + m_refTimeCtx = refCtx; + return; + } + } + else if( idx == (uint8_t)QueueType::ThreadWakeup ) + { + const auto csTime = MemRead( &item->threadWakeup.time ); + if( csTime > timeStop ) + { + timeStop = -1; + m_refTimeCtx = refCtx; + return; + } + int64_t dt = csTime - refCtx; + refCtx = csTime; + MemWrite( &item->threadWakeup.time, dt ); + if( !AppendData( item, QueueDataSize[(int)QueueType::ThreadWakeup] ) ) + { + timeStop = -2; + m_refTimeCtx = refCtx; + return; + } + } + item++; + } + m_refTimeCtx = refCtx; + } + ); + + if( timeStop == -2 ) return DequeueStatus::ConnectionLost; + return ( timeStop == -1 || sz > 0 ) ? DequeueStatus::DataDequeued : DequeueStatus::QueueEmpty; +} + +#define ThreadCtxCheckSerial( _name ) \ + uint32_t thread = MemRead( &item->_name.thread ); \ + switch( ThreadCtxCheck( thread ) ) \ + { \ + case ThreadCtxStatus::Same: break; \ + case ThreadCtxStatus::Changed: assert( m_refTimeThread == 0 ); refThread = 0; break; \ + case ThreadCtxStatus::ConnectionLost: return DequeueStatus::ConnectionLost; \ + default: assert( false ); break; \ + } + +Profiler::DequeueStatus Profiler::DequeueSerial() +{ + { + bool lockHeld = true; + while( !m_serialLock.try_lock() ) + { + if( m_shutdownManual.load( std::memory_order_relaxed ) ) + { + lockHeld = false; + break; + } + } + if( !m_serialQueue.empty() ) m_serialQueue.swap( m_serialDequeue ); + if( lockHeld ) + { + m_serialLock.unlock(); + } + } + + const auto sz = m_serialDequeue.size(); + if( sz > 0 ) + { + InitRpmalloc(); + int64_t refSerial = m_refTimeSerial; + int64_t refGpu = m_refTimeGpu; +#ifdef TRACY_FIBERS + int64_t refThread = m_refTimeThread; +#endif + auto item = m_serialDequeue.data(); + auto end = item + sz; + while( item != end ) + { + uint64_t ptr; + auto idx = MemRead( &item->hdr.idx ); + if( idx < (int)QueueType::Terminate ) + { + switch( (QueueType)idx ) + { + case QueueType::CallstackSerial: + ptr = MemRead( &item->callstackFat.ptr ); + SendCallstackPayload( ptr ); + tracy_free_fast( (void*)ptr ); + break; + case QueueType::LockWait: + case QueueType::LockSharedWait: + { + int64_t t = MemRead( &item->lockWait.time ); + int64_t dt = t - refSerial; + refSerial = t; + MemWrite( &item->lockWait.time, dt ); + break; + } + case QueueType::LockObtain: + case QueueType::LockSharedObtain: + { + int64_t t = MemRead( &item->lockObtain.time ); + int64_t dt = t - refSerial; + refSerial = t; + MemWrite( &item->lockObtain.time, dt ); + break; + } + case QueueType::LockRelease: + case QueueType::LockSharedRelease: + { + int64_t t = MemRead( &item->lockRelease.time ); + int64_t dt = t - refSerial; + refSerial = t; + MemWrite( &item->lockRelease.time, dt ); + break; + } + case QueueType::LockName: + { + ptr = MemRead( &item->lockNameFat.name ); + uint16_t size = MemRead( &item->lockNameFat.size ); + SendSingleString( (const char*)ptr, size ); +#ifndef TRACY_ON_DEMAND + tracy_free_fast( (void*)ptr ); +#endif + break; + } + case QueueType::MemAlloc: + case QueueType::MemAllocNamed: + case QueueType::MemAllocCallstack: + case QueueType::MemAllocCallstackNamed: + { + int64_t t = MemRead( &item->memAlloc.time ); + int64_t dt = t - refSerial; + refSerial = t; + MemWrite( &item->memAlloc.time, dt ); + break; + } + case QueueType::MemFree: + case QueueType::MemFreeNamed: + case QueueType::MemFreeCallstack: + case QueueType::MemFreeCallstackNamed: + { + int64_t t = MemRead( &item->memFree.time ); + int64_t dt = t - refSerial; + refSerial = t; + MemWrite( &item->memFree.time, dt ); + break; + } + case QueueType::GpuZoneBeginSerial: + case QueueType::GpuZoneBeginCallstackSerial: + { + int64_t t = MemRead( &item->gpuZoneBegin.cpuTime ); + int64_t dt = t - refSerial; + refSerial = t; + MemWrite( &item->gpuZoneBegin.cpuTime, dt ); + break; + } + case QueueType::GpuZoneBeginAllocSrcLocSerial: + case QueueType::GpuZoneBeginAllocSrcLocCallstackSerial: + { + int64_t t = MemRead( &item->gpuZoneBegin.cpuTime ); + int64_t dt = t - refSerial; + refSerial = t; + MemWrite( &item->gpuZoneBegin.cpuTime, dt ); + ptr = MemRead( &item->gpuZoneBegin.srcloc ); + SendSourceLocationPayload( ptr ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::GpuZoneEndSerial: + { + int64_t t = MemRead( &item->gpuZoneEnd.cpuTime ); + int64_t dt = t - refSerial; + refSerial = t; + MemWrite( &item->gpuZoneEnd.cpuTime, dt ); + break; + } + case QueueType::GpuTime: + { + int64_t t = MemRead( &item->gpuTime.gpuTime ); + int64_t dt = t - refGpu; + refGpu = t; + MemWrite( &item->gpuTime.gpuTime, dt ); + break; + } + case QueueType::GpuContextName: + { + ptr = MemRead( &item->gpuContextNameFat.ptr ); + uint16_t size = MemRead( &item->gpuContextNameFat.size ); + SendSingleString( (const char*)ptr, size ); +#ifndef TRACY_ON_DEMAND + tracy_free_fast( (void*)ptr ); +#endif + break; + } +#ifdef TRACY_FIBERS + case QueueType::ZoneBegin: + case QueueType::ZoneBeginCallstack: + { + ThreadCtxCheckSerial( zoneBeginThread ); + int64_t t = MemRead( &item->zoneBegin.time ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->zoneBegin.time, dt ); + break; + } + case QueueType::ZoneBeginAllocSrcLoc: + case QueueType::ZoneBeginAllocSrcLocCallstack: + { + ThreadCtxCheckSerial( zoneBeginThread ); + int64_t t = MemRead( &item->zoneBegin.time ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->zoneBegin.time, dt ); + ptr = MemRead( &item->zoneBegin.srcloc ); + SendSourceLocationPayload( ptr ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::ZoneEnd: + { + ThreadCtxCheckSerial( zoneEndThread ); + int64_t t = MemRead( &item->zoneEnd.time ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->zoneEnd.time, dt ); + break; + } + case QueueType::ZoneText: + case QueueType::ZoneName: + { + ThreadCtxCheckSerial( zoneTextFatThread ); + ptr = MemRead( &item->zoneTextFat.text ); + uint16_t size = MemRead( &item->zoneTextFat.size ); + SendSingleString( (const char*)ptr, size ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::Message: + case QueueType::MessageCallstack: + { + ThreadCtxCheckSerial( messageFatThread ); + ptr = MemRead( &item->messageFat.text ); + uint16_t size = MemRead( &item->messageFat.size ); + SendSingleString( (const char*)ptr, size ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::MessageColor: + case QueueType::MessageColorCallstack: + { + ThreadCtxCheckSerial( messageColorFatThread ); + ptr = MemRead( &item->messageColorFat.text ); + uint16_t size = MemRead( &item->messageColorFat.size ); + SendSingleString( (const char*)ptr, size ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::Callstack: + { + ThreadCtxCheckSerial( callstackFatThread ); + ptr = MemRead( &item->callstackFat.ptr ); + SendCallstackPayload( ptr ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::CallstackAlloc: + { + ThreadCtxCheckSerial( callstackAllocFatThread ); + ptr = MemRead( &item->callstackAllocFat.nativePtr ); + if( ptr != 0 ) + { + CutCallstack( (void*)ptr, "lua_pcall" ); + SendCallstackPayload( ptr ); + tracy_free_fast( (void*)ptr ); + } + ptr = MemRead( &item->callstackAllocFat.ptr ); + SendCallstackAlloc( ptr ); + tracy_free_fast( (void*)ptr ); + break; + } + case QueueType::FiberEnter: + { + ThreadCtxCheckSerial( fiberEnter ); + int64_t t = MemRead( &item->fiberEnter.time ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->fiberEnter.time, dt ); + break; + } + case QueueType::FiberLeave: + { + ThreadCtxCheckSerial( fiberLeave ); + int64_t t = MemRead( &item->fiberLeave.time ); + int64_t dt = t - refThread; + refThread = t; + MemWrite( &item->fiberLeave.time, dt ); + break; + } +#endif + default: + assert( false ); + break; + } + } +#ifdef TRACY_FIBERS + else + { + switch( (QueueType)idx ) + { + case QueueType::ZoneColor: + { + ThreadCtxCheckSerial( zoneColorThread ); + break; + } + case QueueType::ZoneValue: + { + ThreadCtxCheckSerial( zoneValueThread ); + break; + } + case QueueType::ZoneValidation: + { + ThreadCtxCheckSerial( zoneValidationThread ); + break; + } + case QueueType::MessageLiteral: + case QueueType::MessageLiteralCallstack: + { + ThreadCtxCheckSerial( messageLiteralThread ); + break; + } + case QueueType::MessageLiteralColor: + case QueueType::MessageLiteralColorCallstack: + { + ThreadCtxCheckSerial( messageColorLiteralThread ); + break; + } + case QueueType::CrashReport: + { + ThreadCtxCheckSerial( crashReportThread ); + break; + } + default: + break; + } + } +#endif + if( !AppendData( item, QueueDataSize[idx] ) ) return DequeueStatus::ConnectionLost; + item++; + } + m_refTimeSerial = refSerial; + m_refTimeGpu = refGpu; +#ifdef TRACY_FIBERS + m_refTimeThread = refThread; +#endif + m_serialDequeue.clear(); + } + else + { + return DequeueStatus::QueueEmpty; + } + return DequeueStatus::DataDequeued; +} + +Profiler::ThreadCtxStatus Profiler::ThreadCtxCheck( uint32_t threadId ) +{ + if( m_threadCtx == threadId ) return ThreadCtxStatus::Same; + QueueItem item; + MemWrite( &item.hdr.type, QueueType::ThreadContext ); + MemWrite( &item.threadCtx.thread, threadId ); + if( !AppendData( &item, QueueDataSize[(int)QueueType::ThreadContext] ) ) return ThreadCtxStatus::ConnectionLost; + m_threadCtx = threadId; + m_refTimeThread = 0; + return ThreadCtxStatus::Changed; +} + +bool Profiler::CommitData() +{ + bool ret = SendData( m_buffer + m_bufferStart, m_bufferOffset - m_bufferStart ); + if( m_bufferOffset > TargetFrameSize * 2 ) m_bufferOffset = 0; + m_bufferStart = m_bufferOffset; + return ret; +} + +bool Profiler::SendData( const char* data, size_t len ) +{ + const lz4sz_t lz4sz = LZ4_compress_fast_continue( (LZ4_stream_t*)m_stream, data, m_lz4Buf + sizeof( lz4sz_t ), (int)len, LZ4Size, 1 ); + memcpy( m_lz4Buf, &lz4sz, sizeof( lz4sz ) ); + return m_sock->Send( m_lz4Buf, lz4sz + sizeof( lz4sz_t ) ) != -1; +} + +void Profiler::SendString( uint64_t str, const char* ptr, size_t len, QueueType type ) +{ + assert( type == QueueType::StringData || + type == QueueType::ThreadName || + type == QueueType::PlotName || + type == QueueType::FrameName || + type == QueueType::ExternalName || + type == QueueType::ExternalThreadName || + type == QueueType::FiberName ); + + QueueItem item; + MemWrite( &item.hdr.type, type ); + MemWrite( &item.stringTransfer.ptr, str ); + + assert( len <= std::numeric_limits::max() ); + auto l16 = uint16_t( len ); + + NeedDataSize( QueueDataSize[(int)type] + sizeof( l16 ) + l16 ); + + AppendDataUnsafe( &item, QueueDataSize[(int)type] ); + AppendDataUnsafe( &l16, sizeof( l16 ) ); + AppendDataUnsafe( ptr, l16 ); +} + +void Profiler::SendSingleString( const char* ptr, size_t len ) +{ + QueueItem item; + MemWrite( &item.hdr.type, QueueType::SingleStringData ); + + assert( len <= std::numeric_limits::max() ); + auto l16 = uint16_t( len ); + + NeedDataSize( QueueDataSize[(int)QueueType::SingleStringData] + sizeof( l16 ) + l16 ); + + AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::SingleStringData] ); + AppendDataUnsafe( &l16, sizeof( l16 ) ); + AppendDataUnsafe( ptr, l16 ); +} + +void Profiler::SendSecondString( const char* ptr, size_t len ) +{ + QueueItem item; + MemWrite( &item.hdr.type, QueueType::SecondStringData ); + + assert( len <= std::numeric_limits::max() ); + auto l16 = uint16_t( len ); + + NeedDataSize( QueueDataSize[(int)QueueType::SecondStringData] + sizeof( l16 ) + l16 ); + + AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::SecondStringData] ); + AppendDataUnsafe( &l16, sizeof( l16 ) ); + AppendDataUnsafe( ptr, l16 ); +} + +void Profiler::SendLongString( uint64_t str, const char* ptr, size_t len, QueueType type ) +{ + assert( type == QueueType::FrameImageData || + type == QueueType::SymbolCode || + type == QueueType::SourceCode ); + + QueueItem item; + MemWrite( &item.hdr.type, type ); + MemWrite( &item.stringTransfer.ptr, str ); + + assert( len <= std::numeric_limits::max() ); + assert( QueueDataSize[(int)type] + sizeof( uint32_t ) + len <= TargetFrameSize ); + auto l32 = uint32_t( len ); + + NeedDataSize( QueueDataSize[(int)type] + sizeof( l32 ) + l32 ); + + AppendDataUnsafe( &item, QueueDataSize[(int)type] ); + AppendDataUnsafe( &l32, sizeof( l32 ) ); + AppendDataUnsafe( ptr, l32 ); +} + +void Profiler::SendSourceLocation( uint64_t ptr ) +{ + auto srcloc = (const SourceLocationData*)ptr; + QueueItem item; + MemWrite( &item.hdr.type, QueueType::SourceLocation ); + MemWrite( &item.srcloc.name, (uint64_t)srcloc->name ); + MemWrite( &item.srcloc.file, (uint64_t)srcloc->file ); + MemWrite( &item.srcloc.function, (uint64_t)srcloc->function ); + MemWrite( &item.srcloc.line, srcloc->line ); + MemWrite( &item.srcloc.b, uint8_t( ( srcloc->color ) & 0xFF ) ); + MemWrite( &item.srcloc.g, uint8_t( ( srcloc->color >> 8 ) & 0xFF ) ); + MemWrite( &item.srcloc.r, uint8_t( ( srcloc->color >> 16 ) & 0xFF ) ); + AppendData( &item, QueueDataSize[(int)QueueType::SourceLocation] ); +} + +void Profiler::SendSourceLocationPayload( uint64_t _ptr ) +{ + auto ptr = (const char*)_ptr; + + QueueItem item; + MemWrite( &item.hdr.type, QueueType::SourceLocationPayload ); + MemWrite( &item.stringTransfer.ptr, _ptr ); + + uint16_t len; + memcpy( &len, ptr, sizeof( len ) ); + assert( len > 2 ); + len -= 2; + ptr += 2; + + NeedDataSize( QueueDataSize[(int)QueueType::SourceLocationPayload] + sizeof( len ) + len ); + + AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::SourceLocationPayload] ); + AppendDataUnsafe( &len, sizeof( len ) ); + AppendDataUnsafe( ptr, len ); +} + +void Profiler::SendCallstackPayload( uint64_t _ptr ) +{ + auto ptr = (uintptr_t*)_ptr; + + QueueItem item; + MemWrite( &item.hdr.type, QueueType::CallstackPayload ); + MemWrite( &item.stringTransfer.ptr, _ptr ); + + const auto sz = *ptr++; + const auto len = sz * sizeof( uint64_t ); + const auto l16 = uint16_t( len ); + + NeedDataSize( QueueDataSize[(int)QueueType::CallstackPayload] + sizeof( l16 ) + l16 ); + + AppendDataUnsafe( &item, QueueDataSize[(int)QueueType::CallstackPayload] ); + AppendDataUnsafe( &l16, sizeof( l16 ) ); + + if( compile_time_condition::value ) + { + AppendDataUnsafe( ptr, sizeof( uint64_t ) * sz ); + } + else + { + for( uintptr_t i=0; i> 63 != 0 ) + { + SendSingleString( "" ); + QueueItem item; + MemWrite( &item.hdr.type, QueueType::SymbolInformation ); + MemWrite( &item.symbolInformation.line, 0 ); + MemWrite( &item.symbolInformation.symAddr, symbol ); + AppendData( &item, QueueDataSize[(int)QueueType::SymbolInformation] ); + } + else + { + m_symbolQueue.emplace( SymbolQueueItem { SymbolQueueItemType::SymbolQuery, symbol } ); + } +#else + AckServerQuery(); +#endif +} + +void Profiler::QueueExternalName( uint64_t ptr ) +{ +#ifdef TRACY_HAS_SYSTEM_TRACING + m_symbolQueue.emplace( SymbolQueueItem { SymbolQueueItemType::ExternalName, ptr } ); +#endif +} + +void Profiler::QueueKernelCode( uint64_t symbol, uint32_t size ) +{ + assert( symbol >> 63 != 0 ); +#ifdef TRACY_HAS_CALLSTACK + m_symbolQueue.emplace( SymbolQueueItem { SymbolQueueItemType::KernelCode, symbol, size } ); +#else + AckSymbolCodeNotAvailable(); +#endif +} + +void Profiler::QueueSourceCodeQuery( uint32_t id ) +{ + assert( m_exectime != 0 ); + assert( m_queryData ); + m_symbolQueue.emplace( SymbolQueueItem { SymbolQueueItemType::SourceCode, uint64_t( m_queryData ), uint64_t( m_queryImage ), id } ); + m_queryData = nullptr; + m_queryImage = nullptr; +} + +#ifdef TRACY_HAS_CALLSTACK +void Profiler::HandleSymbolQueueItem( const SymbolQueueItem& si ) +{ + switch( si.type ) + { + case SymbolQueueItemType::CallstackFrame: + { + const auto frameData = DecodeCallstackPtr( si.ptr ); + auto data = tracy_malloc_fast( sizeof( CallstackEntry ) * frameData.size ); + memcpy( data, frameData.data, sizeof( CallstackEntry ) * frameData.size ); + TracyLfqPrepare( QueueType::CallstackFrameSize ); + MemWrite( &item->callstackFrameSizeFat.ptr, si.ptr ); + MemWrite( &item->callstackFrameSizeFat.size, frameData.size ); + MemWrite( &item->callstackFrameSizeFat.data, (uint64_t)data ); + MemWrite( &item->callstackFrameSizeFat.imageName, (uint64_t)frameData.imageName ); + TracyLfqCommit; + break; + } + case SymbolQueueItemType::SymbolQuery: + { +#ifdef __ANDROID__ + // On Android it's common for code to be in mappings that are only executable + // but not readable. + if( !EnsureReadable( si.ptr ) ) + { + TracyLfqPrepare( QueueType::AckServerQueryNoop ); + TracyLfqCommit; + break; + } +#endif + const auto sym = DecodeSymbolAddress( si.ptr ); + TracyLfqPrepare( QueueType::SymbolInformation ); + MemWrite( &item->symbolInformationFat.line, sym.line ); + MemWrite( &item->symbolInformationFat.symAddr, si.ptr ); + MemWrite( &item->symbolInformationFat.fileString, (uint64_t)sym.file ); + MemWrite( &item->symbolInformationFat.needFree, (uint8_t)sym.needFree ); + TracyLfqCommit; + break; + } +#ifdef TRACY_HAS_SYSTEM_TRACING + case SymbolQueueItemType::ExternalName: + { + const char* threadName; + const char* name; + SysTraceGetExternalName( si.ptr, threadName, name ); + TracyLfqPrepare( QueueType::ExternalNameMetadata ); + MemWrite( &item->externalNameMetadata.thread, si.ptr ); + MemWrite( &item->externalNameMetadata.name, (uint64_t)name ); + MemWrite( &item->externalNameMetadata.threadName, (uint64_t)threadName ); + TracyLfqCommit; + break; + } +#endif + case SymbolQueueItemType::KernelCode: + { +#ifdef _WIN32 + auto mod = GetKernelModulePath( si.ptr ); + if( mod ) + { + auto fn = DecodeCallstackPtrFast( si.ptr ); + if( *fn ) + { + auto hnd = LoadLibraryExA( mod, nullptr, DONT_RESOLVE_DLL_REFERENCES ); + if( hnd ) + { + auto ptr = (const void*)GetProcAddress( hnd, fn ); + if( ptr ) + { + auto buf = (char*)tracy_malloc( si.extra ); + memcpy( buf, ptr, si.extra ); + FreeLibrary( hnd ); + TracyLfqPrepare( QueueType::SymbolCodeMetadata ); + MemWrite( &item->symbolCodeMetadata.symbol, si.ptr ); + MemWrite( &item->symbolCodeMetadata.ptr, (uint64_t)buf ); + MemWrite( &item->symbolCodeMetadata.size, (uint32_t)si.extra ); + TracyLfqCommit; + break; + } + FreeLibrary( hnd ); + } + } + } +#endif + TracyLfqPrepare( QueueType::AckSymbolCodeNotAvailable ); + TracyLfqCommit; + break; + } + case SymbolQueueItemType::SourceCode: + HandleSourceCodeQuery( (char*)si.ptr, (char*)si.extra, si.id ); + break; + default: + assert( false ); + break; + } +} + +void Profiler::SymbolWorker() +{ +#if defined __linux__ && !defined TRACY_NO_CRASH_HANDLER + s_symbolTid = syscall( SYS_gettid ); +#endif + + ThreadExitHandler threadExitHandler; + SetThreadName( "Tracy Symbol Worker" ); +#ifdef TRACY_USE_RPMALLOC + InitRpmalloc(); +#endif + InitCallstack(); + while( m_timeBegin.load( std::memory_order_relaxed ) == 0 ) std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) ); + + for(;;) + { + const auto shouldExit = ShouldExit(); +#ifdef TRACY_ON_DEMAND + if( !IsConnected() ) + { + if( shouldExit ) + { + s_symbolThreadGone.store( true, std::memory_order_release ); + return; + } + while( m_symbolQueue.front() ) m_symbolQueue.pop(); + std::this_thread::sleep_for( std::chrono::milliseconds( 20 ) ); + continue; + } +#endif + auto si = m_symbolQueue.front(); + if( si ) + { + HandleSymbolQueueItem( *si ); + m_symbolQueue.pop(); + } + else + { + if( shouldExit ) + { + s_symbolThreadGone.store( true, std::memory_order_release ); + return; + } + std::this_thread::sleep_for( std::chrono::milliseconds( 20 ) ); + } + } +} +#endif + +bool Profiler::HandleServerQuery() +{ + ServerQueryPacket payload; + if( !m_sock->Read( &payload, sizeof( payload ), 10 ) ) return false; + + uint8_t type; + uint64_t ptr; + memcpy( &type, &payload.type, sizeof( payload.type ) ); + memcpy( &ptr, &payload.ptr, sizeof( payload.ptr ) ); + + switch( type ) + { + case ServerQueryString: + SendString( ptr, (const char*)ptr, QueueType::StringData ); + break; + case ServerQueryThreadString: + if( ptr == m_mainThread ) + { + SendString( ptr, "Main thread", 11, QueueType::ThreadName ); + } + else + { + SendString( ptr, GetThreadName( ptr ), QueueType::ThreadName ); + } + break; + case ServerQuerySourceLocation: + SendSourceLocation( ptr ); + break; + case ServerQueryPlotName: + SendString( ptr, (const char*)ptr, QueueType::PlotName ); + break; + case ServerQueryTerminate: + return false; + case ServerQueryCallstackFrame: + QueueCallstackFrame( ptr ); + break; + case ServerQueryFrameName: + SendString( ptr, (const char*)ptr, QueueType::FrameName ); + break; + case ServerQueryDisconnect: + HandleDisconnect(); + return false; +#ifdef TRACY_HAS_SYSTEM_TRACING + case ServerQueryExternalName: + QueueExternalName( ptr ); + break; +#endif + case ServerQueryParameter: + HandleParameter( ptr ); + break; + case ServerQuerySymbol: + QueueSymbolQuery( ptr ); + break; +#ifndef TRACY_NO_CODE_TRANSFER + case ServerQuerySymbolCode: + HandleSymbolCodeQuery( ptr, payload.extra ); + break; +#endif + case ServerQuerySourceCode: + QueueSourceCodeQuery( uint32_t( ptr ) ); + break; + case ServerQueryDataTransfer: + if( m_queryData ) + { + assert( !m_queryImage ); + m_queryImage = m_queryData; + } + m_queryDataPtr = m_queryData = (char*)tracy_malloc( ptr + 11 ); + AckServerQuery(); + break; + case ServerQueryDataTransferPart: + memcpy( m_queryDataPtr, &ptr, 8 ); + memcpy( m_queryDataPtr+8, &payload.extra, 4 ); + m_queryDataPtr += 12; + AckServerQuery(); + break; +#ifdef TRACY_FIBERS + case ServerQueryFiberName: + SendString( ptr, (const char*)ptr, QueueType::FiberName ); + break; +#endif + default: + assert( false ); + break; + } + + return true; +} + +void Profiler::HandleDisconnect() +{ + moodycamel::ConsumerToken token( GetQueue() ); + +#ifdef TRACY_HAS_SYSTEM_TRACING + if( s_sysTraceThread ) + { + auto timestamp = GetTime(); + for(;;) + { + const auto status = DequeueContextSwitches( token, timestamp ); + if( status == DequeueStatus::ConnectionLost ) + { + return; + } + else if( status == DequeueStatus::QueueEmpty ) + { + if( m_bufferOffset != m_bufferStart ) + { + if( !CommitData() ) return; + } + } + if( timestamp < 0 ) + { + if( m_bufferOffset != m_bufferStart ) + { + if( !CommitData() ) return; + } + break; + } + ClearSerial(); + if( m_sock->HasData() ) + { + while( m_sock->HasData() ) + { + if( !HandleServerQuery() ) return; + } + if( m_bufferOffset != m_bufferStart ) + { + if( !CommitData() ) return; + } + } + else + { + if( m_bufferOffset != m_bufferStart ) + { + if( !CommitData() ) return; + } + std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) ); + } + } + } +#endif + + QueueItem terminate; + MemWrite( &terminate.hdr.type, QueueType::Terminate ); + if( !SendData( (const char*)&terminate, 1 ) ) return; + for(;;) + { + ClearQueues( token ); + if( m_sock->HasData() ) + { + while( m_sock->HasData() ) + { + if( !HandleServerQuery() ) return; + } + if( m_bufferOffset != m_bufferStart ) + { + if( !CommitData() ) return; + } + } + else + { + if( m_bufferOffset != m_bufferStart ) + { + if( !CommitData() ) return; + } + std::this_thread::sleep_for( std::chrono::milliseconds( 10 ) ); + } + } +} + +void Profiler::CalibrateTimer() +{ + m_timerMul = 1.; + +#ifdef TRACY_HW_TIMER + +# if !defined TRACY_TIMER_QPC && defined TRACY_TIMER_FALLBACK + const bool needCalibration = HardwareSupportsInvariantTSC(); +# else + const bool needCalibration = true; +# endif + if( needCalibration ) + { + std::atomic_signal_fence( std::memory_order_acq_rel ); + const auto t0 = std::chrono::high_resolution_clock::now(); + const auto r0 = GetTime(); + std::atomic_signal_fence( std::memory_order_acq_rel ); + std::this_thread::sleep_for( std::chrono::milliseconds( 200 ) ); + std::atomic_signal_fence( std::memory_order_acq_rel ); + const auto t1 = std::chrono::high_resolution_clock::now(); + const auto r1 = GetTime(); + std::atomic_signal_fence( std::memory_order_acq_rel ); + + const auto dt = std::chrono::duration_cast( t1 - t0 ).count(); + const auto dr = r1 - r0; + + m_timerMul = double( dt ) / double( dr ); + } +#endif +} + +void Profiler::CalibrateDelay() +{ + constexpr int Iterations = 50000; + + auto mindiff = std::numeric_limits::max(); + for( int i=0; i 0 && dti < mindiff ) mindiff = dti; + } + m_resolution = mindiff; + +#ifdef TRACY_DELAYED_INIT + m_delay = m_resolution; +#else + constexpr int Events = Iterations * 2; // start + end + static_assert( Events < QueuePrealloc, "Delay calibration loop will allocate memory in queue" ); + + static const tracy::SourceLocationData __tracy_source_location { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; + const auto t0 = GetTime(); + for( int i=0; izoneBegin.time, Profiler::GetTime() ); + MemWrite( &item->zoneBegin.srcloc, (uint64_t)&__tracy_source_location ); + TracyLfqCommit; + } + { + TracyLfqPrepare( QueueType::ZoneEnd ); + MemWrite( &item->zoneEnd.time, GetTime() ); + TracyLfqCommit; + } + } + const auto t1 = GetTime(); + const auto dt = t1 - t0; + m_delay = dt / Events; + + moodycamel::ConsumerToken token( GetQueue() ); + int left = Events; + while( left != 0 ) + { + const auto sz = GetQueue().try_dequeue_bulk_single( token, [](const uint64_t&){}, [](QueueItem* item, size_t sz){} ); + assert( sz > 0 ); + left -= (int)sz; + } + assert( GetQueue().size_approx() == 0 ); +#endif +} + +void Profiler::ReportTopology() +{ +#ifndef TRACY_DELAYED_INIT + struct CpuData + { + uint32_t package; + uint32_t core; + uint32_t thread; + }; + +#if defined _WIN32 +# ifdef TRACY_UWP + t_GetLogicalProcessorInformationEx _GetLogicalProcessorInformationEx = &::GetLogicalProcessorInformationEx; +# else + t_GetLogicalProcessorInformationEx _GetLogicalProcessorInformationEx = (t_GetLogicalProcessorInformationEx)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "GetLogicalProcessorInformationEx" ); +# endif + if( !_GetLogicalProcessorInformationEx ) return; + + DWORD psz = 0; + _GetLogicalProcessorInformationEx( RelationProcessorPackage, nullptr, &psz ); + auto packageInfo = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)tracy_malloc( psz ); + auto res = _GetLogicalProcessorInformationEx( RelationProcessorPackage, packageInfo, &psz ); + assert( res ); + + DWORD csz = 0; + _GetLogicalProcessorInformationEx( RelationProcessorCore, nullptr, &csz ); + auto coreInfo = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)tracy_malloc( csz ); + res = _GetLogicalProcessorInformationEx( RelationProcessorCore, coreInfo, &csz ); + assert( res ); + + SYSTEM_INFO sysinfo; + GetSystemInfo( &sysinfo ); + const uint32_t numcpus = sysinfo.dwNumberOfProcessors; + + auto cpuData = (CpuData*)tracy_malloc( sizeof( CpuData ) * numcpus ); + for( uint32_t i=0; iRelationship == RelationProcessorPackage ); + // FIXME account for GroupCount + auto mask = ptr->Processor.GroupMask[0].Mask; + int core = 0; + while( mask != 0 ) + { + if( mask & 1 ) cpuData[core].package = idx; + core++; + mask >>= 1; + } + ptr = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)(((char*)ptr) + ptr->Size); + idx++; + } + + idx = 0; + ptr = coreInfo; + while( (char*)ptr < ((char*)coreInfo) + csz ) + { + assert( ptr->Relationship == RelationProcessorCore ); + // FIXME account for GroupCount + auto mask = ptr->Processor.GroupMask[0].Mask; + int core = 0; + while( mask != 0 ) + { + if( mask & 1 ) cpuData[core].core = idx; + core++; + mask >>= 1; + } + ptr = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)(((char*)ptr) + ptr->Size); + idx++; + } + + for( uint32_t i=0; icpuTopology.package, data.package ); + MemWrite( &item->cpuTopology.core, data.core ); + MemWrite( &item->cpuTopology.thread, data.thread ); + +#ifdef TRACY_ON_DEMAND + DeferItem( *item ); +#endif + + TracyLfqCommit; + } + + tracy_free( cpuData ); + tracy_free( coreInfo ); + tracy_free( packageInfo ); +#elif defined __linux__ + const int numcpus = std::thread::hardware_concurrency(); + auto cpuData = (CpuData*)tracy_malloc( sizeof( CpuData ) * numcpus ); + memset( cpuData, 0, sizeof( CpuData ) * numcpus ); + + const char* basePath = "/sys/devices/system/cpu/cpu"; + for( int i=0; icpuTopology.package, data.package ); + MemWrite( &item->cpuTopology.core, data.core ); + MemWrite( &item->cpuTopology.thread, data.thread ); + +#ifdef TRACY_ON_DEMAND + DeferItem( *item ); +#endif + + TracyLfqCommit; + } + + tracy_free( cpuData ); +#endif +#endif +} + +void Profiler::SendCallstack( int depth, const char* skipBefore ) +{ +#ifdef TRACY_HAS_CALLSTACK + auto ptr = Callstack( depth ); + CutCallstack( ptr, skipBefore ); + + TracyQueuePrepare( QueueType::Callstack ); + MemWrite( &item->callstackFat.ptr, (uint64_t)ptr ); + TracyQueueCommit( callstackFatThread ); +#endif +} + +void Profiler::CutCallstack( void* callstack, const char* skipBefore ) +{ +#ifdef TRACY_HAS_CALLSTACK + auto data = (uintptr_t*)callstack; + const auto sz = *data++; + uintptr_t i; + for( i=0; i 100000000 ) // 100 ms + { + auto sysTime = m_sysTime.Get(); + if( sysTime >= 0 ) + { + m_sysTimeLast = t; + + TracyLfqPrepare( QueueType::SysTimeReport ); + MemWrite( &item->sysTime.time, GetTime() ); + MemWrite( &item->sysTime.sysTime, sysTime ); + TracyLfqCommit; + } + } +} +#endif + +void Profiler::HandleParameter( uint64_t payload ) +{ + assert( m_paramCallback ); + const auto idx = uint32_t( payload >> 32 ); + const auto val = int32_t( payload & 0xFFFFFFFF ); + m_paramCallback( m_paramCallbackData, idx, val ); + AckServerQuery(); +} + +void Profiler::HandleSymbolCodeQuery( uint64_t symbol, uint32_t size ) +{ + if( symbol >> 63 != 0 ) + { + QueueKernelCode( symbol, size ); + } + else + { +#ifdef __ANDROID__ + // On Android it's common for code to be in mappings that are only executable + // but not readable. + if( !EnsureReadable( symbol ) ) + { + AckSymbolCodeNotAvailable(); + return; + } +#endif + SendLongString( symbol, (const char*)symbol, size, QueueType::SymbolCode ); + } +} + +void Profiler::HandleSourceCodeQuery( char* data, char* image, uint32_t id ) +{ + bool ok = false; + struct stat st; + if( stat( data, &st ) == 0 && (uint64_t)st.st_mtime < m_exectime ) + { + if( st.st_size < ( TargetFrameSize - 16 ) ) + { + FILE* f = fopen( data, "rb" ); + if( f ) + { + auto ptr = (char*)tracy_malloc_fast( st.st_size ); + auto rd = fread( ptr, 1, st.st_size, f ); + fclose( f ); + if( rd == (size_t)st.st_size ) + { + TracyLfqPrepare( QueueType::SourceCodeMetadata ); + MemWrite( &item->sourceCodeMetadata.ptr, (uint64_t)ptr ); + MemWrite( &item->sourceCodeMetadata.size, (uint32_t)rd ); + MemWrite( &item->sourceCodeMetadata.id, id ); + TracyLfqCommit; + ok = true; + } + } + } + } + +#ifdef TRACY_DEBUGINFOD + else if( image && data[0] == '/' ) + { + size_t size; + auto buildid = GetBuildIdForImage( image, size ); + if( buildid ) + { + auto d = debuginfod_find_source( GetDebuginfodClient(), buildid, size, data, nullptr ); + TracyDebug( "DebugInfo source query: %s, fn: %s, image: %s\n", d >= 0 ? " ok " : "fail", data, image ); + if( d >= 0 ) + { + struct stat st; + fstat( d, &st ); + if( st.st_size < ( TargetFrameSize - 16 ) ) + { + lseek( d, 0, SEEK_SET ); + auto ptr = (char*)tracy_malloc_fast( st.st_size ); + auto rd = read( d, ptr, st.st_size ); + if( rd == (size_t)st.st_size ) + { + TracyLfqPrepare( QueueType::SourceCodeMetadata ); + MemWrite( &item->sourceCodeMetadata.ptr, (uint64_t)ptr ); + MemWrite( &item->sourceCodeMetadata.size, (uint32_t)rd ); + MemWrite( &item->sourceCodeMetadata.id, id ); + TracyLfqCommit; + ok = true; + } + } + close( d ); + } + } + } + else + { + TracyDebug( "DebugInfo invalid query fn: %s, image: %s\n", data, image ); + } +#endif + + if( !ok && m_sourceCallback ) + { + size_t sz; + char* ptr = m_sourceCallback( m_sourceCallbackData, data, sz ); + if( ptr ) + { + if( sz < ( TargetFrameSize - 16 ) ) + { + TracyLfqPrepare( QueueType::SourceCodeMetadata ); + MemWrite( &item->sourceCodeMetadata.ptr, (uint64_t)ptr ); + MemWrite( &item->sourceCodeMetadata.size, (uint32_t)sz ); + MemWrite( &item->sourceCodeMetadata.id, id ); + TracyLfqCommit; + ok = true; + } + } + } + + if( !ok ) + { + TracyLfqPrepare( QueueType::AckSourceCodeNotAvailable ); + MemWrite( &item->sourceCodeNotAvailable, id ); + TracyLfqCommit; + } + + tracy_free_fast( data ); + tracy_free_fast( image ); +} + +#if defined _WIN32 && defined TRACY_TIMER_QPC +int64_t Profiler::GetTimeQpc() +{ + LARGE_INTEGER t; + QueryPerformanceCounter( &t ); + return t.QuadPart; +} +#endif + +} + +#ifdef __cplusplus +extern "C" { +#endif + +TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin( const struct ___tracy_source_location_data* srcloc, int active ) +{ + ___tracy_c_zone_context ctx; +#ifdef TRACY_ON_DEMAND + ctx.active = active && tracy::GetProfiler().IsConnected(); +#else + ctx.active = active; +#endif + if( !ctx.active ) return ctx; + const auto id = tracy::GetProfiler().GetNextZoneId(); + ctx.id = id; + +#ifndef TRACY_NO_VERIFY + { + TracyQueuePrepareC( tracy::QueueType::ZoneValidation ); + tracy::MemWrite( &item->zoneValidation.id, id ); + TracyQueueCommitC( zoneValidationThread ); + } +#endif + { + TracyQueuePrepareC( tracy::QueueType::ZoneBegin ); + tracy::MemWrite( &item->zoneBegin.time, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->zoneBegin.srcloc, (uint64_t)srcloc ); + TracyQueueCommitC( zoneBeginThread ); + } + return ctx; +} + +TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin_callstack( const struct ___tracy_source_location_data* srcloc, int depth, int active ) +{ + ___tracy_c_zone_context ctx; +#ifdef TRACY_ON_DEMAND + ctx.active = active && tracy::GetProfiler().IsConnected(); +#else + ctx.active = active; +#endif + if( !ctx.active ) return ctx; + const auto id = tracy::GetProfiler().GetNextZoneId(); + ctx.id = id; + +#ifndef TRACY_NO_VERIFY + { + TracyQueuePrepareC( tracy::QueueType::ZoneValidation ); + tracy::MemWrite( &item->zoneValidation.id, id ); + TracyQueueCommitC( zoneValidationThread ); + } +#endif + tracy::GetProfiler().SendCallstack( depth ); + { + TracyQueuePrepareC( tracy::QueueType::ZoneBeginCallstack ); + tracy::MemWrite( &item->zoneBegin.time, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->zoneBegin.srcloc, (uint64_t)srcloc ); + TracyQueueCommitC( zoneBeginThread ); + } + return ctx; +} + +TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin_alloc( uint64_t srcloc, int active ) +{ + ___tracy_c_zone_context ctx; +#ifdef TRACY_ON_DEMAND + ctx.active = active && tracy::GetProfiler().IsConnected(); +#else + ctx.active = active; +#endif + if( !ctx.active ) + { + tracy::tracy_free( (void*)srcloc ); + return ctx; + } + const auto id = tracy::GetProfiler().GetNextZoneId(); + ctx.id = id; + +#ifndef TRACY_NO_VERIFY + { + TracyQueuePrepareC( tracy::QueueType::ZoneValidation ); + tracy::MemWrite( &item->zoneValidation.id, id ); + TracyQueueCommitC( zoneValidationThread ); + } +#endif + { + TracyQueuePrepareC( tracy::QueueType::ZoneBeginAllocSrcLoc ); + tracy::MemWrite( &item->zoneBegin.time, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->zoneBegin.srcloc, srcloc ); + TracyQueueCommitC( zoneBeginThread ); + } + return ctx; +} + +TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin_alloc_callstack( uint64_t srcloc, int depth, int active ) +{ + ___tracy_c_zone_context ctx; +#ifdef TRACY_ON_DEMAND + ctx.active = active && tracy::GetProfiler().IsConnected(); +#else + ctx.active = active; +#endif + if( !ctx.active ) + { + tracy::tracy_free( (void*)srcloc ); + return ctx; + } + const auto id = tracy::GetProfiler().GetNextZoneId(); + ctx.id = id; + +#ifndef TRACY_NO_VERIFY + { + TracyQueuePrepareC( tracy::QueueType::ZoneValidation ); + tracy::MemWrite( &item->zoneValidation.id, id ); + TracyQueueCommitC( zoneValidationThread ); + } +#endif + tracy::GetProfiler().SendCallstack( depth ); + { + TracyQueuePrepareC( tracy::QueueType::ZoneBeginAllocSrcLocCallstack ); + tracy::MemWrite( &item->zoneBegin.time, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->zoneBegin.srcloc, srcloc ); + TracyQueueCommitC( zoneBeginThread ); + } + return ctx; +} + +TRACY_API void ___tracy_emit_zone_end( TracyCZoneCtx ctx ) +{ + if( !ctx.active ) return; +#ifndef TRACY_NO_VERIFY + { + TracyQueuePrepareC( tracy::QueueType::ZoneValidation ); + tracy::MemWrite( &item->zoneValidation.id, ctx.id ); + TracyQueueCommitC( zoneValidationThread ); + } +#endif + { + TracyQueuePrepareC( tracy::QueueType::ZoneEnd ); + tracy::MemWrite( &item->zoneEnd.time, tracy::Profiler::GetTime() ); + TracyQueueCommitC( zoneEndThread ); + } +} + +TRACY_API void ___tracy_emit_zone_text( TracyCZoneCtx ctx, const char* txt, size_t size ) +{ + assert( size < std::numeric_limits::max() ); + if( !ctx.active ) return; + auto ptr = (char*)tracy::tracy_malloc( size ); + memcpy( ptr, txt, size ); +#ifndef TRACY_NO_VERIFY + { + TracyQueuePrepareC( tracy::QueueType::ZoneValidation ); + tracy::MemWrite( &item->zoneValidation.id, ctx.id ); + TracyQueueCommitC( zoneValidationThread ); + } +#endif + { + TracyQueuePrepareC( tracy::QueueType::ZoneText ); + tracy::MemWrite( &item->zoneTextFat.text, (uint64_t)ptr ); + tracy::MemWrite( &item->zoneTextFat.size, (uint16_t)size ); + TracyQueueCommitC( zoneTextFatThread ); + } +} + +TRACY_API void ___tracy_emit_zone_name( TracyCZoneCtx ctx, const char* txt, size_t size ) +{ + assert( size < std::numeric_limits::max() ); + if( !ctx.active ) return; + auto ptr = (char*)tracy::tracy_malloc( size ); + memcpy( ptr, txt, size ); +#ifndef TRACY_NO_VERIFY + { + TracyQueuePrepareC( tracy::QueueType::ZoneValidation ); + tracy::MemWrite( &item->zoneValidation.id, ctx.id ); + TracyQueueCommitC( zoneValidationThread ); + } +#endif + { + TracyQueuePrepareC( tracy::QueueType::ZoneName ); + tracy::MemWrite( &item->zoneTextFat.text, (uint64_t)ptr ); + tracy::MemWrite( &item->zoneTextFat.size, (uint16_t)size ); + TracyQueueCommitC( zoneTextFatThread ); + } +} + +TRACY_API void ___tracy_emit_zone_color( TracyCZoneCtx ctx, uint32_t color ) { + if( !ctx.active ) return; +#ifndef TRACY_NO_VERIFY + { + TracyQueuePrepareC( tracy::QueueType::ZoneValidation ); + tracy::MemWrite( &item->zoneValidation.id, ctx.id ); + TracyQueueCommitC( zoneValidationThread ); + } +#endif + { + TracyQueuePrepareC( tracy::QueueType::ZoneColor ); + tracy::MemWrite( &item->zoneColor.b, uint8_t( ( color ) & 0xFF ) ); + tracy::MemWrite( &item->zoneColor.g, uint8_t( ( color >> 8 ) & 0xFF ) ); + tracy::MemWrite( &item->zoneColor.r, uint8_t( ( color >> 16 ) & 0xFF ) ); + TracyQueueCommitC( zoneColorThread ); + } +} + +TRACY_API void ___tracy_emit_zone_value( TracyCZoneCtx ctx, uint64_t value ) +{ + if( !ctx.active ) return; +#ifndef TRACY_NO_VERIFY + { + TracyQueuePrepareC( tracy::QueueType::ZoneValidation ); + tracy::MemWrite( &item->zoneValidation.id, ctx.id ); + TracyQueueCommitC( zoneValidationThread ); + } +#endif + { + TracyQueuePrepareC( tracy::QueueType::ZoneValue ); + tracy::MemWrite( &item->zoneValue.value, value ); + TracyQueueCommitC( zoneValueThread ); + } +} + +TRACY_API void ___tracy_emit_memory_alloc( const void* ptr, size_t size, int secure ) { tracy::Profiler::MemAlloc( ptr, size, secure != 0 ); } +TRACY_API void ___tracy_emit_memory_alloc_callstack( const void* ptr, size_t size, int depth, int secure ) { tracy::Profiler::MemAllocCallstack( ptr, size, depth, secure != 0 ); } +TRACY_API void ___tracy_emit_memory_free( const void* ptr, int secure ) { tracy::Profiler::MemFree( ptr, secure != 0 ); } +TRACY_API void ___tracy_emit_memory_free_callstack( const void* ptr, int depth, int secure ) { tracy::Profiler::MemFreeCallstack( ptr, depth, secure != 0 ); } +TRACY_API void ___tracy_emit_memory_alloc_named( const void* ptr, size_t size, int secure, const char* name ) { tracy::Profiler::MemAllocNamed( ptr, size, secure != 0, name ); } +TRACY_API void ___tracy_emit_memory_alloc_callstack_named( const void* ptr, size_t size, int depth, int secure, const char* name ) { tracy::Profiler::MemAllocCallstackNamed( ptr, size, depth, secure != 0, name ); } +TRACY_API void ___tracy_emit_memory_free_named( const void* ptr, int secure, const char* name ) { tracy::Profiler::MemFreeNamed( ptr, secure != 0, name ); } +TRACY_API void ___tracy_emit_memory_free_callstack_named( const void* ptr, int depth, int secure, const char* name ) { tracy::Profiler::MemFreeCallstackNamed( ptr, depth, secure != 0, name ); } +TRACY_API void ___tracy_emit_frame_mark( const char* name ) { tracy::Profiler::SendFrameMark( name ); } +TRACY_API void ___tracy_emit_frame_mark_start( const char* name ) { tracy::Profiler::SendFrameMark( name, tracy::QueueType::FrameMarkMsgStart ); } +TRACY_API void ___tracy_emit_frame_mark_end( const char* name ) { tracy::Profiler::SendFrameMark( name, tracy::QueueType::FrameMarkMsgEnd ); } +TRACY_API void ___tracy_emit_frame_image( const void* image, uint16_t w, uint16_t h, uint8_t offset, int flip ) { tracy::Profiler::SendFrameImage( image, w, h, offset, flip ); } +TRACY_API void ___tracy_emit_plot( const char* name, double val ) { tracy::Profiler::PlotData( name, val ); } +TRACY_API void ___tracy_emit_plot_float( const char* name, float val ) { tracy::Profiler::PlotData( name, val ); } +TRACY_API void ___tracy_emit_plot_int( const char* name, int64_t val ) { tracy::Profiler::PlotData( name, val ); } +TRACY_API void ___tracy_emit_plot_config( const char* name, int type, int step, int fill, uint32_t color ) { tracy::Profiler::ConfigurePlot( name, tracy::PlotFormatType(type), step, fill, color ); } +TRACY_API void ___tracy_emit_message( const char* txt, size_t size, int callstack ) { tracy::Profiler::Message( txt, size, callstack ); } +TRACY_API void ___tracy_emit_messageL( const char* txt, int callstack ) { tracy::Profiler::Message( txt, callstack ); } +TRACY_API void ___tracy_emit_messageC( const char* txt, size_t size, uint32_t color, int callstack ) { tracy::Profiler::MessageColor( txt, size, color, callstack ); } +TRACY_API void ___tracy_emit_messageLC( const char* txt, uint32_t color, int callstack ) { tracy::Profiler::MessageColor( txt, color, callstack ); } +TRACY_API void ___tracy_emit_message_appinfo( const char* txt, size_t size ) { tracy::Profiler::MessageAppInfo( txt, size ); } + +TRACY_API uint64_t ___tracy_alloc_srcloc( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz ) { + return tracy::Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz ); +} + +TRACY_API uint64_t ___tracy_alloc_srcloc_name( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz ) { + return tracy::Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz ); +} + +TRACY_API void ___tracy_emit_gpu_zone_begin( const struct ___tracy_gpu_zone_begin_data data ) +{ + TracyLfqPrepareC( tracy::QueueType::GpuZoneBegin ); + tracy::MemWrite( &item->gpuZoneBegin.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuZoneBegin.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuZoneBegin.srcloc, data.srcloc ); + tracy::MemWrite( &item->gpuZoneBegin.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneBegin.context, data.context ); + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_zone_begin_callstack( const struct ___tracy_gpu_zone_begin_callstack_data data ) +{ + tracy::GetProfiler().SendCallstack( data.depth ); + TracyLfqPrepareC( tracy::QueueType::GpuZoneBeginCallstack ); + tracy::MemWrite( &item->gpuZoneBegin.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuZoneBegin.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuZoneBegin.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneBegin.context, data.context ); + tracy::MemWrite( &item->gpuZoneBegin.srcloc, data.srcloc ); + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_zone_begin_alloc( const struct ___tracy_gpu_zone_begin_data data ) +{ + TracyLfqPrepareC( tracy::QueueType::GpuZoneBeginAllocSrcLoc ); + tracy::MemWrite( &item->gpuZoneBegin.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuZoneBegin.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuZoneBegin.srcloc, data.srcloc ); + tracy::MemWrite( &item->gpuZoneBegin.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneBegin.context, data.context ); + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_zone_begin_alloc_callstack( const struct ___tracy_gpu_zone_begin_callstack_data data ) +{ + tracy::GetProfiler().SendCallstack( data.depth ); + TracyLfqPrepareC( tracy::QueueType::GpuZoneBeginAllocSrcLocCallstack ); + tracy::MemWrite( &item->gpuZoneBegin.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuZoneBegin.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuZoneBegin.srcloc, data.srcloc ); + tracy::MemWrite( &item->gpuZoneBegin.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneBegin.context, data.context ); + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_time( const struct ___tracy_gpu_time_data data ) +{ + TracyLfqPrepareC( tracy::QueueType::GpuTime ); + tracy::MemWrite( &item->gpuTime.gpuTime, data.gpuTime ); + tracy::MemWrite( &item->gpuTime.queryId, data.queryId ); + tracy::MemWrite( &item->gpuTime.context, data.context ); + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_zone_end( const struct ___tracy_gpu_zone_end_data data ) +{ + TracyLfqPrepareC( tracy::QueueType::GpuZoneEnd ); + tracy::MemWrite( &item->gpuZoneEnd.cpuTime, tracy::Profiler::GetTime() ); + memset( &item->gpuZoneEnd.thread, 0, sizeof( item->gpuZoneEnd.thread ) ); + tracy::MemWrite( &item->gpuZoneEnd.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneEnd.context, data.context ); + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_new_context( ___tracy_gpu_new_context_data data ) +{ + TracyLfqPrepareC( tracy::QueueType::GpuNewContext ); + tracy::MemWrite( &item->gpuNewContext.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuNewContext.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuNewContext.gpuTime, data.gpuTime ); + tracy::MemWrite( &item->gpuNewContext.period, data.period ); + tracy::MemWrite( &item->gpuNewContext.context, data.context ); + tracy::MemWrite( &item->gpuNewContext.flags, data.flags ); + tracy::MemWrite( &item->gpuNewContext.type, data.type ); + +#ifdef TRACY_ON_DEMAND + tracy::GetProfiler().DeferItem( *item ); +#endif + + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_context_name( const struct ___tracy_gpu_context_name_data data ) +{ + auto ptr = (char*)tracy::tracy_malloc( data.len ); + memcpy( ptr, data.name, data.len ); + + TracyLfqPrepareC( tracy::QueueType::GpuContextName ); + tracy::MemWrite( &item->gpuContextNameFat.context, data.context ); + tracy::MemWrite( &item->gpuContextNameFat.ptr, (uint64_t)ptr ); + tracy::MemWrite( &item->gpuContextNameFat.size, data.len ); + +#ifdef TRACY_ON_DEMAND + tracy::GetProfiler().DeferItem( *item ); +#endif + + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_calibration( const struct ___tracy_gpu_calibration_data data ) +{ + TracyLfqPrepareC( tracy::QueueType::GpuCalibration ); + tracy::MemWrite( &item->gpuCalibration.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuCalibration.gpuTime, data.gpuTime ); + tracy::MemWrite( &item->gpuCalibration.cpuDelta, data.cpuDelta ); + tracy::MemWrite( &item->gpuCalibration.context, data.context ); + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_time_sync( const struct ___tracy_gpu_time_sync_data data ) +{ + TracyLfqPrepareC( tracy::QueueType::GpuTimeSync ); + tracy::MemWrite( &item->gpuTimeSync.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuTimeSync.gpuTime, data.gpuTime ); + tracy::MemWrite( &item->gpuTimeSync.context, data.context ); + TracyLfqCommitC; +} + +TRACY_API void ___tracy_emit_gpu_zone_begin_serial( const struct ___tracy_gpu_zone_begin_data data ) +{ + auto item = tracy::Profiler::QueueSerial(); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuZoneBeginSerial ); + tracy::MemWrite( &item->gpuZoneBegin.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuZoneBegin.srcloc, data.srcloc ); + tracy::MemWrite( &item->gpuZoneBegin.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuZoneBegin.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneBegin.context, data.context ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API void ___tracy_emit_gpu_zone_begin_callstack_serial( const struct ___tracy_gpu_zone_begin_callstack_data data ) +{ + auto item = tracy::Profiler::QueueSerialCallstack( tracy::Callstack( data.depth ) ); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuZoneBeginCallstackSerial ); + tracy::MemWrite( &item->gpuZoneBegin.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuZoneBegin.srcloc, data.srcloc ); + tracy::MemWrite( &item->gpuZoneBegin.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuZoneBegin.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneBegin.context, data.context ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API void ___tracy_emit_gpu_zone_begin_alloc_serial( const struct ___tracy_gpu_zone_begin_data data ) +{ + auto item = tracy::Profiler::QueueSerial(); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuZoneBeginAllocSrcLocSerial ); + tracy::MemWrite( &item->gpuZoneBegin.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuZoneBegin.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuZoneBegin.srcloc, data.srcloc ); + tracy::MemWrite( &item->gpuZoneBegin.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneBegin.context, data.context ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API void ___tracy_emit_gpu_zone_begin_alloc_callstack_serial( const struct ___tracy_gpu_zone_begin_callstack_data data ) +{ + auto item = tracy::Profiler::QueueSerialCallstack( tracy::Callstack( data.depth ) ); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuZoneBeginAllocSrcLocCallstackSerial ); + tracy::MemWrite( &item->gpuZoneBegin.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuZoneBegin.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuZoneBegin.srcloc, data.srcloc ); + tracy::MemWrite( &item->gpuZoneBegin.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneBegin.context, data.context ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API void ___tracy_emit_gpu_time_serial( const struct ___tracy_gpu_time_data data ) +{ + auto item = tracy::Profiler::QueueSerial(); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuTime ); + tracy::MemWrite( &item->gpuTime.gpuTime, data.gpuTime ); + tracy::MemWrite( &item->gpuTime.queryId, data.queryId ); + tracy::MemWrite( &item->gpuTime.context, data.context ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API void ___tracy_emit_gpu_zone_end_serial( const struct ___tracy_gpu_zone_end_data data ) +{ + auto item = tracy::Profiler::QueueSerial(); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuZoneEndSerial ); + tracy::MemWrite( &item->gpuZoneEnd.cpuTime, tracy::Profiler::GetTime() ); + memset( &item->gpuZoneEnd.thread, 0, sizeof( item->gpuZoneEnd.thread ) ); + tracy::MemWrite( &item->gpuZoneEnd.queryId, data.queryId ); + tracy::MemWrite( &item->gpuZoneEnd.context, data.context ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API void ___tracy_emit_gpu_new_context_serial( ___tracy_gpu_new_context_data data ) +{ + auto item = tracy::Profiler::QueueSerial(); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuNewContext ); + tracy::MemWrite( &item->gpuNewContext.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuNewContext.thread, tracy::GetThreadHandle() ); + tracy::MemWrite( &item->gpuNewContext.gpuTime, data.gpuTime ); + tracy::MemWrite( &item->gpuNewContext.period, data.period ); + tracy::MemWrite( &item->gpuNewContext.context, data.context ); + tracy::MemWrite( &item->gpuNewContext.flags, data.flags ); + tracy::MemWrite( &item->gpuNewContext.type, data.type ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API void ___tracy_emit_gpu_context_name_serial( const struct ___tracy_gpu_context_name_data data ) +{ + auto ptr = (char*)tracy::tracy_malloc( data.len ); + memcpy( ptr, data.name, data.len ); + + auto item = tracy::Profiler::QueueSerial(); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuContextName ); + tracy::MemWrite( &item->gpuContextNameFat.context, data.context ); + tracy::MemWrite( &item->gpuContextNameFat.ptr, (uint64_t)ptr ); + tracy::MemWrite( &item->gpuContextNameFat.size, data.len ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API void ___tracy_emit_gpu_calibration_serial( const struct ___tracy_gpu_calibration_data data ) +{ + auto item = tracy::Profiler::QueueSerial(); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuCalibration ); + tracy::MemWrite( &item->gpuCalibration.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuCalibration.gpuTime, data.gpuTime ); + tracy::MemWrite( &item->gpuCalibration.cpuDelta, data.cpuDelta ); + tracy::MemWrite( &item->gpuCalibration.context, data.context ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API void ___tracy_emit_gpu_time_sync_serial( const struct ___tracy_gpu_time_sync_data data ) +{ + auto item = tracy::Profiler::QueueSerial(); + tracy::MemWrite( &item->hdr.type, tracy::QueueType::GpuTimeSync ); + tracy::MemWrite( &item->gpuTimeSync.cpuTime, tracy::Profiler::GetTime() ); + tracy::MemWrite( &item->gpuTimeSync.gpuTime, data.gpuTime ); + tracy::MemWrite( &item->gpuTimeSync.context, data.context ); + tracy::Profiler::QueueSerialFinish(); +} + +TRACY_API int ___tracy_connected( void ) +{ + return tracy::GetProfiler().IsConnected(); +} + +#ifdef TRACY_FIBERS +TRACY_API void ___tracy_fiber_enter( const char* fiber ){ tracy::Profiler::EnterFiber( fiber ); } +TRACY_API void ___tracy_fiber_leave( void ){ tracy::Profiler::LeaveFiber(); } +#endif + +# ifdef TRACY_MANUAL_LIFETIME +TRACY_API void ___tracy_startup_profiler( void ) +{ + tracy::StartupProfiler(); +} + +TRACY_API void ___tracy_shutdown_profiler( void ) +{ + tracy::ShutdownProfiler(); +} + +TRACY_API int ___tracy_profiler_started( void ) +{ + return tracy::s_isProfilerStarted.load( std::memory_order_seq_cst ); +} +# endif + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/externals/tracy/public/client/TracyProfiler.hpp b/externals/tracy/public/client/TracyProfiler.hpp new file mode 100644 index 000000000..8102da6c5 --- /dev/null +++ b/externals/tracy/public/client/TracyProfiler.hpp @@ -0,0 +1,1009 @@ +#ifndef __TRACYPROFILER_HPP__ +#define __TRACYPROFILER_HPP__ + +#include +#include +#include +#include +#include + +#include "tracy_concurrentqueue.h" +#include "tracy_SPSCQueue.h" +#include "TracyCallstack.hpp" +#include "TracySysPower.hpp" +#include "TracySysTime.hpp" +#include "TracyFastVector.hpp" +#include "../common/TracyQueue.hpp" +#include "../common/TracyAlign.hpp" +#include "../common/TracyAlloc.hpp" +#include "../common/TracyMutex.hpp" +#include "../common/TracyProtocol.hpp" + +#if defined _WIN32 +# include +#endif +#ifdef __APPLE__ +# include +# include +#endif + +#if ( defined _WIN32 || ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) || ( defined TARGET_OS_IOS && TARGET_OS_IOS == 1 ) ) +# define TRACY_HW_TIMER +#endif + +#ifdef __linux__ +# include +#endif + +#if defined TRACY_TIMER_FALLBACK || !defined TRACY_HW_TIMER +# include +#endif + +#ifndef TracyConcat +# define TracyConcat(x,y) TracyConcatIndirect(x,y) +#endif +#ifndef TracyConcatIndirect +# define TracyConcatIndirect(x,y) x##y +#endif + +namespace tracy +{ +#if defined(TRACY_DELAYED_INIT) && defined(TRACY_MANUAL_LIFETIME) +TRACY_API void StartupProfiler(); +TRACY_API void ShutdownProfiler(); +TRACY_API bool IsProfilerStarted(); +# define TracyIsStarted tracy::IsProfilerStarted() +#else +# define TracyIsStarted true +#endif + +class GpuCtx; +class Profiler; +class Socket; +class UdpBroadcast; + +struct GpuCtxWrapper +{ + GpuCtx* ptr; +}; + +TRACY_API moodycamel::ConcurrentQueue::ExplicitProducer* GetToken(); +TRACY_API Profiler& GetProfiler(); +TRACY_API std::atomic& GetLockCounter(); +TRACY_API std::atomic& GetGpuCtxCounter(); +TRACY_API GpuCtxWrapper& GetGpuCtx(); +TRACY_API uint32_t GetThreadHandle(); +TRACY_API bool ProfilerAvailable(); +TRACY_API bool ProfilerAllocatorAvailable(); +TRACY_API int64_t GetFrequencyQpc(); + +#if defined TRACY_TIMER_FALLBACK && defined TRACY_HW_TIMER && ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) +TRACY_API bool HardwareSupportsInvariantTSC(); // check, if we need fallback scenario +#else +# if defined TRACY_HW_TIMER +tracy_force_inline bool HardwareSupportsInvariantTSC() +{ + return true; // this is checked at startup +} +# else +tracy_force_inline bool HardwareSupportsInvariantTSC() +{ + return false; +} +# endif +#endif + + +struct SourceLocationData +{ + const char* name; + const char* function; + const char* file; + uint32_t line; + uint32_t color; +}; + +#ifdef TRACY_ON_DEMAND +struct LuaZoneState +{ + uint32_t counter; + bool active; +}; +#endif + + +#define TracyLfqPrepare( _type ) \ + moodycamel::ConcurrentQueueDefaultTraits::index_t __magic; \ + auto __token = GetToken(); \ + auto& __tail = __token->get_tail_index(); \ + auto item = __token->enqueue_begin( __magic ); \ + MemWrite( &item->hdr.type, _type ); + +#define TracyLfqCommit \ + __tail.store( __magic + 1, std::memory_order_release ); + +#define TracyLfqPrepareC( _type ) \ + tracy::moodycamel::ConcurrentQueueDefaultTraits::index_t __magic; \ + auto __token = tracy::GetToken(); \ + auto& __tail = __token->get_tail_index(); \ + auto item = __token->enqueue_begin( __magic ); \ + tracy::MemWrite( &item->hdr.type, _type ); + +#define TracyLfqCommitC \ + __tail.store( __magic + 1, std::memory_order_release ); + + +#ifdef TRACY_FIBERS +# define TracyQueuePrepare( _type ) \ + auto item = Profiler::QueueSerial(); \ + MemWrite( &item->hdr.type, _type ); +# define TracyQueueCommit( _name ) \ + MemWrite( &item->_name.thread, GetThreadHandle() ); \ + Profiler::QueueSerialFinish(); +# define TracyQueuePrepareC( _type ) \ + auto item = tracy::Profiler::QueueSerial(); \ + tracy::MemWrite( &item->hdr.type, _type ); +# define TracyQueueCommitC( _name ) \ + tracy::MemWrite( &item->_name.thread, tracy::GetThreadHandle() ); \ + tracy::Profiler::QueueSerialFinish(); +#else +# define TracyQueuePrepare( _type ) TracyLfqPrepare( _type ) +# define TracyQueueCommit( _name ) TracyLfqCommit +# define TracyQueuePrepareC( _type ) TracyLfqPrepareC( _type ) +# define TracyQueueCommitC( _name ) TracyLfqCommitC +#endif + + +typedef void(*ParameterCallback)( void* data, uint32_t idx, int32_t val ); +typedef char*(*SourceContentsCallback)( void* data, const char* filename, size_t& size ); + +class Profiler +{ + struct FrameImageQueueItem + { + void* image; + uint32_t frame; + uint16_t w; + uint16_t h; + bool flip; + }; + + enum class SymbolQueueItemType + { + CallstackFrame, + SymbolQuery, + ExternalName, + KernelCode, + SourceCode + }; + + struct SymbolQueueItem + { + SymbolQueueItemType type; + uint64_t ptr; + uint64_t extra; + uint32_t id; + }; + +public: + Profiler(); + ~Profiler(); + + void SpawnWorkerThreads(); + + static tracy_force_inline int64_t GetTime() + { +#ifdef TRACY_HW_TIMER +# if defined TARGET_OS_IOS && TARGET_OS_IOS == 1 + if( HardwareSupportsInvariantTSC() ) return mach_absolute_time(); +# elif defined _WIN32 +# ifdef TRACY_TIMER_QPC + return GetTimeQpc(); +# else + if( HardwareSupportsInvariantTSC() ) return int64_t( __rdtsc() ); +# endif +# elif defined __i386 || defined _M_IX86 + if( HardwareSupportsInvariantTSC() ) + { + uint32_t eax, edx; + asm volatile ( "rdtsc" : "=a" (eax), "=d" (edx) ); + return ( uint64_t( edx ) << 32 ) + uint64_t( eax ); + } +# elif defined __x86_64__ || defined _M_X64 + if( HardwareSupportsInvariantTSC() ) + { + uint64_t rax, rdx; +#ifdef TRACY_PATCHABLE_NOPSLEDS + // Some external tooling (such as rr) wants to patch our rdtsc and replace it by a + // branch to control the external input seen by a program. This kind of patching is + // not generally possible depending on the surrounding code and can lead to significant + // slowdowns if the compiler generated unlucky code and rr and tracy are used together. + // To avoid this, use the rr-safe `nopl 0(%rax, %rax, 1); rdtsc` instruction sequence, + // which rr promises will be patchable independent of the surrounding code. + asm volatile ( + // This is nopl 0(%rax, %rax, 1), but assemblers are inconsistent about whether + // they emit that as a 4 or 5 byte sequence and we need to be guaranteed to use + // the 5 byte one. + ".byte 0x0f, 0x1f, 0x44, 0x00, 0x00\n\t" + "rdtsc" : "=a" (rax), "=d" (rdx) ); +#else + asm volatile ( "rdtsc" : "=a" (rax), "=d" (rdx) ); +#endif + return (int64_t)(( rdx << 32 ) + rax); + } +# else +# error "TRACY_HW_TIMER detection logic needs fixing" +# endif +#endif + +#if !defined TRACY_HW_TIMER || defined TRACY_TIMER_FALLBACK +# if defined __linux__ && defined CLOCK_MONOTONIC_RAW + struct timespec ts; + clock_gettime( CLOCK_MONOTONIC_RAW, &ts ); + return int64_t( ts.tv_sec ) * 1000000000ll + int64_t( ts.tv_nsec ); +# else + return std::chrono::duration_cast( std::chrono::high_resolution_clock::now().time_since_epoch() ).count(); +# endif +#endif + +#if !defined TRACY_TIMER_FALLBACK + return 0; // unreachable branch +#endif + } + + tracy_force_inline uint32_t GetNextZoneId() + { + return m_zoneId.fetch_add( 1, std::memory_order_relaxed ); + } + + static tracy_force_inline QueueItem* QueueSerial() + { + auto& p = GetProfiler(); + p.m_serialLock.lock(); + return p.m_serialQueue.prepare_next(); + } + + static tracy_force_inline QueueItem* QueueSerialCallstack( void* ptr ) + { + auto& p = GetProfiler(); + p.m_serialLock.lock(); + p.SendCallstackSerial( ptr ); + return p.m_serialQueue.prepare_next(); + } + + static tracy_force_inline void QueueSerialFinish() + { + auto& p = GetProfiler(); + p.m_serialQueue.commit_next(); + p.m_serialLock.unlock(); + } + + static tracy_force_inline void SendFrameMark( const char* name ) + { + if( !name ) GetProfiler().m_frameCount.fetch_add( 1, std::memory_order_relaxed ); +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + auto item = QueueSerial(); + MemWrite( &item->hdr.type, QueueType::FrameMarkMsg ); + MemWrite( &item->frameMark.time, GetTime() ); + MemWrite( &item->frameMark.name, uint64_t( name ) ); + QueueSerialFinish(); + } + + static tracy_force_inline void SendFrameMark( const char* name, QueueType type ) + { + assert( type == QueueType::FrameMarkMsgStart || type == QueueType::FrameMarkMsgEnd ); +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + auto item = QueueSerial(); + MemWrite( &item->hdr.type, type ); + MemWrite( &item->frameMark.time, GetTime() ); + MemWrite( &item->frameMark.name, uint64_t( name ) ); + QueueSerialFinish(); + } + + static tracy_force_inline void SendFrameImage( const void* image, uint16_t w, uint16_t h, uint8_t offset, bool flip ) + { +#ifndef TRACY_NO_FRAME_IMAGE + auto& profiler = GetProfiler(); + assert( profiler.m_frameCount.load( std::memory_order_relaxed ) < (std::numeric_limits::max)() ); +# ifdef TRACY_ON_DEMAND + if( !profiler.IsConnected() ) return; +# endif + const auto sz = size_t( w ) * size_t( h ) * 4; + auto ptr = (char*)tracy_malloc( sz ); + memcpy( ptr, image, sz ); + + profiler.m_fiLock.lock(); + auto fi = profiler.m_fiQueue.prepare_next(); + fi->image = ptr; + fi->frame = uint32_t( profiler.m_frameCount.load( std::memory_order_relaxed ) - offset ); + fi->w = w; + fi->h = h; + fi->flip = flip; + profiler.m_fiQueue.commit_next(); + profiler.m_fiLock.unlock(); +#else + static_cast(image); // unused + static_cast(w); // unused + static_cast(h); // unused + static_cast(offset); // unused + static_cast(flip); // unused +#endif + } + + static tracy_force_inline void PlotData( const char* name, int64_t val ) + { +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + TracyLfqPrepare( QueueType::PlotDataInt ); + MemWrite( &item->plotDataInt.name, (uint64_t)name ); + MemWrite( &item->plotDataInt.time, GetTime() ); + MemWrite( &item->plotDataInt.val, val ); + TracyLfqCommit; + } + + static tracy_force_inline void PlotData( const char* name, float val ) + { +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + TracyLfqPrepare( QueueType::PlotDataFloat ); + MemWrite( &item->plotDataFloat.name, (uint64_t)name ); + MemWrite( &item->plotDataFloat.time, GetTime() ); + MemWrite( &item->plotDataFloat.val, val ); + TracyLfqCommit; + } + + static tracy_force_inline void PlotData( const char* name, double val ) + { +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + TracyLfqPrepare( QueueType::PlotDataDouble ); + MemWrite( &item->plotDataDouble.name, (uint64_t)name ); + MemWrite( &item->plotDataDouble.time, GetTime() ); + MemWrite( &item->plotDataDouble.val, val ); + TracyLfqCommit; + } + + static tracy_force_inline void ConfigurePlot( const char* name, PlotFormatType type, bool step, bool fill, uint32_t color ) + { + TracyLfqPrepare( QueueType::PlotConfig ); + MemWrite( &item->plotConfig.name, (uint64_t)name ); + MemWrite( &item->plotConfig.type, (uint8_t)type ); + MemWrite( &item->plotConfig.step, (uint8_t)step ); + MemWrite( &item->plotConfig.fill, (uint8_t)fill ); + MemWrite( &item->plotConfig.color, color ); + +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + + TracyLfqCommit; + } + + static tracy_force_inline void Message( const char* txt, size_t size, int callstack ) + { + assert( size < (std::numeric_limits::max)() ); +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + if( callstack != 0 ) + { + tracy::GetProfiler().SendCallstack( callstack ); + } + + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, txt, size ); + + TracyQueuePrepare( callstack == 0 ? QueueType::Message : QueueType::MessageCallstack ); + MemWrite( &item->messageFat.time, GetTime() ); + MemWrite( &item->messageFat.text, (uint64_t)ptr ); + MemWrite( &item->messageFat.size, (uint16_t)size ); + TracyQueueCommit( messageFatThread ); + } + + static tracy_force_inline void Message( const char* txt, int callstack ) + { +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + if( callstack != 0 ) + { + tracy::GetProfiler().SendCallstack( callstack ); + } + + TracyQueuePrepare( callstack == 0 ? QueueType::MessageLiteral : QueueType::MessageLiteralCallstack ); + MemWrite( &item->messageLiteral.time, GetTime() ); + MemWrite( &item->messageLiteral.text, (uint64_t)txt ); + TracyQueueCommit( messageLiteralThread ); + } + + static tracy_force_inline void MessageColor( const char* txt, size_t size, uint32_t color, int callstack ) + { + assert( size < (std::numeric_limits::max)() ); +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + if( callstack != 0 ) + { + tracy::GetProfiler().SendCallstack( callstack ); + } + + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, txt, size ); + + TracyQueuePrepare( callstack == 0 ? QueueType::MessageColor : QueueType::MessageColorCallstack ); + MemWrite( &item->messageColorFat.time, GetTime() ); + MemWrite( &item->messageColorFat.text, (uint64_t)ptr ); + MemWrite( &item->messageColorFat.b, uint8_t( ( color ) & 0xFF ) ); + MemWrite( &item->messageColorFat.g, uint8_t( ( color >> 8 ) & 0xFF ) ); + MemWrite( &item->messageColorFat.r, uint8_t( ( color >> 16 ) & 0xFF ) ); + MemWrite( &item->messageColorFat.size, (uint16_t)size ); + TracyQueueCommit( messageColorFatThread ); + } + + static tracy_force_inline void MessageColor( const char* txt, uint32_t color, int callstack ) + { +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + if( callstack != 0 ) + { + tracy::GetProfiler().SendCallstack( callstack ); + } + + TracyQueuePrepare( callstack == 0 ? QueueType::MessageLiteralColor : QueueType::MessageLiteralColorCallstack ); + MemWrite( &item->messageColorLiteral.time, GetTime() ); + MemWrite( &item->messageColorLiteral.text, (uint64_t)txt ); + MemWrite( &item->messageColorLiteral.b, uint8_t( ( color ) & 0xFF ) ); + MemWrite( &item->messageColorLiteral.g, uint8_t( ( color >> 8 ) & 0xFF ) ); + MemWrite( &item->messageColorLiteral.r, uint8_t( ( color >> 16 ) & 0xFF ) ); + TracyQueueCommit( messageColorLiteralThread ); + } + + static tracy_force_inline void MessageAppInfo( const char* txt, size_t size ) + { + assert( size < (std::numeric_limits::max)() ); + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, txt, size ); + TracyLfqPrepare( QueueType::MessageAppInfo ); + MemWrite( &item->messageFat.time, GetTime() ); + MemWrite( &item->messageFat.text, (uint64_t)ptr ); + MemWrite( &item->messageFat.size, (uint16_t)size ); + +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + + TracyLfqCommit; + } + + static tracy_force_inline void MemAlloc( const void* ptr, size_t size, bool secure ) + { + if( secure && !ProfilerAvailable() ) return; +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + const auto thread = GetThreadHandle(); + + GetProfiler().m_serialLock.lock(); + SendMemAlloc( QueueType::MemAlloc, thread, ptr, size ); + GetProfiler().m_serialLock.unlock(); + } + + static tracy_force_inline void MemFree( const void* ptr, bool secure ) + { + if( secure && !ProfilerAvailable() ) return; +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + const auto thread = GetThreadHandle(); + + GetProfiler().m_serialLock.lock(); + SendMemFree( QueueType::MemFree, thread, ptr ); + GetProfiler().m_serialLock.unlock(); + } + + static tracy_force_inline void MemAllocCallstack( const void* ptr, size_t size, int depth, bool secure ) + { + if( secure && !ProfilerAvailable() ) return; +#ifdef TRACY_HAS_CALLSTACK + auto& profiler = GetProfiler(); +# ifdef TRACY_ON_DEMAND + if( !profiler.IsConnected() ) return; +# endif + const auto thread = GetThreadHandle(); + + auto callstack = Callstack( depth ); + + profiler.m_serialLock.lock(); + SendCallstackSerial( callstack ); + SendMemAlloc( QueueType::MemAllocCallstack, thread, ptr, size ); + profiler.m_serialLock.unlock(); +#else + static_cast(depth); // unused + MemAlloc( ptr, size, secure ); +#endif + } + + static tracy_force_inline void MemFreeCallstack( const void* ptr, int depth, bool secure ) + { + if( secure && !ProfilerAvailable() ) return; + if( !ProfilerAllocatorAvailable() ) + { + MemFree( ptr, secure ); + return; + } +#ifdef TRACY_HAS_CALLSTACK + auto& profiler = GetProfiler(); +# ifdef TRACY_ON_DEMAND + if( !profiler.IsConnected() ) return; +# endif + const auto thread = GetThreadHandle(); + + auto callstack = Callstack( depth ); + + profiler.m_serialLock.lock(); + SendCallstackSerial( callstack ); + SendMemFree( QueueType::MemFreeCallstack, thread, ptr ); + profiler.m_serialLock.unlock(); +#else + static_cast(depth); // unused + MemFree( ptr, secure ); +#endif + } + + static tracy_force_inline void MemAllocNamed( const void* ptr, size_t size, bool secure, const char* name ) + { + if( secure && !ProfilerAvailable() ) return; +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + const auto thread = GetThreadHandle(); + + GetProfiler().m_serialLock.lock(); + SendMemName( name ); + SendMemAlloc( QueueType::MemAllocNamed, thread, ptr, size ); + GetProfiler().m_serialLock.unlock(); + } + + static tracy_force_inline void MemFreeNamed( const void* ptr, bool secure, const char* name ) + { + if( secure && !ProfilerAvailable() ) return; +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + const auto thread = GetThreadHandle(); + + GetProfiler().m_serialLock.lock(); + SendMemName( name ); + SendMemFree( QueueType::MemFreeNamed, thread, ptr ); + GetProfiler().m_serialLock.unlock(); + } + + static tracy_force_inline void MemAllocCallstackNamed( const void* ptr, size_t size, int depth, bool secure, const char* name ) + { + if( secure && !ProfilerAvailable() ) return; +#ifdef TRACY_HAS_CALLSTACK + auto& profiler = GetProfiler(); +# ifdef TRACY_ON_DEMAND + if( !profiler.IsConnected() ) return; +# endif + const auto thread = GetThreadHandle(); + + auto callstack = Callstack( depth ); + + profiler.m_serialLock.lock(); + SendCallstackSerial( callstack ); + SendMemName( name ); + SendMemAlloc( QueueType::MemAllocCallstackNamed, thread, ptr, size ); + profiler.m_serialLock.unlock(); +#else + static_cast(depth); // unused + static_cast(name); // unused + MemAlloc( ptr, size, secure ); +#endif + } + + static tracy_force_inline void MemFreeCallstackNamed( const void* ptr, int depth, bool secure, const char* name ) + { + if( secure && !ProfilerAvailable() ) return; +#ifdef TRACY_HAS_CALLSTACK + auto& profiler = GetProfiler(); +# ifdef TRACY_ON_DEMAND + if( !profiler.IsConnected() ) return; +# endif + const auto thread = GetThreadHandle(); + + auto callstack = Callstack( depth ); + + profiler.m_serialLock.lock(); + SendCallstackSerial( callstack ); + SendMemName( name ); + SendMemFree( QueueType::MemFreeCallstackNamed, thread, ptr ); + profiler.m_serialLock.unlock(); +#else + static_cast(depth); // unused + static_cast(name); // unused + MemFree( ptr, secure ); +#endif + } + + static tracy_force_inline void SendCallstack( int depth ) + { +#ifdef TRACY_HAS_CALLSTACK + auto ptr = Callstack( depth ); + TracyQueuePrepare( QueueType::Callstack ); + MemWrite( &item->callstackFat.ptr, (uint64_t)ptr ); + TracyQueueCommit( callstackFatThread ); +#else + static_cast(depth); // unused +#endif + } + + static tracy_force_inline void ParameterRegister( ParameterCallback cb, void* data ) + { + auto& profiler = GetProfiler(); + profiler.m_paramCallback = cb; + profiler.m_paramCallbackData = data; + } + + static tracy_force_inline void ParameterSetup( uint32_t idx, const char* name, bool isBool, int32_t val ) + { + TracyLfqPrepare( QueueType::ParamSetup ); + tracy::MemWrite( &item->paramSetup.idx, idx ); + tracy::MemWrite( &item->paramSetup.name, (uint64_t)name ); + tracy::MemWrite( &item->paramSetup.isBool, (uint8_t)isBool ); + tracy::MemWrite( &item->paramSetup.val, val ); + +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + + TracyLfqCommit; + } + + static tracy_force_inline void SourceCallbackRegister( SourceContentsCallback cb, void* data ) + { + auto& profiler = GetProfiler(); + profiler.m_sourceCallback = cb; + profiler.m_sourceCallbackData = data; + } + +#ifdef TRACY_FIBERS + static tracy_force_inline void EnterFiber( const char* fiber ) + { + TracyQueuePrepare( QueueType::FiberEnter ); + MemWrite( &item->fiberEnter.time, GetTime() ); + MemWrite( &item->fiberEnter.fiber, (uint64_t)fiber ); + TracyQueueCommit( fiberEnter ); + } + + static tracy_force_inline void LeaveFiber() + { + TracyQueuePrepare( QueueType::FiberLeave ); + MemWrite( &item->fiberLeave.time, GetTime() ); + TracyQueueCommit( fiberLeave ); + } +#endif + + void SendCallstack( int depth, const char* skipBefore ); + static void CutCallstack( void* callstack, const char* skipBefore ); + + static bool ShouldExit(); + + tracy_force_inline bool IsConnected() const + { + return m_isConnected.load( std::memory_order_acquire ); + } + + tracy_force_inline void SetProgramName( const char* name ) + { + m_programNameLock.lock(); + m_programName = name; + m_programNameLock.unlock(); + } + +#ifdef TRACY_ON_DEMAND + tracy_force_inline uint64_t ConnectionId() const + { + return m_connectionId.load( std::memory_order_acquire ); + } + + tracy_force_inline void DeferItem( const QueueItem& item ) + { + m_deferredLock.lock(); + auto dst = m_deferredQueue.push_next(); + memcpy( dst, &item, sizeof( item ) ); + m_deferredLock.unlock(); + } +#endif + + void RequestShutdown() { m_shutdown.store( true, std::memory_order_relaxed ); m_shutdownManual.store( true, std::memory_order_relaxed ); } + bool HasShutdownFinished() const { return m_shutdownFinished.load( std::memory_order_relaxed ); } + + void SendString( uint64_t str, const char* ptr, QueueType type ) { SendString( str, ptr, strlen( ptr ), type ); } + void SendString( uint64_t str, const char* ptr, size_t len, QueueType type ); + void SendSingleString( const char* ptr ) { SendSingleString( ptr, strlen( ptr ) ); } + void SendSingleString( const char* ptr, size_t len ); + void SendSecondString( const char* ptr ) { SendSecondString( ptr, strlen( ptr ) ); } + void SendSecondString( const char* ptr, size_t len ); + + + // Allocated source location data layout: + // 2b payload size + // 4b color + // 4b source line + // fsz function name + // 1b null terminator + // ssz source file name + // 1b null terminator + // nsz zone name (optional) + + static tracy_force_inline uint64_t AllocSourceLocation( uint32_t line, const char* source, const char* function ) + { + return AllocSourceLocation( line, source, function, nullptr, 0 ); + } + + static tracy_force_inline uint64_t AllocSourceLocation( uint32_t line, const char* source, const char* function, const char* name, size_t nameSz ) + { + return AllocSourceLocation( line, source, strlen(source), function, strlen(function), name, nameSz ); + } + + static tracy_force_inline uint64_t AllocSourceLocation( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz ) + { + return AllocSourceLocation( line, source, sourceSz, function, functionSz, nullptr, 0 ); + } + + static tracy_force_inline uint64_t AllocSourceLocation( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz ) + { + const auto sz32 = uint32_t( 2 + 4 + 4 + functionSz + 1 + sourceSz + 1 + nameSz ); + assert( sz32 <= (std::numeric_limits::max)() ); + const auto sz = uint16_t( sz32 ); + auto ptr = (char*)tracy_malloc( sz ); + memcpy( ptr, &sz, 2 ); + memset( ptr + 2, 0, 4 ); + memcpy( ptr + 6, &line, 4 ); + memcpy( ptr + 10, function, functionSz ); + ptr[10 + functionSz] = '\0'; + memcpy( ptr + 10 + functionSz + 1, source, sourceSz ); + ptr[10 + functionSz + 1 + sourceSz] = '\0'; + if( nameSz != 0 ) + { + memcpy( ptr + 10 + functionSz + 1 + sourceSz + 1, name, nameSz ); + } + return uint64_t( ptr ); + } + +private: + enum class DequeueStatus { DataDequeued, ConnectionLost, QueueEmpty }; + enum class ThreadCtxStatus { Same, Changed, ConnectionLost }; + + static void LaunchWorker( void* ptr ) { ((Profiler*)ptr)->Worker(); } + void Worker(); + +#ifndef TRACY_NO_FRAME_IMAGE + static void LaunchCompressWorker( void* ptr ) { ((Profiler*)ptr)->CompressWorker(); } + void CompressWorker(); +#endif + +#ifdef TRACY_HAS_CALLSTACK + static void LaunchSymbolWorker( void* ptr ) { ((Profiler*)ptr)->SymbolWorker(); } + void SymbolWorker(); + void HandleSymbolQueueItem( const SymbolQueueItem& si ); +#endif + + void InstallCrashHandler(); + void RemoveCrashHandler(); + + void ClearQueues( tracy::moodycamel::ConsumerToken& token ); + void ClearSerial(); + DequeueStatus Dequeue( tracy::moodycamel::ConsumerToken& token ); + DequeueStatus DequeueContextSwitches( tracy::moodycamel::ConsumerToken& token, int64_t& timeStop ); + DequeueStatus DequeueSerial(); + ThreadCtxStatus ThreadCtxCheck( uint32_t threadId ); + bool CommitData(); + + tracy_force_inline bool AppendData( const void* data, size_t len ) + { + const auto ret = NeedDataSize( len ); + AppendDataUnsafe( data, len ); + return ret; + } + + tracy_force_inline bool NeedDataSize( size_t len ) + { + assert( len <= TargetFrameSize ); + bool ret = true; + if( m_bufferOffset - m_bufferStart + (int)len > TargetFrameSize ) + { + ret = CommitData(); + } + return ret; + } + + tracy_force_inline void AppendDataUnsafe( const void* data, size_t len ) + { + memcpy( m_buffer + m_bufferOffset, data, len ); + m_bufferOffset += int( len ); + } + + bool SendData( const char* data, size_t len ); + void SendLongString( uint64_t ptr, const char* str, size_t len, QueueType type ); + void SendSourceLocation( uint64_t ptr ); + void SendSourceLocationPayload( uint64_t ptr ); + void SendCallstackPayload( uint64_t ptr ); + void SendCallstackPayload64( uint64_t ptr ); + void SendCallstackAlloc( uint64_t ptr ); + + void QueueCallstackFrame( uint64_t ptr ); + void QueueSymbolQuery( uint64_t symbol ); + void QueueExternalName( uint64_t ptr ); + void QueueKernelCode( uint64_t symbol, uint32_t size ); + void QueueSourceCodeQuery( uint32_t id ); + + bool HandleServerQuery(); + void HandleDisconnect(); + void HandleParameter( uint64_t payload ); + void HandleSymbolCodeQuery( uint64_t symbol, uint32_t size ); + void HandleSourceCodeQuery( char* data, char* image, uint32_t id ); + + void AckServerQuery(); + void AckSymbolCodeNotAvailable(); + + void CalibrateTimer(); + void CalibrateDelay(); + void ReportTopology(); + + static tracy_force_inline void SendCallstackSerial( void* ptr ) + { +#ifdef TRACY_HAS_CALLSTACK + auto item = GetProfiler().m_serialQueue.prepare_next(); + MemWrite( &item->hdr.type, QueueType::CallstackSerial ); + MemWrite( &item->callstackFat.ptr, (uint64_t)ptr ); + GetProfiler().m_serialQueue.commit_next(); +#else + static_cast(ptr); // unused +#endif + } + + static tracy_force_inline void SendMemAlloc( QueueType type, const uint32_t thread, const void* ptr, size_t size ) + { + assert( type == QueueType::MemAlloc || type == QueueType::MemAllocCallstack || type == QueueType::MemAllocNamed || type == QueueType::MemAllocCallstackNamed ); + + auto item = GetProfiler().m_serialQueue.prepare_next(); + MemWrite( &item->hdr.type, type ); + MemWrite( &item->memAlloc.time, GetTime() ); + MemWrite( &item->memAlloc.thread, thread ); + MemWrite( &item->memAlloc.ptr, (uint64_t)ptr ); + if( compile_time_condition::value ) + { + memcpy( &item->memAlloc.size, &size, 4 ); + memset( &item->memAlloc.size + 4, 0, 2 ); + } + else + { + assert( sizeof( size ) == 8 ); + memcpy( &item->memAlloc.size, &size, 4 ); + memcpy( ((char*)&item->memAlloc.size)+4, ((char*)&size)+4, 2 ); + } + GetProfiler().m_serialQueue.commit_next(); + } + + static tracy_force_inline void SendMemFree( QueueType type, const uint32_t thread, const void* ptr ) + { + assert( type == QueueType::MemFree || type == QueueType::MemFreeCallstack || type == QueueType::MemFreeNamed || type == QueueType::MemFreeCallstackNamed ); + + auto item = GetProfiler().m_serialQueue.prepare_next(); + MemWrite( &item->hdr.type, type ); + MemWrite( &item->memFree.time, GetTime() ); + MemWrite( &item->memFree.thread, thread ); + MemWrite( &item->memFree.ptr, (uint64_t)ptr ); + GetProfiler().m_serialQueue.commit_next(); + } + + static tracy_force_inline void SendMemName( const char* name ) + { + assert( name ); + auto item = GetProfiler().m_serialQueue.prepare_next(); + MemWrite( &item->hdr.type, QueueType::MemNamePayload ); + MemWrite( &item->memName.name, (uint64_t)name ); + GetProfiler().m_serialQueue.commit_next(); + } + +#if defined _WIN32 && defined TRACY_TIMER_QPC + static int64_t GetTimeQpc(); +#endif + + double m_timerMul; + uint64_t m_resolution; + uint64_t m_delay; + std::atomic m_timeBegin; + uint32_t m_mainThread; + uint64_t m_epoch, m_exectime; + std::atomic m_shutdown; + std::atomic m_shutdownManual; + std::atomic m_shutdownFinished; + Socket* m_sock; + UdpBroadcast* m_broadcast; + bool m_noExit; + uint32_t m_userPort; + std::atomic m_zoneId; + int64_t m_samplingPeriod; + + uint32_t m_threadCtx; + int64_t m_refTimeThread; + int64_t m_refTimeSerial; + int64_t m_refTimeCtx; + int64_t m_refTimeGpu; + + void* m_stream; // LZ4_stream_t* + char* m_buffer; + int m_bufferOffset; + int m_bufferStart; + + char* m_lz4Buf; + + FastVector m_serialQueue, m_serialDequeue; + TracyMutex m_serialLock; + +#ifndef TRACY_NO_FRAME_IMAGE + FastVector m_fiQueue, m_fiDequeue; + TracyMutex m_fiLock; +#endif + + SPSCQueue m_symbolQueue; + + std::atomic m_frameCount; + std::atomic m_isConnected; +#ifdef TRACY_ON_DEMAND + std::atomic m_connectionId; + + TracyMutex m_deferredLock; + FastVector m_deferredQueue; +#endif + +#ifdef TRACY_HAS_SYSTIME + void ProcessSysTime(); + + SysTime m_sysTime; + uint64_t m_sysTimeLast = 0; +#else + void ProcessSysTime() {} +#endif + +#ifdef TRACY_HAS_SYSPOWER + SysPower m_sysPower; +#endif + + ParameterCallback m_paramCallback; + void* m_paramCallbackData; + SourceContentsCallback m_sourceCallback; + void* m_sourceCallbackData; + + char* m_queryImage; + char* m_queryData; + char* m_queryDataPtr; + +#if defined _WIN32 + void* m_exceptionHandler; +#endif +#ifdef __linux__ + struct { + struct sigaction pwr, ill, fpe, segv, pipe, bus, abrt; + } m_prevSignal; +#endif + bool m_crashHandlerInstalled; + + const char* m_programName; + TracyMutex m_programNameLock; +}; + +} + +#endif diff --git a/externals/tracy/public/client/TracyRingBuffer.hpp b/externals/tracy/public/client/TracyRingBuffer.hpp new file mode 100644 index 000000000..e9100e2d8 --- /dev/null +++ b/externals/tracy/public/client/TracyRingBuffer.hpp @@ -0,0 +1,141 @@ +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "TracyDebug.hpp" + +namespace tracy +{ + +class RingBuffer +{ +public: + RingBuffer( unsigned int size, int fd, int id, int cpu = -1 ) + : m_size( size ) + , m_id( id ) + , m_cpu( cpu ) + , m_fd( fd ) + { + const auto pageSize = uint32_t( getpagesize() ); + assert( size >= pageSize ); + assert( __builtin_popcount( size ) == 1 ); + m_mapSize = size + pageSize; + auto mapAddr = mmap( nullptr, m_mapSize, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0 ); + if( mapAddr == MAP_FAILED ) + { + TracyDebug( "mmap failed: errno %i (%s)\n", errno, strerror( errno ) ); + m_fd = 0; + m_metadata = nullptr; + close( fd ); + return; + } + m_metadata = (perf_event_mmap_page*)mapAddr; + assert( m_metadata->data_offset == pageSize ); + m_buffer = ((char*)mapAddr) + pageSize; + m_tail = m_metadata->data_tail; + } + + ~RingBuffer() + { + if( m_metadata ) munmap( m_metadata, m_mapSize ); + if( m_fd ) close( m_fd ); + } + + RingBuffer( const RingBuffer& ) = delete; + RingBuffer& operator=( const RingBuffer& ) = delete; + + RingBuffer( RingBuffer&& other ) + { + memcpy( (char*)&other, (char*)this, sizeof( RingBuffer ) ); + m_metadata = nullptr; + m_fd = 0; + } + + RingBuffer& operator=( RingBuffer&& other ) + { + memcpy( (char*)&other, (char*)this, sizeof( RingBuffer ) ); + m_metadata = nullptr; + m_fd = 0; + return *this; + } + + bool IsValid() const { return m_metadata != nullptr; } + int GetId() const { return m_id; } + int GetCpu() const { return m_cpu; } + + void Enable() + { + ioctl( m_fd, PERF_EVENT_IOC_ENABLE, 0 ); + } + + void Read( void* dst, uint64_t offset, uint64_t cnt ) + { + const auto size = m_size; + auto src = ( m_tail + offset ) % size; + if( src + cnt <= size ) + { + memcpy( dst, m_buffer + src, cnt ); + } + else + { + const auto s0 = size - src; + const auto buf = m_buffer; + memcpy( dst, buf + src, s0 ); + memcpy( (char*)dst + s0, buf, cnt - s0 ); + } + } + + void Advance( uint64_t cnt ) + { + m_tail += cnt; + StoreTail(); + } + + bool CheckTscCaps() const + { + return m_metadata->cap_user_time_zero; + } + + int64_t ConvertTimeToTsc( int64_t timestamp ) const + { + if( !m_metadata->cap_user_time_zero ) return 0; + const auto time = timestamp - m_metadata->time_zero; + const auto quot = time / m_metadata->time_mult; + const auto rem = time % m_metadata->time_mult; + return ( quot << m_metadata->time_shift ) + ( rem << m_metadata->time_shift ) / m_metadata->time_mult; + } + + uint64_t LoadHead() const + { + return std::atomic_load_explicit( (const volatile std::atomic*)&m_metadata->data_head, std::memory_order_acquire ); + } + + uint64_t GetTail() const + { + return m_tail; + } + +private: + void StoreTail() + { + std::atomic_store_explicit( (volatile std::atomic*)&m_metadata->data_tail, m_tail, std::memory_order_release ); + } + + unsigned int m_size; + uint64_t m_tail; + char* m_buffer; + int m_id; + int m_cpu; + perf_event_mmap_page* m_metadata; + + size_t m_mapSize; + int m_fd; +}; + +} diff --git a/externals/tracy/public/client/TracyScoped.hpp b/externals/tracy/public/client/TracyScoped.hpp new file mode 100644 index 000000000..d2274e40b --- /dev/null +++ b/externals/tracy/public/client/TracyScoped.hpp @@ -0,0 +1,175 @@ +#ifndef __TRACYSCOPED_HPP__ +#define __TRACYSCOPED_HPP__ + +#include +#include +#include + +#include "../common/TracySystem.hpp" +#include "../common/TracyAlign.hpp" +#include "../common/TracyAlloc.hpp" +#include "TracyProfiler.hpp" + +namespace tracy +{ + +class ScopedZone +{ +public: + ScopedZone( const ScopedZone& ) = delete; + ScopedZone( ScopedZone&& ) = delete; + ScopedZone& operator=( const ScopedZone& ) = delete; + ScopedZone& operator=( ScopedZone&& ) = delete; + + tracy_force_inline ScopedZone( const SourceLocationData* srcloc, bool is_active = true ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; +#ifdef TRACY_ON_DEMAND + m_connectionId = GetProfiler().ConnectionId(); +#endif + TracyQueuePrepare( QueueType::ZoneBegin ); + MemWrite( &item->zoneBegin.time, Profiler::GetTime() ); + MemWrite( &item->zoneBegin.srcloc, (uint64_t)srcloc ); + TracyQueueCommit( zoneBeginThread ); + } + + tracy_force_inline ScopedZone( const SourceLocationData* srcloc, int depth, bool is_active = true ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; +#ifdef TRACY_ON_DEMAND + m_connectionId = GetProfiler().ConnectionId(); +#endif + GetProfiler().SendCallstack( depth ); + + TracyQueuePrepare( QueueType::ZoneBeginCallstack ); + MemWrite( &item->zoneBegin.time, Profiler::GetTime() ); + MemWrite( &item->zoneBegin.srcloc, (uint64_t)srcloc ); + TracyQueueCommit( zoneBeginThread ); + } + + tracy_force_inline ScopedZone( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, bool is_active = true ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; +#ifdef TRACY_ON_DEMAND + m_connectionId = GetProfiler().ConnectionId(); +#endif + TracyQueuePrepare( QueueType::ZoneBeginAllocSrcLoc ); + const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz ); + MemWrite( &item->zoneBegin.time, Profiler::GetTime() ); + MemWrite( &item->zoneBegin.srcloc, srcloc ); + TracyQueueCommit( zoneBeginThread ); + } + + tracy_force_inline ScopedZone( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, int depth, bool is_active = true ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; +#ifdef TRACY_ON_DEMAND + m_connectionId = GetProfiler().ConnectionId(); +#endif + GetProfiler().SendCallstack( depth ); + + TracyQueuePrepare( QueueType::ZoneBeginAllocSrcLocCallstack ); + const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz ); + MemWrite( &item->zoneBegin.time, Profiler::GetTime() ); + MemWrite( &item->zoneBegin.srcloc, srcloc ); + TracyQueueCommit( zoneBeginThread ); + } + + tracy_force_inline ~ScopedZone() + { + if( !m_active ) return; +#ifdef TRACY_ON_DEMAND + if( GetProfiler().ConnectionId() != m_connectionId ) return; +#endif + TracyQueuePrepare( QueueType::ZoneEnd ); + MemWrite( &item->zoneEnd.time, Profiler::GetTime() ); + TracyQueueCommit( zoneEndThread ); + } + + tracy_force_inline void Text( const char* txt, size_t size ) + { + assert( size < (std::numeric_limits::max)() ); + if( !m_active ) return; +#ifdef TRACY_ON_DEMAND + if( GetProfiler().ConnectionId() != m_connectionId ) return; +#endif + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, txt, size ); + TracyQueuePrepare( QueueType::ZoneText ); + MemWrite( &item->zoneTextFat.text, (uint64_t)ptr ); + MemWrite( &item->zoneTextFat.size, (uint16_t)size ); + TracyQueueCommit( zoneTextFatThread ); + } + + tracy_force_inline void Name( const char* txt, size_t size ) + { + assert( size < (std::numeric_limits::max)() ); + if( !m_active ) return; +#ifdef TRACY_ON_DEMAND + if( GetProfiler().ConnectionId() != m_connectionId ) return; +#endif + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, txt, size ); + TracyQueuePrepare( QueueType::ZoneName ); + MemWrite( &item->zoneTextFat.text, (uint64_t)ptr ); + MemWrite( &item->zoneTextFat.size, (uint16_t)size ); + TracyQueueCommit( zoneTextFatThread ); + } + + tracy_force_inline void Color( uint32_t color ) + { + if( !m_active ) return; +#ifdef TRACY_ON_DEMAND + if( GetProfiler().ConnectionId() != m_connectionId ) return; +#endif + TracyQueuePrepare( QueueType::ZoneColor ); + MemWrite( &item->zoneColor.b, uint8_t( ( color ) & 0xFF ) ); + MemWrite( &item->zoneColor.g, uint8_t( ( color >> 8 ) & 0xFF ) ); + MemWrite( &item->zoneColor.r, uint8_t( ( color >> 16 ) & 0xFF ) ); + TracyQueueCommit( zoneColorThread ); + } + + tracy_force_inline void Value( uint64_t value ) + { + if( !m_active ) return; +#ifdef TRACY_ON_DEMAND + if( GetProfiler().ConnectionId() != m_connectionId ) return; +#endif + TracyQueuePrepare( QueueType::ZoneValue ); + MemWrite( &item->zoneValue.value, value ); + TracyQueueCommit( zoneValueThread ); + } + + tracy_force_inline bool IsActive() const { return m_active; } + +private: + const bool m_active; + +#ifdef TRACY_ON_DEMAND + uint64_t m_connectionId = 0; +#endif +}; + +} + +#endif diff --git a/externals/tracy/public/client/TracyStringHelpers.hpp b/externals/tracy/public/client/TracyStringHelpers.hpp new file mode 100644 index 000000000..977be6a3e --- /dev/null +++ b/externals/tracy/public/client/TracyStringHelpers.hpp @@ -0,0 +1,41 @@ +#ifndef __TRACYSTRINGHELPERS_HPP__ +#define __TRACYSTRINGHELPERS_HPP__ + +#include +#include + +#include "../common/TracyAlloc.hpp" +#include "../common/TracyForceInline.hpp" + +namespace tracy +{ + +static tracy_force_inline char* CopyString( const char* src, size_t sz ) +{ + auto dst = (char*)tracy_malloc( sz + 1 ); + memcpy( dst, src, sz ); + dst[sz] = '\0'; + return dst; +} + +static tracy_force_inline char* CopyString( const char* src ) +{ + return CopyString( src, strlen( src ) ); +} + +static tracy_force_inline char* CopyStringFast( const char* src, size_t sz ) +{ + auto dst = (char*)tracy_malloc_fast( sz + 1 ); + memcpy( dst, src, sz ); + dst[sz] = '\0'; + return dst; +} + +static tracy_force_inline char* CopyStringFast( const char* src ) +{ + return CopyStringFast( src, strlen( src ) ); +} + +} + +#endif diff --git a/externals/tracy/public/client/TracySysPower.cpp b/externals/tracy/public/client/TracySysPower.cpp new file mode 100644 index 000000000..b12411d78 --- /dev/null +++ b/externals/tracy/public/client/TracySysPower.cpp @@ -0,0 +1,164 @@ +#include "TracySysPower.hpp" + +#ifdef TRACY_HAS_SYSPOWER + +#include +#include +#include +#include +#include +#include + +#include "TracyDebug.hpp" +#include "TracyProfiler.hpp" +#include "../common/TracyAlloc.hpp" + +namespace tracy +{ + +SysPower::SysPower() + : m_domains( 4 ) + , m_lastTime( 0 ) +{ + ScanDirectory( "/sys/devices/virtual/powercap/intel-rapl", -1 ); +} + +SysPower::~SysPower() +{ + for( auto& v : m_domains ) + { + fclose( v.handle ); + // Do not release v.name, as it may be still needed + } +} + +void SysPower::Tick() +{ + auto t = std::chrono::high_resolution_clock::now().time_since_epoch().count(); + if( t - m_lastTime > 10000000 ) // 10 ms + { + m_lastTime = t; + for( auto& v : m_domains ) + { + char tmp[32]; + if( fread( tmp, 1, 32, v.handle ) > 0 ) + { + rewind( v.handle ); + auto p = (uint64_t)atoll( tmp ); + uint64_t delta; + if( p >= v.value ) + { + delta = p - v.value; + } + else + { + delta = v.overflow - v.value + p; + } + v.value = p; + + TracyLfqPrepare( QueueType::SysPowerReport ); + MemWrite( &item->sysPower.time, Profiler::GetTime() ); + MemWrite( &item->sysPower.delta, delta ); + MemWrite( &item->sysPower.name, (uint64_t)v.name ); + TracyLfqCommit; + } + } + } +} + +void SysPower::ScanDirectory( const char* path, int parent ) +{ + DIR* dir = opendir( path ); + if( !dir ) return; + struct dirent* ent; + uint64_t maxRange = 0; + char* name = nullptr; + FILE* handle = nullptr; + while( ( ent = readdir( dir ) ) ) + { + if( ent->d_type == DT_REG ) + { + if( strcmp( ent->d_name, "max_energy_range_uj" ) == 0 ) + { + char tmp[PATH_MAX]; + snprintf( tmp, PATH_MAX, "%s/max_energy_range_uj", path ); + FILE* f = fopen( tmp, "r" ); + if( f ) + { + void(fscanf( f, "%" PRIu64, &maxRange )); + fclose( f ); + } + } + else if( strcmp( ent->d_name, "name" ) == 0 ) + { + char tmp[PATH_MAX]; + snprintf( tmp, PATH_MAX, "%s/name", path ); + FILE* f = fopen( tmp, "r" ); + if( f ) + { + char ntmp[128]; + if( fgets( ntmp, 128, f ) ) + { + // Last character is newline, skip it + const auto sz = strlen( ntmp ) - 1; + if( parent < 0 ) + { + name = (char*)tracy_malloc( sz + 1 ); + memcpy( name, ntmp, sz ); + name[sz] = '\0'; + } + else + { + const auto p = m_domains[parent]; + const auto psz = strlen( p.name ); + name = (char*)tracy_malloc( psz + sz + 2 ); + memcpy( name, p.name, psz ); + name[psz] = ':'; + memcpy( name+psz+1, ntmp, sz ); + name[psz+sz+1] = '\0'; + } + } + fclose( f ); + } + } + else if( strcmp( ent->d_name, "energy_uj" ) == 0 ) + { + char tmp[PATH_MAX]; + snprintf( tmp, PATH_MAX, "%s/energy_uj", path ); + handle = fopen( tmp, "r" ); + } + } + if( name && handle && maxRange > 0 ) break; + } + if( name && handle && maxRange > 0 ) + { + parent = (int)m_domains.size(); + Domain* domain = m_domains.push_next(); + domain->value = 0; + domain->overflow = maxRange; + domain->handle = handle; + domain->name = name; + TracyDebug( "Power domain id %i, %s found at %s\n", parent, name, path ); + } + else + { + if( name ) tracy_free( name ); + if( handle ) fclose( handle ); + } + + rewinddir( dir ); + while( ( ent = readdir( dir ) ) ) + { + if( ent->d_type == DT_DIR && strncmp( ent->d_name, "intel-rapl:", 11 ) == 0 ) + { + char tmp[PATH_MAX]; + snprintf( tmp, PATH_MAX, "%s/%s", path, ent->d_name ); + ScanDirectory( tmp, parent ); + } + } + closedir( dir ); +} + +} + +#endif diff --git a/externals/tracy/public/client/TracySysPower.hpp b/externals/tracy/public/client/TracySysPower.hpp new file mode 100644 index 000000000..210123bce --- /dev/null +++ b/externals/tracy/public/client/TracySysPower.hpp @@ -0,0 +1,44 @@ +#ifndef __TRACYSYSPOWER_HPP__ +#define __TRACYSYSPOWER_HPP__ + +#if defined __linux__ +# define TRACY_HAS_SYSPOWER +#endif + +#ifdef TRACY_HAS_SYSPOWER + +#include +#include + +#include "TracyFastVector.hpp" + +namespace tracy +{ + +class SysPower +{ + struct Domain + { + uint64_t value; + uint64_t overflow; + FILE* handle; + const char* name; + }; + +public: + SysPower(); + ~SysPower(); + + void Tick(); + +private: + void ScanDirectory( const char* path, int parent ); + + FastVector m_domains; + uint64_t m_lastTime; +}; + +} +#endif + +#endif diff --git a/externals/tracy/public/client/TracySysTime.cpp b/externals/tracy/public/client/TracySysTime.cpp new file mode 100644 index 000000000..b690a9114 --- /dev/null +++ b/externals/tracy/public/client/TracySysTime.cpp @@ -0,0 +1,108 @@ +#include "TracySysTime.hpp" + +#ifdef TRACY_HAS_SYSTIME + +# if defined _WIN32 +# include +# elif defined __linux__ +# include +# include +# elif defined __APPLE__ +# include +# include +# elif defined BSD +# include +# include +# endif + +namespace tracy +{ + +# if defined _WIN32 + +static inline uint64_t ConvertTime( const FILETIME& t ) +{ + return ( uint64_t( t.dwHighDateTime ) << 32 ) | uint64_t( t.dwLowDateTime ); +} + +void SysTime::ReadTimes() +{ + FILETIME idleTime; + FILETIME kernelTime; + FILETIME userTime; + + GetSystemTimes( &idleTime, &kernelTime, &userTime ); + + idle = ConvertTime( idleTime ); + const auto kernel = ConvertTime( kernelTime ); + const auto user = ConvertTime( userTime ); + used = kernel + user; +} + +# elif defined __linux__ + +void SysTime::ReadTimes() +{ + uint64_t user, nice, system; + FILE* f = fopen( "/proc/stat", "r" ); + if( f ) + { + int read = fscanf( f, "cpu %" PRIu64 " %" PRIu64 " %" PRIu64" %" PRIu64, &user, &nice, &system, &idle ); + fclose( f ); + if (read == 4) + { + used = user + nice + system; + } + } +} + +# elif defined __APPLE__ + +void SysTime::ReadTimes() +{ + host_cpu_load_info_data_t info; + mach_msg_type_number_t cnt = HOST_CPU_LOAD_INFO_COUNT; + host_statistics( mach_host_self(), HOST_CPU_LOAD_INFO, reinterpret_cast( &info ), &cnt ); + used = info.cpu_ticks[CPU_STATE_USER] + info.cpu_ticks[CPU_STATE_NICE] + info.cpu_ticks[CPU_STATE_SYSTEM]; + idle = info.cpu_ticks[CPU_STATE_IDLE]; +} + +# elif defined BSD + +void SysTime::ReadTimes() +{ + u_long data[5]; + size_t sz = sizeof( data ); + sysctlbyname( "kern.cp_time", &data, &sz, nullptr, 0 ); + used = data[0] + data[1] + data[2] + data[3]; + idle = data[4]; +} + +#endif + +SysTime::SysTime() +{ + ReadTimes(); +} + +float SysTime::Get() +{ + const auto oldUsed = used; + const auto oldIdle = idle; + + ReadTimes(); + + const auto diffIdle = idle - oldIdle; + const auto diffUsed = used - oldUsed; + +#if defined _WIN32 + return diffUsed == 0 ? -1 : ( diffUsed - diffIdle ) * 100.f / diffUsed; +#elif defined __linux__ || defined __APPLE__ || defined BSD + const auto total = diffUsed + diffIdle; + return total == 0 ? -1 : diffUsed * 100.f / total; +#endif +} + +} + +#endif diff --git a/externals/tracy/public/client/TracySysTime.hpp b/externals/tracy/public/client/TracySysTime.hpp new file mode 100644 index 000000000..cb5ebe736 --- /dev/null +++ b/externals/tracy/public/client/TracySysTime.hpp @@ -0,0 +1,36 @@ +#ifndef __TRACYSYSTIME_HPP__ +#define __TRACYSYSTIME_HPP__ + +#if defined _WIN32 || defined __linux__ || defined __APPLE__ +# define TRACY_HAS_SYSTIME +#else +# include +#endif + +#ifdef BSD +# define TRACY_HAS_SYSTIME +#endif + +#ifdef TRACY_HAS_SYSTIME + +#include + +namespace tracy +{ + +class SysTime +{ +public: + SysTime(); + float Get(); + + void ReadTimes(); + +private: + uint64_t idle, used; +}; + +} +#endif + +#endif diff --git a/externals/tracy/public/client/TracySysTrace.cpp b/externals/tracy/public/client/TracySysTrace.cpp new file mode 100644 index 000000000..af0641fef --- /dev/null +++ b/externals/tracy/public/client/TracySysTrace.cpp @@ -0,0 +1,1602 @@ +#include "TracyDebug.hpp" +#include "TracyStringHelpers.hpp" +#include "TracySysTrace.hpp" +#include "../common/TracySystem.hpp" + +#ifdef TRACY_HAS_SYSTEM_TRACING + +#ifndef TRACY_SAMPLING_HZ +# if defined _WIN32 +# define TRACY_SAMPLING_HZ 8000 +# elif defined __linux__ +# define TRACY_SAMPLING_HZ 10000 +# endif +#endif + +namespace tracy +{ + +static constexpr int GetSamplingFrequency() +{ +#if defined _WIN32 + return TRACY_SAMPLING_HZ > 8000 ? 8000 : ( TRACY_SAMPLING_HZ < 1 ? 1 : TRACY_SAMPLING_HZ ); +#else + return TRACY_SAMPLING_HZ > 1000000 ? 1000000 : ( TRACY_SAMPLING_HZ < 1 ? 1 : TRACY_SAMPLING_HZ ); +#endif +} + +static constexpr int GetSamplingPeriod() +{ + return 1000000000 / GetSamplingFrequency(); +} + +} + +# if defined _WIN32 + +# ifndef NOMINMAX +# define NOMINMAX +# endif + +# define INITGUID +# include +# include +# include +# include +# include +# include +# include +# include + +# include "../common/TracyAlloc.hpp" +# include "../common/TracySystem.hpp" +# include "TracyProfiler.hpp" +# include "TracyThread.hpp" + +namespace tracy +{ + +static const GUID PerfInfoGuid = { 0xce1dbfb4, 0x137e, 0x4da6, { 0x87, 0xb0, 0x3f, 0x59, 0xaa, 0x10, 0x2c, 0xbc } }; +static const GUID DxgKrnlGuid = { 0x802ec45a, 0x1e99, 0x4b83, { 0x99, 0x20, 0x87, 0xc9, 0x82, 0x77, 0xba, 0x9d } }; +static const GUID ThreadV2Guid = { 0x3d6fa8d1, 0xfe05, 0x11d0, { 0x9d, 0xda, 0x00, 0xc0, 0x4f, 0xd7, 0xba, 0x7c } }; + + +static TRACEHANDLE s_traceHandle; +static TRACEHANDLE s_traceHandle2; +static EVENT_TRACE_PROPERTIES* s_prop; +static DWORD s_pid; + +static EVENT_TRACE_PROPERTIES* s_propVsync; +static TRACEHANDLE s_traceHandleVsync; +static TRACEHANDLE s_traceHandleVsync2; +Thread* s_threadVsync = nullptr; + +struct CSwitch +{ + uint32_t newThreadId; + uint32_t oldThreadId; + int8_t newThreadPriority; + int8_t oldThreadPriority; + uint8_t previousCState; + int8_t spareByte; + int8_t oldThreadWaitReason; + int8_t oldThreadWaitMode; + int8_t oldThreadState; + int8_t oldThreadWaitIdealProcessor; + uint32_t newThreadWaitTime; + uint32_t reserved; +}; + +struct ReadyThread +{ + uint32_t threadId; + int8_t adjustReason; + int8_t adjustIncrement; + int8_t flag; + int8_t reserverd; +}; + +struct ThreadTrace +{ + uint32_t processId; + uint32_t threadId; + uint32_t stackBase; + uint32_t stackLimit; + uint32_t userStackBase; + uint32_t userStackLimit; + uint32_t startAddr; + uint32_t win32StartAddr; + uint32_t tebBase; + uint32_t subProcessTag; +}; + +struct StackWalkEvent +{ + uint64_t eventTimeStamp; + uint32_t stackProcess; + uint32_t stackThread; + uint64_t stack[192]; +}; + +struct VSyncInfo +{ + void* dxgAdapter; + uint32_t vidPnTargetId; + uint64_t scannedPhysicalAddress; + uint32_t vidPnSourceId; + uint32_t frameNumber; + int64_t frameQpcTime; + void* hFlipDevice; + uint32_t flipType; + uint64_t flipFenceId; +}; + +extern "C" typedef NTSTATUS (WINAPI *t_NtQueryInformationThread)( HANDLE, THREADINFOCLASS, PVOID, ULONG, PULONG ); +extern "C" typedef BOOL (WINAPI *t_EnumProcessModules)( HANDLE, HMODULE*, DWORD, LPDWORD ); +extern "C" typedef BOOL (WINAPI *t_GetModuleInformation)( HANDLE, HMODULE, LPMODULEINFO, DWORD ); +extern "C" typedef DWORD (WINAPI *t_GetModuleBaseNameA)( HANDLE, HMODULE, LPSTR, DWORD ); +extern "C" typedef HRESULT (WINAPI *t_GetThreadDescription)( HANDLE, PWSTR* ); + +t_NtQueryInformationThread NtQueryInformationThread = (t_NtQueryInformationThread)GetProcAddress( GetModuleHandleA( "ntdll.dll" ), "NtQueryInformationThread" ); +t_EnumProcessModules _EnumProcessModules = (t_EnumProcessModules)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "K32EnumProcessModules" ); +t_GetModuleInformation _GetModuleInformation = (t_GetModuleInformation)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "K32GetModuleInformation" ); +t_GetModuleBaseNameA _GetModuleBaseNameA = (t_GetModuleBaseNameA)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "K32GetModuleBaseNameA" ); + +static t_GetThreadDescription _GetThreadDescription = 0; + + +void WINAPI EventRecordCallback( PEVENT_RECORD record ) +{ +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + + const auto& hdr = record->EventHeader; + switch( hdr.ProviderId.Data1 ) + { + case 0x3d6fa8d1: // Thread Guid + if( hdr.EventDescriptor.Opcode == 36 ) + { + const auto cswitch = (const CSwitch*)record->UserData; + + TracyLfqPrepare( QueueType::ContextSwitch ); + MemWrite( &item->contextSwitch.time, hdr.TimeStamp.QuadPart ); + MemWrite( &item->contextSwitch.oldThread, cswitch->oldThreadId ); + MemWrite( &item->contextSwitch.newThread, cswitch->newThreadId ); + MemWrite( &item->contextSwitch.cpu, record->BufferContext.ProcessorNumber ); + MemWrite( &item->contextSwitch.reason, cswitch->oldThreadWaitReason ); + MemWrite( &item->contextSwitch.state, cswitch->oldThreadState ); + TracyLfqCommit; + } + else if( hdr.EventDescriptor.Opcode == 50 ) + { + const auto rt = (const ReadyThread*)record->UserData; + + TracyLfqPrepare( QueueType::ThreadWakeup ); + MemWrite( &item->threadWakeup.time, hdr.TimeStamp.QuadPart ); + MemWrite( &item->threadWakeup.thread, rt->threadId ); + TracyLfqCommit; + } + else if( hdr.EventDescriptor.Opcode == 1 || hdr.EventDescriptor.Opcode == 3 ) + { + const auto tt = (const ThreadTrace*)record->UserData; + + uint64_t tid = tt->threadId; + if( tid == 0 ) return; + uint64_t pid = tt->processId; + TracyLfqPrepare( QueueType::TidToPid ); + MemWrite( &item->tidToPid.tid, tid ); + MemWrite( &item->tidToPid.pid, pid ); + TracyLfqCommit; + } + break; + case 0xdef2fe46: // StackWalk Guid + if( hdr.EventDescriptor.Opcode == 32 ) + { + const auto sw = (const StackWalkEvent*)record->UserData; + if( sw->stackProcess == s_pid ) + { + const uint64_t sz = ( record->UserDataLength - 16 ) / 8; + if( sz > 0 ) + { + auto trace = (uint64_t*)tracy_malloc( ( 1 + sz ) * sizeof( uint64_t ) ); + memcpy( trace, &sz, sizeof( uint64_t ) ); + memcpy( trace+1, sw->stack, sizeof( uint64_t ) * sz ); + TracyLfqPrepare( QueueType::CallstackSample ); + MemWrite( &item->callstackSampleFat.time, sw->eventTimeStamp ); + MemWrite( &item->callstackSampleFat.thread, sw->stackThread ); + MemWrite( &item->callstackSampleFat.ptr, (uint64_t)trace ); + TracyLfqCommit; + } + } + } + break; + default: + break; + } +} + +void WINAPI EventRecordCallbackVsync( PEVENT_RECORD record ) +{ +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return; +#endif + + const auto& hdr = record->EventHeader; + assert( hdr.ProviderId.Data1 == 0x802EC45A ); + assert( hdr.EventDescriptor.Id == 0x0011 ); + + const auto vs = (const VSyncInfo*)record->UserData; + + TracyLfqPrepare( QueueType::FrameVsync ); + MemWrite( &item->frameVsync.time, hdr.TimeStamp.QuadPart ); + MemWrite( &item->frameVsync.id, vs->vidPnTargetId ); + TracyLfqCommit; +} + +static void SetupVsync() +{ +#if _WIN32_WINNT >= _WIN32_WINNT_WINBLUE && !defined(__MINGW32__) + const auto psz = sizeof( EVENT_TRACE_PROPERTIES ) + MAX_PATH; + s_propVsync = (EVENT_TRACE_PROPERTIES*)tracy_malloc( psz ); + memset( s_propVsync, 0, sizeof( EVENT_TRACE_PROPERTIES ) ); + s_propVsync->LogFileMode = EVENT_TRACE_REAL_TIME_MODE; + s_propVsync->Wnode.BufferSize = psz; +#ifdef TRACY_TIMER_QPC + s_propVsync->Wnode.ClientContext = 1; +#else + s_propVsync->Wnode.ClientContext = 3; +#endif + s_propVsync->LoggerNameOffset = sizeof( EVENT_TRACE_PROPERTIES ); + strcpy( ((char*)s_propVsync) + sizeof( EVENT_TRACE_PROPERTIES ), "TracyVsync" ); + + auto backup = tracy_malloc( psz ); + memcpy( backup, s_propVsync, psz ); + + const auto controlStatus = ControlTraceA( 0, "TracyVsync", s_propVsync, EVENT_TRACE_CONTROL_STOP ); + if( controlStatus != ERROR_SUCCESS && controlStatus != ERROR_WMI_INSTANCE_NOT_FOUND ) + { + tracy_free( backup ); + tracy_free( s_propVsync ); + return; + } + + memcpy( s_propVsync, backup, psz ); + tracy_free( backup ); + + const auto startStatus = StartTraceA( &s_traceHandleVsync, "TracyVsync", s_propVsync ); + if( startStatus != ERROR_SUCCESS ) + { + tracy_free( s_propVsync ); + return; + } + + EVENT_FILTER_EVENT_ID fe = {}; + fe.FilterIn = TRUE; + fe.Count = 1; + fe.Events[0] = 0x0011; // VSyncDPC_Info + + EVENT_FILTER_DESCRIPTOR desc = {}; + desc.Ptr = (ULONGLONG)&fe; + desc.Size = sizeof( fe ); + desc.Type = EVENT_FILTER_TYPE_EVENT_ID; + + ENABLE_TRACE_PARAMETERS params = {}; + params.Version = ENABLE_TRACE_PARAMETERS_VERSION_2; + params.EnableProperty = EVENT_ENABLE_PROPERTY_IGNORE_KEYWORD_0; + params.SourceId = s_propVsync->Wnode.Guid; + params.EnableFilterDesc = &desc; + params.FilterDescCount = 1; + + uint64_t mask = 0x4000000000000001; // Microsoft_Windows_DxgKrnl_Performance | Base + if( EnableTraceEx2( s_traceHandleVsync, &DxgKrnlGuid, EVENT_CONTROL_CODE_ENABLE_PROVIDER, TRACE_LEVEL_INFORMATION, mask, mask, 0, ¶ms ) != ERROR_SUCCESS ) + { + tracy_free( s_propVsync ); + return; + } + + char loggerName[MAX_PATH]; + strcpy( loggerName, "TracyVsync" ); + + EVENT_TRACE_LOGFILEA log = {}; + log.LoggerName = loggerName; + log.ProcessTraceMode = PROCESS_TRACE_MODE_REAL_TIME | PROCESS_TRACE_MODE_EVENT_RECORD | PROCESS_TRACE_MODE_RAW_TIMESTAMP; + log.EventRecordCallback = EventRecordCallbackVsync; + + s_traceHandleVsync2 = OpenTraceA( &log ); + if( s_traceHandleVsync2 == (TRACEHANDLE)INVALID_HANDLE_VALUE ) + { + CloseTrace( s_traceHandleVsync ); + tracy_free( s_propVsync ); + return; + } + + s_threadVsync = (Thread*)tracy_malloc( sizeof( Thread ) ); + new(s_threadVsync) Thread( [] (void*) { + ThreadExitHandler threadExitHandler; + SetThreadPriority( GetCurrentThread(), THREAD_PRIORITY_TIME_CRITICAL ); + SetThreadName( "Tracy Vsync" ); + ProcessTrace( &s_traceHandleVsync2, 1, nullptr, nullptr ); + }, nullptr ); +#endif +} + +static constexpr int GetSamplingInterval() +{ + return GetSamplingPeriod() / 100; +} + +bool SysTraceStart( int64_t& samplingPeriod ) +{ + if( !_GetThreadDescription ) _GetThreadDescription = (t_GetThreadDescription)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "GetThreadDescription" ); + + s_pid = GetCurrentProcessId(); + +#if defined _WIN64 + constexpr bool isOs64Bit = true; +#else + BOOL _iswow64; + IsWow64Process( GetCurrentProcess(), &_iswow64 ); + const bool isOs64Bit = _iswow64; +#endif + + TOKEN_PRIVILEGES priv = {}; + priv.PrivilegeCount = 1; + priv.Privileges[0].Attributes = SE_PRIVILEGE_ENABLED; + if( LookupPrivilegeValue( nullptr, SE_SYSTEM_PROFILE_NAME, &priv.Privileges[0].Luid ) == 0 ) return false; + + HANDLE pt; + if( OpenProcessToken( GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES, &pt ) == 0 ) return false; + const auto adjust = AdjustTokenPrivileges( pt, FALSE, &priv, 0, nullptr, nullptr ); + CloseHandle( pt ); + if( adjust == 0 ) return false; + const auto status = GetLastError(); + if( status != ERROR_SUCCESS ) return false; + + if( isOs64Bit ) + { + TRACE_PROFILE_INTERVAL interval = {}; + interval.Interval = GetSamplingInterval(); + const auto intervalStatus = TraceSetInformation( 0, TraceSampledProfileIntervalInfo, &interval, sizeof( interval ) ); + if( intervalStatus != ERROR_SUCCESS ) return false; + samplingPeriod = GetSamplingPeriod(); + } + + const auto psz = sizeof( EVENT_TRACE_PROPERTIES ) + sizeof( KERNEL_LOGGER_NAME ); + s_prop = (EVENT_TRACE_PROPERTIES*)tracy_malloc( psz ); + memset( s_prop, 0, sizeof( EVENT_TRACE_PROPERTIES ) ); + ULONG flags = 0; +#ifndef TRACY_NO_CONTEXT_SWITCH + flags = EVENT_TRACE_FLAG_CSWITCH | EVENT_TRACE_FLAG_DISPATCHER | EVENT_TRACE_FLAG_THREAD; +#endif +#ifndef TRACY_NO_SAMPLING + if( isOs64Bit ) flags |= EVENT_TRACE_FLAG_PROFILE; +#endif + s_prop->EnableFlags = flags; + s_prop->LogFileMode = EVENT_TRACE_REAL_TIME_MODE; + s_prop->Wnode.BufferSize = psz; + s_prop->Wnode.Flags = WNODE_FLAG_TRACED_GUID; +#ifdef TRACY_TIMER_QPC + s_prop->Wnode.ClientContext = 1; +#else + s_prop->Wnode.ClientContext = 3; +#endif + s_prop->Wnode.Guid = SystemTraceControlGuid; + s_prop->BufferSize = 1024; + s_prop->MinimumBuffers = std::thread::hardware_concurrency() * 4; + s_prop->MaximumBuffers = std::thread::hardware_concurrency() * 6; + s_prop->LoggerNameOffset = sizeof( EVENT_TRACE_PROPERTIES ); + memcpy( ((char*)s_prop) + sizeof( EVENT_TRACE_PROPERTIES ), KERNEL_LOGGER_NAME, sizeof( KERNEL_LOGGER_NAME ) ); + + auto backup = tracy_malloc( psz ); + memcpy( backup, s_prop, psz ); + + const auto controlStatus = ControlTrace( 0, KERNEL_LOGGER_NAME, s_prop, EVENT_TRACE_CONTROL_STOP ); + if( controlStatus != ERROR_SUCCESS && controlStatus != ERROR_WMI_INSTANCE_NOT_FOUND ) + { + tracy_free( backup ); + tracy_free( s_prop ); + return false; + } + + memcpy( s_prop, backup, psz ); + tracy_free( backup ); + + const auto startStatus = StartTrace( &s_traceHandle, KERNEL_LOGGER_NAME, s_prop ); + if( startStatus != ERROR_SUCCESS ) + { + tracy_free( s_prop ); + return false; + } + +#ifndef TRACY_NO_SAMPLING + if( isOs64Bit ) + { + CLASSIC_EVENT_ID stackId[2] = {}; + stackId[0].EventGuid = PerfInfoGuid; + stackId[0].Type = 46; + stackId[1].EventGuid = ThreadV2Guid; + stackId[1].Type = 36; + const auto stackStatus = TraceSetInformation( s_traceHandle, TraceStackTracingInfo, &stackId, sizeof( stackId ) ); + if( stackStatus != ERROR_SUCCESS ) + { + tracy_free( s_prop ); + return false; + } + } +#endif + +#ifdef UNICODE + WCHAR KernelLoggerName[sizeof( KERNEL_LOGGER_NAME )]; +#else + char KernelLoggerName[sizeof( KERNEL_LOGGER_NAME )]; +#endif + memcpy( KernelLoggerName, KERNEL_LOGGER_NAME, sizeof( KERNEL_LOGGER_NAME ) ); + EVENT_TRACE_LOGFILE log = {}; + log.LoggerName = KernelLoggerName; + log.ProcessTraceMode = PROCESS_TRACE_MODE_REAL_TIME | PROCESS_TRACE_MODE_EVENT_RECORD | PROCESS_TRACE_MODE_RAW_TIMESTAMP; + log.EventRecordCallback = EventRecordCallback; + + s_traceHandle2 = OpenTrace( &log ); + if( s_traceHandle2 == (TRACEHANDLE)INVALID_HANDLE_VALUE ) + { + CloseTrace( s_traceHandle ); + tracy_free( s_prop ); + return false; + } + +#ifndef TRACY_NO_VSYNC_CAPTURE + SetupVsync(); +#endif + + return true; +} + +void SysTraceStop() +{ + if( s_threadVsync ) + { + CloseTrace( s_traceHandleVsync2 ); + CloseTrace( s_traceHandleVsync ); + s_threadVsync->~Thread(); + tracy_free( s_threadVsync ); + } + + CloseTrace( s_traceHandle2 ); + CloseTrace( s_traceHandle ); +} + +void SysTraceWorker( void* ptr ) +{ + ThreadExitHandler threadExitHandler; + SetThreadPriority( GetCurrentThread(), THREAD_PRIORITY_TIME_CRITICAL ); + SetThreadName( "Tracy SysTrace" ); + ProcessTrace( &s_traceHandle2, 1, 0, 0 ); + ControlTrace( 0, KERNEL_LOGGER_NAME, s_prop, EVENT_TRACE_CONTROL_STOP ); + tracy_free( s_prop ); +} + +void SysTraceGetExternalName( uint64_t thread, const char*& threadName, const char*& name ) +{ + bool threadSent = false; + auto hnd = OpenThread( THREAD_QUERY_INFORMATION, FALSE, DWORD( thread ) ); + if( hnd == 0 ) + { + hnd = OpenThread( THREAD_QUERY_LIMITED_INFORMATION, FALSE, DWORD( thread ) ); + } + if( hnd != 0 ) + { + if( _GetThreadDescription ) + { + PWSTR tmp; + _GetThreadDescription( hnd, &tmp ); + char buf[256]; + if( tmp ) + { + auto ret = wcstombs( buf, tmp, 256 ); + if( ret != 0 ) + { + threadName = CopyString( buf, ret ); + threadSent = true; + } + } + } + const auto pid = GetProcessIdOfThread( hnd ); + if( !threadSent && NtQueryInformationThread && _EnumProcessModules && _GetModuleInformation && _GetModuleBaseNameA ) + { + void* ptr; + ULONG retlen; + auto status = NtQueryInformationThread( hnd, (THREADINFOCLASS)9 /*ThreadQuerySetWin32StartAddress*/, &ptr, sizeof( &ptr ), &retlen ); + if( status == 0 ) + { + const auto phnd = OpenProcess( PROCESS_QUERY_INFORMATION | PROCESS_VM_READ, FALSE, pid ); + if( phnd != INVALID_HANDLE_VALUE ) + { + HMODULE modules[1024]; + DWORD needed; + if( _EnumProcessModules( phnd, modules, 1024 * sizeof( HMODULE ), &needed ) != 0 ) + { + const auto sz = std::min( DWORD( needed / sizeof( HMODULE ) ), DWORD( 1024 ) ); + for( DWORD i=0; i= (uint64_t)info.lpBaseOfDll && (uint64_t)ptr <= (uint64_t)info.lpBaseOfDll + (uint64_t)info.SizeOfImage ) + { + char buf2[1024]; + const auto modlen = _GetModuleBaseNameA( phnd, modules[i], buf2, 1024 ); + if( modlen != 0 ) + { + threadName = CopyString( buf2, modlen ); + threadSent = true; + } + } + } + } + } + CloseHandle( phnd ); + } + } + } + CloseHandle( hnd ); + if( !threadSent ) + { + threadName = CopyString( "???", 3 ); + threadSent = true; + } + if( pid != 0 ) + { + { + uint64_t _pid = pid; + TracyLfqPrepare( QueueType::TidToPid ); + MemWrite( &item->tidToPid.tid, thread ); + MemWrite( &item->tidToPid.pid, _pid ); + TracyLfqCommit; + } + if( pid == 4 ) + { + name = CopyStringFast( "System", 6 ); + return; + } + else + { + const auto phnd = OpenProcess( PROCESS_QUERY_LIMITED_INFORMATION, FALSE, pid ); + if( phnd != INVALID_HANDLE_VALUE ) + { + char buf2[1024]; + const auto sz = GetProcessImageFileNameA( phnd, buf2, 1024 ); + CloseHandle( phnd ); + if( sz != 0 ) + { + auto ptr = buf2 + sz - 1; + while( ptr > buf2 && *ptr != '\\' ) ptr--; + if( *ptr == '\\' ) ptr++; + name = CopyStringFast( ptr ); + return; + } + } + } + } + } + + if( !threadSent ) + { + threadName = CopyString( "???", 3 ); + } + name = CopyStringFast( "???", 3 ); +} + +} + +# elif defined __linux__ + +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include +# include + +# if defined __i386 || defined __x86_64__ +# include "TracyCpuid.hpp" +# endif + +# include "TracyProfiler.hpp" +# include "TracyRingBuffer.hpp" +# include "TracyThread.hpp" + +namespace tracy +{ + +static std::atomic traceActive { false }; +static int s_numCpus = 0; +static int s_numBuffers = 0; +static int s_ctxBufferIdx = 0; + +static RingBuffer* s_ring = nullptr; + +static const int ThreadHashSize = 4 * 1024; +static uint32_t s_threadHash[ThreadHashSize] = {}; + +static bool CurrentProcOwnsThread( uint32_t tid ) +{ + const auto hash = tid & ( ThreadHashSize-1 ); + const auto hv = s_threadHash[hash]; + if( hv == tid ) return true; + if( hv == -tid ) return false; + + char path[256]; + sprintf( path, "/proc/self/task/%d", tid ); + struct stat st; + if( stat( path, &st ) == 0 ) + { + s_threadHash[hash] = tid; + return true; + } + else + { + s_threadHash[hash] = -tid; + return false; + } +} + +static int perf_event_open( struct perf_event_attr* hw_event, pid_t pid, int cpu, int group_fd, unsigned long flags ) +{ + return syscall( __NR_perf_event_open, hw_event, pid, cpu, group_fd, flags ); +} + +enum TraceEventId +{ + EventCallstack, + EventCpuCycles, + EventInstructionsRetired, + EventCacheReference, + EventCacheMiss, + EventBranchRetired, + EventBranchMiss, + EventVsync, + EventContextSwitch, + EventWakeup, +}; + +static void ProbePreciseIp( perf_event_attr& pe, unsigned long long config0, unsigned long long config1, pid_t pid ) +{ + pe.config = config1; + pe.precise_ip = 3; + while( pe.precise_ip != 0 ) + { + const int fd = perf_event_open( &pe, pid, 0, -1, PERF_FLAG_FD_CLOEXEC ); + if( fd != -1 ) + { + close( fd ); + break; + } + pe.precise_ip--; + } + pe.config = config0; + while( pe.precise_ip != 0 ) + { + const int fd = perf_event_open( &pe, pid, 0, -1, PERF_FLAG_FD_CLOEXEC ); + if( fd != -1 ) + { + close( fd ); + break; + } + pe.precise_ip--; + } + TracyDebug( " Probed precise_ip: %i\n", pe.precise_ip ); +} + +static void ProbePreciseIp( perf_event_attr& pe, pid_t pid ) +{ + pe.precise_ip = 3; + while( pe.precise_ip != 0 ) + { + const int fd = perf_event_open( &pe, pid, 0, -1, PERF_FLAG_FD_CLOEXEC ); + if( fd != -1 ) + { + close( fd ); + break; + } + pe.precise_ip--; + } + TracyDebug( " Probed precise_ip: %i\n", pe.precise_ip ); +} + +static bool IsGenuineIntel() +{ +#if defined __i386 || defined __x86_64__ + uint32_t regs[4] = {}; + __get_cpuid( 0, regs, regs+1, regs+2, regs+3 ); + char manufacturer[12]; + memcpy( manufacturer, regs+1, 4 ); + memcpy( manufacturer+4, regs+3, 4 ); + memcpy( manufacturer+8, regs+2, 4 ); + return memcmp( manufacturer, "GenuineIntel", 12 ) == 0; +#else + return false; +#endif +} + +static const char* ReadFile( const char* path ) +{ + int fd = open( path, O_RDONLY ); + if( fd < 0 ) return nullptr; + + static char tmp[64]; + const auto cnt = read( fd, tmp, 63 ); + close( fd ); + if( cnt < 0 ) return nullptr; + tmp[cnt] = '\0'; + return tmp; +} + +bool SysTraceStart( int64_t& samplingPeriod ) +{ +#ifndef CLOCK_MONOTONIC_RAW + return false; +#endif + + const auto paranoidLevelStr = ReadFile( "/proc/sys/kernel/perf_event_paranoid" ); + if( !paranoidLevelStr ) return false; +#ifdef TRACY_VERBOSE + int paranoidLevel = 2; + paranoidLevel = atoi( paranoidLevelStr ); + TracyDebug( "perf_event_paranoid: %i\n", paranoidLevel ); +#endif + + int switchId = -1, wakeupId = -1, vsyncId = -1; + const auto switchIdStr = ReadFile( "/sys/kernel/debug/tracing/events/sched/sched_switch/id" ); + if( switchIdStr ) switchId = atoi( switchIdStr ); + const auto wakeupIdStr = ReadFile( "/sys/kernel/debug/tracing/events/sched/sched_wakeup/id" ); + if( wakeupIdStr ) wakeupId = atoi( wakeupIdStr ); + const auto vsyncIdStr = ReadFile( "/sys/kernel/debug/tracing/events/drm/drm_vblank_event/id" ); + if( vsyncIdStr ) vsyncId = atoi( vsyncIdStr ); + + TracyDebug( "sched_switch id: %i\n", switchId ); + TracyDebug( "sched_wakeup id: %i\n", wakeupId ); + TracyDebug( "drm_vblank_event id: %i\n", vsyncId ); + +#ifdef TRACY_NO_SAMPLING + const bool noSoftwareSampling = true; +#else + const char* noSoftwareSamplingEnv = GetEnvVar( "TRACY_NO_SAMPLING" ); + const bool noSoftwareSampling = noSoftwareSamplingEnv && noSoftwareSamplingEnv[0] == '1'; +#endif + +#ifdef TRACY_NO_SAMPLE_RETIREMENT + const bool noRetirement = true; +#else + const char* noRetirementEnv = GetEnvVar( "TRACY_NO_SAMPLE_RETIREMENT" ); + const bool noRetirement = noRetirementEnv && noRetirementEnv[0] == '1'; +#endif + +#ifdef TRACY_NO_SAMPLE_CACHE + const bool noCache = true; +#else + const char* noCacheEnv = GetEnvVar( "TRACY_NO_SAMPLE_CACHE" ); + const bool noCache = noCacheEnv && noCacheEnv[0] == '1'; +#endif + +#ifdef TRACY_NO_SAMPLE_BRANCH + const bool noBranch = true; +#else + const char* noBranchEnv = GetEnvVar( "TRACY_NO_SAMPLE_BRANCH" ); + const bool noBranch = noBranchEnv && noBranchEnv[0] == '1'; +#endif + +#ifdef TRACY_NO_CONTEXT_SWITCH + const bool noCtxSwitch = true; +#else + const char* noCtxSwitchEnv = GetEnvVar( "TRACY_NO_CONTEXT_SWITCH" ); + const bool noCtxSwitch = noCtxSwitchEnv && noCtxSwitchEnv[0] == '1'; +#endif + +#ifdef TRACY_NO_VSYNC_CAPTURE + const bool noVsync = true; +#else + const char* noVsyncEnv = GetEnvVar( "TRACY_NO_VSYNC_CAPTURE" ); + const bool noVsync = noVsyncEnv && noVsyncEnv[0] == '1'; +#endif + + samplingPeriod = GetSamplingPeriod(); + uint32_t currentPid = (uint32_t)getpid(); + + s_numCpus = (int)std::thread::hardware_concurrency(); + + const auto maxNumBuffers = s_numCpus * ( + 1 + // software sampling + 2 + // CPU cycles + instructions retired + 2 + // cache reference + miss + 2 + // branch retired + miss + 2 + // context switches + wakeups + 1 // vsync + ); + s_ring = (RingBuffer*)tracy_malloc( sizeof( RingBuffer ) * maxNumBuffers ); + s_numBuffers = 0; + + // software sampling + perf_event_attr pe = {}; + pe.type = PERF_TYPE_SOFTWARE; + pe.size = sizeof( perf_event_attr ); + pe.config = PERF_COUNT_SW_CPU_CLOCK; + pe.sample_freq = GetSamplingFrequency(); + pe.sample_type = PERF_SAMPLE_TID | PERF_SAMPLE_TIME | PERF_SAMPLE_CALLCHAIN; +#if LINUX_VERSION_CODE >= KERNEL_VERSION( 4, 8, 0 ) + pe.sample_max_stack = 127; +#endif + pe.disabled = 1; + pe.freq = 1; + pe.inherit = 1; +#if !defined TRACY_HW_TIMER || !( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) + pe.use_clockid = 1; + pe.clockid = CLOCK_MONOTONIC_RAW; +#endif + + if( !noSoftwareSampling ) + { + TracyDebug( "Setup software sampling\n" ); + ProbePreciseIp( pe, currentPid ); + for( int i=0; i= KERNEL_VERSION( 4, 8, 0 ) + pe.sample_max_stack = 127; +#endif + pe.disabled = 1; + pe.inherit = 1; + pe.config = switchId; +#if !defined TRACY_HW_TIMER || !( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) + pe.use_clockid = 1; + pe.clockid = CLOCK_MONOTONIC_RAW; +#endif + + TracyDebug( "Setup context switch capture\n" ); + for( int i=0; i> 63; + const auto m2 = test >> 47; + if( m1 == m2 ) break; + } + while( --cnt > 0 ); + for( uint64_t j=1; j> 63; + const auto m2 = test >> 47; + if( m1 != m2 ) trace[j] = 0; + } +#endif + + for( uint64_t j=1; j<=cnt; j++ ) + { + if( trace[j] >= (uint64_t)-4095 ) // PERF_CONTEXT_MAX + { + memmove( trace+j, trace+j+1, sizeof( uint64_t ) * ( cnt - j ) ); + cnt--; + } + } + + memcpy( trace, &cnt, sizeof( uint64_t ) ); + return trace; +} + +void SysTraceWorker( void* ptr ) +{ + ThreadExitHandler threadExitHandler; + SetThreadName( "Tracy Sampling" ); + InitRpmalloc(); + sched_param sp = { 99 }; + if( pthread_setschedparam( pthread_self(), SCHED_FIFO, &sp ) != 0 ) TracyDebug( "Failed to increase SysTraceWorker thread priority!\n" ); + auto ctxBufferIdx = s_ctxBufferIdx; + auto ringArray = s_ring; + auto numBuffers = s_numBuffers; + for( int i=0; i tail ); + hadData = true; + + const auto id = ring.GetId(); + assert( id != EventContextSwitch ); + const auto end = head - tail; + uint64_t pos = 0; + if( id == EventCallstack ) + { + while( pos < end ) + { + perf_event_header hdr; + ring.Read( &hdr, pos, sizeof( perf_event_header ) ); + if( hdr.type == PERF_RECORD_SAMPLE ) + { + auto offset = pos + sizeof( perf_event_header ); + + // Layout: + // u32 pid, tid + // u64 time + // u64 cnt + // u64 ip[cnt] + + uint32_t tid; + uint64_t t0; + uint64_t cnt; + + offset += sizeof( uint32_t ); + ring.Read( &tid, offset, sizeof( uint32_t ) ); + offset += sizeof( uint32_t ); + ring.Read( &t0, offset, sizeof( uint64_t ) ); + offset += sizeof( uint64_t ); + ring.Read( &cnt, offset, sizeof( uint64_t ) ); + offset += sizeof( uint64_t ); + + if( cnt > 0 ) + { +#if defined TRACY_HW_TIMER && ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) + t0 = ring.ConvertTimeToTsc( t0 ); +#endif + auto trace = GetCallstackBlock( cnt, ring, offset ); + + TracyLfqPrepare( QueueType::CallstackSample ); + MemWrite( &item->callstackSampleFat.time, t0 ); + MemWrite( &item->callstackSampleFat.thread, tid ); + MemWrite( &item->callstackSampleFat.ptr, (uint64_t)trace ); + TracyLfqCommit; + } + } + pos += hdr.size; + } + } + else + { + while( pos < end ) + { + perf_event_header hdr; + ring.Read( &hdr, pos, sizeof( perf_event_header ) ); + if( hdr.type == PERF_RECORD_SAMPLE ) + { + auto offset = pos + sizeof( perf_event_header ); + + // Layout: + // u64 ip + // u64 time + + uint64_t ip, t0; + ring.Read( &ip, offset, sizeof( uint64_t ) ); + offset += sizeof( uint64_t ); + ring.Read( &t0, offset, sizeof( uint64_t ) ); + +#if defined TRACY_HW_TIMER && ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) + t0 = ring.ConvertTimeToTsc( t0 ); +#endif + QueueType type; + switch( id ) + { + case EventCpuCycles: + type = QueueType::HwSampleCpuCycle; + break; + case EventInstructionsRetired: + type = QueueType::HwSampleInstructionRetired; + break; + case EventCacheReference: + type = QueueType::HwSampleCacheReference; + break; + case EventCacheMiss: + type = QueueType::HwSampleCacheMiss; + break; + case EventBranchRetired: + type = QueueType::HwSampleBranchRetired; + break; + case EventBranchMiss: + type = QueueType::HwSampleBranchMiss; + break; + default: + abort(); + } + + TracyLfqPrepare( type ); + MemWrite( &item->hwSample.ip, ip ); + MemWrite( &item->hwSample.time, t0 ); + TracyLfqCommit; + } + pos += hdr.size; + } + } + assert( pos == end ); + ring.Advance( end ); + } + if( !traceActive.load( std::memory_order_relaxed ) ) break; + + if( ctxBufferIdx != numBuffers ) + { + const auto ctxBufNum = numBuffers - ctxBufferIdx; + + int activeNum = 0; + uint16_t active[512]; + uint32_t end[512]; + uint32_t pos[512]; + for( int i=0; i 0 ) + { + hadData = true; + while( activeNum > 0 ) + { + int sel = -1; + int selPos; + int64_t t0 = std::numeric_limits::max(); + for( int i=0; i= 0 ) + { + auto& ring = ringArray[ctxBufferIdx + sel]; + auto rbPos = pos[sel]; + auto offset = rbPos; + perf_event_header hdr; + ring.Read( &hdr, offset, sizeof( perf_event_header ) ); + +#if defined TRACY_HW_TIMER && ( defined __i386 || defined _M_IX86 || defined __x86_64__ || defined _M_X64 ) + t0 = ring.ConvertTimeToTsc( t0 ); +#endif + + const auto rid = ring.GetId(); + if( rid == EventContextSwitch ) + { + // Layout: + // u64 time + // u64 cnt + // u64 ip[cnt] + // u32 size + // u8 data[size] + // Data (not ABI stable, but has not changed since it was added, in 2009): + // u8 hdr[8] + // u8 prev_comm[16] + // u32 prev_pid + // u32 prev_prio + // lng prev_state + // u8 next_comm[16] + // u32 next_pid + // u32 next_prio + + offset += sizeof( perf_event_header ) + sizeof( uint64_t ); + + uint64_t cnt; + ring.Read( &cnt, offset, sizeof( uint64_t ) ); + offset += sizeof( uint64_t ); + const auto traceOffset = offset; + offset += sizeof( uint64_t ) * cnt + sizeof( uint32_t ) + 8 + 16; + + uint32_t prev_pid, next_pid; + long prev_state; + + ring.Read( &prev_pid, offset, sizeof( uint32_t ) ); + offset += sizeof( uint32_t ) + sizeof( uint32_t ); + ring.Read( &prev_state, offset, sizeof( long ) ); + offset += sizeof( long ) + 16; + ring.Read( &next_pid, offset, sizeof( uint32_t ) ); + + uint8_t reason = 100; + uint8_t state; + + if( prev_state & 0x0001 ) state = 104; + else if( prev_state & 0x0002 ) state = 101; + else if( prev_state & 0x0004 ) state = 105; + else if( prev_state & 0x0008 ) state = 106; + else if( prev_state & 0x0010 ) state = 108; + else if( prev_state & 0x0020 ) state = 109; + else if( prev_state & 0x0040 ) state = 110; + else if( prev_state & 0x0080 ) state = 102; + else state = 103; + + TracyLfqPrepare( QueueType::ContextSwitch ); + MemWrite( &item->contextSwitch.time, t0 ); + MemWrite( &item->contextSwitch.oldThread, prev_pid ); + MemWrite( &item->contextSwitch.newThread, next_pid ); + MemWrite( &item->contextSwitch.cpu, uint8_t( ring.GetCpu() ) ); + MemWrite( &item->contextSwitch.reason, reason ); + MemWrite( &item->contextSwitch.state, state ); + TracyLfqCommit; + + if( cnt > 0 && prev_pid != 0 && CurrentProcOwnsThread( prev_pid ) ) + { + auto trace = GetCallstackBlock( cnt, ring, traceOffset ); + + TracyLfqPrepare( QueueType::CallstackSampleContextSwitch ); + MemWrite( &item->callstackSampleFat.time, t0 ); + MemWrite( &item->callstackSampleFat.thread, prev_pid ); + MemWrite( &item->callstackSampleFat.ptr, (uint64_t)trace ); + TracyLfqCommit; + } + } + else if( rid == EventWakeup ) + { + // Layout: + // u64 time + // u32 size + // u8 data[size] + // Data: + // u8 hdr[8] + // u8 comm[16] + // u32 pid + // u32 prio + // u64 target_cpu + + offset += sizeof( perf_event_header ) + sizeof( uint64_t ) + sizeof( uint32_t ) + 8 + 16; + + uint32_t pid; + ring.Read( &pid, offset, sizeof( uint32_t ) ); + + TracyLfqPrepare( QueueType::ThreadWakeup ); + MemWrite( &item->threadWakeup.time, t0 ); + MemWrite( &item->threadWakeup.thread, pid ); + TracyLfqCommit; + } + else + { + assert( rid == EventVsync ); + // Layout: + // u64 time + // u32 size + // u8 data[size] + // Data (not ABI stable): + // u8 hdr[8] + // i32 crtc + // u32 seq + // i64 ktime + // u8 high precision + + offset += sizeof( perf_event_header ) + sizeof( uint64_t ) + sizeof( uint32_t ) + 8; + + int32_t crtc; + ring.Read( &crtc, offset, sizeof( int32_t ) ); + + // Note: The timestamp value t0 might be off by a number of microseconds from the + // true hardware vblank event. The ktime value should be used instead, but it is + // measured in CLOCK_MONOTONIC time. Tracy only supports the timestamp counter + // register (TSC) or CLOCK_MONOTONIC_RAW clock. +#if 0 + offset += sizeof( uint32_t ) * 2; + int64_t ktime; + ring.Read( &ktime, offset, sizeof( int64_t ) ); +#endif + + TracyLfqPrepare( QueueType::FrameVsync ); + MemWrite( &item->frameVsync.id, crtc ); + MemWrite( &item->frameVsync.time, t0 ); + TracyLfqCommit; + } + + rbPos += hdr.size; + if( rbPos == end[sel] ) + { + memmove( active+selPos, active+selPos+1, sizeof(*active) * ( activeNum - selPos - 1 ) ); + activeNum--; + } + else + { + pos[sel] = rbPos; + } + } + } + for( int i=0; i 0 && buf[sz-1] == '\n' ) buf[sz-1] = '\0'; + threadName = CopyString( buf ); + fclose( f ); + } + else + { + threadName = CopyString( "???", 3 ); + } + + sprintf( fn, "/proc/%" PRIu64 "/status", thread ); + f = fopen( fn, "rb" ); + if( f ) + { + char* tmp = (char*)tracy_malloc_fast( 8*1024 ); + const auto fsz = (ptrdiff_t)fread( tmp, 1, 8*1024, f ); + fclose( f ); + + int pid = -1; + auto line = tmp; + for(;;) + { + if( memcmp( "Tgid:\t", line, 6 ) == 0 ) + { + pid = atoi( line + 6 ); + break; + } + while( line - tmp < fsz && *line != '\n' ) line++; + if( *line != '\n' ) break; + line++; + } + tracy_free_fast( tmp ); + + if( pid >= 0 ) + { + { + uint64_t _pid = pid; + TracyLfqPrepare( QueueType::TidToPid ); + MemWrite( &item->tidToPid.tid, thread ); + MemWrite( &item->tidToPid.pid, _pid ); + TracyLfqCommit; + } + sprintf( fn, "/proc/%i/comm", pid ); + f = fopen( fn, "rb" ); + if( f ) + { + char buf[256]; + const auto sz = fread( buf, 1, 256, f ); + if( sz > 0 && buf[sz-1] == '\n' ) buf[sz-1] = '\0'; + name = CopyStringFast( buf ); + fclose( f ); + return; + } + } + } + name = CopyStringFast( "???", 3 ); +} + +} + +# endif + +#endif diff --git a/externals/tracy/public/client/TracySysTrace.hpp b/externals/tracy/public/client/TracySysTrace.hpp new file mode 100644 index 000000000..8c663cd7a --- /dev/null +++ b/externals/tracy/public/client/TracySysTrace.hpp @@ -0,0 +1,28 @@ +#ifndef __TRACYSYSTRACE_HPP__ +#define __TRACYSYSTRACE_HPP__ + +#if !defined TRACY_NO_SYSTEM_TRACING && ( defined _WIN32 || defined __linux__ ) +# include "../common/TracyUwp.hpp" +# ifndef TRACY_UWP +# define TRACY_HAS_SYSTEM_TRACING +# endif +#endif + +#ifdef TRACY_HAS_SYSTEM_TRACING + +#include + +namespace tracy +{ + +bool SysTraceStart( int64_t& samplingPeriod ); +void SysTraceStop(); +void SysTraceWorker( void* ptr ); + +void SysTraceGetExternalName( uint64_t thread, const char*& threadName, const char*& name ); + +} + +#endif + +#endif diff --git a/externals/tracy/public/client/TracyThread.hpp b/externals/tracy/public/client/TracyThread.hpp new file mode 100644 index 000000000..5638756ac --- /dev/null +++ b/externals/tracy/public/client/TracyThread.hpp @@ -0,0 +1,90 @@ +#ifndef __TRACYTHREAD_HPP__ +#define __TRACYTHREAD_HPP__ + +#if defined _WIN32 +# include +#else +# include +#endif + +#ifdef TRACY_MANUAL_LIFETIME +# include "tracy_rpmalloc.hpp" +#endif + +namespace tracy +{ + +#ifdef TRACY_MANUAL_LIFETIME +extern thread_local bool RpThreadInitDone; +#endif + +class ThreadExitHandler +{ +public: + ~ThreadExitHandler() + { +#ifdef TRACY_MANUAL_LIFETIME + rpmalloc_thread_finalize( 1 ); + RpThreadInitDone = false; +#endif + } +}; + +#if defined _WIN32 + +class Thread +{ +public: + Thread( void(*func)( void* ptr ), void* ptr ) + : m_func( func ) + , m_ptr( ptr ) + , m_hnd( CreateThread( nullptr, 0, Launch, this, 0, nullptr ) ) + {} + + ~Thread() + { + WaitForSingleObject( m_hnd, INFINITE ); + CloseHandle( m_hnd ); + } + + HANDLE Handle() const { return m_hnd; } + +private: + static DWORD WINAPI Launch( void* ptr ) { ((Thread*)ptr)->m_func( ((Thread*)ptr)->m_ptr ); return 0; } + + void(*m_func)( void* ptr ); + void* m_ptr; + HANDLE m_hnd; +}; + +#else + +class Thread +{ +public: + Thread( void(*func)( void* ptr ), void* ptr ) + : m_func( func ) + , m_ptr( ptr ) + { + pthread_create( &m_thread, nullptr, Launch, this ); + } + + ~Thread() + { + pthread_join( m_thread, nullptr ); + } + + pthread_t Handle() const { return m_thread; } + +private: + static void* Launch( void* ptr ) { ((Thread*)ptr)->m_func( ((Thread*)ptr)->m_ptr ); return nullptr; } + void(*m_func)( void* ptr ); + void* m_ptr; + pthread_t m_thread; +}; + +#endif + +} + +#endif diff --git a/externals/tracy/public/client/tracy_SPSCQueue.h b/externals/tracy/public/client/tracy_SPSCQueue.h new file mode 100644 index 000000000..7f1752b56 --- /dev/null +++ b/externals/tracy/public/client/tracy_SPSCQueue.h @@ -0,0 +1,148 @@ +/* +Copyright (c) 2020 Erik Rigtorp + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. + */ + +#pragma once + +#include +#include +#include +#include +#include // std::enable_if, std::is_*_constructible + +#include "../common/TracyAlloc.hpp" + +#if defined (_MSC_VER) +#pragma warning(push) +#pragma warning(disable:4324) +#endif + +namespace tracy { + +template class SPSCQueue { +public: + explicit SPSCQueue(const size_t capacity) + : capacity_(capacity) { + capacity_++; // Needs one slack element + slots_ = (T*)tracy_malloc(sizeof(T) * (capacity_ + 2 * kPadding)); + + static_assert(alignof(SPSCQueue) == kCacheLineSize, ""); + static_assert(sizeof(SPSCQueue) >= 3 * kCacheLineSize, ""); + assert(reinterpret_cast(&readIdx_) - + reinterpret_cast(&writeIdx_) >= + static_cast(kCacheLineSize)); + } + + ~SPSCQueue() { + while (front()) { + pop(); + } + tracy_free(slots_); + } + + // non-copyable and non-movable + SPSCQueue(const SPSCQueue &) = delete; + SPSCQueue &operator=(const SPSCQueue &) = delete; + + template + void emplace(Args &&...args) noexcept( + std::is_nothrow_constructible::value) { + static_assert(std::is_constructible::value, + "T must be constructible with Args&&..."); + auto const writeIdx = writeIdx_.load(std::memory_order_relaxed); + auto nextWriteIdx = writeIdx + 1; + if (nextWriteIdx == capacity_) { + nextWriteIdx = 0; + } + while (nextWriteIdx == readIdxCache_) { + readIdxCache_ = readIdx_.load(std::memory_order_acquire); + } + new (&slots_[writeIdx + kPadding]) T(std::forward(args)...); + writeIdx_.store(nextWriteIdx, std::memory_order_release); + } + + T *front() noexcept { + auto const readIdx = readIdx_.load(std::memory_order_relaxed); + if (readIdx == writeIdxCache_) { + writeIdxCache_ = writeIdx_.load(std::memory_order_acquire); + if (writeIdxCache_ == readIdx) { + return nullptr; + } + } + return &slots_[readIdx + kPadding]; + } + + void pop() noexcept { + static_assert(std::is_nothrow_destructible::value, + "T must be nothrow destructible"); + auto const readIdx = readIdx_.load(std::memory_order_relaxed); + assert(writeIdx_.load(std::memory_order_acquire) != readIdx); + slots_[readIdx + kPadding].~T(); + auto nextReadIdx = readIdx + 1; + if (nextReadIdx == capacity_) { + nextReadIdx = 0; + } + readIdx_.store(nextReadIdx, std::memory_order_release); + } + + size_t size() const noexcept { + std::ptrdiff_t diff = writeIdx_.load(std::memory_order_acquire) - + readIdx_.load(std::memory_order_acquire); + if (diff < 0) { + diff += capacity_; + } + return static_cast(diff); + } + + bool empty() const noexcept { + return writeIdx_.load(std::memory_order_acquire) == + readIdx_.load(std::memory_order_acquire); + } + + size_t capacity() const noexcept { return capacity_ - 1; } + +private: + static constexpr size_t kCacheLineSize = 64; + + // Padding to avoid false sharing between slots_ and adjacent allocations + static constexpr size_t kPadding = (kCacheLineSize - 1) / sizeof(T) + 1; + +private: + size_t capacity_; + T *slots_; + + // Align to cache line size in order to avoid false sharing + // readIdxCache_ and writeIdxCache_ is used to reduce the amount of cache + // coherency traffic + alignas(kCacheLineSize) std::atomic writeIdx_ = {0}; + alignas(kCacheLineSize) size_t readIdxCache_ = 0; + alignas(kCacheLineSize) std::atomic readIdx_ = {0}; + alignas(kCacheLineSize) size_t writeIdxCache_ = 0; + + // Padding to avoid adjacent allocations to share cache line with + // writeIdxCache_ + char padding_[kCacheLineSize - sizeof(SPSCQueue::writeIdxCache_)]; +}; +} // namespace rigtorp + +#if defined (_MSC_VER) +#pragma warning(pop) +#endif diff --git a/externals/tracy/public/client/tracy_concurrentqueue.h b/externals/tracy/public/client/tracy_concurrentqueue.h new file mode 100644 index 000000000..4178d39ea --- /dev/null +++ b/externals/tracy/public/client/tracy_concurrentqueue.h @@ -0,0 +1,1441 @@ +// Provides a C++11 implementation of a multi-producer, multi-consumer lock-free queue. +// An overview, including benchmark results, is provided here: +// http://moodycamel.com/blog/2014/a-fast-general-purpose-lock-free-queue-for-c++ +// The full design is also described in excruciating detail at: +// http://moodycamel.com/blog/2014/detailed-design-of-a-lock-free-queue + +// Simplified BSD license: +// Copyright (c) 2013-2016, Cameron Desrochers. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without modification, +// are permitted provided that the following conditions are met: +// +// - Redistributions of source code must retain the above copyright notice, this list of +// conditions and the following disclaimer. +// - Redistributions in binary form must reproduce the above copyright notice, this list of +// conditions and the following disclaimer in the documentation and/or other materials +// provided with the distribution. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY +// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF +// MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL +// THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT +// OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +// HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR +// TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, +// EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + +#pragma once + +#include "../common/TracyAlloc.hpp" +#include "../common/TracyForceInline.hpp" +#include "../common/TracySystem.hpp" + +#if defined(__GNUC__) +// Disable -Wconversion warnings (spuriously triggered when Traits::size_t and +// Traits::index_t are set to < 32 bits, causing integer promotion, causing warnings +// upon assigning any computed values) +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wconversion" +#endif + +#if defined(__APPLE__) +#include "TargetConditionals.h" +#endif + +#include // Requires C++11. Sorry VS2010. +#include +#include // for max_align_t +#include +#include +#include +#include +#include +#include +#include // for CHAR_BIT +#include +#include // partly for __WINPTHREADS_VERSION if on MinGW-w64 w/ POSIX threading + +namespace tracy +{ + +// Compiler-specific likely/unlikely hints +namespace moodycamel { namespace details { +#if defined(__GNUC__) + inline bool cqLikely(bool x) { return __builtin_expect((x), true); } + inline bool cqUnlikely(bool x) { return __builtin_expect((x), false); } +#else + inline bool cqLikely(bool x) { return x; } + inline bool cqUnlikely(bool x) { return x; } +#endif +} } + +namespace +{ + // to avoid MSVC warning 4127: conditional expression is constant + template + struct compile_time_condition + { + static const bool value = false; + }; + template <> + struct compile_time_condition + { + static const bool value = true; + }; +} + +namespace moodycamel { +namespace details { + template + struct const_numeric_max { + static_assert(std::is_integral::value, "const_numeric_max can only be used with integers"); + static const T value = std::numeric_limits::is_signed + ? (static_cast(1) << (sizeof(T) * CHAR_BIT - 1)) - static_cast(1) + : static_cast(-1); + }; + +#if defined(__GLIBCXX__) + typedef ::max_align_t std_max_align_t; // libstdc++ forgot to add it to std:: for a while +#else + typedef std::max_align_t std_max_align_t; // Others (e.g. MSVC) insist it can *only* be accessed via std:: +#endif + + // Some platforms have incorrectly set max_align_t to a type with <8 bytes alignment even while supporting + // 8-byte aligned scalar values (*cough* 32-bit iOS). Work around this with our own union. See issue #64. + typedef union { + std_max_align_t x; + long long y; + void* z; + } max_align_t; +} + +// Default traits for the ConcurrentQueue. To change some of the +// traits without re-implementing all of them, inherit from this +// struct and shadow the declarations you wish to be different; +// since the traits are used as a template type parameter, the +// shadowed declarations will be used where defined, and the defaults +// otherwise. +struct ConcurrentQueueDefaultTraits +{ + // General-purpose size type. std::size_t is strongly recommended. + typedef std::size_t size_t; + + // The type used for the enqueue and dequeue indices. Must be at least as + // large as size_t. Should be significantly larger than the number of elements + // you expect to hold at once, especially if you have a high turnover rate; + // for example, on 32-bit x86, if you expect to have over a hundred million + // elements or pump several million elements through your queue in a very + // short space of time, using a 32-bit type *may* trigger a race condition. + // A 64-bit int type is recommended in that case, and in practice will + // prevent a race condition no matter the usage of the queue. Note that + // whether the queue is lock-free with a 64-int type depends on the whether + // std::atomic is lock-free, which is platform-specific. + typedef std::size_t index_t; + + // Internally, all elements are enqueued and dequeued from multi-element + // blocks; this is the smallest controllable unit. If you expect few elements + // but many producers, a smaller block size should be favoured. For few producers + // and/or many elements, a larger block size is preferred. A sane default + // is provided. Must be a power of 2. + static const size_t BLOCK_SIZE = 64*1024; + + // For explicit producers (i.e. when using a producer token), the block is + // checked for being empty by iterating through a list of flags, one per element. + // For large block sizes, this is too inefficient, and switching to an atomic + // counter-based approach is faster. The switch is made for block sizes strictly + // larger than this threshold. + static const size_t EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD = 32; + + // How many full blocks can be expected for a single explicit producer? This should + // reflect that number's maximum for optimal performance. Must be a power of 2. + static const size_t EXPLICIT_INITIAL_INDEX_SIZE = 32; + + // Controls the number of items that an explicit consumer (i.e. one with a token) + // must consume before it causes all consumers to rotate and move on to the next + // internal queue. + static const std::uint32_t EXPLICIT_CONSUMER_CONSUMPTION_QUOTA_BEFORE_ROTATE = 256; + + // The maximum number of elements (inclusive) that can be enqueued to a sub-queue. + // Enqueue operations that would cause this limit to be surpassed will fail. Note + // that this limit is enforced at the block level (for performance reasons), i.e. + // it's rounded up to the nearest block size. + static const size_t MAX_SUBQUEUE_SIZE = details::const_numeric_max::value; + + + // Memory allocation can be customized if needed. + // malloc should return nullptr on failure, and handle alignment like std::malloc. +#if defined(malloc) || defined(free) + // Gah, this is 2015, stop defining macros that break standard code already! + // Work around malloc/free being special macros: + static inline void* WORKAROUND_malloc(size_t size) { return malloc(size); } + static inline void WORKAROUND_free(void* ptr) { return free(ptr); } + static inline void* (malloc)(size_t size) { return WORKAROUND_malloc(size); } + static inline void (free)(void* ptr) { return WORKAROUND_free(ptr); } +#else + static inline void* malloc(size_t size) { return tracy::tracy_malloc(size); } + static inline void free(void* ptr) { return tracy::tracy_free(ptr); } +#endif +}; + + +// When producing or consuming many elements, the most efficient way is to: +// 1) Use one of the bulk-operation methods of the queue with a token +// 2) Failing that, use the bulk-operation methods without a token +// 3) Failing that, create a token and use that with the single-item methods +// 4) Failing that, use the single-parameter methods of the queue +// Having said that, don't create tokens willy-nilly -- ideally there should be +// a maximum of one token per thread (of each kind). +struct ProducerToken; +struct ConsumerToken; + +template class ConcurrentQueue; + + +namespace details +{ + struct ConcurrentQueueProducerTypelessBase + { + ConcurrentQueueProducerTypelessBase* next; + std::atomic inactive; + ProducerToken* token; + uint32_t threadId; + + ConcurrentQueueProducerTypelessBase() + : next(nullptr), inactive(false), token(nullptr), threadId(0) + { + } + }; + + template + static inline bool circular_less_than(T a, T b) + { + static_assert(std::is_integral::value && !std::numeric_limits::is_signed, "circular_less_than is intended to be used only with unsigned integer types"); + return static_cast(a - b) > static_cast(static_cast(1) << (static_cast(sizeof(T) * CHAR_BIT - 1))); + // Note: extra parens around rhs of operator<< is MSVC bug: https://developercommunity2.visualstudio.com/t/C4554-triggers-when-both-lhs-and-rhs-is/10034931 + // silencing the bug requires #pragma warning(disable: 4554) around the calling code and has no effect when done here. + } + + template + static inline char* align_for(char* ptr) + { + const std::size_t alignment = std::alignment_of::value; + return ptr + (alignment - (reinterpret_cast(ptr) % alignment)) % alignment; + } + + template + static inline T ceil_to_pow_2(T x) + { + static_assert(std::is_integral::value && !std::numeric_limits::is_signed, "ceil_to_pow_2 is intended to be used only with unsigned integer types"); + + // Adapted from http://graphics.stanford.edu/~seander/bithacks.html#RoundUpPowerOf2 + --x; + x |= x >> 1; + x |= x >> 2; + x |= x >> 4; + for (std::size_t i = 1; i < sizeof(T); i <<= 1) { + x |= x >> (i << 3); + } + ++x; + return x; + } + + template + static inline void swap_relaxed(std::atomic& left, std::atomic& right) + { + T temp = std::move(left.load(std::memory_order_relaxed)); + left.store(std::move(right.load(std::memory_order_relaxed)), std::memory_order_relaxed); + right.store(std::move(temp), std::memory_order_relaxed); + } + + template + static inline T const& nomove(T const& x) + { + return x; + } + + template + struct nomove_if + { + template + static inline T const& eval(T const& x) + { + return x; + } + }; + + template<> + struct nomove_if + { + template + static inline auto eval(U&& x) + -> decltype(std::forward(x)) + { + return std::forward(x); + } + }; + + template + static inline auto deref_noexcept(It& it) noexcept -> decltype(*it) + { + return *it; + } + +#if defined(__clang__) || !defined(__GNUC__) || __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8) + template struct is_trivially_destructible : std::is_trivially_destructible { }; +#else + template struct is_trivially_destructible : std::has_trivial_destructor { }; +#endif + + template struct static_is_lock_free_num { enum { value = 0 }; }; + template<> struct static_is_lock_free_num { enum { value = ATOMIC_CHAR_LOCK_FREE }; }; + template<> struct static_is_lock_free_num { enum { value = ATOMIC_SHORT_LOCK_FREE }; }; + template<> struct static_is_lock_free_num { enum { value = ATOMIC_INT_LOCK_FREE }; }; + template<> struct static_is_lock_free_num { enum { value = ATOMIC_LONG_LOCK_FREE }; }; + template<> struct static_is_lock_free_num { enum { value = ATOMIC_LLONG_LOCK_FREE }; }; + template struct static_is_lock_free : static_is_lock_free_num::type> { }; + template<> struct static_is_lock_free { enum { value = ATOMIC_BOOL_LOCK_FREE }; }; + template struct static_is_lock_free { enum { value = ATOMIC_POINTER_LOCK_FREE }; }; +} + + +struct ProducerToken +{ + template + explicit ProducerToken(ConcurrentQueue& queue); + + ProducerToken(ProducerToken&& other) noexcept + : producer(other.producer) + { + other.producer = nullptr; + if (producer != nullptr) { + producer->token = this; + } + } + + inline ProducerToken& operator=(ProducerToken&& other) noexcept + { + swap(other); + return *this; + } + + void swap(ProducerToken& other) noexcept + { + std::swap(producer, other.producer); + if (producer != nullptr) { + producer->token = this; + } + if (other.producer != nullptr) { + other.producer->token = &other; + } + } + + // A token is always valid unless: + // 1) Memory allocation failed during construction + // 2) It was moved via the move constructor + // (Note: assignment does a swap, leaving both potentially valid) + // 3) The associated queue was destroyed + // Note that if valid() returns true, that only indicates + // that the token is valid for use with a specific queue, + // but not which one; that's up to the user to track. + inline bool valid() const { return producer != nullptr; } + + ~ProducerToken() + { + if (producer != nullptr) { + producer->token = nullptr; + producer->inactive.store(true, std::memory_order_release); + } + } + + // Disable copying and assignment + ProducerToken(ProducerToken const&) = delete; + ProducerToken& operator=(ProducerToken const&) = delete; + +private: + template friend class ConcurrentQueue; + +protected: + details::ConcurrentQueueProducerTypelessBase* producer; +}; + + +struct ConsumerToken +{ + template + explicit ConsumerToken(ConcurrentQueue& q); + + ConsumerToken(ConsumerToken&& other) noexcept + : initialOffset(other.initialOffset), lastKnownGlobalOffset(other.lastKnownGlobalOffset), itemsConsumedFromCurrent(other.itemsConsumedFromCurrent), currentProducer(other.currentProducer), desiredProducer(other.desiredProducer) + { + } + + inline ConsumerToken& operator=(ConsumerToken&& other) noexcept + { + swap(other); + return *this; + } + + void swap(ConsumerToken& other) noexcept + { + std::swap(initialOffset, other.initialOffset); + std::swap(lastKnownGlobalOffset, other.lastKnownGlobalOffset); + std::swap(itemsConsumedFromCurrent, other.itemsConsumedFromCurrent); + std::swap(currentProducer, other.currentProducer); + std::swap(desiredProducer, other.desiredProducer); + } + + // Disable copying and assignment + ConsumerToken(ConsumerToken const&) = delete; + ConsumerToken& operator=(ConsumerToken const&) = delete; + +private: + template friend class ConcurrentQueue; + +private: // but shared with ConcurrentQueue + std::uint32_t initialOffset; + std::uint32_t lastKnownGlobalOffset; + std::uint32_t itemsConsumedFromCurrent; + details::ConcurrentQueueProducerTypelessBase* currentProducer; + details::ConcurrentQueueProducerTypelessBase* desiredProducer; +}; + + +template +class ConcurrentQueue +{ +public: + struct ExplicitProducer; + + typedef moodycamel::ProducerToken producer_token_t; + typedef moodycamel::ConsumerToken consumer_token_t; + + typedef typename Traits::index_t index_t; + typedef typename Traits::size_t size_t; + + static const size_t BLOCK_SIZE = static_cast(Traits::BLOCK_SIZE); + static const size_t EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD = static_cast(Traits::EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD); + static const size_t EXPLICIT_INITIAL_INDEX_SIZE = static_cast(Traits::EXPLICIT_INITIAL_INDEX_SIZE); + static const std::uint32_t EXPLICIT_CONSUMER_CONSUMPTION_QUOTA_BEFORE_ROTATE = static_cast(Traits::EXPLICIT_CONSUMER_CONSUMPTION_QUOTA_BEFORE_ROTATE); +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable: 4307) // + integral constant overflow (that's what the ternary expression is for!) +#pragma warning(disable: 4309) // static_cast: Truncation of constant value +#endif + static const size_t MAX_SUBQUEUE_SIZE = (details::const_numeric_max::value - static_cast(Traits::MAX_SUBQUEUE_SIZE) < BLOCK_SIZE) ? details::const_numeric_max::value : ((static_cast(Traits::MAX_SUBQUEUE_SIZE) + (BLOCK_SIZE - 1)) / BLOCK_SIZE * BLOCK_SIZE); +#ifdef _MSC_VER +#pragma warning(pop) +#endif + + static_assert(!std::numeric_limits::is_signed && std::is_integral::value, "Traits::size_t must be an unsigned integral type"); + static_assert(!std::numeric_limits::is_signed && std::is_integral::value, "Traits::index_t must be an unsigned integral type"); + static_assert(sizeof(index_t) >= sizeof(size_t), "Traits::index_t must be at least as wide as Traits::size_t"); + static_assert((BLOCK_SIZE > 1) && !(BLOCK_SIZE & (BLOCK_SIZE - 1)), "Traits::BLOCK_SIZE must be a power of 2 (and at least 2)"); + static_assert((EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD > 1) && !(EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD & (EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD - 1)), "Traits::EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD must be a power of 2 (and greater than 1)"); + static_assert((EXPLICIT_INITIAL_INDEX_SIZE > 1) && !(EXPLICIT_INITIAL_INDEX_SIZE & (EXPLICIT_INITIAL_INDEX_SIZE - 1)), "Traits::EXPLICIT_INITIAL_INDEX_SIZE must be a power of 2 (and greater than 1)"); + +public: + // Creates a queue with at least `capacity` element slots; note that the + // actual number of elements that can be inserted without additional memory + // allocation depends on the number of producers and the block size (e.g. if + // the block size is equal to `capacity`, only a single block will be allocated + // up-front, which means only a single producer will be able to enqueue elements + // without an extra allocation -- blocks aren't shared between producers). + // This method is not thread safe -- it is up to the user to ensure that the + // queue is fully constructed before it starts being used by other threads (this + // includes making the memory effects of construction visible, possibly with a + // memory barrier). + explicit ConcurrentQueue(size_t capacity = 6 * BLOCK_SIZE) + : producerListTail(nullptr), + producerCount(0), + initialBlockPoolIndex(0), + nextExplicitConsumerId(0), + globalExplicitConsumerOffset(0) + { + populate_initial_block_list(capacity / BLOCK_SIZE + ((capacity & (BLOCK_SIZE - 1)) == 0 ? 0 : 1)); + } + + // Computes the correct amount of pre-allocated blocks for you based + // on the minimum number of elements you want available at any given + // time, and the maximum concurrent number of each type of producer. + ConcurrentQueue(size_t minCapacity, size_t maxExplicitProducers) + : producerListTail(nullptr), + producerCount(0), + initialBlockPoolIndex(0), + nextExplicitConsumerId(0), + globalExplicitConsumerOffset(0) + { + size_t blocks = (((minCapacity + BLOCK_SIZE - 1) / BLOCK_SIZE) - 1) * (maxExplicitProducers + 1) + 2 * (maxExplicitProducers); + populate_initial_block_list(blocks); + } + + // Note: The queue should not be accessed concurrently while it's + // being deleted. It's up to the user to synchronize this. + // This method is not thread safe. + ~ConcurrentQueue() + { + // Destroy producers + auto ptr = producerListTail.load(std::memory_order_relaxed); + while (ptr != nullptr) { + auto next = ptr->next_prod(); + if (ptr->token != nullptr) { + ptr->token->producer = nullptr; + } + destroy(ptr); + ptr = next; + } + + // Destroy global free list + auto block = freeList.head_unsafe(); + while (block != nullptr) { + auto next = block->freeListNext.load(std::memory_order_relaxed); + if (block->dynamicallyAllocated) { + destroy(block); + } + block = next; + } + + // Destroy initial free list + destroy_array(initialBlockPool, initialBlockPoolSize); + } + + // Disable copying and copy assignment + ConcurrentQueue(ConcurrentQueue const&) = delete; + ConcurrentQueue(ConcurrentQueue&& other) = delete; + ConcurrentQueue& operator=(ConcurrentQueue const&) = delete; + ConcurrentQueue& operator=(ConcurrentQueue&& other) = delete; + +public: + tracy_force_inline T* enqueue_begin(producer_token_t const& token, index_t& currentTailIndex) + { + return static_cast(token.producer)->ConcurrentQueue::ExplicitProducer::enqueue_begin(currentTailIndex); + } + + template + size_t try_dequeue_bulk_single(consumer_token_t& token, NotifyThread notifyThread, ProcessData processData ) + { + if (token.desiredProducer == nullptr || token.lastKnownGlobalOffset != globalExplicitConsumerOffset.load(std::memory_order_relaxed)) { + if (!update_current_producer_after_rotation(token)) { + return 0; + } + } + + size_t count = static_cast(token.currentProducer)->dequeue_bulk(notifyThread, processData); + token.itemsConsumedFromCurrent += static_cast(count); + + auto tail = producerListTail.load(std::memory_order_acquire); + auto ptr = static_cast(token.currentProducer)->next_prod(); + if (ptr == nullptr) { + ptr = tail; + } + if( count == 0 ) + { + while (ptr != static_cast(token.currentProducer)) { + auto dequeued = ptr->dequeue_bulk(notifyThread, processData); + if (dequeued != 0) { + token.currentProducer = ptr; + token.itemsConsumedFromCurrent = static_cast(dequeued); + return dequeued; + } + ptr = ptr->next_prod(); + if (ptr == nullptr) { + ptr = tail; + } + } + return 0; + } + else + { + token.currentProducer = ptr; + token.itemsConsumedFromCurrent = 0; + return count; + } + } + + + // Returns an estimate of the total number of elements currently in the queue. This + // estimate is only accurate if the queue has completely stabilized before it is called + // (i.e. all enqueue and dequeue operations have completed and their memory effects are + // visible on the calling thread, and no further operations start while this method is + // being called). + // Thread-safe. + size_t size_approx() const + { + size_t size = 0; + for (auto ptr = producerListTail.load(std::memory_order_acquire); ptr != nullptr; ptr = ptr->next_prod()) { + size += ptr->size_approx(); + } + return size; + } + + + // Returns true if the underlying atomic variables used by + // the queue are lock-free (they should be on most platforms). + // Thread-safe. + static bool is_lock_free() + { + return + details::static_is_lock_free::value == 2 && + details::static_is_lock_free::value == 2 && + details::static_is_lock_free::value == 2 && + details::static_is_lock_free::value == 2 && + details::static_is_lock_free::value == 2; + } + + +private: + friend struct ProducerToken; + friend struct ConsumerToken; + friend struct ExplicitProducer; + + + /////////////////////////////// + // Queue methods + /////////////////////////////// + + inline bool update_current_producer_after_rotation(consumer_token_t& token) + { + // Ah, there's been a rotation, figure out where we should be! + auto tail = producerListTail.load(std::memory_order_acquire); + if (token.desiredProducer == nullptr && tail == nullptr) { + return false; + } + auto prodCount = producerCount.load(std::memory_order_relaxed); + auto globalOffset = globalExplicitConsumerOffset.load(std::memory_order_relaxed); + if (details::cqUnlikely(token.desiredProducer == nullptr)) { + // Aha, first time we're dequeueing anything. + // Figure out our local position + // Note: offset is from start, not end, but we're traversing from end -- subtract from count first + std::uint32_t offset = prodCount - 1 - (token.initialOffset % prodCount); + token.desiredProducer = tail; + for (std::uint32_t i = 0; i != offset; ++i) { + token.desiredProducer = static_cast(token.desiredProducer)->next_prod(); + if (token.desiredProducer == nullptr) { + token.desiredProducer = tail; + } + } + } + + std::uint32_t delta = globalOffset - token.lastKnownGlobalOffset; + if (delta >= prodCount) { + delta = delta % prodCount; + } + for (std::uint32_t i = 0; i != delta; ++i) { + token.desiredProducer = static_cast(token.desiredProducer)->next_prod(); + if (token.desiredProducer == nullptr) { + token.desiredProducer = tail; + } + } + + token.lastKnownGlobalOffset = globalOffset; + token.currentProducer = token.desiredProducer; + token.itemsConsumedFromCurrent = 0; + return true; + } + + + /////////////////////////// + // Free list + /////////////////////////// + + template + struct FreeListNode + { + FreeListNode() : freeListRefs(0), freeListNext(nullptr) { } + + std::atomic freeListRefs; + std::atomic freeListNext; + }; + + // A simple CAS-based lock-free free list. Not the fastest thing in the world under heavy contention, but + // simple and correct (assuming nodes are never freed until after the free list is destroyed), and fairly + // speedy under low contention. + template // N must inherit FreeListNode or have the same fields (and initialization of them) + struct FreeList + { + FreeList() : freeListHead(nullptr) { } + FreeList(FreeList&& other) : freeListHead(other.freeListHead.load(std::memory_order_relaxed)) { other.freeListHead.store(nullptr, std::memory_order_relaxed); } + void swap(FreeList& other) { details::swap_relaxed(freeListHead, other.freeListHead); } + + FreeList(FreeList const&) = delete; + FreeList& operator=(FreeList const&) = delete; + + inline void add(N* node) + { + // We know that the should-be-on-freelist bit is 0 at this point, so it's safe to + // set it using a fetch_add + if (node->freeListRefs.fetch_add(SHOULD_BE_ON_FREELIST, std::memory_order_acq_rel) == 0) { + // Oh look! We were the last ones referencing this node, and we know + // we want to add it to the free list, so let's do it! + add_knowing_refcount_is_zero(node); + } + } + + inline N* try_get() + { + auto head = freeListHead.load(std::memory_order_acquire); + while (head != nullptr) { + auto prevHead = head; + auto refs = head->freeListRefs.load(std::memory_order_relaxed); + if ((refs & REFS_MASK) == 0 || !head->freeListRefs.compare_exchange_strong(refs, refs + 1, std::memory_order_acquire, std::memory_order_relaxed)) { + head = freeListHead.load(std::memory_order_acquire); + continue; + } + + // Good, reference count has been incremented (it wasn't at zero), which means we can read the + // next and not worry about it changing between now and the time we do the CAS + auto next = head->freeListNext.load(std::memory_order_relaxed); + if (freeListHead.compare_exchange_strong(head, next, std::memory_order_acquire, std::memory_order_relaxed)) { + // Yay, got the node. This means it was on the list, which means shouldBeOnFreeList must be false no + // matter the refcount (because nobody else knows it's been taken off yet, it can't have been put back on). + assert((head->freeListRefs.load(std::memory_order_relaxed) & SHOULD_BE_ON_FREELIST) == 0); + + // Decrease refcount twice, once for our ref, and once for the list's ref + head->freeListRefs.fetch_sub(2, std::memory_order_release); + return head; + } + + // OK, the head must have changed on us, but we still need to decrease the refcount we increased. + // Note that we don't need to release any memory effects, but we do need to ensure that the reference + // count decrement happens-after the CAS on the head. + refs = prevHead->freeListRefs.fetch_sub(1, std::memory_order_acq_rel); + if (refs == SHOULD_BE_ON_FREELIST + 1) { + add_knowing_refcount_is_zero(prevHead); + } + } + + return nullptr; + } + + // Useful for traversing the list when there's no contention (e.g. to destroy remaining nodes) + N* head_unsafe() const { return freeListHead.load(std::memory_order_relaxed); } + + private: + inline void add_knowing_refcount_is_zero(N* node) + { + // Since the refcount is zero, and nobody can increase it once it's zero (except us, and we run + // only one copy of this method per node at a time, i.e. the single thread case), then we know + // we can safely change the next pointer of the node; however, once the refcount is back above + // zero, then other threads could increase it (happens under heavy contention, when the refcount + // goes to zero in between a load and a refcount increment of a node in try_get, then back up to + // something non-zero, then the refcount increment is done by the other thread) -- so, if the CAS + // to add the node to the actual list fails, decrease the refcount and leave the add operation to + // the next thread who puts the refcount back at zero (which could be us, hence the loop). + auto head = freeListHead.load(std::memory_order_relaxed); + while (true) { + node->freeListNext.store(head, std::memory_order_relaxed); + node->freeListRefs.store(1, std::memory_order_release); + if (!freeListHead.compare_exchange_strong(head, node, std::memory_order_release, std::memory_order_relaxed)) { + // Hmm, the add failed, but we can only try again when the refcount goes back to zero + if (node->freeListRefs.fetch_add(SHOULD_BE_ON_FREELIST - 1, std::memory_order_release) == 1) { + continue; + } + } + return; + } + } + + private: + // Implemented like a stack, but where node order doesn't matter (nodes are inserted out of order under contention) + std::atomic freeListHead; + + static const std::uint32_t REFS_MASK = 0x7FFFFFFF; + static const std::uint32_t SHOULD_BE_ON_FREELIST = 0x80000000; + }; + + + /////////////////////////// + // Block + /////////////////////////// + + struct Block + { + Block() + : next(nullptr), elementsCompletelyDequeued(0), freeListRefs(0), freeListNext(nullptr), shouldBeOnFreeList(false), dynamicallyAllocated(true) + { + } + + inline bool is_empty() const + { + if (compile_time_condition::value) { + // Check flags + for (size_t i = 0; i < BLOCK_SIZE; ++i) { + if (!emptyFlags[i].load(std::memory_order_relaxed)) { + return false; + } + } + + // Aha, empty; make sure we have all other memory effects that happened before the empty flags were set + std::atomic_thread_fence(std::memory_order_acquire); + return true; + } + else { + // Check counter + if (elementsCompletelyDequeued.load(std::memory_order_relaxed) == BLOCK_SIZE) { + std::atomic_thread_fence(std::memory_order_acquire); + return true; + } + assert(elementsCompletelyDequeued.load(std::memory_order_relaxed) <= BLOCK_SIZE); + return false; + } + } + + // Returns true if the block is now empty (does not apply in explicit context) + inline bool set_empty(index_t i) + { + if (BLOCK_SIZE <= EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD) { + // Set flag + assert(!emptyFlags[BLOCK_SIZE - 1 - static_cast(i & static_cast(BLOCK_SIZE - 1))].load(std::memory_order_relaxed)); + emptyFlags[BLOCK_SIZE - 1 - static_cast(i & static_cast(BLOCK_SIZE - 1))].store(true, std::memory_order_release); + return false; + } + else { + // Increment counter + auto prevVal = elementsCompletelyDequeued.fetch_add(1, std::memory_order_release); + assert(prevVal < BLOCK_SIZE); + return prevVal == BLOCK_SIZE - 1; + } + } + + // Sets multiple contiguous item statuses to 'empty' (assumes no wrapping and count > 0). + // Returns true if the block is now empty (does not apply in explicit context). + inline bool set_many_empty(index_t i, size_t count) + { + if (compile_time_condition::value) { + // Set flags + std::atomic_thread_fence(std::memory_order_release); + i = BLOCK_SIZE - 1 - static_cast(i & static_cast(BLOCK_SIZE - 1)) - count + 1; + for (size_t j = 0; j != count; ++j) { + assert(!emptyFlags[i + j].load(std::memory_order_relaxed)); + emptyFlags[i + j].store(true, std::memory_order_relaxed); + } + return false; + } + else { + // Increment counter + auto prevVal = elementsCompletelyDequeued.fetch_add(count, std::memory_order_release); + assert(prevVal + count <= BLOCK_SIZE); + return prevVal + count == BLOCK_SIZE; + } + } + + inline void set_all_empty() + { + if (BLOCK_SIZE <= EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD) { + // Set all flags + for (size_t i = 0; i != BLOCK_SIZE; ++i) { + emptyFlags[i].store(true, std::memory_order_relaxed); + } + } + else { + // Reset counter + elementsCompletelyDequeued.store(BLOCK_SIZE, std::memory_order_relaxed); + } + } + + inline void reset_empty() + { + if (compile_time_condition::value) { + // Reset flags + for (size_t i = 0; i != BLOCK_SIZE; ++i) { + emptyFlags[i].store(false, std::memory_order_relaxed); + } + } + else { + // Reset counter + elementsCompletelyDequeued.store(0, std::memory_order_relaxed); + } + } + + inline T* operator[](index_t idx) noexcept { return static_cast(static_cast(elements)) + static_cast(idx & static_cast(BLOCK_SIZE - 1)); } + inline T const* operator[](index_t idx) const noexcept { return static_cast(static_cast(elements)) + static_cast(idx & static_cast(BLOCK_SIZE - 1)); } + + private: + // IMPORTANT: This must be the first member in Block, so that if T depends on the alignment of + // addresses returned by malloc, that alignment will be preserved. Apparently clang actually + // generates code that uses this assumption for AVX instructions in some cases. Ideally, we + // should also align Block to the alignment of T in case it's higher than malloc's 16-byte + // alignment, but this is hard to do in a cross-platform way. Assert for this case: + static_assert(std::alignment_of::value <= std::alignment_of::value, "The queue does not support super-aligned types at this time"); + // Additionally, we need the alignment of Block itself to be a multiple of max_align_t since + // otherwise the appropriate padding will not be added at the end of Block in order to make + // arrays of Blocks all be properly aligned (not just the first one). We use a union to force + // this. + union { + char elements[sizeof(T) * BLOCK_SIZE]; + details::max_align_t dummy; + }; + public: + Block* next; + std::atomic elementsCompletelyDequeued; + std::atomic emptyFlags[BLOCK_SIZE <= EXPLICIT_BLOCK_EMPTY_COUNTER_THRESHOLD ? BLOCK_SIZE : 1]; + public: + std::atomic freeListRefs; + std::atomic freeListNext; + std::atomic shouldBeOnFreeList; + bool dynamicallyAllocated; // Perhaps a better name for this would be 'isNotPartOfInitialBlockPool' + }; + static_assert(std::alignment_of::value >= std::alignment_of::value, "Internal error: Blocks must be at least as aligned as the type they are wrapping"); + + + /////////////////////////// + // Producer base + /////////////////////////// + + struct ProducerBase : public details::ConcurrentQueueProducerTypelessBase + { + ProducerBase(ConcurrentQueue* parent_) : + tailIndex(0), + headIndex(0), + dequeueOptimisticCount(0), + dequeueOvercommit(0), + tailBlock(nullptr), + parent(parent_) + { + } + + virtual ~ProducerBase() { }; + + template + inline size_t dequeue_bulk(NotifyThread notifyThread, ProcessData processData) + { + return static_cast(this)->dequeue_bulk(notifyThread, processData); + } + + inline ProducerBase* next_prod() const { return static_cast(next); } + + inline size_t size_approx() const + { + auto tail = tailIndex.load(std::memory_order_relaxed); + auto head = headIndex.load(std::memory_order_relaxed); + return details::circular_less_than(head, tail) ? static_cast(tail - head) : 0; + } + + inline index_t getTail() const { return tailIndex.load(std::memory_order_relaxed); } + protected: + std::atomic tailIndex; // Where to enqueue to next + std::atomic headIndex; // Where to dequeue from next + + std::atomic dequeueOptimisticCount; + std::atomic dequeueOvercommit; + + Block* tailBlock; + + public: + ConcurrentQueue* parent; + }; + + + public: + /////////////////////////// + // Explicit queue + /////////////////////////// + struct ExplicitProducer : public ProducerBase + { + explicit ExplicitProducer(ConcurrentQueue* _parent) : + ProducerBase(_parent), + blockIndex(nullptr), + pr_blockIndexSlotsUsed(0), + pr_blockIndexSize(EXPLICIT_INITIAL_INDEX_SIZE >> 1), + pr_blockIndexFront(0), + pr_blockIndexEntries(nullptr), + pr_blockIndexRaw(nullptr) + { + size_t poolBasedIndexSize = details::ceil_to_pow_2(_parent->initialBlockPoolSize) >> 1; + if (poolBasedIndexSize > pr_blockIndexSize) { + pr_blockIndexSize = poolBasedIndexSize; + } + + new_block_index(0); // This creates an index with double the number of current entries, i.e. EXPLICIT_INITIAL_INDEX_SIZE + } + + ~ExplicitProducer() + { + // Destruct any elements not yet dequeued. + // Since we're in the destructor, we can assume all elements + // are either completely dequeued or completely not (no halfways). + if (this->tailBlock != nullptr) { // Note this means there must be a block index too + // First find the block that's partially dequeued, if any + Block* halfDequeuedBlock = nullptr; + if ((this->headIndex.load(std::memory_order_relaxed) & static_cast(BLOCK_SIZE - 1)) != 0) { + // The head's not on a block boundary, meaning a block somewhere is partially dequeued + // (or the head block is the tail block and was fully dequeued, but the head/tail are still not on a boundary) + size_t i = (pr_blockIndexFront - pr_blockIndexSlotsUsed) & (pr_blockIndexSize - 1); + while (details::circular_less_than(pr_blockIndexEntries[i].base + BLOCK_SIZE, this->headIndex.load(std::memory_order_relaxed))) { + i = (i + 1) & (pr_blockIndexSize - 1); + } + assert(details::circular_less_than(pr_blockIndexEntries[i].base, this->headIndex.load(std::memory_order_relaxed))); + halfDequeuedBlock = pr_blockIndexEntries[i].block; + } + + // Start at the head block (note the first line in the loop gives us the head from the tail on the first iteration) + auto block = this->tailBlock; + do { + block = block->next; + if (block->ConcurrentQueue::Block::is_empty()) { + continue; + } + + size_t i = 0; // Offset into block + if (block == halfDequeuedBlock) { + i = static_cast(this->headIndex.load(std::memory_order_relaxed) & static_cast(BLOCK_SIZE - 1)); + } + + // Walk through all the items in the block; if this is the tail block, we need to stop when we reach the tail index + auto lastValidIndex = (this->tailIndex.load(std::memory_order_relaxed) & static_cast(BLOCK_SIZE - 1)) == 0 ? BLOCK_SIZE : static_cast(this->tailIndex.load(std::memory_order_relaxed) & static_cast(BLOCK_SIZE - 1)); + while (i != BLOCK_SIZE && (block != this->tailBlock || i != lastValidIndex)) { + (*block)[i++]->~T(); + } + } while (block != this->tailBlock); + } + + // Destroy all blocks that we own + if (this->tailBlock != nullptr) { + auto block = this->tailBlock; + do { + auto nextBlock = block->next; + if (block->dynamicallyAllocated) { + destroy(block); + } + else { + this->parent->add_block_to_free_list(block); + } + block = nextBlock; + } while (block != this->tailBlock); + } + + // Destroy the block indices + auto header = static_cast(pr_blockIndexRaw); + while (header != nullptr) { + auto prev = static_cast(header->prev); + header->~BlockIndexHeader(); + (Traits::free)(header); + header = prev; + } + } + + inline void enqueue_begin_alloc(index_t currentTailIndex) + { + // We reached the end of a block, start a new one + if (this->tailBlock != nullptr && this->tailBlock->next->ConcurrentQueue::Block::is_empty()) { + // We can re-use the block ahead of us, it's empty! + this->tailBlock = this->tailBlock->next; + this->tailBlock->ConcurrentQueue::Block::reset_empty(); + + // We'll put the block on the block index (guaranteed to be room since we're conceptually removing the + // last block from it first -- except instead of removing then adding, we can just overwrite). + // Note that there must be a valid block index here, since even if allocation failed in the ctor, + // it would have been re-attempted when adding the first block to the queue; since there is such + // a block, a block index must have been successfully allocated. + } + else { + // We're going to need a new block; check that the block index has room + if (pr_blockIndexRaw == nullptr || pr_blockIndexSlotsUsed == pr_blockIndexSize) { + // Hmm, the circular block index is already full -- we'll need + // to allocate a new index. Note pr_blockIndexRaw can only be nullptr if + // the initial allocation failed in the constructor. + new_block_index(pr_blockIndexSlotsUsed); + } + + // Insert a new block in the circular linked list + auto newBlock = this->parent->ConcurrentQueue::requisition_block(); + newBlock->ConcurrentQueue::Block::reset_empty(); + if (this->tailBlock == nullptr) { + newBlock->next = newBlock; + } + else { + newBlock->next = this->tailBlock->next; + this->tailBlock->next = newBlock; + } + this->tailBlock = newBlock; + ++pr_blockIndexSlotsUsed; + } + + // Add block to block index + auto& entry = blockIndex.load(std::memory_order_relaxed)->entries[pr_blockIndexFront]; + entry.base = currentTailIndex; + entry.block = this->tailBlock; + blockIndex.load(std::memory_order_relaxed)->front.store(pr_blockIndexFront, std::memory_order_release); + pr_blockIndexFront = (pr_blockIndexFront + 1) & (pr_blockIndexSize - 1); + } + + tracy_force_inline T* enqueue_begin(index_t& currentTailIndex) + { + currentTailIndex = this->tailIndex.load(std::memory_order_relaxed); + if (details::cqUnlikely((currentTailIndex & static_cast(BLOCK_SIZE - 1)) == 0)) { + this->enqueue_begin_alloc(currentTailIndex); + } + return (*this->tailBlock)[currentTailIndex]; + } + + tracy_force_inline std::atomic& get_tail_index() + { + return this->tailIndex; + } + + template + size_t dequeue_bulk(NotifyThread notifyThread, ProcessData processData) + { + auto tail = this->tailIndex.load(std::memory_order_relaxed); + auto overcommit = this->dequeueOvercommit.load(std::memory_order_relaxed); + auto desiredCount = static_cast(tail - (this->dequeueOptimisticCount.load(std::memory_order_relaxed) - overcommit)); + if (details::circular_less_than(0, desiredCount)) { + desiredCount = desiredCount < 8192 ? desiredCount : 8192; + std::atomic_thread_fence(std::memory_order_acquire); + + auto myDequeueCount = this->dequeueOptimisticCount.fetch_add(desiredCount, std::memory_order_relaxed); + assert(overcommit <= myDequeueCount); + + tail = this->tailIndex.load(std::memory_order_acquire); + auto actualCount = static_cast(tail - (myDequeueCount - overcommit)); + if (details::circular_less_than(0, actualCount)) { + actualCount = desiredCount < actualCount ? desiredCount : actualCount; + if (actualCount < desiredCount) { + this->dequeueOvercommit.fetch_add(desiredCount - actualCount, std::memory_order_release); + } + + // Get the first index. Note that since there's guaranteed to be at least actualCount elements, this + // will never exceed tail. + auto firstIndex = this->headIndex.fetch_add(actualCount, std::memory_order_acq_rel); + + // Determine which block the first element is in + auto localBlockIndex = blockIndex.load(std::memory_order_acquire); + auto localBlockIndexHead = localBlockIndex->front.load(std::memory_order_acquire); + + auto headBase = localBlockIndex->entries[localBlockIndexHead].base; + auto firstBlockBaseIndex = firstIndex & ~static_cast(BLOCK_SIZE - 1); + auto offset = static_cast(static_cast::type>(firstBlockBaseIndex - headBase) / BLOCK_SIZE); + auto indexIndex = (localBlockIndexHead + offset) & (localBlockIndex->size - 1); + + notifyThread( this->threadId ); + + // Iterate the blocks and dequeue + auto index = firstIndex; + do { + auto firstIndexInBlock = index; + auto endIndex = (index & ~static_cast(BLOCK_SIZE - 1)) + static_cast(BLOCK_SIZE); + endIndex = details::circular_less_than(firstIndex + static_cast(actualCount), endIndex) ? firstIndex + static_cast(actualCount) : endIndex; + auto block = localBlockIndex->entries[indexIndex].block; + + const auto sz = endIndex - index; + processData( (*block)[index], sz ); + index += sz; + + block->ConcurrentQueue::Block::set_many_empty(firstIndexInBlock, static_cast(endIndex - firstIndexInBlock)); + indexIndex = (indexIndex + 1) & (localBlockIndex->size - 1); + } while (index != firstIndex + actualCount); + + return actualCount; + } + else { + // Wasn't anything to dequeue after all; make the effective dequeue count eventually consistent + this->dequeueOvercommit.fetch_add(desiredCount, std::memory_order_release); + } + } + + return 0; + } + + private: + struct BlockIndexEntry + { + index_t base; + Block* block; + }; + + struct BlockIndexHeader + { + size_t size; + std::atomic front; // Current slot (not next, like pr_blockIndexFront) + BlockIndexEntry* entries; + void* prev; + }; + + + bool new_block_index(size_t numberOfFilledSlotsToExpose) + { + auto prevBlockSizeMask = pr_blockIndexSize - 1; + + // Create the new block + pr_blockIndexSize <<= 1; + auto newRawPtr = static_cast((Traits::malloc)(sizeof(BlockIndexHeader) + std::alignment_of::value - 1 + sizeof(BlockIndexEntry) * pr_blockIndexSize)); + if (newRawPtr == nullptr) { + pr_blockIndexSize >>= 1; // Reset to allow graceful retry + return false; + } + + auto newBlockIndexEntries = reinterpret_cast(details::align_for(newRawPtr + sizeof(BlockIndexHeader))); + + // Copy in all the old indices, if any + size_t j = 0; + if (pr_blockIndexSlotsUsed != 0) { + auto i = (pr_blockIndexFront - pr_blockIndexSlotsUsed) & prevBlockSizeMask; + do { + newBlockIndexEntries[j++] = pr_blockIndexEntries[i]; + i = (i + 1) & prevBlockSizeMask; + } while (i != pr_blockIndexFront); + } + + // Update everything + auto header = new (newRawPtr) BlockIndexHeader; + header->size = pr_blockIndexSize; + header->front.store(numberOfFilledSlotsToExpose - 1, std::memory_order_relaxed); + header->entries = newBlockIndexEntries; + header->prev = pr_blockIndexRaw; // we link the new block to the old one so we can free it later + + pr_blockIndexFront = j; + pr_blockIndexEntries = newBlockIndexEntries; + pr_blockIndexRaw = newRawPtr; + blockIndex.store(header, std::memory_order_release); + + return true; + } + + private: + std::atomic blockIndex; + + // To be used by producer only -- consumer must use the ones in referenced by blockIndex + size_t pr_blockIndexSlotsUsed; + size_t pr_blockIndexSize; + size_t pr_blockIndexFront; // Next slot (not current) + BlockIndexEntry* pr_blockIndexEntries; + void* pr_blockIndexRaw; + }; + + ExplicitProducer* get_explicit_producer(producer_token_t const& token) + { + return static_cast(token.producer); + } + + private: + + ////////////////////////////////// + // Block pool manipulation + ////////////////////////////////// + + void populate_initial_block_list(size_t blockCount) + { + initialBlockPoolSize = blockCount; + if (initialBlockPoolSize == 0) { + initialBlockPool = nullptr; + return; + } + + initialBlockPool = create_array(blockCount); + if (initialBlockPool == nullptr) { + initialBlockPoolSize = 0; + } + for (size_t i = 0; i < initialBlockPoolSize; ++i) { + initialBlockPool[i].dynamicallyAllocated = false; + } + } + + inline Block* try_get_block_from_initial_pool() + { + if (initialBlockPoolIndex.load(std::memory_order_relaxed) >= initialBlockPoolSize) { + return nullptr; + } + + auto index = initialBlockPoolIndex.fetch_add(1, std::memory_order_relaxed); + + return index < initialBlockPoolSize ? (initialBlockPool + index) : nullptr; + } + + inline void add_block_to_free_list(Block* block) + { + freeList.add(block); + } + + inline void add_blocks_to_free_list(Block* block) + { + while (block != nullptr) { + auto next = block->next; + add_block_to_free_list(block); + block = next; + } + } + + inline Block* try_get_block_from_free_list() + { + return freeList.try_get(); + } + + // Gets a free block from one of the memory pools, or allocates a new one (if applicable) + Block* requisition_block() + { + auto block = try_get_block_from_initial_pool(); + if (block != nullptr) { + return block; + } + + block = try_get_block_from_free_list(); + if (block != nullptr) { + return block; + } + + return create(); + } + + + ////////////////////////////////// + // Producer list manipulation + ////////////////////////////////// + + ProducerBase* recycle_or_create_producer() + { + bool recycled; + return recycle_or_create_producer(recycled); + } + + ProducerBase* recycle_or_create_producer(bool& recycled) + { + // Try to re-use one first + for (auto ptr = producerListTail.load(std::memory_order_acquire); ptr != nullptr; ptr = ptr->next_prod()) { + if (ptr->inactive.load(std::memory_order_relaxed)) { + if( ptr->size_approx() == 0 ) + { + bool expected = true; + if (ptr->inactive.compare_exchange_strong(expected, /* desired */ false, std::memory_order_acquire, std::memory_order_relaxed)) { + // We caught one! It's been marked as activated, the caller can have it + recycled = true; + return ptr; + } + } + } + } + + recycled = false; + return add_producer(static_cast(create(this))); + } + + ProducerBase* add_producer(ProducerBase* producer) + { + // Handle failed memory allocation + if (producer == nullptr) { + return nullptr; + } + + producerCount.fetch_add(1, std::memory_order_relaxed); + + // Add it to the lock-free list + auto prevTail = producerListTail.load(std::memory_order_relaxed); + do { + producer->next = prevTail; + } while (!producerListTail.compare_exchange_weak(prevTail, producer, std::memory_order_release, std::memory_order_relaxed)); + + return producer; + } + + void reown_producers() + { + // After another instance is moved-into/swapped-with this one, all the + // producers we stole still think their parents are the other queue. + // So fix them up! + for (auto ptr = producerListTail.load(std::memory_order_relaxed); ptr != nullptr; ptr = ptr->next_prod()) { + ptr->parent = this; + } + } + + ////////////////////////////////// + // Utility functions + ////////////////////////////////// + + template + static inline U* create_array(size_t count) + { + assert(count > 0); + return static_cast((Traits::malloc)(sizeof(U) * count)); + } + + template + static inline void destroy_array(U* p, size_t count) + { + ((void)count); + if (p != nullptr) { + assert(count > 0); + (Traits::free)(p); + } + } + + template + static inline U* create() + { + auto p = (Traits::malloc)(sizeof(U)); + return new (p) U; + } + + template + static inline U* create(A1&& a1) + { + auto p = (Traits::malloc)(sizeof(U)); + return new (p) U(std::forward(a1)); + } + + template + static inline void destroy(U* p) + { + if (p != nullptr) { + p->~U(); + } + (Traits::free)(p); + } + +private: + std::atomic producerListTail; + std::atomic producerCount; + + std::atomic initialBlockPoolIndex; + Block* initialBlockPool; + size_t initialBlockPoolSize; + + FreeList freeList; + + std::atomic nextExplicitConsumerId; + std::atomic globalExplicitConsumerOffset; +}; + + +template +ProducerToken::ProducerToken(ConcurrentQueue& queue) + : producer(queue.recycle_or_create_producer()) +{ + if (producer != nullptr) { + producer->token = this; + producer->threadId = detail::GetThreadHandleImpl(); + } +} + +template +ConsumerToken::ConsumerToken(ConcurrentQueue& queue) + : itemsConsumedFromCurrent(0), currentProducer(nullptr), desiredProducer(nullptr) +{ + initialOffset = queue.nextExplicitConsumerId.fetch_add(1, std::memory_order_release); + lastKnownGlobalOffset = static_cast(-1); +} + +template +inline void swap(ConcurrentQueue& a, ConcurrentQueue& b) noexcept +{ + a.swap(b); +} + +inline void swap(ProducerToken& a, ProducerToken& b) noexcept +{ + a.swap(b); +} + +inline void swap(ConsumerToken& a, ConsumerToken& b) noexcept +{ + a.swap(b); +} + +} + +} /* namespace tracy */ + +#if defined(__GNUC__) +#pragma GCC diagnostic pop +#endif diff --git a/externals/tracy/public/client/tracy_rpmalloc.cpp b/externals/tracy/public/client/tracy_rpmalloc.cpp new file mode 100644 index 000000000..711505d21 --- /dev/null +++ b/externals/tracy/public/client/tracy_rpmalloc.cpp @@ -0,0 +1,3518 @@ +#ifdef TRACY_ENABLE + +/* rpmalloc.c - Memory allocator - Public Domain - 2016-2020 Mattias Jansson + * + * This library provides a cross-platform lock free thread caching malloc implementation in C11. + * The latest source code is always available at + * + * https://github.com/mjansson/rpmalloc + * + * This library is put in the public domain; you can redistribute it and/or modify it without any restrictions. + * + */ + +#include "tracy_rpmalloc.hpp" + +#define BUILD_DYNAMIC_LINK 1 + +//////////// +/// +/// Build time configurable limits +/// +////// + +#if defined(__clang__) +#pragma clang diagnostic ignored "-Wunused-macros" +#pragma clang diagnostic ignored "-Wunused-function" +#if __has_warning("-Wreserved-identifier") +#pragma clang diagnostic ignored "-Wreserved-identifier" +#endif +#elif defined(__GNUC__) +#pragma GCC diagnostic ignored "-Wunused-macros" +#pragma GCC diagnostic ignored "-Wunused-function" +#pragma GCC diagnostic ignored "-Warray-bounds" +#endif + +#ifndef HEAP_ARRAY_SIZE +//! Size of heap hashmap +#define HEAP_ARRAY_SIZE 47 +#endif +#ifndef ENABLE_THREAD_CACHE +//! Enable per-thread cache +#define ENABLE_THREAD_CACHE 1 +#endif +#ifndef ENABLE_GLOBAL_CACHE +//! Enable global cache shared between all threads, requires thread cache +#define ENABLE_GLOBAL_CACHE 1 +#endif +#ifndef ENABLE_VALIDATE_ARGS +//! Enable validation of args to public entry points +#define ENABLE_VALIDATE_ARGS 0 +#endif +#ifndef ENABLE_STATISTICS +//! Enable statistics collection +#define ENABLE_STATISTICS 0 +#endif +#ifndef ENABLE_ASSERTS +//! Enable asserts +#define ENABLE_ASSERTS 0 +#endif +#ifndef ENABLE_OVERRIDE +//! Override standard library malloc/free and new/delete entry points +#define ENABLE_OVERRIDE 0 +#endif +#ifndef ENABLE_PRELOAD +//! Support preloading +#define ENABLE_PRELOAD 0 +#endif +#ifndef DISABLE_UNMAP +//! Disable unmapping memory pages (also enables unlimited cache) +#define DISABLE_UNMAP 0 +#endif +#ifndef ENABLE_UNLIMITED_CACHE +//! Enable unlimited global cache (no unmapping until finalization) +#define ENABLE_UNLIMITED_CACHE 0 +#endif +#ifndef ENABLE_ADAPTIVE_THREAD_CACHE +//! Enable adaptive thread cache size based on use heuristics +#define ENABLE_ADAPTIVE_THREAD_CACHE 0 +#endif +#ifndef DEFAULT_SPAN_MAP_COUNT +//! Default number of spans to map in call to map more virtual memory (default values yield 4MiB here) +#define DEFAULT_SPAN_MAP_COUNT 64 +#endif +#ifndef GLOBAL_CACHE_MULTIPLIER +//! Multiplier for global cache +#define GLOBAL_CACHE_MULTIPLIER 8 +#endif + +#if DISABLE_UNMAP && !ENABLE_GLOBAL_CACHE +#error Must use global cache if unmap is disabled +#endif + +#if DISABLE_UNMAP +#undef ENABLE_UNLIMITED_CACHE +#define ENABLE_UNLIMITED_CACHE 1 +#endif + +#if !ENABLE_GLOBAL_CACHE +#undef ENABLE_UNLIMITED_CACHE +#define ENABLE_UNLIMITED_CACHE 0 +#endif + +#if !ENABLE_THREAD_CACHE +#undef ENABLE_ADAPTIVE_THREAD_CACHE +#define ENABLE_ADAPTIVE_THREAD_CACHE 0 +#endif + +#if defined(_WIN32) || defined(__WIN32__) || defined(_WIN64) +# define PLATFORM_WINDOWS 1 +# define PLATFORM_POSIX 0 +#else +# define PLATFORM_WINDOWS 0 +# define PLATFORM_POSIX 1 +#endif + +/// Platform and arch specifics +#if defined(_MSC_VER) && !defined(__clang__) +# pragma warning (disable: 5105) +# ifndef FORCEINLINE +# define FORCEINLINE inline __forceinline +# endif +#else +# ifndef FORCEINLINE +# define FORCEINLINE inline __attribute__((__always_inline__)) +# endif +#endif +#if PLATFORM_WINDOWS +# ifndef WIN32_LEAN_AND_MEAN +# define WIN32_LEAN_AND_MEAN +# endif +# include +# if ENABLE_VALIDATE_ARGS +# include +# endif +#else +# include +# include +# include +# include +# if defined(__linux__) || defined(__ANDROID__) +# include +# if !defined(PR_SET_VMA) +# define PR_SET_VMA 0x53564d41 +# define PR_SET_VMA_ANON_NAME 0 +# endif +# endif +# if defined(__APPLE__) +# include +# if !TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR +# include +# include +# endif +# include +# endif +# if defined(__HAIKU__) || defined(__TINYC__) +# include +# endif +#endif + +#include +#include +#include + +#if defined(_WIN32) && (!defined(BUILD_DYNAMIC_LINK) || !BUILD_DYNAMIC_LINK) +#include +static DWORD fls_key; +#endif + +#if PLATFORM_POSIX +# include +# include +# ifdef __FreeBSD__ +# include +# define MAP_HUGETLB MAP_ALIGNED_SUPER +# ifndef PROT_MAX +# define PROT_MAX(f) 0 +# endif +# else +# define PROT_MAX(f) 0 +# endif +# ifdef __sun +extern int madvise(caddr_t, size_t, int); +# endif +# ifndef MAP_UNINITIALIZED +# define MAP_UNINITIALIZED 0 +# endif +#endif +#include + +#if ENABLE_ASSERTS +# undef NDEBUG +# if defined(_MSC_VER) && !defined(_DEBUG) +# define _DEBUG +# endif +# include +#define RPMALLOC_TOSTRING_M(x) #x +#define RPMALLOC_TOSTRING(x) RPMALLOC_TOSTRING_M(x) +#define rpmalloc_assert(truth, message) \ + do { \ + if (!(truth)) { \ + if (_memory_config.error_callback) { \ + _memory_config.error_callback( \ + message " (" RPMALLOC_TOSTRING(truth) ") at " __FILE__ ":" RPMALLOC_TOSTRING(__LINE__)); \ + } else { \ + assert((truth) && message); \ + } \ + } \ + } while (0) +#else +# define rpmalloc_assert(truth, message) do {} while(0) +#endif +#if ENABLE_STATISTICS +# include +#endif + +////// +/// +/// Atomic access abstraction (since MSVC does not do C11 yet) +/// +////// + +#include + +typedef std::atomic atomic32_t; +typedef std::atomic atomic64_t; +typedef std::atomic atomicptr_t; + +static FORCEINLINE int32_t atomic_load32(atomic32_t* src) { return std::atomic_load_explicit(src, std::memory_order_relaxed); } +static FORCEINLINE void atomic_store32(atomic32_t* dst, int32_t val) { std::atomic_store_explicit(dst, val, std::memory_order_relaxed); } +static FORCEINLINE int32_t atomic_incr32(atomic32_t* val) { return std::atomic_fetch_add_explicit(val, 1, std::memory_order_relaxed) + 1; } +static FORCEINLINE int32_t atomic_decr32(atomic32_t* val) { return std::atomic_fetch_add_explicit(val, -1, std::memory_order_relaxed) - 1; } +static FORCEINLINE int32_t atomic_add32(atomic32_t* val, int32_t add) { return std::atomic_fetch_add_explicit(val, add, std::memory_order_relaxed) + add; } +static FORCEINLINE int atomic_cas32_acquire(atomic32_t* dst, int32_t val, int32_t ref) { return std::atomic_compare_exchange_weak_explicit(dst, &ref, val, std::memory_order_acquire, std::memory_order_relaxed); } +static FORCEINLINE void atomic_store32_release(atomic32_t* dst, int32_t val) { std::atomic_store_explicit(dst, val, std::memory_order_release); } +static FORCEINLINE int64_t atomic_load64(atomic64_t* val) { return std::atomic_load_explicit(val, std::memory_order_relaxed); } +static FORCEINLINE int64_t atomic_add64(atomic64_t* val, int64_t add) { return std::atomic_fetch_add_explicit(val, add, std::memory_order_relaxed) + add; } +static FORCEINLINE void* atomic_load_ptr(atomicptr_t* src) { return std::atomic_load_explicit(src, std::memory_order_relaxed); } +static FORCEINLINE void atomic_store_ptr(atomicptr_t* dst, void* val) { std::atomic_store_explicit(dst, val, std::memory_order_relaxed); } +static FORCEINLINE void atomic_store_ptr_release(atomicptr_t* dst, void* val) { std::atomic_store_explicit(dst, val, std::memory_order_release); } +static FORCEINLINE void* atomic_exchange_ptr_acquire(atomicptr_t* dst, void* val) { return std::atomic_exchange_explicit(dst, val, std::memory_order_acquire); } +static FORCEINLINE int atomic_cas_ptr(atomicptr_t* dst, void* val, void* ref) { return std::atomic_compare_exchange_weak_explicit(dst, &ref, val, std::memory_order_relaxed, std::memory_order_relaxed); } + +#if defined(_MSC_VER) && !defined(__clang__) + +#define EXPECTED(x) (x) +#define UNEXPECTED(x) (x) + +#else + +#define EXPECTED(x) __builtin_expect((x), 1) +#define UNEXPECTED(x) __builtin_expect((x), 0) + +#endif + +//////////// +/// +/// Statistics related functions (evaluate to nothing when statistics not enabled) +/// +////// + +#if ENABLE_STATISTICS +# define _rpmalloc_stat_inc(counter) atomic_incr32(counter) +# define _rpmalloc_stat_dec(counter) atomic_decr32(counter) +# define _rpmalloc_stat_add(counter, value) atomic_add32(counter, (int32_t)(value)) +# define _rpmalloc_stat_add64(counter, value) atomic_add64(counter, (int64_t)(value)) +# define _rpmalloc_stat_add_peak(counter, value, peak) do { int32_t _cur_count = atomic_add32(counter, (int32_t)(value)); if (_cur_count > (peak)) peak = _cur_count; } while (0) +# define _rpmalloc_stat_sub(counter, value) atomic_add32(counter, -(int32_t)(value)) +# define _rpmalloc_stat_inc_alloc(heap, class_idx) do { \ + int32_t alloc_current = atomic_incr32(&heap->size_class_use[class_idx].alloc_current); \ + if (alloc_current > heap->size_class_use[class_idx].alloc_peak) \ + heap->size_class_use[class_idx].alloc_peak = alloc_current; \ + atomic_incr32(&heap->size_class_use[class_idx].alloc_total); \ +} while(0) +# define _rpmalloc_stat_inc_free(heap, class_idx) do { \ + atomic_decr32(&heap->size_class_use[class_idx].alloc_current); \ + atomic_incr32(&heap->size_class_use[class_idx].free_total); \ +} while(0) +#else +# define _rpmalloc_stat_inc(counter) do {} while(0) +# define _rpmalloc_stat_dec(counter) do {} while(0) +# define _rpmalloc_stat_add(counter, value) do {} while(0) +# define _rpmalloc_stat_add64(counter, value) do {} while(0) +# define _rpmalloc_stat_add_peak(counter, value, peak) do {} while (0) +# define _rpmalloc_stat_sub(counter, value) do {} while(0) +# define _rpmalloc_stat_inc_alloc(heap, class_idx) do {} while(0) +# define _rpmalloc_stat_inc_free(heap, class_idx) do {} while(0) +#endif + + +/// +/// Preconfigured limits and sizes +/// + +//! Granularity of a small allocation block (must be power of two) +#define SMALL_GRANULARITY 16 +//! Small granularity shift count +#define SMALL_GRANULARITY_SHIFT 4 +//! Number of small block size classes +#define SMALL_CLASS_COUNT 65 +//! Maximum size of a small block +#define SMALL_SIZE_LIMIT (SMALL_GRANULARITY * (SMALL_CLASS_COUNT - 1)) +//! Granularity of a medium allocation block +#define MEDIUM_GRANULARITY 512 +//! Medium granularity shift count +#define MEDIUM_GRANULARITY_SHIFT 9 +//! Number of medium block size classes +#define MEDIUM_CLASS_COUNT 61 +//! Total number of small + medium size classes +#define SIZE_CLASS_COUNT (SMALL_CLASS_COUNT + MEDIUM_CLASS_COUNT) +//! Number of large block size classes +#define LARGE_CLASS_COUNT 63 +//! Maximum size of a medium block +#define MEDIUM_SIZE_LIMIT (SMALL_SIZE_LIMIT + (MEDIUM_GRANULARITY * MEDIUM_CLASS_COUNT)) +//! Maximum size of a large block +#define LARGE_SIZE_LIMIT ((LARGE_CLASS_COUNT * _memory_span_size) - SPAN_HEADER_SIZE) +//! Size of a span header (must be a multiple of SMALL_GRANULARITY and a power of two) +#define SPAN_HEADER_SIZE 128 +//! Number of spans in thread cache +#define MAX_THREAD_SPAN_CACHE 400 +//! Number of spans to transfer between thread and global cache +#define THREAD_SPAN_CACHE_TRANSFER 64 +//! Number of spans in thread cache for large spans (must be greater than LARGE_CLASS_COUNT / 2) +#define MAX_THREAD_SPAN_LARGE_CACHE 100 +//! Number of spans to transfer between thread and global cache for large spans +#define THREAD_SPAN_LARGE_CACHE_TRANSFER 6 + +static_assert((SMALL_GRANULARITY & (SMALL_GRANULARITY - 1)) == 0, "Small granularity must be power of two"); +static_assert((SPAN_HEADER_SIZE & (SPAN_HEADER_SIZE - 1)) == 0, "Span header size must be power of two"); + +#if ENABLE_VALIDATE_ARGS +//! Maximum allocation size to avoid integer overflow +#undef MAX_ALLOC_SIZE +#define MAX_ALLOC_SIZE (((size_t)-1) - _memory_span_size) +#endif + +#define pointer_offset(ptr, ofs) (void*)((char*)(ptr) + (ptrdiff_t)(ofs)) +#define pointer_diff(first, second) (ptrdiff_t)((const char*)(first) - (const char*)(second)) + +#define INVALID_POINTER ((void*)((uintptr_t)-1)) + +#define SIZE_CLASS_LARGE SIZE_CLASS_COUNT +#define SIZE_CLASS_HUGE ((uint32_t)-1) + +//////////// +/// +/// Data types +/// +////// + +namespace tracy +{ + +//! A memory heap, per thread +typedef struct heap_t heap_t; +//! Span of memory pages +typedef struct span_t span_t; +//! Span list +typedef struct span_list_t span_list_t; +//! Span active data +typedef struct span_active_t span_active_t; +//! Size class definition +typedef struct size_class_t size_class_t; +//! Global cache +typedef struct global_cache_t global_cache_t; + +//! Flag indicating span is the first (master) span of a split superspan +#define SPAN_FLAG_MASTER 1U +//! Flag indicating span is a secondary (sub) span of a split superspan +#define SPAN_FLAG_SUBSPAN 2U +//! Flag indicating span has blocks with increased alignment +#define SPAN_FLAG_ALIGNED_BLOCKS 4U +//! Flag indicating an unmapped master span +#define SPAN_FLAG_UNMAPPED_MASTER 8U + +#if ENABLE_ADAPTIVE_THREAD_CACHE || ENABLE_STATISTICS +struct span_use_t { + //! Current number of spans used (actually used, not in cache) + atomic32_t current; + //! High water mark of spans used + atomic32_t high; +#if ENABLE_STATISTICS + //! Number of spans in deferred list + atomic32_t spans_deferred; + //! Number of spans transitioned to global cache + atomic32_t spans_to_global; + //! Number of spans transitioned from global cache + atomic32_t spans_from_global; + //! Number of spans transitioned to thread cache + atomic32_t spans_to_cache; + //! Number of spans transitioned from thread cache + atomic32_t spans_from_cache; + //! Number of spans transitioned to reserved state + atomic32_t spans_to_reserved; + //! Number of spans transitioned from reserved state + atomic32_t spans_from_reserved; + //! Number of raw memory map calls + atomic32_t spans_map_calls; +#endif +}; +typedef struct span_use_t span_use_t; +#endif + +#if ENABLE_STATISTICS +struct size_class_use_t { + //! Current number of allocations + atomic32_t alloc_current; + //! Peak number of allocations + int32_t alloc_peak; + //! Total number of allocations + atomic32_t alloc_total; + //! Total number of frees + atomic32_t free_total; + //! Number of spans in use + atomic32_t spans_current; + //! Number of spans transitioned to cache + int32_t spans_peak; + //! Number of spans transitioned to cache + atomic32_t spans_to_cache; + //! Number of spans transitioned from cache + atomic32_t spans_from_cache; + //! Number of spans transitioned from reserved state + atomic32_t spans_from_reserved; + //! Number of spans mapped + atomic32_t spans_map_calls; + int32_t unused; +}; +typedef struct size_class_use_t size_class_use_t; +#endif + +// A span can either represent a single span of memory pages with size declared by span_map_count configuration variable, +// or a set of spans in a continuous region, a super span. Any reference to the term "span" usually refers to both a single +// span or a super span. A super span can further be divided into multiple spans (or this, super spans), where the first +// (super)span is the master and subsequent (super)spans are subspans. The master span keeps track of how many subspans +// that are still alive and mapped in virtual memory, and once all subspans and master have been unmapped the entire +// superspan region is released and unmapped (on Windows for example, the entire superspan range has to be released +// in the same call to release the virtual memory range, but individual subranges can be decommitted individually +// to reduce physical memory use). +struct span_t { + //! Free list + void* free_list; + //! Total block count of size class + uint32_t block_count; + //! Size class + uint32_t size_class; + //! Index of last block initialized in free list + uint32_t free_list_limit; + //! Number of used blocks remaining when in partial state + uint32_t used_count; + //! Deferred free list + atomicptr_t free_list_deferred; + //! Size of deferred free list, or list of spans when part of a cache list + uint32_t list_size; + //! Size of a block + uint32_t block_size; + //! Flags and counters + uint32_t flags; + //! Number of spans + uint32_t span_count; + //! Total span counter for master spans + uint32_t total_spans; + //! Offset from master span for subspans + uint32_t offset_from_master; + //! Remaining span counter, for master spans + atomic32_t remaining_spans; + //! Alignment offset + uint32_t align_offset; + //! Owning heap + heap_t* heap; + //! Next span + span_t* next; + //! Previous span + span_t* prev; +}; +static_assert(sizeof(span_t) <= SPAN_HEADER_SIZE, "span size mismatch"); + +struct span_cache_t { + size_t count; + span_t* span[MAX_THREAD_SPAN_CACHE]; +}; +typedef struct span_cache_t span_cache_t; + +struct span_large_cache_t { + size_t count; + span_t* span[MAX_THREAD_SPAN_LARGE_CACHE]; +}; +typedef struct span_large_cache_t span_large_cache_t; + +struct heap_size_class_t { + //! Free list of active span + void* free_list; + //! Double linked list of partially used spans with free blocks. + // Previous span pointer in head points to tail span of list. + span_t* partial_span; + //! Early level cache of fully free spans + span_t* cache; +}; +typedef struct heap_size_class_t heap_size_class_t; + +// Control structure for a heap, either a thread heap or a first class heap if enabled +struct heap_t { + //! Owning thread ID + uintptr_t owner_thread; + //! Free lists for each size class + heap_size_class_t size_class[SIZE_CLASS_COUNT]; +#if ENABLE_THREAD_CACHE + //! Arrays of fully freed spans, single span + span_cache_t span_cache; +#endif + //! List of deferred free spans (single linked list) + atomicptr_t span_free_deferred; + //! Number of full spans + size_t full_span_count; + //! Mapped but unused spans + span_t* span_reserve; + //! Master span for mapped but unused spans + span_t* span_reserve_master; + //! Number of mapped but unused spans + uint32_t spans_reserved; + //! Child count + atomic32_t child_count; + //! Next heap in id list + heap_t* next_heap; + //! Next heap in orphan list + heap_t* next_orphan; + //! Heap ID + int32_t id; + //! Finalization state flag + int finalize; + //! Master heap owning the memory pages + heap_t* master_heap; +#if ENABLE_THREAD_CACHE + //! Arrays of fully freed spans, large spans with > 1 span count + span_large_cache_t span_large_cache[LARGE_CLASS_COUNT - 1]; +#endif +#if RPMALLOC_FIRST_CLASS_HEAPS + //! Double linked list of fully utilized spans with free blocks for each size class. + // Previous span pointer in head points to tail span of list. + span_t* full_span[SIZE_CLASS_COUNT]; + //! Double linked list of large and huge spans allocated by this heap + span_t* large_huge_span; +#endif +#if ENABLE_ADAPTIVE_THREAD_CACHE || ENABLE_STATISTICS + //! Current and high water mark of spans used per span count + span_use_t span_use[LARGE_CLASS_COUNT]; +#endif +#if ENABLE_STATISTICS + //! Allocation stats per size class + size_class_use_t size_class_use[SIZE_CLASS_COUNT + 1]; + //! Number of bytes transitioned thread -> global + atomic64_t thread_to_global; + //! Number of bytes transitioned global -> thread + atomic64_t global_to_thread; +#endif +}; + +// Size class for defining a block size bucket +struct size_class_t { + //! Size of blocks in this class + uint32_t block_size; + //! Number of blocks in each chunk + uint16_t block_count; + //! Class index this class is merged with + uint16_t class_idx; +}; +static_assert(sizeof(size_class_t) == 8, "Size class size mismatch"); + +struct global_cache_t { + //! Cache lock + atomic32_t lock; + //! Cache count + uint32_t count; +#if ENABLE_STATISTICS + //! Insert count + size_t insert_count; + //! Extract count + size_t extract_count; +#endif + //! Cached spans + span_t* span[GLOBAL_CACHE_MULTIPLIER * MAX_THREAD_SPAN_CACHE]; + //! Unlimited cache overflow + span_t* overflow; +}; + +//////////// +/// +/// Global data +/// +////// + +//! Default span size (64KiB) +#define _memory_default_span_size (64 * 1024) +#define _memory_default_span_size_shift 16 +#define _memory_default_span_mask (~((uintptr_t)(_memory_span_size - 1))) + +//! Initialized flag +static int _rpmalloc_initialized; +//! Main thread ID +static uintptr_t _rpmalloc_main_thread_id; +//! Configuration +static rpmalloc_config_t _memory_config; +//! Memory page size +static size_t _memory_page_size; +//! Shift to divide by page size +static size_t _memory_page_size_shift; +//! Granularity at which memory pages are mapped by OS +static size_t _memory_map_granularity; +#if RPMALLOC_CONFIGURABLE +//! Size of a span of memory pages +static size_t _memory_span_size; +//! Shift to divide by span size +static size_t _memory_span_size_shift; +//! Mask to get to start of a memory span +static uintptr_t _memory_span_mask; +#else +//! Hardwired span size +#define _memory_span_size _memory_default_span_size +#define _memory_span_size_shift _memory_default_span_size_shift +#define _memory_span_mask _memory_default_span_mask +#endif +//! Number of spans to map in each map call +static size_t _memory_span_map_count; +//! Number of spans to keep reserved in each heap +static size_t _memory_heap_reserve_count; +//! Global size classes +static size_class_t _memory_size_class[SIZE_CLASS_COUNT]; +//! Run-time size limit of medium blocks +static size_t _memory_medium_size_limit; +//! Heap ID counter +static atomic32_t _memory_heap_id; +//! Huge page support +static int _memory_huge_pages; +#if ENABLE_GLOBAL_CACHE +//! Global span cache +static global_cache_t _memory_span_cache[LARGE_CLASS_COUNT]; +#endif +//! Global reserved spans +static span_t* _memory_global_reserve; +//! Global reserved count +static size_t _memory_global_reserve_count; +//! Global reserved master +static span_t* _memory_global_reserve_master; +//! All heaps +static heap_t* _memory_heaps[HEAP_ARRAY_SIZE]; +//! Used to restrict access to mapping memory for huge pages +static atomic32_t _memory_global_lock; +//! Orphaned heaps +static heap_t* _memory_orphan_heaps; +#if RPMALLOC_FIRST_CLASS_HEAPS +//! Orphaned heaps (first class heaps) +static heap_t* _memory_first_class_orphan_heaps; +#endif +#if ENABLE_STATISTICS +//! Allocations counter +static atomic64_t _allocation_counter; +//! Deallocations counter +static atomic64_t _deallocation_counter; +//! Active heap count +static atomic32_t _memory_active_heaps; +//! Number of currently mapped memory pages +static atomic32_t _mapped_pages; +//! Peak number of concurrently mapped memory pages +static int32_t _mapped_pages_peak; +//! Number of mapped master spans +static atomic32_t _master_spans; +//! Number of unmapped dangling master spans +static atomic32_t _unmapped_master_spans; +//! Running counter of total number of mapped memory pages since start +static atomic32_t _mapped_total; +//! Running counter of total number of unmapped memory pages since start +static atomic32_t _unmapped_total; +//! Number of currently mapped memory pages in OS calls +static atomic32_t _mapped_pages_os; +//! Number of currently allocated pages in huge allocations +static atomic32_t _huge_pages_current; +//! Peak number of currently allocated pages in huge allocations +static int32_t _huge_pages_peak; +#endif + +//////////// +/// +/// Thread local heap and ID +/// +////// + +//! Current thread heap +#if ((defined(__APPLE__) || defined(__HAIKU__)) && ENABLE_PRELOAD) || defined(__TINYC__) +static pthread_key_t _memory_thread_heap; +#else +# ifdef _MSC_VER +# define _Thread_local __declspec(thread) +# define TLS_MODEL +# else +# ifndef __HAIKU__ +# define TLS_MODEL __attribute__((tls_model("initial-exec"))) +# else +# define TLS_MODEL +# endif +# if !defined(__clang__) && defined(__GNUC__) +# define _Thread_local __thread +# endif +# endif +static _Thread_local heap_t* _memory_thread_heap TLS_MODEL; +#endif + +static inline heap_t* +get_thread_heap_raw(void) { +#if (defined(__APPLE__) || defined(__HAIKU__)) && ENABLE_PRELOAD + return pthread_getspecific(_memory_thread_heap); +#else + return _memory_thread_heap; +#endif +} + +//! Get the current thread heap +static inline heap_t* +get_thread_heap(void) { + heap_t* heap = get_thread_heap_raw(); +#if ENABLE_PRELOAD + if (EXPECTED(heap != 0)) + return heap; + rpmalloc_initialize(); + return get_thread_heap_raw(); +#else + return heap; +#endif +} + +//! Fast thread ID +static inline uintptr_t +get_thread_id(void) { +#if defined(_WIN32) + return (uintptr_t)((void*)NtCurrentTeb()); +#elif (defined(__GNUC__) || defined(__clang__)) && !defined(__CYGWIN__) + uintptr_t tid; +# if defined(__i386__) + __asm__("movl %%gs:0, %0" : "=r" (tid) : : ); +# elif defined(__x86_64__) +# if defined(__MACH__) + __asm__("movq %%gs:0, %0" : "=r" (tid) : : ); +# else + __asm__("movq %%fs:0, %0" : "=r" (tid) : : ); +# endif +# elif defined(__arm__) + __asm__ volatile ("mrc p15, 0, %0, c13, c0, 3" : "=r" (tid)); +# elif defined(__aarch64__) +# if defined(__MACH__) + // tpidr_el0 likely unused, always return 0 on iOS + __asm__ volatile ("mrs %0, tpidrro_el0" : "=r" (tid)); +# else + __asm__ volatile ("mrs %0, tpidr_el0" : "=r" (tid)); +# endif +# else + tid = (uintptr_t)((void*)get_thread_heap_raw()); +# endif + return tid; +#else + return (uintptr_t)((void*)get_thread_heap_raw()); +#endif +} + +//! Set the current thread heap +static void +set_thread_heap(heap_t* heap) { +#if ((defined(__APPLE__) || defined(__HAIKU__)) && ENABLE_PRELOAD) || defined(__TINYC__) + pthread_setspecific(_memory_thread_heap, heap); +#else + _memory_thread_heap = heap; +#endif + if (heap) + heap->owner_thread = get_thread_id(); +} + +//! Set main thread ID +extern void +rpmalloc_set_main_thread(void); + +void +rpmalloc_set_main_thread(void) { + _rpmalloc_main_thread_id = get_thread_id(); +} + +static void +_rpmalloc_spin(void) { +#if defined(_MSC_VER) + _mm_pause(); +#elif defined(__x86_64__) || defined(__i386__) + __asm__ volatile("pause" ::: "memory"); +#elif defined(__aarch64__) || (defined(__arm__) && __ARM_ARCH >= 7) + __asm__ volatile("yield" ::: "memory"); +#elif defined(__powerpc__) || defined(__powerpc64__) + // No idea if ever been compiled in such archs but ... as precaution + __asm__ volatile("or 27,27,27"); +#elif defined(__sparc__) + __asm__ volatile("rd %ccr, %g0 \n\trd %ccr, %g0 \n\trd %ccr, %g0"); +#else + struct timespec ts = {0}; + nanosleep(&ts, 0); +#endif +} + +#if defined(_WIN32) && (!defined(BUILD_DYNAMIC_LINK) || !BUILD_DYNAMIC_LINK) +static void NTAPI +_rpmalloc_thread_destructor(void* value) { +#if ENABLE_OVERRIDE + // If this is called on main thread it means rpmalloc_finalize + // has not been called and shutdown is forced (through _exit) or unclean + if (get_thread_id() == _rpmalloc_main_thread_id) + return; +#endif + if (value) + rpmalloc_thread_finalize(1); +} +#endif + + +//////////// +/// +/// Low level memory map/unmap +/// +////// + +static void +_rpmalloc_set_name(void* address, size_t size) { +#if defined(__linux__) || defined(__ANDROID__) + const char *name = _memory_huge_pages ? _memory_config.huge_page_name : _memory_config.page_name; + if (address == MAP_FAILED || !name) + return; + // If the kernel does not support CONFIG_ANON_VMA_NAME or if the call fails + // (e.g. invalid name) it is a no-op basically. + (void)prctl(PR_SET_VMA, PR_SET_VMA_ANON_NAME, (uintptr_t)address, size, (uintptr_t)name); +#else + (void)sizeof(size); + (void)sizeof(address); +#endif +} + + +//! Map more virtual memory +// size is number of bytes to map +// offset receives the offset in bytes from start of mapped region +// returns address to start of mapped region to use +static void* +_rpmalloc_mmap(size_t size, size_t* offset) { + rpmalloc_assert(!(size % _memory_page_size), "Invalid mmap size"); + rpmalloc_assert(size >= _memory_page_size, "Invalid mmap size"); + void* address = _memory_config.memory_map(size, offset); + if (EXPECTED(address != 0)) { + _rpmalloc_stat_add_peak(&_mapped_pages, (size >> _memory_page_size_shift), _mapped_pages_peak); + _rpmalloc_stat_add(&_mapped_total, (size >> _memory_page_size_shift)); + } + return address; +} + +//! Unmap virtual memory +// address is the memory address to unmap, as returned from _memory_map +// size is the number of bytes to unmap, which might be less than full region for a partial unmap +// offset is the offset in bytes to the actual mapped region, as set by _memory_map +// release is set to 0 for partial unmap, or size of entire range for a full unmap +static void +_rpmalloc_unmap(void* address, size_t size, size_t offset, size_t release) { + rpmalloc_assert(!release || (release >= size), "Invalid unmap size"); + rpmalloc_assert(!release || (release >= _memory_page_size), "Invalid unmap size"); + if (release) { + rpmalloc_assert(!(release % _memory_page_size), "Invalid unmap size"); + _rpmalloc_stat_sub(&_mapped_pages, (release >> _memory_page_size_shift)); + _rpmalloc_stat_add(&_unmapped_total, (release >> _memory_page_size_shift)); + } + _memory_config.memory_unmap(address, size, offset, release); +} + +//! Default implementation to map new pages to virtual memory +static void* +_rpmalloc_mmap_os(size_t size, size_t* offset) { + //Either size is a heap (a single page) or a (multiple) span - we only need to align spans, and only if larger than map granularity + size_t padding = ((size >= _memory_span_size) && (_memory_span_size > _memory_map_granularity)) ? _memory_span_size : 0; + rpmalloc_assert(size >= _memory_page_size, "Invalid mmap size"); +#if PLATFORM_WINDOWS + //Ok to MEM_COMMIT - according to MSDN, "actual physical pages are not allocated unless/until the virtual addresses are actually accessed" + void* ptr = VirtualAlloc(0, size + padding, (_memory_huge_pages ? MEM_LARGE_PAGES : 0) | MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE); + if (!ptr) { + if (_memory_config.map_fail_callback) { + if (_memory_config.map_fail_callback(size + padding)) + return _rpmalloc_mmap_os(size, offset); + } else { + rpmalloc_assert(ptr, "Failed to map virtual memory block"); + } + return 0; + } +#else + int flags = MAP_PRIVATE | MAP_ANONYMOUS | MAP_UNINITIALIZED; +# if defined(__APPLE__) && !TARGET_OS_IPHONE && !TARGET_OS_SIMULATOR + int fd = (int)VM_MAKE_TAG(240U); + if (_memory_huge_pages) + fd |= VM_FLAGS_SUPERPAGE_SIZE_2MB; + void* ptr = mmap(0, size + padding, PROT_READ | PROT_WRITE, flags, fd, 0); +# elif defined(MAP_HUGETLB) + void* ptr = mmap(0, size + padding, PROT_READ | PROT_WRITE | PROT_MAX(PROT_READ | PROT_WRITE), (_memory_huge_pages ? MAP_HUGETLB : 0) | flags, -1, 0); +# if defined(MADV_HUGEPAGE) + // In some configurations, huge pages allocations might fail thus + // we fallback to normal allocations and promote the region as transparent huge page + if ((ptr == MAP_FAILED || !ptr) && _memory_huge_pages) { + ptr = mmap(0, size + padding, PROT_READ | PROT_WRITE, flags, -1, 0); + if (ptr && ptr != MAP_FAILED) { + int prm = madvise(ptr, size + padding, MADV_HUGEPAGE); + (void)prm; + rpmalloc_assert((prm == 0), "Failed to promote the page to THP"); + } + } +# endif + _rpmalloc_set_name(ptr, size + padding); +# elif defined(MAP_ALIGNED) + const size_t align = (sizeof(size_t) * 8) - (size_t)(__builtin_clzl(size - 1)); + void* ptr = mmap(0, size + padding, PROT_READ | PROT_WRITE, (_memory_huge_pages ? MAP_ALIGNED(align) : 0) | flags, -1, 0); +# elif defined(MAP_ALIGN) + caddr_t base = (_memory_huge_pages ? (caddr_t)(4 << 20) : 0); + void* ptr = mmap(base, size + padding, PROT_READ | PROT_WRITE, (_memory_huge_pages ? MAP_ALIGN : 0) | flags, -1, 0); +# else + void* ptr = mmap(0, size + padding, PROT_READ | PROT_WRITE, flags, -1, 0); +# endif + if ((ptr == MAP_FAILED) || !ptr) { + if (_memory_config.map_fail_callback) { + if (_memory_config.map_fail_callback(size + padding)) + return _rpmalloc_mmap_os(size, offset); + } else if (errno != ENOMEM) { + rpmalloc_assert((ptr != MAP_FAILED) && ptr, "Failed to map virtual memory block"); + } + return 0; + } +#endif + _rpmalloc_stat_add(&_mapped_pages_os, (int32_t)((size + padding) >> _memory_page_size_shift)); + if (padding) { + size_t final_padding = padding - ((uintptr_t)ptr & ~_memory_span_mask); + rpmalloc_assert(final_padding <= _memory_span_size, "Internal failure in padding"); + rpmalloc_assert(final_padding <= padding, "Internal failure in padding"); + rpmalloc_assert(!(final_padding % 8), "Internal failure in padding"); + ptr = pointer_offset(ptr, final_padding); + *offset = final_padding >> 3; + } + rpmalloc_assert((size < _memory_span_size) || !((uintptr_t)ptr & ~_memory_span_mask), "Internal failure in padding"); + return ptr; +} + +//! Default implementation to unmap pages from virtual memory +static void +_rpmalloc_unmap_os(void* address, size_t size, size_t offset, size_t release) { + rpmalloc_assert(release || (offset == 0), "Invalid unmap size"); + rpmalloc_assert(!release || (release >= _memory_page_size), "Invalid unmap size"); + rpmalloc_assert(size >= _memory_page_size, "Invalid unmap size"); + if (release && offset) { + offset <<= 3; + address = pointer_offset(address, -(int32_t)offset); + if ((release >= _memory_span_size) && (_memory_span_size > _memory_map_granularity)) { + //Padding is always one span size + release += _memory_span_size; + } + } +#if !DISABLE_UNMAP +#if PLATFORM_WINDOWS + if (!VirtualFree(address, release ? 0 : size, release ? MEM_RELEASE : MEM_DECOMMIT)) { + rpmalloc_assert(0, "Failed to unmap virtual memory block"); + } +#else + if (release) { + if (munmap(address, release)) { + rpmalloc_assert(0, "Failed to unmap virtual memory block"); + } + } else { +#if defined(MADV_FREE_REUSABLE) + int ret; + while ((ret = madvise(address, size, MADV_FREE_REUSABLE)) == -1 && (errno == EAGAIN)) + errno = 0; + if ((ret == -1) && (errno != 0)) { +#elif defined(MADV_DONTNEED) + if (madvise(address, size, MADV_DONTNEED)) { +#elif defined(MADV_PAGEOUT) + if (madvise(address, size, MADV_PAGEOUT)) { +#elif defined(MADV_FREE) + if (madvise(address, size, MADV_FREE)) { +#else + if (posix_madvise(address, size, POSIX_MADV_DONTNEED)) { +#endif + rpmalloc_assert(0, "Failed to madvise virtual memory block as free"); + } + } +#endif +#endif + if (release) + _rpmalloc_stat_sub(&_mapped_pages_os, release >> _memory_page_size_shift); +} + +static void +_rpmalloc_span_mark_as_subspan_unless_master(span_t* master, span_t* subspan, size_t span_count); + +//! Use global reserved spans to fulfill a memory map request (reserve size must be checked by caller) +static span_t* +_rpmalloc_global_get_reserved_spans(size_t span_count) { + span_t* span = _memory_global_reserve; + _rpmalloc_span_mark_as_subspan_unless_master(_memory_global_reserve_master, span, span_count); + _memory_global_reserve_count -= span_count; + if (_memory_global_reserve_count) + _memory_global_reserve = (span_t*)pointer_offset(span, span_count << _memory_span_size_shift); + else + _memory_global_reserve = 0; + return span; +} + +//! Store the given spans as global reserve (must only be called from within new heap allocation, not thread safe) +static void +_rpmalloc_global_set_reserved_spans(span_t* master, span_t* reserve, size_t reserve_span_count) { + _memory_global_reserve_master = master; + _memory_global_reserve_count = reserve_span_count; + _memory_global_reserve = reserve; +} + + +//////////// +/// +/// Span linked list management +/// +////// + +//! Add a span to double linked list at the head +static void +_rpmalloc_span_double_link_list_add(span_t** head, span_t* span) { + if (*head) + (*head)->prev = span; + span->next = *head; + *head = span; +} + +//! Pop head span from double linked list +static void +_rpmalloc_span_double_link_list_pop_head(span_t** head, span_t* span) { + rpmalloc_assert(*head == span, "Linked list corrupted"); + span = *head; + *head = span->next; +} + +//! Remove a span from double linked list +static void +_rpmalloc_span_double_link_list_remove(span_t** head, span_t* span) { + rpmalloc_assert(*head, "Linked list corrupted"); + if (*head == span) { + *head = span->next; + } else { + span_t* next_span = span->next; + span_t* prev_span = span->prev; + prev_span->next = next_span; + if (EXPECTED(next_span != 0)) + next_span->prev = prev_span; + } +} + + +//////////// +/// +/// Span control +/// +////// + +static void +_rpmalloc_heap_cache_insert(heap_t* heap, span_t* span); + +static void +_rpmalloc_heap_finalize(heap_t* heap); + +static void +_rpmalloc_heap_set_reserved_spans(heap_t* heap, span_t* master, span_t* reserve, size_t reserve_span_count); + +//! Declare the span to be a subspan and store distance from master span and span count +static void +_rpmalloc_span_mark_as_subspan_unless_master(span_t* master, span_t* subspan, size_t span_count) { + rpmalloc_assert((subspan != master) || (subspan->flags & SPAN_FLAG_MASTER), "Span master pointer and/or flag mismatch"); + if (subspan != master) { + subspan->flags = SPAN_FLAG_SUBSPAN; + subspan->offset_from_master = (uint32_t)((uintptr_t)pointer_diff(subspan, master) >> _memory_span_size_shift); + subspan->align_offset = 0; + } + subspan->span_count = (uint32_t)span_count; +} + +//! Use reserved spans to fulfill a memory map request (reserve size must be checked by caller) +static span_t* +_rpmalloc_span_map_from_reserve(heap_t* heap, size_t span_count) { + //Update the heap span reserve + span_t* span = heap->span_reserve; + heap->span_reserve = (span_t*)pointer_offset(span, span_count * _memory_span_size); + heap->spans_reserved -= (uint32_t)span_count; + + _rpmalloc_span_mark_as_subspan_unless_master(heap->span_reserve_master, span, span_count); + if (span_count <= LARGE_CLASS_COUNT) + _rpmalloc_stat_inc(&heap->span_use[span_count - 1].spans_from_reserved); + + return span; +} + +//! Get the aligned number of spans to map in based on wanted count, configured mapping granularity and the page size +static size_t +_rpmalloc_span_align_count(size_t span_count) { + size_t request_count = (span_count > _memory_span_map_count) ? span_count : _memory_span_map_count; + if ((_memory_page_size > _memory_span_size) && ((request_count * _memory_span_size) % _memory_page_size)) + request_count += _memory_span_map_count - (request_count % _memory_span_map_count); + return request_count; +} + +//! Setup a newly mapped span +static void +_rpmalloc_span_initialize(span_t* span, size_t total_span_count, size_t span_count, size_t align_offset) { + span->total_spans = (uint32_t)total_span_count; + span->span_count = (uint32_t)span_count; + span->align_offset = (uint32_t)align_offset; + span->flags = SPAN_FLAG_MASTER; + atomic_store32(&span->remaining_spans, (int32_t)total_span_count); +} + +static void +_rpmalloc_span_unmap(span_t* span); + +//! Map an aligned set of spans, taking configured mapping granularity and the page size into account +static span_t* +_rpmalloc_span_map_aligned_count(heap_t* heap, size_t span_count) { + //If we already have some, but not enough, reserved spans, release those to heap cache and map a new + //full set of spans. Otherwise we would waste memory if page size > span size (huge pages) + size_t aligned_span_count = _rpmalloc_span_align_count(span_count); + size_t align_offset = 0; + span_t* span = (span_t*)_rpmalloc_mmap(aligned_span_count * _memory_span_size, &align_offset); + if (!span) + return 0; + _rpmalloc_span_initialize(span, aligned_span_count, span_count, align_offset); + _rpmalloc_stat_inc(&_master_spans); + if (span_count <= LARGE_CLASS_COUNT) + _rpmalloc_stat_inc(&heap->span_use[span_count - 1].spans_map_calls); + if (aligned_span_count > span_count) { + span_t* reserved_spans = (span_t*)pointer_offset(span, span_count * _memory_span_size); + size_t reserved_count = aligned_span_count - span_count; + if (heap->spans_reserved) { + _rpmalloc_span_mark_as_subspan_unless_master(heap->span_reserve_master, heap->span_reserve, heap->spans_reserved); + _rpmalloc_heap_cache_insert(heap, heap->span_reserve); + } + if (reserved_count > _memory_heap_reserve_count) { + // If huge pages or eager spam map count, the global reserve spin lock is held by caller, _rpmalloc_span_map + rpmalloc_assert(atomic_load32(&_memory_global_lock) == 1, "Global spin lock not held as expected"); + size_t remain_count = reserved_count - _memory_heap_reserve_count; + reserved_count = _memory_heap_reserve_count; + span_t* remain_span = (span_t*)pointer_offset(reserved_spans, reserved_count * _memory_span_size); + if (_memory_global_reserve) { + _rpmalloc_span_mark_as_subspan_unless_master(_memory_global_reserve_master, _memory_global_reserve, _memory_global_reserve_count); + _rpmalloc_span_unmap(_memory_global_reserve); + } + _rpmalloc_global_set_reserved_spans(span, remain_span, remain_count); + } + _rpmalloc_heap_set_reserved_spans(heap, span, reserved_spans, reserved_count); + } + return span; +} + +//! Map in memory pages for the given number of spans (or use previously reserved pages) +static span_t* +_rpmalloc_span_map(heap_t* heap, size_t span_count) { + if (span_count <= heap->spans_reserved) + return _rpmalloc_span_map_from_reserve(heap, span_count); + span_t* span = 0; + int use_global_reserve = (_memory_page_size > _memory_span_size) || (_memory_span_map_count > _memory_heap_reserve_count); + if (use_global_reserve) { + // If huge pages, make sure only one thread maps more memory to avoid bloat + while (!atomic_cas32_acquire(&_memory_global_lock, 1, 0)) + _rpmalloc_spin(); + if (_memory_global_reserve_count >= span_count) { + size_t reserve_count = (!heap->spans_reserved ? _memory_heap_reserve_count : span_count); + if (_memory_global_reserve_count < reserve_count) + reserve_count = _memory_global_reserve_count; + span = _rpmalloc_global_get_reserved_spans(reserve_count); + if (span) { + if (reserve_count > span_count) { + span_t* reserved_span = (span_t*)pointer_offset(span, span_count << _memory_span_size_shift); + _rpmalloc_heap_set_reserved_spans(heap, _memory_global_reserve_master, reserved_span, reserve_count - span_count); + } + // Already marked as subspan in _rpmalloc_global_get_reserved_spans + span->span_count = (uint32_t)span_count; + } + } + } + if (!span) + span = _rpmalloc_span_map_aligned_count(heap, span_count); + if (use_global_reserve) + atomic_store32_release(&_memory_global_lock, 0); + return span; +} + +//! Unmap memory pages for the given number of spans (or mark as unused if no partial unmappings) +static void +_rpmalloc_span_unmap(span_t* span) { + rpmalloc_assert((span->flags & SPAN_FLAG_MASTER) || (span->flags & SPAN_FLAG_SUBSPAN), "Span flag corrupted"); + rpmalloc_assert(!(span->flags & SPAN_FLAG_MASTER) || !(span->flags & SPAN_FLAG_SUBSPAN), "Span flag corrupted"); + + int is_master = !!(span->flags & SPAN_FLAG_MASTER); + span_t* master = is_master ? span : ((span_t*)pointer_offset(span, -(intptr_t)((uintptr_t)span->offset_from_master * _memory_span_size))); + rpmalloc_assert(is_master || (span->flags & SPAN_FLAG_SUBSPAN), "Span flag corrupted"); + rpmalloc_assert(master->flags & SPAN_FLAG_MASTER, "Span flag corrupted"); + + size_t span_count = span->span_count; + if (!is_master) { + //Directly unmap subspans (unless huge pages, in which case we defer and unmap entire page range with master) + rpmalloc_assert(span->align_offset == 0, "Span align offset corrupted"); + if (_memory_span_size >= _memory_page_size) + _rpmalloc_unmap(span, span_count * _memory_span_size, 0, 0); + } else { + //Special double flag to denote an unmapped master + //It must be kept in memory since span header must be used + span->flags |= SPAN_FLAG_MASTER | SPAN_FLAG_SUBSPAN | SPAN_FLAG_UNMAPPED_MASTER; + _rpmalloc_stat_add(&_unmapped_master_spans, 1); + } + + if (atomic_add32(&master->remaining_spans, -(int32_t)span_count) <= 0) { + //Everything unmapped, unmap the master span with release flag to unmap the entire range of the super span + rpmalloc_assert(!!(master->flags & SPAN_FLAG_MASTER) && !!(master->flags & SPAN_FLAG_SUBSPAN), "Span flag corrupted"); + size_t unmap_count = master->span_count; + if (_memory_span_size < _memory_page_size) + unmap_count = master->total_spans; + _rpmalloc_stat_sub(&_master_spans, 1); + _rpmalloc_stat_sub(&_unmapped_master_spans, 1); + _rpmalloc_unmap(master, unmap_count * _memory_span_size, master->align_offset, (size_t)master->total_spans * _memory_span_size); + } +} + +//! Move the span (used for small or medium allocations) to the heap thread cache +static void +_rpmalloc_span_release_to_cache(heap_t* heap, span_t* span) { + rpmalloc_assert(heap == span->heap, "Span heap pointer corrupted"); + rpmalloc_assert(span->size_class < SIZE_CLASS_COUNT, "Invalid span size class"); + rpmalloc_assert(span->span_count == 1, "Invalid span count"); +#if ENABLE_ADAPTIVE_THREAD_CACHE || ENABLE_STATISTICS + atomic_decr32(&heap->span_use[0].current); +#endif + _rpmalloc_stat_dec(&heap->size_class_use[span->size_class].spans_current); + if (!heap->finalize) { + _rpmalloc_stat_inc(&heap->span_use[0].spans_to_cache); + _rpmalloc_stat_inc(&heap->size_class_use[span->size_class].spans_to_cache); + if (heap->size_class[span->size_class].cache) + _rpmalloc_heap_cache_insert(heap, heap->size_class[span->size_class].cache); + heap->size_class[span->size_class].cache = span; + } else { + _rpmalloc_span_unmap(span); + } +} + +//! Initialize a (partial) free list up to next system memory page, while reserving the first block +//! as allocated, returning number of blocks in list +static uint32_t +free_list_partial_init(void** list, void** first_block, void* page_start, void* block_start, uint32_t block_count, uint32_t block_size) { + rpmalloc_assert(block_count, "Internal failure"); + *first_block = block_start; + if (block_count > 1) { + void* free_block = pointer_offset(block_start, block_size); + void* block_end = pointer_offset(block_start, (size_t)block_size * block_count); + //If block size is less than half a memory page, bound init to next memory page boundary + if (block_size < (_memory_page_size >> 1)) { + void* page_end = pointer_offset(page_start, _memory_page_size); + if (page_end < block_end) + block_end = page_end; + } + *list = free_block; + block_count = 2; + void* next_block = pointer_offset(free_block, block_size); + while (next_block < block_end) { + *((void**)free_block) = next_block; + free_block = next_block; + ++block_count; + next_block = pointer_offset(next_block, block_size); + } + *((void**)free_block) = 0; + } else { + *list = 0; + } + return block_count; +} + +//! Initialize an unused span (from cache or mapped) to be new active span, putting the initial free list in heap class free list +static void* +_rpmalloc_span_initialize_new(heap_t* heap, heap_size_class_t* heap_size_class, span_t* span, uint32_t class_idx) { + rpmalloc_assert(span->span_count == 1, "Internal failure"); + size_class_t* size_class = _memory_size_class + class_idx; + span->size_class = class_idx; + span->heap = heap; + span->flags &= ~SPAN_FLAG_ALIGNED_BLOCKS; + span->block_size = size_class->block_size; + span->block_count = size_class->block_count; + span->free_list = 0; + span->list_size = 0; + atomic_store_ptr_release(&span->free_list_deferred, 0); + + //Setup free list. Only initialize one system page worth of free blocks in list + void* block; + span->free_list_limit = free_list_partial_init(&heap_size_class->free_list, &block, + span, pointer_offset(span, SPAN_HEADER_SIZE), size_class->block_count, size_class->block_size); + //Link span as partial if there remains blocks to be initialized as free list, or full if fully initialized + if (span->free_list_limit < span->block_count) { + _rpmalloc_span_double_link_list_add(&heap_size_class->partial_span, span); + span->used_count = span->free_list_limit; + } else { +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_add(&heap->full_span[class_idx], span); +#endif + ++heap->full_span_count; + span->used_count = span->block_count; + } + return block; +} + +static void +_rpmalloc_span_extract_free_list_deferred(span_t* span) { + // We need acquire semantics on the CAS operation since we are interested in the list size + // Refer to _rpmalloc_deallocate_defer_small_or_medium for further comments on this dependency + do { + span->free_list = atomic_exchange_ptr_acquire(&span->free_list_deferred, INVALID_POINTER); + } while (span->free_list == INVALID_POINTER); + span->used_count -= span->list_size; + span->list_size = 0; + atomic_store_ptr_release(&span->free_list_deferred, 0); +} + +static int +_rpmalloc_span_is_fully_utilized(span_t* span) { + rpmalloc_assert(span->free_list_limit <= span->block_count, "Span free list corrupted"); + return !span->free_list && (span->free_list_limit >= span->block_count); +} + +static int +_rpmalloc_span_finalize(heap_t* heap, size_t iclass, span_t* span, span_t** list_head) { + void* free_list = heap->size_class[iclass].free_list; + span_t* class_span = (span_t*)((uintptr_t)free_list & _memory_span_mask); + if (span == class_span) { + // Adopt the heap class free list back into the span free list + void* block = span->free_list; + void* last_block = 0; + while (block) { + last_block = block; + block = *((void**)block); + } + uint32_t free_count = 0; + block = free_list; + while (block) { + ++free_count; + block = *((void**)block); + } + if (last_block) { + *((void**)last_block) = free_list; + } else { + span->free_list = free_list; + } + heap->size_class[iclass].free_list = 0; + span->used_count -= free_count; + } + //If this assert triggers you have memory leaks + rpmalloc_assert(span->list_size == span->used_count, "Memory leak detected"); + if (span->list_size == span->used_count) { + _rpmalloc_stat_dec(&heap->span_use[0].current); + _rpmalloc_stat_dec(&heap->size_class_use[iclass].spans_current); + // This function only used for spans in double linked lists + if (list_head) + _rpmalloc_span_double_link_list_remove(list_head, span); + _rpmalloc_span_unmap(span); + return 1; + } + return 0; +} + + +//////////// +/// +/// Global cache +/// +////// + +#if ENABLE_GLOBAL_CACHE + +//! Finalize a global cache +static void +_rpmalloc_global_cache_finalize(global_cache_t* cache) { + while (!atomic_cas32_acquire(&cache->lock, 1, 0)) + _rpmalloc_spin(); + + for (size_t ispan = 0; ispan < cache->count; ++ispan) + _rpmalloc_span_unmap(cache->span[ispan]); + cache->count = 0; + + while (cache->overflow) { + span_t* span = cache->overflow; + cache->overflow = span->next; + _rpmalloc_span_unmap(span); + } + + atomic_store32_release(&cache->lock, 0); +} + +static void +_rpmalloc_global_cache_insert_spans(span_t** span, size_t span_count, size_t count) { + const size_t cache_limit = (span_count == 1) ? + GLOBAL_CACHE_MULTIPLIER * MAX_THREAD_SPAN_CACHE : + GLOBAL_CACHE_MULTIPLIER * (MAX_THREAD_SPAN_LARGE_CACHE - (span_count >> 1)); + + global_cache_t* cache = &_memory_span_cache[span_count - 1]; + + size_t insert_count = count; + while (!atomic_cas32_acquire(&cache->lock, 1, 0)) + _rpmalloc_spin(); + +#if ENABLE_STATISTICS + cache->insert_count += count; +#endif + if ((cache->count + insert_count) > cache_limit) + insert_count = cache_limit - cache->count; + + memcpy(cache->span + cache->count, span, sizeof(span_t*) * insert_count); + cache->count += (uint32_t)insert_count; + +#if ENABLE_UNLIMITED_CACHE + while (insert_count < count) { +#else + // Enable unlimited cache if huge pages, or we will leak since it is unlikely that an entire huge page + // will be unmapped, and we're unable to partially decommit a huge page + while ((_memory_page_size > _memory_span_size) && (insert_count < count)) { +#endif + span_t* current_span = span[insert_count++]; + current_span->next = cache->overflow; + cache->overflow = current_span; + } + atomic_store32_release(&cache->lock, 0); + + span_t* keep = 0; + for (size_t ispan = insert_count; ispan < count; ++ispan) { + span_t* current_span = span[ispan]; + // Keep master spans that has remaining subspans to avoid dangling them + if ((current_span->flags & SPAN_FLAG_MASTER) && + (atomic_load32(¤t_span->remaining_spans) > (int32_t)current_span->span_count)) { + current_span->next = keep; + keep = current_span; + } else { + _rpmalloc_span_unmap(current_span); + } + } + + if (keep) { + while (!atomic_cas32_acquire(&cache->lock, 1, 0)) + _rpmalloc_spin(); + + size_t islot = 0; + while (keep) { + for (; islot < cache->count; ++islot) { + span_t* current_span = cache->span[islot]; + if (!(current_span->flags & SPAN_FLAG_MASTER) || ((current_span->flags & SPAN_FLAG_MASTER) && + (atomic_load32(¤t_span->remaining_spans) <= (int32_t)current_span->span_count))) { + _rpmalloc_span_unmap(current_span); + cache->span[islot] = keep; + break; + } + } + if (islot == cache->count) + break; + keep = keep->next; + } + + if (keep) { + span_t* tail = keep; + while (tail->next) + tail = tail->next; + tail->next = cache->overflow; + cache->overflow = keep; + } + + atomic_store32_release(&cache->lock, 0); + } +} + +static size_t +_rpmalloc_global_cache_extract_spans(span_t** span, size_t span_count, size_t count) { + global_cache_t* cache = &_memory_span_cache[span_count - 1]; + + size_t extract_count = 0; + while (!atomic_cas32_acquire(&cache->lock, 1, 0)) + _rpmalloc_spin(); + +#if ENABLE_STATISTICS + cache->extract_count += count; +#endif + size_t want = count - extract_count; + if (want > cache->count) + want = cache->count; + + memcpy(span + extract_count, cache->span + (cache->count - want), sizeof(span_t*) * want); + cache->count -= (uint32_t)want; + extract_count += want; + + while ((extract_count < count) && cache->overflow) { + span_t* current_span = cache->overflow; + span[extract_count++] = current_span; + cache->overflow = current_span->next; + } + +#if ENABLE_ASSERTS + for (size_t ispan = 0; ispan < extract_count; ++ispan) { + assert(span[ispan]->span_count == span_count); + } +#endif + + atomic_store32_release(&cache->lock, 0); + + return extract_count; +} + +#endif + +//////////// +/// +/// Heap control +/// +////// + +static void _rpmalloc_deallocate_huge(span_t*); + +//! Store the given spans as reserve in the given heap +static void +_rpmalloc_heap_set_reserved_spans(heap_t* heap, span_t* master, span_t* reserve, size_t reserve_span_count) { + heap->span_reserve_master = master; + heap->span_reserve = reserve; + heap->spans_reserved = (uint32_t)reserve_span_count; +} + +//! Adopt the deferred span cache list, optionally extracting the first single span for immediate re-use +static void +_rpmalloc_heap_cache_adopt_deferred(heap_t* heap, span_t** single_span) { + span_t* span = (span_t*)((void*)atomic_exchange_ptr_acquire(&heap->span_free_deferred, 0)); + while (span) { + span_t* next_span = (span_t*)span->free_list; + rpmalloc_assert(span->heap == heap, "Span heap pointer corrupted"); + if (EXPECTED(span->size_class < SIZE_CLASS_COUNT)) { + rpmalloc_assert(heap->full_span_count, "Heap span counter corrupted"); + --heap->full_span_count; + _rpmalloc_stat_dec(&heap->span_use[0].spans_deferred); +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_remove(&heap->full_span[span->size_class], span); +#endif + _rpmalloc_stat_dec(&heap->span_use[0].current); + _rpmalloc_stat_dec(&heap->size_class_use[span->size_class].spans_current); + if (single_span && !*single_span) + *single_span = span; + else + _rpmalloc_heap_cache_insert(heap, span); + } else { + if (span->size_class == SIZE_CLASS_HUGE) { + _rpmalloc_deallocate_huge(span); + } else { + rpmalloc_assert(span->size_class == SIZE_CLASS_LARGE, "Span size class invalid"); + rpmalloc_assert(heap->full_span_count, "Heap span counter corrupted"); + --heap->full_span_count; +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_remove(&heap->large_huge_span, span); +#endif + uint32_t idx = span->span_count - 1; + _rpmalloc_stat_dec(&heap->span_use[idx].spans_deferred); + _rpmalloc_stat_dec(&heap->span_use[idx].current); + if (!idx && single_span && !*single_span) + *single_span = span; + else + _rpmalloc_heap_cache_insert(heap, span); + } + } + span = next_span; + } +} + +static void +_rpmalloc_heap_unmap(heap_t* heap) { + if (!heap->master_heap) { + if ((heap->finalize > 1) && !atomic_load32(&heap->child_count)) { + span_t* span = (span_t*)((uintptr_t)heap & _memory_span_mask); + _rpmalloc_span_unmap(span); + } + } else { + if (atomic_decr32(&heap->master_heap->child_count) == 0) { + _rpmalloc_heap_unmap(heap->master_heap); + } + } +} + +static void +_rpmalloc_heap_global_finalize(heap_t* heap) { + if (heap->finalize++ > 1) { + --heap->finalize; + return; + } + + _rpmalloc_heap_finalize(heap); + +#if ENABLE_THREAD_CACHE + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) { + span_cache_t* span_cache; + if (!iclass) + span_cache = &heap->span_cache; + else + span_cache = (span_cache_t*)(heap->span_large_cache + (iclass - 1)); + for (size_t ispan = 0; ispan < span_cache->count; ++ispan) + _rpmalloc_span_unmap(span_cache->span[ispan]); + span_cache->count = 0; + } +#endif + + if (heap->full_span_count) { + --heap->finalize; + return; + } + + for (size_t iclass = 0; iclass < SIZE_CLASS_COUNT; ++iclass) { + if (heap->size_class[iclass].free_list || heap->size_class[iclass].partial_span) { + --heap->finalize; + return; + } + } + //Heap is now completely free, unmap and remove from heap list + size_t list_idx = (size_t)heap->id % HEAP_ARRAY_SIZE; + heap_t* list_heap = _memory_heaps[list_idx]; + if (list_heap == heap) { + _memory_heaps[list_idx] = heap->next_heap; + } else { + while (list_heap->next_heap != heap) + list_heap = list_heap->next_heap; + list_heap->next_heap = heap->next_heap; + } + + _rpmalloc_heap_unmap(heap); +} + +//! Insert a single span into thread heap cache, releasing to global cache if overflow +static void +_rpmalloc_heap_cache_insert(heap_t* heap, span_t* span) { + if (UNEXPECTED(heap->finalize != 0)) { + _rpmalloc_span_unmap(span); + _rpmalloc_heap_global_finalize(heap); + return; + } +#if ENABLE_THREAD_CACHE + size_t span_count = span->span_count; + _rpmalloc_stat_inc(&heap->span_use[span_count - 1].spans_to_cache); + if (span_count == 1) { + span_cache_t* span_cache = &heap->span_cache; + span_cache->span[span_cache->count++] = span; + if (span_cache->count == MAX_THREAD_SPAN_CACHE) { + const size_t remain_count = MAX_THREAD_SPAN_CACHE - THREAD_SPAN_CACHE_TRANSFER; +#if ENABLE_GLOBAL_CACHE + _rpmalloc_stat_add64(&heap->thread_to_global, THREAD_SPAN_CACHE_TRANSFER * _memory_span_size); + _rpmalloc_stat_add(&heap->span_use[span_count - 1].spans_to_global, THREAD_SPAN_CACHE_TRANSFER); + _rpmalloc_global_cache_insert_spans(span_cache->span + remain_count, span_count, THREAD_SPAN_CACHE_TRANSFER); +#else + for (size_t ispan = 0; ispan < THREAD_SPAN_CACHE_TRANSFER; ++ispan) + _rpmalloc_span_unmap(span_cache->span[remain_count + ispan]); +#endif + span_cache->count = remain_count; + } + } else { + size_t cache_idx = span_count - 2; + span_large_cache_t* span_cache = heap->span_large_cache + cache_idx; + span_cache->span[span_cache->count++] = span; + const size_t cache_limit = (MAX_THREAD_SPAN_LARGE_CACHE - (span_count >> 1)); + if (span_cache->count == cache_limit) { + const size_t transfer_limit = 2 + (cache_limit >> 2); + const size_t transfer_count = (THREAD_SPAN_LARGE_CACHE_TRANSFER <= transfer_limit ? THREAD_SPAN_LARGE_CACHE_TRANSFER : transfer_limit); + const size_t remain_count = cache_limit - transfer_count; +#if ENABLE_GLOBAL_CACHE + _rpmalloc_stat_add64(&heap->thread_to_global, transfer_count * span_count * _memory_span_size); + _rpmalloc_stat_add(&heap->span_use[span_count - 1].spans_to_global, transfer_count); + _rpmalloc_global_cache_insert_spans(span_cache->span + remain_count, span_count, transfer_count); +#else + for (size_t ispan = 0; ispan < transfer_count; ++ispan) + _rpmalloc_span_unmap(span_cache->span[remain_count + ispan]); +#endif + span_cache->count = remain_count; + } + } +#else + (void)sizeof(heap); + _rpmalloc_span_unmap(span); +#endif +} + +//! Extract the given number of spans from the different cache levels +static span_t* +_rpmalloc_heap_thread_cache_extract(heap_t* heap, size_t span_count) { + span_t* span = 0; +#if ENABLE_THREAD_CACHE + span_cache_t* span_cache; + if (span_count == 1) + span_cache = &heap->span_cache; + else + span_cache = (span_cache_t*)(heap->span_large_cache + (span_count - 2)); + if (span_cache->count) { + _rpmalloc_stat_inc(&heap->span_use[span_count - 1].spans_from_cache); + return span_cache->span[--span_cache->count]; + } +#endif + return span; +} + +static span_t* +_rpmalloc_heap_thread_cache_deferred_extract(heap_t* heap, size_t span_count) { + span_t* span = 0; + if (span_count == 1) { + _rpmalloc_heap_cache_adopt_deferred(heap, &span); + } else { + _rpmalloc_heap_cache_adopt_deferred(heap, 0); + span = _rpmalloc_heap_thread_cache_extract(heap, span_count); + } + return span; +} + +static span_t* +_rpmalloc_heap_reserved_extract(heap_t* heap, size_t span_count) { + if (heap->spans_reserved >= span_count) + return _rpmalloc_span_map(heap, span_count); + return 0; +} + +//! Extract a span from the global cache +static span_t* +_rpmalloc_heap_global_cache_extract(heap_t* heap, size_t span_count) { +#if ENABLE_GLOBAL_CACHE +#if ENABLE_THREAD_CACHE + span_cache_t* span_cache; + size_t wanted_count; + if (span_count == 1) { + span_cache = &heap->span_cache; + wanted_count = THREAD_SPAN_CACHE_TRANSFER; + } else { + span_cache = (span_cache_t*)(heap->span_large_cache + (span_count - 2)); + wanted_count = THREAD_SPAN_LARGE_CACHE_TRANSFER; + } + span_cache->count = _rpmalloc_global_cache_extract_spans(span_cache->span, span_count, wanted_count); + if (span_cache->count) { + _rpmalloc_stat_add64(&heap->global_to_thread, span_count * span_cache->count * _memory_span_size); + _rpmalloc_stat_add(&heap->span_use[span_count - 1].spans_from_global, span_cache->count); + return span_cache->span[--span_cache->count]; + } +#else + span_t* span = 0; + size_t count = _rpmalloc_global_cache_extract_spans(&span, span_count, 1); + if (count) { + _rpmalloc_stat_add64(&heap->global_to_thread, span_count * count * _memory_span_size); + _rpmalloc_stat_add(&heap->span_use[span_count - 1].spans_from_global, count); + return span; + } +#endif +#endif + (void)sizeof(heap); + (void)sizeof(span_count); + return 0; +} + +static void +_rpmalloc_inc_span_statistics(heap_t* heap, size_t span_count, uint32_t class_idx) { + (void)sizeof(heap); + (void)sizeof(span_count); + (void)sizeof(class_idx); +#if ENABLE_ADAPTIVE_THREAD_CACHE || ENABLE_STATISTICS + uint32_t idx = (uint32_t)span_count - 1; + uint32_t current_count = (uint32_t)atomic_incr32(&heap->span_use[idx].current); + if (current_count > (uint32_t)atomic_load32(&heap->span_use[idx].high)) + atomic_store32(&heap->span_use[idx].high, (int32_t)current_count); + _rpmalloc_stat_add_peak(&heap->size_class_use[class_idx].spans_current, 1, heap->size_class_use[class_idx].spans_peak); +#endif +} + +//! Get a span from one of the cache levels (thread cache, reserved, global cache) or fallback to mapping more memory +static span_t* +_rpmalloc_heap_extract_new_span(heap_t* heap, heap_size_class_t* heap_size_class, size_t span_count, uint32_t class_idx) { + span_t* span; +#if ENABLE_THREAD_CACHE + if (heap_size_class && heap_size_class->cache) { + span = heap_size_class->cache; + heap_size_class->cache = (heap->span_cache.count ? heap->span_cache.span[--heap->span_cache.count] : 0); + _rpmalloc_inc_span_statistics(heap, span_count, class_idx); + return span; + } +#endif + (void)sizeof(class_idx); + // Allow 50% overhead to increase cache hits + size_t base_span_count = span_count; + size_t limit_span_count = (span_count > 2) ? (span_count + (span_count >> 1)) : span_count; + if (limit_span_count > LARGE_CLASS_COUNT) + limit_span_count = LARGE_CLASS_COUNT; + do { + span = _rpmalloc_heap_thread_cache_extract(heap, span_count); + if (EXPECTED(span != 0)) { + _rpmalloc_stat_inc(&heap->size_class_use[class_idx].spans_from_cache); + _rpmalloc_inc_span_statistics(heap, span_count, class_idx); + return span; + } + span = _rpmalloc_heap_thread_cache_deferred_extract(heap, span_count); + if (EXPECTED(span != 0)) { + _rpmalloc_stat_inc(&heap->size_class_use[class_idx].spans_from_cache); + _rpmalloc_inc_span_statistics(heap, span_count, class_idx); + return span; + } + span = _rpmalloc_heap_reserved_extract(heap, span_count); + if (EXPECTED(span != 0)) { + _rpmalloc_stat_inc(&heap->size_class_use[class_idx].spans_from_reserved); + _rpmalloc_inc_span_statistics(heap, span_count, class_idx); + return span; + } + span = _rpmalloc_heap_global_cache_extract(heap, span_count); + if (EXPECTED(span != 0)) { + _rpmalloc_stat_inc(&heap->size_class_use[class_idx].spans_from_cache); + _rpmalloc_inc_span_statistics(heap, span_count, class_idx); + return span; + } + ++span_count; + } while (span_count <= limit_span_count); + //Final fallback, map in more virtual memory + span = _rpmalloc_span_map(heap, base_span_count); + _rpmalloc_inc_span_statistics(heap, base_span_count, class_idx); + _rpmalloc_stat_inc(&heap->size_class_use[class_idx].spans_map_calls); + return span; +} + +static void +_rpmalloc_heap_initialize(heap_t* heap) { + memset((void*)heap, 0, sizeof(heap_t)); + //Get a new heap ID + heap->id = 1 + atomic_incr32(&_memory_heap_id); + + //Link in heap in heap ID map + size_t list_idx = (size_t)heap->id % HEAP_ARRAY_SIZE; + heap->next_heap = _memory_heaps[list_idx]; + _memory_heaps[list_idx] = heap; +} + +static void +_rpmalloc_heap_orphan(heap_t* heap, int first_class) { + heap->owner_thread = (uintptr_t)-1; +#if RPMALLOC_FIRST_CLASS_HEAPS + heap_t** heap_list = (first_class ? &_memory_first_class_orphan_heaps : &_memory_orphan_heaps); +#else + (void)sizeof(first_class); + heap_t** heap_list = &_memory_orphan_heaps; +#endif + heap->next_orphan = *heap_list; + *heap_list = heap; +} + +//! Allocate a new heap from newly mapped memory pages +static heap_t* +_rpmalloc_heap_allocate_new(void) { + // Map in pages for a 16 heaps. If page size is greater than required size for this, map a page and + // use first part for heaps and remaining part for spans for allocations. Adds a lot of complexity, + // but saves a lot of memory on systems where page size > 64 spans (4MiB) + size_t heap_size = sizeof(heap_t); + size_t aligned_heap_size = 16 * ((heap_size + 15) / 16); + size_t request_heap_count = 16; + size_t heap_span_count = ((aligned_heap_size * request_heap_count) + sizeof(span_t) + _memory_span_size - 1) / _memory_span_size; + size_t block_size = _memory_span_size * heap_span_count; + size_t span_count = heap_span_count; + span_t* span = 0; + // If there are global reserved spans, use these first + if (_memory_global_reserve_count >= heap_span_count) { + span = _rpmalloc_global_get_reserved_spans(heap_span_count); + } + if (!span) { + if (_memory_page_size > block_size) { + span_count = _memory_page_size / _memory_span_size; + block_size = _memory_page_size; + // If using huge pages, make sure to grab enough heaps to avoid reallocating a huge page just to serve new heaps + size_t possible_heap_count = (block_size - sizeof(span_t)) / aligned_heap_size; + if (possible_heap_count >= (request_heap_count * 16)) + request_heap_count *= 16; + else if (possible_heap_count < request_heap_count) + request_heap_count = possible_heap_count; + heap_span_count = ((aligned_heap_size * request_heap_count) + sizeof(span_t) + _memory_span_size - 1) / _memory_span_size; + } + + size_t align_offset = 0; + span = (span_t*)_rpmalloc_mmap(block_size, &align_offset); + if (!span) + return 0; + + // Master span will contain the heaps + _rpmalloc_stat_inc(&_master_spans); + _rpmalloc_span_initialize(span, span_count, heap_span_count, align_offset); + } + + size_t remain_size = _memory_span_size - sizeof(span_t); + heap_t* heap = (heap_t*)pointer_offset(span, sizeof(span_t)); + _rpmalloc_heap_initialize(heap); + + // Put extra heaps as orphans + size_t num_heaps = remain_size / aligned_heap_size; + if (num_heaps < request_heap_count) + num_heaps = request_heap_count; + atomic_store32(&heap->child_count, (int32_t)num_heaps - 1); + heap_t* extra_heap = (heap_t*)pointer_offset(heap, aligned_heap_size); + while (num_heaps > 1) { + _rpmalloc_heap_initialize(extra_heap); + extra_heap->master_heap = heap; + _rpmalloc_heap_orphan(extra_heap, 1); + extra_heap = (heap_t*)pointer_offset(extra_heap, aligned_heap_size); + --num_heaps; + } + + if (span_count > heap_span_count) { + // Cap reserved spans + size_t remain_count = span_count - heap_span_count; + size_t reserve_count = (remain_count > _memory_heap_reserve_count ? _memory_heap_reserve_count : remain_count); + span_t* remain_span = (span_t*)pointer_offset(span, heap_span_count * _memory_span_size); + _rpmalloc_heap_set_reserved_spans(heap, span, remain_span, reserve_count); + + if (remain_count > reserve_count) { + // Set to global reserved spans + remain_span = (span_t*)pointer_offset(remain_span, reserve_count * _memory_span_size); + reserve_count = remain_count - reserve_count; + _rpmalloc_global_set_reserved_spans(span, remain_span, reserve_count); + } + } + + return heap; +} + +static heap_t* +_rpmalloc_heap_extract_orphan(heap_t** heap_list) { + heap_t* heap = *heap_list; + *heap_list = (heap ? heap->next_orphan : 0); + return heap; +} + +//! Allocate a new heap, potentially reusing a previously orphaned heap +static heap_t* +_rpmalloc_heap_allocate(int first_class) { + heap_t* heap = 0; + while (!atomic_cas32_acquire(&_memory_global_lock, 1, 0)) + _rpmalloc_spin(); + if (first_class == 0) + heap = _rpmalloc_heap_extract_orphan(&_memory_orphan_heaps); +#if RPMALLOC_FIRST_CLASS_HEAPS + if (!heap) + heap = _rpmalloc_heap_extract_orphan(&_memory_first_class_orphan_heaps); +#endif + if (!heap) + heap = _rpmalloc_heap_allocate_new(); + atomic_store32_release(&_memory_global_lock, 0); + _rpmalloc_heap_cache_adopt_deferred(heap, 0); + return heap; +} + +extern thread_local bool RpThreadShutdown; + +static void +_rpmalloc_heap_release(void* heapptr, int first_class, int release_cache) { + heap_t* heap = (heap_t*)heapptr; + if (!heap) + return; + RpThreadShutdown = true; + //Release thread cache spans back to global cache + _rpmalloc_heap_cache_adopt_deferred(heap, 0); + if (release_cache || heap->finalize) { +#if ENABLE_THREAD_CACHE + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) { + span_cache_t* span_cache; + if (!iclass) + span_cache = &heap->span_cache; + else + span_cache = (span_cache_t*)(heap->span_large_cache + (iclass - 1)); + if (!span_cache->count) + continue; +#if ENABLE_GLOBAL_CACHE + if (heap->finalize) { + for (size_t ispan = 0; ispan < span_cache->count; ++ispan) + _rpmalloc_span_unmap(span_cache->span[ispan]); + } else { + _rpmalloc_stat_add64(&heap->thread_to_global, span_cache->count * (iclass + 1) * _memory_span_size); + _rpmalloc_stat_add(&heap->span_use[iclass].spans_to_global, span_cache->count); + _rpmalloc_global_cache_insert_spans(span_cache->span, iclass + 1, span_cache->count); + } +#else + for (size_t ispan = 0; ispan < span_cache->count; ++ispan) + _rpmalloc_span_unmap(span_cache->span[ispan]); +#endif + span_cache->count = 0; + } +#endif + } + + if (get_thread_heap_raw() == heap) + set_thread_heap(0); + +#if ENABLE_STATISTICS + atomic_decr32(&_memory_active_heaps); + rpmalloc_assert(atomic_load32(&_memory_active_heaps) >= 0, "Still active heaps during finalization"); +#endif + + // If we are forcibly terminating with _exit the state of the + // lock atomic is unknown and it's best to just go ahead and exit + if (get_thread_id() != _rpmalloc_main_thread_id) { + while (!atomic_cas32_acquire(&_memory_global_lock, 1, 0)) + _rpmalloc_spin(); + } + _rpmalloc_heap_orphan(heap, first_class); + atomic_store32_release(&_memory_global_lock, 0); +} + +static void +_rpmalloc_heap_release_raw(void* heapptr, int release_cache) { + _rpmalloc_heap_release(heapptr, 0, release_cache); +} + +static void +_rpmalloc_heap_release_raw_fc(void* heapptr) { + _rpmalloc_heap_release_raw(heapptr, 1); +} + +static void +_rpmalloc_heap_finalize(heap_t* heap) { + if (heap->spans_reserved) { + span_t* span = _rpmalloc_span_map(heap, heap->spans_reserved); + _rpmalloc_span_unmap(span); + heap->spans_reserved = 0; + } + + _rpmalloc_heap_cache_adopt_deferred(heap, 0); + + for (size_t iclass = 0; iclass < SIZE_CLASS_COUNT; ++iclass) { + if (heap->size_class[iclass].cache) + _rpmalloc_span_unmap(heap->size_class[iclass].cache); + heap->size_class[iclass].cache = 0; + span_t* span = heap->size_class[iclass].partial_span; + while (span) { + span_t* next = span->next; + _rpmalloc_span_finalize(heap, iclass, span, &heap->size_class[iclass].partial_span); + span = next; + } + // If class still has a free list it must be a full span + if (heap->size_class[iclass].free_list) { + span_t* class_span = (span_t*)((uintptr_t)heap->size_class[iclass].free_list & _memory_span_mask); + span_t** list = 0; +#if RPMALLOC_FIRST_CLASS_HEAPS + list = &heap->full_span[iclass]; +#endif + --heap->full_span_count; + if (!_rpmalloc_span_finalize(heap, iclass, class_span, list)) { + if (list) + _rpmalloc_span_double_link_list_remove(list, class_span); + _rpmalloc_span_double_link_list_add(&heap->size_class[iclass].partial_span, class_span); + } + } + } + +#if ENABLE_THREAD_CACHE + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) { + span_cache_t* span_cache; + if (!iclass) + span_cache = &heap->span_cache; + else + span_cache = (span_cache_t*)(heap->span_large_cache + (iclass - 1)); + for (size_t ispan = 0; ispan < span_cache->count; ++ispan) + _rpmalloc_span_unmap(span_cache->span[ispan]); + span_cache->count = 0; + } +#endif + rpmalloc_assert(!atomic_load_ptr(&heap->span_free_deferred), "Heaps still active during finalization"); +} + + +//////////// +/// +/// Allocation entry points +/// +////// + +//! Pop first block from a free list +static void* +free_list_pop(void** list) { + void* block = *list; + *list = *((void**)block); + return block; +} + +//! Allocate a small/medium sized memory block from the given heap +static void* +_rpmalloc_allocate_from_heap_fallback(heap_t* heap, heap_size_class_t* heap_size_class, uint32_t class_idx) { + span_t* span = heap_size_class->partial_span; + if (EXPECTED(span != 0)) { + rpmalloc_assert(span->block_count == _memory_size_class[span->size_class].block_count, "Span block count corrupted"); + rpmalloc_assert(!_rpmalloc_span_is_fully_utilized(span), "Internal failure"); + void* block; + if (span->free_list) { + //Span local free list is not empty, swap to size class free list + block = free_list_pop(&span->free_list); + heap_size_class->free_list = span->free_list; + span->free_list = 0; + } else { + //If the span did not fully initialize free list, link up another page worth of blocks + void* block_start = pointer_offset(span, SPAN_HEADER_SIZE + ((size_t)span->free_list_limit * span->block_size)); + span->free_list_limit += free_list_partial_init(&heap_size_class->free_list, &block, + (void*)((uintptr_t)block_start & ~(_memory_page_size - 1)), block_start, + span->block_count - span->free_list_limit, span->block_size); + } + rpmalloc_assert(span->free_list_limit <= span->block_count, "Span block count corrupted"); + span->used_count = span->free_list_limit; + + //Swap in deferred free list if present + if (atomic_load_ptr(&span->free_list_deferred)) + _rpmalloc_span_extract_free_list_deferred(span); + + //If span is still not fully utilized keep it in partial list and early return block + if (!_rpmalloc_span_is_fully_utilized(span)) + return block; + + //The span is fully utilized, unlink from partial list and add to fully utilized list + _rpmalloc_span_double_link_list_pop_head(&heap_size_class->partial_span, span); +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_add(&heap->full_span[class_idx], span); +#endif + ++heap->full_span_count; + return block; + } + + //Find a span in one of the cache levels + span = _rpmalloc_heap_extract_new_span(heap, heap_size_class, 1, class_idx); + if (EXPECTED(span != 0)) { + //Mark span as owned by this heap and set base data, return first block + return _rpmalloc_span_initialize_new(heap, heap_size_class, span, class_idx); + } + + return 0; +} + +//! Allocate a small sized memory block from the given heap +static void* +_rpmalloc_allocate_small(heap_t* heap, size_t size) { + rpmalloc_assert(heap, "No thread heap"); + //Small sizes have unique size classes + const uint32_t class_idx = (uint32_t)((size + (SMALL_GRANULARITY - 1)) >> SMALL_GRANULARITY_SHIFT); + heap_size_class_t* heap_size_class = heap->size_class + class_idx; + _rpmalloc_stat_inc_alloc(heap, class_idx); + if (EXPECTED(heap_size_class->free_list != 0)) + return free_list_pop(&heap_size_class->free_list); + return _rpmalloc_allocate_from_heap_fallback(heap, heap_size_class, class_idx); +} + +//! Allocate a medium sized memory block from the given heap +static void* +_rpmalloc_allocate_medium(heap_t* heap, size_t size) { + rpmalloc_assert(heap, "No thread heap"); + //Calculate the size class index and do a dependent lookup of the final class index (in case of merged classes) + const uint32_t base_idx = (uint32_t)(SMALL_CLASS_COUNT + ((size - (SMALL_SIZE_LIMIT + 1)) >> MEDIUM_GRANULARITY_SHIFT)); + const uint32_t class_idx = _memory_size_class[base_idx].class_idx; + heap_size_class_t* heap_size_class = heap->size_class + class_idx; + _rpmalloc_stat_inc_alloc(heap, class_idx); + if (EXPECTED(heap_size_class->free_list != 0)) + return free_list_pop(&heap_size_class->free_list); + return _rpmalloc_allocate_from_heap_fallback(heap, heap_size_class, class_idx); +} + +//! Allocate a large sized memory block from the given heap +static void* +_rpmalloc_allocate_large(heap_t* heap, size_t size) { + rpmalloc_assert(heap, "No thread heap"); + //Calculate number of needed max sized spans (including header) + //Since this function is never called if size > LARGE_SIZE_LIMIT + //the span_count is guaranteed to be <= LARGE_CLASS_COUNT + size += SPAN_HEADER_SIZE; + size_t span_count = size >> _memory_span_size_shift; + if (size & (_memory_span_size - 1)) + ++span_count; + + //Find a span in one of the cache levels + span_t* span = _rpmalloc_heap_extract_new_span(heap, 0, span_count, SIZE_CLASS_LARGE); + if (!span) + return span; + + //Mark span as owned by this heap and set base data + rpmalloc_assert(span->span_count >= span_count, "Internal failure"); + span->size_class = SIZE_CLASS_LARGE; + span->heap = heap; + +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_add(&heap->large_huge_span, span); +#endif + ++heap->full_span_count; + + return pointer_offset(span, SPAN_HEADER_SIZE); +} + +//! Allocate a huge block by mapping memory pages directly +static void* +_rpmalloc_allocate_huge(heap_t* heap, size_t size) { + rpmalloc_assert(heap, "No thread heap"); + _rpmalloc_heap_cache_adopt_deferred(heap, 0); + size += SPAN_HEADER_SIZE; + size_t num_pages = size >> _memory_page_size_shift; + if (size & (_memory_page_size - 1)) + ++num_pages; + size_t align_offset = 0; + span_t* span = (span_t*)_rpmalloc_mmap(num_pages * _memory_page_size, &align_offset); + if (!span) + return span; + + //Store page count in span_count + span->size_class = SIZE_CLASS_HUGE; + span->span_count = (uint32_t)num_pages; + span->align_offset = (uint32_t)align_offset; + span->heap = heap; + _rpmalloc_stat_add_peak(&_huge_pages_current, num_pages, _huge_pages_peak); + +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_add(&heap->large_huge_span, span); +#endif + ++heap->full_span_count; + + return pointer_offset(span, SPAN_HEADER_SIZE); +} + +//! Allocate a block of the given size +static void* +_rpmalloc_allocate(heap_t* heap, size_t size) { + _rpmalloc_stat_add64(&_allocation_counter, 1); + if (EXPECTED(size <= SMALL_SIZE_LIMIT)) + return _rpmalloc_allocate_small(heap, size); + else if (size <= _memory_medium_size_limit) + return _rpmalloc_allocate_medium(heap, size); + else if (size <= LARGE_SIZE_LIMIT) + return _rpmalloc_allocate_large(heap, size); + return _rpmalloc_allocate_huge(heap, size); +} + +static void* +_rpmalloc_aligned_allocate(heap_t* heap, size_t alignment, size_t size) { + if (alignment <= SMALL_GRANULARITY) + return _rpmalloc_allocate(heap, size); + +#if ENABLE_VALIDATE_ARGS + if ((size + alignment) < size) { + errno = EINVAL; + return 0; + } + if (alignment & (alignment - 1)) { + errno = EINVAL; + return 0; + } +#endif + + if ((alignment <= SPAN_HEADER_SIZE) && (size < _memory_medium_size_limit)) { + // If alignment is less or equal to span header size (which is power of two), + // and size aligned to span header size multiples is less than size + alignment, + // then use natural alignment of blocks to provide alignment + size_t multiple_size = size ? (size + (SPAN_HEADER_SIZE - 1)) & ~(uintptr_t)(SPAN_HEADER_SIZE - 1) : SPAN_HEADER_SIZE; + rpmalloc_assert(!(multiple_size % SPAN_HEADER_SIZE), "Failed alignment calculation"); + if (multiple_size <= (size + alignment)) + return _rpmalloc_allocate(heap, multiple_size); + } + + void* ptr = 0; + size_t align_mask = alignment - 1; + if (alignment <= _memory_page_size) { + ptr = _rpmalloc_allocate(heap, size + alignment); + if ((uintptr_t)ptr & align_mask) { + ptr = (void*)(((uintptr_t)ptr & ~(uintptr_t)align_mask) + alignment); + //Mark as having aligned blocks + span_t* span = (span_t*)((uintptr_t)ptr & _memory_span_mask); + span->flags |= SPAN_FLAG_ALIGNED_BLOCKS; + } + return ptr; + } + + // Fallback to mapping new pages for this request. Since pointers passed + // to rpfree must be able to reach the start of the span by bitmasking of + // the address with the span size, the returned aligned pointer from this + // function must be with a span size of the start of the mapped area. + // In worst case this requires us to loop and map pages until we get a + // suitable memory address. It also means we can never align to span size + // or greater, since the span header will push alignment more than one + // span size away from span start (thus causing pointer mask to give us + // an invalid span start on free) + if (alignment & align_mask) { + errno = EINVAL; + return 0; + } + if (alignment >= _memory_span_size) { + errno = EINVAL; + return 0; + } + + size_t extra_pages = alignment / _memory_page_size; + + // Since each span has a header, we will at least need one extra memory page + size_t num_pages = 1 + (size / _memory_page_size); + if (size & (_memory_page_size - 1)) + ++num_pages; + + if (extra_pages > num_pages) + num_pages = 1 + extra_pages; + + size_t original_pages = num_pages; + size_t limit_pages = (_memory_span_size / _memory_page_size) * 2; + if (limit_pages < (original_pages * 2)) + limit_pages = original_pages * 2; + + size_t mapped_size, align_offset; + span_t* span; + +retry: + align_offset = 0; + mapped_size = num_pages * _memory_page_size; + + span = (span_t*)_rpmalloc_mmap(mapped_size, &align_offset); + if (!span) { + errno = ENOMEM; + return 0; + } + ptr = pointer_offset(span, SPAN_HEADER_SIZE); + + if ((uintptr_t)ptr & align_mask) + ptr = (void*)(((uintptr_t)ptr & ~(uintptr_t)align_mask) + alignment); + + if (((size_t)pointer_diff(ptr, span) >= _memory_span_size) || + (pointer_offset(ptr, size) > pointer_offset(span, mapped_size)) || + (((uintptr_t)ptr & _memory_span_mask) != (uintptr_t)span)) { + _rpmalloc_unmap(span, mapped_size, align_offset, mapped_size); + ++num_pages; + if (num_pages > limit_pages) { + errno = EINVAL; + return 0; + } + goto retry; + } + + //Store page count in span_count + span->size_class = SIZE_CLASS_HUGE; + span->span_count = (uint32_t)num_pages; + span->align_offset = (uint32_t)align_offset; + span->heap = heap; + _rpmalloc_stat_add_peak(&_huge_pages_current, num_pages, _huge_pages_peak); + +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_add(&heap->large_huge_span, span); +#endif + ++heap->full_span_count; + + _rpmalloc_stat_add64(&_allocation_counter, 1); + + return ptr; +} + + +//////////// +/// +/// Deallocation entry points +/// +////// + +//! Deallocate the given small/medium memory block in the current thread local heap +static void +_rpmalloc_deallocate_direct_small_or_medium(span_t* span, void* block) { + heap_t* heap = span->heap; + rpmalloc_assert(heap->owner_thread == get_thread_id() || !heap->owner_thread || heap->finalize, "Internal failure"); + //Add block to free list + if (UNEXPECTED(_rpmalloc_span_is_fully_utilized(span))) { + span->used_count = span->block_count; +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_remove(&heap->full_span[span->size_class], span); +#endif + _rpmalloc_span_double_link_list_add(&heap->size_class[span->size_class].partial_span, span); + --heap->full_span_count; + } + *((void**)block) = span->free_list; + --span->used_count; + span->free_list = block; + if (UNEXPECTED(span->used_count == span->list_size)) { + // If there are no used blocks it is guaranteed that no other external thread is accessing the span + if (span->used_count) { + // Make sure we have synchronized the deferred list and list size by using acquire semantics + // and guarantee that no external thread is accessing span concurrently + void* free_list; + do { + free_list = atomic_exchange_ptr_acquire(&span->free_list_deferred, INVALID_POINTER); + } while (free_list == INVALID_POINTER); + atomic_store_ptr_release(&span->free_list_deferred, free_list); + } + _rpmalloc_span_double_link_list_remove(&heap->size_class[span->size_class].partial_span, span); + _rpmalloc_span_release_to_cache(heap, span); + } +} + +static void +_rpmalloc_deallocate_defer_free_span(heap_t* heap, span_t* span) { + if (span->size_class != SIZE_CLASS_HUGE) + _rpmalloc_stat_inc(&heap->span_use[span->span_count - 1].spans_deferred); + //This list does not need ABA protection, no mutable side state + do { + span->free_list = (void*)atomic_load_ptr(&heap->span_free_deferred); + } while (!atomic_cas_ptr(&heap->span_free_deferred, span, span->free_list)); +} + +//! Put the block in the deferred free list of the owning span +static void +_rpmalloc_deallocate_defer_small_or_medium(span_t* span, void* block) { + // The memory ordering here is a bit tricky, to avoid having to ABA protect + // the deferred free list to avoid desynchronization of list and list size + // we need to have acquire semantics on successful CAS of the pointer to + // guarantee the list_size variable validity + release semantics on pointer store + void* free_list; + do { + free_list = atomic_exchange_ptr_acquire(&span->free_list_deferred, INVALID_POINTER); + } while (free_list == INVALID_POINTER); + *((void**)block) = free_list; + uint32_t free_count = ++span->list_size; + int all_deferred_free = (free_count == span->block_count); + atomic_store_ptr_release(&span->free_list_deferred, block); + if (all_deferred_free) { + // Span was completely freed by this block. Due to the INVALID_POINTER spin lock + // no other thread can reach this state simultaneously on this span. + // Safe to move to owner heap deferred cache + _rpmalloc_deallocate_defer_free_span(span->heap, span); + } +} + +static void +_rpmalloc_deallocate_small_or_medium(span_t* span, void* p) { + _rpmalloc_stat_inc_free(span->heap, span->size_class); + if (span->flags & SPAN_FLAG_ALIGNED_BLOCKS) { + //Realign pointer to block start + void* blocks_start = pointer_offset(span, SPAN_HEADER_SIZE); + uint32_t block_offset = (uint32_t)pointer_diff(p, blocks_start); + p = pointer_offset(p, -(int32_t)(block_offset % span->block_size)); + } + //Check if block belongs to this heap or if deallocation should be deferred +#if RPMALLOC_FIRST_CLASS_HEAPS + int defer = (span->heap->owner_thread && (span->heap->owner_thread != get_thread_id()) && !span->heap->finalize); +#else + int defer = ((span->heap->owner_thread != get_thread_id()) && !span->heap->finalize); +#endif + if (!defer) + _rpmalloc_deallocate_direct_small_or_medium(span, p); + else + _rpmalloc_deallocate_defer_small_or_medium(span, p); +} + +//! Deallocate the given large memory block to the current heap +static void +_rpmalloc_deallocate_large(span_t* span) { + rpmalloc_assert(span->size_class == SIZE_CLASS_LARGE, "Bad span size class"); + rpmalloc_assert(!(span->flags & SPAN_FLAG_MASTER) || !(span->flags & SPAN_FLAG_SUBSPAN), "Span flag corrupted"); + rpmalloc_assert((span->flags & SPAN_FLAG_MASTER) || (span->flags & SPAN_FLAG_SUBSPAN), "Span flag corrupted"); + //We must always defer (unless finalizing) if from another heap since we cannot touch the list or counters of another heap +#if RPMALLOC_FIRST_CLASS_HEAPS + int defer = (span->heap->owner_thread && (span->heap->owner_thread != get_thread_id()) && !span->heap->finalize); +#else + int defer = ((span->heap->owner_thread != get_thread_id()) && !span->heap->finalize); +#endif + if (defer) { + _rpmalloc_deallocate_defer_free_span(span->heap, span); + return; + } + rpmalloc_assert(span->heap->full_span_count, "Heap span counter corrupted"); + --span->heap->full_span_count; +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_remove(&span->heap->large_huge_span, span); +#endif +#if ENABLE_ADAPTIVE_THREAD_CACHE || ENABLE_STATISTICS + //Decrease counter + size_t idx = span->span_count - 1; + atomic_decr32(&span->heap->span_use[idx].current); +#endif + heap_t* heap = span->heap; + rpmalloc_assert(heap, "No thread heap"); +#if ENABLE_THREAD_CACHE + const int set_as_reserved = ((span->span_count > 1) && (heap->span_cache.count == 0) && !heap->finalize && !heap->spans_reserved); +#else + const int set_as_reserved = ((span->span_count > 1) && !heap->finalize && !heap->spans_reserved); +#endif + if (set_as_reserved) { + heap->span_reserve = span; + heap->spans_reserved = span->span_count; + if (span->flags & SPAN_FLAG_MASTER) { + heap->span_reserve_master = span; + } else { //SPAN_FLAG_SUBSPAN + span_t* master = (span_t*)pointer_offset(span, -(intptr_t)((size_t)span->offset_from_master * _memory_span_size)); + heap->span_reserve_master = master; + rpmalloc_assert(master->flags & SPAN_FLAG_MASTER, "Span flag corrupted"); + rpmalloc_assert(atomic_load32(&master->remaining_spans) >= (int32_t)span->span_count, "Master span count corrupted"); + } + _rpmalloc_stat_inc(&heap->span_use[idx].spans_to_reserved); + } else { + //Insert into cache list + _rpmalloc_heap_cache_insert(heap, span); + } +} + +//! Deallocate the given huge span +static void +_rpmalloc_deallocate_huge(span_t* span) { + rpmalloc_assert(span->heap, "No span heap"); +#if RPMALLOC_FIRST_CLASS_HEAPS + int defer = (span->heap->owner_thread && (span->heap->owner_thread != get_thread_id()) && !span->heap->finalize); +#else + int defer = ((span->heap->owner_thread != get_thread_id()) && !span->heap->finalize); +#endif + if (defer) { + _rpmalloc_deallocate_defer_free_span(span->heap, span); + return; + } + rpmalloc_assert(span->heap->full_span_count, "Heap span counter corrupted"); + --span->heap->full_span_count; +#if RPMALLOC_FIRST_CLASS_HEAPS + _rpmalloc_span_double_link_list_remove(&span->heap->large_huge_span, span); +#endif + + //Oversized allocation, page count is stored in span_count + size_t num_pages = span->span_count; + _rpmalloc_unmap(span, num_pages * _memory_page_size, span->align_offset, num_pages * _memory_page_size); + _rpmalloc_stat_sub(&_huge_pages_current, num_pages); +} + +//! Deallocate the given block +static void +_rpmalloc_deallocate(void* p) { + _rpmalloc_stat_add64(&_deallocation_counter, 1); + //Grab the span (always at start of span, using span alignment) + span_t* span = (span_t*)((uintptr_t)p & _memory_span_mask); + if (UNEXPECTED(!span)) + return; + if (EXPECTED(span->size_class < SIZE_CLASS_COUNT)) + _rpmalloc_deallocate_small_or_medium(span, p); + else if (span->size_class == SIZE_CLASS_LARGE) + _rpmalloc_deallocate_large(span); + else + _rpmalloc_deallocate_huge(span); +} + +//////////// +/// +/// Reallocation entry points +/// +////// + +static size_t +_rpmalloc_usable_size(void* p); + +//! Reallocate the given block to the given size +static void* +_rpmalloc_reallocate(heap_t* heap, void* p, size_t size, size_t oldsize, unsigned int flags) { + if (p) { + //Grab the span using guaranteed span alignment + span_t* span = (span_t*)((uintptr_t)p & _memory_span_mask); + if (EXPECTED(span->size_class < SIZE_CLASS_COUNT)) { + //Small/medium sized block + rpmalloc_assert(span->span_count == 1, "Span counter corrupted"); + void* blocks_start = pointer_offset(span, SPAN_HEADER_SIZE); + uint32_t block_offset = (uint32_t)pointer_diff(p, blocks_start); + uint32_t block_idx = block_offset / span->block_size; + void* block = pointer_offset(blocks_start, (size_t)block_idx * span->block_size); + if (!oldsize) + oldsize = (size_t)((ptrdiff_t)span->block_size - pointer_diff(p, block)); + if ((size_t)span->block_size >= size) { + //Still fits in block, never mind trying to save memory, but preserve data if alignment changed + if ((p != block) && !(flags & RPMALLOC_NO_PRESERVE)) + memmove(block, p, oldsize); + return block; + } + } else if (span->size_class == SIZE_CLASS_LARGE) { + //Large block + size_t total_size = size + SPAN_HEADER_SIZE; + size_t num_spans = total_size >> _memory_span_size_shift; + if (total_size & (_memory_span_mask - 1)) + ++num_spans; + size_t current_spans = span->span_count; + void* block = pointer_offset(span, SPAN_HEADER_SIZE); + if (!oldsize) + oldsize = (current_spans * _memory_span_size) - (size_t)pointer_diff(p, block) - SPAN_HEADER_SIZE; + if ((current_spans >= num_spans) && (total_size >= (oldsize / 2))) { + //Still fits in block, never mind trying to save memory, but preserve data if alignment changed + if ((p != block) && !(flags & RPMALLOC_NO_PRESERVE)) + memmove(block, p, oldsize); + return block; + } + } else { + //Oversized block + size_t total_size = size + SPAN_HEADER_SIZE; + size_t num_pages = total_size >> _memory_page_size_shift; + if (total_size & (_memory_page_size - 1)) + ++num_pages; + //Page count is stored in span_count + size_t current_pages = span->span_count; + void* block = pointer_offset(span, SPAN_HEADER_SIZE); + if (!oldsize) + oldsize = (current_pages * _memory_page_size) - (size_t)pointer_diff(p, block) - SPAN_HEADER_SIZE; + if ((current_pages >= num_pages) && (num_pages >= (current_pages / 2))) { + //Still fits in block, never mind trying to save memory, but preserve data if alignment changed + if ((p != block) && !(flags & RPMALLOC_NO_PRESERVE)) + memmove(block, p, oldsize); + return block; + } + } + } else { + oldsize = 0; + } + + if (!!(flags & RPMALLOC_GROW_OR_FAIL)) + return 0; + + //Size is greater than block size, need to allocate a new block and deallocate the old + //Avoid hysteresis by overallocating if increase is small (below 37%) + size_t lower_bound = oldsize + (oldsize >> 2) + (oldsize >> 3); + size_t new_size = (size > lower_bound) ? size : ((size > oldsize) ? lower_bound : size); + void* block = _rpmalloc_allocate(heap, new_size); + if (p && block) { + if (!(flags & RPMALLOC_NO_PRESERVE)) + memcpy(block, p, oldsize < new_size ? oldsize : new_size); + _rpmalloc_deallocate(p); + } + + return block; +} + +static void* +_rpmalloc_aligned_reallocate(heap_t* heap, void* ptr, size_t alignment, size_t size, size_t oldsize, + unsigned int flags) { + if (alignment <= SMALL_GRANULARITY) + return _rpmalloc_reallocate(heap, ptr, size, oldsize, flags); + + int no_alloc = !!(flags & RPMALLOC_GROW_OR_FAIL); + size_t usablesize = (ptr ? _rpmalloc_usable_size(ptr) : 0); + if ((usablesize >= size) && !((uintptr_t)ptr & (alignment - 1))) { + if (no_alloc || (size >= (usablesize / 2))) + return ptr; + } + // Aligned alloc marks span as having aligned blocks + void* block = (!no_alloc ? _rpmalloc_aligned_allocate(heap, alignment, size) : 0); + if (EXPECTED(block != 0)) { + if (!(flags & RPMALLOC_NO_PRESERVE) && ptr) { + if (!oldsize) + oldsize = usablesize; + memcpy(block, ptr, oldsize < size ? oldsize : size); + } + _rpmalloc_deallocate(ptr); + } + return block; +} + + +//////////// +/// +/// Initialization, finalization and utility +/// +////// + +//! Get the usable size of the given block +static size_t +_rpmalloc_usable_size(void* p) { + //Grab the span using guaranteed span alignment + span_t* span = (span_t*)((uintptr_t)p & _memory_span_mask); + if (span->size_class < SIZE_CLASS_COUNT) { + //Small/medium block + void* blocks_start = pointer_offset(span, SPAN_HEADER_SIZE); + return span->block_size - ((size_t)pointer_diff(p, blocks_start) % span->block_size); + } + if (span->size_class == SIZE_CLASS_LARGE) { + //Large block + size_t current_spans = span->span_count; + return (current_spans * _memory_span_size) - (size_t)pointer_diff(p, span); + } + //Oversized block, page count is stored in span_count + size_t current_pages = span->span_count; + return (current_pages * _memory_page_size) - (size_t)pointer_diff(p, span); +} + +//! Adjust and optimize the size class properties for the given class +static void +_rpmalloc_adjust_size_class(size_t iclass) { + size_t block_size = _memory_size_class[iclass].block_size; + size_t block_count = (_memory_span_size - SPAN_HEADER_SIZE) / block_size; + + _memory_size_class[iclass].block_count = (uint16_t)block_count; + _memory_size_class[iclass].class_idx = (uint16_t)iclass; + + //Check if previous size classes can be merged + if (iclass >= SMALL_CLASS_COUNT) { + size_t prevclass = iclass; + while (prevclass > 0) { + --prevclass; + //A class can be merged if number of pages and number of blocks are equal + if (_memory_size_class[prevclass].block_count == _memory_size_class[iclass].block_count) + memcpy(_memory_size_class + prevclass, _memory_size_class + iclass, sizeof(_memory_size_class[iclass])); + else + break; + } + } +} + +//! Initialize the allocator and setup global data +TRACY_API int +rpmalloc_initialize(void) { + if (_rpmalloc_initialized) { + rpmalloc_thread_initialize(); + return 0; + } + return rpmalloc_initialize_config(0); +} + +int +rpmalloc_initialize_config(const rpmalloc_config_t* config) { + if (_rpmalloc_initialized) { + rpmalloc_thread_initialize(); + return 0; + } + _rpmalloc_initialized = 1; + + if (config) + memcpy(&_memory_config, config, sizeof(rpmalloc_config_t)); + else + memset(&_memory_config, 0, sizeof(rpmalloc_config_t)); + + if (!_memory_config.memory_map || !_memory_config.memory_unmap) { + _memory_config.memory_map = _rpmalloc_mmap_os; + _memory_config.memory_unmap = _rpmalloc_unmap_os; + } + +#if PLATFORM_WINDOWS + SYSTEM_INFO system_info; + memset(&system_info, 0, sizeof(system_info)); + GetSystemInfo(&system_info); + _memory_map_granularity = system_info.dwAllocationGranularity; +#else + _memory_map_granularity = (size_t)sysconf(_SC_PAGESIZE); +#endif + +#if RPMALLOC_CONFIGURABLE + _memory_page_size = _memory_config.page_size; +#else + _memory_page_size = 0; +#endif + _memory_huge_pages = 0; + if (!_memory_page_size) { +#if PLATFORM_WINDOWS + _memory_page_size = system_info.dwPageSize; +#else + _memory_page_size = _memory_map_granularity; + if (_memory_config.enable_huge_pages) { +#if defined(__linux__) + size_t huge_page_size = 0; + FILE* meminfo = fopen("/proc/meminfo", "r"); + if (meminfo) { + char line[128]; + while (!huge_page_size && fgets(line, sizeof(line) - 1, meminfo)) { + line[sizeof(line) - 1] = 0; + if (strstr(line, "Hugepagesize:")) + huge_page_size = (size_t)strtol(line + 13, 0, 10) * 1024; + } + fclose(meminfo); + } + if (huge_page_size) { + _memory_huge_pages = 1; + _memory_page_size = huge_page_size; + _memory_map_granularity = huge_page_size; + } +#elif defined(__FreeBSD__) + int rc; + size_t sz = sizeof(rc); + + if (sysctlbyname("vm.pmap.pg_ps_enabled", &rc, &sz, NULL, 0) == 0 && rc == 1) { + _memory_huge_pages = 1; + _memory_page_size = 2 * 1024 * 1024; + _memory_map_granularity = _memory_page_size; + } +#elif defined(__APPLE__) || defined(__NetBSD__) + _memory_huge_pages = 1; + _memory_page_size = 2 * 1024 * 1024; + _memory_map_granularity = _memory_page_size; +#endif + } +#endif + } else { + if (_memory_config.enable_huge_pages) + _memory_huge_pages = 1; + } + +#if PLATFORM_WINDOWS + if (_memory_config.enable_huge_pages) { + HANDLE token = 0; + size_t large_page_minimum = GetLargePageMinimum(); + if (large_page_minimum) + OpenProcessToken(GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES | TOKEN_QUERY, &token); + if (token) { + LUID luid; + if (LookupPrivilegeValue(0, SE_LOCK_MEMORY_NAME, &luid)) { + TOKEN_PRIVILEGES token_privileges; + memset(&token_privileges, 0, sizeof(token_privileges)); + token_privileges.PrivilegeCount = 1; + token_privileges.Privileges[0].Luid = luid; + token_privileges.Privileges[0].Attributes = SE_PRIVILEGE_ENABLED; + if (AdjustTokenPrivileges(token, FALSE, &token_privileges, 0, 0, 0)) { + if (GetLastError() == ERROR_SUCCESS) + _memory_huge_pages = 1; + } + } + CloseHandle(token); + } + if (_memory_huge_pages) { + if (large_page_minimum > _memory_page_size) + _memory_page_size = large_page_minimum; + if (large_page_minimum > _memory_map_granularity) + _memory_map_granularity = large_page_minimum; + } + } +#endif + + size_t min_span_size = 256; + size_t max_page_size; +#if UINTPTR_MAX > 0xFFFFFFFF + max_page_size = 4096ULL * 1024ULL * 1024ULL; +#else + max_page_size = 4 * 1024 * 1024; +#endif + if (_memory_page_size < min_span_size) + _memory_page_size = min_span_size; + if (_memory_page_size > max_page_size) + _memory_page_size = max_page_size; + _memory_page_size_shift = 0; + size_t page_size_bit = _memory_page_size; + while (page_size_bit != 1) { + ++_memory_page_size_shift; + page_size_bit >>= 1; + } + _memory_page_size = ((size_t)1 << _memory_page_size_shift); + +#if RPMALLOC_CONFIGURABLE + if (!_memory_config.span_size) { + _memory_span_size = _memory_default_span_size; + _memory_span_size_shift = _memory_default_span_size_shift; + _memory_span_mask = _memory_default_span_mask; + } else { + size_t span_size = _memory_config.span_size; + if (span_size > (256 * 1024)) + span_size = (256 * 1024); + _memory_span_size = 4096; + _memory_span_size_shift = 12; + while (_memory_span_size < span_size) { + _memory_span_size <<= 1; + ++_memory_span_size_shift; + } + _memory_span_mask = ~(uintptr_t)(_memory_span_size - 1); + } +#endif + + _memory_span_map_count = ( _memory_config.span_map_count ? _memory_config.span_map_count : DEFAULT_SPAN_MAP_COUNT); + if ((_memory_span_size * _memory_span_map_count) < _memory_page_size) + _memory_span_map_count = (_memory_page_size / _memory_span_size); + if ((_memory_page_size >= _memory_span_size) && ((_memory_span_map_count * _memory_span_size) % _memory_page_size)) + _memory_span_map_count = (_memory_page_size / _memory_span_size); + _memory_heap_reserve_count = (_memory_span_map_count > DEFAULT_SPAN_MAP_COUNT) ? DEFAULT_SPAN_MAP_COUNT : _memory_span_map_count; + + _memory_config.page_size = _memory_page_size; + _memory_config.span_size = _memory_span_size; + _memory_config.span_map_count = _memory_span_map_count; + _memory_config.enable_huge_pages = _memory_huge_pages; + +#if ((defined(__APPLE__) || defined(__HAIKU__)) && ENABLE_PRELOAD) || defined(__TINYC__) + if (pthread_key_create(&_memory_thread_heap, _rpmalloc_heap_release_raw_fc)) + return -1; +#endif +#if defined(_WIN32) && (!defined(BUILD_DYNAMIC_LINK) || !BUILD_DYNAMIC_LINK) + fls_key = FlsAlloc(&_rpmalloc_thread_destructor); +#endif + + //Setup all small and medium size classes + size_t iclass = 0; + _memory_size_class[iclass].block_size = SMALL_GRANULARITY; + _rpmalloc_adjust_size_class(iclass); + for (iclass = 1; iclass < SMALL_CLASS_COUNT; ++iclass) { + size_t size = iclass * SMALL_GRANULARITY; + _memory_size_class[iclass].block_size = (uint32_t)size; + _rpmalloc_adjust_size_class(iclass); + } + //At least two blocks per span, then fall back to large allocations + _memory_medium_size_limit = (_memory_span_size - SPAN_HEADER_SIZE) >> 1; + if (_memory_medium_size_limit > MEDIUM_SIZE_LIMIT) + _memory_medium_size_limit = MEDIUM_SIZE_LIMIT; + for (iclass = 0; iclass < MEDIUM_CLASS_COUNT; ++iclass) { + size_t size = SMALL_SIZE_LIMIT + ((iclass + 1) * MEDIUM_GRANULARITY); + if (size > _memory_medium_size_limit) + break; + _memory_size_class[SMALL_CLASS_COUNT + iclass].block_size = (uint32_t)size; + _rpmalloc_adjust_size_class(SMALL_CLASS_COUNT + iclass); + } + + _memory_orphan_heaps = 0; +#if RPMALLOC_FIRST_CLASS_HEAPS + _memory_first_class_orphan_heaps = 0; +#endif +#if ENABLE_STATISTICS + atomic_store32(&_memory_active_heaps, 0); + atomic_store32(&_mapped_pages, 0); + _mapped_pages_peak = 0; + atomic_store32(&_master_spans, 0); + atomic_store32(&_mapped_total, 0); + atomic_store32(&_unmapped_total, 0); + atomic_store32(&_mapped_pages_os, 0); + atomic_store32(&_huge_pages_current, 0); + _huge_pages_peak = 0; +#endif + memset(_memory_heaps, 0, sizeof(_memory_heaps)); + atomic_store32_release(&_memory_global_lock, 0); + + //Initialize this thread + rpmalloc_thread_initialize(); + return 0; +} + +//! Finalize the allocator +TRACY_API void +rpmalloc_finalize(void) { + rpmalloc_thread_finalize(1); + //rpmalloc_dump_statistics(stdout); + + if (_memory_global_reserve) { + atomic_add32(&_memory_global_reserve_master->remaining_spans, -(int32_t)_memory_global_reserve_count); + _memory_global_reserve_master = 0; + _memory_global_reserve_count = 0; + _memory_global_reserve = 0; + } + atomic_store32_release(&_memory_global_lock, 0); + + //Free all thread caches and fully free spans + for (size_t list_idx = 0; list_idx < HEAP_ARRAY_SIZE; ++list_idx) { + heap_t* heap = _memory_heaps[list_idx]; + while (heap) { + heap_t* next_heap = heap->next_heap; + heap->finalize = 1; + _rpmalloc_heap_global_finalize(heap); + heap = next_heap; + } + } + +#if ENABLE_GLOBAL_CACHE + //Free global caches + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) + _rpmalloc_global_cache_finalize(&_memory_span_cache[iclass]); +#endif + +#if (defined(__APPLE__) || defined(__HAIKU__)) && ENABLE_PRELOAD + pthread_key_delete(_memory_thread_heap); +#endif +#if defined(_WIN32) && (!defined(BUILD_DYNAMIC_LINK) || !BUILD_DYNAMIC_LINK) + FlsFree(fls_key); + fls_key = 0; +#endif +#if ENABLE_STATISTICS + //If you hit these asserts you probably have memory leaks (perhaps global scope data doing dynamic allocations) or double frees in your code + rpmalloc_assert(atomic_load32(&_mapped_pages) == 0, "Memory leak detected"); + rpmalloc_assert(atomic_load32(&_mapped_pages_os) == 0, "Memory leak detected"); +#endif + + _rpmalloc_initialized = 0; +} + +//! Initialize thread, assign heap +TRACY_API void +rpmalloc_thread_initialize(void) { + if (!get_thread_heap_raw()) { + heap_t* heap = _rpmalloc_heap_allocate(0); + if (heap) { + _rpmalloc_stat_inc(&_memory_active_heaps); + set_thread_heap(heap); +#if defined(_WIN32) && (!defined(BUILD_DYNAMIC_LINK) || !BUILD_DYNAMIC_LINK) + FlsSetValue(fls_key, heap); +#endif + } + } +} + +//! Finalize thread, orphan heap +TRACY_API void +rpmalloc_thread_finalize(int release_caches) { + heap_t* heap = get_thread_heap_raw(); + if (heap) + _rpmalloc_heap_release_raw(heap, release_caches); + set_thread_heap(0); +#if defined(_WIN32) && (!defined(BUILD_DYNAMIC_LINK) || !BUILD_DYNAMIC_LINK) + FlsSetValue(fls_key, 0); +#endif +} + +int +rpmalloc_is_thread_initialized(void) { + return (get_thread_heap_raw() != 0) ? 1 : 0; +} + +const rpmalloc_config_t* +rpmalloc_config(void) { + return &_memory_config; +} + +// Extern interface + +TRACY_API RPMALLOC_ALLOCATOR void* +rpmalloc(size_t size) { +#if ENABLE_VALIDATE_ARGS + if (size >= MAX_ALLOC_SIZE) { + errno = EINVAL; + return 0; + } +#endif + heap_t* heap = get_thread_heap(); + return _rpmalloc_allocate(heap, size); +} + +TRACY_API void +rpfree(void* ptr) { + _rpmalloc_deallocate(ptr); +} + +extern inline RPMALLOC_ALLOCATOR void* +rpcalloc(size_t num, size_t size) { + size_t total; +#if ENABLE_VALIDATE_ARGS +#if PLATFORM_WINDOWS + int err = SizeTMult(num, size, &total); + if ((err != S_OK) || (total >= MAX_ALLOC_SIZE)) { + errno = EINVAL; + return 0; + } +#else + int err = __builtin_umull_overflow(num, size, &total); + if (err || (total >= MAX_ALLOC_SIZE)) { + errno = EINVAL; + return 0; + } +#endif +#else + total = num * size; +#endif + heap_t* heap = get_thread_heap(); + void* block = _rpmalloc_allocate(heap, total); + if (block) + memset(block, 0, total); + return block; +} + +TRACY_API RPMALLOC_ALLOCATOR void* +rprealloc(void* ptr, size_t size) { +#if ENABLE_VALIDATE_ARGS + if (size >= MAX_ALLOC_SIZE) { + errno = EINVAL; + return ptr; + } +#endif + heap_t* heap = get_thread_heap(); + return _rpmalloc_reallocate(heap, ptr, size, 0, 0); +} + +extern RPMALLOC_ALLOCATOR void* +rpaligned_realloc(void* ptr, size_t alignment, size_t size, size_t oldsize, + unsigned int flags) { +#if ENABLE_VALIDATE_ARGS + if ((size + alignment < size) || (alignment > _memory_page_size)) { + errno = EINVAL; + return 0; + } +#endif + heap_t* heap = get_thread_heap(); + return _rpmalloc_aligned_reallocate(heap, ptr, alignment, size, oldsize, flags); +} + +extern RPMALLOC_ALLOCATOR void* +rpaligned_alloc(size_t alignment, size_t size) { + heap_t* heap = get_thread_heap(); + return _rpmalloc_aligned_allocate(heap, alignment, size); +} + +extern inline RPMALLOC_ALLOCATOR void* +rpaligned_calloc(size_t alignment, size_t num, size_t size) { + size_t total; +#if ENABLE_VALIDATE_ARGS +#if PLATFORM_WINDOWS + int err = SizeTMult(num, size, &total); + if ((err != S_OK) || (total >= MAX_ALLOC_SIZE)) { + errno = EINVAL; + return 0; + } +#else + int err = __builtin_umull_overflow(num, size, &total); + if (err || (total >= MAX_ALLOC_SIZE)) { + errno = EINVAL; + return 0; + } +#endif +#else + total = num * size; +#endif + void* block = rpaligned_alloc(alignment, total); + if (block) + memset(block, 0, total); + return block; +} + +extern inline RPMALLOC_ALLOCATOR void* +rpmemalign(size_t alignment, size_t size) { + return rpaligned_alloc(alignment, size); +} + +extern inline int +rpposix_memalign(void **memptr, size_t alignment, size_t size) { + if (memptr) + *memptr = rpaligned_alloc(alignment, size); + else + return EINVAL; + return *memptr ? 0 : ENOMEM; +} + +extern inline size_t +rpmalloc_usable_size(void* ptr) { + return (ptr ? _rpmalloc_usable_size(ptr) : 0); +} + +extern inline void +rpmalloc_thread_collect(void) { +} + +void +rpmalloc_thread_statistics(rpmalloc_thread_statistics_t* stats) { + memset(stats, 0, sizeof(rpmalloc_thread_statistics_t)); + heap_t* heap = get_thread_heap_raw(); + if (!heap) + return; + + for (size_t iclass = 0; iclass < SIZE_CLASS_COUNT; ++iclass) { + size_class_t* size_class = _memory_size_class + iclass; + span_t* span = heap->size_class[iclass].partial_span; + while (span) { + size_t free_count = span->list_size; + size_t block_count = size_class->block_count; + if (span->free_list_limit < block_count) + block_count = span->free_list_limit; + free_count += (block_count - span->used_count); + stats->sizecache = free_count * size_class->block_size; + span = span->next; + } + } + +#if ENABLE_THREAD_CACHE + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) { + span_cache_t* span_cache; + if (!iclass) + span_cache = &heap->span_cache; + else + span_cache = (span_cache_t*)(heap->span_large_cache + (iclass - 1)); + stats->spancache = span_cache->count * (iclass + 1) * _memory_span_size; + } +#endif + + span_t* deferred = (span_t*)atomic_load_ptr(&heap->span_free_deferred); + while (deferred) { + if (deferred->size_class != SIZE_CLASS_HUGE) + stats->spancache = (size_t)deferred->span_count * _memory_span_size; + deferred = (span_t*)deferred->free_list; + } + +#if ENABLE_STATISTICS + stats->thread_to_global = (size_t)atomic_load64(&heap->thread_to_global); + stats->global_to_thread = (size_t)atomic_load64(&heap->global_to_thread); + + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) { + stats->span_use[iclass].current = (size_t)atomic_load32(&heap->span_use[iclass].current); + stats->span_use[iclass].peak = (size_t)atomic_load32(&heap->span_use[iclass].high); + stats->span_use[iclass].to_global = (size_t)atomic_load32(&heap->span_use[iclass].spans_to_global); + stats->span_use[iclass].from_global = (size_t)atomic_load32(&heap->span_use[iclass].spans_from_global); + stats->span_use[iclass].to_cache = (size_t)atomic_load32(&heap->span_use[iclass].spans_to_cache); + stats->span_use[iclass].from_cache = (size_t)atomic_load32(&heap->span_use[iclass].spans_from_cache); + stats->span_use[iclass].to_reserved = (size_t)atomic_load32(&heap->span_use[iclass].spans_to_reserved); + stats->span_use[iclass].from_reserved = (size_t)atomic_load32(&heap->span_use[iclass].spans_from_reserved); + stats->span_use[iclass].map_calls = (size_t)atomic_load32(&heap->span_use[iclass].spans_map_calls); + } + for (size_t iclass = 0; iclass < SIZE_CLASS_COUNT; ++iclass) { + stats->size_use[iclass].alloc_current = (size_t)atomic_load32(&heap->size_class_use[iclass].alloc_current); + stats->size_use[iclass].alloc_peak = (size_t)heap->size_class_use[iclass].alloc_peak; + stats->size_use[iclass].alloc_total = (size_t)atomic_load32(&heap->size_class_use[iclass].alloc_total); + stats->size_use[iclass].free_total = (size_t)atomic_load32(&heap->size_class_use[iclass].free_total); + stats->size_use[iclass].spans_to_cache = (size_t)atomic_load32(&heap->size_class_use[iclass].spans_to_cache); + stats->size_use[iclass].spans_from_cache = (size_t)atomic_load32(&heap->size_class_use[iclass].spans_from_cache); + stats->size_use[iclass].spans_from_reserved = (size_t)atomic_load32(&heap->size_class_use[iclass].spans_from_reserved); + stats->size_use[iclass].map_calls = (size_t)atomic_load32(&heap->size_class_use[iclass].spans_map_calls); + } +#endif +} + +void +rpmalloc_global_statistics(rpmalloc_global_statistics_t* stats) { + memset(stats, 0, sizeof(rpmalloc_global_statistics_t)); +#if ENABLE_STATISTICS + stats->mapped = (size_t)atomic_load32(&_mapped_pages) * _memory_page_size; + stats->mapped_peak = (size_t)_mapped_pages_peak * _memory_page_size; + stats->mapped_total = (size_t)atomic_load32(&_mapped_total) * _memory_page_size; + stats->unmapped_total = (size_t)atomic_load32(&_unmapped_total) * _memory_page_size; + stats->huge_alloc = (size_t)atomic_load32(&_huge_pages_current) * _memory_page_size; + stats->huge_alloc_peak = (size_t)_huge_pages_peak * _memory_page_size; +#endif +#if ENABLE_GLOBAL_CACHE + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) + stats->cached += _memory_span_cache[iclass].count * (iclass + 1) * _memory_span_size; +#endif +} + +#if ENABLE_STATISTICS + +static void +_memory_heap_dump_statistics(heap_t* heap, void* file) { + fprintf(file, "Heap %d stats:\n", heap->id); + fprintf(file, "Class CurAlloc PeakAlloc TotAlloc TotFree BlkSize BlkCount SpansCur SpansPeak PeakAllocMiB ToCacheMiB FromCacheMiB FromReserveMiB MmapCalls\n"); + for (size_t iclass = 0; iclass < SIZE_CLASS_COUNT; ++iclass) { + if (!atomic_load32(&heap->size_class_use[iclass].alloc_total)) + continue; + fprintf(file, "%3u: %10u %10u %10u %10u %8u %8u %8d %9d %13zu %11zu %12zu %14zu %9u\n", (uint32_t)iclass, + atomic_load32(&heap->size_class_use[iclass].alloc_current), + heap->size_class_use[iclass].alloc_peak, + atomic_load32(&heap->size_class_use[iclass].alloc_total), + atomic_load32(&heap->size_class_use[iclass].free_total), + _memory_size_class[iclass].block_size, + _memory_size_class[iclass].block_count, + atomic_load32(&heap->size_class_use[iclass].spans_current), + heap->size_class_use[iclass].spans_peak, + ((size_t)heap->size_class_use[iclass].alloc_peak * (size_t)_memory_size_class[iclass].block_size) / (size_t)(1024 * 1024), + ((size_t)atomic_load32(&heap->size_class_use[iclass].spans_to_cache) * _memory_span_size) / (size_t)(1024 * 1024), + ((size_t)atomic_load32(&heap->size_class_use[iclass].spans_from_cache) * _memory_span_size) / (size_t)(1024 * 1024), + ((size_t)atomic_load32(&heap->size_class_use[iclass].spans_from_reserved) * _memory_span_size) / (size_t)(1024 * 1024), + atomic_load32(&heap->size_class_use[iclass].spans_map_calls)); + } + fprintf(file, "Spans Current Peak Deferred PeakMiB Cached ToCacheMiB FromCacheMiB ToReserveMiB FromReserveMiB ToGlobalMiB FromGlobalMiB MmapCalls\n"); + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) { + if (!atomic_load32(&heap->span_use[iclass].high) && !atomic_load32(&heap->span_use[iclass].spans_map_calls)) + continue; + fprintf(file, "%4u: %8d %8u %8u %8zu %7u %11zu %12zu %12zu %14zu %11zu %13zu %10u\n", (uint32_t)(iclass + 1), + atomic_load32(&heap->span_use[iclass].current), + atomic_load32(&heap->span_use[iclass].high), + atomic_load32(&heap->span_use[iclass].spans_deferred), + ((size_t)atomic_load32(&heap->span_use[iclass].high) * (size_t)_memory_span_size * (iclass + 1)) / (size_t)(1024 * 1024), +#if ENABLE_THREAD_CACHE + (unsigned int)(!iclass ? heap->span_cache.count : heap->span_large_cache[iclass - 1].count), + ((size_t)atomic_load32(&heap->span_use[iclass].spans_to_cache) * (iclass + 1) * _memory_span_size) / (size_t)(1024 * 1024), + ((size_t)atomic_load32(&heap->span_use[iclass].spans_from_cache) * (iclass + 1) * _memory_span_size) / (size_t)(1024 * 1024), +#else + 0, (size_t)0, (size_t)0, +#endif + ((size_t)atomic_load32(&heap->span_use[iclass].spans_to_reserved) * (iclass + 1) * _memory_span_size) / (size_t)(1024 * 1024), + ((size_t)atomic_load32(&heap->span_use[iclass].spans_from_reserved) * (iclass + 1) * _memory_span_size) / (size_t)(1024 * 1024), + ((size_t)atomic_load32(&heap->span_use[iclass].spans_to_global) * (size_t)_memory_span_size * (iclass + 1)) / (size_t)(1024 * 1024), + ((size_t)atomic_load32(&heap->span_use[iclass].spans_from_global) * (size_t)_memory_span_size * (iclass + 1)) / (size_t)(1024 * 1024), + atomic_load32(&heap->span_use[iclass].spans_map_calls)); + } + fprintf(file, "Full spans: %zu\n", heap->full_span_count); + fprintf(file, "ThreadToGlobalMiB GlobalToThreadMiB\n"); + fprintf(file, "%17zu %17zu\n", (size_t)atomic_load64(&heap->thread_to_global) / (size_t)(1024 * 1024), (size_t)atomic_load64(&heap->global_to_thread) / (size_t)(1024 * 1024)); +} + +#endif + +void +rpmalloc_dump_statistics(void* file) { +#if ENABLE_STATISTICS + for (size_t list_idx = 0; list_idx < HEAP_ARRAY_SIZE; ++list_idx) { + heap_t* heap = _memory_heaps[list_idx]; + while (heap) { + int need_dump = 0; + for (size_t iclass = 0; !need_dump && (iclass < SIZE_CLASS_COUNT); ++iclass) { + if (!atomic_load32(&heap->size_class_use[iclass].alloc_total)) { + rpmalloc_assert(!atomic_load32(&heap->size_class_use[iclass].free_total), "Heap statistics counter mismatch"); + rpmalloc_assert(!atomic_load32(&heap->size_class_use[iclass].spans_map_calls), "Heap statistics counter mismatch"); + continue; + } + need_dump = 1; + } + for (size_t iclass = 0; !need_dump && (iclass < LARGE_CLASS_COUNT); ++iclass) { + if (!atomic_load32(&heap->span_use[iclass].high) && !atomic_load32(&heap->span_use[iclass].spans_map_calls)) + continue; + need_dump = 1; + } + if (need_dump) + _memory_heap_dump_statistics(heap, file); + heap = heap->next_heap; + } + } + fprintf(file, "Global stats:\n"); + size_t huge_current = (size_t)atomic_load32(&_huge_pages_current) * _memory_page_size; + size_t huge_peak = (size_t)_huge_pages_peak * _memory_page_size; + fprintf(file, "HugeCurrentMiB HugePeakMiB\n"); + fprintf(file, "%14zu %11zu\n", huge_current / (size_t)(1024 * 1024), huge_peak / (size_t)(1024 * 1024)); + + fprintf(file, "GlobalCacheMiB\n"); + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) { + global_cache_t* cache = _memory_span_cache + iclass; + size_t global_cache = (size_t)cache->count * iclass * _memory_span_size; + + size_t global_overflow_cache = 0; + span_t* span = cache->overflow; + while (span) { + global_overflow_cache += iclass * _memory_span_size; + span = span->next; + } + if (global_cache || global_overflow_cache || cache->insert_count || cache->extract_count) + fprintf(file, "%4zu: %8zuMiB (%8zuMiB overflow) %14zu insert %14zu extract\n", iclass + 1, global_cache / (size_t)(1024 * 1024), global_overflow_cache / (size_t)(1024 * 1024), cache->insert_count, cache->extract_count); + } + + size_t mapped = (size_t)atomic_load32(&_mapped_pages) * _memory_page_size; + size_t mapped_os = (size_t)atomic_load32(&_mapped_pages_os) * _memory_page_size; + size_t mapped_peak = (size_t)_mapped_pages_peak * _memory_page_size; + size_t mapped_total = (size_t)atomic_load32(&_mapped_total) * _memory_page_size; + size_t unmapped_total = (size_t)atomic_load32(&_unmapped_total) * _memory_page_size; + fprintf(file, "MappedMiB MappedOSMiB MappedPeakMiB MappedTotalMiB UnmappedTotalMiB\n"); + fprintf(file, "%9zu %11zu %13zu %14zu %16zu\n", + mapped / (size_t)(1024 * 1024), + mapped_os / (size_t)(1024 * 1024), + mapped_peak / (size_t)(1024 * 1024), + mapped_total / (size_t)(1024 * 1024), + unmapped_total / (size_t)(1024 * 1024)); + + fprintf(file, "\n"); +#if 0 + int64_t allocated = atomic_load64(&_allocation_counter); + int64_t deallocated = atomic_load64(&_deallocation_counter); + fprintf(file, "Allocation count: %lli\n", allocated); + fprintf(file, "Deallocation count: %lli\n", deallocated); + fprintf(file, "Current allocations: %lli\n", (allocated - deallocated)); + fprintf(file, "Master spans: %d\n", atomic_load32(&_master_spans)); + fprintf(file, "Dangling master spans: %d\n", atomic_load32(&_unmapped_master_spans)); +#endif +#endif + (void)sizeof(file); +} + +#if RPMALLOC_FIRST_CLASS_HEAPS + +extern inline rpmalloc_heap_t* +rpmalloc_heap_acquire(void) { + // Must be a pristine heap from newly mapped memory pages, or else memory blocks + // could already be allocated from the heap which would (wrongly) be released when + // heap is cleared with rpmalloc_heap_free_all(). Also heaps guaranteed to be + // pristine from the dedicated orphan list can be used. + heap_t* heap = _rpmalloc_heap_allocate(1); + heap->owner_thread = 0; + _rpmalloc_stat_inc(&_memory_active_heaps); + return heap; +} + +extern inline void +rpmalloc_heap_release(rpmalloc_heap_t* heap) { + if (heap) + _rpmalloc_heap_release(heap, 1, 1); +} + +extern inline RPMALLOC_ALLOCATOR void* +rpmalloc_heap_alloc(rpmalloc_heap_t* heap, size_t size) { +#if ENABLE_VALIDATE_ARGS + if (size >= MAX_ALLOC_SIZE) { + errno = EINVAL; + return 0; + } +#endif + return _rpmalloc_allocate(heap, size); +} + +extern inline RPMALLOC_ALLOCATOR void* +rpmalloc_heap_aligned_alloc(rpmalloc_heap_t* heap, size_t alignment, size_t size) { +#if ENABLE_VALIDATE_ARGS + if (size >= MAX_ALLOC_SIZE) { + errno = EINVAL; + return 0; + } +#endif + return _rpmalloc_aligned_allocate(heap, alignment, size); +} + +extern inline RPMALLOC_ALLOCATOR void* +rpmalloc_heap_calloc(rpmalloc_heap_t* heap, size_t num, size_t size) { + return rpmalloc_heap_aligned_calloc(heap, 0, num, size); +} + +extern inline RPMALLOC_ALLOCATOR void* +rpmalloc_heap_aligned_calloc(rpmalloc_heap_t* heap, size_t alignment, size_t num, size_t size) { + size_t total; +#if ENABLE_VALIDATE_ARGS +#if PLATFORM_WINDOWS + int err = SizeTMult(num, size, &total); + if ((err != S_OK) || (total >= MAX_ALLOC_SIZE)) { + errno = EINVAL; + return 0; + } +#else + int err = __builtin_umull_overflow(num, size, &total); + if (err || (total >= MAX_ALLOC_SIZE)) { + errno = EINVAL; + return 0; + } +#endif +#else + total = num * size; +#endif + void* block = _rpmalloc_aligned_allocate(heap, alignment, total); + if (block) + memset(block, 0, total); + return block; +} + +extern inline RPMALLOC_ALLOCATOR void* +rpmalloc_heap_realloc(rpmalloc_heap_t* heap, void* ptr, size_t size, unsigned int flags) { +#if ENABLE_VALIDATE_ARGS + if (size >= MAX_ALLOC_SIZE) { + errno = EINVAL; + return ptr; + } +#endif + return _rpmalloc_reallocate(heap, ptr, size, 0, flags); +} + +extern inline RPMALLOC_ALLOCATOR void* +rpmalloc_heap_aligned_realloc(rpmalloc_heap_t* heap, void* ptr, size_t alignment, size_t size, unsigned int flags) { +#if ENABLE_VALIDATE_ARGS + if ((size + alignment < size) || (alignment > _memory_page_size)) { + errno = EINVAL; + return 0; + } +#endif + return _rpmalloc_aligned_reallocate(heap, ptr, alignment, size, 0, flags); +} + +extern inline void +rpmalloc_heap_free(rpmalloc_heap_t* heap, void* ptr) { + (void)sizeof(heap); + _rpmalloc_deallocate(ptr); +} + +extern inline void +rpmalloc_heap_free_all(rpmalloc_heap_t* heap) { + span_t* span; + span_t* next_span; + + _rpmalloc_heap_cache_adopt_deferred(heap, 0); + + for (size_t iclass = 0; iclass < SIZE_CLASS_COUNT; ++iclass) { + span = heap->size_class[iclass].partial_span; + while (span) { + next_span = span->next; + _rpmalloc_heap_cache_insert(heap, span); + span = next_span; + } + heap->size_class[iclass].partial_span = 0; + span = heap->full_span[iclass]; + while (span) { + next_span = span->next; + _rpmalloc_heap_cache_insert(heap, span); + span = next_span; + } + } + memset(heap->size_class, 0, sizeof(heap->size_class)); + memset(heap->full_span, 0, sizeof(heap->full_span)); + + span = heap->large_huge_span; + while (span) { + next_span = span->next; + if (UNEXPECTED(span->size_class == SIZE_CLASS_HUGE)) + _rpmalloc_deallocate_huge(span); + else + _rpmalloc_heap_cache_insert(heap, span); + span = next_span; + } + heap->large_huge_span = 0; + heap->full_span_count = 0; + +#if ENABLE_THREAD_CACHE + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) { + span_cache_t* span_cache; + if (!iclass) + span_cache = &heap->span_cache; + else + span_cache = (span_cache_t*)(heap->span_large_cache + (iclass - 1)); + if (!span_cache->count) + continue; +#if ENABLE_GLOBAL_CACHE + _rpmalloc_stat_add64(&heap->thread_to_global, span_cache->count * (iclass + 1) * _memory_span_size); + _rpmalloc_stat_add(&heap->span_use[iclass].spans_to_global, span_cache->count); + _rpmalloc_global_cache_insert_spans(span_cache->span, iclass + 1, span_cache->count); +#else + for (size_t ispan = 0; ispan < span_cache->count; ++ispan) + _rpmalloc_span_unmap(span_cache->span[ispan]); +#endif + span_cache->count = 0; + } +#endif + +#if ENABLE_STATISTICS + for (size_t iclass = 0; iclass < SIZE_CLASS_COUNT; ++iclass) { + atomic_store32(&heap->size_class_use[iclass].alloc_current, 0); + atomic_store32(&heap->size_class_use[iclass].spans_current, 0); + } + for (size_t iclass = 0; iclass < LARGE_CLASS_COUNT; ++iclass) { + atomic_store32(&heap->span_use[iclass].current, 0); + } +#endif +} + +extern inline void +rpmalloc_heap_thread_set_current(rpmalloc_heap_t* heap) { + heap_t* prev_heap = get_thread_heap_raw(); + if (prev_heap != heap) { + set_thread_heap(heap); + if (prev_heap) + rpmalloc_heap_release(prev_heap); + } +} + +#endif + +} + +#endif diff --git a/externals/tracy/public/client/tracy_rpmalloc.hpp b/externals/tracy/public/client/tracy_rpmalloc.hpp new file mode 100644 index 000000000..51216a21b --- /dev/null +++ b/externals/tracy/public/client/tracy_rpmalloc.hpp @@ -0,0 +1,363 @@ +/* rpmalloc.h - Memory allocator - Public Domain - 2016 Mattias Jansson + * + * This library provides a cross-platform lock free thread caching malloc implementation in C11. + * The latest source code is always available at + * + * https://github.com/mjansson/rpmalloc + * + * This library is put in the public domain; you can redistribute it and/or modify it without any restrictions. + * + */ + +#pragma once + +#include +#include "../common/TracyApi.h" + +namespace tracy +{ + +#if defined(__clang__) || defined(__GNUC__) +# define RPMALLOC_EXPORT __attribute__((visibility("default"))) +# define RPMALLOC_ALLOCATOR +# if (defined(__clang_major__) && (__clang_major__ < 4)) || (defined(__GNUC__) && defined(ENABLE_PRELOAD) && ENABLE_PRELOAD) +# define RPMALLOC_ATTRIB_MALLOC +# define RPMALLOC_ATTRIB_ALLOC_SIZE(size) +# define RPMALLOC_ATTRIB_ALLOC_SIZE2(count, size) +# else +# define RPMALLOC_ATTRIB_MALLOC __attribute__((__malloc__)) +# define RPMALLOC_ATTRIB_ALLOC_SIZE(size) __attribute__((alloc_size(size))) +# define RPMALLOC_ATTRIB_ALLOC_SIZE2(count, size) __attribute__((alloc_size(count, size))) +# endif +# define RPMALLOC_CDECL +#elif defined(_MSC_VER) +# define RPMALLOC_EXPORT +# define RPMALLOC_ALLOCATOR __declspec(allocator) __declspec(restrict) +# define RPMALLOC_ATTRIB_MALLOC +# define RPMALLOC_ATTRIB_ALLOC_SIZE(size) +# define RPMALLOC_ATTRIB_ALLOC_SIZE2(count,size) +# define RPMALLOC_CDECL __cdecl +#else +# define RPMALLOC_EXPORT +# define RPMALLOC_ALLOCATOR +# define RPMALLOC_ATTRIB_MALLOC +# define RPMALLOC_ATTRIB_ALLOC_SIZE(size) +# define RPMALLOC_ATTRIB_ALLOC_SIZE2(count,size) +# define RPMALLOC_CDECL +#endif + +//! Define RPMALLOC_CONFIGURABLE to enable configuring sizes. Will introduce +// a very small overhead due to some size calculations not being compile time constants +#ifndef RPMALLOC_CONFIGURABLE +#define RPMALLOC_CONFIGURABLE 0 +#endif + +//! Define RPMALLOC_FIRST_CLASS_HEAPS to enable heap based API (rpmalloc_heap_* functions). +// Will introduce a very small overhead to track fully allocated spans in heaps +#ifndef RPMALLOC_FIRST_CLASS_HEAPS +#define RPMALLOC_FIRST_CLASS_HEAPS 0 +#endif + +//! Flag to rpaligned_realloc to not preserve content in reallocation +#define RPMALLOC_NO_PRESERVE 1 +//! Flag to rpaligned_realloc to fail and return null pointer if grow cannot be done in-place, +// in which case the original pointer is still valid (just like a call to realloc which failes to allocate +// a new block). +#define RPMALLOC_GROW_OR_FAIL 2 + +typedef struct rpmalloc_global_statistics_t { + //! Current amount of virtual memory mapped, all of which might not have been committed (only if ENABLE_STATISTICS=1) + size_t mapped; + //! Peak amount of virtual memory mapped, all of which might not have been committed (only if ENABLE_STATISTICS=1) + size_t mapped_peak; + //! Current amount of memory in global caches for small and medium sizes (<32KiB) + size_t cached; + //! Current amount of memory allocated in huge allocations, i.e larger than LARGE_SIZE_LIMIT which is 2MiB by default (only if ENABLE_STATISTICS=1) + size_t huge_alloc; + //! Peak amount of memory allocated in huge allocations, i.e larger than LARGE_SIZE_LIMIT which is 2MiB by default (only if ENABLE_STATISTICS=1) + size_t huge_alloc_peak; + //! Total amount of memory mapped since initialization (only if ENABLE_STATISTICS=1) + size_t mapped_total; + //! Total amount of memory unmapped since initialization (only if ENABLE_STATISTICS=1) + size_t unmapped_total; +} rpmalloc_global_statistics_t; + +typedef struct rpmalloc_thread_statistics_t { + //! Current number of bytes available in thread size class caches for small and medium sizes (<32KiB) + size_t sizecache; + //! Current number of bytes available in thread span caches for small and medium sizes (<32KiB) + size_t spancache; + //! Total number of bytes transitioned from thread cache to global cache (only if ENABLE_STATISTICS=1) + size_t thread_to_global; + //! Total number of bytes transitioned from global cache to thread cache (only if ENABLE_STATISTICS=1) + size_t global_to_thread; + //! Per span count statistics (only if ENABLE_STATISTICS=1) + struct { + //! Currently used number of spans + size_t current; + //! High water mark of spans used + size_t peak; + //! Number of spans transitioned to global cache + size_t to_global; + //! Number of spans transitioned from global cache + size_t from_global; + //! Number of spans transitioned to thread cache + size_t to_cache; + //! Number of spans transitioned from thread cache + size_t from_cache; + //! Number of spans transitioned to reserved state + size_t to_reserved; + //! Number of spans transitioned from reserved state + size_t from_reserved; + //! Number of raw memory map calls (not hitting the reserve spans but resulting in actual OS mmap calls) + size_t map_calls; + } span_use[64]; + //! Per size class statistics (only if ENABLE_STATISTICS=1) + struct { + //! Current number of allocations + size_t alloc_current; + //! Peak number of allocations + size_t alloc_peak; + //! Total number of allocations + size_t alloc_total; + //! Total number of frees + size_t free_total; + //! Number of spans transitioned to cache + size_t spans_to_cache; + //! Number of spans transitioned from cache + size_t spans_from_cache; + //! Number of spans transitioned from reserved state + size_t spans_from_reserved; + //! Number of raw memory map calls (not hitting the reserve spans but resulting in actual OS mmap calls) + size_t map_calls; + } size_use[128]; +} rpmalloc_thread_statistics_t; + +typedef struct rpmalloc_config_t { + //! Map memory pages for the given number of bytes. The returned address MUST be + // aligned to the rpmalloc span size, which will always be a power of two. + // Optionally the function can store an alignment offset in the offset variable + // in case it performs alignment and the returned pointer is offset from the + // actual start of the memory region due to this alignment. The alignment offset + // will be passed to the memory unmap function. The alignment offset MUST NOT be + // larger than 65535 (storable in an uint16_t), if it is you must use natural + // alignment to shift it into 16 bits. If you set a memory_map function, you + // must also set a memory_unmap function or else the default implementation will + // be used for both. This function must be thread safe, it can be called by + // multiple threads simultaneously. + void* (*memory_map)(size_t size, size_t* offset); + //! Unmap the memory pages starting at address and spanning the given number of bytes. + // If release is set to non-zero, the unmap is for an entire span range as returned by + // a previous call to memory_map and that the entire range should be released. The + // release argument holds the size of the entire span range. If release is set to 0, + // the unmap is a partial decommit of a subset of the mapped memory range. + // If you set a memory_unmap function, you must also set a memory_map function or + // else the default implementation will be used for both. This function must be thread + // safe, it can be called by multiple threads simultaneously. + void (*memory_unmap)(void* address, size_t size, size_t offset, size_t release); + //! Called when an assert fails, if asserts are enabled. Will use the standard assert() + // if this is not set. + void (*error_callback)(const char* message); + //! Called when a call to map memory pages fails (out of memory). If this callback is + // not set or returns zero the library will return a null pointer in the allocation + // call. If this callback returns non-zero the map call will be retried. The argument + // passed is the number of bytes that was requested in the map call. Only used if + // the default system memory map function is used (memory_map callback is not set). + int (*map_fail_callback)(size_t size); + //! Size of memory pages. The page size MUST be a power of two. All memory mapping + // requests to memory_map will be made with size set to a multiple of the page size. + // Used if RPMALLOC_CONFIGURABLE is defined to 1, otherwise system page size is used. + size_t page_size; + //! Size of a span of memory blocks. MUST be a power of two, and in [4096,262144] + // range (unless 0 - set to 0 to use the default span size). Used if RPMALLOC_CONFIGURABLE + // is defined to 1. + size_t span_size; + //! Number of spans to map at each request to map new virtual memory blocks. This can + // be used to minimize the system call overhead at the cost of virtual memory address + // space. The extra mapped pages will not be written until actually used, so physical + // committed memory should not be affected in the default implementation. Will be + // aligned to a multiple of spans that match memory page size in case of huge pages. + size_t span_map_count; + //! Enable use of large/huge pages. If this flag is set to non-zero and page size is + // zero, the allocator will try to enable huge pages and auto detect the configuration. + // If this is set to non-zero and page_size is also non-zero, the allocator will + // assume huge pages have been configured and enabled prior to initializing the + // allocator. + // For Windows, see https://docs.microsoft.com/en-us/windows/desktop/memory/large-page-support + // For Linux, see https://www.kernel.org/doc/Documentation/vm/hugetlbpage.txt + int enable_huge_pages; + //! Respectively allocated pages and huge allocated pages names for systems + // supporting it to be able to distinguish among anonymous regions. + const char *page_name; + const char *huge_page_name; +} rpmalloc_config_t; + +//! Initialize allocator with default configuration +TRACY_API int +rpmalloc_initialize(void); + +//! Initialize allocator with given configuration +RPMALLOC_EXPORT int +rpmalloc_initialize_config(const rpmalloc_config_t* config); + +//! Get allocator configuration +RPMALLOC_EXPORT const rpmalloc_config_t* +rpmalloc_config(void); + +//! Finalize allocator +TRACY_API void +rpmalloc_finalize(void); + +//! Initialize allocator for calling thread +TRACY_API void +rpmalloc_thread_initialize(void); + +//! Finalize allocator for calling thread +TRACY_API void +rpmalloc_thread_finalize(int release_caches); + +//! Perform deferred deallocations pending for the calling thread heap +RPMALLOC_EXPORT void +rpmalloc_thread_collect(void); + +//! Query if allocator is initialized for calling thread +RPMALLOC_EXPORT int +rpmalloc_is_thread_initialized(void); + +//! Get per-thread statistics +RPMALLOC_EXPORT void +rpmalloc_thread_statistics(rpmalloc_thread_statistics_t* stats); + +//! Get global statistics +RPMALLOC_EXPORT void +rpmalloc_global_statistics(rpmalloc_global_statistics_t* stats); + +//! Dump all statistics in human readable format to file (should be a FILE*) +RPMALLOC_EXPORT void +rpmalloc_dump_statistics(void* file); + +//! Allocate a memory block of at least the given size +TRACY_API RPMALLOC_ALLOCATOR void* +rpmalloc(size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE(1); + +//! Free the given memory block +TRACY_API void +rpfree(void* ptr); + +//! Allocate a memory block of at least the given size and zero initialize it +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpcalloc(size_t num, size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE2(1, 2); + +//! Reallocate the given block to at least the given size +TRACY_API RPMALLOC_ALLOCATOR void* +rprealloc(void* ptr, size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE(2); + +//! Reallocate the given block to at least the given size and alignment, +// with optional control flags (see RPMALLOC_NO_PRESERVE). +// Alignment must be a power of two and a multiple of sizeof(void*), +// and should ideally be less than memory page size. A caveat of rpmalloc +// internals is that this must also be strictly less than the span size (default 64KiB) +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpaligned_realloc(void* ptr, size_t alignment, size_t size, size_t oldsize, unsigned int flags) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE(3); + +//! Allocate a memory block of at least the given size and alignment. +// Alignment must be a power of two and a multiple of sizeof(void*), +// and should ideally be less than memory page size. A caveat of rpmalloc +// internals is that this must also be strictly less than the span size (default 64KiB) +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpaligned_alloc(size_t alignment, size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE(2); + +//! Allocate a memory block of at least the given size and alignment, and zero initialize it. +// Alignment must be a power of two and a multiple of sizeof(void*), +// and should ideally be less than memory page size. A caveat of rpmalloc +// internals is that this must also be strictly less than the span size (default 64KiB) +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpaligned_calloc(size_t alignment, size_t num, size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE2(2, 3); + +//! Allocate a memory block of at least the given size and alignment. +// Alignment must be a power of two and a multiple of sizeof(void*), +// and should ideally be less than memory page size. A caveat of rpmalloc +// internals is that this must also be strictly less than the span size (default 64KiB) +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpmemalign(size_t alignment, size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE(2); + +//! Allocate a memory block of at least the given size and alignment. +// Alignment must be a power of two and a multiple of sizeof(void*), +// and should ideally be less than memory page size. A caveat of rpmalloc +// internals is that this must also be strictly less than the span size (default 64KiB) +RPMALLOC_EXPORT int +rpposix_memalign(void** memptr, size_t alignment, size_t size); + +//! Query the usable size of the given memory block (from given pointer to the end of block) +RPMALLOC_EXPORT size_t +rpmalloc_usable_size(void* ptr); + +#if RPMALLOC_FIRST_CLASS_HEAPS + +//! Heap type +typedef struct heap_t rpmalloc_heap_t; + +//! Acquire a new heap. Will reuse existing released heaps or allocate memory for a new heap +// if none available. Heap API is implemented with the strict assumption that only one single +// thread will call heap functions for a given heap at any given time, no functions are thread safe. +RPMALLOC_EXPORT rpmalloc_heap_t* +rpmalloc_heap_acquire(void); + +//! Release a heap (does NOT free the memory allocated by the heap, use rpmalloc_heap_free_all before destroying the heap). +// Releasing a heap will enable it to be reused by other threads. Safe to pass a null pointer. +RPMALLOC_EXPORT void +rpmalloc_heap_release(rpmalloc_heap_t* heap); + +//! Allocate a memory block of at least the given size using the given heap. +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpmalloc_heap_alloc(rpmalloc_heap_t* heap, size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE(2); + +//! Allocate a memory block of at least the given size using the given heap. The returned +// block will have the requested alignment. Alignment must be a power of two and a multiple of sizeof(void*), +// and should ideally be less than memory page size. A caveat of rpmalloc +// internals is that this must also be strictly less than the span size (default 64KiB). +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpmalloc_heap_aligned_alloc(rpmalloc_heap_t* heap, size_t alignment, size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE(3); + +//! Allocate a memory block of at least the given size using the given heap and zero initialize it. +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpmalloc_heap_calloc(rpmalloc_heap_t* heap, size_t num, size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE2(2, 3); + +//! Allocate a memory block of at least the given size using the given heap and zero initialize it. The returned +// block will have the requested alignment. Alignment must either be zero, or a power of two and a multiple of sizeof(void*), +// and should ideally be less than memory page size. A caveat of rpmalloc +// internals is that this must also be strictly less than the span size (default 64KiB). +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpmalloc_heap_aligned_calloc(rpmalloc_heap_t* heap, size_t alignment, size_t num, size_t size) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE2(2, 3); + +//! Reallocate the given block to at least the given size. The memory block MUST be allocated +// by the same heap given to this function. +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpmalloc_heap_realloc(rpmalloc_heap_t* heap, void* ptr, size_t size, unsigned int flags) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE(3); + +//! Reallocate the given block to at least the given size. The memory block MUST be allocated +// by the same heap given to this function. The returned block will have the requested alignment. +// Alignment must be either zero, or a power of two and a multiple of sizeof(void*), and should ideally be +// less than memory page size. A caveat of rpmalloc internals is that this must also be strictly less than +// the span size (default 64KiB). +RPMALLOC_EXPORT RPMALLOC_ALLOCATOR void* +rpmalloc_heap_aligned_realloc(rpmalloc_heap_t* heap, void* ptr, size_t alignment, size_t size, unsigned int flags) RPMALLOC_ATTRIB_MALLOC RPMALLOC_ATTRIB_ALLOC_SIZE(4); + +//! Free the given memory block from the given heap. The memory block MUST be allocated +// by the same heap given to this function. +RPMALLOC_EXPORT void +rpmalloc_heap_free(rpmalloc_heap_t* heap, void* ptr); + +//! Free all memory allocated by the heap +RPMALLOC_EXPORT void +rpmalloc_heap_free_all(rpmalloc_heap_t* heap); + +//! Set the given heap as the current heap for the calling thread. A heap MUST only be current heap +// for a single thread, a heap can never be shared between multiple threads. The previous +// current heap for the calling thread is released to be reused by other threads. +RPMALLOC_EXPORT void +rpmalloc_heap_thread_set_current(rpmalloc_heap_t* heap); + +#endif + +} diff --git a/externals/tracy/public/common/TracyAlign.hpp b/externals/tracy/public/common/TracyAlign.hpp new file mode 100644 index 000000000..c3531ba0d --- /dev/null +++ b/externals/tracy/public/common/TracyAlign.hpp @@ -0,0 +1,27 @@ +#ifndef __TRACYALIGN_HPP__ +#define __TRACYALIGN_HPP__ + +#include + +#include "TracyForceInline.hpp" + +namespace tracy +{ + +template +tracy_force_inline T MemRead( const void* ptr ) +{ + T val; + memcpy( &val, ptr, sizeof( T ) ); + return val; +} + +template +tracy_force_inline void MemWrite( void* ptr, T val ) +{ + memcpy( ptr, &val, sizeof( T ) ); +} + +} + +#endif diff --git a/externals/tracy/public/common/TracyAlloc.hpp b/externals/tracy/public/common/TracyAlloc.hpp new file mode 100644 index 000000000..ddb0e5df6 --- /dev/null +++ b/externals/tracy/public/common/TracyAlloc.hpp @@ -0,0 +1,72 @@ +#ifndef __TRACYALLOC_HPP__ +#define __TRACYALLOC_HPP__ + +#include + +#if defined TRACY_ENABLE && !defined __EMSCRIPTEN__ +# include "TracyApi.h" +# include "TracyForceInline.hpp" +# include "../client/tracy_rpmalloc.hpp" +# define TRACY_USE_RPMALLOC +#endif + +namespace tracy +{ + +#ifdef TRACY_USE_RPMALLOC +TRACY_API void InitRpmalloc(); +#else +static inline void InitRpmalloc() {} +#endif + +static inline void* tracy_malloc( size_t size ) +{ +#ifdef TRACY_USE_RPMALLOC + InitRpmalloc(); + return rpmalloc( size ); +#else + return malloc( size ); +#endif +} + +static inline void* tracy_malloc_fast( size_t size ) +{ +#ifdef TRACY_USE_RPMALLOC + return rpmalloc( size ); +#else + return malloc( size ); +#endif +} + +static inline void tracy_free( void* ptr ) +{ +#ifdef TRACY_USE_RPMALLOC + InitRpmalloc(); + rpfree( ptr ); +#else + free( ptr ); +#endif +} + +static inline void tracy_free_fast( void* ptr ) +{ +#ifdef TRACY_USE_RPMALLOC + rpfree( ptr ); +#else + free( ptr ); +#endif +} + +static inline void* tracy_realloc( void* ptr, size_t size ) +{ +#ifdef TRACY_USE_RPMALLOC + InitRpmalloc(); + return rprealloc( ptr, size ); +#else + return realloc( ptr, size ); +#endif +} + +} + +#endif diff --git a/externals/tracy/public/common/TracyApi.h b/externals/tracy/public/common/TracyApi.h new file mode 100644 index 000000000..f396ce0c6 --- /dev/null +++ b/externals/tracy/public/common/TracyApi.h @@ -0,0 +1,16 @@ +#ifndef __TRACYAPI_H__ +#define __TRACYAPI_H__ + +#if defined _WIN32 +# if defined TRACY_EXPORTS +# define TRACY_API __declspec(dllexport) +# elif defined TRACY_IMPORTS +# define TRACY_API __declspec(dllimport) +# else +# define TRACY_API +# endif +#else +# define TRACY_API __attribute__((visibility("default"))) +#endif + +#endif // __TRACYAPI_H__ diff --git a/externals/tracy/public/common/TracyColor.hpp b/externals/tracy/public/common/TracyColor.hpp new file mode 100644 index 000000000..4825c0fba --- /dev/null +++ b/externals/tracy/public/common/TracyColor.hpp @@ -0,0 +1,690 @@ +#ifndef __TRACYCOLOR_HPP__ +#define __TRACYCOLOR_HPP__ + +namespace tracy +{ +struct Color +{ +enum ColorType +{ + Snow = 0xfffafa, + GhostWhite = 0xf8f8ff, + WhiteSmoke = 0xf5f5f5, + Gainsboro = 0xdcdcdc, + FloralWhite = 0xfffaf0, + OldLace = 0xfdf5e6, + Linen = 0xfaf0e6, + AntiqueWhite = 0xfaebd7, + PapayaWhip = 0xffefd5, + BlanchedAlmond = 0xffebcd, + Bisque = 0xffe4c4, + PeachPuff = 0xffdab9, + NavajoWhite = 0xffdead, + Moccasin = 0xffe4b5, + Cornsilk = 0xfff8dc, + Ivory = 0xfffff0, + LemonChiffon = 0xfffacd, + Seashell = 0xfff5ee, + Honeydew = 0xf0fff0, + MintCream = 0xf5fffa, + Azure = 0xf0ffff, + AliceBlue = 0xf0f8ff, + Lavender = 0xe6e6fa, + LavenderBlush = 0xfff0f5, + MistyRose = 0xffe4e1, + White = 0xffffff, + Black = 0x000000, + DarkSlateGray = 0x2f4f4f, + DarkSlateGrey = 0x2f4f4f, + DimGray = 0x696969, + DimGrey = 0x696969, + SlateGray = 0x708090, + SlateGrey = 0x708090, + LightSlateGray = 0x778899, + LightSlateGrey = 0x778899, + Gray = 0xbebebe, + Grey = 0xbebebe, + X11Gray = 0xbebebe, + X11Grey = 0xbebebe, + WebGray = 0x808080, + WebGrey = 0x808080, + LightGrey = 0xd3d3d3, + LightGray = 0xd3d3d3, + MidnightBlue = 0x191970, + Navy = 0x000080, + NavyBlue = 0x000080, + CornflowerBlue = 0x6495ed, + DarkSlateBlue = 0x483d8b, + SlateBlue = 0x6a5acd, + MediumSlateBlue = 0x7b68ee, + LightSlateBlue = 0x8470ff, + MediumBlue = 0x0000cd, + RoyalBlue = 0x4169e1, + Blue = 0x0000ff, + DodgerBlue = 0x1e90ff, + DeepSkyBlue = 0x00bfff, + SkyBlue = 0x87ceeb, + LightSkyBlue = 0x87cefa, + SteelBlue = 0x4682b4, + LightSteelBlue = 0xb0c4de, + LightBlue = 0xadd8e6, + PowderBlue = 0xb0e0e6, + PaleTurquoise = 0xafeeee, + DarkTurquoise = 0x00ced1, + MediumTurquoise = 0x48d1cc, + Turquoise = 0x40e0d0, + Cyan = 0x00ffff, + Aqua = 0x00ffff, + LightCyan = 0xe0ffff, + CadetBlue = 0x5f9ea0, + MediumAquamarine = 0x66cdaa, + Aquamarine = 0x7fffd4, + DarkGreen = 0x006400, + DarkOliveGreen = 0x556b2f, + DarkSeaGreen = 0x8fbc8f, + SeaGreen = 0x2e8b57, + MediumSeaGreen = 0x3cb371, + LightSeaGreen = 0x20b2aa, + PaleGreen = 0x98fb98, + SpringGreen = 0x00ff7f, + LawnGreen = 0x7cfc00, + Green = 0x00ff00, + Lime = 0x00ff00, + X11Green = 0x00ff00, + WebGreen = 0x008000, + Chartreuse = 0x7fff00, + MediumSpringGreen = 0x00fa9a, + GreenYellow = 0xadff2f, + LimeGreen = 0x32cd32, + YellowGreen = 0x9acd32, + ForestGreen = 0x228b22, + OliveDrab = 0x6b8e23, + DarkKhaki = 0xbdb76b, + Khaki = 0xf0e68c, + PaleGoldenrod = 0xeee8aa, + LightGoldenrodYellow = 0xfafad2, + LightYellow = 0xffffe0, + Yellow = 0xffff00, + Gold = 0xffd700, + LightGoldenrod = 0xeedd82, + Goldenrod = 0xdaa520, + DarkGoldenrod = 0xb8860b, + RosyBrown = 0xbc8f8f, + IndianRed = 0xcd5c5c, + SaddleBrown = 0x8b4513, + Sienna = 0xa0522d, + Peru = 0xcd853f, + Burlywood = 0xdeb887, + Beige = 0xf5f5dc, + Wheat = 0xf5deb3, + SandyBrown = 0xf4a460, + Tan = 0xd2b48c, + Chocolate = 0xd2691e, + Firebrick = 0xb22222, + Brown = 0xa52a2a, + DarkSalmon = 0xe9967a, + Salmon = 0xfa8072, + LightSalmon = 0xffa07a, + Orange = 0xffa500, + DarkOrange = 0xff8c00, + Coral = 0xff7f50, + LightCoral = 0xf08080, + Tomato = 0xff6347, + OrangeRed = 0xff4500, + Red = 0xff0000, + HotPink = 0xff69b4, + DeepPink = 0xff1493, + Pink = 0xffc0cb, + LightPink = 0xffb6c1, + PaleVioletRed = 0xdb7093, + Maroon = 0xb03060, + X11Maroon = 0xb03060, + WebMaroon = 0x800000, + MediumVioletRed = 0xc71585, + VioletRed = 0xd02090, + Magenta = 0xff00ff, + Fuchsia = 0xff00ff, + Violet = 0xee82ee, + Plum = 0xdda0dd, + Orchid = 0xda70d6, + MediumOrchid = 0xba55d3, + DarkOrchid = 0x9932cc, + DarkViolet = 0x9400d3, + BlueViolet = 0x8a2be2, + Purple = 0xa020f0, + X11Purple = 0xa020f0, + WebPurple = 0x800080, + MediumPurple = 0x9370db, + Thistle = 0xd8bfd8, + Snow1 = 0xfffafa, + Snow2 = 0xeee9e9, + Snow3 = 0xcdc9c9, + Snow4 = 0x8b8989, + Seashell1 = 0xfff5ee, + Seashell2 = 0xeee5de, + Seashell3 = 0xcdc5bf, + Seashell4 = 0x8b8682, + AntiqueWhite1 = 0xffefdb, + AntiqueWhite2 = 0xeedfcc, + AntiqueWhite3 = 0xcdc0b0, + AntiqueWhite4 = 0x8b8378, + Bisque1 = 0xffe4c4, + Bisque2 = 0xeed5b7, + Bisque3 = 0xcdb79e, + Bisque4 = 0x8b7d6b, + PeachPuff1 = 0xffdab9, + PeachPuff2 = 0xeecbad, + PeachPuff3 = 0xcdaf95, + PeachPuff4 = 0x8b7765, + NavajoWhite1 = 0xffdead, + NavajoWhite2 = 0xeecfa1, + NavajoWhite3 = 0xcdb38b, + NavajoWhite4 = 0x8b795e, + LemonChiffon1 = 0xfffacd, + LemonChiffon2 = 0xeee9bf, + LemonChiffon3 = 0xcdc9a5, + LemonChiffon4 = 0x8b8970, + Cornsilk1 = 0xfff8dc, + Cornsilk2 = 0xeee8cd, + Cornsilk3 = 0xcdc8b1, + Cornsilk4 = 0x8b8878, + Ivory1 = 0xfffff0, + Ivory2 = 0xeeeee0, + Ivory3 = 0xcdcdc1, + Ivory4 = 0x8b8b83, + Honeydew1 = 0xf0fff0, + Honeydew2 = 0xe0eee0, + Honeydew3 = 0xc1cdc1, + Honeydew4 = 0x838b83, + LavenderBlush1 = 0xfff0f5, + LavenderBlush2 = 0xeee0e5, + LavenderBlush3 = 0xcdc1c5, + LavenderBlush4 = 0x8b8386, + MistyRose1 = 0xffe4e1, + MistyRose2 = 0xeed5d2, + MistyRose3 = 0xcdb7b5, + MistyRose4 = 0x8b7d7b, + Azure1 = 0xf0ffff, + Azure2 = 0xe0eeee, + Azure3 = 0xc1cdcd, + Azure4 = 0x838b8b, + SlateBlue1 = 0x836fff, + SlateBlue2 = 0x7a67ee, + SlateBlue3 = 0x6959cd, + SlateBlue4 = 0x473c8b, + RoyalBlue1 = 0x4876ff, + RoyalBlue2 = 0x436eee, + RoyalBlue3 = 0x3a5fcd, + RoyalBlue4 = 0x27408b, + Blue1 = 0x0000ff, + Blue2 = 0x0000ee, + Blue3 = 0x0000cd, + Blue4 = 0x00008b, + DodgerBlue1 = 0x1e90ff, + DodgerBlue2 = 0x1c86ee, + DodgerBlue3 = 0x1874cd, + DodgerBlue4 = 0x104e8b, + SteelBlue1 = 0x63b8ff, + SteelBlue2 = 0x5cacee, + SteelBlue3 = 0x4f94cd, + SteelBlue4 = 0x36648b, + DeepSkyBlue1 = 0x00bfff, + DeepSkyBlue2 = 0x00b2ee, + DeepSkyBlue3 = 0x009acd, + DeepSkyBlue4 = 0x00688b, + SkyBlue1 = 0x87ceff, + SkyBlue2 = 0x7ec0ee, + SkyBlue3 = 0x6ca6cd, + SkyBlue4 = 0x4a708b, + LightSkyBlue1 = 0xb0e2ff, + LightSkyBlue2 = 0xa4d3ee, + LightSkyBlue3 = 0x8db6cd, + LightSkyBlue4 = 0x607b8b, + SlateGray1 = 0xc6e2ff, + SlateGray2 = 0xb9d3ee, + SlateGray3 = 0x9fb6cd, + SlateGray4 = 0x6c7b8b, + LightSteelBlue1 = 0xcae1ff, + LightSteelBlue2 = 0xbcd2ee, + LightSteelBlue3 = 0xa2b5cd, + LightSteelBlue4 = 0x6e7b8b, + LightBlue1 = 0xbfefff, + LightBlue2 = 0xb2dfee, + LightBlue3 = 0x9ac0cd, + LightBlue4 = 0x68838b, + LightCyan1 = 0xe0ffff, + LightCyan2 = 0xd1eeee, + LightCyan3 = 0xb4cdcd, + LightCyan4 = 0x7a8b8b, + PaleTurquoise1 = 0xbbffff, + PaleTurquoise2 = 0xaeeeee, + PaleTurquoise3 = 0x96cdcd, + PaleTurquoise4 = 0x668b8b, + CadetBlue1 = 0x98f5ff, + CadetBlue2 = 0x8ee5ee, + CadetBlue3 = 0x7ac5cd, + CadetBlue4 = 0x53868b, + Turquoise1 = 0x00f5ff, + Turquoise2 = 0x00e5ee, + Turquoise3 = 0x00c5cd, + Turquoise4 = 0x00868b, + Cyan1 = 0x00ffff, + Cyan2 = 0x00eeee, + Cyan3 = 0x00cdcd, + Cyan4 = 0x008b8b, + DarkSlateGray1 = 0x97ffff, + DarkSlateGray2 = 0x8deeee, + DarkSlateGray3 = 0x79cdcd, + DarkSlateGray4 = 0x528b8b, + Aquamarine1 = 0x7fffd4, + Aquamarine2 = 0x76eec6, + Aquamarine3 = 0x66cdaa, + Aquamarine4 = 0x458b74, + DarkSeaGreen1 = 0xc1ffc1, + DarkSeaGreen2 = 0xb4eeb4, + DarkSeaGreen3 = 0x9bcd9b, + DarkSeaGreen4 = 0x698b69, + SeaGreen1 = 0x54ff9f, + SeaGreen2 = 0x4eee94, + SeaGreen3 = 0x43cd80, + SeaGreen4 = 0x2e8b57, + PaleGreen1 = 0x9aff9a, + PaleGreen2 = 0x90ee90, + PaleGreen3 = 0x7ccd7c, + PaleGreen4 = 0x548b54, + SpringGreen1 = 0x00ff7f, + SpringGreen2 = 0x00ee76, + SpringGreen3 = 0x00cd66, + SpringGreen4 = 0x008b45, + Green1 = 0x00ff00, + Green2 = 0x00ee00, + Green3 = 0x00cd00, + Green4 = 0x008b00, + Chartreuse1 = 0x7fff00, + Chartreuse2 = 0x76ee00, + Chartreuse3 = 0x66cd00, + Chartreuse4 = 0x458b00, + OliveDrab1 = 0xc0ff3e, + OliveDrab2 = 0xb3ee3a, + OliveDrab3 = 0x9acd32, + OliveDrab4 = 0x698b22, + DarkOliveGreen1 = 0xcaff70, + DarkOliveGreen2 = 0xbcee68, + DarkOliveGreen3 = 0xa2cd5a, + DarkOliveGreen4 = 0x6e8b3d, + Khaki1 = 0xfff68f, + Khaki2 = 0xeee685, + Khaki3 = 0xcdc673, + Khaki4 = 0x8b864e, + LightGoldenrod1 = 0xffec8b, + LightGoldenrod2 = 0xeedc82, + LightGoldenrod3 = 0xcdbe70, + LightGoldenrod4 = 0x8b814c, + LightYellow1 = 0xffffe0, + LightYellow2 = 0xeeeed1, + LightYellow3 = 0xcdcdb4, + LightYellow4 = 0x8b8b7a, + Yellow1 = 0xffff00, + Yellow2 = 0xeeee00, + Yellow3 = 0xcdcd00, + Yellow4 = 0x8b8b00, + Gold1 = 0xffd700, + Gold2 = 0xeec900, + Gold3 = 0xcdad00, + Gold4 = 0x8b7500, + Goldenrod1 = 0xffc125, + Goldenrod2 = 0xeeb422, + Goldenrod3 = 0xcd9b1d, + Goldenrod4 = 0x8b6914, + DarkGoldenrod1 = 0xffb90f, + DarkGoldenrod2 = 0xeead0e, + DarkGoldenrod3 = 0xcd950c, + DarkGoldenrod4 = 0x8b6508, + RosyBrown1 = 0xffc1c1, + RosyBrown2 = 0xeeb4b4, + RosyBrown3 = 0xcd9b9b, + RosyBrown4 = 0x8b6969, + IndianRed1 = 0xff6a6a, + IndianRed2 = 0xee6363, + IndianRed3 = 0xcd5555, + IndianRed4 = 0x8b3a3a, + Sienna1 = 0xff8247, + Sienna2 = 0xee7942, + Sienna3 = 0xcd6839, + Sienna4 = 0x8b4726, + Burlywood1 = 0xffd39b, + Burlywood2 = 0xeec591, + Burlywood3 = 0xcdaa7d, + Burlywood4 = 0x8b7355, + Wheat1 = 0xffe7ba, + Wheat2 = 0xeed8ae, + Wheat3 = 0xcdba96, + Wheat4 = 0x8b7e66, + Tan1 = 0xffa54f, + Tan2 = 0xee9a49, + Tan3 = 0xcd853f, + Tan4 = 0x8b5a2b, + Chocolate1 = 0xff7f24, + Chocolate2 = 0xee7621, + Chocolate3 = 0xcd661d, + Chocolate4 = 0x8b4513, + Firebrick1 = 0xff3030, + Firebrick2 = 0xee2c2c, + Firebrick3 = 0xcd2626, + Firebrick4 = 0x8b1a1a, + Brown1 = 0xff4040, + Brown2 = 0xee3b3b, + Brown3 = 0xcd3333, + Brown4 = 0x8b2323, + Salmon1 = 0xff8c69, + Salmon2 = 0xee8262, + Salmon3 = 0xcd7054, + Salmon4 = 0x8b4c39, + LightSalmon1 = 0xffa07a, + LightSalmon2 = 0xee9572, + LightSalmon3 = 0xcd8162, + LightSalmon4 = 0x8b5742, + Orange1 = 0xffa500, + Orange2 = 0xee9a00, + Orange3 = 0xcd8500, + Orange4 = 0x8b5a00, + DarkOrange1 = 0xff7f00, + DarkOrange2 = 0xee7600, + DarkOrange3 = 0xcd6600, + DarkOrange4 = 0x8b4500, + Coral1 = 0xff7256, + Coral2 = 0xee6a50, + Coral3 = 0xcd5b45, + Coral4 = 0x8b3e2f, + Tomato1 = 0xff6347, + Tomato2 = 0xee5c42, + Tomato3 = 0xcd4f39, + Tomato4 = 0x8b3626, + OrangeRed1 = 0xff4500, + OrangeRed2 = 0xee4000, + OrangeRed3 = 0xcd3700, + OrangeRed4 = 0x8b2500, + Red1 = 0xff0000, + Red2 = 0xee0000, + Red3 = 0xcd0000, + Red4 = 0x8b0000, + DeepPink1 = 0xff1493, + DeepPink2 = 0xee1289, + DeepPink3 = 0xcd1076, + DeepPink4 = 0x8b0a50, + HotPink1 = 0xff6eb4, + HotPink2 = 0xee6aa7, + HotPink3 = 0xcd6090, + HotPink4 = 0x8b3a62, + Pink1 = 0xffb5c5, + Pink2 = 0xeea9b8, + Pink3 = 0xcd919e, + Pink4 = 0x8b636c, + LightPink1 = 0xffaeb9, + LightPink2 = 0xeea2ad, + LightPink3 = 0xcd8c95, + LightPink4 = 0x8b5f65, + PaleVioletRed1 = 0xff82ab, + PaleVioletRed2 = 0xee799f, + PaleVioletRed3 = 0xcd6889, + PaleVioletRed4 = 0x8b475d, + Maroon1 = 0xff34b3, + Maroon2 = 0xee30a7, + Maroon3 = 0xcd2990, + Maroon4 = 0x8b1c62, + VioletRed1 = 0xff3e96, + VioletRed2 = 0xee3a8c, + VioletRed3 = 0xcd3278, + VioletRed4 = 0x8b2252, + Magenta1 = 0xff00ff, + Magenta2 = 0xee00ee, + Magenta3 = 0xcd00cd, + Magenta4 = 0x8b008b, + Orchid1 = 0xff83fa, + Orchid2 = 0xee7ae9, + Orchid3 = 0xcd69c9, + Orchid4 = 0x8b4789, + Plum1 = 0xffbbff, + Plum2 = 0xeeaeee, + Plum3 = 0xcd96cd, + Plum4 = 0x8b668b, + MediumOrchid1 = 0xe066ff, + MediumOrchid2 = 0xd15fee, + MediumOrchid3 = 0xb452cd, + MediumOrchid4 = 0x7a378b, + DarkOrchid1 = 0xbf3eff, + DarkOrchid2 = 0xb23aee, + DarkOrchid3 = 0x9a32cd, + DarkOrchid4 = 0x68228b, + Purple1 = 0x9b30ff, + Purple2 = 0x912cee, + Purple3 = 0x7d26cd, + Purple4 = 0x551a8b, + MediumPurple1 = 0xab82ff, + MediumPurple2 = 0x9f79ee, + MediumPurple3 = 0x8968cd, + MediumPurple4 = 0x5d478b, + Thistle1 = 0xffe1ff, + Thistle2 = 0xeed2ee, + Thistle3 = 0xcdb5cd, + Thistle4 = 0x8b7b8b, + Gray0 = 0x000000, + Grey0 = 0x000000, + Gray1 = 0x030303, + Grey1 = 0x030303, + Gray2 = 0x050505, + Grey2 = 0x050505, + Gray3 = 0x080808, + Grey3 = 0x080808, + Gray4 = 0x0a0a0a, + Grey4 = 0x0a0a0a, + Gray5 = 0x0d0d0d, + Grey5 = 0x0d0d0d, + Gray6 = 0x0f0f0f, + Grey6 = 0x0f0f0f, + Gray7 = 0x121212, + Grey7 = 0x121212, + Gray8 = 0x141414, + Grey8 = 0x141414, + Gray9 = 0x171717, + Grey9 = 0x171717, + Gray10 = 0x1a1a1a, + Grey10 = 0x1a1a1a, + Gray11 = 0x1c1c1c, + Grey11 = 0x1c1c1c, + Gray12 = 0x1f1f1f, + Grey12 = 0x1f1f1f, + Gray13 = 0x212121, + Grey13 = 0x212121, + Gray14 = 0x242424, + Grey14 = 0x242424, + Gray15 = 0x262626, + Grey15 = 0x262626, + Gray16 = 0x292929, + Grey16 = 0x292929, + Gray17 = 0x2b2b2b, + Grey17 = 0x2b2b2b, + Gray18 = 0x2e2e2e, + Grey18 = 0x2e2e2e, + Gray19 = 0x303030, + Grey19 = 0x303030, + Gray20 = 0x333333, + Grey20 = 0x333333, + Gray21 = 0x363636, + Grey21 = 0x363636, + Gray22 = 0x383838, + Grey22 = 0x383838, + Gray23 = 0x3b3b3b, + Grey23 = 0x3b3b3b, + Gray24 = 0x3d3d3d, + Grey24 = 0x3d3d3d, + Gray25 = 0x404040, + Grey25 = 0x404040, + Gray26 = 0x424242, + Grey26 = 0x424242, + Gray27 = 0x454545, + Grey27 = 0x454545, + Gray28 = 0x474747, + Grey28 = 0x474747, + Gray29 = 0x4a4a4a, + Grey29 = 0x4a4a4a, + Gray30 = 0x4d4d4d, + Grey30 = 0x4d4d4d, + Gray31 = 0x4f4f4f, + Grey31 = 0x4f4f4f, + Gray32 = 0x525252, + Grey32 = 0x525252, + Gray33 = 0x545454, + Grey33 = 0x545454, + Gray34 = 0x575757, + Grey34 = 0x575757, + Gray35 = 0x595959, + Grey35 = 0x595959, + Gray36 = 0x5c5c5c, + Grey36 = 0x5c5c5c, + Gray37 = 0x5e5e5e, + Grey37 = 0x5e5e5e, + Gray38 = 0x616161, + Grey38 = 0x616161, + Gray39 = 0x636363, + Grey39 = 0x636363, + Gray40 = 0x666666, + Grey40 = 0x666666, + Gray41 = 0x696969, + Grey41 = 0x696969, + Gray42 = 0x6b6b6b, + Grey42 = 0x6b6b6b, + Gray43 = 0x6e6e6e, + Grey43 = 0x6e6e6e, + Gray44 = 0x707070, + Grey44 = 0x707070, + Gray45 = 0x737373, + Grey45 = 0x737373, + Gray46 = 0x757575, + Grey46 = 0x757575, + Gray47 = 0x787878, + Grey47 = 0x787878, + Gray48 = 0x7a7a7a, + Grey48 = 0x7a7a7a, + Gray49 = 0x7d7d7d, + Grey49 = 0x7d7d7d, + Gray50 = 0x7f7f7f, + Grey50 = 0x7f7f7f, + Gray51 = 0x828282, + Grey51 = 0x828282, + Gray52 = 0x858585, + Grey52 = 0x858585, + Gray53 = 0x878787, + Grey53 = 0x878787, + Gray54 = 0x8a8a8a, + Grey54 = 0x8a8a8a, + Gray55 = 0x8c8c8c, + Grey55 = 0x8c8c8c, + Gray56 = 0x8f8f8f, + Grey56 = 0x8f8f8f, + Gray57 = 0x919191, + Grey57 = 0x919191, + Gray58 = 0x949494, + Grey58 = 0x949494, + Gray59 = 0x969696, + Grey59 = 0x969696, + Gray60 = 0x999999, + Grey60 = 0x999999, + Gray61 = 0x9c9c9c, + Grey61 = 0x9c9c9c, + Gray62 = 0x9e9e9e, + Grey62 = 0x9e9e9e, + Gray63 = 0xa1a1a1, + Grey63 = 0xa1a1a1, + Gray64 = 0xa3a3a3, + Grey64 = 0xa3a3a3, + Gray65 = 0xa6a6a6, + Grey65 = 0xa6a6a6, + Gray66 = 0xa8a8a8, + Grey66 = 0xa8a8a8, + Gray67 = 0xababab, + Grey67 = 0xababab, + Gray68 = 0xadadad, + Grey68 = 0xadadad, + Gray69 = 0xb0b0b0, + Grey69 = 0xb0b0b0, + Gray70 = 0xb3b3b3, + Grey70 = 0xb3b3b3, + Gray71 = 0xb5b5b5, + Grey71 = 0xb5b5b5, + Gray72 = 0xb8b8b8, + Grey72 = 0xb8b8b8, + Gray73 = 0xbababa, + Grey73 = 0xbababa, + Gray74 = 0xbdbdbd, + Grey74 = 0xbdbdbd, + Gray75 = 0xbfbfbf, + Grey75 = 0xbfbfbf, + Gray76 = 0xc2c2c2, + Grey76 = 0xc2c2c2, + Gray77 = 0xc4c4c4, + Grey77 = 0xc4c4c4, + Gray78 = 0xc7c7c7, + Grey78 = 0xc7c7c7, + Gray79 = 0xc9c9c9, + Grey79 = 0xc9c9c9, + Gray80 = 0xcccccc, + Grey80 = 0xcccccc, + Gray81 = 0xcfcfcf, + Grey81 = 0xcfcfcf, + Gray82 = 0xd1d1d1, + Grey82 = 0xd1d1d1, + Gray83 = 0xd4d4d4, + Grey83 = 0xd4d4d4, + Gray84 = 0xd6d6d6, + Grey84 = 0xd6d6d6, + Gray85 = 0xd9d9d9, + Grey85 = 0xd9d9d9, + Gray86 = 0xdbdbdb, + Grey86 = 0xdbdbdb, + Gray87 = 0xdedede, + Grey87 = 0xdedede, + Gray88 = 0xe0e0e0, + Grey88 = 0xe0e0e0, + Gray89 = 0xe3e3e3, + Grey89 = 0xe3e3e3, + Gray90 = 0xe5e5e5, + Grey90 = 0xe5e5e5, + Gray91 = 0xe8e8e8, + Grey91 = 0xe8e8e8, + Gray92 = 0xebebeb, + Grey92 = 0xebebeb, + Gray93 = 0xededed, + Grey93 = 0xededed, + Gray94 = 0xf0f0f0, + Grey94 = 0xf0f0f0, + Gray95 = 0xf2f2f2, + Grey95 = 0xf2f2f2, + Gray96 = 0xf5f5f5, + Grey96 = 0xf5f5f5, + Gray97 = 0xf7f7f7, + Grey97 = 0xf7f7f7, + Gray98 = 0xfafafa, + Grey98 = 0xfafafa, + Gray99 = 0xfcfcfc, + Grey99 = 0xfcfcfc, + Gray100 = 0xffffff, + Grey100 = 0xffffff, + DarkGrey = 0xa9a9a9, + DarkGray = 0xa9a9a9, + DarkBlue = 0x00008b, + DarkCyan = 0x008b8b, + DarkMagenta = 0x8b008b, + DarkRed = 0x8b0000, + LightGreen = 0x90ee90, + Crimson = 0xdc143c, + Indigo = 0x4b0082, + Olive = 0x808000, + RebeccaPurple = 0x663399, + Silver = 0xc0c0c0, + Teal = 0x008080, +}; +}; +} + +#endif diff --git a/externals/tracy/public/common/TracyForceInline.hpp b/externals/tracy/public/common/TracyForceInline.hpp new file mode 100644 index 000000000..b6a5833e5 --- /dev/null +++ b/externals/tracy/public/common/TracyForceInline.hpp @@ -0,0 +1,20 @@ +#ifndef __TRACYFORCEINLINE_HPP__ +#define __TRACYFORCEINLINE_HPP__ + +#if defined(__GNUC__) +# define tracy_force_inline __attribute__((always_inline)) inline +#elif defined(_MSC_VER) +# define tracy_force_inline __forceinline +#else +# define tracy_force_inline inline +#endif + +#if defined(__GNUC__) +# define tracy_no_inline __attribute__((noinline)) +#elif defined(_MSC_VER) +# define tracy_no_inline __declspec(noinline) +#else +# define tracy_no_inline +#endif + +#endif diff --git a/externals/tracy/public/common/TracyMutex.hpp b/externals/tracy/public/common/TracyMutex.hpp new file mode 100644 index 000000000..57fb01a0c --- /dev/null +++ b/externals/tracy/public/common/TracyMutex.hpp @@ -0,0 +1,24 @@ +#ifndef __TRACYMUTEX_HPP__ +#define __TRACYMUTEX_HPP__ + +#if defined _MSC_VER + +# include + +namespace tracy +{ +using TracyMutex = std::shared_mutex; +} + +#else + +#include + +namespace tracy +{ +using TracyMutex = std::mutex; +} + +#endif + +#endif diff --git a/externals/tracy/public/common/TracyProtocol.hpp b/externals/tracy/public/common/TracyProtocol.hpp new file mode 100644 index 000000000..51399b09c --- /dev/null +++ b/externals/tracy/public/common/TracyProtocol.hpp @@ -0,0 +1,169 @@ +#ifndef __TRACYPROTOCOL_HPP__ +#define __TRACYPROTOCOL_HPP__ + +#include +#include + +namespace tracy +{ + +constexpr unsigned Lz4CompressBound( unsigned isize ) { return isize + ( isize / 255 ) + 16; } + +enum : uint32_t { ProtocolVersion = 65 }; +enum : uint16_t { BroadcastVersion = 3 }; + +using lz4sz_t = uint32_t; + +enum { TargetFrameSize = 256 * 1024 }; +enum { LZ4Size = Lz4CompressBound( TargetFrameSize ) }; +static_assert( LZ4Size <= (std::numeric_limits::max)(), "LZ4Size greater than lz4sz_t" ); +static_assert( TargetFrameSize * 2 >= 64 * 1024, "Not enough space for LZ4 stream buffer" ); + +enum { HandshakeShibbolethSize = 8 }; +static const char HandshakeShibboleth[HandshakeShibbolethSize] = { 'T', 'r', 'a', 'c', 'y', 'P', 'r', 'f' }; + +enum HandshakeStatus : uint8_t +{ + HandshakePending, + HandshakeWelcome, + HandshakeProtocolMismatch, + HandshakeNotAvailable, + HandshakeDropped +}; + +enum { WelcomeMessageProgramNameSize = 64 }; +enum { WelcomeMessageHostInfoSize = 1024 }; + +#pragma pack( push, 1 ) + +// Must increase left query space after handling! +enum ServerQuery : uint8_t +{ + ServerQueryTerminate, + ServerQueryString, + ServerQueryThreadString, + ServerQuerySourceLocation, + ServerQueryPlotName, + ServerQueryFrameName, + ServerQueryParameter, + ServerQueryFiberName, + // Items above are high priority. Split order must be preserved. See IsQueryPrio(). + ServerQueryDisconnect, + ServerQueryCallstackFrame, + ServerQueryExternalName, + ServerQuerySymbol, + ServerQuerySymbolCode, + ServerQuerySourceCode, + ServerQueryDataTransfer, + ServerQueryDataTransferPart +}; + +struct ServerQueryPacket +{ + ServerQuery type; + uint64_t ptr; + uint32_t extra; +}; + +enum { ServerQueryPacketSize = sizeof( ServerQueryPacket ) }; + + +enum CpuArchitecture : uint8_t +{ + CpuArchUnknown, + CpuArchX86, + CpuArchX64, + CpuArchArm32, + CpuArchArm64 +}; + + +struct WelcomeFlag +{ + enum _t : uint8_t + { + OnDemand = 1 << 0, + IsApple = 1 << 1, + CodeTransfer = 1 << 2, + CombineSamples = 1 << 3, + IdentifySamples = 1 << 4, + }; +}; + +struct WelcomeMessage +{ + double timerMul; + int64_t initBegin; + int64_t initEnd; + uint64_t delay; + uint64_t resolution; + uint64_t epoch; + uint64_t exectime; + uint64_t pid; + int64_t samplingPeriod; + uint8_t flags; + uint8_t cpuArch; + char cpuManufacturer[12]; + uint32_t cpuId; + char programName[WelcomeMessageProgramNameSize]; + char hostInfo[WelcomeMessageHostInfoSize]; +}; + +enum { WelcomeMessageSize = sizeof( WelcomeMessage ) }; + + +struct OnDemandPayloadMessage +{ + uint64_t frames; + uint64_t currentTime; +}; + +enum { OnDemandPayloadMessageSize = sizeof( OnDemandPayloadMessage ) }; + + +struct BroadcastMessage +{ + uint16_t broadcastVersion; + uint16_t listenPort; + uint32_t protocolVersion; + uint64_t pid; + int32_t activeTime; // in seconds + char programName[WelcomeMessageProgramNameSize]; +}; + +struct BroadcastMessage_v2 +{ + uint16_t broadcastVersion; + uint16_t listenPort; + uint32_t protocolVersion; + int32_t activeTime; + char programName[WelcomeMessageProgramNameSize]; +}; + +struct BroadcastMessage_v1 +{ + uint32_t broadcastVersion; + uint32_t protocolVersion; + uint32_t listenPort; + uint32_t activeTime; + char programName[WelcomeMessageProgramNameSize]; +}; + +struct BroadcastMessage_v0 +{ + uint32_t broadcastVersion; + uint32_t protocolVersion; + uint32_t activeTime; + char programName[WelcomeMessageProgramNameSize]; +}; + +enum { BroadcastMessageSize = sizeof( BroadcastMessage ) }; +enum { BroadcastMessageSize_v2 = sizeof( BroadcastMessage_v2 ) }; +enum { BroadcastMessageSize_v1 = sizeof( BroadcastMessage_v1 ) }; +enum { BroadcastMessageSize_v0 = sizeof( BroadcastMessage_v0 ) }; + +#pragma pack( pop ) + +} + +#endif diff --git a/externals/tracy/public/common/TracyQueue.hpp b/externals/tracy/public/common/TracyQueue.hpp new file mode 100644 index 000000000..df886e416 --- /dev/null +++ b/externals/tracy/public/common/TracyQueue.hpp @@ -0,0 +1,894 @@ +#ifndef __TRACYQUEUE_HPP__ +#define __TRACYQUEUE_HPP__ + +#include +#include + +namespace tracy +{ + +enum class QueueType : uint8_t +{ + ZoneText, + ZoneName, + Message, + MessageColor, + MessageCallstack, + MessageColorCallstack, + MessageAppInfo, + ZoneBeginAllocSrcLoc, + ZoneBeginAllocSrcLocCallstack, + CallstackSerial, + Callstack, + CallstackAlloc, + CallstackSample, + CallstackSampleContextSwitch, + FrameImage, + ZoneBegin, + ZoneBeginCallstack, + ZoneEnd, + LockWait, + LockObtain, + LockRelease, + LockSharedWait, + LockSharedObtain, + LockSharedRelease, + LockName, + MemAlloc, + MemAllocNamed, + MemFree, + MemFreeNamed, + MemAllocCallstack, + MemAllocCallstackNamed, + MemFreeCallstack, + MemFreeCallstackNamed, + GpuZoneBegin, + GpuZoneBeginCallstack, + GpuZoneBeginAllocSrcLoc, + GpuZoneBeginAllocSrcLocCallstack, + GpuZoneEnd, + GpuZoneBeginSerial, + GpuZoneBeginCallstackSerial, + GpuZoneBeginAllocSrcLocSerial, + GpuZoneBeginAllocSrcLocCallstackSerial, + GpuZoneEndSerial, + PlotDataInt, + PlotDataFloat, + PlotDataDouble, + ContextSwitch, + ThreadWakeup, + GpuTime, + GpuContextName, + CallstackFrameSize, + SymbolInformation, + ExternalNameMetadata, + SymbolCodeMetadata, + SourceCodeMetadata, + FiberEnter, + FiberLeave, + Terminate, + KeepAlive, + ThreadContext, + GpuCalibration, + GpuTimeSync, + Crash, + CrashReport, + ZoneValidation, + ZoneColor, + ZoneValue, + FrameMarkMsg, + FrameMarkMsgStart, + FrameMarkMsgEnd, + FrameVsync, + SourceLocation, + LockAnnounce, + LockTerminate, + LockMark, + MessageLiteral, + MessageLiteralColor, + MessageLiteralCallstack, + MessageLiteralColorCallstack, + GpuNewContext, + CallstackFrame, + SysTimeReport, + SysPowerReport, + TidToPid, + HwSampleCpuCycle, + HwSampleInstructionRetired, + HwSampleCacheReference, + HwSampleCacheMiss, + HwSampleBranchRetired, + HwSampleBranchMiss, + PlotConfig, + ParamSetup, + AckServerQueryNoop, + AckSourceCodeNotAvailable, + AckSymbolCodeNotAvailable, + CpuTopology, + SingleStringData, + SecondStringData, + MemNamePayload, + StringData, + ThreadName, + PlotName, + SourceLocationPayload, + CallstackPayload, + CallstackAllocPayload, + FrameName, + FrameImageData, + ExternalName, + ExternalThreadName, + SymbolCode, + SourceCode, + FiberName, + NUM_TYPES +}; + +#pragma pack( push, 1 ) + +struct QueueThreadContext +{ + uint32_t thread; +}; + +struct QueueZoneBeginLean +{ + int64_t time; +}; + +struct QueueZoneBegin : public QueueZoneBeginLean +{ + uint64_t srcloc; // ptr +}; + +struct QueueZoneBeginThread : public QueueZoneBegin +{ + uint32_t thread; +}; + +struct QueueZoneEnd +{ + int64_t time; +}; + +struct QueueZoneEndThread : public QueueZoneEnd +{ + uint32_t thread; +}; + +struct QueueZoneValidation +{ + uint32_t id; +}; + +struct QueueZoneValidationThread : public QueueZoneValidation +{ + uint32_t thread; +}; + +struct QueueZoneColor +{ + uint8_t b; + uint8_t g; + uint8_t r; +}; + +struct QueueZoneColorThread : public QueueZoneColor +{ + uint32_t thread; +}; + +struct QueueZoneValue +{ + uint64_t value; +}; + +struct QueueZoneValueThread : public QueueZoneValue +{ + uint32_t thread; +}; + +struct QueueStringTransfer +{ + uint64_t ptr; +}; + +struct QueueFrameMark +{ + int64_t time; + uint64_t name; // ptr +}; + +struct QueueFrameVsync +{ + int64_t time; + uint32_t id; +}; + +struct QueueFrameImage +{ + uint32_t frame; + uint16_t w; + uint16_t h; + uint8_t flip; +}; + +struct QueueFrameImageFat : public QueueFrameImage +{ + uint64_t image; // ptr +}; + +struct QueueSourceLocation +{ + uint64_t name; + uint64_t function; // ptr + uint64_t file; // ptr + uint32_t line; + uint8_t b; + uint8_t g; + uint8_t r; +}; + +struct QueueZoneTextFat +{ + uint64_t text; // ptr + uint16_t size; +}; + +struct QueueZoneTextFatThread : public QueueZoneTextFat +{ + uint32_t thread; +}; + +enum class LockType : uint8_t +{ + Lockable, + SharedLockable +}; + +struct QueueLockAnnounce +{ + uint32_t id; + int64_t time; + uint64_t lckloc; // ptr + LockType type; +}; + +struct QueueFiberEnter +{ + int64_t time; + uint64_t fiber; // ptr + uint32_t thread; +}; + +struct QueueFiberLeave +{ + int64_t time; + uint32_t thread; +}; + +struct QueueLockTerminate +{ + uint32_t id; + int64_t time; +}; + +struct QueueLockWait +{ + uint32_t thread; + uint32_t id; + int64_t time; +}; + +struct QueueLockObtain +{ + uint32_t thread; + uint32_t id; + int64_t time; +}; + +struct QueueLockRelease +{ + uint32_t id; + int64_t time; +}; + +struct QueueLockReleaseShared : public QueueLockRelease +{ + uint32_t thread; +}; + +struct QueueLockMark +{ + uint32_t thread; + uint32_t id; + uint64_t srcloc; // ptr +}; + +struct QueueLockName +{ + uint32_t id; +}; + +struct QueueLockNameFat : public QueueLockName +{ + uint64_t name; // ptr + uint16_t size; +}; + +struct QueuePlotDataBase +{ + uint64_t name; // ptr + int64_t time; +}; + +struct QueuePlotDataInt : public QueuePlotDataBase +{ + int64_t val; +}; + +struct QueuePlotDataFloat : public QueuePlotDataBase +{ + float val; +}; + +struct QueuePlotDataDouble : public QueuePlotDataBase +{ + double val; +}; + +struct QueueMessage +{ + int64_t time; +}; + +struct QueueMessageColor : public QueueMessage +{ + uint8_t b; + uint8_t g; + uint8_t r; +}; + +struct QueueMessageLiteral : public QueueMessage +{ + uint64_t text; // ptr +}; + +struct QueueMessageLiteralThread : public QueueMessageLiteral +{ + uint32_t thread; +}; + +struct QueueMessageColorLiteral : public QueueMessageColor +{ + uint64_t text; // ptr +}; + +struct QueueMessageColorLiteralThread : public QueueMessageColorLiteral +{ + uint32_t thread; +}; + +struct QueueMessageFat : public QueueMessage +{ + uint64_t text; // ptr + uint16_t size; +}; + +struct QueueMessageFatThread : public QueueMessageFat +{ + uint32_t thread; +}; + +struct QueueMessageColorFat : public QueueMessageColor +{ + uint64_t text; // ptr + uint16_t size; +}; + +struct QueueMessageColorFatThread : public QueueMessageColorFat +{ + uint32_t thread; +}; + +// Don't change order, only add new entries at the end, this is also used on trace dumps! +enum class GpuContextType : uint8_t +{ + Invalid, + OpenGl, + Vulkan, + OpenCL, + Direct3D12, + Direct3D11 +}; + +enum GpuContextFlags : uint8_t +{ + GpuContextCalibration = 1 << 0 +}; + +struct QueueGpuNewContext +{ + int64_t cpuTime; + int64_t gpuTime; + uint32_t thread; + float period; + uint8_t context; + GpuContextFlags flags; + GpuContextType type; +}; + +struct QueueGpuZoneBeginLean +{ + int64_t cpuTime; + uint32_t thread; + uint16_t queryId; + uint8_t context; +}; + +struct QueueGpuZoneBegin : public QueueGpuZoneBeginLean +{ + uint64_t srcloc; +}; + +struct QueueGpuZoneEnd +{ + int64_t cpuTime; + uint32_t thread; + uint16_t queryId; + uint8_t context; +}; + +struct QueueGpuTime +{ + int64_t gpuTime; + uint16_t queryId; + uint8_t context; +}; + +struct QueueGpuCalibration +{ + int64_t gpuTime; + int64_t cpuTime; + int64_t cpuDelta; + uint8_t context; +}; + +struct QueueGpuTimeSync +{ + int64_t gpuTime; + int64_t cpuTime; + uint8_t context; +}; + +struct QueueGpuContextName +{ + uint8_t context; +}; + +struct QueueGpuContextNameFat : public QueueGpuContextName +{ + uint64_t ptr; + uint16_t size; +}; + +struct QueueMemNamePayload +{ + uint64_t name; +}; + +struct QueueMemAlloc +{ + int64_t time; + uint32_t thread; + uint64_t ptr; + char size[6]; +}; + +struct QueueMemFree +{ + int64_t time; + uint32_t thread; + uint64_t ptr; +}; + +struct QueueCallstackFat +{ + uint64_t ptr; +}; + +struct QueueCallstackFatThread : public QueueCallstackFat +{ + uint32_t thread; +}; + +struct QueueCallstackAllocFat +{ + uint64_t ptr; + uint64_t nativePtr; +}; + +struct QueueCallstackAllocFatThread : public QueueCallstackAllocFat +{ + uint32_t thread; +}; + +struct QueueCallstackSample +{ + int64_t time; + uint32_t thread; +}; + +struct QueueCallstackSampleFat : public QueueCallstackSample +{ + uint64_t ptr; +}; + +struct QueueCallstackFrameSize +{ + uint64_t ptr; + uint8_t size; +}; + +struct QueueCallstackFrameSizeFat : public QueueCallstackFrameSize +{ + uint64_t data; + uint64_t imageName; +}; + +struct QueueCallstackFrame +{ + uint32_t line; + uint64_t symAddr; + uint32_t symLen; +}; + +struct QueueSymbolInformation +{ + uint32_t line; + uint64_t symAddr; +}; + +struct QueueSymbolInformationFat : public QueueSymbolInformation +{ + uint64_t fileString; + uint8_t needFree; +}; + +struct QueueCrashReport +{ + int64_t time; + uint64_t text; // ptr +}; + +struct QueueCrashReportThread +{ + uint32_t thread; +}; + +struct QueueSysTime +{ + int64_t time; + float sysTime; +}; + +struct QueueSysPower +{ + int64_t time; + uint64_t delta; + uint64_t name; // ptr +}; + +struct QueueContextSwitch +{ + int64_t time; + uint32_t oldThread; + uint32_t newThread; + uint8_t cpu; + uint8_t reason; + uint8_t state; +}; + +struct QueueThreadWakeup +{ + int64_t time; + uint32_t thread; +}; + +struct QueueTidToPid +{ + uint64_t tid; + uint64_t pid; +}; + +struct QueueHwSample +{ + uint64_t ip; + int64_t time; +}; + +enum class PlotFormatType : uint8_t +{ + Number, + Memory, + Percentage +}; + +struct QueuePlotConfig +{ + uint64_t name; // ptr + uint8_t type; + uint8_t step; + uint8_t fill; + uint32_t color; +}; + +struct QueueParamSetup +{ + uint32_t idx; + uint64_t name; // ptr + uint8_t isBool; + int32_t val; +}; + +struct QueueSourceCodeNotAvailable +{ + uint32_t id; +}; + +struct QueueCpuTopology +{ + uint32_t package; + uint32_t core; + uint32_t thread; +}; + +struct QueueExternalNameMetadata +{ + uint64_t thread; + uint64_t name; + uint64_t threadName; +}; + +struct QueueSymbolCodeMetadata +{ + uint64_t symbol; + uint64_t ptr; + uint32_t size; +}; + +struct QueueSourceCodeMetadata +{ + uint64_t ptr; + uint32_t size; + uint32_t id; +}; + +struct QueueHeader +{ + union + { + QueueType type; + uint8_t idx; + }; +}; + +struct QueueItem +{ + QueueHeader hdr; + union + { + QueueThreadContext threadCtx; + QueueZoneBegin zoneBegin; + QueueZoneBeginLean zoneBeginLean; + QueueZoneBeginThread zoneBeginThread; + QueueZoneEnd zoneEnd; + QueueZoneEndThread zoneEndThread; + QueueZoneValidation zoneValidation; + QueueZoneValidationThread zoneValidationThread; + QueueZoneColor zoneColor; + QueueZoneColorThread zoneColorThread; + QueueZoneValue zoneValue; + QueueZoneValueThread zoneValueThread; + QueueStringTransfer stringTransfer; + QueueFrameMark frameMark; + QueueFrameVsync frameVsync; + QueueFrameImage frameImage; + QueueFrameImageFat frameImageFat; + QueueSourceLocation srcloc; + QueueZoneTextFat zoneTextFat; + QueueZoneTextFatThread zoneTextFatThread; + QueueLockAnnounce lockAnnounce; + QueueLockTerminate lockTerminate; + QueueLockWait lockWait; + QueueLockObtain lockObtain; + QueueLockRelease lockRelease; + QueueLockReleaseShared lockReleaseShared; + QueueLockMark lockMark; + QueueLockName lockName; + QueueLockNameFat lockNameFat; + QueuePlotDataInt plotDataInt; + QueuePlotDataFloat plotDataFloat; + QueuePlotDataDouble plotDataDouble; + QueueMessage message; + QueueMessageColor messageColor; + QueueMessageLiteral messageLiteral; + QueueMessageLiteralThread messageLiteralThread; + QueueMessageColorLiteral messageColorLiteral; + QueueMessageColorLiteralThread messageColorLiteralThread; + QueueMessageFat messageFat; + QueueMessageFatThread messageFatThread; + QueueMessageColorFat messageColorFat; + QueueMessageColorFatThread messageColorFatThread; + QueueGpuNewContext gpuNewContext; + QueueGpuZoneBegin gpuZoneBegin; + QueueGpuZoneBeginLean gpuZoneBeginLean; + QueueGpuZoneEnd gpuZoneEnd; + QueueGpuTime gpuTime; + QueueGpuCalibration gpuCalibration; + QueueGpuTimeSync gpuTimeSync; + QueueGpuContextName gpuContextName; + QueueGpuContextNameFat gpuContextNameFat; + QueueMemAlloc memAlloc; + QueueMemFree memFree; + QueueMemNamePayload memName; + QueueCallstackFat callstackFat; + QueueCallstackFatThread callstackFatThread; + QueueCallstackAllocFat callstackAllocFat; + QueueCallstackAllocFatThread callstackAllocFatThread; + QueueCallstackSample callstackSample; + QueueCallstackSampleFat callstackSampleFat; + QueueCallstackFrameSize callstackFrameSize; + QueueCallstackFrameSizeFat callstackFrameSizeFat; + QueueCallstackFrame callstackFrame; + QueueSymbolInformation symbolInformation; + QueueSymbolInformationFat symbolInformationFat; + QueueCrashReport crashReport; + QueueCrashReportThread crashReportThread; + QueueSysTime sysTime; + QueueSysPower sysPower; + QueueContextSwitch contextSwitch; + QueueThreadWakeup threadWakeup; + QueueTidToPid tidToPid; + QueueHwSample hwSample; + QueuePlotConfig plotConfig; + QueueParamSetup paramSetup; + QueueCpuTopology cpuTopology; + QueueExternalNameMetadata externalNameMetadata; + QueueSymbolCodeMetadata symbolCodeMetadata; + QueueSourceCodeMetadata sourceCodeMetadata; + QueueSourceCodeNotAvailable sourceCodeNotAvailable; + QueueFiberEnter fiberEnter; + QueueFiberLeave fiberLeave; + }; +}; +#pragma pack( pop ) + + +enum { QueueItemSize = sizeof( QueueItem ) }; + +static constexpr size_t QueueDataSize[] = { + sizeof( QueueHeader ), // zone text + sizeof( QueueHeader ), // zone name + sizeof( QueueHeader ) + sizeof( QueueMessage ), + sizeof( QueueHeader ) + sizeof( QueueMessageColor ), + sizeof( QueueHeader ) + sizeof( QueueMessage ), // callstack + sizeof( QueueHeader ) + sizeof( QueueMessageColor ), // callstack + sizeof( QueueHeader ) + sizeof( QueueMessage ), // app info + sizeof( QueueHeader ) + sizeof( QueueZoneBeginLean ), // allocated source location + sizeof( QueueHeader ) + sizeof( QueueZoneBeginLean ), // allocated source location, callstack + sizeof( QueueHeader ), // callstack memory + sizeof( QueueHeader ), // callstack + sizeof( QueueHeader ), // callstack alloc + sizeof( QueueHeader ) + sizeof( QueueCallstackSample ), + sizeof( QueueHeader ) + sizeof( QueueCallstackSample ), // context switch + sizeof( QueueHeader ) + sizeof( QueueFrameImage ), + sizeof( QueueHeader ) + sizeof( QueueZoneBegin ), + sizeof( QueueHeader ) + sizeof( QueueZoneBegin ), // callstack + sizeof( QueueHeader ) + sizeof( QueueZoneEnd ), + sizeof( QueueHeader ) + sizeof( QueueLockWait ), + sizeof( QueueHeader ) + sizeof( QueueLockObtain ), + sizeof( QueueHeader ) + sizeof( QueueLockRelease ), + sizeof( QueueHeader ) + sizeof( QueueLockWait ), // shared + sizeof( QueueHeader ) + sizeof( QueueLockObtain ), // shared + sizeof( QueueHeader ) + sizeof( QueueLockReleaseShared ), + sizeof( QueueHeader ) + sizeof( QueueLockName ), + sizeof( QueueHeader ) + sizeof( QueueMemAlloc ), + sizeof( QueueHeader ) + sizeof( QueueMemAlloc ), // named + sizeof( QueueHeader ) + sizeof( QueueMemFree ), + sizeof( QueueHeader ) + sizeof( QueueMemFree ), // named + sizeof( QueueHeader ) + sizeof( QueueMemAlloc ), // callstack + sizeof( QueueHeader ) + sizeof( QueueMemAlloc ), // callstack, named + sizeof( QueueHeader ) + sizeof( QueueMemFree ), // callstack + sizeof( QueueHeader ) + sizeof( QueueMemFree ), // callstack, named + sizeof( QueueHeader ) + sizeof( QueueGpuZoneBegin ), + sizeof( QueueHeader ) + sizeof( QueueGpuZoneBegin ), // callstack + sizeof( QueueHeader ) + sizeof( QueueGpuZoneBeginLean ),// allocated source location + sizeof( QueueHeader ) + sizeof( QueueGpuZoneBeginLean ),// allocated source location, callstack + sizeof( QueueHeader ) + sizeof( QueueGpuZoneEnd ), + sizeof( QueueHeader ) + sizeof( QueueGpuZoneBegin ), // serial + sizeof( QueueHeader ) + sizeof( QueueGpuZoneBegin ), // serial, callstack + sizeof( QueueHeader ) + sizeof( QueueGpuZoneBeginLean ),// serial, allocated source location + sizeof( QueueHeader ) + sizeof( QueueGpuZoneBeginLean ),// serial, allocated source location, callstack + sizeof( QueueHeader ) + sizeof( QueueGpuZoneEnd ), // serial + sizeof( QueueHeader ) + sizeof( QueuePlotDataInt ), + sizeof( QueueHeader ) + sizeof( QueuePlotDataFloat ), + sizeof( QueueHeader ) + sizeof( QueuePlotDataDouble ), + sizeof( QueueHeader ) + sizeof( QueueContextSwitch ), + sizeof( QueueHeader ) + sizeof( QueueThreadWakeup ), + sizeof( QueueHeader ) + sizeof( QueueGpuTime ), + sizeof( QueueHeader ) + sizeof( QueueGpuContextName ), + sizeof( QueueHeader ) + sizeof( QueueCallstackFrameSize ), + sizeof( QueueHeader ) + sizeof( QueueSymbolInformation ), + sizeof( QueueHeader ), // ExternalNameMetadata - not for wire transfer + sizeof( QueueHeader ), // SymbolCodeMetadata - not for wire transfer + sizeof( QueueHeader ), // SourceCodeMetadata - not for wire transfer + sizeof( QueueHeader ) + sizeof( QueueFiberEnter ), + sizeof( QueueHeader ) + sizeof( QueueFiberLeave ), + // above items must be first + sizeof( QueueHeader ), // terminate + sizeof( QueueHeader ), // keep alive + sizeof( QueueHeader ) + sizeof( QueueThreadContext ), + sizeof( QueueHeader ) + sizeof( QueueGpuCalibration ), + sizeof( QueueHeader ) + sizeof( QueueGpuTimeSync ), + sizeof( QueueHeader ), // crash + sizeof( QueueHeader ) + sizeof( QueueCrashReport ), + sizeof( QueueHeader ) + sizeof( QueueZoneValidation ), + sizeof( QueueHeader ) + sizeof( QueueZoneColor ), + sizeof( QueueHeader ) + sizeof( QueueZoneValue ), + sizeof( QueueHeader ) + sizeof( QueueFrameMark ), // continuous frames + sizeof( QueueHeader ) + sizeof( QueueFrameMark ), // start + sizeof( QueueHeader ) + sizeof( QueueFrameMark ), // end + sizeof( QueueHeader ) + sizeof( QueueFrameVsync ), + sizeof( QueueHeader ) + sizeof( QueueSourceLocation ), + sizeof( QueueHeader ) + sizeof( QueueLockAnnounce ), + sizeof( QueueHeader ) + sizeof( QueueLockTerminate ), + sizeof( QueueHeader ) + sizeof( QueueLockMark ), + sizeof( QueueHeader ) + sizeof( QueueMessageLiteral ), + sizeof( QueueHeader ) + sizeof( QueueMessageColorLiteral ), + sizeof( QueueHeader ) + sizeof( QueueMessageLiteral ), // callstack + sizeof( QueueHeader ) + sizeof( QueueMessageColorLiteral ), // callstack + sizeof( QueueHeader ) + sizeof( QueueGpuNewContext ), + sizeof( QueueHeader ) + sizeof( QueueCallstackFrame ), + sizeof( QueueHeader ) + sizeof( QueueSysTime ), + sizeof( QueueHeader ) + sizeof( QueueSysPower ), + sizeof( QueueHeader ) + sizeof( QueueTidToPid ), + sizeof( QueueHeader ) + sizeof( QueueHwSample ), // cpu cycle + sizeof( QueueHeader ) + sizeof( QueueHwSample ), // instruction retired + sizeof( QueueHeader ) + sizeof( QueueHwSample ), // cache reference + sizeof( QueueHeader ) + sizeof( QueueHwSample ), // cache miss + sizeof( QueueHeader ) + sizeof( QueueHwSample ), // branch retired + sizeof( QueueHeader ) + sizeof( QueueHwSample ), // branch miss + sizeof( QueueHeader ) + sizeof( QueuePlotConfig ), + sizeof( QueueHeader ) + sizeof( QueueParamSetup ), + sizeof( QueueHeader ), // server query acknowledgement + sizeof( QueueHeader ) + sizeof( QueueSourceCodeNotAvailable ), + sizeof( QueueHeader ), // symbol code not available + sizeof( QueueHeader ) + sizeof( QueueCpuTopology ), + sizeof( QueueHeader ), // single string data + sizeof( QueueHeader ), // second string data + sizeof( QueueHeader ) + sizeof( QueueMemNamePayload ), + // keep all QueueStringTransfer below + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // string data + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // thread name + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // plot name + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // allocated source location payload + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // callstack payload + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // callstack alloc payload + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // frame name + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // frame image data + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // external name + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // external thread name + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // symbol code + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // source code + sizeof( QueueHeader ) + sizeof( QueueStringTransfer ), // fiber name +}; + +static_assert( QueueItemSize == 32, "Queue item size not 32 bytes" ); +static_assert( sizeof( QueueDataSize ) / sizeof( size_t ) == (uint8_t)QueueType::NUM_TYPES, "QueueDataSize mismatch" ); +static_assert( sizeof( void* ) <= sizeof( uint64_t ), "Pointer size > 8 bytes" ); +static_assert( sizeof( void* ) == sizeof( uintptr_t ), "Pointer size != uintptr_t" ); + +} + +#endif diff --git a/externals/tracy/public/common/TracySocket.cpp b/externals/tracy/public/common/TracySocket.cpp new file mode 100644 index 000000000..259678989 --- /dev/null +++ b/externals/tracy/public/common/TracySocket.cpp @@ -0,0 +1,749 @@ +#include +#include +#include +#include +#include +#include +#include + +#include "TracyAlloc.hpp" +#include "TracySocket.hpp" +#include "TracySystem.hpp" + +#ifdef _WIN32 +# ifndef NOMINMAX +# define NOMINMAX +# endif +# include +# include +# ifdef _MSC_VER +# pragma warning(disable:4244) +# pragma warning(disable:4267) +# endif +# define poll WSAPoll +#else +# include +# include +# include +# include +# include +# include +# include +# include +# include +#endif + +#ifndef MSG_NOSIGNAL +# define MSG_NOSIGNAL 0 +#endif + +namespace tracy +{ + +#ifdef _WIN32 +typedef SOCKET socket_t; +#else +typedef int socket_t; +#endif + +#ifdef _WIN32 +struct __wsinit +{ + __wsinit() + { + WSADATA wsaData; + if( WSAStartup( MAKEWORD( 2, 2 ), &wsaData ) != 0 ) + { + fprintf( stderr, "Cannot init winsock.\n" ); + exit( 1 ); + } + } +}; + +void InitWinSock() +{ + static __wsinit init; +} +#endif + + +enum { BufSize = 128 * 1024 }; + +Socket::Socket() + : m_buf( (char*)tracy_malloc( BufSize ) ) + , m_bufPtr( nullptr ) + , m_sock( -1 ) + , m_bufLeft( 0 ) + , m_ptr( nullptr ) +{ +#ifdef _WIN32 + InitWinSock(); +#endif +} + +Socket::Socket( int sock ) + : m_buf( (char*)tracy_malloc( BufSize ) ) + , m_bufPtr( nullptr ) + , m_sock( sock ) + , m_bufLeft( 0 ) + , m_ptr( nullptr ) +{ +} + +Socket::~Socket() +{ + tracy_free( m_buf ); + if( m_sock.load( std::memory_order_relaxed ) != -1 ) + { + Close(); + } + if( m_ptr ) + { + freeaddrinfo( m_res ); +#ifdef _WIN32 + closesocket( m_connSock ); +#else + close( m_connSock ); +#endif + } +} + +bool Socket::Connect( const char* addr, uint16_t port ) +{ + assert( !IsValid() ); + + if( m_ptr ) + { + const auto c = connect( m_connSock, m_ptr->ai_addr, m_ptr->ai_addrlen ); + if( c == -1 ) + { +#if defined _WIN32 + const auto err = WSAGetLastError(); + if( err == WSAEALREADY || err == WSAEINPROGRESS ) return false; + if( err != WSAEISCONN ) + { + freeaddrinfo( m_res ); + closesocket( m_connSock ); + m_ptr = nullptr; + return false; + } +#else + const auto err = errno; + if( err == EALREADY || err == EINPROGRESS ) return false; + if( err != EISCONN ) + { + freeaddrinfo( m_res ); + close( m_connSock ); + m_ptr = nullptr; + return false; + } +#endif + } + +#if defined _WIN32 + u_long nonblocking = 0; + ioctlsocket( m_connSock, FIONBIO, &nonblocking ); +#else + int flags = fcntl( m_connSock, F_GETFL, 0 ); + fcntl( m_connSock, F_SETFL, flags & ~O_NONBLOCK ); +#endif + m_sock.store( m_connSock, std::memory_order_relaxed ); + freeaddrinfo( m_res ); + m_ptr = nullptr; + return true; + } + + struct addrinfo hints; + struct addrinfo *res, *ptr; + + memset( &hints, 0, sizeof( hints ) ); + hints.ai_family = AF_UNSPEC; + hints.ai_socktype = SOCK_STREAM; + + char portbuf[32]; + sprintf( portbuf, "%" PRIu16, port ); + + if( getaddrinfo( addr, portbuf, &hints, &res ) != 0 ) return false; + int sock = 0; + for( ptr = res; ptr; ptr = ptr->ai_next ) + { + if( ( sock = socket( ptr->ai_family, ptr->ai_socktype, ptr->ai_protocol ) ) == -1 ) continue; +#if defined __APPLE__ + int val = 1; + setsockopt( sock, SOL_SOCKET, SO_NOSIGPIPE, &val, sizeof( val ) ); +#endif +#if defined _WIN32 + u_long nonblocking = 1; + ioctlsocket( sock, FIONBIO, &nonblocking ); +#else + int flags = fcntl( sock, F_GETFL, 0 ); + fcntl( sock, F_SETFL, flags | O_NONBLOCK ); +#endif + if( connect( sock, ptr->ai_addr, ptr->ai_addrlen ) == 0 ) + { + break; + } + else + { +#if defined _WIN32 + const auto err = WSAGetLastError(); + if( err != WSAEWOULDBLOCK ) + { + closesocket( sock ); + continue; + } +#else + if( errno != EINPROGRESS ) + { + close( sock ); + continue; + } +#endif + } + m_res = res; + m_ptr = ptr; + m_connSock = sock; + return false; + } + freeaddrinfo( res ); + if( !ptr ) return false; + +#if defined _WIN32 + u_long nonblocking = 0; + ioctlsocket( sock, FIONBIO, &nonblocking ); +#else + int flags = fcntl( sock, F_GETFL, 0 ); + fcntl( sock, F_SETFL, flags & ~O_NONBLOCK ); +#endif + + m_sock.store( sock, std::memory_order_relaxed ); + return true; +} + +bool Socket::ConnectBlocking( const char* addr, uint16_t port ) +{ + assert( !IsValid() ); + assert( !m_ptr ); + + struct addrinfo hints; + struct addrinfo *res, *ptr; + + memset( &hints, 0, sizeof( hints ) ); + hints.ai_family = AF_UNSPEC; + hints.ai_socktype = SOCK_STREAM; + + char portbuf[32]; + sprintf( portbuf, "%" PRIu16, port ); + + if( getaddrinfo( addr, portbuf, &hints, &res ) != 0 ) return false; + int sock = 0; + for( ptr = res; ptr; ptr = ptr->ai_next ) + { + if( ( sock = socket( ptr->ai_family, ptr->ai_socktype, ptr->ai_protocol ) ) == -1 ) continue; +#if defined __APPLE__ + int val = 1; + setsockopt( sock, SOL_SOCKET, SO_NOSIGPIPE, &val, sizeof( val ) ); +#endif + if( connect( sock, ptr->ai_addr, ptr->ai_addrlen ) == -1 ) + { +#ifdef _WIN32 + closesocket( sock ); +#else + close( sock ); +#endif + continue; + } + break; + } + freeaddrinfo( res ); + if( !ptr ) return false; + + m_sock.store( sock, std::memory_order_relaxed ); + return true; +} + +void Socket::Close() +{ + const auto sock = m_sock.load( std::memory_order_relaxed ); + assert( sock != -1 ); +#ifdef _WIN32 + closesocket( sock ); +#else + close( sock ); +#endif + m_sock.store( -1, std::memory_order_relaxed ); +} + +int Socket::Send( const void* _buf, int len ) +{ + const auto sock = m_sock.load( std::memory_order_relaxed ); + auto buf = (const char*)_buf; + assert( sock != -1 ); + auto start = buf; + while( len > 0 ) + { + auto ret = send( sock, buf, len, MSG_NOSIGNAL ); + if( ret == -1 ) return -1; + len -= ret; + buf += ret; + } + return int( buf - start ); +} + +int Socket::GetSendBufSize() +{ + const auto sock = m_sock.load( std::memory_order_relaxed ); + int bufSize; +#if defined _WIN32 + int sz = sizeof( bufSize ); + getsockopt( sock, SOL_SOCKET, SO_SNDBUF, (char*)&bufSize, &sz ); +#else + socklen_t sz = sizeof( bufSize ); + getsockopt( sock, SOL_SOCKET, SO_SNDBUF, &bufSize, &sz ); +#endif + return bufSize; +} + +int Socket::RecvBuffered( void* buf, int len, int timeout ) +{ + if( len <= m_bufLeft ) + { + memcpy( buf, m_bufPtr, len ); + m_bufPtr += len; + m_bufLeft -= len; + return len; + } + + if( m_bufLeft > 0 ) + { + memcpy( buf, m_bufPtr, m_bufLeft ); + const auto ret = m_bufLeft; + m_bufLeft = 0; + return ret; + } + + if( len >= BufSize ) return Recv( buf, len, timeout ); + + m_bufLeft = Recv( m_buf, BufSize, timeout ); + if( m_bufLeft <= 0 ) return m_bufLeft; + + const auto sz = len < m_bufLeft ? len : m_bufLeft; + memcpy( buf, m_buf, sz ); + m_bufPtr = m_buf + sz; + m_bufLeft -= sz; + return sz; +} + +int Socket::Recv( void* _buf, int len, int timeout ) +{ + const auto sock = m_sock.load( std::memory_order_relaxed ); + auto buf = (char*)_buf; + + struct pollfd fd; + fd.fd = (socket_t)sock; + fd.events = POLLIN; + + if( poll( &fd, 1, timeout ) > 0 ) + { + return recv( sock, buf, len, 0 ); + } + else + { + return -1; + } +} + +int Socket::ReadUpTo( void* _buf, int len ) +{ + const auto sock = m_sock.load( std::memory_order_relaxed ); + auto buf = (char*)_buf; + + int rd = 0; + while( len > 0 ) + { + const auto res = recv( sock, buf, len, 0 ); + if( res == 0 ) break; + if( res == -1 ) return -1; + len -= res; + rd += res; + buf += res; + } + return rd; +} + +bool Socket::Read( void* buf, int len, int timeout ) +{ + auto cbuf = (char*)buf; + while( len > 0 ) + { + if( !ReadImpl( cbuf, len, timeout ) ) return false; + } + return true; +} + +bool Socket::ReadImpl( char*& buf, int& len, int timeout ) +{ + const auto sz = RecvBuffered( buf, len, timeout ); + switch( sz ) + { + case 0: + return false; + case -1: +#ifdef _WIN32 + { + auto err = WSAGetLastError(); + if( err == WSAECONNABORTED || err == WSAECONNRESET ) return false; + } +#endif + break; + default: + len -= sz; + buf += sz; + break; + } + return true; +} + +bool Socket::ReadRaw( void* _buf, int len, int timeout ) +{ + auto buf = (char*)_buf; + while( len > 0 ) + { + const auto sz = Recv( buf, len, timeout ); + if( sz <= 0 ) return false; + len -= sz; + buf += sz; + } + return true; +} + +bool Socket::HasData() +{ + const auto sock = m_sock.load( std::memory_order_relaxed ); + if( m_bufLeft > 0 ) return true; + + struct pollfd fd; + fd.fd = (socket_t)sock; + fd.events = POLLIN; + + return poll( &fd, 1, 0 ) > 0; +} + +bool Socket::IsValid() const +{ + return m_sock.load( std::memory_order_relaxed ) >= 0; +} + + +ListenSocket::ListenSocket() + : m_sock( -1 ) +{ +#ifdef _WIN32 + InitWinSock(); +#endif +} + +ListenSocket::~ListenSocket() +{ + if( m_sock != -1 ) Close(); +} + +static int addrinfo_and_socket_for_family( uint16_t port, int ai_family, struct addrinfo** res ) +{ + struct addrinfo hints; + memset( &hints, 0, sizeof( hints ) ); + hints.ai_family = ai_family; + hints.ai_socktype = SOCK_STREAM; +#ifndef TRACY_ONLY_LOCALHOST + const char* onlyLocalhost = GetEnvVar( "TRACY_ONLY_LOCALHOST" ); + if( !onlyLocalhost || onlyLocalhost[0] != '1' ) + { + hints.ai_flags = AI_PASSIVE; + } +#endif + char portbuf[32]; + sprintf( portbuf, "%" PRIu16, port ); + if( getaddrinfo( nullptr, portbuf, &hints, res ) != 0 ) return -1; + int sock = socket( (*res)->ai_family, (*res)->ai_socktype, (*res)->ai_protocol ); + if (sock == -1) freeaddrinfo( *res ); + return sock; +} + +bool ListenSocket::Listen( uint16_t port, int backlog ) +{ + assert( m_sock == -1 ); + + struct addrinfo* res = nullptr; + +#if !defined TRACY_ONLY_IPV4 && !defined TRACY_ONLY_LOCALHOST + const char* onlyIPv4 = GetEnvVar( "TRACY_ONLY_IPV4" ); + if( !onlyIPv4 || onlyIPv4[0] != '1' ) + { + m_sock = addrinfo_and_socket_for_family( port, AF_INET6, &res ); + } +#endif + if (m_sock == -1) + { + // IPV6 protocol may not be available/is disabled. Try to create a socket + // with the IPV4 protocol + m_sock = addrinfo_and_socket_for_family( port, AF_INET, &res ); + if( m_sock == -1 ) return false; + } +#if defined _WIN32 + unsigned long val = 0; + setsockopt( m_sock, IPPROTO_IPV6, IPV6_V6ONLY, (const char*)&val, sizeof( val ) ); +#elif defined BSD + int val = 0; + setsockopt( m_sock, IPPROTO_IPV6, IPV6_V6ONLY, (const char*)&val, sizeof( val ) ); + val = 1; + setsockopt( m_sock, SOL_SOCKET, SO_REUSEADDR, &val, sizeof( val ) ); +#else + int val = 1; + setsockopt( m_sock, SOL_SOCKET, SO_REUSEADDR, &val, sizeof( val ) ); +#endif + if( bind( m_sock, res->ai_addr, res->ai_addrlen ) == -1 ) { freeaddrinfo( res ); Close(); return false; } + if( listen( m_sock, backlog ) == -1 ) { freeaddrinfo( res ); Close(); return false; } + freeaddrinfo( res ); + return true; +} + +Socket* ListenSocket::Accept() +{ + struct sockaddr_storage remote; + socklen_t sz = sizeof( remote ); + + struct pollfd fd; + fd.fd = (socket_t)m_sock; + fd.events = POLLIN; + + if( poll( &fd, 1, 10 ) > 0 ) + { + int sock = accept( m_sock, (sockaddr*)&remote, &sz); + if( sock == -1 ) return nullptr; + +#if defined __APPLE__ + int val = 1; + setsockopt( sock, SOL_SOCKET, SO_NOSIGPIPE, &val, sizeof( val ) ); +#endif + + auto ptr = (Socket*)tracy_malloc( sizeof( Socket ) ); + new(ptr) Socket( sock ); + return ptr; + } + else + { + return nullptr; + } +} + +void ListenSocket::Close() +{ + assert( m_sock != -1 ); +#ifdef _WIN32 + closesocket( m_sock ); +#else + close( m_sock ); +#endif + m_sock = -1; +} + +UdpBroadcast::UdpBroadcast() + : m_sock( -1 ) +{ +#ifdef _WIN32 + InitWinSock(); +#endif +} + +UdpBroadcast::~UdpBroadcast() +{ + if( m_sock != -1 ) Close(); +} + +bool UdpBroadcast::Open( const char* addr, uint16_t port ) +{ + assert( m_sock == -1 ); + + struct addrinfo hints; + struct addrinfo *res, *ptr; + + memset( &hints, 0, sizeof( hints ) ); + hints.ai_family = AF_INET; + hints.ai_socktype = SOCK_DGRAM; + + char portbuf[32]; + sprintf( portbuf, "%" PRIu16, port ); + + if( getaddrinfo( addr, portbuf, &hints, &res ) != 0 ) return false; + int sock = 0; + for( ptr = res; ptr; ptr = ptr->ai_next ) + { + if( ( sock = socket( ptr->ai_family, ptr->ai_socktype, ptr->ai_protocol ) ) == -1 ) continue; +#if defined __APPLE__ + int val = 1; + setsockopt( sock, SOL_SOCKET, SO_NOSIGPIPE, &val, sizeof( val ) ); +#endif +#if defined _WIN32 + unsigned long broadcast = 1; + if( setsockopt( sock, SOL_SOCKET, SO_BROADCAST, (const char*)&broadcast, sizeof( broadcast ) ) == -1 ) +#else + int broadcast = 1; + if( setsockopt( sock, SOL_SOCKET, SO_BROADCAST, &broadcast, sizeof( broadcast ) ) == -1 ) +#endif + { +#ifdef _WIN32 + closesocket( sock ); +#else + close( sock ); +#endif + continue; + } + break; + } + freeaddrinfo( res ); + if( !ptr ) return false; + + m_sock = sock; + inet_pton( AF_INET, addr, &m_addr ); + return true; +} + +void UdpBroadcast::Close() +{ + assert( m_sock != -1 ); +#ifdef _WIN32 + closesocket( m_sock ); +#else + close( m_sock ); +#endif + m_sock = -1; +} + +int UdpBroadcast::Send( uint16_t port, const void* data, int len ) +{ + assert( m_sock != -1 ); + struct sockaddr_in addr; + addr.sin_family = AF_INET; + addr.sin_port = htons( port ); + addr.sin_addr.s_addr = m_addr; + return sendto( m_sock, (const char*)data, len, MSG_NOSIGNAL, (sockaddr*)&addr, sizeof( addr ) ); +} + +IpAddress::IpAddress() + : m_number( 0 ) +{ + *m_text = '\0'; +} + +IpAddress::~IpAddress() +{ +} + +void IpAddress::Set( const struct sockaddr& addr ) +{ +#if defined _WIN32 && ( !defined NTDDI_WIN10 || NTDDI_VERSION < NTDDI_WIN10 ) + struct sockaddr_in tmp; + memcpy( &tmp, &addr, sizeof( tmp ) ); + auto ai = &tmp; +#else + auto ai = (const struct sockaddr_in*)&addr; +#endif + inet_ntop( AF_INET, &ai->sin_addr, m_text, 17 ); + m_number = ai->sin_addr.s_addr; +} + +UdpListen::UdpListen() + : m_sock( -1 ) +{ +#ifdef _WIN32 + InitWinSock(); +#endif +} + +UdpListen::~UdpListen() +{ + if( m_sock != -1 ) Close(); +} + +bool UdpListen::Listen( uint16_t port ) +{ + assert( m_sock == -1 ); + + int sock; + if( ( sock = socket( AF_INET, SOCK_DGRAM, 0 ) ) == -1 ) return false; + +#if defined __APPLE__ + int val = 1; + setsockopt( sock, SOL_SOCKET, SO_NOSIGPIPE, &val, sizeof( val ) ); +#endif +#if defined _WIN32 + unsigned long reuse = 1; + setsockopt( sock, SOL_SOCKET, SO_REUSEADDR, (const char*)&reuse, sizeof( reuse ) ); +#else + int reuse = 1; + setsockopt( sock, SOL_SOCKET, SO_REUSEADDR, &reuse, sizeof( reuse ) ); +#endif +#if defined _WIN32 + unsigned long broadcast = 1; + if( setsockopt( sock, SOL_SOCKET, SO_BROADCAST, (const char*)&broadcast, sizeof( broadcast ) ) == -1 ) +#else + int broadcast = 1; + if( setsockopt( sock, SOL_SOCKET, SO_BROADCAST, &broadcast, sizeof( broadcast ) ) == -1 ) +#endif + { +#ifdef _WIN32 + closesocket( sock ); +#else + close( sock ); +#endif + return false; + } + + struct sockaddr_in addr; + addr.sin_family = AF_INET; + addr.sin_port = htons( port ); + addr.sin_addr.s_addr = INADDR_ANY; + + if( bind( sock, (sockaddr*)&addr, sizeof( addr ) ) == -1 ) + { +#ifdef _WIN32 + closesocket( sock ); +#else + close( sock ); +#endif + return false; + } + + m_sock = sock; + return true; +} + +void UdpListen::Close() +{ + assert( m_sock != -1 ); +#ifdef _WIN32 + closesocket( m_sock ); +#else + close( m_sock ); +#endif + m_sock = -1; +} + +const char* UdpListen::Read( size_t& len, IpAddress& addr, int timeout ) +{ + static char buf[2048]; + + struct pollfd fd; + fd.fd = (socket_t)m_sock; + fd.events = POLLIN; + if( poll( &fd, 1, timeout ) <= 0 ) return nullptr; + + sockaddr sa; + socklen_t salen = sizeof( struct sockaddr ); + len = (size_t)recvfrom( m_sock, buf, 2048, 0, &sa, &salen ); + addr.Set( sa ); + + return buf; +} + +} diff --git a/externals/tracy/public/common/TracySocket.hpp b/externals/tracy/public/common/TracySocket.hpp new file mode 100644 index 000000000..f7713aac6 --- /dev/null +++ b/externals/tracy/public/common/TracySocket.hpp @@ -0,0 +1,155 @@ +#ifndef __TRACYSOCKET_HPP__ +#define __TRACYSOCKET_HPP__ + +#include +#include +#include + +struct addrinfo; +struct sockaddr; + +namespace tracy +{ + +#ifdef _WIN32 +void InitWinSock(); +#endif + +class Socket +{ +public: + Socket(); + Socket( int sock ); + ~Socket(); + + bool Connect( const char* addr, uint16_t port ); + bool ConnectBlocking( const char* addr, uint16_t port ); + void Close(); + + int Send( const void* buf, int len ); + int GetSendBufSize(); + + int ReadUpTo( void* buf, int len ); + bool Read( void* buf, int len, int timeout ); + + template + bool Read( void* buf, int len, int timeout, ShouldExit exitCb ) + { + auto cbuf = (char*)buf; + while( len > 0 ) + { + if( exitCb() ) return false; + if( !ReadImpl( cbuf, len, timeout ) ) return false; + } + return true; + } + + bool ReadRaw( void* buf, int len, int timeout ); + bool HasData(); + bool IsValid() const; + + Socket( const Socket& ) = delete; + Socket( Socket&& ) = delete; + Socket& operator=( const Socket& ) = delete; + Socket& operator=( Socket&& ) = delete; + +private: + int RecvBuffered( void* buf, int len, int timeout ); + int Recv( void* buf, int len, int timeout ); + + bool ReadImpl( char*& buf, int& len, int timeout ); + + char* m_buf; + char* m_bufPtr; + std::atomic m_sock; + int m_bufLeft; + + struct addrinfo *m_res; + struct addrinfo *m_ptr; + int m_connSock; +}; + +class ListenSocket +{ +public: + ListenSocket(); + ~ListenSocket(); + + bool Listen( uint16_t port, int backlog ); + Socket* Accept(); + void Close(); + + ListenSocket( const ListenSocket& ) = delete; + ListenSocket( ListenSocket&& ) = delete; + ListenSocket& operator=( const ListenSocket& ) = delete; + ListenSocket& operator=( ListenSocket&& ) = delete; + +private: + int m_sock; +}; + +class UdpBroadcast +{ +public: + UdpBroadcast(); + ~UdpBroadcast(); + + bool Open( const char* addr, uint16_t port ); + void Close(); + + int Send( uint16_t port, const void* data, int len ); + + UdpBroadcast( const UdpBroadcast& ) = delete; + UdpBroadcast( UdpBroadcast&& ) = delete; + UdpBroadcast& operator=( const UdpBroadcast& ) = delete; + UdpBroadcast& operator=( UdpBroadcast&& ) = delete; + +private: + int m_sock; + uint32_t m_addr; +}; + +class IpAddress +{ +public: + IpAddress(); + ~IpAddress(); + + void Set( const struct sockaddr& addr ); + + uint32_t GetNumber() const { return m_number; } + const char* GetText() const { return m_text; } + + IpAddress( const IpAddress& ) = delete; + IpAddress( IpAddress&& ) = delete; + IpAddress& operator=( const IpAddress& ) = delete; + IpAddress& operator=( IpAddress&& ) = delete; + +private: + uint32_t m_number; + char m_text[17]; +}; + +class UdpListen +{ +public: + UdpListen(); + ~UdpListen(); + + bool Listen( uint16_t port ); + void Close(); + + const char* Read( size_t& len, IpAddress& addr, int timeout ); + + UdpListen( const UdpListen& ) = delete; + UdpListen( UdpListen&& ) = delete; + UdpListen& operator=( const UdpListen& ) = delete; + UdpListen& operator=( UdpListen&& ) = delete; + +private: + int m_sock; +}; + +} + +#endif diff --git a/externals/tracy/public/common/TracyStackFrames.cpp b/externals/tracy/public/common/TracyStackFrames.cpp new file mode 100644 index 000000000..7b0abace3 --- /dev/null +++ b/externals/tracy/public/common/TracyStackFrames.cpp @@ -0,0 +1,122 @@ +#include "TracyStackFrames.hpp" + +namespace tracy +{ + +const char* s_tracyStackFrames_[] = { + "tracy::Callstack", + "tracy::Callstack(int)", + "tracy::GpuCtxScope::{ctor}", + "tracy::Profiler::SendCallstack", + "tracy::Profiler::SendCallstack(int)", + "tracy::Profiler::SendCallstack(int, unsigned long)", + "tracy::Profiler::MemAllocCallstack", + "tracy::Profiler::MemAllocCallstack(void const*, unsigned long, int)", + "tracy::Profiler::MemFreeCallstack", + "tracy::Profiler::MemFreeCallstack(void const*, int)", + "tracy::ScopedZone::{ctor}", + "tracy::ScopedZone::ScopedZone(tracy::SourceLocationData const*, int, bool)", + "tracy::Profiler::Message", + nullptr +}; + +const char** s_tracyStackFrames = s_tracyStackFrames_; + +const StringMatch s_tracySkipSubframes_[] = { + { "/include/arm_neon.h", 19 }, + { "/include/adxintrin.h", 20 }, + { "/include/ammintrin.h", 20 }, + { "/include/amxbf16intrin.h", 24 }, + { "/include/amxint8intrin.h", 24 }, + { "/include/amxtileintrin.h", 24 }, + { "/include/avx2intrin.h", 21 }, + { "/include/avx5124fmapsintrin.h", 29 }, + { "/include/avx5124vnniwintrin.h", 29 }, + { "/include/avx512bf16intrin.h", 27 }, + { "/include/avx512bf16vlintrin.h", 29 }, + { "/include/avx512bitalgintrin.h", 29 }, + { "/include/avx512bwintrin.h", 25 }, + { "/include/avx512cdintrin.h", 25 }, + { "/include/avx512dqintrin.h", 25 }, + { "/include/avx512erintrin.h", 25 }, + { "/include/avx512fintrin.h", 24 }, + { "/include/avx512ifmaintrin.h", 27 }, + { "/include/avx512ifmavlintrin.h", 29 }, + { "/include/avx512pfintrin.h", 25 }, + { "/include/avx512vbmi2intrin.h", 28 }, + { "/include/avx512vbmi2vlintrin.h", 30 }, + { "/include/avx512vbmiintrin.h", 27 }, + { "/include/avx512vbmivlintrin.h", 29 }, + { "/include/avx512vlbwintrin.h", 27 }, + { "/include/avx512vldqintrin.h", 27 }, + { "/include/avx512vlintrin.h", 25 }, + { "/include/avx512vnniintrin.h", 27 }, + { "/include/avx512vnnivlintrin.h", 29 }, + { "/include/avx512vp2intersectintrin.h", 35 }, + { "/include/avx512vp2intersectvlintrin.h", 37 }, + { "/include/avx512vpopcntdqintrin.h", 32 }, + { "/include/avx512vpopcntdqvlintrin.h", 34 }, + { "/include/avxintrin.h", 20 }, + { "/include/avxvnniintrin.h", 24 }, + { "/include/bmi2intrin.h", 21 }, + { "/include/bmiintrin.h", 20 }, + { "/include/bmmintrin.h", 20 }, + { "/include/cetintrin.h", 20 }, + { "/include/cldemoteintrin.h", 25 }, + { "/include/clflushoptintrin.h", 27 }, + { "/include/clwbintrin.h", 21 }, + { "/include/clzerointrin.h", 23 }, + { "/include/emmintrin.h", 20 }, + { "/include/enqcmdintrin.h", 23 }, + { "/include/f16cintrin.h", 21 }, + { "/include/fma4intrin.h", 21 }, + { "/include/fmaintrin.h", 20 }, + { "/include/fxsrintrin.h", 21 }, + { "/include/gfniintrin.h", 21 }, + { "/include/hresetintrin.h", 23 }, + { "/include/ia32intrin.h", 21 }, + { "/include/immintrin.h", 20 }, + { "/include/keylockerintrin.h", 26 }, + { "/include/lwpintrin.h", 20 }, + { "/include/lzcntintrin.h", 22 }, + { "/include/mmintrin.h", 19 }, + { "/include/movdirintrin.h", 23 }, + { "/include/mwaitxintrin.h", 23 }, + { "/include/nmmintrin.h", 20 }, + { "/include/pconfigintrin.h", 24 }, + { "/include/pkuintrin.h", 20 }, + { "/include/pmmintrin.h", 20 }, + { "/include/popcntintrin.h", 23 }, + { "/include/prfchwintrin.h", 23 }, + { "/include/rdseedintrin.h", 23 }, + { "/include/rtmintrin.h", 20 }, + { "/include/serializeintrin.h", 26 }, + { "/include/sgxintrin.h", 20 }, + { "/include/shaintrin.h", 20 }, + { "/include/smmintrin.h", 20 }, + { "/include/tbmintrin.h", 20 }, + { "/include/tmmintrin.h", 20 }, + { "/include/tsxldtrkintrin.h", 25 }, + { "/include/uintrintrin.h", 22 }, + { "/include/vaesintrin.h", 21 }, + { "/include/vpclmulqdqintrin.h", 27 }, + { "/include/waitpkgintrin.h", 24 }, + { "/include/wbnoinvdintrin.h", 25 }, + { "/include/wmmintrin.h", 20 }, + { "/include/x86gprintrin.h", 23 }, + { "/include/x86intrin.h", 20 }, + { "/include/xmmintrin.h", 20 }, + { "/include/xopintrin.h", 20 }, + { "/include/xsavecintrin.h", 23 }, + { "/include/xsaveintrin.h", 22 }, + { "/include/xsaveoptintrin.h", 25 }, + { "/include/xsavesintrin.h", 23 }, + { "/include/xtestintrin.h", 22 }, + { "/bits/atomic_base.h", 19 }, + { "/atomic", 7 }, + {} +}; + +const StringMatch* s_tracySkipSubframes = s_tracySkipSubframes_; + +} diff --git a/externals/tracy/public/common/TracyStackFrames.hpp b/externals/tracy/public/common/TracyStackFrames.hpp new file mode 100644 index 000000000..9d4262c00 --- /dev/null +++ b/externals/tracy/public/common/TracyStackFrames.hpp @@ -0,0 +1,22 @@ +#ifndef __TRACYSTACKFRAMES_HPP__ +#define __TRACYSTACKFRAMES_HPP__ + +#include + +namespace tracy +{ + +struct StringMatch +{ + const char* str; + size_t len; +}; + +extern const char** s_tracyStackFrames; +extern const StringMatch* s_tracySkipSubframes; + +static constexpr int s_tracySkipSubframesMinLen = 7; + +} + +#endif diff --git a/externals/tracy/public/common/TracySystem.cpp b/externals/tracy/public/common/TracySystem.cpp new file mode 100644 index 000000000..0e26aecae --- /dev/null +++ b/externals/tracy/public/common/TracySystem.cpp @@ -0,0 +1,331 @@ +#ifdef _MSC_VER +# pragma warning(disable:4996) +#endif +#if defined _WIN32 +# ifndef WIN32_LEAN_AND_MEAN +# define WIN32_LEAN_AND_MEAN +# endif +# ifndef NOMINMAX +# define NOMINMAX +# endif +# include +# include +# include "TracyUwp.hpp" +#else +# include +# include +# include +#endif + +#ifdef __linux__ +# ifdef __ANDROID__ +# include +# else +# include +# endif +# include +#elif defined __FreeBSD__ +# include +#elif defined __NetBSD__ || defined __DragonFly__ +# include +#elif defined __QNX__ +# include +# include +#endif + +#ifdef __MINGW32__ +# define __STDC_FORMAT_MACROS +#endif +#include +#include +#include + +#include "TracySystem.hpp" + +#if defined _WIN32 +extern "C" typedef HRESULT (WINAPI *t_SetThreadDescription)( HANDLE, PCWSTR ); +extern "C" typedef HRESULT (WINAPI *t_GetThreadDescription)( HANDLE, PWSTR* ); +#endif + +#ifdef TRACY_ENABLE +# include +# include "TracyAlloc.hpp" +#endif + +namespace tracy +{ + +namespace detail +{ + +TRACY_API uint32_t GetThreadHandleImpl() +{ +#if defined _WIN32 + static_assert( sizeof( decltype( GetCurrentThreadId() ) ) <= sizeof( uint32_t ), "Thread handle too big to fit in protocol" ); + return uint32_t( GetCurrentThreadId() ); +#elif defined __APPLE__ + uint64_t id; + pthread_threadid_np( pthread_self(), &id ); + return uint32_t( id ); +#elif defined __ANDROID__ + return (uint32_t)gettid(); +#elif defined __linux__ + return (uint32_t)syscall( SYS_gettid ); +#elif defined __FreeBSD__ + long id; + thr_self( &id ); + return id; +#elif defined __NetBSD__ + return _lwp_self(); +#elif defined __DragonFly__ + return lwp_gettid(); +#elif defined __OpenBSD__ + return getthrid(); +#elif defined __QNX__ + return (uint32_t) gettid(); +#elif defined __EMSCRIPTEN__ + // Not supported, but let it compile. + return 0; +#else + // To add support for a platform, retrieve and return the kernel thread identifier here. + // + // Note that pthread_t (as for example returned by pthread_self()) is *not* a kernel + // thread identifier. It is a pointer to a library-allocated data structure instead. + // Such pointers will be reused heavily, making the pthread_t non-unique. Additionally + // a 64-bit pointer cannot be reliably truncated to 32 bits. + #error "Unsupported platform!" +#endif + +} + +} + +#ifdef TRACY_ENABLE +struct ThreadNameData +{ + uint32_t id; + const char* name; + ThreadNameData* next; +}; +std::atomic& GetThreadNameData(); +#endif + +#ifdef _MSC_VER +# pragma pack( push, 8 ) +struct THREADNAME_INFO +{ + DWORD dwType; + LPCSTR szName; + DWORD dwThreadID; + DWORD dwFlags; +}; +# pragma pack( pop ) + +void ThreadNameMsvcMagic( const THREADNAME_INFO& info ) +{ + __try + { + RaiseException( 0x406D1388, 0, sizeof(info)/sizeof(ULONG_PTR), (ULONG_PTR*)&info ); + } + __except(EXCEPTION_EXECUTE_HANDLER) + { + } +} +#endif + +TRACY_API void SetThreadName( const char* name ) +{ +#if defined _WIN32 +# ifdef TRACY_UWP + static auto _SetThreadDescription = &::SetThreadDescription; +# else + static auto _SetThreadDescription = (t_SetThreadDescription)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "SetThreadDescription" ); +# endif + if( _SetThreadDescription ) + { + wchar_t buf[256]; + mbstowcs( buf, name, 256 ); + _SetThreadDescription( GetCurrentThread(), buf ); + } + else + { +# if defined _MSC_VER + THREADNAME_INFO info; + info.dwType = 0x1000; + info.szName = name; + info.dwThreadID = GetCurrentThreadId(); + info.dwFlags = 0; + ThreadNameMsvcMagic( info ); +# endif + } +#elif defined _GNU_SOURCE && !defined __EMSCRIPTEN__ + { + const auto sz = strlen( name ); + if( sz <= 15 ) + { +#if defined __APPLE__ + pthread_setname_np( name ); +#else + pthread_setname_np( pthread_self(), name ); +#endif + } + else + { + char buf[16]; + memcpy( buf, name, 15 ); + buf[15] = '\0'; +#if defined __APPLE__ + pthread_setname_np( buf ); +#else + pthread_setname_np( pthread_self(), buf ); +#endif + } + } +#elif defined __QNX__ + { + const auto sz = strlen( name ); + if( sz <= _NTO_THREAD_NAME_MAX ) + { + pthread_setname_np( pthread_self(), name ); + } + else + { + char buf[_NTO_THREAD_NAME_MAX + 1]; + memcpy( buf, name, _NTO_THREAD_NAME_MAX ); + buf[_NTO_THREAD_NAME_MAX] = '\0'; + pthread_setname_np( pthread_self(), buf ); + } + }; +#endif +#ifdef TRACY_ENABLE + { + const auto sz = strlen( name ); + char* buf = (char*)tracy_malloc( sz+1 ); + memcpy( buf, name, sz ); + buf[sz] = '\0'; + auto data = (ThreadNameData*)tracy_malloc_fast( sizeof( ThreadNameData ) ); + data->id = detail::GetThreadHandleImpl(); + data->name = buf; + data->next = GetThreadNameData().load( std::memory_order_relaxed ); + while( !GetThreadNameData().compare_exchange_weak( data->next, data, std::memory_order_release, std::memory_order_relaxed ) ) {} + } +#endif +} + +TRACY_API const char* GetThreadName( uint32_t id ) +{ + static char buf[256]; +#ifdef TRACY_ENABLE + auto ptr = GetThreadNameData().load( std::memory_order_relaxed ); + while( ptr ) + { + if( ptr->id == id ) + { + return ptr->name; + } + ptr = ptr->next; + } +#endif + +#if defined _WIN32 +# ifdef TRACY_UWP + static auto _GetThreadDescription = &::GetThreadDescription; +# else + static auto _GetThreadDescription = (t_GetThreadDescription)GetProcAddress( GetModuleHandleA( "kernel32.dll" ), "GetThreadDescription" ); +# endif + if( _GetThreadDescription ) + { + auto hnd = OpenThread( THREAD_QUERY_LIMITED_INFORMATION, FALSE, (DWORD)id ); + if( hnd != 0 ) + { + PWSTR tmp; + if( SUCCEEDED( _GetThreadDescription( hnd, &tmp ) ) ) + { + auto ret = wcstombs( buf, tmp, 256 ); + CloseHandle( hnd ); + LocalFree( tmp ); + if( ret != static_cast( -1 ) ) + { + return buf; + } + } + } + } +#elif defined __linux__ + int cs, fd; + char path[32]; + snprintf( path, sizeof( path ), "/proc/self/task/%d/comm", id ); + sprintf( buf, "%" PRIu32, id ); +# ifndef __ANDROID__ + pthread_setcancelstate( PTHREAD_CANCEL_DISABLE, &cs ); +# endif + if ( ( fd = open( path, O_RDONLY ) ) > 0) { + int len = read( fd, buf, 255 ); + if( len > 0 ) + { + buf[len] = 0; + if( len > 1 && buf[len-1] == '\n' ) + { + buf[len-1] = 0; + } + } + close( fd ); + } +# ifndef __ANDROID__ + pthread_setcancelstate( cs, 0 ); +# endif + return buf; +#elif defined __QNX__ + static char qnxNameBuf[_NTO_THREAD_NAME_MAX + 1] = {0}; + if (pthread_getname_np(static_cast(id), qnxNameBuf, _NTO_THREAD_NAME_MAX) == 0) { + return qnxNameBuf; + }; +#endif + + sprintf( buf, "%" PRIu32, id ); + return buf; +} + +TRACY_API const char* GetEnvVar( const char* name ) +{ +#if defined _WIN32 + // unfortunately getenv() on Windows is just fundamentally broken. It caches the entire + // environment block once on startup, then never refreshes it again. If any environment + // strings are added or modified after startup of the CRT, those changes will not be + // seen by getenv(). This removes the possibility of an app using this SDK from + // programmatically setting any of the behaviour controlling envvars here. + // + // To work around this, we'll instead go directly to the Win32 environment strings APIs + // to get the current value. + static char buffer[1024]; + DWORD const kBufferSize = DWORD(sizeof(buffer) / sizeof(buffer[0])); + DWORD count = GetEnvironmentVariableA(name, buffer, kBufferSize); + + if( count == 0 ) + return nullptr; + + if( count >= kBufferSize ) + { + char* buf = reinterpret_cast(_alloca(count + 1)); + count = GetEnvironmentVariableA(name, buf, count + 1); + memcpy(buffer, buf, kBufferSize); + buffer[kBufferSize - 1] = 0; + } + + return buffer; +#else + return getenv(name); +#endif +} + +} + +#ifdef __cplusplus +extern "C" { +#endif + +TRACY_API void ___tracy_set_thread_name( const char* name ) { tracy::SetThreadName( name ); } + +#ifdef __cplusplus +} +#endif diff --git a/externals/tracy/public/common/TracySystem.hpp b/externals/tracy/public/common/TracySystem.hpp new file mode 100644 index 000000000..aa5a88c10 --- /dev/null +++ b/externals/tracy/public/common/TracySystem.hpp @@ -0,0 +1,32 @@ +#ifndef __TRACYSYSTEM_HPP__ +#define __TRACYSYSTEM_HPP__ + +#include + +#include "TracyApi.h" + +namespace tracy +{ + +namespace detail +{ +TRACY_API uint32_t GetThreadHandleImpl(); +} + +#ifdef TRACY_ENABLE +TRACY_API uint32_t GetThreadHandle(); +#else +static inline uint32_t GetThreadHandle() +{ + return detail::GetThreadHandleImpl(); +} +#endif + +TRACY_API void SetThreadName( const char* name ); +TRACY_API const char* GetThreadName( uint32_t id ); + +TRACY_API const char* GetEnvVar( const char* name ); + +} + +#endif diff --git a/externals/tracy/public/common/TracyUwp.hpp b/externals/tracy/public/common/TracyUwp.hpp new file mode 100644 index 000000000..7dce96b96 --- /dev/null +++ b/externals/tracy/public/common/TracyUwp.hpp @@ -0,0 +1,11 @@ +#ifndef __TRACYUWP_HPP__ +#define __TRACYUWP_HPP__ + +#ifdef _WIN32 +# include +# if WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP) && !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) +# define TRACY_UWP +# endif +#endif + +#endif diff --git a/externals/tracy/public/common/TracyVersion.hpp b/externals/tracy/public/common/TracyVersion.hpp new file mode 100644 index 000000000..2355279f7 --- /dev/null +++ b/externals/tracy/public/common/TracyVersion.hpp @@ -0,0 +1,14 @@ +#ifndef __TRACYVERSION_HPP__ +#define __TRACYVERSION_HPP__ + +namespace tracy +{ +namespace Version +{ +enum { Major = 0 }; +enum { Minor = 10 }; +enum { Patch = 0 }; +} +} + +#endif diff --git a/externals/tracy/public/common/TracyYield.hpp b/externals/tracy/public/common/TracyYield.hpp new file mode 100644 index 000000000..035836cdb --- /dev/null +++ b/externals/tracy/public/common/TracyYield.hpp @@ -0,0 +1,28 @@ +#ifndef __TRACYYIELD_HPP__ +#define __TRACYYIELD_HPP__ + +#if defined __SSE2__ || defined _M_AMD64 || (defined _M_IX86_FP && _M_IX86_FP == 2) +# include +#else +# include +#endif + +#include "TracyForceInline.hpp" + +namespace tracy +{ + +static tracy_force_inline void YieldThread() +{ +#if defined __SSE2__ || defined _M_AMD64 || (defined _M_IX86_FP && _M_IX86_FP == 2) + _mm_pause(); +#elif defined __aarch64__ + asm volatile( "isb" : : ); +#else + std::this_thread::yield(); +#endif +} + +} + +#endif diff --git a/externals/tracy/public/common/tracy_lz4.cpp b/externals/tracy/public/common/tracy_lz4.cpp new file mode 100644 index 000000000..6c26639c5 --- /dev/null +++ b/externals/tracy/public/common/tracy_lz4.cpp @@ -0,0 +1,2720 @@ +/* + LZ4 - Fast LZ compression algorithm + Copyright (C) 2011-2020, Yann Collet. + + BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are + met: + + * Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above + copyright notice, this list of conditions and the following disclaimer + in the documentation and/or other materials provided with the + distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + You can contact the author at : + - LZ4 homepage : http://www.lz4.org + - LZ4 source repository : https://github.com/lz4/lz4 +*/ + +/*-************************************ +* Tuning parameters +**************************************/ +/* + * LZ4_HEAPMODE : + * Select how default compression functions will allocate memory for their hash table, + * in memory stack (0:default, fastest), or in memory heap (1:requires malloc()). + */ +#ifndef LZ4_HEAPMODE +# define LZ4_HEAPMODE 0 +#endif + +/* + * LZ4_ACCELERATION_DEFAULT : + * Select "acceleration" for LZ4_compress_fast() when parameter value <= 0 + */ +#define LZ4_ACCELERATION_DEFAULT 1 +/* + * LZ4_ACCELERATION_MAX : + * Any "acceleration" value higher than this threshold + * get treated as LZ4_ACCELERATION_MAX instead (fix #876) + */ +#define LZ4_ACCELERATION_MAX 65537 + + +/*-************************************ +* CPU Feature Detection +**************************************/ +/* LZ4_FORCE_MEMORY_ACCESS + * By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable. + * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal. + * The below switch allow to select different access method for improved performance. + * Method 0 (default) : use `memcpy()`. Safe and portable. + * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable). + * This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`. + * Method 2 : direct access. This method is portable but violate C standard. + * It can generate buggy code on targets which assembly generation depends on alignment. + * But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6) + * See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details. + * Prefer these methods in priority order (0 > 1 > 2) + */ +#ifndef LZ4_FORCE_MEMORY_ACCESS /* can be defined externally */ +# if defined(__GNUC__) && \ + ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) \ + || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) ) +# define LZ4_FORCE_MEMORY_ACCESS 2 +# elif (defined(__INTEL_COMPILER) && !defined(_WIN32)) || defined(__GNUC__) +# define LZ4_FORCE_MEMORY_ACCESS 1 +# endif +#endif + +/* + * LZ4_FORCE_SW_BITCOUNT + * Define this parameter if your target system or compiler does not support hardware bit count + */ +#if defined(_MSC_VER) && defined(_WIN32_WCE) /* Visual Studio for WinCE doesn't support Hardware bit count */ +# undef LZ4_FORCE_SW_BITCOUNT /* avoid double def */ +# define LZ4_FORCE_SW_BITCOUNT +#endif + + + +/*-************************************ +* Dependency +**************************************/ +/* + * LZ4_SRC_INCLUDED: + * Amalgamation flag, whether lz4.c is included + */ +#ifndef LZ4_SRC_INCLUDED +# define LZ4_SRC_INCLUDED 1 +#endif + +#ifndef LZ4_STATIC_LINKING_ONLY +#define LZ4_STATIC_LINKING_ONLY +#endif + +#ifndef LZ4_DISABLE_DEPRECATE_WARNINGS +#define LZ4_DISABLE_DEPRECATE_WARNINGS /* due to LZ4_decompress_safe_withPrefix64k */ +#endif + +#define LZ4_STATIC_LINKING_ONLY /* LZ4_DISTANCE_MAX */ +#include "tracy_lz4.hpp" +/* see also "memory routines" below */ + + +/*-************************************ +* Compiler Options +**************************************/ +#if defined(_MSC_VER) && (_MSC_VER >= 1400) /* Visual Studio 2005+ */ +# include /* only present in VS2005+ */ +# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ +# pragma warning(disable : 6237) /* disable: C6237: conditional expression is always 0 */ +#endif /* _MSC_VER */ + +#ifndef LZ4_FORCE_INLINE +# ifdef _MSC_VER /* Visual Studio */ +# define LZ4_FORCE_INLINE static __forceinline +# else +# if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */ +# ifdef __GNUC__ +# define LZ4_FORCE_INLINE static inline __attribute__((always_inline)) +# else +# define LZ4_FORCE_INLINE static inline +# endif +# else +# define LZ4_FORCE_INLINE static +# endif /* __STDC_VERSION__ */ +# endif /* _MSC_VER */ +#endif /* LZ4_FORCE_INLINE */ + +/* LZ4_FORCE_O2 and LZ4_FORCE_INLINE + * gcc on ppc64le generates an unrolled SIMDized loop for LZ4_wildCopy8, + * together with a simple 8-byte copy loop as a fall-back path. + * However, this optimization hurts the decompression speed by >30%, + * because the execution does not go to the optimized loop + * for typical compressible data, and all of the preamble checks + * before going to the fall-back path become useless overhead. + * This optimization happens only with the -O3 flag, and -O2 generates + * a simple 8-byte copy loop. + * With gcc on ppc64le, all of the LZ4_decompress_* and LZ4_wildCopy8 + * functions are annotated with __attribute__((optimize("O2"))), + * and also LZ4_wildCopy8 is forcibly inlined, so that the O2 attribute + * of LZ4_wildCopy8 does not affect the compression speed. + */ +#if defined(__PPC64__) && defined(__LITTLE_ENDIAN__) && defined(__GNUC__) && !defined(__clang__) +# define LZ4_FORCE_O2 __attribute__((optimize("O2"))) +# undef LZ4_FORCE_INLINE +# define LZ4_FORCE_INLINE static __inline __attribute__((optimize("O2"),always_inline)) +#else +# define LZ4_FORCE_O2 +#endif + +#if (defined(__GNUC__) && (__GNUC__ >= 3)) || (defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 800)) || defined(__clang__) +# define expect(expr,value) (__builtin_expect ((expr),(value)) ) +#else +# define expect(expr,value) (expr) +#endif + +#ifndef likely +#define likely(expr) expect((expr) != 0, 1) +#endif +#ifndef unlikely +#define unlikely(expr) expect((expr) != 0, 0) +#endif + +/* Should the alignment test prove unreliable, for some reason, + * it can be disabled by setting LZ4_ALIGN_TEST to 0 */ +#ifndef LZ4_ALIGN_TEST /* can be externally provided */ +# define LZ4_ALIGN_TEST 1 +#endif + + +/*-************************************ +* Memory routines +**************************************/ + +/*! LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION : + * Disable relatively high-level LZ4/HC functions that use dynamic memory + * allocation functions (malloc(), calloc(), free()). + * + * Note that this is a compile-time switch. And since it disables + * public/stable LZ4 v1 API functions, we don't recommend using this + * symbol to generate a library for distribution. + * + * The following public functions are removed when this symbol is defined. + * - lz4 : LZ4_createStream, LZ4_freeStream, + * LZ4_createStreamDecode, LZ4_freeStreamDecode, LZ4_create (deprecated) + * - lz4hc : LZ4_createStreamHC, LZ4_freeStreamHC, + * LZ4_createHC (deprecated), LZ4_freeHC (deprecated) + * - lz4frame, lz4file : All LZ4F_* functions + */ +#if defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) +# define ALLOC(s) lz4_error_memory_allocation_is_disabled +# define ALLOC_AND_ZERO(s) lz4_error_memory_allocation_is_disabled +# define FREEMEM(p) lz4_error_memory_allocation_is_disabled +#elif defined(LZ4_USER_MEMORY_FUNCTIONS) +/* memory management functions can be customized by user project. + * Below functions must exist somewhere in the Project + * and be available at link time */ +void* LZ4_malloc(size_t s); +void* LZ4_calloc(size_t n, size_t s); +void LZ4_free(void* p); +# define ALLOC(s) LZ4_malloc(s) +# define ALLOC_AND_ZERO(s) LZ4_calloc(1,s) +# define FREEMEM(p) LZ4_free(p) +#else +# include /* malloc, calloc, free */ +# define ALLOC(s) malloc(s) +# define ALLOC_AND_ZERO(s) calloc(1,s) +# define FREEMEM(p) free(p) +#endif + +#if ! LZ4_FREESTANDING +# include /* memset, memcpy */ +#endif +#if !defined(LZ4_memset) +# define LZ4_memset(p,v,s) memset((p),(v),(s)) +#endif +#define MEM_INIT(p,v,s) LZ4_memset((p),(v),(s)) + + +/*-************************************ +* Common Constants +**************************************/ +#define MINMATCH 4 + +#define WILDCOPYLENGTH 8 +#define LASTLITERALS 5 /* see ../doc/lz4_Block_format.md#parsing-restrictions */ +#define MFLIMIT 12 /* see ../doc/lz4_Block_format.md#parsing-restrictions */ +#define MATCH_SAFEGUARD_DISTANCE ((2*WILDCOPYLENGTH) - MINMATCH) /* ensure it's possible to write 2 x wildcopyLength without overflowing output buffer */ +#define FASTLOOP_SAFE_DISTANCE 64 +static const int LZ4_minLength = (MFLIMIT+1); + +#define KB *(1 <<10) +#define MB *(1 <<20) +#define GB *(1U<<30) + +#define LZ4_DISTANCE_ABSOLUTE_MAX 65535 +#if (LZ4_DISTANCE_MAX > LZ4_DISTANCE_ABSOLUTE_MAX) /* max supported by LZ4 format */ +# error "LZ4_DISTANCE_MAX is too big : must be <= 65535" +#endif + +#define ML_BITS 4 +#define ML_MASK ((1U<=1) +# include +#else +# ifndef assert +# define assert(condition) ((void)0) +# endif +#endif + +#define LZ4_STATIC_ASSERT(c) { enum { LZ4_static_assert = 1/(int)(!!(c)) }; } /* use after variable declarations */ + +#if defined(LZ4_DEBUG) && (LZ4_DEBUG>=2) +# include + static int g_debuglog_enable = 1; +# define DEBUGLOG(l, ...) { \ + if ((g_debuglog_enable) && (l<=LZ4_DEBUG)) { \ + fprintf(stderr, __FILE__ ": "); \ + fprintf(stderr, __VA_ARGS__); \ + fprintf(stderr, " \n"); \ + } } +#else +# define DEBUGLOG(l, ...) {} /* disabled */ +#endif + +static int LZ4_isAligned(const void* ptr, size_t alignment) +{ + return ((size_t)ptr & (alignment -1)) == 0; +} + + +/*-************************************ +* Types +**************************************/ +#include +#if defined(__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) +# include + typedef uint8_t BYTE; + typedef uint16_t U16; + typedef uint32_t U32; + typedef int32_t S32; + typedef uint64_t U64; + typedef uintptr_t uptrval; +#else +# if UINT_MAX != 4294967295UL +# error "LZ4 code (when not C++ or C99) assumes that sizeof(int) == 4" +# endif + typedef unsigned char BYTE; + typedef unsigned short U16; + typedef unsigned int U32; + typedef signed int S32; + typedef unsigned long long U64; + typedef size_t uptrval; /* generally true, except OpenVMS-64 */ +#endif + +#if defined(__x86_64__) + typedef U64 reg_t; /* 64-bits in x32 mode */ +#else + typedef size_t reg_t; /* 32-bits in x32 mode */ +#endif + +typedef enum { + notLimited = 0, + limitedOutput = 1, + fillOutput = 2 +} limitedOutput_directive; + +namespace tracy +{ + +/*-************************************ +* Reading and writing into memory +**************************************/ + +/** + * LZ4 relies on memcpy with a constant size being inlined. In freestanding + * environments, the compiler can't assume the implementation of memcpy() is + * standard compliant, so it can't apply its specialized memcpy() inlining + * logic. When possible, use __builtin_memcpy() to tell the compiler to analyze + * memcpy() as if it were standard compliant, so it can inline it in freestanding + * environments. This is needed when decompressing the Linux Kernel, for example. + */ +#if !defined(LZ4_memcpy) +# if defined(__GNUC__) && (__GNUC__ >= 4) +# define LZ4_memcpy(dst, src, size) __builtin_memcpy(dst, src, size) +# else +# define LZ4_memcpy(dst, src, size) memcpy(dst, src, size) +# endif +#endif + +#if !defined(LZ4_memmove) +# if defined(__GNUC__) && (__GNUC__ >= 4) +# define LZ4_memmove __builtin_memmove +# else +# define LZ4_memmove memmove +# endif +#endif + +static unsigned LZ4_isLittleEndian(void) +{ + const union { U32 u; BYTE c[4]; } one = { 1 }; /* don't use static : performance detrimental */ + return one.c[0]; +} + + +#if defined(LZ4_FORCE_MEMORY_ACCESS) && (LZ4_FORCE_MEMORY_ACCESS==2) +/* lie to the compiler about data alignment; use with caution */ + +static U16 LZ4_read16(const void* memPtr) { return *(const U16*) memPtr; } +static U32 LZ4_read32(const void* memPtr) { return *(const U32*) memPtr; } +static reg_t LZ4_read_ARCH(const void* memPtr) { return *(const reg_t*) memPtr; } + +static void LZ4_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; } +static void LZ4_write32(void* memPtr, U32 value) { *(U32*)memPtr = value; } + +#elif defined(LZ4_FORCE_MEMORY_ACCESS) && (LZ4_FORCE_MEMORY_ACCESS==1) + +/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */ +/* currently only defined for gcc and icc */ +typedef union { U16 u16; U32 u32; reg_t uArch; } __attribute__((packed)) LZ4_unalign; + +static U16 LZ4_read16(const void* ptr) { return ((const LZ4_unalign*)ptr)->u16; } +static U32 LZ4_read32(const void* ptr) { return ((const LZ4_unalign*)ptr)->u32; } +static reg_t LZ4_read_ARCH(const void* ptr) { return ((const LZ4_unalign*)ptr)->uArch; } + +static void LZ4_write16(void* memPtr, U16 value) { ((LZ4_unalign*)memPtr)->u16 = value; } +static void LZ4_write32(void* memPtr, U32 value) { ((LZ4_unalign*)memPtr)->u32 = value; } + +#else /* safe and portable access using memcpy() */ + +static U16 LZ4_read16(const void* memPtr) +{ + U16 val; LZ4_memcpy(&val, memPtr, sizeof(val)); return val; +} + +static U32 LZ4_read32(const void* memPtr) +{ + U32 val; LZ4_memcpy(&val, memPtr, sizeof(val)); return val; +} + +static reg_t LZ4_read_ARCH(const void* memPtr) +{ + reg_t val; LZ4_memcpy(&val, memPtr, sizeof(val)); return val; +} + +static void LZ4_write16(void* memPtr, U16 value) +{ + LZ4_memcpy(memPtr, &value, sizeof(value)); +} + +static void LZ4_write32(void* memPtr, U32 value) +{ + LZ4_memcpy(memPtr, &value, sizeof(value)); +} + +#endif /* LZ4_FORCE_MEMORY_ACCESS */ + + +static U16 LZ4_readLE16(const void* memPtr) +{ + if (LZ4_isLittleEndian()) { + return LZ4_read16(memPtr); + } else { + const BYTE* p = (const BYTE*)memPtr; + return (U16)((U16)p[0] + (p[1]<<8)); + } +} + +static void LZ4_writeLE16(void* memPtr, U16 value) +{ + if (LZ4_isLittleEndian()) { + LZ4_write16(memPtr, value); + } else { + BYTE* p = (BYTE*)memPtr; + p[0] = (BYTE) value; + p[1] = (BYTE)(value>>8); + } +} + +/* customized variant of memcpy, which can overwrite up to 8 bytes beyond dstEnd */ +LZ4_FORCE_INLINE +void LZ4_wildCopy8(void* dstPtr, const void* srcPtr, void* dstEnd) +{ + BYTE* d = (BYTE*)dstPtr; + const BYTE* s = (const BYTE*)srcPtr; + BYTE* const e = (BYTE*)dstEnd; + + do { LZ4_memcpy(d,s,8); d+=8; s+=8; } while (d= 16. */ +LZ4_FORCE_INLINE void +LZ4_wildCopy32(void* dstPtr, const void* srcPtr, void* dstEnd) +{ + BYTE* d = (BYTE*)dstPtr; + const BYTE* s = (const BYTE*)srcPtr; + BYTE* const e = (BYTE*)dstEnd; + + do { LZ4_memcpy(d,s,16); LZ4_memcpy(d+16,s+16,16); d+=32; s+=32; } while (d= dstPtr + MINMATCH + * - there is at least 8 bytes available to write after dstEnd */ +LZ4_FORCE_INLINE void +LZ4_memcpy_using_offset(BYTE* dstPtr, const BYTE* srcPtr, BYTE* dstEnd, const size_t offset) +{ + BYTE v[8]; + + assert(dstEnd >= dstPtr + MINMATCH); + + switch(offset) { + case 1: + MEM_INIT(v, *srcPtr, 8); + break; + case 2: + LZ4_memcpy(v, srcPtr, 2); + LZ4_memcpy(&v[2], srcPtr, 2); +#if defined(_MSC_VER) && (_MSC_VER <= 1933) /* MSVC 2022 ver 17.3 or earlier */ +# pragma warning(push) +# pragma warning(disable : 6385) /* warning C6385: Reading invalid data from 'v'. */ +#endif + LZ4_memcpy(&v[4], v, 4); +#if defined(_MSC_VER) && (_MSC_VER <= 1933) /* MSVC 2022 ver 17.3 or earlier */ +# pragma warning(pop) +#endif + break; + case 4: + LZ4_memcpy(v, srcPtr, 4); + LZ4_memcpy(&v[4], srcPtr, 4); + break; + default: + LZ4_memcpy_using_offset_base(dstPtr, srcPtr, dstEnd, offset); + return; + } + + LZ4_memcpy(dstPtr, v, 8); + dstPtr += 8; + while (dstPtr < dstEnd) { + LZ4_memcpy(dstPtr, v, 8); + dstPtr += 8; + } +} +#endif + + +/*-************************************ +* Common functions +**************************************/ +LZ4_FORCE_INLINE unsigned LZ4_NbCommonBytes (reg_t val) +{ + assert(val != 0); + if (LZ4_isLittleEndian()) { + if (sizeof(val) == 8) { +# if defined(_MSC_VER) && (_MSC_VER >= 1800) && (defined(_M_AMD64) && !defined(_M_ARM64EC)) && !defined(LZ4_FORCE_SW_BITCOUNT) +/*-************************************************************************************************* +* ARM64EC is a Microsoft-designed ARM64 ABI compatible with AMD64 applications on ARM64 Windows 11. +* The ARM64EC ABI does not support AVX/AVX2/AVX512 instructions, nor their relevant intrinsics +* including _tzcnt_u64. Therefore, we need to neuter the _tzcnt_u64 code path for ARM64EC. +****************************************************************************************************/ +# if defined(__clang__) && (__clang_major__ < 10) + /* Avoid undefined clang-cl intrinsics issue. + * See https://github.com/lz4/lz4/pull/1017 for details. */ + return (unsigned)__builtin_ia32_tzcnt_u64(val) >> 3; +# else + /* x64 CPUS without BMI support interpret `TZCNT` as `REP BSF` */ + return (unsigned)_tzcnt_u64(val) >> 3; +# endif +# elif defined(_MSC_VER) && defined(_WIN64) && !defined(LZ4_FORCE_SW_BITCOUNT) + unsigned long r = 0; + _BitScanForward64(&r, (U64)val); + return (unsigned)r >> 3; +# elif (defined(__clang__) || (defined(__GNUC__) && ((__GNUC__ > 3) || \ + ((__GNUC__ == 3) && (__GNUC_MINOR__ >= 4))))) && \ + !defined(LZ4_FORCE_SW_BITCOUNT) + return (unsigned)__builtin_ctzll((U64)val) >> 3; +# else + const U64 m = 0x0101010101010101ULL; + val ^= val - 1; + return (unsigned)(((U64)((val & (m - 1)) * m)) >> 56); +# endif + } else /* 32 bits */ { +# if defined(_MSC_VER) && (_MSC_VER >= 1400) && !defined(LZ4_FORCE_SW_BITCOUNT) + unsigned long r; + _BitScanForward(&r, (U32)val); + return (unsigned)r >> 3; +# elif (defined(__clang__) || (defined(__GNUC__) && ((__GNUC__ > 3) || \ + ((__GNUC__ == 3) && (__GNUC_MINOR__ >= 4))))) && \ + !defined(__TINYC__) && !defined(LZ4_FORCE_SW_BITCOUNT) + return (unsigned)__builtin_ctz((U32)val) >> 3; +# else + const U32 m = 0x01010101; + return (unsigned)((((val - 1) ^ val) & (m - 1)) * m) >> 24; +# endif + } + } else /* Big Endian CPU */ { + if (sizeof(val)==8) { +# if (defined(__clang__) || (defined(__GNUC__) && ((__GNUC__ > 3) || \ + ((__GNUC__ == 3) && (__GNUC_MINOR__ >= 4))))) && \ + !defined(__TINYC__) && !defined(LZ4_FORCE_SW_BITCOUNT) + return (unsigned)__builtin_clzll((U64)val) >> 3; +# else +#if 1 + /* this method is probably faster, + * but adds a 128 bytes lookup table */ + static const unsigned char ctz7_tab[128] = { + 7, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, + 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, + 5, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, + 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, + 6, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, + 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, + 5, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, + 4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0, + }; + U64 const mask = 0x0101010101010101ULL; + U64 const t = (((val >> 8) - mask) | val) & mask; + return ctz7_tab[(t * 0x0080402010080402ULL) >> 57]; +#else + /* this method doesn't consume memory space like the previous one, + * but it contains several branches, + * that may end up slowing execution */ + static const U32 by32 = sizeof(val)*4; /* 32 on 64 bits (goal), 16 on 32 bits. + Just to avoid some static analyzer complaining about shift by 32 on 32-bits target. + Note that this code path is never triggered in 32-bits mode. */ + unsigned r; + if (!(val>>by32)) { r=4; } else { r=0; val>>=by32; } + if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; } + r += (!val); + return r; +#endif +# endif + } else /* 32 bits */ { +# if (defined(__clang__) || (defined(__GNUC__) && ((__GNUC__ > 3) || \ + ((__GNUC__ == 3) && (__GNUC_MINOR__ >= 4))))) && \ + !defined(LZ4_FORCE_SW_BITCOUNT) + return (unsigned)__builtin_clz((U32)val) >> 3; +# else + val >>= 8; + val = ((((val + 0x00FFFF00) | 0x00FFFFFF) + val) | + (val + 0x00FF0000)) >> 24; + return (unsigned)val ^ 3; +# endif + } + } +} + + +#define STEPSIZE sizeof(reg_t) +LZ4_FORCE_INLINE +unsigned LZ4_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* pInLimit) +{ + const BYTE* const pStart = pIn; + + if (likely(pIn < pInLimit-(STEPSIZE-1))) { + reg_t const diff = LZ4_read_ARCH(pMatch) ^ LZ4_read_ARCH(pIn); + if (!diff) { + pIn+=STEPSIZE; pMatch+=STEPSIZE; + } else { + return LZ4_NbCommonBytes(diff); + } } + + while (likely(pIn < pInLimit-(STEPSIZE-1))) { + reg_t const diff = LZ4_read_ARCH(pMatch) ^ LZ4_read_ARCH(pIn); + if (!diff) { pIn+=STEPSIZE; pMatch+=STEPSIZE; continue; } + pIn += LZ4_NbCommonBytes(diff); + return (unsigned)(pIn - pStart); + } + + if ((STEPSIZE==8) && (pIn<(pInLimit-3)) && (LZ4_read32(pMatch) == LZ4_read32(pIn))) { pIn+=4; pMatch+=4; } + if ((pIn<(pInLimit-1)) && (LZ4_read16(pMatch) == LZ4_read16(pIn))) { pIn+=2; pMatch+=2; } + if ((pIn compression run slower on incompressible data */ + + +/*-************************************ +* Local Structures and types +**************************************/ +typedef enum { clearedTable = 0, byPtr, byU32, byU16 } tableType_t; + +/** + * This enum distinguishes several different modes of accessing previous + * content in the stream. + * + * - noDict : There is no preceding content. + * - withPrefix64k : Table entries up to ctx->dictSize before the current blob + * blob being compressed are valid and refer to the preceding + * content (of length ctx->dictSize), which is available + * contiguously preceding in memory the content currently + * being compressed. + * - usingExtDict : Like withPrefix64k, but the preceding content is somewhere + * else in memory, starting at ctx->dictionary with length + * ctx->dictSize. + * - usingDictCtx : Everything concerning the preceding content is + * in a separate context, pointed to by ctx->dictCtx. + * ctx->dictionary, ctx->dictSize, and table entries + * in the current context that refer to positions + * preceding the beginning of the current compression are + * ignored. Instead, ctx->dictCtx->dictionary and ctx->dictCtx + * ->dictSize describe the location and size of the preceding + * content, and matches are found by looking in the ctx + * ->dictCtx->hashTable. + */ +typedef enum { noDict = 0, withPrefix64k, usingExtDict, usingDictCtx } dict_directive; +typedef enum { noDictIssue = 0, dictSmall } dictIssue_directive; + + +/*-************************************ +* Local Utils +**************************************/ +int LZ4_versionNumber (void) { return LZ4_VERSION_NUMBER; } +const char* LZ4_versionString(void) { return LZ4_VERSION_STRING; } +int LZ4_compressBound(int isize) { return LZ4_COMPRESSBOUND(isize); } +int LZ4_sizeofState(void) { return sizeof(LZ4_stream_t); } + + +/*-**************************************** +* Internal Definitions, used only in Tests +*******************************************/ + +int LZ4_compress_forceExtDict (LZ4_stream_t* LZ4_dict, const char* source, char* dest, int srcSize); + +int LZ4_decompress_safe_forceExtDict(const char* source, char* dest, + int compressedSize, int maxOutputSize, + const void* dictStart, size_t dictSize); +int LZ4_decompress_safe_partial_forceExtDict(const char* source, char* dest, + int compressedSize, int targetOutputSize, int dstCapacity, + const void* dictStart, size_t dictSize); + +/*-****************************** +* Compression functions +********************************/ +LZ4_FORCE_INLINE U32 LZ4_hash4(U32 sequence, tableType_t const tableType) +{ + if (tableType == byU16) + return ((sequence * 2654435761U) >> ((MINMATCH*8)-(LZ4_HASHLOG+1))); + else + return ((sequence * 2654435761U) >> ((MINMATCH*8)-LZ4_HASHLOG)); +} + +LZ4_FORCE_INLINE U32 LZ4_hash5(U64 sequence, tableType_t const tableType) +{ + const U32 hashLog = (tableType == byU16) ? LZ4_HASHLOG+1 : LZ4_HASHLOG; + if (LZ4_isLittleEndian()) { + const U64 prime5bytes = 889523592379ULL; + return (U32)(((sequence << 24) * prime5bytes) >> (64 - hashLog)); + } else { + const U64 prime8bytes = 11400714785074694791ULL; + return (U32)(((sequence >> 24) * prime8bytes) >> (64 - hashLog)); + } +} + +LZ4_FORCE_INLINE U32 LZ4_hashPosition(const void* const p, tableType_t const tableType) +{ + if ((sizeof(reg_t)==8) && (tableType != byU16)) return LZ4_hash5(LZ4_read_ARCH(p), tableType); + return LZ4_hash4(LZ4_read32(p), tableType); +} + +LZ4_FORCE_INLINE void LZ4_clearHash(U32 h, void* tableBase, tableType_t const tableType) +{ + switch (tableType) + { + default: /* fallthrough */ + case clearedTable: { /* illegal! */ assert(0); return; } + case byPtr: { const BYTE** hashTable = (const BYTE**)tableBase; hashTable[h] = NULL; return; } + case byU32: { U32* hashTable = (U32*) tableBase; hashTable[h] = 0; return; } + case byU16: { U16* hashTable = (U16*) tableBase; hashTable[h] = 0; return; } + } +} + +LZ4_FORCE_INLINE void LZ4_putIndexOnHash(U32 idx, U32 h, void* tableBase, tableType_t const tableType) +{ + switch (tableType) + { + default: /* fallthrough */ + case clearedTable: /* fallthrough */ + case byPtr: { /* illegal! */ assert(0); return; } + case byU32: { U32* hashTable = (U32*) tableBase; hashTable[h] = idx; return; } + case byU16: { U16* hashTable = (U16*) tableBase; assert(idx < 65536); hashTable[h] = (U16)idx; return; } + } +} + +LZ4_FORCE_INLINE void LZ4_putPositionOnHash(const BYTE* p, U32 h, + void* tableBase, tableType_t const tableType, + const BYTE* srcBase) +{ + switch (tableType) + { + case clearedTable: { /* illegal! */ assert(0); return; } + case byPtr: { const BYTE** hashTable = (const BYTE**)tableBase; hashTable[h] = p; return; } + case byU32: { U32* hashTable = (U32*) tableBase; hashTable[h] = (U32)(p-srcBase); return; } + case byU16: { U16* hashTable = (U16*) tableBase; hashTable[h] = (U16)(p-srcBase); return; } + } +} + +LZ4_FORCE_INLINE void LZ4_putPosition(const BYTE* p, void* tableBase, tableType_t tableType, const BYTE* srcBase) +{ + U32 const h = LZ4_hashPosition(p, tableType); + LZ4_putPositionOnHash(p, h, tableBase, tableType, srcBase); +} + +/* LZ4_getIndexOnHash() : + * Index of match position registered in hash table. + * hash position must be calculated by using base+index, or dictBase+index. + * Assumption 1 : only valid if tableType == byU32 or byU16. + * Assumption 2 : h is presumed valid (within limits of hash table) + */ +LZ4_FORCE_INLINE U32 LZ4_getIndexOnHash(U32 h, const void* tableBase, tableType_t tableType) +{ + LZ4_STATIC_ASSERT(LZ4_MEMORY_USAGE > 2); + if (tableType == byU32) { + const U32* const hashTable = (const U32*) tableBase; + assert(h < (1U << (LZ4_MEMORY_USAGE-2))); + return hashTable[h]; + } + if (tableType == byU16) { + const U16* const hashTable = (const U16*) tableBase; + assert(h < (1U << (LZ4_MEMORY_USAGE-1))); + return hashTable[h]; + } + assert(0); return 0; /* forbidden case */ +} + +static const BYTE* LZ4_getPositionOnHash(U32 h, const void* tableBase, tableType_t tableType, const BYTE* srcBase) +{ + if (tableType == byPtr) { const BYTE* const* hashTable = (const BYTE* const*) tableBase; return hashTable[h]; } + if (tableType == byU32) { const U32* const hashTable = (const U32*) tableBase; return hashTable[h] + srcBase; } + { const U16* const hashTable = (const U16*) tableBase; return hashTable[h] + srcBase; } /* default, to ensure a return */ +} + +LZ4_FORCE_INLINE const BYTE* +LZ4_getPosition(const BYTE* p, + const void* tableBase, tableType_t tableType, + const BYTE* srcBase) +{ + U32 const h = LZ4_hashPosition(p, tableType); + return LZ4_getPositionOnHash(h, tableBase, tableType, srcBase); +} + +LZ4_FORCE_INLINE void +LZ4_prepareTable(LZ4_stream_t_internal* const cctx, + const int inputSize, + const tableType_t tableType) { + /* If the table hasn't been used, it's guaranteed to be zeroed out, and is + * therefore safe to use no matter what mode we're in. Otherwise, we figure + * out if it's safe to leave as is or whether it needs to be reset. + */ + if ((tableType_t)cctx->tableType != clearedTable) { + assert(inputSize >= 0); + if ((tableType_t)cctx->tableType != tableType + || ((tableType == byU16) && cctx->currentOffset + (unsigned)inputSize >= 0xFFFFU) + || ((tableType == byU32) && cctx->currentOffset > 1 GB) + || tableType == byPtr + || inputSize >= 4 KB) + { + DEBUGLOG(4, "LZ4_prepareTable: Resetting table in %p", cctx); + MEM_INIT(cctx->hashTable, 0, LZ4_HASHTABLESIZE); + cctx->currentOffset = 0; + cctx->tableType = (U32)clearedTable; + } else { + DEBUGLOG(4, "LZ4_prepareTable: Re-use hash table (no reset)"); + } + } + + /* Adding a gap, so all previous entries are > LZ4_DISTANCE_MAX back, + * is faster than compressing without a gap. + * However, compressing with currentOffset == 0 is faster still, + * so we preserve that case. + */ + if (cctx->currentOffset != 0 && tableType == byU32) { + DEBUGLOG(5, "LZ4_prepareTable: adding 64KB to currentOffset"); + cctx->currentOffset += 64 KB; + } + + /* Finally, clear history */ + cctx->dictCtx = NULL; + cctx->dictionary = NULL; + cctx->dictSize = 0; +} + +/** LZ4_compress_generic() : + * inlined, to ensure branches are decided at compilation time. + * Presumed already validated at this stage: + * - source != NULL + * - inputSize > 0 + */ +LZ4_FORCE_INLINE int LZ4_compress_generic_validated( + LZ4_stream_t_internal* const cctx, + const char* const source, + char* const dest, + const int inputSize, + int* inputConsumed, /* only written when outputDirective == fillOutput */ + const int maxOutputSize, + const limitedOutput_directive outputDirective, + const tableType_t tableType, + const dict_directive dictDirective, + const dictIssue_directive dictIssue, + const int acceleration) +{ + int result; + const BYTE* ip = (const BYTE*) source; + + U32 const startIndex = cctx->currentOffset; + const BYTE* base = (const BYTE*) source - startIndex; + const BYTE* lowLimit; + + const LZ4_stream_t_internal* dictCtx = (const LZ4_stream_t_internal*) cctx->dictCtx; + const BYTE* const dictionary = + dictDirective == usingDictCtx ? dictCtx->dictionary : cctx->dictionary; + const U32 dictSize = + dictDirective == usingDictCtx ? dictCtx->dictSize : cctx->dictSize; + const U32 dictDelta = (dictDirective == usingDictCtx) ? startIndex - dictCtx->currentOffset : 0; /* make indexes in dictCtx comparable with index in current context */ + + int const maybe_extMem = (dictDirective == usingExtDict) || (dictDirective == usingDictCtx); + U32 const prefixIdxLimit = startIndex - dictSize; /* used when dictDirective == dictSmall */ + const BYTE* const dictEnd = dictionary ? dictionary + dictSize : dictionary; + const BYTE* anchor = (const BYTE*) source; + const BYTE* const iend = ip + inputSize; + const BYTE* const mflimitPlusOne = iend - MFLIMIT + 1; + const BYTE* const matchlimit = iend - LASTLITERALS; + + /* the dictCtx currentOffset is indexed on the start of the dictionary, + * while a dictionary in the current context precedes the currentOffset */ + const BYTE* dictBase = (dictionary == NULL) ? NULL : + (dictDirective == usingDictCtx) ? + dictionary + dictSize - dictCtx->currentOffset : + dictionary + dictSize - startIndex; + + BYTE* op = (BYTE*) dest; + BYTE* const olimit = op + maxOutputSize; + + U32 offset = 0; + U32 forwardH; + + DEBUGLOG(5, "LZ4_compress_generic_validated: srcSize=%i, tableType=%u", inputSize, tableType); + assert(ip != NULL); + /* If init conditions are not met, we don't have to mark stream + * as having dirty context, since no action was taken yet */ + if (outputDirective == fillOutput && maxOutputSize < 1) { return 0; } /* Impossible to store anything */ + if ((tableType == byU16) && (inputSize>=LZ4_64Klimit)) { return 0; } /* Size too large (not within 64K limit) */ + if (tableType==byPtr) assert(dictDirective==noDict); /* only supported use case with byPtr */ + assert(acceleration >= 1); + + lowLimit = (const BYTE*)source - (dictDirective == withPrefix64k ? dictSize : 0); + + /* Update context state */ + if (dictDirective == usingDictCtx) { + /* Subsequent linked blocks can't use the dictionary. */ + /* Instead, they use the block we just compressed. */ + cctx->dictCtx = NULL; + cctx->dictSize = (U32)inputSize; + } else { + cctx->dictSize += (U32)inputSize; + } + cctx->currentOffset += (U32)inputSize; + cctx->tableType = (U32)tableType; + + if (inputSizehashTable, tableType, base); + ip++; forwardH = LZ4_hashPosition(ip, tableType); + + /* Main Loop */ + for ( ; ; ) { + const BYTE* match; + BYTE* token; + const BYTE* filledIp; + + /* Find a match */ + if (tableType == byPtr) { + const BYTE* forwardIp = ip; + int step = 1; + int searchMatchNb = acceleration << LZ4_skipTrigger; + do { + U32 const h = forwardH; + ip = forwardIp; + forwardIp += step; + step = (searchMatchNb++ >> LZ4_skipTrigger); + + if (unlikely(forwardIp > mflimitPlusOne)) goto _last_literals; + assert(ip < mflimitPlusOne); + + match = LZ4_getPositionOnHash(h, cctx->hashTable, tableType, base); + forwardH = LZ4_hashPosition(forwardIp, tableType); + LZ4_putPositionOnHash(ip, h, cctx->hashTable, tableType, base); + + } while ( (match+LZ4_DISTANCE_MAX < ip) + || (LZ4_read32(match) != LZ4_read32(ip)) ); + + } else { /* byU32, byU16 */ + + const BYTE* forwardIp = ip; + int step = 1; + int searchMatchNb = acceleration << LZ4_skipTrigger; + do { + U32 const h = forwardH; + U32 const current = (U32)(forwardIp - base); + U32 matchIndex = LZ4_getIndexOnHash(h, cctx->hashTable, tableType); + assert(matchIndex <= current); + assert(forwardIp - base < (ptrdiff_t)(2 GB - 1)); + ip = forwardIp; + forwardIp += step; + step = (searchMatchNb++ >> LZ4_skipTrigger); + + if (unlikely(forwardIp > mflimitPlusOne)) goto _last_literals; + assert(ip < mflimitPlusOne); + + if (dictDirective == usingDictCtx) { + if (matchIndex < startIndex) { + /* there was no match, try the dictionary */ + assert(tableType == byU32); + matchIndex = LZ4_getIndexOnHash(h, dictCtx->hashTable, byU32); + match = dictBase + matchIndex; + matchIndex += dictDelta; /* make dictCtx index comparable with current context */ + lowLimit = dictionary; + } else { + match = base + matchIndex; + lowLimit = (const BYTE*)source; + } + } else if (dictDirective == usingExtDict) { + if (matchIndex < startIndex) { + DEBUGLOG(7, "extDict candidate: matchIndex=%5u < startIndex=%5u", matchIndex, startIndex); + assert(startIndex - matchIndex >= MINMATCH); + assert(dictBase); + match = dictBase + matchIndex; + lowLimit = dictionary; + } else { + match = base + matchIndex; + lowLimit = (const BYTE*)source; + } + } else { /* single continuous memory segment */ + match = base + matchIndex; + } + forwardH = LZ4_hashPosition(forwardIp, tableType); + LZ4_putIndexOnHash(current, h, cctx->hashTable, tableType); + + DEBUGLOG(7, "candidate at pos=%u (offset=%u \n", matchIndex, current - matchIndex); + if ((dictIssue == dictSmall) && (matchIndex < prefixIdxLimit)) { continue; } /* match outside of valid area */ + assert(matchIndex < current); + if ( ((tableType != byU16) || (LZ4_DISTANCE_MAX < LZ4_DISTANCE_ABSOLUTE_MAX)) + && (matchIndex+LZ4_DISTANCE_MAX < current)) { + continue; + } /* too far */ + assert((current - matchIndex) <= LZ4_DISTANCE_MAX); /* match now expected within distance */ + + if (LZ4_read32(match) == LZ4_read32(ip)) { + if (maybe_extMem) offset = current - matchIndex; + break; /* match found */ + } + + } while(1); + } + + /* Catch up */ + filledIp = ip; + while (((ip>anchor) & (match > lowLimit)) && (unlikely(ip[-1]==match[-1]))) { ip--; match--; } + + /* Encode Literals */ + { unsigned const litLength = (unsigned)(ip - anchor); + token = op++; + if ((outputDirective == limitedOutput) && /* Check output buffer overflow */ + (unlikely(op + litLength + (2 + 1 + LASTLITERALS) + (litLength/255) > olimit)) ) { + return 0; /* cannot compress within `dst` budget. Stored indexes in hash table are nonetheless fine */ + } + if ((outputDirective == fillOutput) && + (unlikely(op + (litLength+240)/255 /* litlen */ + litLength /* literals */ + 2 /* offset */ + 1 /* token */ + MFLIMIT - MINMATCH /* min last literals so last match is <= end - MFLIMIT */ > olimit))) { + op--; + goto _last_literals; + } + if (litLength >= RUN_MASK) { + int len = (int)(litLength - RUN_MASK); + *token = (RUN_MASK<= 255 ; len-=255) *op++ = 255; + *op++ = (BYTE)len; + } + else *token = (BYTE)(litLength< olimit)) { + /* the match was too close to the end, rewind and go to last literals */ + op = token; + goto _last_literals; + } + + /* Encode Offset */ + if (maybe_extMem) { /* static test */ + DEBUGLOG(6, " with offset=%u (ext if > %i)", offset, (int)(ip - (const BYTE*)source)); + assert(offset <= LZ4_DISTANCE_MAX && offset > 0); + LZ4_writeLE16(op, (U16)offset); op+=2; + } else { + DEBUGLOG(6, " with offset=%u (same segment)", (U32)(ip - match)); + assert(ip-match <= LZ4_DISTANCE_MAX); + LZ4_writeLE16(op, (U16)(ip - match)); op+=2; + } + + /* Encode MatchLength */ + { unsigned matchCode; + + if ( (dictDirective==usingExtDict || dictDirective==usingDictCtx) + && (lowLimit==dictionary) /* match within extDict */ ) { + const BYTE* limit = ip + (dictEnd-match); + assert(dictEnd > match); + if (limit > matchlimit) limit = matchlimit; + matchCode = LZ4_count(ip+MINMATCH, match+MINMATCH, limit); + ip += (size_t)matchCode + MINMATCH; + if (ip==limit) { + unsigned const more = LZ4_count(limit, (const BYTE*)source, matchlimit); + matchCode += more; + ip += more; + } + DEBUGLOG(6, " with matchLength=%u starting in extDict", matchCode+MINMATCH); + } else { + matchCode = LZ4_count(ip+MINMATCH, match+MINMATCH, matchlimit); + ip += (size_t)matchCode + MINMATCH; + DEBUGLOG(6, " with matchLength=%u", matchCode+MINMATCH); + } + + if ((outputDirective) && /* Check output buffer overflow */ + (unlikely(op + (1 + LASTLITERALS) + (matchCode+240)/255 > olimit)) ) { + if (outputDirective == fillOutput) { + /* Match description too long : reduce it */ + U32 newMatchCode = 15 /* in token */ - 1 /* to avoid needing a zero byte */ + ((U32)(olimit - op) - 1 - LASTLITERALS) * 255; + ip -= matchCode - newMatchCode; + assert(newMatchCode < matchCode); + matchCode = newMatchCode; + if (unlikely(ip <= filledIp)) { + /* We have already filled up to filledIp so if ip ends up less than filledIp + * we have positions in the hash table beyond the current position. This is + * a problem if we reuse the hash table. So we have to remove these positions + * from the hash table. + */ + const BYTE* ptr; + DEBUGLOG(5, "Clearing %u positions", (U32)(filledIp - ip)); + for (ptr = ip; ptr <= filledIp; ++ptr) { + U32 const h = LZ4_hashPosition(ptr, tableType); + LZ4_clearHash(h, cctx->hashTable, tableType); + } + } + } else { + assert(outputDirective == limitedOutput); + return 0; /* cannot compress within `dst` budget. Stored indexes in hash table are nonetheless fine */ + } + } + if (matchCode >= ML_MASK) { + *token += ML_MASK; + matchCode -= ML_MASK; + LZ4_write32(op, 0xFFFFFFFF); + while (matchCode >= 4*255) { + op+=4; + LZ4_write32(op, 0xFFFFFFFF); + matchCode -= 4*255; + } + op += matchCode / 255; + *op++ = (BYTE)(matchCode % 255); + } else + *token += (BYTE)(matchCode); + } + /* Ensure we have enough space for the last literals. */ + assert(!(outputDirective == fillOutput && op + 1 + LASTLITERALS > olimit)); + + anchor = ip; + + /* Test end of chunk */ + if (ip >= mflimitPlusOne) break; + + /* Fill table */ + LZ4_putPosition(ip-2, cctx->hashTable, tableType, base); + + /* Test next position */ + if (tableType == byPtr) { + + match = LZ4_getPosition(ip, cctx->hashTable, tableType, base); + LZ4_putPosition(ip, cctx->hashTable, tableType, base); + if ( (match+LZ4_DISTANCE_MAX >= ip) + && (LZ4_read32(match) == LZ4_read32(ip)) ) + { token=op++; *token=0; goto _next_match; } + + } else { /* byU32, byU16 */ + + U32 const h = LZ4_hashPosition(ip, tableType); + U32 const current = (U32)(ip-base); + U32 matchIndex = LZ4_getIndexOnHash(h, cctx->hashTable, tableType); + assert(matchIndex < current); + if (dictDirective == usingDictCtx) { + if (matchIndex < startIndex) { + /* there was no match, try the dictionary */ + matchIndex = LZ4_getIndexOnHash(h, dictCtx->hashTable, byU32); + match = dictBase + matchIndex; + lowLimit = dictionary; /* required for match length counter */ + matchIndex += dictDelta; + } else { + match = base + matchIndex; + lowLimit = (const BYTE*)source; /* required for match length counter */ + } + } else if (dictDirective==usingExtDict) { + if (matchIndex < startIndex) { + assert(dictBase); + match = dictBase + matchIndex; + lowLimit = dictionary; /* required for match length counter */ + } else { + match = base + matchIndex; + lowLimit = (const BYTE*)source; /* required for match length counter */ + } + } else { /* single memory segment */ + match = base + matchIndex; + } + LZ4_putIndexOnHash(current, h, cctx->hashTable, tableType); + assert(matchIndex < current); + if ( ((dictIssue==dictSmall) ? (matchIndex >= prefixIdxLimit) : 1) + && (((tableType==byU16) && (LZ4_DISTANCE_MAX == LZ4_DISTANCE_ABSOLUTE_MAX)) ? 1 : (matchIndex+LZ4_DISTANCE_MAX >= current)) + && (LZ4_read32(match) == LZ4_read32(ip)) ) { + token=op++; + *token=0; + if (maybe_extMem) offset = current - matchIndex; + DEBUGLOG(6, "seq.start:%i, literals=%u, match.start:%i", + (int)(anchor-(const BYTE*)source), 0, (int)(ip-(const BYTE*)source)); + goto _next_match; + } + } + + /* Prepare next loop */ + forwardH = LZ4_hashPosition(++ip, tableType); + + } + +_last_literals: + /* Encode Last Literals */ + { size_t lastRun = (size_t)(iend - anchor); + if ( (outputDirective) && /* Check output buffer overflow */ + (op + lastRun + 1 + ((lastRun+255-RUN_MASK)/255) > olimit)) { + if (outputDirective == fillOutput) { + /* adapt lastRun to fill 'dst' */ + assert(olimit >= op); + lastRun = (size_t)(olimit-op) - 1/*token*/; + lastRun -= (lastRun + 256 - RUN_MASK) / 256; /*additional length tokens*/ + } else { + assert(outputDirective == limitedOutput); + return 0; /* cannot compress within `dst` budget. Stored indexes in hash table are nonetheless fine */ + } + } + DEBUGLOG(6, "Final literal run : %i literals", (int)lastRun); + if (lastRun >= RUN_MASK) { + size_t accumulator = lastRun - RUN_MASK; + *op++ = RUN_MASK << ML_BITS; + for(; accumulator >= 255 ; accumulator-=255) *op++ = 255; + *op++ = (BYTE) accumulator; + } else { + *op++ = (BYTE)(lastRun< 0); + DEBUGLOG(5, "LZ4_compress_generic: compressed %i bytes into %i bytes", inputSize, result); + return result; +} + +/** LZ4_compress_generic() : + * inlined, to ensure branches are decided at compilation time; + * takes care of src == (NULL, 0) + * and forward the rest to LZ4_compress_generic_validated */ +LZ4_FORCE_INLINE int LZ4_compress_generic( + LZ4_stream_t_internal* const cctx, + const char* const src, + char* const dst, + const int srcSize, + int *inputConsumed, /* only written when outputDirective == fillOutput */ + const int dstCapacity, + const limitedOutput_directive outputDirective, + const tableType_t tableType, + const dict_directive dictDirective, + const dictIssue_directive dictIssue, + const int acceleration) +{ + DEBUGLOG(5, "LZ4_compress_generic: srcSize=%i, dstCapacity=%i", + srcSize, dstCapacity); + + if ((U32)srcSize > (U32)LZ4_MAX_INPUT_SIZE) { return 0; } /* Unsupported srcSize, too large (or negative) */ + if (srcSize == 0) { /* src == NULL supported if srcSize == 0 */ + if (outputDirective != notLimited && dstCapacity <= 0) return 0; /* no output, can't write anything */ + DEBUGLOG(5, "Generating an empty block"); + assert(outputDirective == notLimited || dstCapacity >= 1); + assert(dst != NULL); + dst[0] = 0; + if (outputDirective == fillOutput) { + assert (inputConsumed != NULL); + *inputConsumed = 0; + } + return 1; + } + assert(src != NULL); + + return LZ4_compress_generic_validated(cctx, src, dst, srcSize, + inputConsumed, /* only written into if outputDirective == fillOutput */ + dstCapacity, outputDirective, + tableType, dictDirective, dictIssue, acceleration); +} + + +int LZ4_compress_fast_extState(void* state, const char* source, char* dest, int inputSize, int maxOutputSize, int acceleration) +{ + LZ4_stream_t_internal* const ctx = & LZ4_initStream(state, sizeof(LZ4_stream_t)) -> internal_donotuse; + assert(ctx != NULL); + if (acceleration < 1) acceleration = LZ4_ACCELERATION_DEFAULT; + if (acceleration > LZ4_ACCELERATION_MAX) acceleration = LZ4_ACCELERATION_MAX; + if (maxOutputSize >= LZ4_compressBound(inputSize)) { + if (inputSize < LZ4_64Klimit) { + return LZ4_compress_generic(ctx, source, dest, inputSize, NULL, 0, notLimited, byU16, noDict, noDictIssue, acceleration); + } else { + const tableType_t tableType = ((sizeof(void*)==4) && ((uptrval)source > LZ4_DISTANCE_MAX)) ? byPtr : byU32; + return LZ4_compress_generic(ctx, source, dest, inputSize, NULL, 0, notLimited, tableType, noDict, noDictIssue, acceleration); + } + } else { + if (inputSize < LZ4_64Klimit) { + return LZ4_compress_generic(ctx, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, byU16, noDict, noDictIssue, acceleration); + } else { + const tableType_t tableType = ((sizeof(void*)==4) && ((uptrval)source > LZ4_DISTANCE_MAX)) ? byPtr : byU32; + return LZ4_compress_generic(ctx, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, noDict, noDictIssue, acceleration); + } + } +} + +/** + * LZ4_compress_fast_extState_fastReset() : + * A variant of LZ4_compress_fast_extState(). + * + * Using this variant avoids an expensive initialization step. It is only safe + * to call if the state buffer is known to be correctly initialized already + * (see comment in lz4.h on LZ4_resetStream_fast() for a definition of + * "correctly initialized"). + */ +int LZ4_compress_fast_extState_fastReset(void* state, const char* src, char* dst, int srcSize, int dstCapacity, int acceleration) +{ + LZ4_stream_t_internal* ctx = &((LZ4_stream_t*)state)->internal_donotuse; + if (acceleration < 1) acceleration = LZ4_ACCELERATION_DEFAULT; + if (acceleration > LZ4_ACCELERATION_MAX) acceleration = LZ4_ACCELERATION_MAX; + + if (dstCapacity >= LZ4_compressBound(srcSize)) { + if (srcSize < LZ4_64Klimit) { + const tableType_t tableType = byU16; + LZ4_prepareTable(ctx, srcSize, tableType); + if (ctx->currentOffset) { + return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, 0, notLimited, tableType, noDict, dictSmall, acceleration); + } else { + return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, 0, notLimited, tableType, noDict, noDictIssue, acceleration); + } + } else { + const tableType_t tableType = ((sizeof(void*)==4) && ((uptrval)src > LZ4_DISTANCE_MAX)) ? byPtr : byU32; + LZ4_prepareTable(ctx, srcSize, tableType); + return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, 0, notLimited, tableType, noDict, noDictIssue, acceleration); + } + } else { + if (srcSize < LZ4_64Klimit) { + const tableType_t tableType = byU16; + LZ4_prepareTable(ctx, srcSize, tableType); + if (ctx->currentOffset) { + return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, dstCapacity, limitedOutput, tableType, noDict, dictSmall, acceleration); + } else { + return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, dstCapacity, limitedOutput, tableType, noDict, noDictIssue, acceleration); + } + } else { + const tableType_t tableType = ((sizeof(void*)==4) && ((uptrval)src > LZ4_DISTANCE_MAX)) ? byPtr : byU32; + LZ4_prepareTable(ctx, srcSize, tableType); + return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, dstCapacity, limitedOutput, tableType, noDict, noDictIssue, acceleration); + } + } +} + + +int LZ4_compress_fast(const char* source, char* dest, int inputSize, int maxOutputSize, int acceleration) +{ + int result; +#if (LZ4_HEAPMODE) + LZ4_stream_t* ctxPtr = (LZ4_stream_t*)ALLOC(sizeof(LZ4_stream_t)); /* malloc-calloc always properly aligned */ + if (ctxPtr == NULL) return 0; +#else + LZ4_stream_t ctx; + LZ4_stream_t* const ctxPtr = &ctx; +#endif + result = LZ4_compress_fast_extState(ctxPtr, source, dest, inputSize, maxOutputSize, acceleration); + +#if (LZ4_HEAPMODE) + FREEMEM(ctxPtr); +#endif + return result; +} + + +int LZ4_compress_default(const char* src, char* dst, int srcSize, int maxOutputSize) +{ + return LZ4_compress_fast(src, dst, srcSize, maxOutputSize, 1); +} + + +/* Note!: This function leaves the stream in an unclean/broken state! + * It is not safe to subsequently use the same state with a _fastReset() or + * _continue() call without resetting it. */ +static int LZ4_compress_destSize_extState (LZ4_stream_t* state, const char* src, char* dst, int* srcSizePtr, int targetDstSize) +{ + void* const s = LZ4_initStream(state, sizeof (*state)); + assert(s != NULL); (void)s; + + if (targetDstSize >= LZ4_compressBound(*srcSizePtr)) { /* compression success is guaranteed */ + return LZ4_compress_fast_extState(state, src, dst, *srcSizePtr, targetDstSize, 1); + } else { + if (*srcSizePtr < LZ4_64Klimit) { + return LZ4_compress_generic(&state->internal_donotuse, src, dst, *srcSizePtr, srcSizePtr, targetDstSize, fillOutput, byU16, noDict, noDictIssue, 1); + } else { + tableType_t const addrMode = ((sizeof(void*)==4) && ((uptrval)src > LZ4_DISTANCE_MAX)) ? byPtr : byU32; + return LZ4_compress_generic(&state->internal_donotuse, src, dst, *srcSizePtr, srcSizePtr, targetDstSize, fillOutput, addrMode, noDict, noDictIssue, 1); + } } +} + + +int LZ4_compress_destSize(const char* src, char* dst, int* srcSizePtr, int targetDstSize) +{ +#if (LZ4_HEAPMODE) + LZ4_stream_t* ctx = (LZ4_stream_t*)ALLOC(sizeof(LZ4_stream_t)); /* malloc-calloc always properly aligned */ + if (ctx == NULL) return 0; +#else + LZ4_stream_t ctxBody; + LZ4_stream_t* ctx = &ctxBody; +#endif + + int result = LZ4_compress_destSize_extState(ctx, src, dst, srcSizePtr, targetDstSize); + +#if (LZ4_HEAPMODE) + FREEMEM(ctx); +#endif + return result; +} + + + +/*-****************************** +* Streaming functions +********************************/ + +#if !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) +LZ4_stream_t* LZ4_createStream(void) +{ + LZ4_stream_t* const lz4s = (LZ4_stream_t*)ALLOC(sizeof(LZ4_stream_t)); + LZ4_STATIC_ASSERT(sizeof(LZ4_stream_t) >= sizeof(LZ4_stream_t_internal)); + DEBUGLOG(4, "LZ4_createStream %p", lz4s); + if (lz4s == NULL) return NULL; + LZ4_initStream(lz4s, sizeof(*lz4s)); + return lz4s; +} +#endif + +static size_t LZ4_stream_t_alignment(void) +{ +#if LZ4_ALIGN_TEST + typedef struct { char c; LZ4_stream_t t; } t_a; + return sizeof(t_a) - sizeof(LZ4_stream_t); +#else + return 1; /* effectively disabled */ +#endif +} + +LZ4_stream_t* LZ4_initStream (void* buffer, size_t size) +{ + DEBUGLOG(5, "LZ4_initStream"); + if (buffer == NULL) { return NULL; } + if (size < sizeof(LZ4_stream_t)) { return NULL; } + if (!LZ4_isAligned(buffer, LZ4_stream_t_alignment())) return NULL; + MEM_INIT(buffer, 0, sizeof(LZ4_stream_t_internal)); + return (LZ4_stream_t*)buffer; +} + +/* resetStream is now deprecated, + * prefer initStream() which is more general */ +void LZ4_resetStream (LZ4_stream_t* LZ4_stream) +{ + DEBUGLOG(5, "LZ4_resetStream (ctx:%p)", LZ4_stream); + MEM_INIT(LZ4_stream, 0, sizeof(LZ4_stream_t_internal)); +} + +void LZ4_resetStream_fast(LZ4_stream_t* ctx) { + LZ4_prepareTable(&(ctx->internal_donotuse), 0, byU32); +} + +#if !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) +int LZ4_freeStream (LZ4_stream_t* LZ4_stream) +{ + if (!LZ4_stream) return 0; /* support free on NULL */ + DEBUGLOG(5, "LZ4_freeStream %p", LZ4_stream); + FREEMEM(LZ4_stream); + return (0); +} +#endif + + +#define HASH_UNIT sizeof(reg_t) +int LZ4_loadDict (LZ4_stream_t* LZ4_dict, const char* dictionary, int dictSize) +{ + LZ4_stream_t_internal* dict = &LZ4_dict->internal_donotuse; + const tableType_t tableType = byU32; + const BYTE* p = (const BYTE*)dictionary; + const BYTE* const dictEnd = p + dictSize; + const BYTE* base; + + DEBUGLOG(4, "LZ4_loadDict (%i bytes from %p into %p)", dictSize, dictionary, LZ4_dict); + + /* It's necessary to reset the context, + * and not just continue it with prepareTable() + * to avoid any risk of generating overflowing matchIndex + * when compressing using this dictionary */ + LZ4_resetStream(LZ4_dict); + + /* We always increment the offset by 64 KB, since, if the dict is longer, + * we truncate it to the last 64k, and if it's shorter, we still want to + * advance by a whole window length so we can provide the guarantee that + * there are only valid offsets in the window, which allows an optimization + * in LZ4_compress_fast_continue() where it uses noDictIssue even when the + * dictionary isn't a full 64k. */ + dict->currentOffset += 64 KB; + + if (dictSize < (int)HASH_UNIT) { + return 0; + } + + if ((dictEnd - p) > 64 KB) p = dictEnd - 64 KB; + base = dictEnd - dict->currentOffset; + dict->dictionary = p; + dict->dictSize = (U32)(dictEnd - p); + dict->tableType = (U32)tableType; + + while (p <= dictEnd-HASH_UNIT) { + LZ4_putPosition(p, dict->hashTable, tableType, base); + p+=3; + } + + return (int)dict->dictSize; +} + +void LZ4_attach_dictionary(LZ4_stream_t* workingStream, const LZ4_stream_t* dictionaryStream) +{ + const LZ4_stream_t_internal* dictCtx = (dictionaryStream == NULL) ? NULL : + &(dictionaryStream->internal_donotuse); + + DEBUGLOG(4, "LZ4_attach_dictionary (%p, %p, size %u)", + workingStream, dictionaryStream, + dictCtx != NULL ? dictCtx->dictSize : 0); + + if (dictCtx != NULL) { + /* If the current offset is zero, we will never look in the + * external dictionary context, since there is no value a table + * entry can take that indicate a miss. In that case, we need + * to bump the offset to something non-zero. + */ + if (workingStream->internal_donotuse.currentOffset == 0) { + workingStream->internal_donotuse.currentOffset = 64 KB; + } + + /* Don't actually attach an empty dictionary. + */ + if (dictCtx->dictSize == 0) { + dictCtx = NULL; + } + } + workingStream->internal_donotuse.dictCtx = dictCtx; +} + + +static void LZ4_renormDictT(LZ4_stream_t_internal* LZ4_dict, int nextSize) +{ + assert(nextSize >= 0); + if (LZ4_dict->currentOffset + (unsigned)nextSize > 0x80000000) { /* potential ptrdiff_t overflow (32-bits mode) */ + /* rescale hash table */ + U32 const delta = LZ4_dict->currentOffset - 64 KB; + const BYTE* dictEnd = LZ4_dict->dictionary + LZ4_dict->dictSize; + int i; + DEBUGLOG(4, "LZ4_renormDictT"); + for (i=0; ihashTable[i] < delta) LZ4_dict->hashTable[i]=0; + else LZ4_dict->hashTable[i] -= delta; + } + LZ4_dict->currentOffset = 64 KB; + if (LZ4_dict->dictSize > 64 KB) LZ4_dict->dictSize = 64 KB; + LZ4_dict->dictionary = dictEnd - LZ4_dict->dictSize; + } +} + + +int LZ4_compress_fast_continue (LZ4_stream_t* LZ4_stream, + const char* source, char* dest, + int inputSize, int maxOutputSize, + int acceleration) +{ + const tableType_t tableType = byU32; + LZ4_stream_t_internal* const streamPtr = &LZ4_stream->internal_donotuse; + const char* dictEnd = streamPtr->dictSize ? (const char*)streamPtr->dictionary + streamPtr->dictSize : NULL; + + DEBUGLOG(5, "LZ4_compress_fast_continue (inputSize=%i, dictSize=%u)", inputSize, streamPtr->dictSize); + + LZ4_renormDictT(streamPtr, inputSize); /* fix index overflow */ + if (acceleration < 1) acceleration = LZ4_ACCELERATION_DEFAULT; + if (acceleration > LZ4_ACCELERATION_MAX) acceleration = LZ4_ACCELERATION_MAX; + + /* invalidate tiny dictionaries */ + if ( (streamPtr->dictSize < 4) /* tiny dictionary : not enough for a hash */ + && (dictEnd != source) /* prefix mode */ + && (inputSize > 0) /* tolerance : don't lose history, in case next invocation would use prefix mode */ + && (streamPtr->dictCtx == NULL) /* usingDictCtx */ + ) { + DEBUGLOG(5, "LZ4_compress_fast_continue: dictSize(%u) at addr:%p is too small", streamPtr->dictSize, streamPtr->dictionary); + /* remove dictionary existence from history, to employ faster prefix mode */ + streamPtr->dictSize = 0; + streamPtr->dictionary = (const BYTE*)source; + dictEnd = source; + } + + /* Check overlapping input/dictionary space */ + { const char* const sourceEnd = source + inputSize; + if ((sourceEnd > (const char*)streamPtr->dictionary) && (sourceEnd < dictEnd)) { + streamPtr->dictSize = (U32)(dictEnd - sourceEnd); + if (streamPtr->dictSize > 64 KB) streamPtr->dictSize = 64 KB; + if (streamPtr->dictSize < 4) streamPtr->dictSize = 0; + streamPtr->dictionary = (const BYTE*)dictEnd - streamPtr->dictSize; + } + } + + /* prefix mode : source data follows dictionary */ + if (dictEnd == source) { + if ((streamPtr->dictSize < 64 KB) && (streamPtr->dictSize < streamPtr->currentOffset)) + return LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, withPrefix64k, dictSmall, acceleration); + else + return LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, withPrefix64k, noDictIssue, acceleration); + } + + /* external dictionary mode */ + { int result; + if (streamPtr->dictCtx) { + /* We depend here on the fact that dictCtx'es (produced by + * LZ4_loadDict) guarantee that their tables contain no references + * to offsets between dictCtx->currentOffset - 64 KB and + * dictCtx->currentOffset - dictCtx->dictSize. This makes it safe + * to use noDictIssue even when the dict isn't a full 64 KB. + */ + if (inputSize > 4 KB) { + /* For compressing large blobs, it is faster to pay the setup + * cost to copy the dictionary's tables into the active context, + * so that the compression loop is only looking into one table. + */ + LZ4_memcpy(streamPtr, streamPtr->dictCtx, sizeof(*streamPtr)); + result = LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, usingExtDict, noDictIssue, acceleration); + } else { + result = LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, usingDictCtx, noDictIssue, acceleration); + } + } else { /* small data <= 4 KB */ + if ((streamPtr->dictSize < 64 KB) && (streamPtr->dictSize < streamPtr->currentOffset)) { + result = LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, usingExtDict, dictSmall, acceleration); + } else { + result = LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, usingExtDict, noDictIssue, acceleration); + } + } + streamPtr->dictionary = (const BYTE*)source; + streamPtr->dictSize = (U32)inputSize; + return result; + } +} + + +/* Hidden debug function, to force-test external dictionary mode */ +int LZ4_compress_forceExtDict (LZ4_stream_t* LZ4_dict, const char* source, char* dest, int srcSize) +{ + LZ4_stream_t_internal* streamPtr = &LZ4_dict->internal_donotuse; + int result; + + LZ4_renormDictT(streamPtr, srcSize); + + if ((streamPtr->dictSize < 64 KB) && (streamPtr->dictSize < streamPtr->currentOffset)) { + result = LZ4_compress_generic(streamPtr, source, dest, srcSize, NULL, 0, notLimited, byU32, usingExtDict, dictSmall, 1); + } else { + result = LZ4_compress_generic(streamPtr, source, dest, srcSize, NULL, 0, notLimited, byU32, usingExtDict, noDictIssue, 1); + } + + streamPtr->dictionary = (const BYTE*)source; + streamPtr->dictSize = (U32)srcSize; + + return result; +} + + +/*! LZ4_saveDict() : + * If previously compressed data block is not guaranteed to remain available at its memory location, + * save it into a safer place (char* safeBuffer). + * Note : no need to call LZ4_loadDict() afterwards, dictionary is immediately usable, + * one can therefore call LZ4_compress_fast_continue() right after. + * @return : saved dictionary size in bytes (necessarily <= dictSize), or 0 if error. + */ +int LZ4_saveDict (LZ4_stream_t* LZ4_dict, char* safeBuffer, int dictSize) +{ + LZ4_stream_t_internal* const dict = &LZ4_dict->internal_donotuse; + + DEBUGLOG(5, "LZ4_saveDict : dictSize=%i, safeBuffer=%p", dictSize, safeBuffer); + + if ((U32)dictSize > 64 KB) { dictSize = 64 KB; } /* useless to define a dictionary > 64 KB */ + if ((U32)dictSize > dict->dictSize) { dictSize = (int)dict->dictSize; } + + if (safeBuffer == NULL) assert(dictSize == 0); + if (dictSize > 0) { + const BYTE* const previousDictEnd = dict->dictionary + dict->dictSize; + assert(dict->dictionary); + LZ4_memmove(safeBuffer, previousDictEnd - dictSize, (size_t)dictSize); + } + + dict->dictionary = (const BYTE*)safeBuffer; + dict->dictSize = (U32)dictSize; + + return dictSize; +} + + + +/*-******************************* + * Decompression functions + ********************************/ + +typedef enum { decode_full_block = 0, partial_decode = 1 } earlyEnd_directive; + +#undef MIN +#define MIN(a,b) ( (a) < (b) ? (a) : (b) ) + + +/* variant for decompress_unsafe() + * does not know end of input + * presumes input is well formed + * note : will consume at least one byte */ +size_t read_long_length_no_check(const BYTE** pp) +{ + size_t b, l = 0; + do { b = **pp; (*pp)++; l += b; } while (b==255); + DEBUGLOG(6, "read_long_length_no_check: +length=%zu using %zu input bytes", l, l/255 + 1) + return l; +} + +/* core decoder variant for LZ4_decompress_fast*() + * for legacy support only : these entry points are deprecated. + * - Presumes input is correctly formed (no defense vs malformed inputs) + * - Does not know input size (presume input buffer is "large enough") + * - Decompress a full block (only) + * @return : nb of bytes read from input. + * Note : this variant is not optimized for speed, just for maintenance. + * the goal is to remove support of decompress_fast*() variants by v2.0 +**/ +LZ4_FORCE_INLINE int +LZ4_decompress_unsafe_generic( + const BYTE* const istart, + BYTE* const ostart, + int decompressedSize, + + size_t prefixSize, + const BYTE* const dictStart, /* only if dict==usingExtDict */ + const size_t dictSize /* note: =0 if dictStart==NULL */ + ) +{ + const BYTE* ip = istart; + BYTE* op = (BYTE*)ostart; + BYTE* const oend = ostart + decompressedSize; + const BYTE* const prefixStart = ostart - prefixSize; + + DEBUGLOG(5, "LZ4_decompress_unsafe_generic"); + if (dictStart == NULL) assert(dictSize == 0); + + while (1) { + /* start new sequence */ + unsigned token = *ip++; + + /* literals */ + { size_t ll = token >> ML_BITS; + if (ll==15) { + /* long literal length */ + ll += read_long_length_no_check(&ip); + } + if ((size_t)(oend-op) < ll) return -1; /* output buffer overflow */ + LZ4_memmove(op, ip, ll); /* support in-place decompression */ + op += ll; + ip += ll; + if ((size_t)(oend-op) < MFLIMIT) { + if (op==oend) break; /* end of block */ + DEBUGLOG(5, "invalid: literals end at distance %zi from end of block", oend-op); + /* incorrect end of block : + * last match must start at least MFLIMIT==12 bytes before end of output block */ + return -1; + } } + + /* match */ + { size_t ml = token & 15; + size_t const offset = LZ4_readLE16(ip); + ip+=2; + + if (ml==15) { + /* long literal length */ + ml += read_long_length_no_check(&ip); + } + ml += MINMATCH; + + if ((size_t)(oend-op) < ml) return -1; /* output buffer overflow */ + + { const BYTE* match = op - offset; + + /* out of range */ + if (offset > (size_t)(op - prefixStart) + dictSize) { + DEBUGLOG(6, "offset out of range"); + return -1; + } + + /* check special case : extDict */ + if (offset > (size_t)(op - prefixStart)) { + /* extDict scenario */ + const BYTE* const dictEnd = dictStart + dictSize; + const BYTE* extMatch = dictEnd - (offset - (size_t)(op-prefixStart)); + size_t const extml = (size_t)(dictEnd - extMatch); + if (extml > ml) { + /* match entirely within extDict */ + LZ4_memmove(op, extMatch, ml); + op += ml; + ml = 0; + } else { + /* match split between extDict & prefix */ + LZ4_memmove(op, extMatch, extml); + op += extml; + ml -= extml; + } + match = prefixStart; + } + + /* match copy - slow variant, supporting overlap copy */ + { size_t u; + for (u=0; u= ipmax before start of loop. Returns initial_error if so. + * @error (output) - error code. Must be set to 0 before call. +**/ +typedef size_t Rvl_t; +static const Rvl_t rvl_error = (Rvl_t)(-1); +LZ4_FORCE_INLINE Rvl_t +read_variable_length(const BYTE** ip, const BYTE* ilimit, + int initial_check) +{ + Rvl_t s, length = 0; + assert(ip != NULL); + assert(*ip != NULL); + assert(ilimit != NULL); + if (initial_check && unlikely((*ip) >= ilimit)) { /* read limit reached */ + return rvl_error; + } + do { + s = **ip; + (*ip)++; + length += s; + if (unlikely((*ip) > ilimit)) { /* read limit reached */ + return rvl_error; + } + /* accumulator overflow detection (32-bit mode only) */ + if ((sizeof(length)<8) && unlikely(length > ((Rvl_t)(-1)/2)) ) { + return rvl_error; + } + } while (s==255); + + return length; +} + +/*! LZ4_decompress_generic() : + * This generic decompression function covers all use cases. + * It shall be instantiated several times, using different sets of directives. + * Note that it is important for performance that this function really get inlined, + * in order to remove useless branches during compilation optimization. + */ +LZ4_FORCE_INLINE int +LZ4_decompress_generic( + const char* const src, + char* const dst, + int srcSize, + int outputSize, /* If endOnInput==endOnInputSize, this value is `dstCapacity` */ + + earlyEnd_directive partialDecoding, /* full, partial */ + dict_directive dict, /* noDict, withPrefix64k, usingExtDict */ + const BYTE* const lowPrefix, /* always <= dst, == dst when no prefix */ + const BYTE* const dictStart, /* only if dict==usingExtDict */ + const size_t dictSize /* note : = 0 if noDict */ + ) +{ + if ((src == NULL) || (outputSize < 0)) { return -1; } + + { const BYTE* ip = (const BYTE*) src; + const BYTE* const iend = ip + srcSize; + + BYTE* op = (BYTE*) dst; + BYTE* const oend = op + outputSize; + BYTE* cpy; + + const BYTE* const dictEnd = (dictStart == NULL) ? NULL : dictStart + dictSize; + + const int checkOffset = (dictSize < (int)(64 KB)); + + + /* Set up the "end" pointers for the shortcut. */ + const BYTE* const shortiend = iend - 14 /*maxLL*/ - 2 /*offset*/; + const BYTE* const shortoend = oend - 14 /*maxLL*/ - 18 /*maxML*/; + + const BYTE* match; + size_t offset; + unsigned token; + size_t length; + + + DEBUGLOG(5, "LZ4_decompress_generic (srcSize:%i, dstSize:%i)", srcSize, outputSize); + + /* Special cases */ + assert(lowPrefix <= op); + if (unlikely(outputSize==0)) { + /* Empty output buffer */ + if (partialDecoding) return 0; + return ((srcSize==1) && (*ip==0)) ? 0 : -1; + } + if (unlikely(srcSize==0)) { return -1; } + + /* LZ4_FAST_DEC_LOOP: + * designed for modern OoO performance cpus, + * where copying reliably 32-bytes is preferable to an unpredictable branch. + * note : fast loop may show a regression for some client arm chips. */ +#if LZ4_FAST_DEC_LOOP + if ((oend - op) < FASTLOOP_SAFE_DISTANCE) { + DEBUGLOG(6, "skip fast decode loop"); + goto safe_decode; + } + + /* Fast loop : decode sequences as long as output < oend-FASTLOOP_SAFE_DISTANCE */ + while (1) { + /* Main fastloop assertion: We can always wildcopy FASTLOOP_SAFE_DISTANCE */ + assert(oend - op >= FASTLOOP_SAFE_DISTANCE); + assert(ip < iend); + token = *ip++; + length = token >> ML_BITS; /* literal length */ + + /* decode literal length */ + if (length == RUN_MASK) { + size_t const addl = read_variable_length(&ip, iend-RUN_MASK, 1); + if (addl == rvl_error) { goto _output_error; } + length += addl; + if (unlikely((uptrval)(op)+length<(uptrval)(op))) { goto _output_error; } /* overflow detection */ + if (unlikely((uptrval)(ip)+length<(uptrval)(ip))) { goto _output_error; } /* overflow detection */ + + /* copy literals */ + cpy = op+length; + LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH); + if ((cpy>oend-32) || (ip+length>iend-32)) { goto safe_literal_copy; } + LZ4_wildCopy32(op, ip, cpy); + ip += length; op = cpy; + } else { + cpy = op+length; + DEBUGLOG(7, "copy %u bytes in a 16-bytes stripe", (unsigned)length); + /* We don't need to check oend, since we check it once for each loop below */ + if (ip > iend-(16 + 1/*max lit + offset + nextToken*/)) { goto safe_literal_copy; } + /* Literals can only be <= 14, but hope compilers optimize better when copy by a register size */ + LZ4_memcpy(op, ip, 16); + ip += length; op = cpy; + } + + /* get offset */ + offset = LZ4_readLE16(ip); ip+=2; + match = op - offset; + assert(match <= op); /* overflow check */ + + /* get matchlength */ + length = token & ML_MASK; + + if (length == ML_MASK) { + size_t const addl = read_variable_length(&ip, iend - LASTLITERALS + 1, 0); + if (addl == rvl_error) { goto _output_error; } + length += addl; + length += MINMATCH; + if (unlikely((uptrval)(op)+length<(uptrval)op)) { goto _output_error; } /* overflow detection */ + if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) { goto _output_error; } /* Error : offset outside buffers */ + if (op + length >= oend - FASTLOOP_SAFE_DISTANCE) { + goto safe_match_copy; + } + } else { + length += MINMATCH; + if (op + length >= oend - FASTLOOP_SAFE_DISTANCE) { + goto safe_match_copy; + } + + /* Fastpath check: skip LZ4_wildCopy32 when true */ + if ((dict == withPrefix64k) || (match >= lowPrefix)) { + if (offset >= 8) { + assert(match >= lowPrefix); + assert(match <= op); + assert(op + 18 <= oend); + + LZ4_memcpy(op, match, 8); + LZ4_memcpy(op+8, match+8, 8); + LZ4_memcpy(op+16, match+16, 2); + op += length; + continue; + } } } + + if (checkOffset && (unlikely(match + dictSize < lowPrefix))) { goto _output_error; } /* Error : offset outside buffers */ + /* match starting within external dictionary */ + if ((dict==usingExtDict) && (match < lowPrefix)) { + assert(dictEnd != NULL); + if (unlikely(op+length > oend-LASTLITERALS)) { + if (partialDecoding) { + DEBUGLOG(7, "partialDecoding: dictionary match, close to dstEnd"); + length = MIN(length, (size_t)(oend-op)); + } else { + goto _output_error; /* end-of-block condition violated */ + } } + + if (length <= (size_t)(lowPrefix-match)) { + /* match fits entirely within external dictionary : just copy */ + LZ4_memmove(op, dictEnd - (lowPrefix-match), length); + op += length; + } else { + /* match stretches into both external dictionary and current block */ + size_t const copySize = (size_t)(lowPrefix - match); + size_t const restSize = length - copySize; + LZ4_memcpy(op, dictEnd - copySize, copySize); + op += copySize; + if (restSize > (size_t)(op - lowPrefix)) { /* overlap copy */ + BYTE* const endOfMatch = op + restSize; + const BYTE* copyFrom = lowPrefix; + while (op < endOfMatch) { *op++ = *copyFrom++; } + } else { + LZ4_memcpy(op, lowPrefix, restSize); + op += restSize; + } } + continue; + } + + /* copy match within block */ + cpy = op + length; + + assert((op <= oend) && (oend-op >= 32)); + if (unlikely(offset<16)) { + LZ4_memcpy_using_offset(op, match, cpy, offset); + } else { + LZ4_wildCopy32(op, match, cpy); + } + + op = cpy; /* wildcopy correction */ + } + safe_decode: +#endif + + /* Main Loop : decode remaining sequences where output < FASTLOOP_SAFE_DISTANCE */ + while (1) { + assert(ip < iend); + token = *ip++; + length = token >> ML_BITS; /* literal length */ + + /* A two-stage shortcut for the most common case: + * 1) If the literal length is 0..14, and there is enough space, + * enter the shortcut and copy 16 bytes on behalf of the literals + * (in the fast mode, only 8 bytes can be safely copied this way). + * 2) Further if the match length is 4..18, copy 18 bytes in a similar + * manner; but we ensure that there's enough space in the output for + * those 18 bytes earlier, upon entering the shortcut (in other words, + * there is a combined check for both stages). + */ + if ( (length != RUN_MASK) + /* strictly "less than" on input, to re-enter the loop with at least one byte */ + && likely((ip < shortiend) & (op <= shortoend)) ) { + /* Copy the literals */ + LZ4_memcpy(op, ip, 16); + op += length; ip += length; + + /* The second stage: prepare for match copying, decode full info. + * If it doesn't work out, the info won't be wasted. */ + length = token & ML_MASK; /* match length */ + offset = LZ4_readLE16(ip); ip += 2; + match = op - offset; + assert(match <= op); /* check overflow */ + + /* Do not deal with overlapping matches. */ + if ( (length != ML_MASK) + && (offset >= 8) + && (dict==withPrefix64k || match >= lowPrefix) ) { + /* Copy the match. */ + LZ4_memcpy(op + 0, match + 0, 8); + LZ4_memcpy(op + 8, match + 8, 8); + LZ4_memcpy(op +16, match +16, 2); + op += length + MINMATCH; + /* Both stages worked, load the next token. */ + continue; + } + + /* The second stage didn't work out, but the info is ready. + * Propel it right to the point of match copying. */ + goto _copy_match; + } + + /* decode literal length */ + if (length == RUN_MASK) { + size_t const addl = read_variable_length(&ip, iend-RUN_MASK, 1); + if (addl == rvl_error) { goto _output_error; } + length += addl; + if (unlikely((uptrval)(op)+length<(uptrval)(op))) { goto _output_error; } /* overflow detection */ + if (unlikely((uptrval)(ip)+length<(uptrval)(ip))) { goto _output_error; } /* overflow detection */ + } + + /* copy literals */ + cpy = op+length; +#if LZ4_FAST_DEC_LOOP + safe_literal_copy: +#endif + LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH); + if ((cpy>oend-MFLIMIT) || (ip+length>iend-(2+1+LASTLITERALS))) { + /* We've either hit the input parsing restriction or the output parsing restriction. + * In the normal scenario, decoding a full block, it must be the last sequence, + * otherwise it's an error (invalid input or dimensions). + * In partialDecoding scenario, it's necessary to ensure there is no buffer overflow. + */ + if (partialDecoding) { + /* Since we are partial decoding we may be in this block because of the output parsing + * restriction, which is not valid since the output buffer is allowed to be undersized. + */ + DEBUGLOG(7, "partialDecoding: copying literals, close to input or output end") + DEBUGLOG(7, "partialDecoding: literal length = %u", (unsigned)length); + DEBUGLOG(7, "partialDecoding: remaining space in dstBuffer : %i", (int)(oend - op)); + DEBUGLOG(7, "partialDecoding: remaining space in srcBuffer : %i", (int)(iend - ip)); + /* Finishing in the middle of a literals segment, + * due to lack of input. + */ + if (ip+length > iend) { + length = (size_t)(iend-ip); + cpy = op + length; + } + /* Finishing in the middle of a literals segment, + * due to lack of output space. + */ + if (cpy > oend) { + cpy = oend; + assert(op<=oend); + length = (size_t)(oend-op); + } + } else { + /* We must be on the last sequence (or invalid) because of the parsing limitations + * so check that we exactly consume the input and don't overrun the output buffer. + */ + if ((ip+length != iend) || (cpy > oend)) { + DEBUGLOG(6, "should have been last run of literals") + DEBUGLOG(6, "ip(%p) + length(%i) = %p != iend (%p)", ip, (int)length, ip+length, iend); + DEBUGLOG(6, "or cpy(%p) > oend(%p)", cpy, oend); + goto _output_error; + } + } + LZ4_memmove(op, ip, length); /* supports overlapping memory regions, for in-place decompression scenarios */ + ip += length; + op += length; + /* Necessarily EOF when !partialDecoding. + * When partialDecoding, it is EOF if we've either + * filled the output buffer or + * can't proceed with reading an offset for following match. + */ + if (!partialDecoding || (cpy == oend) || (ip >= (iend-2))) { + break; + } + } else { + LZ4_wildCopy8(op, ip, cpy); /* can overwrite up to 8 bytes beyond cpy */ + ip += length; op = cpy; + } + + /* get offset */ + offset = LZ4_readLE16(ip); ip+=2; + match = op - offset; + + /* get matchlength */ + length = token & ML_MASK; + + _copy_match: + if (length == ML_MASK) { + size_t const addl = read_variable_length(&ip, iend - LASTLITERALS + 1, 0); + if (addl == rvl_error) { goto _output_error; } + length += addl; + if (unlikely((uptrval)(op)+length<(uptrval)op)) goto _output_error; /* overflow detection */ + } + length += MINMATCH; + +#if LZ4_FAST_DEC_LOOP + safe_match_copy: +#endif + if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) goto _output_error; /* Error : offset outside buffers */ + /* match starting within external dictionary */ + if ((dict==usingExtDict) && (match < lowPrefix)) { + assert(dictEnd != NULL); + if (unlikely(op+length > oend-LASTLITERALS)) { + if (partialDecoding) length = MIN(length, (size_t)(oend-op)); + else goto _output_error; /* doesn't respect parsing restriction */ + } + + if (length <= (size_t)(lowPrefix-match)) { + /* match fits entirely within external dictionary : just copy */ + LZ4_memmove(op, dictEnd - (lowPrefix-match), length); + op += length; + } else { + /* match stretches into both external dictionary and current block */ + size_t const copySize = (size_t)(lowPrefix - match); + size_t const restSize = length - copySize; + LZ4_memcpy(op, dictEnd - copySize, copySize); + op += copySize; + if (restSize > (size_t)(op - lowPrefix)) { /* overlap copy */ + BYTE* const endOfMatch = op + restSize; + const BYTE* copyFrom = lowPrefix; + while (op < endOfMatch) *op++ = *copyFrom++; + } else { + LZ4_memcpy(op, lowPrefix, restSize); + op += restSize; + } } + continue; + } + assert(match >= lowPrefix); + + /* copy match within block */ + cpy = op + length; + + /* partialDecoding : may end anywhere within the block */ + assert(op<=oend); + if (partialDecoding && (cpy > oend-MATCH_SAFEGUARD_DISTANCE)) { + size_t const mlen = MIN(length, (size_t)(oend-op)); + const BYTE* const matchEnd = match + mlen; + BYTE* const copyEnd = op + mlen; + if (matchEnd > op) { /* overlap copy */ + while (op < copyEnd) { *op++ = *match++; } + } else { + LZ4_memcpy(op, match, mlen); + } + op = copyEnd; + if (op == oend) { break; } + continue; + } + + if (unlikely(offset<8)) { + LZ4_write32(op, 0); /* silence msan warning when offset==0 */ + op[0] = match[0]; + op[1] = match[1]; + op[2] = match[2]; + op[3] = match[3]; + match += inc32table[offset]; + LZ4_memcpy(op+4, match, 4); + match -= dec64table[offset]; + } else { + LZ4_memcpy(op, match, 8); + match += 8; + } + op += 8; + + if (unlikely(cpy > oend-MATCH_SAFEGUARD_DISTANCE)) { + BYTE* const oCopyLimit = oend - (WILDCOPYLENGTH-1); + if (cpy > oend-LASTLITERALS) { goto _output_error; } /* Error : last LASTLITERALS bytes must be literals (uncompressed) */ + if (op < oCopyLimit) { + LZ4_wildCopy8(op, match, oCopyLimit); + match += oCopyLimit - op; + op = oCopyLimit; + } + while (op < cpy) { *op++ = *match++; } + } else { + LZ4_memcpy(op, match, 8); + if (length > 16) { LZ4_wildCopy8(op+8, match+8, cpy); } + } + op = cpy; /* wildcopy correction */ + } + + /* end of decoding */ + DEBUGLOG(5, "decoded %i bytes", (int) (((char*)op)-dst)); + return (int) (((char*)op)-dst); /* Nb of output bytes decoded */ + + /* Overflow error detected */ + _output_error: + return (int) (-(((const char*)ip)-src))-1; + } +} + + +/*===== Instantiate the API decoding functions. =====*/ + +LZ4_FORCE_O2 +int LZ4_decompress_safe(const char* source, char* dest, int compressedSize, int maxDecompressedSize) +{ + return LZ4_decompress_generic(source, dest, compressedSize, maxDecompressedSize, + decode_full_block, noDict, + (BYTE*)dest, NULL, 0); +} + +LZ4_FORCE_O2 +int LZ4_decompress_safe_partial(const char* src, char* dst, int compressedSize, int targetOutputSize, int dstCapacity) +{ + dstCapacity = MIN(targetOutputSize, dstCapacity); + return LZ4_decompress_generic(src, dst, compressedSize, dstCapacity, + partial_decode, + noDict, (BYTE*)dst, NULL, 0); +} + +LZ4_FORCE_O2 +int LZ4_decompress_fast(const char* source, char* dest, int originalSize) +{ + DEBUGLOG(5, "LZ4_decompress_fast"); + return LZ4_decompress_unsafe_generic( + (const BYTE*)source, (BYTE*)dest, originalSize, + 0, NULL, 0); +} + +/*===== Instantiate a few more decoding cases, used more than once. =====*/ + +LZ4_FORCE_O2 /* Exported, an obsolete API function. */ +int LZ4_decompress_safe_withPrefix64k(const char* source, char* dest, int compressedSize, int maxOutputSize) +{ + return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, + decode_full_block, withPrefix64k, + (BYTE*)dest - 64 KB, NULL, 0); +} + +LZ4_FORCE_O2 +static int LZ4_decompress_safe_partial_withPrefix64k(const char* source, char* dest, int compressedSize, int targetOutputSize, int dstCapacity) +{ + dstCapacity = MIN(targetOutputSize, dstCapacity); + return LZ4_decompress_generic(source, dest, compressedSize, dstCapacity, + partial_decode, withPrefix64k, + (BYTE*)dest - 64 KB, NULL, 0); +} + +/* Another obsolete API function, paired with the previous one. */ +int LZ4_decompress_fast_withPrefix64k(const char* source, char* dest, int originalSize) +{ + return LZ4_decompress_unsafe_generic( + (const BYTE*)source, (BYTE*)dest, originalSize, + 64 KB, NULL, 0); +} + +LZ4_FORCE_O2 +static int LZ4_decompress_safe_withSmallPrefix(const char* source, char* dest, int compressedSize, int maxOutputSize, + size_t prefixSize) +{ + return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, + decode_full_block, noDict, + (BYTE*)dest-prefixSize, NULL, 0); +} + +LZ4_FORCE_O2 +static int LZ4_decompress_safe_partial_withSmallPrefix(const char* source, char* dest, int compressedSize, int targetOutputSize, int dstCapacity, + size_t prefixSize) +{ + dstCapacity = MIN(targetOutputSize, dstCapacity); + return LZ4_decompress_generic(source, dest, compressedSize, dstCapacity, + partial_decode, noDict, + (BYTE*)dest-prefixSize, NULL, 0); +} + +LZ4_FORCE_O2 +int LZ4_decompress_safe_forceExtDict(const char* source, char* dest, + int compressedSize, int maxOutputSize, + const void* dictStart, size_t dictSize) +{ + return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, + decode_full_block, usingExtDict, + (BYTE*)dest, (const BYTE*)dictStart, dictSize); +} + +LZ4_FORCE_O2 +int LZ4_decompress_safe_partial_forceExtDict(const char* source, char* dest, + int compressedSize, int targetOutputSize, int dstCapacity, + const void* dictStart, size_t dictSize) +{ + dstCapacity = MIN(targetOutputSize, dstCapacity); + return LZ4_decompress_generic(source, dest, compressedSize, dstCapacity, + partial_decode, usingExtDict, + (BYTE*)dest, (const BYTE*)dictStart, dictSize); +} + +LZ4_FORCE_O2 +static int LZ4_decompress_fast_extDict(const char* source, char* dest, int originalSize, + const void* dictStart, size_t dictSize) +{ + return LZ4_decompress_unsafe_generic( + (const BYTE*)source, (BYTE*)dest, originalSize, + 0, (const BYTE*)dictStart, dictSize); +} + +/* The "double dictionary" mode, for use with e.g. ring buffers: the first part + * of the dictionary is passed as prefix, and the second via dictStart + dictSize. + * These routines are used only once, in LZ4_decompress_*_continue(). + */ +LZ4_FORCE_INLINE +int LZ4_decompress_safe_doubleDict(const char* source, char* dest, int compressedSize, int maxOutputSize, + size_t prefixSize, const void* dictStart, size_t dictSize) +{ + return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, + decode_full_block, usingExtDict, + (BYTE*)dest-prefixSize, (const BYTE*)dictStart, dictSize); +} + +/*===== streaming decompression functions =====*/ + +#if !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) +LZ4_streamDecode_t* LZ4_createStreamDecode(void) +{ + LZ4_STATIC_ASSERT(sizeof(LZ4_streamDecode_t) >= sizeof(LZ4_streamDecode_t_internal)); + return (LZ4_streamDecode_t*) ALLOC_AND_ZERO(sizeof(LZ4_streamDecode_t)); +} + +int LZ4_freeStreamDecode (LZ4_streamDecode_t* LZ4_stream) +{ + if (LZ4_stream == NULL) { return 0; } /* support free on NULL */ + FREEMEM(LZ4_stream); + return 0; +} +#endif + +/*! LZ4_setStreamDecode() : + * Use this function to instruct where to find the dictionary. + * This function is not necessary if previous data is still available where it was decoded. + * Loading a size of 0 is allowed (same effect as no dictionary). + * @return : 1 if OK, 0 if error + */ +int LZ4_setStreamDecode (LZ4_streamDecode_t* LZ4_streamDecode, const char* dictionary, int dictSize) +{ + LZ4_streamDecode_t_internal* lz4sd = &LZ4_streamDecode->internal_donotuse; + lz4sd->prefixSize = (size_t)dictSize; + if (dictSize) { + assert(dictionary != NULL); + lz4sd->prefixEnd = (const BYTE*) dictionary + dictSize; + } else { + lz4sd->prefixEnd = (const BYTE*) dictionary; + } + lz4sd->externalDict = NULL; + lz4sd->extDictSize = 0; + return 1; +} + +/*! LZ4_decoderRingBufferSize() : + * when setting a ring buffer for streaming decompression (optional scenario), + * provides the minimum size of this ring buffer + * to be compatible with any source respecting maxBlockSize condition. + * Note : in a ring buffer scenario, + * blocks are presumed decompressed next to each other. + * When not enough space remains for next block (remainingSize < maxBlockSize), + * decoding resumes from beginning of ring buffer. + * @return : minimum ring buffer size, + * or 0 if there is an error (invalid maxBlockSize). + */ +int LZ4_decoderRingBufferSize(int maxBlockSize) +{ + if (maxBlockSize < 0) return 0; + if (maxBlockSize > LZ4_MAX_INPUT_SIZE) return 0; + if (maxBlockSize < 16) maxBlockSize = 16; + return LZ4_DECODER_RING_BUFFER_SIZE(maxBlockSize); +} + +/* +*_continue() : + These decoding functions allow decompression of multiple blocks in "streaming" mode. + Previously decoded blocks must still be available at the memory position where they were decoded. + If it's not possible, save the relevant part of decoded data into a safe buffer, + and indicate where it stands using LZ4_setStreamDecode() +*/ +LZ4_FORCE_O2 +int LZ4_decompress_safe_continue (LZ4_streamDecode_t* LZ4_streamDecode, const char* source, char* dest, int compressedSize, int maxOutputSize) +{ + LZ4_streamDecode_t_internal* lz4sd = &LZ4_streamDecode->internal_donotuse; + int result; + + if (lz4sd->prefixSize == 0) { + /* The first call, no dictionary yet. */ + assert(lz4sd->extDictSize == 0); + result = LZ4_decompress_safe(source, dest, compressedSize, maxOutputSize); + if (result <= 0) return result; + lz4sd->prefixSize = (size_t)result; + lz4sd->prefixEnd = (BYTE*)dest + result; + } else if (lz4sd->prefixEnd == (BYTE*)dest) { + /* They're rolling the current segment. */ + if (lz4sd->prefixSize >= 64 KB - 1) + result = LZ4_decompress_safe_withPrefix64k(source, dest, compressedSize, maxOutputSize); + else if (lz4sd->extDictSize == 0) + result = LZ4_decompress_safe_withSmallPrefix(source, dest, compressedSize, maxOutputSize, + lz4sd->prefixSize); + else + result = LZ4_decompress_safe_doubleDict(source, dest, compressedSize, maxOutputSize, + lz4sd->prefixSize, lz4sd->externalDict, lz4sd->extDictSize); + if (result <= 0) return result; + lz4sd->prefixSize += (size_t)result; + lz4sd->prefixEnd += result; + } else { + /* The buffer wraps around, or they're switching to another buffer. */ + lz4sd->extDictSize = lz4sd->prefixSize; + lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize; + result = LZ4_decompress_safe_forceExtDict(source, dest, compressedSize, maxOutputSize, + lz4sd->externalDict, lz4sd->extDictSize); + if (result <= 0) return result; + lz4sd->prefixSize = (size_t)result; + lz4sd->prefixEnd = (BYTE*)dest + result; + } + + return result; +} + +LZ4_FORCE_O2 int +LZ4_decompress_fast_continue (LZ4_streamDecode_t* LZ4_streamDecode, + const char* source, char* dest, int originalSize) +{ + LZ4_streamDecode_t_internal* const lz4sd = + (assert(LZ4_streamDecode!=NULL), &LZ4_streamDecode->internal_donotuse); + int result; + + DEBUGLOG(5, "LZ4_decompress_fast_continue (toDecodeSize=%i)", originalSize); + assert(originalSize >= 0); + + if (lz4sd->prefixSize == 0) { + DEBUGLOG(5, "first invocation : no prefix nor extDict"); + assert(lz4sd->extDictSize == 0); + result = LZ4_decompress_fast(source, dest, originalSize); + if (result <= 0) return result; + lz4sd->prefixSize = (size_t)originalSize; + lz4sd->prefixEnd = (BYTE*)dest + originalSize; + } else if (lz4sd->prefixEnd == (BYTE*)dest) { + DEBUGLOG(5, "continue using existing prefix"); + result = LZ4_decompress_unsafe_generic( + (const BYTE*)source, (BYTE*)dest, originalSize, + lz4sd->prefixSize, + lz4sd->externalDict, lz4sd->extDictSize); + if (result <= 0) return result; + lz4sd->prefixSize += (size_t)originalSize; + lz4sd->prefixEnd += originalSize; + } else { + DEBUGLOG(5, "prefix becomes extDict"); + lz4sd->extDictSize = lz4sd->prefixSize; + lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize; + result = LZ4_decompress_fast_extDict(source, dest, originalSize, + lz4sd->externalDict, lz4sd->extDictSize); + if (result <= 0) return result; + lz4sd->prefixSize = (size_t)originalSize; + lz4sd->prefixEnd = (BYTE*)dest + originalSize; + } + + return result; +} + + +/* +Advanced decoding functions : +*_usingDict() : + These decoding functions work the same as "_continue" ones, + the dictionary must be explicitly provided within parameters +*/ + +int LZ4_decompress_safe_usingDict(const char* source, char* dest, int compressedSize, int maxOutputSize, const char* dictStart, int dictSize) +{ + if (dictSize==0) + return LZ4_decompress_safe(source, dest, compressedSize, maxOutputSize); + if (dictStart+dictSize == dest) { + if (dictSize >= 64 KB - 1) { + return LZ4_decompress_safe_withPrefix64k(source, dest, compressedSize, maxOutputSize); + } + assert(dictSize >= 0); + return LZ4_decompress_safe_withSmallPrefix(source, dest, compressedSize, maxOutputSize, (size_t)dictSize); + } + assert(dictSize >= 0); + return LZ4_decompress_safe_forceExtDict(source, dest, compressedSize, maxOutputSize, dictStart, (size_t)dictSize); +} + +int LZ4_decompress_safe_partial_usingDict(const char* source, char* dest, int compressedSize, int targetOutputSize, int dstCapacity, const char* dictStart, int dictSize) +{ + if (dictSize==0) + return LZ4_decompress_safe_partial(source, dest, compressedSize, targetOutputSize, dstCapacity); + if (dictStart+dictSize == dest) { + if (dictSize >= 64 KB - 1) { + return LZ4_decompress_safe_partial_withPrefix64k(source, dest, compressedSize, targetOutputSize, dstCapacity); + } + assert(dictSize >= 0); + return LZ4_decompress_safe_partial_withSmallPrefix(source, dest, compressedSize, targetOutputSize, dstCapacity, (size_t)dictSize); + } + assert(dictSize >= 0); + return LZ4_decompress_safe_partial_forceExtDict(source, dest, compressedSize, targetOutputSize, dstCapacity, dictStart, (size_t)dictSize); +} + +int LZ4_decompress_fast_usingDict(const char* source, char* dest, int originalSize, const char* dictStart, int dictSize) +{ + if (dictSize==0 || dictStart+dictSize == dest) + return LZ4_decompress_unsafe_generic( + (const BYTE*)source, (BYTE*)dest, originalSize, + (size_t)dictSize, NULL, 0); + assert(dictSize >= 0); + return LZ4_decompress_fast_extDict(source, dest, originalSize, dictStart, (size_t)dictSize); +} + + +/*=************************************************* +* Obsolete Functions +***************************************************/ +/* obsolete compression functions */ +int LZ4_compress_limitedOutput(const char* source, char* dest, int inputSize, int maxOutputSize) +{ + return LZ4_compress_default(source, dest, inputSize, maxOutputSize); +} +int LZ4_compress(const char* src, char* dest, int srcSize) +{ + return LZ4_compress_default(src, dest, srcSize, LZ4_compressBound(srcSize)); +} +int LZ4_compress_limitedOutput_withState (void* state, const char* src, char* dst, int srcSize, int dstSize) +{ + return LZ4_compress_fast_extState(state, src, dst, srcSize, dstSize, 1); +} +int LZ4_compress_withState (void* state, const char* src, char* dst, int srcSize) +{ + return LZ4_compress_fast_extState(state, src, dst, srcSize, LZ4_compressBound(srcSize), 1); +} +int LZ4_compress_limitedOutput_continue (LZ4_stream_t* LZ4_stream, const char* src, char* dst, int srcSize, int dstCapacity) +{ + return LZ4_compress_fast_continue(LZ4_stream, src, dst, srcSize, dstCapacity, 1); +} +int LZ4_compress_continue (LZ4_stream_t* LZ4_stream, const char* source, char* dest, int inputSize) +{ + return LZ4_compress_fast_continue(LZ4_stream, source, dest, inputSize, LZ4_compressBound(inputSize), 1); +} + +/* +These decompression functions are deprecated and should no longer be used. +They are only provided here for compatibility with older user programs. +- LZ4_uncompress is totally equivalent to LZ4_decompress_fast +- LZ4_uncompress_unknownOutputSize is totally equivalent to LZ4_decompress_safe +*/ +int LZ4_uncompress (const char* source, char* dest, int outputSize) +{ + return LZ4_decompress_fast(source, dest, outputSize); +} +int LZ4_uncompress_unknownOutputSize (const char* source, char* dest, int isize, int maxOutputSize) +{ + return LZ4_decompress_safe(source, dest, isize, maxOutputSize); +} + +/* Obsolete Streaming functions */ + +int LZ4_sizeofStreamState(void) { return sizeof(LZ4_stream_t); } + +int LZ4_resetStreamState(void* state, char* inputBuffer) +{ + (void)inputBuffer; + LZ4_resetStream((LZ4_stream_t*)state); + return 0; +} + +#if !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) +void* LZ4_create (char* inputBuffer) +{ + (void)inputBuffer; + return LZ4_createStream(); +} +#endif + +char* LZ4_slideInputBuffer (void* state) +{ + /* avoid const char * -> char * conversion warning */ + return (char *)(uptrval)((LZ4_stream_t*)state)->internal_donotuse.dictionary; +} + +#endif /* LZ4_COMMONDEFS_ONLY */ + +} diff --git a/externals/tracy/public/common/tracy_lz4.hpp b/externals/tracy/public/common/tracy_lz4.hpp new file mode 100644 index 000000000..672c2feb2 --- /dev/null +++ b/externals/tracy/public/common/tracy_lz4.hpp @@ -0,0 +1,847 @@ +/* + * LZ4 - Fast LZ compression algorithm + * Header File + * Copyright (C) 2011-2020, Yann Collet. + + BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are + met: + + * Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above + copyright notice, this list of conditions and the following disclaimer + in the documentation and/or other materials provided with the + distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + You can contact the author at : + - LZ4 homepage : http://www.lz4.org + - LZ4 source repository : https://github.com/lz4/lz4 +*/ + +#ifndef TRACY_LZ4_H_2983827168210 +#define TRACY_LZ4_H_2983827168210 + +/* --- Dependency --- */ +#include /* size_t */ +#include + + +/** + Introduction + + LZ4 is lossless compression algorithm, providing compression speed >500 MB/s per core, + scalable with multi-cores CPU. It features an extremely fast decoder, with speed in + multiple GB/s per core, typically reaching RAM speed limits on multi-core systems. + + The LZ4 compression library provides in-memory compression and decompression functions. + It gives full buffer control to user. + Compression can be done in: + - a single step (described as Simple Functions) + - a single step, reusing a context (described in Advanced Functions) + - unbounded multiple steps (described as Streaming compression) + + lz4.h generates and decodes LZ4-compressed blocks (doc/lz4_Block_format.md). + Decompressing such a compressed block requires additional metadata. + Exact metadata depends on exact decompression function. + For the typical case of LZ4_decompress_safe(), + metadata includes block's compressed size, and maximum bound of decompressed size. + Each application is free to encode and pass such metadata in whichever way it wants. + + lz4.h only handle blocks, it can not generate Frames. + + Blocks are different from Frames (doc/lz4_Frame_format.md). + Frames bundle both blocks and metadata in a specified manner. + Embedding metadata is required for compressed data to be self-contained and portable. + Frame format is delivered through a companion API, declared in lz4frame.h. + The `lz4` CLI can only manage frames. +*/ + +/*^*************************************************************** +* Export parameters +*****************************************************************/ +/* +* LZ4_DLL_EXPORT : +* Enable exporting of functions when building a Windows DLL +* LZ4LIB_VISIBILITY : +* Control library symbols visibility. +*/ +#ifndef LZ4LIB_VISIBILITY +# if defined(__GNUC__) && (__GNUC__ >= 4) +# define LZ4LIB_VISIBILITY __attribute__ ((visibility ("default"))) +# else +# define LZ4LIB_VISIBILITY +# endif +#endif +#if defined(LZ4_DLL_EXPORT) && (LZ4_DLL_EXPORT==1) +# define LZ4LIB_API __declspec(dllexport) LZ4LIB_VISIBILITY +#elif defined(LZ4_DLL_IMPORT) && (LZ4_DLL_IMPORT==1) +# define LZ4LIB_API __declspec(dllimport) LZ4LIB_VISIBILITY /* It isn't required but allows to generate better code, saving a function pointer load from the IAT and an indirect jump.*/ +#else +# define LZ4LIB_API LZ4LIB_VISIBILITY +#endif + +/*! LZ4_FREESTANDING : + * When this macro is set to 1, it enables "freestanding mode" that is + * suitable for typical freestanding environment which doesn't support + * standard C library. + * + * - LZ4_FREESTANDING is a compile-time switch. + * - It requires the following macros to be defined: + * LZ4_memcpy, LZ4_memmove, LZ4_memset. + * - It only enables LZ4/HC functions which don't use heap. + * All LZ4F_* functions are not supported. + * - See tests/freestanding.c to check its basic setup. + */ +#if defined(LZ4_FREESTANDING) && (LZ4_FREESTANDING == 1) +# define LZ4_HEAPMODE 0 +# define LZ4HC_HEAPMODE 0 +# define LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION 1 +# if !defined(LZ4_memcpy) +# error "LZ4_FREESTANDING requires macro 'LZ4_memcpy'." +# endif +# if !defined(LZ4_memset) +# error "LZ4_FREESTANDING requires macro 'LZ4_memset'." +# endif +# if !defined(LZ4_memmove) +# error "LZ4_FREESTANDING requires macro 'LZ4_memmove'." +# endif +#elif ! defined(LZ4_FREESTANDING) +# define LZ4_FREESTANDING 0 +#endif + + +/*------ Version ------*/ +#define LZ4_VERSION_MAJOR 1 /* for breaking interface changes */ +#define LZ4_VERSION_MINOR 9 /* for new (non-breaking) interface capabilities */ +#define LZ4_VERSION_RELEASE 4 /* for tweaks, bug-fixes, or development */ + +#define LZ4_VERSION_NUMBER (LZ4_VERSION_MAJOR *100*100 + LZ4_VERSION_MINOR *100 + LZ4_VERSION_RELEASE) + +#define LZ4_LIB_VERSION LZ4_VERSION_MAJOR.LZ4_VERSION_MINOR.LZ4_VERSION_RELEASE +#define LZ4_QUOTE(str) #str +#define LZ4_EXPAND_AND_QUOTE(str) LZ4_QUOTE(str) +#define LZ4_VERSION_STRING LZ4_EXPAND_AND_QUOTE(LZ4_LIB_VERSION) /* requires v1.7.3+ */ + +namespace tracy +{ + +LZ4LIB_API int LZ4_versionNumber (void); /**< library version number; useful to check dll version; requires v1.3.0+ */ +LZ4LIB_API const char* LZ4_versionString (void); /**< library version string; useful to check dll version; requires v1.7.5+ */ + + +/*-************************************ +* Tuning parameter +**************************************/ +#define LZ4_MEMORY_USAGE_MIN 10 +#define LZ4_MEMORY_USAGE_DEFAULT 14 +#define LZ4_MEMORY_USAGE_MAX 20 + +/*! + * LZ4_MEMORY_USAGE : + * Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; ) + * Increasing memory usage improves compression ratio, at the cost of speed. + * Reduced memory usage may improve speed at the cost of ratio, thanks to better cache locality. + * Default value is 14, for 16KB, which nicely fits into Intel x86 L1 cache + */ +#ifndef LZ4_MEMORY_USAGE +# define LZ4_MEMORY_USAGE LZ4_MEMORY_USAGE_DEFAULT +#endif + +#if (LZ4_MEMORY_USAGE < LZ4_MEMORY_USAGE_MIN) +# error "LZ4_MEMORY_USAGE is too small !" +#endif + +#if (LZ4_MEMORY_USAGE > LZ4_MEMORY_USAGE_MAX) +# error "LZ4_MEMORY_USAGE is too large !" +#endif + +/*-************************************ +* Simple Functions +**************************************/ +/*! LZ4_compress_default() : + * Compresses 'srcSize' bytes from buffer 'src' + * into already allocated 'dst' buffer of size 'dstCapacity'. + * Compression is guaranteed to succeed if 'dstCapacity' >= LZ4_compressBound(srcSize). + * It also runs faster, so it's a recommended setting. + * If the function cannot compress 'src' into a more limited 'dst' budget, + * compression stops *immediately*, and the function result is zero. + * In which case, 'dst' content is undefined (invalid). + * srcSize : max supported value is LZ4_MAX_INPUT_SIZE. + * dstCapacity : size of buffer 'dst' (which must be already allocated) + * @return : the number of bytes written into buffer 'dst' (necessarily <= dstCapacity) + * or 0 if compression fails + * Note : This function is protected against buffer overflow scenarios (never writes outside 'dst' buffer, nor read outside 'source' buffer). + */ +LZ4LIB_API int LZ4_compress_default(const char* src, char* dst, int srcSize, int dstCapacity); + +/*! LZ4_decompress_safe() : + * compressedSize : is the exact complete size of the compressed block. + * dstCapacity : is the size of destination buffer (which must be already allocated), presumed an upper bound of decompressed size. + * @return : the number of bytes decompressed into destination buffer (necessarily <= dstCapacity) + * If destination buffer is not large enough, decoding will stop and output an error code (negative value). + * If the source stream is detected malformed, the function will stop decoding and return a negative result. + * Note 1 : This function is protected against malicious data packets : + * it will never writes outside 'dst' buffer, nor read outside 'source' buffer, + * even if the compressed block is maliciously modified to order the decoder to do these actions. + * In such case, the decoder stops immediately, and considers the compressed block malformed. + * Note 2 : compressedSize and dstCapacity must be provided to the function, the compressed block does not contain them. + * The implementation is free to send / store / derive this information in whichever way is most beneficial. + * If there is a need for a different format which bundles together both compressed data and its metadata, consider looking at lz4frame.h instead. + */ +LZ4LIB_API int LZ4_decompress_safe (const char* src, char* dst, int compressedSize, int dstCapacity); + + +/*-************************************ +* Advanced Functions +**************************************/ +#define LZ4_MAX_INPUT_SIZE 0x7E000000 /* 2 113 929 216 bytes */ +#define LZ4_COMPRESSBOUND(isize) ((unsigned)(isize) > (unsigned)LZ4_MAX_INPUT_SIZE ? 0 : (isize) + ((isize)/255) + 16) + +/*! LZ4_compressBound() : + Provides the maximum size that LZ4 compression may output in a "worst case" scenario (input data not compressible) + This function is primarily useful for memory allocation purposes (destination buffer size). + Macro LZ4_COMPRESSBOUND() is also provided for compilation-time evaluation (stack memory allocation for example). + Note that LZ4_compress_default() compresses faster when dstCapacity is >= LZ4_compressBound(srcSize) + inputSize : max supported value is LZ4_MAX_INPUT_SIZE + return : maximum output size in a "worst case" scenario + or 0, if input size is incorrect (too large or negative) +*/ +LZ4LIB_API int LZ4_compressBound(int inputSize); + +/*! LZ4_compress_fast() : + Same as LZ4_compress_default(), but allows selection of "acceleration" factor. + The larger the acceleration value, the faster the algorithm, but also the lesser the compression. + It's a trade-off. It can be fine tuned, with each successive value providing roughly +~3% to speed. + An acceleration value of "1" is the same as regular LZ4_compress_default() + Values <= 0 will be replaced by LZ4_ACCELERATION_DEFAULT (currently == 1, see lz4.c). + Values > LZ4_ACCELERATION_MAX will be replaced by LZ4_ACCELERATION_MAX (currently == 65537, see lz4.c). +*/ +LZ4LIB_API int LZ4_compress_fast (const char* src, char* dst, int srcSize, int dstCapacity, int acceleration); + + +/*! LZ4_compress_fast_extState() : + * Same as LZ4_compress_fast(), using an externally allocated memory space for its state. + * Use LZ4_sizeofState() to know how much memory must be allocated, + * and allocate it on 8-bytes boundaries (using `malloc()` typically). + * Then, provide this buffer as `void* state` to compression function. + */ +LZ4LIB_API int LZ4_sizeofState(void); +LZ4LIB_API int LZ4_compress_fast_extState (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int acceleration); + + +/*! LZ4_compress_destSize() : + * Reverse the logic : compresses as much data as possible from 'src' buffer + * into already allocated buffer 'dst', of size >= 'targetDestSize'. + * This function either compresses the entire 'src' content into 'dst' if it's large enough, + * or fill 'dst' buffer completely with as much data as possible from 'src'. + * note: acceleration parameter is fixed to "default". + * + * *srcSizePtr : will be modified to indicate how many bytes where read from 'src' to fill 'dst'. + * New value is necessarily <= input value. + * @return : Nb bytes written into 'dst' (necessarily <= targetDestSize) + * or 0 if compression fails. + * + * Note : from v1.8.2 to v1.9.1, this function had a bug (fixed un v1.9.2+): + * the produced compressed content could, in specific circumstances, + * require to be decompressed into a destination buffer larger + * by at least 1 byte than the content to decompress. + * If an application uses `LZ4_compress_destSize()`, + * it's highly recommended to update liblz4 to v1.9.2 or better. + * If this can't be done or ensured, + * the receiving decompression function should provide + * a dstCapacity which is > decompressedSize, by at least 1 byte. + * See https://github.com/lz4/lz4/issues/859 for details + */ +LZ4LIB_API int LZ4_compress_destSize (const char* src, char* dst, int* srcSizePtr, int targetDstSize); + + +/*! LZ4_decompress_safe_partial() : + * Decompress an LZ4 compressed block, of size 'srcSize' at position 'src', + * into destination buffer 'dst' of size 'dstCapacity'. + * Up to 'targetOutputSize' bytes will be decoded. + * The function stops decoding on reaching this objective. + * This can be useful to boost performance + * whenever only the beginning of a block is required. + * + * @return : the number of bytes decoded in `dst` (necessarily <= targetOutputSize) + * If source stream is detected malformed, function returns a negative result. + * + * Note 1 : @return can be < targetOutputSize, if compressed block contains less data. + * + * Note 2 : targetOutputSize must be <= dstCapacity + * + * Note 3 : this function effectively stops decoding on reaching targetOutputSize, + * so dstCapacity is kind of redundant. + * This is because in older versions of this function, + * decoding operation would still write complete sequences. + * Therefore, there was no guarantee that it would stop writing at exactly targetOutputSize, + * it could write more bytes, though only up to dstCapacity. + * Some "margin" used to be required for this operation to work properly. + * Thankfully, this is no longer necessary. + * The function nonetheless keeps the same signature, in an effort to preserve API compatibility. + * + * Note 4 : If srcSize is the exact size of the block, + * then targetOutputSize can be any value, + * including larger than the block's decompressed size. + * The function will, at most, generate block's decompressed size. + * + * Note 5 : If srcSize is _larger_ than block's compressed size, + * then targetOutputSize **MUST** be <= block's decompressed size. + * Otherwise, *silent corruption will occur*. + */ +LZ4LIB_API int LZ4_decompress_safe_partial (const char* src, char* dst, int srcSize, int targetOutputSize, int dstCapacity); + + +/*-********************************************* +* Streaming Compression Functions +***********************************************/ +typedef union LZ4_stream_u LZ4_stream_t; /* incomplete type (defined later) */ + +/** + Note about RC_INVOKED + + - RC_INVOKED is predefined symbol of rc.exe (the resource compiler which is part of MSVC/Visual Studio). + https://docs.microsoft.com/en-us/windows/win32/menurc/predefined-macros + + - Since rc.exe is a legacy compiler, it truncates long symbol (> 30 chars) + and reports warning "RC4011: identifier truncated". + + - To eliminate the warning, we surround long preprocessor symbol with + "#if !defined(RC_INVOKED) ... #endif" block that means + "skip this block when rc.exe is trying to read it". +*/ +#if !defined(RC_INVOKED) /* https://docs.microsoft.com/en-us/windows/win32/menurc/predefined-macros */ +#if !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) +LZ4LIB_API LZ4_stream_t* LZ4_createStream(void); +LZ4LIB_API int LZ4_freeStream (LZ4_stream_t* streamPtr); +#endif /* !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) */ +#endif + +/*! LZ4_resetStream_fast() : v1.9.0+ + * Use this to prepare an LZ4_stream_t for a new chain of dependent blocks + * (e.g., LZ4_compress_fast_continue()). + * + * An LZ4_stream_t must be initialized once before usage. + * This is automatically done when created by LZ4_createStream(). + * However, should the LZ4_stream_t be simply declared on stack (for example), + * it's necessary to initialize it first, using LZ4_initStream(). + * + * After init, start any new stream with LZ4_resetStream_fast(). + * A same LZ4_stream_t can be re-used multiple times consecutively + * and compress multiple streams, + * provided that it starts each new stream with LZ4_resetStream_fast(). + * + * LZ4_resetStream_fast() is much faster than LZ4_initStream(), + * but is not compatible with memory regions containing garbage data. + * + * Note: it's only useful to call LZ4_resetStream_fast() + * in the context of streaming compression. + * The *extState* functions perform their own resets. + * Invoking LZ4_resetStream_fast() before is redundant, and even counterproductive. + */ +LZ4LIB_API void LZ4_resetStream_fast (LZ4_stream_t* streamPtr); + +/*! LZ4_loadDict() : + * Use this function to reference a static dictionary into LZ4_stream_t. + * The dictionary must remain available during compression. + * LZ4_loadDict() triggers a reset, so any previous data will be forgotten. + * The same dictionary will have to be loaded on decompression side for successful decoding. + * Dictionary are useful for better compression of small data (KB range). + * While LZ4 accept any input as dictionary, + * results are generally better when using Zstandard's Dictionary Builder. + * Loading a size of 0 is allowed, and is the same as reset. + * @return : loaded dictionary size, in bytes (necessarily <= 64 KB) + */ +LZ4LIB_API int LZ4_loadDict (LZ4_stream_t* streamPtr, const char* dictionary, int dictSize); + +/*! LZ4_compress_fast_continue() : + * Compress 'src' content using data from previously compressed blocks, for better compression ratio. + * 'dst' buffer must be already allocated. + * If dstCapacity >= LZ4_compressBound(srcSize), compression is guaranteed to succeed, and runs faster. + * + * @return : size of compressed block + * or 0 if there is an error (typically, cannot fit into 'dst'). + * + * Note 1 : Each invocation to LZ4_compress_fast_continue() generates a new block. + * Each block has precise boundaries. + * Each block must be decompressed separately, calling LZ4_decompress_*() with relevant metadata. + * It's not possible to append blocks together and expect a single invocation of LZ4_decompress_*() to decompress them together. + * + * Note 2 : The previous 64KB of source data is __assumed__ to remain present, unmodified, at same address in memory ! + * + * Note 3 : When input is structured as a double-buffer, each buffer can have any size, including < 64 KB. + * Make sure that buffers are separated, by at least one byte. + * This construction ensures that each block only depends on previous block. + * + * Note 4 : If input buffer is a ring-buffer, it can have any size, including < 64 KB. + * + * Note 5 : After an error, the stream status is undefined (invalid), it can only be reset or freed. + */ +LZ4LIB_API int LZ4_compress_fast_continue (LZ4_stream_t* streamPtr, const char* src, char* dst, int srcSize, int dstCapacity, int acceleration); + +/*! LZ4_saveDict() : + * If last 64KB data cannot be guaranteed to remain available at its current memory location, + * save it into a safer place (char* safeBuffer). + * This is schematically equivalent to a memcpy() followed by LZ4_loadDict(), + * but is much faster, because LZ4_saveDict() doesn't need to rebuild tables. + * @return : saved dictionary size in bytes (necessarily <= maxDictSize), or 0 if error. + */ +LZ4LIB_API int LZ4_saveDict (LZ4_stream_t* streamPtr, char* safeBuffer, int maxDictSize); + + +/*-********************************************** +* Streaming Decompression Functions +* Bufferless synchronous API +************************************************/ +typedef union LZ4_streamDecode_u LZ4_streamDecode_t; /* tracking context */ + +/*! LZ4_createStreamDecode() and LZ4_freeStreamDecode() : + * creation / destruction of streaming decompression tracking context. + * A tracking context can be re-used multiple times. + */ +#if !defined(RC_INVOKED) /* https://docs.microsoft.com/en-us/windows/win32/menurc/predefined-macros */ +#if !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) +LZ4LIB_API LZ4_streamDecode_t* LZ4_createStreamDecode(void); +LZ4LIB_API int LZ4_freeStreamDecode (LZ4_streamDecode_t* LZ4_stream); +#endif /* !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) */ +#endif + +/*! LZ4_setStreamDecode() : + * An LZ4_streamDecode_t context can be allocated once and re-used multiple times. + * Use this function to start decompression of a new stream of blocks. + * A dictionary can optionally be set. Use NULL or size 0 for a reset order. + * Dictionary is presumed stable : it must remain accessible and unmodified during next decompression. + * @return : 1 if OK, 0 if error + */ +LZ4LIB_API int LZ4_setStreamDecode (LZ4_streamDecode_t* LZ4_streamDecode, const char* dictionary, int dictSize); + +/*! LZ4_decoderRingBufferSize() : v1.8.2+ + * Note : in a ring buffer scenario (optional), + * blocks are presumed decompressed next to each other + * up to the moment there is not enough remaining space for next block (remainingSize < maxBlockSize), + * at which stage it resumes from beginning of ring buffer. + * When setting such a ring buffer for streaming decompression, + * provides the minimum size of this ring buffer + * to be compatible with any source respecting maxBlockSize condition. + * @return : minimum ring buffer size, + * or 0 if there is an error (invalid maxBlockSize). + */ +LZ4LIB_API int LZ4_decoderRingBufferSize(int maxBlockSize); +#define LZ4_DECODER_RING_BUFFER_SIZE(maxBlockSize) (65536 + 14 + (maxBlockSize)) /* for static allocation; maxBlockSize presumed valid */ + +/*! LZ4_decompress_*_continue() : + * These decoding functions allow decompression of consecutive blocks in "streaming" mode. + * A block is an unsplittable entity, it must be presented entirely to a decompression function. + * Decompression functions only accepts one block at a time. + * The last 64KB of previously decoded data *must* remain available and unmodified at the memory position where they were decoded. + * If less than 64KB of data has been decoded, all the data must be present. + * + * Special : if decompression side sets a ring buffer, it must respect one of the following conditions : + * - Decompression buffer size is _at least_ LZ4_decoderRingBufferSize(maxBlockSize). + * maxBlockSize is the maximum size of any single block. It can have any value > 16 bytes. + * In which case, encoding and decoding buffers do not need to be synchronized. + * Actually, data can be produced by any source compliant with LZ4 format specification, and respecting maxBlockSize. + * - Synchronized mode : + * Decompression buffer size is _exactly_ the same as compression buffer size, + * and follows exactly same update rule (block boundaries at same positions), + * and decoding function is provided with exact decompressed size of each block (exception for last block of the stream), + * _then_ decoding & encoding ring buffer can have any size, including small ones ( < 64 KB). + * - Decompression buffer is larger than encoding buffer, by a minimum of maxBlockSize more bytes. + * In which case, encoding and decoding buffers do not need to be synchronized, + * and encoding ring buffer can have any size, including small ones ( < 64 KB). + * + * Whenever these conditions are not possible, + * save the last 64KB of decoded data into a safe buffer where it can't be modified during decompression, + * then indicate where this data is saved using LZ4_setStreamDecode(), before decompressing next block. +*/ +LZ4LIB_API int +LZ4_decompress_safe_continue (LZ4_streamDecode_t* LZ4_streamDecode, + const char* src, char* dst, + int srcSize, int dstCapacity); + + +/*! LZ4_decompress_*_usingDict() : + * These decoding functions work the same as + * a combination of LZ4_setStreamDecode() followed by LZ4_decompress_*_continue() + * They are stand-alone, and don't need an LZ4_streamDecode_t structure. + * Dictionary is presumed stable : it must remain accessible and unmodified during decompression. + * Performance tip : Decompression speed can be substantially increased + * when dst == dictStart + dictSize. + */ +LZ4LIB_API int +LZ4_decompress_safe_usingDict(const char* src, char* dst, + int srcSize, int dstCapacity, + const char* dictStart, int dictSize); + +LZ4LIB_API int +LZ4_decompress_safe_partial_usingDict(const char* src, char* dst, + int compressedSize, + int targetOutputSize, int maxOutputSize, + const char* dictStart, int dictSize); + +} + +#endif /* LZ4_H_2983827168210 */ + + +/*^************************************* + * !!!!!! STATIC LINKING ONLY !!!!!! + ***************************************/ + +/*-**************************************************************************** + * Experimental section + * + * Symbols declared in this section must be considered unstable. Their + * signatures or semantics may change, or they may be removed altogether in the + * future. They are therefore only safe to depend on when the caller is + * statically linked against the library. + * + * To protect against unsafe usage, not only are the declarations guarded, + * the definitions are hidden by default + * when building LZ4 as a shared/dynamic library. + * + * In order to access these declarations, + * define LZ4_STATIC_LINKING_ONLY in your application + * before including LZ4's headers. + * + * In order to make their implementations accessible dynamically, you must + * define LZ4_PUBLISH_STATIC_FUNCTIONS when building the LZ4 library. + ******************************************************************************/ + +#ifdef LZ4_STATIC_LINKING_ONLY + +#ifndef TRACY_LZ4_STATIC_3504398509 +#define TRACY_LZ4_STATIC_3504398509 + +#ifdef LZ4_PUBLISH_STATIC_FUNCTIONS +#define LZ4LIB_STATIC_API LZ4LIB_API +#else +#define LZ4LIB_STATIC_API +#endif + +namespace tracy +{ + +/*! LZ4_compress_fast_extState_fastReset() : + * A variant of LZ4_compress_fast_extState(). + * + * Using this variant avoids an expensive initialization step. + * It is only safe to call if the state buffer is known to be correctly initialized already + * (see above comment on LZ4_resetStream_fast() for a definition of "correctly initialized"). + * From a high level, the difference is that + * this function initializes the provided state with a call to something like LZ4_resetStream_fast() + * while LZ4_compress_fast_extState() starts with a call to LZ4_resetStream(). + */ +LZ4LIB_STATIC_API int LZ4_compress_fast_extState_fastReset (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int acceleration); + +/*! LZ4_attach_dictionary() : + * This is an experimental API that allows + * efficient use of a static dictionary many times. + * + * Rather than re-loading the dictionary buffer into a working context before + * each compression, or copying a pre-loaded dictionary's LZ4_stream_t into a + * working LZ4_stream_t, this function introduces a no-copy setup mechanism, + * in which the working stream references the dictionary stream in-place. + * + * Several assumptions are made about the state of the dictionary stream. + * Currently, only streams which have been prepared by LZ4_loadDict() should + * be expected to work. + * + * Alternatively, the provided dictionaryStream may be NULL, + * in which case any existing dictionary stream is unset. + * + * If a dictionary is provided, it replaces any pre-existing stream history. + * The dictionary contents are the only history that can be referenced and + * logically immediately precede the data compressed in the first subsequent + * compression call. + * + * The dictionary will only remain attached to the working stream through the + * first compression call, at the end of which it is cleared. The dictionary + * stream (and source buffer) must remain in-place / accessible / unchanged + * through the completion of the first compression call on the stream. + */ +LZ4LIB_STATIC_API void +LZ4_attach_dictionary(LZ4_stream_t* workingStream, + const LZ4_stream_t* dictionaryStream); + + +/*! In-place compression and decompression + * + * It's possible to have input and output sharing the same buffer, + * for highly constrained memory environments. + * In both cases, it requires input to lay at the end of the buffer, + * and decompression to start at beginning of the buffer. + * Buffer size must feature some margin, hence be larger than final size. + * + * |<------------------------buffer--------------------------------->| + * |<-----------compressed data--------->| + * |<-----------decompressed size------------------>| + * |<----margin---->| + * + * This technique is more useful for decompression, + * since decompressed size is typically larger, + * and margin is short. + * + * In-place decompression will work inside any buffer + * which size is >= LZ4_DECOMPRESS_INPLACE_BUFFER_SIZE(decompressedSize). + * This presumes that decompressedSize > compressedSize. + * Otherwise, it means compression actually expanded data, + * and it would be more efficient to store such data with a flag indicating it's not compressed. + * This can happen when data is not compressible (already compressed, or encrypted). + * + * For in-place compression, margin is larger, as it must be able to cope with both + * history preservation, requiring input data to remain unmodified up to LZ4_DISTANCE_MAX, + * and data expansion, which can happen when input is not compressible. + * As a consequence, buffer size requirements are much higher, + * and memory savings offered by in-place compression are more limited. + * + * There are ways to limit this cost for compression : + * - Reduce history size, by modifying LZ4_DISTANCE_MAX. + * Note that it is a compile-time constant, so all compressions will apply this limit. + * Lower values will reduce compression ratio, except when input_size < LZ4_DISTANCE_MAX, + * so it's a reasonable trick when inputs are known to be small. + * - Require the compressor to deliver a "maximum compressed size". + * This is the `dstCapacity` parameter in `LZ4_compress*()`. + * When this size is < LZ4_COMPRESSBOUND(inputSize), then compression can fail, + * in which case, the return code will be 0 (zero). + * The caller must be ready for these cases to happen, + * and typically design a backup scheme to send data uncompressed. + * The combination of both techniques can significantly reduce + * the amount of margin required for in-place compression. + * + * In-place compression can work in any buffer + * which size is >= (maxCompressedSize) + * with maxCompressedSize == LZ4_COMPRESSBOUND(srcSize) for guaranteed compression success. + * LZ4_COMPRESS_INPLACE_BUFFER_SIZE() depends on both maxCompressedSize and LZ4_DISTANCE_MAX, + * so it's possible to reduce memory requirements by playing with them. + */ + +#define LZ4_DECOMPRESS_INPLACE_MARGIN(compressedSize) (((compressedSize) >> 8) + 32) +#define LZ4_DECOMPRESS_INPLACE_BUFFER_SIZE(decompressedSize) ((decompressedSize) + LZ4_DECOMPRESS_INPLACE_MARGIN(decompressedSize)) /**< note: presumes that compressedSize < decompressedSize. note2: margin is overestimated a bit, since it could use compressedSize instead */ + +#ifndef LZ4_DISTANCE_MAX /* history window size; can be user-defined at compile time */ +# define LZ4_DISTANCE_MAX 65535 /* set to maximum value by default */ +#endif + +#define LZ4_COMPRESS_INPLACE_MARGIN (LZ4_DISTANCE_MAX + 32) /* LZ4_DISTANCE_MAX can be safely replaced by srcSize when it's smaller */ +#define LZ4_COMPRESS_INPLACE_BUFFER_SIZE(maxCompressedSize) ((maxCompressedSize) + LZ4_COMPRESS_INPLACE_MARGIN) /**< maxCompressedSize is generally LZ4_COMPRESSBOUND(inputSize), but can be set to any lower value, with the risk that compression can fail (return code 0(zero)) */ + +} + +#endif /* LZ4_STATIC_3504398509 */ +#endif /* LZ4_STATIC_LINKING_ONLY */ + + + +#ifndef TRACY_LZ4_H_98237428734687 +#define TRACY_LZ4_H_98237428734687 + +namespace tracy +{ + +/*-************************************************************ + * Private Definitions + ************************************************************** + * Do not use these definitions directly. + * They are only exposed to allow static allocation of `LZ4_stream_t` and `LZ4_streamDecode_t`. + * Accessing members will expose user code to API and/or ABI break in future versions of the library. + **************************************************************/ +#define LZ4_HASHLOG (LZ4_MEMORY_USAGE-2) +#define LZ4_HASHTABLESIZE (1 << LZ4_MEMORY_USAGE) +#define LZ4_HASH_SIZE_U32 (1 << LZ4_HASHLOG) /* required as macro for static allocation */ + +#if defined(__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) + typedef int8_t LZ4_i8; + typedef uint8_t LZ4_byte; + typedef uint16_t LZ4_u16; + typedef uint32_t LZ4_u32; +#else + typedef signed char LZ4_i8; + typedef unsigned char LZ4_byte; + typedef unsigned short LZ4_u16; + typedef unsigned int LZ4_u32; +#endif + +/*! LZ4_stream_t : + * Never ever use below internal definitions directly ! + * These definitions are not API/ABI safe, and may change in future versions. + * If you need static allocation, declare or allocate an LZ4_stream_t object. +**/ + +typedef struct LZ4_stream_t_internal LZ4_stream_t_internal; +struct LZ4_stream_t_internal { + LZ4_u32 hashTable[LZ4_HASH_SIZE_U32]; + const LZ4_byte* dictionary; + const LZ4_stream_t_internal* dictCtx; + LZ4_u32 currentOffset; + LZ4_u32 tableType; + LZ4_u32 dictSize; + /* Implicit padding to ensure structure is aligned */ +}; + +#define LZ4_STREAM_MINSIZE ((1UL << LZ4_MEMORY_USAGE) + 32) /* static size, for inter-version compatibility */ +union LZ4_stream_u { + char minStateSize[LZ4_STREAM_MINSIZE]; + LZ4_stream_t_internal internal_donotuse; +}; /* previously typedef'd to LZ4_stream_t */ + + +/*! LZ4_initStream() : v1.9.0+ + * An LZ4_stream_t structure must be initialized at least once. + * This is automatically done when invoking LZ4_createStream(), + * but it's not when the structure is simply declared on stack (for example). + * + * Use LZ4_initStream() to properly initialize a newly declared LZ4_stream_t. + * It can also initialize any arbitrary buffer of sufficient size, + * and will @return a pointer of proper type upon initialization. + * + * Note : initialization fails if size and alignment conditions are not respected. + * In which case, the function will @return NULL. + * Note2: An LZ4_stream_t structure guarantees correct alignment and size. + * Note3: Before v1.9.0, use LZ4_resetStream() instead +**/ +LZ4LIB_API LZ4_stream_t* LZ4_initStream (void* buffer, size_t size); + + +/*! LZ4_streamDecode_t : + * Never ever use below internal definitions directly ! + * These definitions are not API/ABI safe, and may change in future versions. + * If you need static allocation, declare or allocate an LZ4_streamDecode_t object. +**/ +typedef struct { + const LZ4_byte* externalDict; + const LZ4_byte* prefixEnd; + size_t extDictSize; + size_t prefixSize; +} LZ4_streamDecode_t_internal; + +#define LZ4_STREAMDECODE_MINSIZE 32 +union LZ4_streamDecode_u { + char minStateSize[LZ4_STREAMDECODE_MINSIZE]; + LZ4_streamDecode_t_internal internal_donotuse; +} ; /* previously typedef'd to LZ4_streamDecode_t */ + + + +/*-************************************ +* Obsolete Functions +**************************************/ + +/*! Deprecation warnings + * + * Deprecated functions make the compiler generate a warning when invoked. + * This is meant to invite users to update their source code. + * Should deprecation warnings be a problem, it is generally possible to disable them, + * typically with -Wno-deprecated-declarations for gcc + * or _CRT_SECURE_NO_WARNINGS in Visual. + * + * Another method is to define LZ4_DISABLE_DEPRECATE_WARNINGS + * before including the header file. + */ +#ifdef LZ4_DISABLE_DEPRECATE_WARNINGS +# define LZ4_DEPRECATED(message) /* disable deprecation warnings */ +#else +# if defined (__cplusplus) && (__cplusplus >= 201402) /* C++14 or greater */ +# define LZ4_DEPRECATED(message) [[deprecated(message)]] +# elif defined(_MSC_VER) +# define LZ4_DEPRECATED(message) __declspec(deprecated(message)) +# elif defined(__clang__) || (defined(__GNUC__) && (__GNUC__ * 10 + __GNUC_MINOR__ >= 45)) +# define LZ4_DEPRECATED(message) __attribute__((deprecated(message))) +# elif defined(__GNUC__) && (__GNUC__ * 10 + __GNUC_MINOR__ >= 31) +# define LZ4_DEPRECATED(message) __attribute__((deprecated)) +# else +# pragma message("WARNING: LZ4_DEPRECATED needs custom implementation for this compiler") +# define LZ4_DEPRECATED(message) /* disabled */ +# endif +#endif /* LZ4_DISABLE_DEPRECATE_WARNINGS */ + +/*! Obsolete compression functions (since v1.7.3) */ +LZ4_DEPRECATED("use LZ4_compress_default() instead") LZ4LIB_API int LZ4_compress (const char* src, char* dest, int srcSize); +LZ4_DEPRECATED("use LZ4_compress_default() instead") LZ4LIB_API int LZ4_compress_limitedOutput (const char* src, char* dest, int srcSize, int maxOutputSize); +LZ4_DEPRECATED("use LZ4_compress_fast_extState() instead") LZ4LIB_API int LZ4_compress_withState (void* state, const char* source, char* dest, int inputSize); +LZ4_DEPRECATED("use LZ4_compress_fast_extState() instead") LZ4LIB_API int LZ4_compress_limitedOutput_withState (void* state, const char* source, char* dest, int inputSize, int maxOutputSize); +LZ4_DEPRECATED("use LZ4_compress_fast_continue() instead") LZ4LIB_API int LZ4_compress_continue (LZ4_stream_t* LZ4_streamPtr, const char* source, char* dest, int inputSize); +LZ4_DEPRECATED("use LZ4_compress_fast_continue() instead") LZ4LIB_API int LZ4_compress_limitedOutput_continue (LZ4_stream_t* LZ4_streamPtr, const char* source, char* dest, int inputSize, int maxOutputSize); + +/*! Obsolete decompression functions (since v1.8.0) */ +LZ4_DEPRECATED("use LZ4_decompress_fast() instead") LZ4LIB_API int LZ4_uncompress (const char* source, char* dest, int outputSize); +LZ4_DEPRECATED("use LZ4_decompress_safe() instead") LZ4LIB_API int LZ4_uncompress_unknownOutputSize (const char* source, char* dest, int isize, int maxOutputSize); + +/* Obsolete streaming functions (since v1.7.0) + * degraded functionality; do not use! + * + * In order to perform streaming compression, these functions depended on data + * that is no longer tracked in the state. They have been preserved as well as + * possible: using them will still produce a correct output. However, they don't + * actually retain any history between compression calls. The compression ratio + * achieved will therefore be no better than compressing each chunk + * independently. + */ +LZ4_DEPRECATED("Use LZ4_createStream() instead") LZ4LIB_API void* LZ4_create (char* inputBuffer); +LZ4_DEPRECATED("Use LZ4_createStream() instead") LZ4LIB_API int LZ4_sizeofStreamState(void); +LZ4_DEPRECATED("Use LZ4_resetStream() instead") LZ4LIB_API int LZ4_resetStreamState(void* state, char* inputBuffer); +LZ4_DEPRECATED("Use LZ4_saveDict() instead") LZ4LIB_API char* LZ4_slideInputBuffer (void* state); + +/*! Obsolete streaming decoding functions (since v1.7.0) */ +LZ4_DEPRECATED("use LZ4_decompress_safe_usingDict() instead") LZ4LIB_API int LZ4_decompress_safe_withPrefix64k (const char* src, char* dst, int compressedSize, int maxDstSize); +LZ4_DEPRECATED("use LZ4_decompress_fast_usingDict() instead") LZ4LIB_API int LZ4_decompress_fast_withPrefix64k (const char* src, char* dst, int originalSize); + +/*! Obsolete LZ4_decompress_fast variants (since v1.9.0) : + * These functions used to be faster than LZ4_decompress_safe(), + * but this is no longer the case. They are now slower. + * This is because LZ4_decompress_fast() doesn't know the input size, + * and therefore must progress more cautiously into the input buffer to not read beyond the end of block. + * On top of that `LZ4_decompress_fast()` is not protected vs malformed or malicious inputs, making it a security liability. + * As a consequence, LZ4_decompress_fast() is strongly discouraged, and deprecated. + * + * The last remaining LZ4_decompress_fast() specificity is that + * it can decompress a block without knowing its compressed size. + * Such functionality can be achieved in a more secure manner + * by employing LZ4_decompress_safe_partial(). + * + * Parameters: + * originalSize : is the uncompressed size to regenerate. + * `dst` must be already allocated, its size must be >= 'originalSize' bytes. + * @return : number of bytes read from source buffer (== compressed size). + * The function expects to finish at block's end exactly. + * If the source stream is detected malformed, the function stops decoding and returns a negative result. + * note : LZ4_decompress_fast*() requires originalSize. Thanks to this information, it never writes past the output buffer. + * However, since it doesn't know its 'src' size, it may read an unknown amount of input, past input buffer bounds. + * Also, since match offsets are not validated, match reads from 'src' may underflow too. + * These issues never happen if input (compressed) data is correct. + * But they may happen if input data is invalid (error or intentional tampering). + * As a consequence, use these functions in trusted environments with trusted data **only**. + */ +LZ4_DEPRECATED("This function is deprecated and unsafe. Consider using LZ4_decompress_safe() instead") +LZ4LIB_API int LZ4_decompress_fast (const char* src, char* dst, int originalSize); +LZ4_DEPRECATED("This function is deprecated and unsafe. Consider using LZ4_decompress_safe_continue() instead") +LZ4LIB_API int LZ4_decompress_fast_continue (LZ4_streamDecode_t* LZ4_streamDecode, const char* src, char* dst, int originalSize); +LZ4_DEPRECATED("This function is deprecated and unsafe. Consider using LZ4_decompress_safe_usingDict() instead") +LZ4LIB_API int LZ4_decompress_fast_usingDict (const char* src, char* dst, int originalSize, const char* dictStart, int dictSize); + +/*! LZ4_resetStream() : + * An LZ4_stream_t structure must be initialized at least once. + * This is done with LZ4_initStream(), or LZ4_resetStream(). + * Consider switching to LZ4_initStream(), + * invoking LZ4_resetStream() will trigger deprecation warnings in the future. + */ +LZ4LIB_API void LZ4_resetStream (LZ4_stream_t* streamPtr); + +} + +#endif /* LZ4_H_98237428734687 */ diff --git a/externals/tracy/public/common/tracy_lz4hc.cpp b/externals/tracy/public/common/tracy_lz4hc.cpp new file mode 100644 index 000000000..eec7239e0 --- /dev/null +++ b/externals/tracy/public/common/tracy_lz4hc.cpp @@ -0,0 +1,1636 @@ +/* + LZ4 HC - High Compression Mode of LZ4 + Copyright (C) 2011-2020, Yann Collet. + + BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are + met: + + * Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above + copyright notice, this list of conditions and the following disclaimer + in the documentation and/or other materials provided with the + distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + You can contact the author at : + - LZ4 source repository : https://github.com/lz4/lz4 + - LZ4 public forum : https://groups.google.com/forum/#!forum/lz4c +*/ +/* note : lz4hc is not an independent module, it requires lz4.h/lz4.c for proper compilation */ + + +/* ************************************* +* Tuning Parameter +***************************************/ + +/*! HEAPMODE : + * Select how default compression function will allocate workplace memory, + * in stack (0:fastest), or in heap (1:requires malloc()). + * Since workplace is rather large, heap mode is recommended. +**/ +#ifndef LZ4HC_HEAPMODE +# define LZ4HC_HEAPMODE 1 +#endif + + +/*=== Dependency ===*/ +#define LZ4_HC_STATIC_LINKING_ONLY +#include "tracy_lz4hc.hpp" + + +/*=== Common definitions ===*/ +#if defined(__GNUC__) +# pragma GCC diagnostic ignored "-Wunused-function" +#endif +#if defined (__clang__) +# pragma clang diagnostic ignored "-Wunused-function" +#endif + +#define LZ4_COMMONDEFS_ONLY +#ifndef LZ4_SRC_INCLUDED +#include "tracy_lz4.cpp" /* LZ4_count, constants, mem */ +#endif + + +/*=== Enums ===*/ +typedef enum { noDictCtx, usingDictCtxHc } dictCtx_directive; + + +/*=== Constants ===*/ +#define OPTIMAL_ML (int)((ML_MASK-1)+MINMATCH) +#define LZ4_OPT_NUM (1<<12) + + +/*=== Macros ===*/ +#define MIN(a,b) ( (a) < (b) ? (a) : (b) ) +#define MAX(a,b) ( (a) > (b) ? (a) : (b) ) +#define HASH_FUNCTION(i) (((i) * 2654435761U) >> ((MINMATCH*8)-LZ4HC_HASH_LOG)) +#define DELTANEXTMAXD(p) chainTable[(p) & LZ4HC_MAXD_MASK] /* flexible, LZ4HC_MAXD dependent */ +#define DELTANEXTU16(table, pos) table[(U16)(pos)] /* faster */ +/* Make fields passed to, and updated by LZ4HC_encodeSequence explicit */ +#define UPDATABLE(ip, op, anchor) &ip, &op, &anchor + +namespace tracy +{ + +static U32 LZ4HC_hashPtr(const void* ptr) { return HASH_FUNCTION(LZ4_read32(ptr)); } + + +/************************************** +* HC Compression +**************************************/ +static void LZ4HC_clearTables (LZ4HC_CCtx_internal* hc4) +{ + MEM_INIT(hc4->hashTable, 0, sizeof(hc4->hashTable)); + MEM_INIT(hc4->chainTable, 0xFF, sizeof(hc4->chainTable)); +} + +static void LZ4HC_init_internal (LZ4HC_CCtx_internal* hc4, const BYTE* start) +{ + size_t const bufferSize = (size_t)(hc4->end - hc4->prefixStart); + size_t newStartingOffset = bufferSize + hc4->dictLimit; + assert(newStartingOffset >= bufferSize); /* check overflow */ + if (newStartingOffset > 1 GB) { + LZ4HC_clearTables(hc4); + newStartingOffset = 0; + } + newStartingOffset += 64 KB; + hc4->nextToUpdate = (U32)newStartingOffset; + hc4->prefixStart = start; + hc4->end = start; + hc4->dictStart = start; + hc4->dictLimit = (U32)newStartingOffset; + hc4->lowLimit = (U32)newStartingOffset; +} + + +/* Update chains up to ip (excluded) */ +LZ4_FORCE_INLINE void LZ4HC_Insert (LZ4HC_CCtx_internal* hc4, const BYTE* ip) +{ + U16* const chainTable = hc4->chainTable; + U32* const hashTable = hc4->hashTable; + const BYTE* const prefixPtr = hc4->prefixStart; + U32 const prefixIdx = hc4->dictLimit; + U32 const target = (U32)(ip - prefixPtr) + prefixIdx; + U32 idx = hc4->nextToUpdate; + assert(ip >= prefixPtr); + assert(target >= prefixIdx); + + while (idx < target) { + U32 const h = LZ4HC_hashPtr(prefixPtr+idx-prefixIdx); + size_t delta = idx - hashTable[h]; + if (delta>LZ4_DISTANCE_MAX) delta = LZ4_DISTANCE_MAX; + DELTANEXTU16(chainTable, idx) = (U16)delta; + hashTable[h] = idx; + idx++; + } + + hc4->nextToUpdate = target; +} + +/** LZ4HC_countBack() : + * @return : negative value, nb of common bytes before ip/match */ +LZ4_FORCE_INLINE +int LZ4HC_countBack(const BYTE* const ip, const BYTE* const match, + const BYTE* const iMin, const BYTE* const mMin) +{ + int back = 0; + int const min = (int)MAX(iMin - ip, mMin - match); + assert(min <= 0); + assert(ip >= iMin); assert((size_t)(ip-iMin) < (1U<<31)); + assert(match >= mMin); assert((size_t)(match - mMin) < (1U<<31)); + while ( (back > min) + && (ip[back-1] == match[back-1]) ) + back--; + return back; +} + +#if defined(_MSC_VER) +# define LZ4HC_rotl32(x,r) _rotl(x,r) +#else +# define LZ4HC_rotl32(x,r) ((x << r) | (x >> (32 - r))) +#endif + + +static U32 LZ4HC_rotatePattern(size_t const rotate, U32 const pattern) +{ + size_t const bitsToRotate = (rotate & (sizeof(pattern) - 1)) << 3; + if (bitsToRotate == 0) return pattern; + return LZ4HC_rotl32(pattern, (int)bitsToRotate); +} + +/* LZ4HC_countPattern() : + * pattern32 must be a sample of repetitive pattern of length 1, 2 or 4 (but not 3!) */ +static unsigned +LZ4HC_countPattern(const BYTE* ip, const BYTE* const iEnd, U32 const pattern32) +{ + const BYTE* const iStart = ip; + reg_t const pattern = (sizeof(pattern)==8) ? + (reg_t)pattern32 + (((reg_t)pattern32) << (sizeof(pattern)*4)) : pattern32; + + while (likely(ip < iEnd-(sizeof(pattern)-1))) { + reg_t const diff = LZ4_read_ARCH(ip) ^ pattern; + if (!diff) { ip+=sizeof(pattern); continue; } + ip += LZ4_NbCommonBytes(diff); + return (unsigned)(ip - iStart); + } + + if (LZ4_isLittleEndian()) { + reg_t patternByte = pattern; + while ((ip>= 8; + } + } else { /* big endian */ + U32 bitOffset = (sizeof(pattern)*8) - 8; + while (ip < iEnd) { + BYTE const byte = (BYTE)(pattern >> bitOffset); + if (*ip != byte) break; + ip ++; bitOffset -= 8; + } } + + return (unsigned)(ip - iStart); +} + +/* LZ4HC_reverseCountPattern() : + * pattern must be a sample of repetitive pattern of length 1, 2 or 4 (but not 3!) + * read using natural platform endianness */ +static unsigned +LZ4HC_reverseCountPattern(const BYTE* ip, const BYTE* const iLow, U32 pattern) +{ + const BYTE* const iStart = ip; + + while (likely(ip >= iLow+4)) { + if (LZ4_read32(ip-4) != pattern) break; + ip -= 4; + } + { const BYTE* bytePtr = (const BYTE*)(&pattern) + 3; /* works for any endianness */ + while (likely(ip>iLow)) { + if (ip[-1] != *bytePtr) break; + ip--; bytePtr--; + } } + return (unsigned)(iStart - ip); +} + +/* LZ4HC_protectDictEnd() : + * Checks if the match is in the last 3 bytes of the dictionary, so reading the + * 4 byte MINMATCH would overflow. + * @returns true if the match index is okay. + */ +static int LZ4HC_protectDictEnd(U32 const dictLimit, U32 const matchIndex) +{ + return ((U32)((dictLimit - 1) - matchIndex) >= 3); +} + +typedef enum { rep_untested, rep_not, rep_confirmed } repeat_state_e; +typedef enum { favorCompressionRatio=0, favorDecompressionSpeed } HCfavor_e; + +LZ4_FORCE_INLINE int +LZ4HC_InsertAndGetWiderMatch ( + LZ4HC_CCtx_internal* const hc4, + const BYTE* const ip, + const BYTE* const iLowLimit, const BYTE* const iHighLimit, + int longest, + const BYTE** matchpos, + const BYTE** startpos, + const int maxNbAttempts, + const int patternAnalysis, const int chainSwap, + const dictCtx_directive dict, + const HCfavor_e favorDecSpeed) +{ + U16* const chainTable = hc4->chainTable; + U32* const HashTable = hc4->hashTable; + const LZ4HC_CCtx_internal * const dictCtx = hc4->dictCtx; + const BYTE* const prefixPtr = hc4->prefixStart; + const U32 prefixIdx = hc4->dictLimit; + const U32 ipIndex = (U32)(ip - prefixPtr) + prefixIdx; + const int withinStartDistance = (hc4->lowLimit + (LZ4_DISTANCE_MAX + 1) > ipIndex); + const U32 lowestMatchIndex = (withinStartDistance) ? hc4->lowLimit : ipIndex - LZ4_DISTANCE_MAX; + const BYTE* const dictStart = hc4->dictStart; + const U32 dictIdx = hc4->lowLimit; + const BYTE* const dictEnd = dictStart + prefixIdx - dictIdx; + int const lookBackLength = (int)(ip-iLowLimit); + int nbAttempts = maxNbAttempts; + U32 matchChainPos = 0; + U32 const pattern = LZ4_read32(ip); + U32 matchIndex; + repeat_state_e repeat = rep_untested; + size_t srcPatternLength = 0; + + DEBUGLOG(7, "LZ4HC_InsertAndGetWiderMatch"); + /* First Match */ + LZ4HC_Insert(hc4, ip); + matchIndex = HashTable[LZ4HC_hashPtr(ip)]; + DEBUGLOG(7, "First match at index %u / %u (lowestMatchIndex)", + matchIndex, lowestMatchIndex); + + while ((matchIndex>=lowestMatchIndex) && (nbAttempts>0)) { + int matchLength=0; + nbAttempts--; + assert(matchIndex < ipIndex); + if (favorDecSpeed && (ipIndex - matchIndex < 8)) { + /* do nothing */ + } else if (matchIndex >= prefixIdx) { /* within current Prefix */ + const BYTE* const matchPtr = prefixPtr + matchIndex - prefixIdx; + assert(matchPtr < ip); + assert(longest >= 1); + if (LZ4_read16(iLowLimit + longest - 1) == LZ4_read16(matchPtr - lookBackLength + longest - 1)) { + if (LZ4_read32(matchPtr) == pattern) { + int const back = lookBackLength ? LZ4HC_countBack(ip, matchPtr, iLowLimit, prefixPtr) : 0; + matchLength = MINMATCH + (int)LZ4_count(ip+MINMATCH, matchPtr+MINMATCH, iHighLimit); + matchLength -= back; + if (matchLength > longest) { + longest = matchLength; + *matchpos = matchPtr + back; + *startpos = ip + back; + } } } + } else { /* lowestMatchIndex <= matchIndex < dictLimit */ + const BYTE* const matchPtr = dictStart + (matchIndex - dictIdx); + assert(matchIndex >= dictIdx); + if ( likely(matchIndex <= prefixIdx - 4) + && (LZ4_read32(matchPtr) == pattern) ) { + int back = 0; + const BYTE* vLimit = ip + (prefixIdx - matchIndex); + if (vLimit > iHighLimit) vLimit = iHighLimit; + matchLength = (int)LZ4_count(ip+MINMATCH, matchPtr+MINMATCH, vLimit) + MINMATCH; + if ((ip+matchLength == vLimit) && (vLimit < iHighLimit)) + matchLength += LZ4_count(ip+matchLength, prefixPtr, iHighLimit); + back = lookBackLength ? LZ4HC_countBack(ip, matchPtr, iLowLimit, dictStart) : 0; + matchLength -= back; + if (matchLength > longest) { + longest = matchLength; + *matchpos = prefixPtr - prefixIdx + matchIndex + back; /* virtual pos, relative to ip, to retrieve offset */ + *startpos = ip + back; + } } } + + if (chainSwap && matchLength==longest) { /* better match => select a better chain */ + assert(lookBackLength==0); /* search forward only */ + if (matchIndex + (U32)longest <= ipIndex) { + int const kTrigger = 4; + U32 distanceToNextMatch = 1; + int const end = longest - MINMATCH + 1; + int step = 1; + int accel = 1 << kTrigger; + int pos; + for (pos = 0; pos < end; pos += step) { + U32 const candidateDist = DELTANEXTU16(chainTable, matchIndex + (U32)pos); + step = (accel++ >> kTrigger); + if (candidateDist > distanceToNextMatch) { + distanceToNextMatch = candidateDist; + matchChainPos = (U32)pos; + accel = 1 << kTrigger; + } } + if (distanceToNextMatch > 1) { + if (distanceToNextMatch > matchIndex) break; /* avoid overflow */ + matchIndex -= distanceToNextMatch; + continue; + } } } + + { U32 const distNextMatch = DELTANEXTU16(chainTable, matchIndex); + if (patternAnalysis && distNextMatch==1 && matchChainPos==0) { + U32 const matchCandidateIdx = matchIndex-1; + /* may be a repeated pattern */ + if (repeat == rep_untested) { + if ( ((pattern & 0xFFFF) == (pattern >> 16)) + & ((pattern & 0xFF) == (pattern >> 24)) ) { + repeat = rep_confirmed; + srcPatternLength = LZ4HC_countPattern(ip+sizeof(pattern), iHighLimit, pattern) + sizeof(pattern); + } else { + repeat = rep_not; + } } + if ( (repeat == rep_confirmed) && (matchCandidateIdx >= lowestMatchIndex) + && LZ4HC_protectDictEnd(prefixIdx, matchCandidateIdx) ) { + const int extDict = matchCandidateIdx < prefixIdx; + const BYTE* const matchPtr = (extDict ? dictStart - dictIdx : prefixPtr - prefixIdx) + matchCandidateIdx; + if (LZ4_read32(matchPtr) == pattern) { /* good candidate */ + const BYTE* const iLimit = extDict ? dictEnd : iHighLimit; + size_t forwardPatternLength = LZ4HC_countPattern(matchPtr+sizeof(pattern), iLimit, pattern) + sizeof(pattern); + if (extDict && matchPtr + forwardPatternLength == iLimit) { + U32 const rotatedPattern = LZ4HC_rotatePattern(forwardPatternLength, pattern); + forwardPatternLength += LZ4HC_countPattern(prefixPtr, iHighLimit, rotatedPattern); + } + { const BYTE* const lowestMatchPtr = extDict ? dictStart : prefixPtr; + size_t backLength = LZ4HC_reverseCountPattern(matchPtr, lowestMatchPtr, pattern); + size_t currentSegmentLength; + if (!extDict + && matchPtr - backLength == prefixPtr + && dictIdx < prefixIdx) { + U32 const rotatedPattern = LZ4HC_rotatePattern((U32)(-(int)backLength), pattern); + backLength += LZ4HC_reverseCountPattern(dictEnd, dictStart, rotatedPattern); + } + /* Limit backLength not go further than lowestMatchIndex */ + backLength = matchCandidateIdx - MAX(matchCandidateIdx - (U32)backLength, lowestMatchIndex); + assert(matchCandidateIdx - backLength >= lowestMatchIndex); + currentSegmentLength = backLength + forwardPatternLength; + /* Adjust to end of pattern if the source pattern fits, otherwise the beginning of the pattern */ + if ( (currentSegmentLength >= srcPatternLength) /* current pattern segment large enough to contain full srcPatternLength */ + && (forwardPatternLength <= srcPatternLength) ) { /* haven't reached this position yet */ + U32 const newMatchIndex = matchCandidateIdx + (U32)forwardPatternLength - (U32)srcPatternLength; /* best position, full pattern, might be followed by more match */ + if (LZ4HC_protectDictEnd(prefixIdx, newMatchIndex)) + matchIndex = newMatchIndex; + else { + /* Can only happen if started in the prefix */ + assert(newMatchIndex >= prefixIdx - 3 && newMatchIndex < prefixIdx && !extDict); + matchIndex = prefixIdx; + } + } else { + U32 const newMatchIndex = matchCandidateIdx - (U32)backLength; /* farthest position in current segment, will find a match of length currentSegmentLength + maybe some back */ + if (!LZ4HC_protectDictEnd(prefixIdx, newMatchIndex)) { + assert(newMatchIndex >= prefixIdx - 3 && newMatchIndex < prefixIdx && !extDict); + matchIndex = prefixIdx; + } else { + matchIndex = newMatchIndex; + if (lookBackLength==0) { /* no back possible */ + size_t const maxML = MIN(currentSegmentLength, srcPatternLength); + if ((size_t)longest < maxML) { + assert(prefixPtr - prefixIdx + matchIndex != ip); + if ((size_t)(ip - prefixPtr) + prefixIdx - matchIndex > LZ4_DISTANCE_MAX) break; + assert(maxML < 2 GB); + longest = (int)maxML; + *matchpos = prefixPtr - prefixIdx + matchIndex; /* virtual pos, relative to ip, to retrieve offset */ + *startpos = ip; + } + { U32 const distToNextPattern = DELTANEXTU16(chainTable, matchIndex); + if (distToNextPattern > matchIndex) break; /* avoid overflow */ + matchIndex -= distToNextPattern; + } } } } } + continue; + } } + } } /* PA optimization */ + + /* follow current chain */ + matchIndex -= DELTANEXTU16(chainTable, matchIndex + matchChainPos); + + } /* while ((matchIndex>=lowestMatchIndex) && (nbAttempts)) */ + + if ( dict == usingDictCtxHc + && nbAttempts > 0 + && ipIndex - lowestMatchIndex < LZ4_DISTANCE_MAX) { + size_t const dictEndOffset = (size_t)(dictCtx->end - dictCtx->prefixStart) + dictCtx->dictLimit; + U32 dictMatchIndex = dictCtx->hashTable[LZ4HC_hashPtr(ip)]; + assert(dictEndOffset <= 1 GB); + matchIndex = dictMatchIndex + lowestMatchIndex - (U32)dictEndOffset; + while (ipIndex - matchIndex <= LZ4_DISTANCE_MAX && nbAttempts--) { + const BYTE* const matchPtr = dictCtx->prefixStart - dictCtx->dictLimit + dictMatchIndex; + + if (LZ4_read32(matchPtr) == pattern) { + int mlt; + int back = 0; + const BYTE* vLimit = ip + (dictEndOffset - dictMatchIndex); + if (vLimit > iHighLimit) vLimit = iHighLimit; + mlt = (int)LZ4_count(ip+MINMATCH, matchPtr+MINMATCH, vLimit) + MINMATCH; + back = lookBackLength ? LZ4HC_countBack(ip, matchPtr, iLowLimit, dictCtx->prefixStart) : 0; + mlt -= back; + if (mlt > longest) { + longest = mlt; + *matchpos = prefixPtr - prefixIdx + matchIndex + back; + *startpos = ip + back; + } } + + { U32 const nextOffset = DELTANEXTU16(dictCtx->chainTable, dictMatchIndex); + dictMatchIndex -= nextOffset; + matchIndex -= nextOffset; + } } } + + return longest; +} + +LZ4_FORCE_INLINE int +LZ4HC_InsertAndFindBestMatch(LZ4HC_CCtx_internal* const hc4, /* Index table will be updated */ + const BYTE* const ip, const BYTE* const iLimit, + const BYTE** matchpos, + const int maxNbAttempts, + const int patternAnalysis, + const dictCtx_directive dict) +{ + const BYTE* uselessPtr = ip; + /* note : LZ4HC_InsertAndGetWiderMatch() is able to modify the starting position of a match (*startpos), + * but this won't be the case here, as we define iLowLimit==ip, + * so LZ4HC_InsertAndGetWiderMatch() won't be allowed to search past ip */ + return LZ4HC_InsertAndGetWiderMatch(hc4, ip, ip, iLimit, MINMATCH-1, matchpos, &uselessPtr, maxNbAttempts, patternAnalysis, 0 /*chainSwap*/, dict, favorCompressionRatio); +} + +/* LZ4HC_encodeSequence() : + * @return : 0 if ok, + * 1 if buffer issue detected */ +LZ4_FORCE_INLINE int LZ4HC_encodeSequence ( + const BYTE** _ip, + BYTE** _op, + const BYTE** _anchor, + int matchLength, + const BYTE* const match, + limitedOutput_directive limit, + BYTE* oend) +{ +#define ip (*_ip) +#define op (*_op) +#define anchor (*_anchor) + + size_t length; + BYTE* const token = op++; + +#if defined(LZ4_DEBUG) && (LZ4_DEBUG >= 6) + static const BYTE* start = NULL; + static U32 totalCost = 0; + U32 const pos = (start==NULL) ? 0 : (U32)(anchor - start); + U32 const ll = (U32)(ip - anchor); + U32 const llAdd = (ll>=15) ? ((ll-15) / 255) + 1 : 0; + U32 const mlAdd = (matchLength>=19) ? ((matchLength-19) / 255) + 1 : 0; + U32 const cost = 1 + llAdd + ll + 2 + mlAdd; + if (start==NULL) start = anchor; /* only works for single segment */ + /* g_debuglog_enable = (pos >= 2228) & (pos <= 2262); */ + DEBUGLOG(6, "pos:%7u -- literals:%4u, match:%4i, offset:%5u, cost:%4u + %5u", + pos, + (U32)(ip - anchor), matchLength, (U32)(ip-match), + cost, totalCost); + totalCost += cost; +#endif + + /* Encode Literal length */ + length = (size_t)(ip - anchor); + LZ4_STATIC_ASSERT(notLimited == 0); + /* Check output limit */ + if (limit && ((op + (length / 255) + length + (2 + 1 + LASTLITERALS)) > oend)) { + DEBUGLOG(6, "Not enough room to write %i literals (%i bytes remaining)", + (int)length, (int)(oend - op)); + return 1; + } + if (length >= RUN_MASK) { + size_t len = length - RUN_MASK; + *token = (RUN_MASK << ML_BITS); + for(; len >= 255 ; len -= 255) *op++ = 255; + *op++ = (BYTE)len; + } else { + *token = (BYTE)(length << ML_BITS); + } + + /* Copy Literals */ + LZ4_wildCopy8(op, anchor, op + length); + op += length; + + /* Encode Offset */ + assert( (ip - match) <= LZ4_DISTANCE_MAX ); /* note : consider providing offset as a value, rather than as a pointer difference */ + LZ4_writeLE16(op, (U16)(ip - match)); op += 2; + + /* Encode MatchLength */ + assert(matchLength >= MINMATCH); + length = (size_t)matchLength - MINMATCH; + if (limit && (op + (length / 255) + (1 + LASTLITERALS) > oend)) { + DEBUGLOG(6, "Not enough room to write match length"); + return 1; /* Check output limit */ + } + if (length >= ML_MASK) { + *token += ML_MASK; + length -= ML_MASK; + for(; length >= 510 ; length -= 510) { *op++ = 255; *op++ = 255; } + if (length >= 255) { length -= 255; *op++ = 255; } + *op++ = (BYTE)length; + } else { + *token += (BYTE)(length); + } + + /* Prepare next loop */ + ip += matchLength; + anchor = ip; + + return 0; +} +#undef ip +#undef op +#undef anchor + +LZ4_FORCE_INLINE int LZ4HC_compress_hashChain ( + LZ4HC_CCtx_internal* const ctx, + const char* const source, + char* const dest, + int* srcSizePtr, + int const maxOutputSize, + int maxNbAttempts, + const limitedOutput_directive limit, + const dictCtx_directive dict + ) +{ + const int inputSize = *srcSizePtr; + const int patternAnalysis = (maxNbAttempts > 128); /* levels 9+ */ + + const BYTE* ip = (const BYTE*) source; + const BYTE* anchor = ip; + const BYTE* const iend = ip + inputSize; + const BYTE* const mflimit = iend - MFLIMIT; + const BYTE* const matchlimit = (iend - LASTLITERALS); + + BYTE* optr = (BYTE*) dest; + BYTE* op = (BYTE*) dest; + BYTE* oend = op + maxOutputSize; + + int ml0, ml, ml2, ml3; + const BYTE* start0; + const BYTE* ref0; + const BYTE* ref = NULL; + const BYTE* start2 = NULL; + const BYTE* ref2 = NULL; + const BYTE* start3 = NULL; + const BYTE* ref3 = NULL; + + /* init */ + *srcSizePtr = 0; + if (limit == fillOutput) oend -= LASTLITERALS; /* Hack for support LZ4 format restriction */ + if (inputSize < LZ4_minLength) goto _last_literals; /* Input too small, no compression (all literals) */ + + /* Main Loop */ + while (ip <= mflimit) { + ml = LZ4HC_InsertAndFindBestMatch(ctx, ip, matchlimit, &ref, maxNbAttempts, patternAnalysis, dict); + if (ml encode ML1 */ + optr = op; + if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; + continue; + } + + if (start0 < ip) { /* first match was skipped at least once */ + if (start2 < ip + ml0) { /* squeezing ML1 between ML0(original ML1) and ML2 */ + ip = start0; ref = ref0; ml = ml0; /* restore initial ML1 */ + } } + + /* Here, start0==ip */ + if ((start2 - ip) < 3) { /* First Match too small : removed */ + ml = ml2; + ip = start2; + ref =ref2; + goto _Search2; + } + +_Search3: + /* At this stage, we have : + * ml2 > ml1, and + * ip1+3 <= ip2 (usually < ip1+ml1) */ + if ((start2 - ip) < OPTIMAL_ML) { + int correction; + int new_ml = ml; + if (new_ml > OPTIMAL_ML) new_ml = OPTIMAL_ML; + if (ip+new_ml > start2 + ml2 - MINMATCH) new_ml = (int)(start2 - ip) + ml2 - MINMATCH; + correction = new_ml - (int)(start2 - ip); + if (correction > 0) { + start2 += correction; + ref2 += correction; + ml2 -= correction; + } + } + /* Now, we have start2 = ip+new_ml, with new_ml = min(ml, OPTIMAL_ML=18) */ + + if (start2 + ml2 <= mflimit) { + ml3 = LZ4HC_InsertAndGetWiderMatch(ctx, + start2 + ml2 - 3, start2, matchlimit, ml2, &ref3, &start3, + maxNbAttempts, patternAnalysis, 0, dict, favorCompressionRatio); + } else { + ml3 = ml2; + } + + if (ml3 == ml2) { /* No better match => encode ML1 and ML2 */ + /* ip & ref are known; Now for ml */ + if (start2 < ip+ml) ml = (int)(start2 - ip); + /* Now, encode 2 sequences */ + optr = op; + if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; + ip = start2; + optr = op; + if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml2, ref2, limit, oend)) { + ml = ml2; + ref = ref2; + goto _dest_overflow; + } + continue; + } + + if (start3 < ip+ml+3) { /* Not enough space for match 2 : remove it */ + if (start3 >= (ip+ml)) { /* can write Seq1 immediately ==> Seq2 is removed, so Seq3 becomes Seq1 */ + if (start2 < ip+ml) { + int correction = (int)(ip+ml - start2); + start2 += correction; + ref2 += correction; + ml2 -= correction; + if (ml2 < MINMATCH) { + start2 = start3; + ref2 = ref3; + ml2 = ml3; + } + } + + optr = op; + if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; + ip = start3; + ref = ref3; + ml = ml3; + + start0 = start2; + ref0 = ref2; + ml0 = ml2; + goto _Search2; + } + + start2 = start3; + ref2 = ref3; + ml2 = ml3; + goto _Search3; + } + + /* + * OK, now we have 3 ascending matches; + * let's write the first one ML1. + * ip & ref are known; Now decide ml. + */ + if (start2 < ip+ml) { + if ((start2 - ip) < OPTIMAL_ML) { + int correction; + if (ml > OPTIMAL_ML) ml = OPTIMAL_ML; + if (ip + ml > start2 + ml2 - MINMATCH) ml = (int)(start2 - ip) + ml2 - MINMATCH; + correction = ml - (int)(start2 - ip); + if (correction > 0) { + start2 += correction; + ref2 += correction; + ml2 -= correction; + } + } else { + ml = (int)(start2 - ip); + } + } + optr = op; + if (LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, limit, oend)) goto _dest_overflow; + + /* ML2 becomes ML1 */ + ip = start2; ref = ref2; ml = ml2; + + /* ML3 becomes ML2 */ + start2 = start3; ref2 = ref3; ml2 = ml3; + + /* let's find a new ML3 */ + goto _Search3; + } + +_last_literals: + /* Encode Last Literals */ + { size_t lastRunSize = (size_t)(iend - anchor); /* literals */ + size_t llAdd = (lastRunSize + 255 - RUN_MASK) / 255; + size_t const totalSize = 1 + llAdd + lastRunSize; + if (limit == fillOutput) oend += LASTLITERALS; /* restore correct value */ + if (limit && (op + totalSize > oend)) { + if (limit == limitedOutput) return 0; + /* adapt lastRunSize to fill 'dest' */ + lastRunSize = (size_t)(oend - op) - 1 /*token*/; + llAdd = (lastRunSize + 256 - RUN_MASK) / 256; + lastRunSize -= llAdd; + } + DEBUGLOG(6, "Final literal run : %i literals", (int)lastRunSize); + ip = anchor + lastRunSize; /* can be != iend if limit==fillOutput */ + + if (lastRunSize >= RUN_MASK) { + size_t accumulator = lastRunSize - RUN_MASK; + *op++ = (RUN_MASK << ML_BITS); + for(; accumulator >= 255 ; accumulator -= 255) *op++ = 255; + *op++ = (BYTE) accumulator; + } else { + *op++ = (BYTE)(lastRunSize << ML_BITS); + } + LZ4_memcpy(op, anchor, lastRunSize); + op += lastRunSize; + } + + /* End */ + *srcSizePtr = (int) (((const char*)ip) - source); + return (int) (((char*)op)-dest); + +_dest_overflow: + if (limit == fillOutput) { + /* Assumption : ip, anchor, ml and ref must be set correctly */ + size_t const ll = (size_t)(ip - anchor); + size_t const ll_addbytes = (ll + 240) / 255; + size_t const ll_totalCost = 1 + ll_addbytes + ll; + BYTE* const maxLitPos = oend - 3; /* 2 for offset, 1 for token */ + DEBUGLOG(6, "Last sequence overflowing"); + op = optr; /* restore correct out pointer */ + if (op + ll_totalCost <= maxLitPos) { + /* ll validated; now adjust match length */ + size_t const bytesLeftForMl = (size_t)(maxLitPos - (op+ll_totalCost)); + size_t const maxMlSize = MINMATCH + (ML_MASK-1) + (bytesLeftForMl * 255); + assert(maxMlSize < INT_MAX); assert(ml >= 0); + if ((size_t)ml > maxMlSize) ml = (int)maxMlSize; + if ((oend + LASTLITERALS) - (op + ll_totalCost + 2) - 1 + ml >= MFLIMIT) { + LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ref, notLimited, oend); + } } + goto _last_literals; + } + /* compression failed */ + return 0; +} + + +static int LZ4HC_compress_optimal( LZ4HC_CCtx_internal* ctx, + const char* const source, char* dst, + int* srcSizePtr, int dstCapacity, + int const nbSearches, size_t sufficient_len, + const limitedOutput_directive limit, int const fullUpdate, + const dictCtx_directive dict, + const HCfavor_e favorDecSpeed); + + +LZ4_FORCE_INLINE int LZ4HC_compress_generic_internal ( + LZ4HC_CCtx_internal* const ctx, + const char* const src, + char* const dst, + int* const srcSizePtr, + int const dstCapacity, + int cLevel, + const limitedOutput_directive limit, + const dictCtx_directive dict + ) +{ + typedef enum { lz4hc, lz4opt } lz4hc_strat_e; + typedef struct { + lz4hc_strat_e strat; + int nbSearches; + U32 targetLength; + } cParams_t; + static const cParams_t clTable[LZ4HC_CLEVEL_MAX+1] = { + { lz4hc, 2, 16 }, /* 0, unused */ + { lz4hc, 2, 16 }, /* 1, unused */ + { lz4hc, 2, 16 }, /* 2, unused */ + { lz4hc, 4, 16 }, /* 3 */ + { lz4hc, 8, 16 }, /* 4 */ + { lz4hc, 16, 16 }, /* 5 */ + { lz4hc, 32, 16 }, /* 6 */ + { lz4hc, 64, 16 }, /* 7 */ + { lz4hc, 128, 16 }, /* 8 */ + { lz4hc, 256, 16 }, /* 9 */ + { lz4opt, 96, 64 }, /*10==LZ4HC_CLEVEL_OPT_MIN*/ + { lz4opt, 512,128 }, /*11 */ + { lz4opt,16384,LZ4_OPT_NUM }, /* 12==LZ4HC_CLEVEL_MAX */ + }; + + DEBUGLOG(4, "LZ4HC_compress_generic(ctx=%p, src=%p, srcSize=%d, limit=%d)", + ctx, src, *srcSizePtr, limit); + + if (limit == fillOutput && dstCapacity < 1) return 0; /* Impossible to store anything */ + if ((U32)*srcSizePtr > (U32)LZ4_MAX_INPUT_SIZE) return 0; /* Unsupported input size (too large or negative) */ + + ctx->end += *srcSizePtr; + if (cLevel < 1) cLevel = LZ4HC_CLEVEL_DEFAULT; /* note : convention is different from lz4frame, maybe something to review */ + cLevel = MIN(LZ4HC_CLEVEL_MAX, cLevel); + { cParams_t const cParam = clTable[cLevel]; + HCfavor_e const favor = ctx->favorDecSpeed ? favorDecompressionSpeed : favorCompressionRatio; + int result; + + if (cParam.strat == lz4hc) { + result = LZ4HC_compress_hashChain(ctx, + src, dst, srcSizePtr, dstCapacity, + cParam.nbSearches, limit, dict); + } else { + assert(cParam.strat == lz4opt); + result = LZ4HC_compress_optimal(ctx, + src, dst, srcSizePtr, dstCapacity, + cParam.nbSearches, cParam.targetLength, limit, + cLevel == LZ4HC_CLEVEL_MAX, /* ultra mode */ + dict, favor); + } + if (result <= 0) ctx->dirty = 1; + return result; + } +} + +static void LZ4HC_setExternalDict(LZ4HC_CCtx_internal* ctxPtr, const BYTE* newBlock); + +static int +LZ4HC_compress_generic_noDictCtx ( + LZ4HC_CCtx_internal* const ctx, + const char* const src, + char* const dst, + int* const srcSizePtr, + int const dstCapacity, + int cLevel, + limitedOutput_directive limit + ) +{ + assert(ctx->dictCtx == NULL); + return LZ4HC_compress_generic_internal(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit, noDictCtx); +} + +static int +LZ4HC_compress_generic_dictCtx ( + LZ4HC_CCtx_internal* const ctx, + const char* const src, + char* const dst, + int* const srcSizePtr, + int const dstCapacity, + int cLevel, + limitedOutput_directive limit + ) +{ + const size_t position = (size_t)(ctx->end - ctx->prefixStart) + (ctx->dictLimit - ctx->lowLimit); + assert(ctx->dictCtx != NULL); + if (position >= 64 KB) { + ctx->dictCtx = NULL; + return LZ4HC_compress_generic_noDictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); + } else if (position == 0 && *srcSizePtr > 4 KB) { + LZ4_memcpy(ctx, ctx->dictCtx, sizeof(LZ4HC_CCtx_internal)); + LZ4HC_setExternalDict(ctx, (const BYTE *)src); + ctx->compressionLevel = (short)cLevel; + return LZ4HC_compress_generic_noDictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); + } else { + return LZ4HC_compress_generic_internal(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit, usingDictCtxHc); + } +} + +static int +LZ4HC_compress_generic ( + LZ4HC_CCtx_internal* const ctx, + const char* const src, + char* const dst, + int* const srcSizePtr, + int const dstCapacity, + int cLevel, + limitedOutput_directive limit + ) +{ + if (ctx->dictCtx == NULL) { + return LZ4HC_compress_generic_noDictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); + } else { + return LZ4HC_compress_generic_dictCtx(ctx, src, dst, srcSizePtr, dstCapacity, cLevel, limit); + } +} + + +int LZ4_sizeofStateHC(void) { return (int)sizeof(LZ4_streamHC_t); } + +static size_t LZ4_streamHC_t_alignment(void) +{ +#if LZ4_ALIGN_TEST + typedef struct { char c; LZ4_streamHC_t t; } t_a; + return sizeof(t_a) - sizeof(LZ4_streamHC_t); +#else + return 1; /* effectively disabled */ +#endif +} + +/* state is presumed correctly initialized, + * in which case its size and alignment have already been validate */ +int LZ4_compress_HC_extStateHC_fastReset (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel) +{ + LZ4HC_CCtx_internal* const ctx = &((LZ4_streamHC_t*)state)->internal_donotuse; + if (!LZ4_isAligned(state, LZ4_streamHC_t_alignment())) return 0; + LZ4_resetStreamHC_fast((LZ4_streamHC_t*)state, compressionLevel); + LZ4HC_init_internal (ctx, (const BYTE*)src); + if (dstCapacity < LZ4_compressBound(srcSize)) + return LZ4HC_compress_generic (ctx, src, dst, &srcSize, dstCapacity, compressionLevel, limitedOutput); + else + return LZ4HC_compress_generic (ctx, src, dst, &srcSize, dstCapacity, compressionLevel, notLimited); +} + +int LZ4_compress_HC_extStateHC (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel) +{ + LZ4_streamHC_t* const ctx = LZ4_initStreamHC(state, sizeof(*ctx)); + if (ctx==NULL) return 0; /* init failure */ + return LZ4_compress_HC_extStateHC_fastReset(state, src, dst, srcSize, dstCapacity, compressionLevel); +} + +int LZ4_compress_HC(const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel) +{ + int cSize; +#if defined(LZ4HC_HEAPMODE) && LZ4HC_HEAPMODE==1 + LZ4_streamHC_t* const statePtr = (LZ4_streamHC_t*)ALLOC(sizeof(LZ4_streamHC_t)); + if (statePtr==NULL) return 0; +#else + LZ4_streamHC_t state; + LZ4_streamHC_t* const statePtr = &state; +#endif + cSize = LZ4_compress_HC_extStateHC(statePtr, src, dst, srcSize, dstCapacity, compressionLevel); +#if defined(LZ4HC_HEAPMODE) && LZ4HC_HEAPMODE==1 + FREEMEM(statePtr); +#endif + return cSize; +} + +/* state is presumed sized correctly (>= sizeof(LZ4_streamHC_t)) */ +int LZ4_compress_HC_destSize(void* state, const char* source, char* dest, int* sourceSizePtr, int targetDestSize, int cLevel) +{ + LZ4_streamHC_t* const ctx = LZ4_initStreamHC(state, sizeof(*ctx)); + if (ctx==NULL) return 0; /* init failure */ + LZ4HC_init_internal(&ctx->internal_donotuse, (const BYTE*) source); + LZ4_setCompressionLevel(ctx, cLevel); + return LZ4HC_compress_generic(&ctx->internal_donotuse, source, dest, sourceSizePtr, targetDestSize, cLevel, fillOutput); +} + + + +/************************************** +* Streaming Functions +**************************************/ +/* allocation */ +#if !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) +LZ4_streamHC_t* LZ4_createStreamHC(void) +{ + LZ4_streamHC_t* const state = + (LZ4_streamHC_t*)ALLOC_AND_ZERO(sizeof(LZ4_streamHC_t)); + if (state == NULL) return NULL; + LZ4_setCompressionLevel(state, LZ4HC_CLEVEL_DEFAULT); + return state; +} + +int LZ4_freeStreamHC (LZ4_streamHC_t* LZ4_streamHCPtr) +{ + DEBUGLOG(4, "LZ4_freeStreamHC(%p)", LZ4_streamHCPtr); + if (!LZ4_streamHCPtr) return 0; /* support free on NULL */ + FREEMEM(LZ4_streamHCPtr); + return 0; +} +#endif + + +LZ4_streamHC_t* LZ4_initStreamHC (void* buffer, size_t size) +{ + LZ4_streamHC_t* const LZ4_streamHCPtr = (LZ4_streamHC_t*)buffer; + DEBUGLOG(4, "LZ4_initStreamHC(%p, %u)", buffer, (unsigned)size); + /* check conditions */ + if (buffer == NULL) return NULL; + if (size < sizeof(LZ4_streamHC_t)) return NULL; + if (!LZ4_isAligned(buffer, LZ4_streamHC_t_alignment())) return NULL; + /* init */ + { LZ4HC_CCtx_internal* const hcstate = &(LZ4_streamHCPtr->internal_donotuse); + MEM_INIT(hcstate, 0, sizeof(*hcstate)); } + LZ4_setCompressionLevel(LZ4_streamHCPtr, LZ4HC_CLEVEL_DEFAULT); + return LZ4_streamHCPtr; +} + +/* just a stub */ +void LZ4_resetStreamHC (LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel) +{ + LZ4_initStreamHC(LZ4_streamHCPtr, sizeof(*LZ4_streamHCPtr)); + LZ4_setCompressionLevel(LZ4_streamHCPtr, compressionLevel); +} + +void LZ4_resetStreamHC_fast (LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel) +{ + DEBUGLOG(4, "LZ4_resetStreamHC_fast(%p, %d)", LZ4_streamHCPtr, compressionLevel); + if (LZ4_streamHCPtr->internal_donotuse.dirty) { + LZ4_initStreamHC(LZ4_streamHCPtr, sizeof(*LZ4_streamHCPtr)); + } else { + /* preserve end - prefixStart : can trigger clearTable's threshold */ + if (LZ4_streamHCPtr->internal_donotuse.end != NULL) { + LZ4_streamHCPtr->internal_donotuse.end -= (uptrval)LZ4_streamHCPtr->internal_donotuse.prefixStart; + } else { + assert(LZ4_streamHCPtr->internal_donotuse.prefixStart == NULL); + } + LZ4_streamHCPtr->internal_donotuse.prefixStart = NULL; + LZ4_streamHCPtr->internal_donotuse.dictCtx = NULL; + } + LZ4_setCompressionLevel(LZ4_streamHCPtr, compressionLevel); +} + +void LZ4_setCompressionLevel(LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel) +{ + DEBUGLOG(5, "LZ4_setCompressionLevel(%p, %d)", LZ4_streamHCPtr, compressionLevel); + if (compressionLevel < 1) compressionLevel = LZ4HC_CLEVEL_DEFAULT; + if (compressionLevel > LZ4HC_CLEVEL_MAX) compressionLevel = LZ4HC_CLEVEL_MAX; + LZ4_streamHCPtr->internal_donotuse.compressionLevel = (short)compressionLevel; +} + +void LZ4_favorDecompressionSpeed(LZ4_streamHC_t* LZ4_streamHCPtr, int favor) +{ + LZ4_streamHCPtr->internal_donotuse.favorDecSpeed = (favor!=0); +} + +/* LZ4_loadDictHC() : + * LZ4_streamHCPtr is presumed properly initialized */ +int LZ4_loadDictHC (LZ4_streamHC_t* LZ4_streamHCPtr, + const char* dictionary, int dictSize) +{ + LZ4HC_CCtx_internal* const ctxPtr = &LZ4_streamHCPtr->internal_donotuse; + DEBUGLOG(4, "LZ4_loadDictHC(ctx:%p, dict:%p, dictSize:%d)", LZ4_streamHCPtr, dictionary, dictSize); + assert(LZ4_streamHCPtr != NULL); + if (dictSize > 64 KB) { + dictionary += (size_t)dictSize - 64 KB; + dictSize = 64 KB; + } + /* need a full initialization, there are bad side-effects when using resetFast() */ + { int const cLevel = ctxPtr->compressionLevel; + LZ4_initStreamHC(LZ4_streamHCPtr, sizeof(*LZ4_streamHCPtr)); + LZ4_setCompressionLevel(LZ4_streamHCPtr, cLevel); + } + LZ4HC_init_internal (ctxPtr, (const BYTE*)dictionary); + ctxPtr->end = (const BYTE*)dictionary + dictSize; + if (dictSize >= 4) LZ4HC_Insert (ctxPtr, ctxPtr->end-3); + return dictSize; +} + +void LZ4_attach_HC_dictionary(LZ4_streamHC_t *working_stream, const LZ4_streamHC_t *dictionary_stream) { + working_stream->internal_donotuse.dictCtx = dictionary_stream != NULL ? &(dictionary_stream->internal_donotuse) : NULL; +} + +/* compression */ + +static void LZ4HC_setExternalDict(LZ4HC_CCtx_internal* ctxPtr, const BYTE* newBlock) +{ + DEBUGLOG(4, "LZ4HC_setExternalDict(%p, %p)", ctxPtr, newBlock); + if (ctxPtr->end >= ctxPtr->prefixStart + 4) + LZ4HC_Insert (ctxPtr, ctxPtr->end-3); /* Referencing remaining dictionary content */ + + /* Only one memory segment for extDict, so any previous extDict is lost at this stage */ + ctxPtr->lowLimit = ctxPtr->dictLimit; + ctxPtr->dictStart = ctxPtr->prefixStart; + ctxPtr->dictLimit += (U32)(ctxPtr->end - ctxPtr->prefixStart); + ctxPtr->prefixStart = newBlock; + ctxPtr->end = newBlock; + ctxPtr->nextToUpdate = ctxPtr->dictLimit; /* match referencing will resume from there */ + + /* cannot reference an extDict and a dictCtx at the same time */ + ctxPtr->dictCtx = NULL; +} + +static int +LZ4_compressHC_continue_generic (LZ4_streamHC_t* LZ4_streamHCPtr, + const char* src, char* dst, + int* srcSizePtr, int dstCapacity, + limitedOutput_directive limit) +{ + LZ4HC_CCtx_internal* const ctxPtr = &LZ4_streamHCPtr->internal_donotuse; + DEBUGLOG(5, "LZ4_compressHC_continue_generic(ctx=%p, src=%p, srcSize=%d, limit=%d)", + LZ4_streamHCPtr, src, *srcSizePtr, limit); + assert(ctxPtr != NULL); + /* auto-init if forgotten */ + if (ctxPtr->prefixStart == NULL) LZ4HC_init_internal (ctxPtr, (const BYTE*) src); + + /* Check overflow */ + if ((size_t)(ctxPtr->end - ctxPtr->prefixStart) + ctxPtr->dictLimit > 2 GB) { + size_t dictSize = (size_t)(ctxPtr->end - ctxPtr->prefixStart); + if (dictSize > 64 KB) dictSize = 64 KB; + LZ4_loadDictHC(LZ4_streamHCPtr, (const char*)(ctxPtr->end) - dictSize, (int)dictSize); + } + + /* Check if blocks follow each other */ + if ((const BYTE*)src != ctxPtr->end) + LZ4HC_setExternalDict(ctxPtr, (const BYTE*)src); + + /* Check overlapping input/dictionary space */ + { const BYTE* sourceEnd = (const BYTE*) src + *srcSizePtr; + const BYTE* const dictBegin = ctxPtr->dictStart; + const BYTE* const dictEnd = ctxPtr->dictStart + (ctxPtr->dictLimit - ctxPtr->lowLimit); + if ((sourceEnd > dictBegin) && ((const BYTE*)src < dictEnd)) { + if (sourceEnd > dictEnd) sourceEnd = dictEnd; + ctxPtr->lowLimit += (U32)(sourceEnd - ctxPtr->dictStart); + ctxPtr->dictStart += (U32)(sourceEnd - ctxPtr->dictStart); + if (ctxPtr->dictLimit - ctxPtr->lowLimit < 4) { + ctxPtr->lowLimit = ctxPtr->dictLimit; + ctxPtr->dictStart = ctxPtr->prefixStart; + } } } + + return LZ4HC_compress_generic (ctxPtr, src, dst, srcSizePtr, dstCapacity, ctxPtr->compressionLevel, limit); +} + +int LZ4_compress_HC_continue (LZ4_streamHC_t* LZ4_streamHCPtr, const char* src, char* dst, int srcSize, int dstCapacity) +{ + if (dstCapacity < LZ4_compressBound(srcSize)) + return LZ4_compressHC_continue_generic (LZ4_streamHCPtr, src, dst, &srcSize, dstCapacity, limitedOutput); + else + return LZ4_compressHC_continue_generic (LZ4_streamHCPtr, src, dst, &srcSize, dstCapacity, notLimited); +} + +int LZ4_compress_HC_continue_destSize (LZ4_streamHC_t* LZ4_streamHCPtr, const char* src, char* dst, int* srcSizePtr, int targetDestSize) +{ + return LZ4_compressHC_continue_generic(LZ4_streamHCPtr, src, dst, srcSizePtr, targetDestSize, fillOutput); +} + + + +/* LZ4_saveDictHC : + * save history content + * into a user-provided buffer + * which is then used to continue compression + */ +int LZ4_saveDictHC (LZ4_streamHC_t* LZ4_streamHCPtr, char* safeBuffer, int dictSize) +{ + LZ4HC_CCtx_internal* const streamPtr = &LZ4_streamHCPtr->internal_donotuse; + int const prefixSize = (int)(streamPtr->end - streamPtr->prefixStart); + DEBUGLOG(5, "LZ4_saveDictHC(%p, %p, %d)", LZ4_streamHCPtr, safeBuffer, dictSize); + assert(prefixSize >= 0); + if (dictSize > 64 KB) dictSize = 64 KB; + if (dictSize < 4) dictSize = 0; + if (dictSize > prefixSize) dictSize = prefixSize; + if (safeBuffer == NULL) assert(dictSize == 0); + if (dictSize > 0) + LZ4_memmove(safeBuffer, streamPtr->end - dictSize, dictSize); + { U32 const endIndex = (U32)(streamPtr->end - streamPtr->prefixStart) + streamPtr->dictLimit; + streamPtr->end = (const BYTE*)safeBuffer + dictSize; + streamPtr->prefixStart = streamPtr->end - dictSize; + streamPtr->dictLimit = endIndex - (U32)dictSize; + streamPtr->lowLimit = endIndex - (U32)dictSize; + streamPtr->dictStart = streamPtr->prefixStart; + if (streamPtr->nextToUpdate < streamPtr->dictLimit) + streamPtr->nextToUpdate = streamPtr->dictLimit; + } + return dictSize; +} + + +/*************************************************** +* Deprecated Functions +***************************************************/ + +/* These functions currently generate deprecation warnings */ + +/* Wrappers for deprecated compression functions */ +int LZ4_compressHC(const char* src, char* dst, int srcSize) { return LZ4_compress_HC (src, dst, srcSize, LZ4_compressBound(srcSize), 0); } +int LZ4_compressHC_limitedOutput(const char* src, char* dst, int srcSize, int maxDstSize) { return LZ4_compress_HC(src, dst, srcSize, maxDstSize, 0); } +int LZ4_compressHC2(const char* src, char* dst, int srcSize, int cLevel) { return LZ4_compress_HC (src, dst, srcSize, LZ4_compressBound(srcSize), cLevel); } +int LZ4_compressHC2_limitedOutput(const char* src, char* dst, int srcSize, int maxDstSize, int cLevel) { return LZ4_compress_HC(src, dst, srcSize, maxDstSize, cLevel); } +int LZ4_compressHC_withStateHC (void* state, const char* src, char* dst, int srcSize) { return LZ4_compress_HC_extStateHC (state, src, dst, srcSize, LZ4_compressBound(srcSize), 0); } +int LZ4_compressHC_limitedOutput_withStateHC (void* state, const char* src, char* dst, int srcSize, int maxDstSize) { return LZ4_compress_HC_extStateHC (state, src, dst, srcSize, maxDstSize, 0); } +int LZ4_compressHC2_withStateHC (void* state, const char* src, char* dst, int srcSize, int cLevel) { return LZ4_compress_HC_extStateHC(state, src, dst, srcSize, LZ4_compressBound(srcSize), cLevel); } +int LZ4_compressHC2_limitedOutput_withStateHC (void* state, const char* src, char* dst, int srcSize, int maxDstSize, int cLevel) { return LZ4_compress_HC_extStateHC(state, src, dst, srcSize, maxDstSize, cLevel); } +int LZ4_compressHC_continue (LZ4_streamHC_t* ctx, const char* src, char* dst, int srcSize) { return LZ4_compress_HC_continue (ctx, src, dst, srcSize, LZ4_compressBound(srcSize)); } +int LZ4_compressHC_limitedOutput_continue (LZ4_streamHC_t* ctx, const char* src, char* dst, int srcSize, int maxDstSize) { return LZ4_compress_HC_continue (ctx, src, dst, srcSize, maxDstSize); } + + +/* Deprecated streaming functions */ +int LZ4_sizeofStreamStateHC(void) { return sizeof(LZ4_streamHC_t); } + +/* state is presumed correctly sized, aka >= sizeof(LZ4_streamHC_t) + * @return : 0 on success, !=0 if error */ +int LZ4_resetStreamStateHC(void* state, char* inputBuffer) +{ + LZ4_streamHC_t* const hc4 = LZ4_initStreamHC(state, sizeof(*hc4)); + if (hc4 == NULL) return 1; /* init failed */ + LZ4HC_init_internal (&hc4->internal_donotuse, (const BYTE*)inputBuffer); + return 0; +} + +#if !defined(LZ4_STATIC_LINKING_ONLY_DISABLE_MEMORY_ALLOCATION) +void* LZ4_createHC (const char* inputBuffer) +{ + LZ4_streamHC_t* const hc4 = LZ4_createStreamHC(); + if (hc4 == NULL) return NULL; /* not enough memory */ + LZ4HC_init_internal (&hc4->internal_donotuse, (const BYTE*)inputBuffer); + return hc4; +} + +int LZ4_freeHC (void* LZ4HC_Data) +{ + if (!LZ4HC_Data) return 0; /* support free on NULL */ + FREEMEM(LZ4HC_Data); + return 0; +} +#endif + +int LZ4_compressHC2_continue (void* LZ4HC_Data, const char* src, char* dst, int srcSize, int cLevel) +{ + return LZ4HC_compress_generic (&((LZ4_streamHC_t*)LZ4HC_Data)->internal_donotuse, src, dst, &srcSize, 0, cLevel, notLimited); +} + +int LZ4_compressHC2_limitedOutput_continue (void* LZ4HC_Data, const char* src, char* dst, int srcSize, int dstCapacity, int cLevel) +{ + return LZ4HC_compress_generic (&((LZ4_streamHC_t*)LZ4HC_Data)->internal_donotuse, src, dst, &srcSize, dstCapacity, cLevel, limitedOutput); +} + +char* LZ4_slideInputBufferHC(void* LZ4HC_Data) +{ + LZ4_streamHC_t* const ctx = (LZ4_streamHC_t*)LZ4HC_Data; + const BYTE* bufferStart = ctx->internal_donotuse.prefixStart - ctx->internal_donotuse.dictLimit + ctx->internal_donotuse.lowLimit; + LZ4_resetStreamHC_fast(ctx, ctx->internal_donotuse.compressionLevel); + /* avoid const char * -> char * conversion warning :( */ + return (char*)(uptrval)bufferStart; +} + + +/* ================================================ + * LZ4 Optimal parser (levels [LZ4HC_CLEVEL_OPT_MIN - LZ4HC_CLEVEL_MAX]) + * ===============================================*/ +typedef struct { + int price; + int off; + int mlen; + int litlen; +} LZ4HC_optimal_t; + +/* price in bytes */ +LZ4_FORCE_INLINE int LZ4HC_literalsPrice(int const litlen) +{ + int price = litlen; + assert(litlen >= 0); + if (litlen >= (int)RUN_MASK) + price += 1 + ((litlen-(int)RUN_MASK) / 255); + return price; +} + + +/* requires mlen >= MINMATCH */ +LZ4_FORCE_INLINE int LZ4HC_sequencePrice(int litlen, int mlen) +{ + int price = 1 + 2 ; /* token + 16-bit offset */ + assert(litlen >= 0); + assert(mlen >= MINMATCH); + + price += LZ4HC_literalsPrice(litlen); + + if (mlen >= (int)(ML_MASK+MINMATCH)) + price += 1 + ((mlen-(int)(ML_MASK+MINMATCH)) / 255); + + return price; +} + + +typedef struct { + int off; + int len; +} LZ4HC_match_t; + +LZ4_FORCE_INLINE LZ4HC_match_t +LZ4HC_FindLongerMatch(LZ4HC_CCtx_internal* const ctx, + const BYTE* ip, const BYTE* const iHighLimit, + int minLen, int nbSearches, + const dictCtx_directive dict, + const HCfavor_e favorDecSpeed) +{ + LZ4HC_match_t match = { 0 , 0 }; + const BYTE* matchPtr = NULL; + /* note : LZ4HC_InsertAndGetWiderMatch() is able to modify the starting position of a match (*startpos), + * but this won't be the case here, as we define iLowLimit==ip, + * so LZ4HC_InsertAndGetWiderMatch() won't be allowed to search past ip */ + int matchLength = LZ4HC_InsertAndGetWiderMatch(ctx, ip, ip, iHighLimit, minLen, &matchPtr, &ip, nbSearches, 1 /*patternAnalysis*/, 1 /*chainSwap*/, dict, favorDecSpeed); + if (matchLength <= minLen) return match; + if (favorDecSpeed) { + if ((matchLength>18) & (matchLength<=36)) matchLength=18; /* favor shortcut */ + } + match.len = matchLength; + match.off = (int)(ip-matchPtr); + return match; +} + + +static int LZ4HC_compress_optimal ( LZ4HC_CCtx_internal* ctx, + const char* const source, + char* dst, + int* srcSizePtr, + int dstCapacity, + int const nbSearches, + size_t sufficient_len, + const limitedOutput_directive limit, + int const fullUpdate, + const dictCtx_directive dict, + const HCfavor_e favorDecSpeed) +{ + int retval = 0; +#define TRAILING_LITERALS 3 +#if defined(LZ4HC_HEAPMODE) && LZ4HC_HEAPMODE==1 + LZ4HC_optimal_t* const opt = (LZ4HC_optimal_t*)ALLOC(sizeof(LZ4HC_optimal_t) * (LZ4_OPT_NUM + TRAILING_LITERALS)); +#else + LZ4HC_optimal_t opt[LZ4_OPT_NUM + TRAILING_LITERALS]; /* ~64 KB, which is a bit large for stack... */ +#endif + + const BYTE* ip = (const BYTE*) source; + const BYTE* anchor = ip; + const BYTE* const iend = ip + *srcSizePtr; + const BYTE* const mflimit = iend - MFLIMIT; + const BYTE* const matchlimit = iend - LASTLITERALS; + BYTE* op = (BYTE*) dst; + BYTE* opSaved = (BYTE*) dst; + BYTE* oend = op + dstCapacity; + int ovml = MINMATCH; /* overflow - last sequence */ + const BYTE* ovref = NULL; + + /* init */ +#if defined(LZ4HC_HEAPMODE) && LZ4HC_HEAPMODE==1 + if (opt == NULL) goto _return_label; +#endif + DEBUGLOG(5, "LZ4HC_compress_optimal(dst=%p, dstCapa=%u)", dst, (unsigned)dstCapacity); + *srcSizePtr = 0; + if (limit == fillOutput) oend -= LASTLITERALS; /* Hack for support LZ4 format restriction */ + if (sufficient_len >= LZ4_OPT_NUM) sufficient_len = LZ4_OPT_NUM-1; + + /* Main Loop */ + while (ip <= mflimit) { + int const llen = (int)(ip - anchor); + int best_mlen, best_off; + int cur, last_match_pos = 0; + + LZ4HC_match_t const firstMatch = LZ4HC_FindLongerMatch(ctx, ip, matchlimit, MINMATCH-1, nbSearches, dict, favorDecSpeed); + if (firstMatch.len==0) { ip++; continue; } + + if ((size_t)firstMatch.len > sufficient_len) { + /* good enough solution : immediate encoding */ + int const firstML = firstMatch.len; + const BYTE* const matchPos = ip - firstMatch.off; + opSaved = op; + if ( LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), firstML, matchPos, limit, oend) ) { /* updates ip, op and anchor */ + ovml = firstML; + ovref = matchPos; + goto _dest_overflow; + } + continue; + } + + /* set prices for first positions (literals) */ + { int rPos; + for (rPos = 0 ; rPos < MINMATCH ; rPos++) { + int const cost = LZ4HC_literalsPrice(llen + rPos); + opt[rPos].mlen = 1; + opt[rPos].off = 0; + opt[rPos].litlen = llen + rPos; + opt[rPos].price = cost; + DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i) -- initial setup", + rPos, cost, opt[rPos].litlen); + } } + /* set prices using initial match */ + { int mlen = MINMATCH; + int const matchML = firstMatch.len; /* necessarily < sufficient_len < LZ4_OPT_NUM */ + int const offset = firstMatch.off; + assert(matchML < LZ4_OPT_NUM); + for ( ; mlen <= matchML ; mlen++) { + int const cost = LZ4HC_sequencePrice(llen, mlen); + opt[mlen].mlen = mlen; + opt[mlen].off = offset; + opt[mlen].litlen = llen; + opt[mlen].price = cost; + DEBUGLOG(7, "rPos:%3i => price:%3i (matchlen=%i) -- initial setup", + mlen, cost, mlen); + } } + last_match_pos = firstMatch.len; + { int addLit; + for (addLit = 1; addLit <= TRAILING_LITERALS; addLit ++) { + opt[last_match_pos+addLit].mlen = 1; /* literal */ + opt[last_match_pos+addLit].off = 0; + opt[last_match_pos+addLit].litlen = addLit; + opt[last_match_pos+addLit].price = opt[last_match_pos].price + LZ4HC_literalsPrice(addLit); + DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i) -- initial setup", + last_match_pos+addLit, opt[last_match_pos+addLit].price, addLit); + } } + + /* check further positions */ + for (cur = 1; cur < last_match_pos; cur++) { + const BYTE* const curPtr = ip + cur; + LZ4HC_match_t newMatch; + + if (curPtr > mflimit) break; + DEBUGLOG(7, "rPos:%u[%u] vs [%u]%u", + cur, opt[cur].price, opt[cur+1].price, cur+1); + if (fullUpdate) { + /* not useful to search here if next position has same (or lower) cost */ + if ( (opt[cur+1].price <= opt[cur].price) + /* in some cases, next position has same cost, but cost rises sharply after, so a small match would still be beneficial */ + && (opt[cur+MINMATCH].price < opt[cur].price + 3/*min seq price*/) ) + continue; + } else { + /* not useful to search here if next position has same (or lower) cost */ + if (opt[cur+1].price <= opt[cur].price) continue; + } + + DEBUGLOG(7, "search at rPos:%u", cur); + if (fullUpdate) + newMatch = LZ4HC_FindLongerMatch(ctx, curPtr, matchlimit, MINMATCH-1, nbSearches, dict, favorDecSpeed); + else + /* only test matches of minimum length; slightly faster, but misses a few bytes */ + newMatch = LZ4HC_FindLongerMatch(ctx, curPtr, matchlimit, last_match_pos - cur, nbSearches, dict, favorDecSpeed); + if (!newMatch.len) continue; + + if ( ((size_t)newMatch.len > sufficient_len) + || (newMatch.len + cur >= LZ4_OPT_NUM) ) { + /* immediate encoding */ + best_mlen = newMatch.len; + best_off = newMatch.off; + last_match_pos = cur + 1; + goto encode; + } + + /* before match : set price with literals at beginning */ + { int const baseLitlen = opt[cur].litlen; + int litlen; + for (litlen = 1; litlen < MINMATCH; litlen++) { + int const price = opt[cur].price - LZ4HC_literalsPrice(baseLitlen) + LZ4HC_literalsPrice(baseLitlen+litlen); + int const pos = cur + litlen; + if (price < opt[pos].price) { + opt[pos].mlen = 1; /* literal */ + opt[pos].off = 0; + opt[pos].litlen = baseLitlen+litlen; + opt[pos].price = price; + DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i)", + pos, price, opt[pos].litlen); + } } } + + /* set prices using match at position = cur */ + { int const matchML = newMatch.len; + int ml = MINMATCH; + + assert(cur + newMatch.len < LZ4_OPT_NUM); + for ( ; ml <= matchML ; ml++) { + int const pos = cur + ml; + int const offset = newMatch.off; + int price; + int ll; + DEBUGLOG(7, "testing price rPos %i (last_match_pos=%i)", + pos, last_match_pos); + if (opt[cur].mlen == 1) { + ll = opt[cur].litlen; + price = ((cur > ll) ? opt[cur - ll].price : 0) + + LZ4HC_sequencePrice(ll, ml); + } else { + ll = 0; + price = opt[cur].price + LZ4HC_sequencePrice(0, ml); + } + + assert((U32)favorDecSpeed <= 1); + if (pos > last_match_pos+TRAILING_LITERALS + || price <= opt[pos].price - (int)favorDecSpeed) { + DEBUGLOG(7, "rPos:%3i => price:%3i (matchlen=%i)", + pos, price, ml); + assert(pos < LZ4_OPT_NUM); + if ( (ml == matchML) /* last pos of last match */ + && (last_match_pos < pos) ) + last_match_pos = pos; + opt[pos].mlen = ml; + opt[pos].off = offset; + opt[pos].litlen = ll; + opt[pos].price = price; + } } } + /* complete following positions with literals */ + { int addLit; + for (addLit = 1; addLit <= TRAILING_LITERALS; addLit ++) { + opt[last_match_pos+addLit].mlen = 1; /* literal */ + opt[last_match_pos+addLit].off = 0; + opt[last_match_pos+addLit].litlen = addLit; + opt[last_match_pos+addLit].price = opt[last_match_pos].price + LZ4HC_literalsPrice(addLit); + DEBUGLOG(7, "rPos:%3i => price:%3i (litlen=%i)", last_match_pos+addLit, opt[last_match_pos+addLit].price, addLit); + } } + } /* for (cur = 1; cur <= last_match_pos; cur++) */ + + assert(last_match_pos < LZ4_OPT_NUM + TRAILING_LITERALS); + best_mlen = opt[last_match_pos].mlen; + best_off = opt[last_match_pos].off; + cur = last_match_pos - best_mlen; + +encode: /* cur, last_match_pos, best_mlen, best_off must be set */ + assert(cur < LZ4_OPT_NUM); + assert(last_match_pos >= 1); /* == 1 when only one candidate */ + DEBUGLOG(6, "reverse traversal, looking for shortest path (last_match_pos=%i)", last_match_pos); + { int candidate_pos = cur; + int selected_matchLength = best_mlen; + int selected_offset = best_off; + while (1) { /* from end to beginning */ + int const next_matchLength = opt[candidate_pos].mlen; /* can be 1, means literal */ + int const next_offset = opt[candidate_pos].off; + DEBUGLOG(7, "pos %i: sequence length %i", candidate_pos, selected_matchLength); + opt[candidate_pos].mlen = selected_matchLength; + opt[candidate_pos].off = selected_offset; + selected_matchLength = next_matchLength; + selected_offset = next_offset; + if (next_matchLength > candidate_pos) break; /* last match elected, first match to encode */ + assert(next_matchLength > 0); /* can be 1, means literal */ + candidate_pos -= next_matchLength; + } } + + /* encode all recorded sequences in order */ + { int rPos = 0; /* relative position (to ip) */ + while (rPos < last_match_pos) { + int const ml = opt[rPos].mlen; + int const offset = opt[rPos].off; + if (ml == 1) { ip++; rPos++; continue; } /* literal; note: can end up with several literals, in which case, skip them */ + rPos += ml; + assert(ml >= MINMATCH); + assert((offset >= 1) && (offset <= LZ4_DISTANCE_MAX)); + opSaved = op; + if ( LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ml, ip - offset, limit, oend) ) { /* updates ip, op and anchor */ + ovml = ml; + ovref = ip - offset; + goto _dest_overflow; + } } } + } /* while (ip <= mflimit) */ + +_last_literals: + /* Encode Last Literals */ + { size_t lastRunSize = (size_t)(iend - anchor); /* literals */ + size_t llAdd = (lastRunSize + 255 - RUN_MASK) / 255; + size_t const totalSize = 1 + llAdd + lastRunSize; + if (limit == fillOutput) oend += LASTLITERALS; /* restore correct value */ + if (limit && (op + totalSize > oend)) { + if (limit == limitedOutput) { /* Check output limit */ + retval = 0; + goto _return_label; + } + /* adapt lastRunSize to fill 'dst' */ + lastRunSize = (size_t)(oend - op) - 1 /*token*/; + llAdd = (lastRunSize + 256 - RUN_MASK) / 256; + lastRunSize -= llAdd; + } + DEBUGLOG(6, "Final literal run : %i literals", (int)lastRunSize); + ip = anchor + lastRunSize; /* can be != iend if limit==fillOutput */ + + if (lastRunSize >= RUN_MASK) { + size_t accumulator = lastRunSize - RUN_MASK; + *op++ = (RUN_MASK << ML_BITS); + for(; accumulator >= 255 ; accumulator -= 255) *op++ = 255; + *op++ = (BYTE) accumulator; + } else { + *op++ = (BYTE)(lastRunSize << ML_BITS); + } + LZ4_memcpy(op, anchor, lastRunSize); + op += lastRunSize; + } + + /* End */ + *srcSizePtr = (int) (((const char*)ip) - source); + retval = (int) ((char*)op-dst); + goto _return_label; + +_dest_overflow: +if (limit == fillOutput) { + /* Assumption : ip, anchor, ovml and ovref must be set correctly */ + size_t const ll = (size_t)(ip - anchor); + size_t const ll_addbytes = (ll + 240) / 255; + size_t const ll_totalCost = 1 + ll_addbytes + ll; + BYTE* const maxLitPos = oend - 3; /* 2 for offset, 1 for token */ + DEBUGLOG(6, "Last sequence overflowing (only %i bytes remaining)", (int)(oend-1-opSaved)); + op = opSaved; /* restore correct out pointer */ + if (op + ll_totalCost <= maxLitPos) { + /* ll validated; now adjust match length */ + size_t const bytesLeftForMl = (size_t)(maxLitPos - (op+ll_totalCost)); + size_t const maxMlSize = MINMATCH + (ML_MASK-1) + (bytesLeftForMl * 255); + assert(maxMlSize < INT_MAX); assert(ovml >= 0); + if ((size_t)ovml > maxMlSize) ovml = (int)maxMlSize; + if ((oend + LASTLITERALS) - (op + ll_totalCost + 2) - 1 + ovml >= MFLIMIT) { + DEBUGLOG(6, "Space to end : %i + ml (%i)", (int)((oend + LASTLITERALS) - (op + ll_totalCost + 2) - 1), ovml); + DEBUGLOG(6, "Before : ip = %p, anchor = %p", ip, anchor); + LZ4HC_encodeSequence(UPDATABLE(ip, op, anchor), ovml, ovref, notLimited, oend); + DEBUGLOG(6, "After : ip = %p, anchor = %p", ip, anchor); + } } + goto _last_literals; +} +_return_label: +#if defined(LZ4HC_HEAPMODE) && LZ4HC_HEAPMODE==1 + FREEMEM(opt); +#endif + return retval; +} + +} diff --git a/externals/tracy/public/common/tracy_lz4hc.hpp b/externals/tracy/public/common/tracy_lz4hc.hpp new file mode 100644 index 000000000..460cbae7f --- /dev/null +++ b/externals/tracy/public/common/tracy_lz4hc.hpp @@ -0,0 +1,405 @@ +/* + LZ4 HC - High Compression Mode of LZ4 + Header File + Copyright (C) 2011-2020, Yann Collet. + BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are + met: + + * Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above + copyright notice, this list of conditions and the following disclaimer + in the documentation and/or other materials provided with the + distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + You can contact the author at : + - LZ4 source repository : https://github.com/lz4/lz4 + - LZ4 public forum : https://groups.google.com/forum/#!forum/lz4c +*/ +#ifndef TRACY_LZ4_HC_H_19834876238432 +#define TRACY_LZ4_HC_H_19834876238432 + +/* --- Dependency --- */ +/* note : lz4hc requires lz4.h/lz4.c for compilation */ +#include "tracy_lz4.hpp" /* stddef, LZ4LIB_API, LZ4_DEPRECATED */ + + +/* --- Useful constants --- */ +#define LZ4HC_CLEVEL_MIN 3 +#define LZ4HC_CLEVEL_DEFAULT 9 +#define LZ4HC_CLEVEL_OPT_MIN 10 +#define LZ4HC_CLEVEL_MAX 12 + +namespace tracy +{ + +/*-************************************ + * Block Compression + **************************************/ +/*! LZ4_compress_HC() : + * Compress data from `src` into `dst`, using the powerful but slower "HC" algorithm. + * `dst` must be already allocated. + * Compression is guaranteed to succeed if `dstCapacity >= LZ4_compressBound(srcSize)` (see "lz4.h") + * Max supported `srcSize` value is LZ4_MAX_INPUT_SIZE (see "lz4.h") + * `compressionLevel` : any value between 1 and LZ4HC_CLEVEL_MAX will work. + * Values > LZ4HC_CLEVEL_MAX behave the same as LZ4HC_CLEVEL_MAX. + * @return : the number of bytes written into 'dst' + * or 0 if compression fails. + */ +LZ4LIB_API int LZ4_compress_HC (const char* src, char* dst, int srcSize, int dstCapacity, int compressionLevel); + + +/* Note : + * Decompression functions are provided within "lz4.h" (BSD license) + */ + + +/*! LZ4_compress_HC_extStateHC() : + * Same as LZ4_compress_HC(), but using an externally allocated memory segment for `state`. + * `state` size is provided by LZ4_sizeofStateHC(). + * Memory segment must be aligned on 8-bytes boundaries (which a normal malloc() should do properly). + */ +LZ4LIB_API int LZ4_sizeofStateHC(void); +LZ4LIB_API int LZ4_compress_HC_extStateHC(void* stateHC, const char* src, char* dst, int srcSize, int maxDstSize, int compressionLevel); + + +/*! LZ4_compress_HC_destSize() : v1.9.0+ + * Will compress as much data as possible from `src` + * to fit into `targetDstSize` budget. + * Result is provided in 2 parts : + * @return : the number of bytes written into 'dst' (necessarily <= targetDstSize) + * or 0 if compression fails. + * `srcSizePtr` : on success, *srcSizePtr is updated to indicate how much bytes were read from `src` + */ +LZ4LIB_API int LZ4_compress_HC_destSize(void* stateHC, + const char* src, char* dst, + int* srcSizePtr, int targetDstSize, + int compressionLevel); + + +/*-************************************ + * Streaming Compression + * Bufferless synchronous API + **************************************/ + typedef union LZ4_streamHC_u LZ4_streamHC_t; /* incomplete type (defined later) */ + +/*! LZ4_createStreamHC() and LZ4_freeStreamHC() : + * These functions create and release memory for LZ4 HC streaming state. + * Newly created states are automatically initialized. + * A same state can be used multiple times consecutively, + * starting with LZ4_resetStreamHC_fast() to start a new stream of blocks. + */ +LZ4LIB_API LZ4_streamHC_t* LZ4_createStreamHC(void); +LZ4LIB_API int LZ4_freeStreamHC (LZ4_streamHC_t* streamHCPtr); + +/* + These functions compress data in successive blocks of any size, + using previous blocks as dictionary, to improve compression ratio. + One key assumption is that previous blocks (up to 64 KB) remain read-accessible while compressing next blocks. + There is an exception for ring buffers, which can be smaller than 64 KB. + Ring-buffer scenario is automatically detected and handled within LZ4_compress_HC_continue(). + + Before starting compression, state must be allocated and properly initialized. + LZ4_createStreamHC() does both, though compression level is set to LZ4HC_CLEVEL_DEFAULT. + + Selecting the compression level can be done with LZ4_resetStreamHC_fast() (starts a new stream) + or LZ4_setCompressionLevel() (anytime, between blocks in the same stream) (experimental). + LZ4_resetStreamHC_fast() only works on states which have been properly initialized at least once, + which is automatically the case when state is created using LZ4_createStreamHC(). + + After reset, a first "fictional block" can be designated as initial dictionary, + using LZ4_loadDictHC() (Optional). + + Invoke LZ4_compress_HC_continue() to compress each successive block. + The number of blocks is unlimited. + Previous input blocks, including initial dictionary when present, + must remain accessible and unmodified during compression. + + It's allowed to update compression level anytime between blocks, + using LZ4_setCompressionLevel() (experimental). + + 'dst' buffer should be sized to handle worst case scenarios + (see LZ4_compressBound(), it ensures compression success). + In case of failure, the API does not guarantee recovery, + so the state _must_ be reset. + To ensure compression success + whenever `dst` buffer size cannot be made >= LZ4_compressBound(), + consider using LZ4_compress_HC_continue_destSize(). + + Whenever previous input blocks can't be preserved unmodified in-place during compression of next blocks, + it's possible to copy the last blocks into a more stable memory space, using LZ4_saveDictHC(). + Return value of LZ4_saveDictHC() is the size of dictionary effectively saved into 'safeBuffer' (<= 64 KB) + + After completing a streaming compression, + it's possible to start a new stream of blocks, using the same LZ4_streamHC_t state, + just by resetting it, using LZ4_resetStreamHC_fast(). +*/ + +LZ4LIB_API void LZ4_resetStreamHC_fast(LZ4_streamHC_t* streamHCPtr, int compressionLevel); /* v1.9.0+ */ +LZ4LIB_API int LZ4_loadDictHC (LZ4_streamHC_t* streamHCPtr, const char* dictionary, int dictSize); + +LZ4LIB_API int LZ4_compress_HC_continue (LZ4_streamHC_t* streamHCPtr, + const char* src, char* dst, + int srcSize, int maxDstSize); + +/*! LZ4_compress_HC_continue_destSize() : v1.9.0+ + * Similar to LZ4_compress_HC_continue(), + * but will read as much data as possible from `src` + * to fit into `targetDstSize` budget. + * Result is provided into 2 parts : + * @return : the number of bytes written into 'dst' (necessarily <= targetDstSize) + * or 0 if compression fails. + * `srcSizePtr` : on success, *srcSizePtr will be updated to indicate how much bytes were read from `src`. + * Note that this function may not consume the entire input. + */ +LZ4LIB_API int LZ4_compress_HC_continue_destSize(LZ4_streamHC_t* LZ4_streamHCPtr, + const char* src, char* dst, + int* srcSizePtr, int targetDstSize); + +LZ4LIB_API int LZ4_saveDictHC (LZ4_streamHC_t* streamHCPtr, char* safeBuffer, int maxDictSize); + + + +/*^********************************************** + * !!!!!! STATIC LINKING ONLY !!!!!! + ***********************************************/ + +/*-****************************************************************** + * PRIVATE DEFINITIONS : + * Do not use these definitions directly. + * They are merely exposed to allow static allocation of `LZ4_streamHC_t`. + * Declare an `LZ4_streamHC_t` directly, rather than any type below. + * Even then, only do so in the context of static linking, as definitions may change between versions. + ********************************************************************/ + +#define LZ4HC_DICTIONARY_LOGSIZE 16 +#define LZ4HC_MAXD (1<= LZ4HC_CLEVEL_OPT_MIN. + */ +LZ4LIB_STATIC_API void LZ4_favorDecompressionSpeed( + LZ4_streamHC_t* LZ4_streamHCPtr, int favor); + +/*! LZ4_resetStreamHC_fast() : v1.9.0+ + * When an LZ4_streamHC_t is known to be in a internally coherent state, + * it can often be prepared for a new compression with almost no work, only + * sometimes falling back to the full, expensive reset that is always required + * when the stream is in an indeterminate state (i.e., the reset performed by + * LZ4_resetStreamHC()). + * + * LZ4_streamHCs are guaranteed to be in a valid state when: + * - returned from LZ4_createStreamHC() + * - reset by LZ4_resetStreamHC() + * - memset(stream, 0, sizeof(LZ4_streamHC_t)) + * - the stream was in a valid state and was reset by LZ4_resetStreamHC_fast() + * - the stream was in a valid state and was then used in any compression call + * that returned success + * - the stream was in an indeterminate state and was used in a compression + * call that fully reset the state (LZ4_compress_HC_extStateHC()) and that + * returned success + * + * Note: + * A stream that was last used in a compression call that returned an error + * may be passed to this function. However, it will be fully reset, which will + * clear any existing history and settings from the context. + */ +LZ4LIB_STATIC_API void LZ4_resetStreamHC_fast( + LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel); + +/*! LZ4_compress_HC_extStateHC_fastReset() : + * A variant of LZ4_compress_HC_extStateHC(). + * + * Using this variant avoids an expensive initialization step. It is only safe + * to call if the state buffer is known to be correctly initialized already + * (see above comment on LZ4_resetStreamHC_fast() for a definition of + * "correctly initialized"). From a high level, the difference is that this + * function initializes the provided state with a call to + * LZ4_resetStreamHC_fast() while LZ4_compress_HC_extStateHC() starts with a + * call to LZ4_resetStreamHC(). + */ +LZ4LIB_STATIC_API int LZ4_compress_HC_extStateHC_fastReset ( + void* state, + const char* src, char* dst, + int srcSize, int dstCapacity, + int compressionLevel); + +/*! LZ4_attach_HC_dictionary() : + * This is an experimental API that allows for the efficient use of a + * static dictionary many times. + * + * Rather than re-loading the dictionary buffer into a working context before + * each compression, or copying a pre-loaded dictionary's LZ4_streamHC_t into a + * working LZ4_streamHC_t, this function introduces a no-copy setup mechanism, + * in which the working stream references the dictionary stream in-place. + * + * Several assumptions are made about the state of the dictionary stream. + * Currently, only streams which have been prepared by LZ4_loadDictHC() should + * be expected to work. + * + * Alternatively, the provided dictionary stream pointer may be NULL, in which + * case any existing dictionary stream is unset. + * + * A dictionary should only be attached to a stream without any history (i.e., + * a stream that has just been reset). + * + * The dictionary will remain attached to the working stream only for the + * current stream session. Calls to LZ4_resetStreamHC(_fast) will remove the + * dictionary context association from the working stream. The dictionary + * stream (and source buffer) must remain in-place / accessible / unchanged + * through the lifetime of the stream session. + */ +LZ4LIB_STATIC_API void LZ4_attach_HC_dictionary( + LZ4_streamHC_t *working_stream, + const LZ4_streamHC_t *dictionary_stream); + +} + +#endif /* LZ4_HC_SLO_098092834 */ +#endif /* LZ4_HC_STATIC_LINKING_ONLY */ diff --git a/externals/tracy/public/libbacktrace/LICENSE b/externals/tracy/public/libbacktrace/LICENSE new file mode 100644 index 000000000..097d2774e --- /dev/null +++ b/externals/tracy/public/libbacktrace/LICENSE @@ -0,0 +1,29 @@ +# Copyright (C) 2012-2016 Free Software Foundation, Inc. + +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are +# met: + +# (1) Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. + +# (2) Redistributions in binary form must reproduce the above copyright +# notice, this list of conditions and the following disclaimer in +# the documentation and/or other materials provided with the +# distribution. + +# (3) The name of the author may not be used to +# endorse or promote products derived from this software without +# specific prior written permission. + +# THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +# IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +# DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +# INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +# HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +# STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +# IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +# POSSIBILITY OF SUCH DAMAGE. diff --git a/externals/tracy/public/libbacktrace/alloc.cpp b/externals/tracy/public/libbacktrace/alloc.cpp new file mode 100644 index 000000000..a365a4860 --- /dev/null +++ b/externals/tracy/public/libbacktrace/alloc.cpp @@ -0,0 +1,174 @@ +/* alloc.c -- Memory allocation without mmap. + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#include "config.h" + +#include +#include +#include + +#include "backtrace.hpp" +#include "internal.hpp" + +#include "../common/TracyAlloc.hpp" + +namespace tracy +{ + +/* Allocation routines to use on systems that do not support anonymous + mmap. This implementation just uses malloc, which means that the + backtrace functions may not be safely invoked from a signal + handler. */ + +/* Allocate memory like malloc. If ERROR_CALLBACK is NULL, don't + report an error. */ + +void * +backtrace_alloc (struct backtrace_state *state ATTRIBUTE_UNUSED, + size_t size, backtrace_error_callback error_callback, + void *data) +{ + void *ret; + + ret = tracy_malloc (size); + if (ret == NULL) + { + if (error_callback) + error_callback (data, "malloc", errno); + } + return ret; +} + +/* Free memory. */ + +void +backtrace_free (struct backtrace_state *state ATTRIBUTE_UNUSED, + void *p, size_t size ATTRIBUTE_UNUSED, + backtrace_error_callback error_callback ATTRIBUTE_UNUSED, + void *data ATTRIBUTE_UNUSED) +{ + tracy_free (p); +} + +/* Grow VEC by SIZE bytes. */ + +void * +backtrace_vector_grow (struct backtrace_state *state ATTRIBUTE_UNUSED, + size_t size, backtrace_error_callback error_callback, + void *data, struct backtrace_vector *vec) +{ + void *ret; + + if (size > vec->alc) + { + size_t alc; + void *base; + + if (vec->size == 0) + alc = 32 * size; + else if (vec->size >= 4096) + alc = vec->size + 4096; + else + alc = 2 * vec->size; + + if (alc < vec->size + size) + alc = vec->size + size; + + base = tracy_realloc (vec->base, alc); + if (base == NULL) + { + error_callback (data, "realloc", errno); + return NULL; + } + + vec->base = base; + vec->alc = alc - vec->size; + } + + ret = (char *) vec->base + vec->size; + vec->size += size; + vec->alc -= size; + return ret; +} + +/* Finish the current allocation on VEC. */ + +void * +backtrace_vector_finish (struct backtrace_state *state, + struct backtrace_vector *vec, + backtrace_error_callback error_callback, + void *data) +{ + void *ret; + + /* With this allocator we call realloc in backtrace_vector_grow, + which means we can't easily reuse the memory here. So just + release it. */ + if (!backtrace_vector_release (state, vec, error_callback, data)) + return NULL; + ret = vec->base; + vec->base = NULL; + vec->size = 0; + vec->alc = 0; + return ret; +} + +/* Release any extra space allocated for VEC. */ + +int +backtrace_vector_release (struct backtrace_state *state ATTRIBUTE_UNUSED, + struct backtrace_vector *vec, + backtrace_error_callback error_callback, + void *data) +{ + vec->alc = 0; + + if (vec->size == 0) + { + /* As of C17, realloc with size 0 is marked as an obsolescent feature, use + free instead. */ + tracy_free (vec->base); + vec->base = NULL; + return 1; + } + + vec->base = tracy_realloc (vec->base, vec->size); + if (vec->base == NULL) + { + error_callback (data, "realloc", errno); + return 0; + } + + return 1; +} + +} diff --git a/externals/tracy/public/libbacktrace/backtrace.hpp b/externals/tracy/public/libbacktrace/backtrace.hpp new file mode 100644 index 000000000..e4be297a9 --- /dev/null +++ b/externals/tracy/public/libbacktrace/backtrace.hpp @@ -0,0 +1,186 @@ +/* backtrace.h -- Public header file for stack backtrace library. + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#ifndef BACKTRACE_H +#define BACKTRACE_H + +#include +#include +#include + +namespace tracy +{ + +/* The backtrace state. This struct is intentionally not defined in + the public interface. */ + +struct backtrace_state; + +/* The type of the error callback argument to backtrace functions. + This function, if not NULL, will be called for certain error cases. + The DATA argument is passed to the function that calls this one. + The MSG argument is an error message. The ERRNUM argument, if + greater than 0, holds an errno value. The MSG buffer may become + invalid after this function returns. + + As a special case, the ERRNUM argument will be passed as -1 if no + debug info can be found for the executable, or if the debug info + exists but has an unsupported version, but the function requires + debug info (e.g., backtrace_full, backtrace_pcinfo). The MSG in + this case will be something along the lines of "no debug info". + Similarly, ERRNUM will be passed as -1 if there is no symbol table, + but the function requires a symbol table (e.g., backtrace_syminfo). + This may be used as a signal that some other approach should be + tried. */ + +typedef void (*backtrace_error_callback) (void *data, const char *msg, + int errnum); + +/* Create state information for the backtrace routines. This must be + called before any of the other routines, and its return value must + be passed to all of the other routines. FILENAME is the path name + of the executable file; if it is NULL the library will try + system-specific path names. If not NULL, FILENAME must point to a + permanent buffer. If THREADED is non-zero the state may be + accessed by multiple threads simultaneously, and the library will + use appropriate atomic operations. If THREADED is zero the state + may only be accessed by one thread at a time. This returns a state + pointer on success, NULL on error. If an error occurs, this will + call the ERROR_CALLBACK routine. + + Calling this function allocates resources that cannot be freed. + There is no backtrace_free_state function. The state is used to + cache information that is expensive to recompute. Programs are + expected to call this function at most once and to save the return + value for all later calls to backtrace functions. */ + +extern struct backtrace_state *backtrace_create_state ( + const char *filename, int threaded, + backtrace_error_callback error_callback, void *data); + +/* The type of the callback argument to the backtrace_full function. + DATA is the argument passed to backtrace_full. PC is the program + counter. FILENAME is the name of the file containing PC, or NULL + if not available. LINENO is the line number in FILENAME containing + PC, or 0 if not available. FUNCTION is the name of the function + containing PC, or NULL if not available. This should return 0 to + continuing tracing. The FILENAME and FUNCTION buffers may become + invalid after this function returns. */ + +typedef int (*backtrace_full_callback) (void *data, uintptr_t pc, uintptr_t lowaddr, + const char *filename, int lineno, + const char *function); + +/* Get a full stack backtrace. SKIP is the number of frames to skip; + passing 0 will start the trace with the function calling + backtrace_full. DATA is passed to the callback routine. If any + call to CALLBACK returns a non-zero value, the stack backtrace + stops, and backtrace returns that value; this may be used to limit + the number of stack frames desired. If all calls to CALLBACK + return 0, backtrace returns 0. The backtrace_full function will + make at least one call to either CALLBACK or ERROR_CALLBACK. This + function requires debug info for the executable. */ + +extern int backtrace_full (struct backtrace_state *state, int skip, + backtrace_full_callback callback, + backtrace_error_callback error_callback, + void *data); + +/* The type of the callback argument to the backtrace_simple function. + DATA is the argument passed to simple_backtrace. PC is the program + counter. This should return 0 to continue tracing. */ + +typedef int (*backtrace_simple_callback) (void *data, uintptr_t pc); + +/* Get a simple backtrace. SKIP is the number of frames to skip, as + in backtrace. DATA is passed to the callback routine. If any call + to CALLBACK returns a non-zero value, the stack backtrace stops, + and backtrace_simple returns that value. Otherwise + backtrace_simple returns 0. The backtrace_simple function will + make at least one call to either CALLBACK or ERROR_CALLBACK. This + function does not require any debug info for the executable. */ + +extern int backtrace_simple (struct backtrace_state *state, int skip, + backtrace_simple_callback callback, + backtrace_error_callback error_callback, + void *data); + +/* Print the current backtrace in a user readable format to a FILE. + SKIP is the number of frames to skip, as in backtrace_full. Any + error messages are printed to stderr. This function requires debug + info for the executable. */ + +extern void backtrace_print (struct backtrace_state *state, int skip, FILE *); + +/* Given PC, a program counter in the current program, call the + callback function with filename, line number, and function name + information. This will normally call the callback function exactly + once. However, if the PC happens to describe an inlined call, and + the debugging information contains the necessary information, then + this may call the callback function multiple times. This will make + at least one call to either CALLBACK or ERROR_CALLBACK. This + returns the first non-zero value returned by CALLBACK, or 0. */ + +extern int backtrace_pcinfo (struct backtrace_state *state, uintptr_t pc, + backtrace_full_callback callback, + backtrace_error_callback error_callback, + void *data); + +/* The type of the callback argument to backtrace_syminfo. DATA and + PC are the arguments passed to backtrace_syminfo. SYMNAME is the + name of the symbol for the corresponding code. SYMVAL is the + value and SYMSIZE is the size of the symbol. SYMNAME will be NULL + if no error occurred but the symbol could not be found. */ + +typedef void (*backtrace_syminfo_callback) (void *data, uintptr_t pc, + const char *symname, + uintptr_t symval, + uintptr_t symsize); + +/* Given ADDR, an address or program counter in the current program, + call the callback information with the symbol name and value + describing the function or variable in which ADDR may be found. + This will call either CALLBACK or ERROR_CALLBACK exactly once. + This returns 1 on success, 0 on failure. This function requires + the symbol table but does not require the debug info. Note that if + the symbol table is present but ADDR could not be found in the + table, CALLBACK will be called with a NULL SYMNAME argument. + Returns 1 on success, 0 on error. */ + +extern int backtrace_syminfo (struct backtrace_state *state, uintptr_t addr, + backtrace_syminfo_callback callback, + backtrace_error_callback error_callback, + void *data); + +} + +#endif diff --git a/externals/tracy/public/libbacktrace/config.h b/externals/tracy/public/libbacktrace/config.h new file mode 100644 index 000000000..87e38a95b --- /dev/null +++ b/externals/tracy/public/libbacktrace/config.h @@ -0,0 +1,26 @@ +#include +#if defined(__linux__) && !defined(__GLIBC__) && !defined(__WORDSIZE) +// include __WORDSIZE headers for musl +# include +#endif +#if __WORDSIZE == 64 +# define BACKTRACE_ELF_SIZE 64 +#else +# define BACKTRACE_ELF_SIZE 32 +#endif + +#define HAVE_DLFCN_H 1 +#define HAVE_FCNTL 1 +#define HAVE_INTTYPES_H 1 +#define HAVE_LSTAT 1 +#define HAVE_READLINK 1 +#define HAVE_DL_ITERATE_PHDR 1 +#define HAVE_ATOMIC_FUNCTIONS 1 +#define HAVE_DECL_STRNLEN 1 + +#ifdef __APPLE__ +# define HAVE_MACH_O_DYLD_H 1 +#elif defined BSD +# define HAVE_KERN_PROC 1 +# define HAVE_KERN_PROC_ARGS 1 +#endif diff --git a/externals/tracy/public/libbacktrace/dwarf.cpp b/externals/tracy/public/libbacktrace/dwarf.cpp new file mode 100644 index 000000000..b6d681aa9 --- /dev/null +++ b/externals/tracy/public/libbacktrace/dwarf.cpp @@ -0,0 +1,4448 @@ +/* dwarf.c -- Get file/line information from DWARF for backtraces. + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#include "config.h" + +#include +#include +#include +#include + +#include "filenames.hpp" + +#include "backtrace.hpp" +#include "internal.hpp" + +namespace tracy +{ + +/* DWARF constants. */ + +enum dwarf_tag { + DW_TAG_entry_point = 0x3, + DW_TAG_compile_unit = 0x11, + DW_TAG_inlined_subroutine = 0x1d, + DW_TAG_subprogram = 0x2e, + DW_TAG_skeleton_unit = 0x4a, +}; + +enum dwarf_form { + DW_FORM_addr = 0x01, + DW_FORM_block2 = 0x03, + DW_FORM_block4 = 0x04, + DW_FORM_data2 = 0x05, + DW_FORM_data4 = 0x06, + DW_FORM_data8 = 0x07, + DW_FORM_string = 0x08, + DW_FORM_block = 0x09, + DW_FORM_block1 = 0x0a, + DW_FORM_data1 = 0x0b, + DW_FORM_flag = 0x0c, + DW_FORM_sdata = 0x0d, + DW_FORM_strp = 0x0e, + DW_FORM_udata = 0x0f, + DW_FORM_ref_addr = 0x10, + DW_FORM_ref1 = 0x11, + DW_FORM_ref2 = 0x12, + DW_FORM_ref4 = 0x13, + DW_FORM_ref8 = 0x14, + DW_FORM_ref_udata = 0x15, + DW_FORM_indirect = 0x16, + DW_FORM_sec_offset = 0x17, + DW_FORM_exprloc = 0x18, + DW_FORM_flag_present = 0x19, + DW_FORM_ref_sig8 = 0x20, + DW_FORM_strx = 0x1a, + DW_FORM_addrx = 0x1b, + DW_FORM_ref_sup4 = 0x1c, + DW_FORM_strp_sup = 0x1d, + DW_FORM_data16 = 0x1e, + DW_FORM_line_strp = 0x1f, + DW_FORM_implicit_const = 0x21, + DW_FORM_loclistx = 0x22, + DW_FORM_rnglistx = 0x23, + DW_FORM_ref_sup8 = 0x24, + DW_FORM_strx1 = 0x25, + DW_FORM_strx2 = 0x26, + DW_FORM_strx3 = 0x27, + DW_FORM_strx4 = 0x28, + DW_FORM_addrx1 = 0x29, + DW_FORM_addrx2 = 0x2a, + DW_FORM_addrx3 = 0x2b, + DW_FORM_addrx4 = 0x2c, + DW_FORM_GNU_addr_index = 0x1f01, + DW_FORM_GNU_str_index = 0x1f02, + DW_FORM_GNU_ref_alt = 0x1f20, + DW_FORM_GNU_strp_alt = 0x1f21 +}; + +enum dwarf_attribute { + DW_AT_sibling = 0x01, + DW_AT_location = 0x02, + DW_AT_name = 0x03, + DW_AT_ordering = 0x09, + DW_AT_subscr_data = 0x0a, + DW_AT_byte_size = 0x0b, + DW_AT_bit_offset = 0x0c, + DW_AT_bit_size = 0x0d, + DW_AT_element_list = 0x0f, + DW_AT_stmt_list = 0x10, + DW_AT_low_pc = 0x11, + DW_AT_high_pc = 0x12, + DW_AT_language = 0x13, + DW_AT_member = 0x14, + DW_AT_discr = 0x15, + DW_AT_discr_value = 0x16, + DW_AT_visibility = 0x17, + DW_AT_import = 0x18, + DW_AT_string_length = 0x19, + DW_AT_common_reference = 0x1a, + DW_AT_comp_dir = 0x1b, + DW_AT_const_value = 0x1c, + DW_AT_containing_type = 0x1d, + DW_AT_default_value = 0x1e, + DW_AT_inline = 0x20, + DW_AT_is_optional = 0x21, + DW_AT_lower_bound = 0x22, + DW_AT_producer = 0x25, + DW_AT_prototyped = 0x27, + DW_AT_return_addr = 0x2a, + DW_AT_start_scope = 0x2c, + DW_AT_bit_stride = 0x2e, + DW_AT_upper_bound = 0x2f, + DW_AT_abstract_origin = 0x31, + DW_AT_accessibility = 0x32, + DW_AT_address_class = 0x33, + DW_AT_artificial = 0x34, + DW_AT_base_types = 0x35, + DW_AT_calling_convention = 0x36, + DW_AT_count = 0x37, + DW_AT_data_member_location = 0x38, + DW_AT_decl_column = 0x39, + DW_AT_decl_file = 0x3a, + DW_AT_decl_line = 0x3b, + DW_AT_declaration = 0x3c, + DW_AT_discr_list = 0x3d, + DW_AT_encoding = 0x3e, + DW_AT_external = 0x3f, + DW_AT_frame_base = 0x40, + DW_AT_friend = 0x41, + DW_AT_identifier_case = 0x42, + DW_AT_macro_info = 0x43, + DW_AT_namelist_items = 0x44, + DW_AT_priority = 0x45, + DW_AT_segment = 0x46, + DW_AT_specification = 0x47, + DW_AT_static_link = 0x48, + DW_AT_type = 0x49, + DW_AT_use_location = 0x4a, + DW_AT_variable_parameter = 0x4b, + DW_AT_virtuality = 0x4c, + DW_AT_vtable_elem_location = 0x4d, + DW_AT_allocated = 0x4e, + DW_AT_associated = 0x4f, + DW_AT_data_location = 0x50, + DW_AT_byte_stride = 0x51, + DW_AT_entry_pc = 0x52, + DW_AT_use_UTF8 = 0x53, + DW_AT_extension = 0x54, + DW_AT_ranges = 0x55, + DW_AT_trampoline = 0x56, + DW_AT_call_column = 0x57, + DW_AT_call_file = 0x58, + DW_AT_call_line = 0x59, + DW_AT_description = 0x5a, + DW_AT_binary_scale = 0x5b, + DW_AT_decimal_scale = 0x5c, + DW_AT_small = 0x5d, + DW_AT_decimal_sign = 0x5e, + DW_AT_digit_count = 0x5f, + DW_AT_picture_string = 0x60, + DW_AT_mutable = 0x61, + DW_AT_threads_scaled = 0x62, + DW_AT_explicit = 0x63, + DW_AT_object_pointer = 0x64, + DW_AT_endianity = 0x65, + DW_AT_elemental = 0x66, + DW_AT_pure = 0x67, + DW_AT_recursive = 0x68, + DW_AT_signature = 0x69, + DW_AT_main_subprogram = 0x6a, + DW_AT_data_bit_offset = 0x6b, + DW_AT_const_expr = 0x6c, + DW_AT_enum_class = 0x6d, + DW_AT_linkage_name = 0x6e, + DW_AT_string_length_bit_size = 0x6f, + DW_AT_string_length_byte_size = 0x70, + DW_AT_rank = 0x71, + DW_AT_str_offsets_base = 0x72, + DW_AT_addr_base = 0x73, + DW_AT_rnglists_base = 0x74, + DW_AT_dwo_name = 0x76, + DW_AT_reference = 0x77, + DW_AT_rvalue_reference = 0x78, + DW_AT_macros = 0x79, + DW_AT_call_all_calls = 0x7a, + DW_AT_call_all_source_calls = 0x7b, + DW_AT_call_all_tail_calls = 0x7c, + DW_AT_call_return_pc = 0x7d, + DW_AT_call_value = 0x7e, + DW_AT_call_origin = 0x7f, + DW_AT_call_parameter = 0x80, + DW_AT_call_pc = 0x81, + DW_AT_call_tail_call = 0x82, + DW_AT_call_target = 0x83, + DW_AT_call_target_clobbered = 0x84, + DW_AT_call_data_location = 0x85, + DW_AT_call_data_value = 0x86, + DW_AT_noreturn = 0x87, + DW_AT_alignment = 0x88, + DW_AT_export_symbols = 0x89, + DW_AT_deleted = 0x8a, + DW_AT_defaulted = 0x8b, + DW_AT_loclists_base = 0x8c, + DW_AT_lo_user = 0x2000, + DW_AT_hi_user = 0x3fff, + DW_AT_MIPS_fde = 0x2001, + DW_AT_MIPS_loop_begin = 0x2002, + DW_AT_MIPS_tail_loop_begin = 0x2003, + DW_AT_MIPS_epilog_begin = 0x2004, + DW_AT_MIPS_loop_unroll_factor = 0x2005, + DW_AT_MIPS_software_pipeline_depth = 0x2006, + DW_AT_MIPS_linkage_name = 0x2007, + DW_AT_MIPS_stride = 0x2008, + DW_AT_MIPS_abstract_name = 0x2009, + DW_AT_MIPS_clone_origin = 0x200a, + DW_AT_MIPS_has_inlines = 0x200b, + DW_AT_HP_block_index = 0x2000, + DW_AT_HP_unmodifiable = 0x2001, + DW_AT_HP_prologue = 0x2005, + DW_AT_HP_epilogue = 0x2008, + DW_AT_HP_actuals_stmt_list = 0x2010, + DW_AT_HP_proc_per_section = 0x2011, + DW_AT_HP_raw_data_ptr = 0x2012, + DW_AT_HP_pass_by_reference = 0x2013, + DW_AT_HP_opt_level = 0x2014, + DW_AT_HP_prof_version_id = 0x2015, + DW_AT_HP_opt_flags = 0x2016, + DW_AT_HP_cold_region_low_pc = 0x2017, + DW_AT_HP_cold_region_high_pc = 0x2018, + DW_AT_HP_all_variables_modifiable = 0x2019, + DW_AT_HP_linkage_name = 0x201a, + DW_AT_HP_prof_flags = 0x201b, + DW_AT_HP_unit_name = 0x201f, + DW_AT_HP_unit_size = 0x2020, + DW_AT_HP_widened_byte_size = 0x2021, + DW_AT_HP_definition_points = 0x2022, + DW_AT_HP_default_location = 0x2023, + DW_AT_HP_is_result_param = 0x2029, + DW_AT_sf_names = 0x2101, + DW_AT_src_info = 0x2102, + DW_AT_mac_info = 0x2103, + DW_AT_src_coords = 0x2104, + DW_AT_body_begin = 0x2105, + DW_AT_body_end = 0x2106, + DW_AT_GNU_vector = 0x2107, + DW_AT_GNU_guarded_by = 0x2108, + DW_AT_GNU_pt_guarded_by = 0x2109, + DW_AT_GNU_guarded = 0x210a, + DW_AT_GNU_pt_guarded = 0x210b, + DW_AT_GNU_locks_excluded = 0x210c, + DW_AT_GNU_exclusive_locks_required = 0x210d, + DW_AT_GNU_shared_locks_required = 0x210e, + DW_AT_GNU_odr_signature = 0x210f, + DW_AT_GNU_template_name = 0x2110, + DW_AT_GNU_call_site_value = 0x2111, + DW_AT_GNU_call_site_data_value = 0x2112, + DW_AT_GNU_call_site_target = 0x2113, + DW_AT_GNU_call_site_target_clobbered = 0x2114, + DW_AT_GNU_tail_call = 0x2115, + DW_AT_GNU_all_tail_call_sites = 0x2116, + DW_AT_GNU_all_call_sites = 0x2117, + DW_AT_GNU_all_source_call_sites = 0x2118, + DW_AT_GNU_macros = 0x2119, + DW_AT_GNU_deleted = 0x211a, + DW_AT_GNU_dwo_name = 0x2130, + DW_AT_GNU_dwo_id = 0x2131, + DW_AT_GNU_ranges_base = 0x2132, + DW_AT_GNU_addr_base = 0x2133, + DW_AT_GNU_pubnames = 0x2134, + DW_AT_GNU_pubtypes = 0x2135, + DW_AT_GNU_discriminator = 0x2136, + DW_AT_GNU_locviews = 0x2137, + DW_AT_GNU_entry_view = 0x2138, + DW_AT_VMS_rtnbeg_pd_address = 0x2201, + DW_AT_use_GNAT_descriptive_type = 0x2301, + DW_AT_GNAT_descriptive_type = 0x2302, + DW_AT_GNU_numerator = 0x2303, + DW_AT_GNU_denominator = 0x2304, + DW_AT_GNU_bias = 0x2305, + DW_AT_upc_threads_scaled = 0x3210, + DW_AT_PGI_lbase = 0x3a00, + DW_AT_PGI_soffset = 0x3a01, + DW_AT_PGI_lstride = 0x3a02, + DW_AT_APPLE_optimized = 0x3fe1, + DW_AT_APPLE_flags = 0x3fe2, + DW_AT_APPLE_isa = 0x3fe3, + DW_AT_APPLE_block = 0x3fe4, + DW_AT_APPLE_major_runtime_vers = 0x3fe5, + DW_AT_APPLE_runtime_class = 0x3fe6, + DW_AT_APPLE_omit_frame_ptr = 0x3fe7, + DW_AT_APPLE_property_name = 0x3fe8, + DW_AT_APPLE_property_getter = 0x3fe9, + DW_AT_APPLE_property_setter = 0x3fea, + DW_AT_APPLE_property_attribute = 0x3feb, + DW_AT_APPLE_objc_complete_type = 0x3fec, + DW_AT_APPLE_property = 0x3fed +}; + +enum dwarf_line_number_op { + DW_LNS_extended_op = 0x0, + DW_LNS_copy = 0x1, + DW_LNS_advance_pc = 0x2, + DW_LNS_advance_line = 0x3, + DW_LNS_set_file = 0x4, + DW_LNS_set_column = 0x5, + DW_LNS_negate_stmt = 0x6, + DW_LNS_set_basic_block = 0x7, + DW_LNS_const_add_pc = 0x8, + DW_LNS_fixed_advance_pc = 0x9, + DW_LNS_set_prologue_end = 0xa, + DW_LNS_set_epilogue_begin = 0xb, + DW_LNS_set_isa = 0xc, +}; + +enum dwarf_extended_line_number_op { + DW_LNE_end_sequence = 0x1, + DW_LNE_set_address = 0x2, + DW_LNE_define_file = 0x3, + DW_LNE_set_discriminator = 0x4, +}; + +enum dwarf_line_number_content_type { + DW_LNCT_path = 0x1, + DW_LNCT_directory_index = 0x2, + DW_LNCT_timestamp = 0x3, + DW_LNCT_size = 0x4, + DW_LNCT_MD5 = 0x5, + DW_LNCT_lo_user = 0x2000, + DW_LNCT_hi_user = 0x3fff +}; + +enum dwarf_range_list_entry { + DW_RLE_end_of_list = 0x00, + DW_RLE_base_addressx = 0x01, + DW_RLE_startx_endx = 0x02, + DW_RLE_startx_length = 0x03, + DW_RLE_offset_pair = 0x04, + DW_RLE_base_address = 0x05, + DW_RLE_start_end = 0x06, + DW_RLE_start_length = 0x07 +}; + +enum dwarf_unit_type { + DW_UT_compile = 0x01, + DW_UT_type = 0x02, + DW_UT_partial = 0x03, + DW_UT_skeleton = 0x04, + DW_UT_split_compile = 0x05, + DW_UT_split_type = 0x06, + DW_UT_lo_user = 0x80, + DW_UT_hi_user = 0xff +}; + +#if !defined(HAVE_DECL_STRNLEN) || !HAVE_DECL_STRNLEN + +/* If strnlen is not declared, provide our own version. */ + +static size_t +xstrnlen (const char *s, size_t maxlen) +{ + size_t i; + + for (i = 0; i < maxlen; ++i) + if (s[i] == '\0') + break; + return i; +} + +#define strnlen xstrnlen + +#endif + +/* A buffer to read DWARF info. */ + +struct dwarf_buf +{ + /* Buffer name for error messages. */ + const char *name; + /* Start of the buffer. */ + const unsigned char *start; + /* Next byte to read. */ + const unsigned char *buf; + /* The number of bytes remaining. */ + size_t left; + /* Whether the data is big-endian. */ + int is_bigendian; + /* Error callback routine. */ + backtrace_error_callback error_callback; + /* Data for error_callback. */ + void *data; + /* Non-zero if we've reported an underflow error. */ + int reported_underflow; +}; + +/* A single attribute in a DWARF abbreviation. */ + +struct attr +{ + /* The attribute name. */ + enum dwarf_attribute name; + /* The attribute form. */ + enum dwarf_form form; + /* The attribute value, for DW_FORM_implicit_const. */ + int64_t val; +}; + +/* A single DWARF abbreviation. */ + +struct abbrev +{ + /* The abbrev code--the number used to refer to the abbrev. */ + uint64_t code; + /* The entry tag. */ + enum dwarf_tag tag; + /* Non-zero if this abbrev has child entries. */ + int has_children; + /* The number of attributes. */ + size_t num_attrs; + /* The attributes. */ + struct attr *attrs; +}; + +/* The DWARF abbreviations for a compilation unit. This structure + only exists while reading the compilation unit. Most DWARF readers + seem to a hash table to map abbrev ID's to abbrev entries. + However, we primarily care about GCC, and GCC simply issues ID's in + numerical order starting at 1. So we simply keep a sorted vector, + and try to just look up the code. */ + +struct abbrevs +{ + /* The number of abbrevs in the vector. */ + size_t num_abbrevs; + /* The abbrevs, sorted by the code field. */ + struct abbrev *abbrevs; +}; + +/* The different kinds of attribute values. */ + +enum attr_val_encoding +{ + /* No attribute value. */ + ATTR_VAL_NONE, + /* An address. */ + ATTR_VAL_ADDRESS, + /* An index into the .debug_addr section, whose value is relative to + the DW_AT_addr_base attribute of the compilation unit. */ + ATTR_VAL_ADDRESS_INDEX, + /* A unsigned integer. */ + ATTR_VAL_UINT, + /* A sigd integer. */ + ATTR_VAL_SINT, + /* A string. */ + ATTR_VAL_STRING, + /* An index into the .debug_str_offsets section. */ + ATTR_VAL_STRING_INDEX, + /* An offset to other data in the containing unit. */ + ATTR_VAL_REF_UNIT, + /* An offset to other data within the .debug_info section. */ + ATTR_VAL_REF_INFO, + /* An offset to other data within the alt .debug_info section. */ + ATTR_VAL_REF_ALT_INFO, + /* An offset to data in some other section. */ + ATTR_VAL_REF_SECTION, + /* A type signature. */ + ATTR_VAL_REF_TYPE, + /* An index into the .debug_rnglists section. */ + ATTR_VAL_RNGLISTS_INDEX, + /* A block of data (not represented). */ + ATTR_VAL_BLOCK, + /* An expression (not represented). */ + ATTR_VAL_EXPR, +}; + +/* An attribute value. */ + +struct attr_val +{ + /* How the value is stored in the field u. */ + enum attr_val_encoding encoding; + union + { + /* ATTR_VAL_ADDRESS*, ATTR_VAL_UINT, ATTR_VAL_REF*. */ + uint64_t uint; + /* ATTR_VAL_SINT. */ + int64_t sint; + /* ATTR_VAL_STRING. */ + const char *string; + /* ATTR_VAL_BLOCK not stored. */ + } u; +}; + +/* The line number program header. */ + +struct line_header +{ + /* The version of the line number information. */ + int version; + /* Address size. */ + int addrsize; + /* The minimum instruction length. */ + unsigned int min_insn_len; + /* The maximum number of ops per instruction. */ + unsigned int max_ops_per_insn; + /* The line base for special opcodes. */ + int line_base; + /* The line range for special opcodes. */ + unsigned int line_range; + /* The opcode base--the first special opcode. */ + unsigned int opcode_base; + /* Opcode lengths, indexed by opcode - 1. */ + const unsigned char *opcode_lengths; + /* The number of directory entries. */ + size_t dirs_count; + /* The directory entries. */ + const char **dirs; + /* The number of filenames. */ + size_t filenames_count; + /* The filenames. */ + const char **filenames; +}; + +/* A format description from a line header. */ + +struct line_header_format +{ + int lnct; /* LNCT code. */ + enum dwarf_form form; /* Form of entry data. */ +}; + +/* Map a single PC value to a file/line. We will keep a vector of + these sorted by PC value. Each file/line will be correct from the + PC up to the PC of the next entry if there is one. We allocate one + extra entry at the end so that we can use bsearch. */ + +struct line +{ + /* PC. */ + uintptr_t pc; + /* File name. Many entries in the array are expected to point to + the same file name. */ + const char *filename; + /* Line number. */ + int lineno; + /* Index of the object in the original array read from the DWARF + section, before it has been sorted. The index makes it possible + to use Quicksort and maintain stability. */ + int idx; +}; + +/* A growable vector of line number information. This is used while + reading the line numbers. */ + +struct line_vector +{ + /* Memory. This is an array of struct line. */ + struct backtrace_vector vec; + /* Number of valid mappings. */ + size_t count; +}; + +/* A function described in the debug info. */ + +struct function +{ + /* The name of the function. */ + const char *name; + /* If this is an inlined function, the filename of the call + site. */ + const char *caller_filename; + /* If this is an inlined function, the line number of the call + site. */ + int caller_lineno; + /* Map PC ranges to inlined functions. */ + struct function_addrs *function_addrs; + size_t function_addrs_count; +}; + +/* An address range for a function. This maps a PC value to a + specific function. */ + +struct function_addrs +{ + /* Range is LOW <= PC < HIGH. */ + uintptr_t low; + uintptr_t high; + /* Function for this address range. */ + struct function *function; +}; + +/* A growable vector of function address ranges. */ + +struct function_vector +{ + /* Memory. This is an array of struct function_addrs. */ + struct backtrace_vector vec; + /* Number of address ranges present. */ + size_t count; +}; + +/* A DWARF compilation unit. This only holds the information we need + to map a PC to a file and line. */ + +struct unit +{ + /* The first entry for this compilation unit. */ + const unsigned char *unit_data; + /* The length of the data for this compilation unit. */ + size_t unit_data_len; + /* The offset of UNIT_DATA from the start of the information for + this compilation unit. */ + size_t unit_data_offset; + /* Offset of the start of the compilation unit from the start of the + .debug_info section. */ + size_t low_offset; + /* Offset of the end of the compilation unit from the start of the + .debug_info section. */ + size_t high_offset; + /* DWARF version. */ + int version; + /* Whether unit is DWARF64. */ + int is_dwarf64; + /* Address size. */ + int addrsize; + /* Offset into line number information. */ + off_t lineoff; + /* Offset of compilation unit in .debug_str_offsets. */ + uint64_t str_offsets_base; + /* Offset of compilation unit in .debug_addr. */ + uint64_t addr_base; + /* Offset of compilation unit in .debug_rnglists. */ + uint64_t rnglists_base; + /* Primary source file. */ + const char *filename; + /* Compilation command working directory. */ + const char *comp_dir; + /* Absolute file name, only set if needed. */ + const char *abs_filename; + /* The abbreviations for this unit. */ + struct abbrevs abbrevs; + + /* The fields above this point are read in during initialization and + may be accessed freely. The fields below this point are read in + as needed, and therefore require care, as different threads may + try to initialize them simultaneously. */ + + /* PC to line number mapping. This is NULL if the values have not + been read. This is (struct line *) -1 if there was an error + reading the values. */ + struct line *lines; + /* Number of entries in lines. */ + size_t lines_count; + /* PC ranges to function. */ + struct function_addrs *function_addrs; + size_t function_addrs_count; +}; + +/* An address range for a compilation unit. This maps a PC value to a + specific compilation unit. Note that we invert the representation + in DWARF: instead of listing the units and attaching a list of + ranges, we list the ranges and have each one point to the unit. + This lets us do a binary search to find the unit. */ + +struct unit_addrs +{ + /* Range is LOW <= PC < HIGH. */ + uintptr_t low; + uintptr_t high; + /* Compilation unit for this address range. */ + struct unit *u; +}; + +/* A growable vector of compilation unit address ranges. */ + +struct unit_addrs_vector +{ + /* Memory. This is an array of struct unit_addrs. */ + struct backtrace_vector vec; + /* Number of address ranges present. */ + size_t count; +}; + +/* A growable vector of compilation unit pointer. */ + +struct unit_vector +{ + struct backtrace_vector vec; + size_t count; +}; + +/* The information we need to map a PC to a file and line. */ + +struct dwarf_data +{ + /* The data for the next file we know about. */ + struct dwarf_data *next; + /* The data for .gnu_debugaltlink. */ + struct dwarf_data *altlink; + /* The base address for this file. */ + uintptr_t base_address; + /* A sorted list of address ranges. */ + struct unit_addrs *addrs; + /* Number of address ranges in list. */ + size_t addrs_count; + /* A sorted list of units. */ + struct unit **units; + /* Number of units in the list. */ + size_t units_count; + /* The unparsed DWARF debug data. */ + struct dwarf_sections dwarf_sections; + /* Whether the data is big-endian or not. */ + int is_bigendian; + /* A vector used for function addresses. We keep this here so that + we can grow the vector as we read more functions. */ + struct function_vector fvec; +}; + +/* Report an error for a DWARF buffer. */ + +static void +dwarf_buf_error (struct dwarf_buf *buf, const char *msg, int errnum) +{ + char b[200]; + + snprintf (b, sizeof b, "%s in %s at %d", + msg, buf->name, (int) (buf->buf - buf->start)); + buf->error_callback (buf->data, b, errnum); +} + +/* Require at least COUNT bytes in BUF. Return 1 if all is well, 0 on + error. */ + +static int +require (struct dwarf_buf *buf, size_t count) +{ + if (buf->left >= count) + return 1; + + if (!buf->reported_underflow) + { + dwarf_buf_error (buf, "DWARF underflow", 0); + buf->reported_underflow = 1; + } + + return 0; +} + +/* Advance COUNT bytes in BUF. Return 1 if all is well, 0 on + error. */ + +static int +advance (struct dwarf_buf *buf, size_t count) +{ + if (!require (buf, count)) + return 0; + buf->buf += count; + buf->left -= count; + return 1; +} + +/* Read one zero-terminated string from BUF and advance past the string. */ + +static const char * +read_string (struct dwarf_buf *buf) +{ + const char *p = (const char *)buf->buf; + size_t len = strnlen (p, buf->left); + + /* - If len == left, we ran out of buffer before finding the zero terminator. + Generate an error by advancing len + 1. + - If len < left, advance by len + 1 to skip past the zero terminator. */ + size_t count = len + 1; + + if (!advance (buf, count)) + return NULL; + + return p; +} + +/* Read one byte from BUF and advance 1 byte. */ + +static unsigned char +read_byte (struct dwarf_buf *buf) +{ + const unsigned char *p = buf->buf; + + if (!advance (buf, 1)) + return 0; + return p[0]; +} + +/* Read a signed char from BUF and advance 1 byte. */ + +static signed char +read_sbyte (struct dwarf_buf *buf) +{ + const unsigned char *p = buf->buf; + + if (!advance (buf, 1)) + return 0; + return (*p ^ 0x80) - 0x80; +} + +/* Read a uint16 from BUF and advance 2 bytes. */ + +static uint16_t +read_uint16 (struct dwarf_buf *buf) +{ + const unsigned char *p = buf->buf; + + if (!advance (buf, 2)) + return 0; + if (buf->is_bigendian) + return ((uint16_t) p[0] << 8) | (uint16_t) p[1]; + else + return ((uint16_t) p[1] << 8) | (uint16_t) p[0]; +} + +/* Read a 24 bit value from BUF and advance 3 bytes. */ + +static uint32_t +read_uint24 (struct dwarf_buf *buf) +{ + const unsigned char *p = buf->buf; + + if (!advance (buf, 3)) + return 0; + if (buf->is_bigendian) + return (((uint32_t) p[0] << 16) | ((uint32_t) p[1] << 8) + | (uint32_t) p[2]); + else + return (((uint32_t) p[2] << 16) | ((uint32_t) p[1] << 8) + | (uint32_t) p[0]); +} + +/* Read a uint32 from BUF and advance 4 bytes. */ + +static uint32_t +read_uint32 (struct dwarf_buf *buf) +{ + const unsigned char *p = buf->buf; + + if (!advance (buf, 4)) + return 0; + if (buf->is_bigendian) + return (((uint32_t) p[0] << 24) | ((uint32_t) p[1] << 16) + | ((uint32_t) p[2] << 8) | (uint32_t) p[3]); + else + return (((uint32_t) p[3] << 24) | ((uint32_t) p[2] << 16) + | ((uint32_t) p[1] << 8) | (uint32_t) p[0]); +} + +/* Read a uint64 from BUF and advance 8 bytes. */ + +static uint64_t +read_uint64 (struct dwarf_buf *buf) +{ + const unsigned char *p = buf->buf; + + if (!advance (buf, 8)) + return 0; + if (buf->is_bigendian) + return (((uint64_t) p[0] << 56) | ((uint64_t) p[1] << 48) + | ((uint64_t) p[2] << 40) | ((uint64_t) p[3] << 32) + | ((uint64_t) p[4] << 24) | ((uint64_t) p[5] << 16) + | ((uint64_t) p[6] << 8) | (uint64_t) p[7]); + else + return (((uint64_t) p[7] << 56) | ((uint64_t) p[6] << 48) + | ((uint64_t) p[5] << 40) | ((uint64_t) p[4] << 32) + | ((uint64_t) p[3] << 24) | ((uint64_t) p[2] << 16) + | ((uint64_t) p[1] << 8) | (uint64_t) p[0]); +} + +/* Read an offset from BUF and advance the appropriate number of + bytes. */ + +static uint64_t +read_offset (struct dwarf_buf *buf, int is_dwarf64) +{ + if (is_dwarf64) + return read_uint64 (buf); + else + return read_uint32 (buf); +} + +/* Read an address from BUF and advance the appropriate number of + bytes. */ + +static uint64_t +read_address (struct dwarf_buf *buf, int addrsize) +{ + switch (addrsize) + { + case 1: + return read_byte (buf); + case 2: + return read_uint16 (buf); + case 4: + return read_uint32 (buf); + case 8: + return read_uint64 (buf); + default: + dwarf_buf_error (buf, "unrecognized address size", 0); + return 0; + } +} + +/* Return whether a value is the highest possible address, given the + address size. */ + +static int +is_highest_address (uint64_t address, int addrsize) +{ + switch (addrsize) + { + case 1: + return address == (unsigned char) -1; + case 2: + return address == (uint16_t) -1; + case 4: + return address == (uint32_t) -1; + case 8: + return address == (uint64_t) -1; + default: + return 0; + } +} + +/* Read an unsigned LEB128 number. */ + +static uint64_t +read_uleb128 (struct dwarf_buf *buf) +{ + uint64_t ret; + unsigned int shift; + int overflow; + unsigned char b; + + ret = 0; + shift = 0; + overflow = 0; + do + { + const unsigned char *p; + + p = buf->buf; + if (!advance (buf, 1)) + return 0; + b = *p; + if (shift < 64) + ret |= ((uint64_t) (b & 0x7f)) << shift; + else if (!overflow) + { + dwarf_buf_error (buf, "LEB128 overflows uint64_t", 0); + overflow = 1; + } + shift += 7; + } + while ((b & 0x80) != 0); + + return ret; +} + +/* Read a signed LEB128 number. */ + +static int64_t +read_sleb128 (struct dwarf_buf *buf) +{ + uint64_t val; + unsigned int shift; + int overflow; + unsigned char b; + + val = 0; + shift = 0; + overflow = 0; + do + { + const unsigned char *p; + + p = buf->buf; + if (!advance (buf, 1)) + return 0; + b = *p; + if (shift < 64) + val |= ((uint64_t) (b & 0x7f)) << shift; + else if (!overflow) + { + dwarf_buf_error (buf, "signed LEB128 overflows uint64_t", 0); + overflow = 1; + } + shift += 7; + } + while ((b & 0x80) != 0); + + if ((b & 0x40) != 0 && shift < 64) + val |= ((uint64_t) -1) << shift; + + return (int64_t) val; +} + +/* Return the length of an LEB128 number. */ + +static size_t +leb128_len (const unsigned char *p) +{ + size_t ret; + + ret = 1; + while ((*p & 0x80) != 0) + { + ++p; + ++ret; + } + return ret; +} + +/* Read initial_length from BUF and advance the appropriate number of bytes. */ + +static uint64_t +read_initial_length (struct dwarf_buf *buf, int *is_dwarf64) +{ + uint64_t len; + + len = read_uint32 (buf); + if (len == 0xffffffff) + { + len = read_uint64 (buf); + *is_dwarf64 = 1; + } + else + *is_dwarf64 = 0; + + return len; +} + +/* Free an abbreviations structure. */ + +static void +free_abbrevs (struct backtrace_state *state, struct abbrevs *abbrevs, + backtrace_error_callback error_callback, void *data) +{ + size_t i; + + for (i = 0; i < abbrevs->num_abbrevs; ++i) + backtrace_free (state, abbrevs->abbrevs[i].attrs, + abbrevs->abbrevs[i].num_attrs * sizeof (struct attr), + error_callback, data); + backtrace_free (state, abbrevs->abbrevs, + abbrevs->num_abbrevs * sizeof (struct abbrev), + error_callback, data); + abbrevs->num_abbrevs = 0; + abbrevs->abbrevs = NULL; +} + +/* Read an attribute value. Returns 1 on success, 0 on failure. If + the value can be represented as a uint64_t, sets *VAL and sets + *IS_VALID to 1. We don't try to store the value of other attribute + forms, because we don't care about them. */ + +static int +read_attribute (enum dwarf_form form, uint64_t implicit_val, + struct dwarf_buf *buf, int is_dwarf64, int version, + int addrsize, const struct dwarf_sections *dwarf_sections, + struct dwarf_data *altlink, struct attr_val *val) +{ + /* Avoid warnings about val.u.FIELD may be used uninitialized if + this function is inlined. The warnings aren't valid but can + occur because the different fields are set and used + conditionally. */ + memset (val, 0, sizeof *val); + + switch (form) + { + case DW_FORM_addr: + val->encoding = ATTR_VAL_ADDRESS; + val->u.uint = read_address (buf, addrsize); + return 1; + case DW_FORM_block2: + val->encoding = ATTR_VAL_BLOCK; + return advance (buf, read_uint16 (buf)); + case DW_FORM_block4: + val->encoding = ATTR_VAL_BLOCK; + return advance (buf, read_uint32 (buf)); + case DW_FORM_data2: + val->encoding = ATTR_VAL_UINT; + val->u.uint = read_uint16 (buf); + return 1; + case DW_FORM_data4: + val->encoding = ATTR_VAL_UINT; + val->u.uint = read_uint32 (buf); + return 1; + case DW_FORM_data8: + val->encoding = ATTR_VAL_UINT; + val->u.uint = read_uint64 (buf); + return 1; + case DW_FORM_data16: + val->encoding = ATTR_VAL_BLOCK; + return advance (buf, 16); + case DW_FORM_string: + val->encoding = ATTR_VAL_STRING; + val->u.string = read_string (buf); + return val->u.string == NULL ? 0 : 1; + case DW_FORM_block: + val->encoding = ATTR_VAL_BLOCK; + return advance (buf, read_uleb128 (buf)); + case DW_FORM_block1: + val->encoding = ATTR_VAL_BLOCK; + return advance (buf, read_byte (buf)); + case DW_FORM_data1: + val->encoding = ATTR_VAL_UINT; + val->u.uint = read_byte (buf); + return 1; + case DW_FORM_flag: + val->encoding = ATTR_VAL_UINT; + val->u.uint = read_byte (buf); + return 1; + case DW_FORM_sdata: + val->encoding = ATTR_VAL_SINT; + val->u.sint = read_sleb128 (buf); + return 1; + case DW_FORM_strp: + { + uint64_t offset; + + offset = read_offset (buf, is_dwarf64); + if (offset >= dwarf_sections->size[DEBUG_STR]) + { + dwarf_buf_error (buf, "DW_FORM_strp out of range", 0); + return 0; + } + val->encoding = ATTR_VAL_STRING; + val->u.string = + (const char *) dwarf_sections->data[DEBUG_STR] + offset; + return 1; + } + case DW_FORM_line_strp: + { + uint64_t offset; + + offset = read_offset (buf, is_dwarf64); + if (offset >= dwarf_sections->size[DEBUG_LINE_STR]) + { + dwarf_buf_error (buf, "DW_FORM_line_strp out of range", 0); + return 0; + } + val->encoding = ATTR_VAL_STRING; + val->u.string = + (const char *) dwarf_sections->data[DEBUG_LINE_STR] + offset; + return 1; + } + case DW_FORM_udata: + val->encoding = ATTR_VAL_UINT; + val->u.uint = read_uleb128 (buf); + return 1; + case DW_FORM_ref_addr: + val->encoding = ATTR_VAL_REF_INFO; + if (version == 2) + val->u.uint = read_address (buf, addrsize); + else + val->u.uint = read_offset (buf, is_dwarf64); + return 1; + case DW_FORM_ref1: + val->encoding = ATTR_VAL_REF_UNIT; + val->u.uint = read_byte (buf); + return 1; + case DW_FORM_ref2: + val->encoding = ATTR_VAL_REF_UNIT; + val->u.uint = read_uint16 (buf); + return 1; + case DW_FORM_ref4: + val->encoding = ATTR_VAL_REF_UNIT; + val->u.uint = read_uint32 (buf); + return 1; + case DW_FORM_ref8: + val->encoding = ATTR_VAL_REF_UNIT; + val->u.uint = read_uint64 (buf); + return 1; + case DW_FORM_ref_udata: + val->encoding = ATTR_VAL_REF_UNIT; + val->u.uint = read_uleb128 (buf); + return 1; + case DW_FORM_indirect: + { + uint64_t form; + + form = read_uleb128 (buf); + if (form == DW_FORM_implicit_const) + { + dwarf_buf_error (buf, + "DW_FORM_indirect to DW_FORM_implicit_const", + 0); + return 0; + } + return read_attribute ((enum dwarf_form) form, 0, buf, is_dwarf64, + version, addrsize, dwarf_sections, altlink, + val); + } + case DW_FORM_sec_offset: + val->encoding = ATTR_VAL_REF_SECTION; + val->u.uint = read_offset (buf, is_dwarf64); + return 1; + case DW_FORM_exprloc: + val->encoding = ATTR_VAL_EXPR; + return advance (buf, read_uleb128 (buf)); + case DW_FORM_flag_present: + val->encoding = ATTR_VAL_UINT; + val->u.uint = 1; + return 1; + case DW_FORM_ref_sig8: + val->encoding = ATTR_VAL_REF_TYPE; + val->u.uint = read_uint64 (buf); + return 1; + case DW_FORM_strx: case DW_FORM_strx1: case DW_FORM_strx2: + case DW_FORM_strx3: case DW_FORM_strx4: + { + uint64_t offset; + + switch (form) + { + case DW_FORM_strx: + offset = read_uleb128 (buf); + break; + case DW_FORM_strx1: + offset = read_byte (buf); + break; + case DW_FORM_strx2: + offset = read_uint16 (buf); + break; + case DW_FORM_strx3: + offset = read_uint24 (buf); + break; + case DW_FORM_strx4: + offset = read_uint32 (buf); + break; + default: + /* This case can't happen. */ + return 0; + } + val->encoding = ATTR_VAL_STRING_INDEX; + val->u.uint = offset; + return 1; + } + case DW_FORM_addrx: case DW_FORM_addrx1: case DW_FORM_addrx2: + case DW_FORM_addrx3: case DW_FORM_addrx4: + { + uint64_t offset; + + switch (form) + { + case DW_FORM_addrx: + offset = read_uleb128 (buf); + break; + case DW_FORM_addrx1: + offset = read_byte (buf); + break; + case DW_FORM_addrx2: + offset = read_uint16 (buf); + break; + case DW_FORM_addrx3: + offset = read_uint24 (buf); + break; + case DW_FORM_addrx4: + offset = read_uint32 (buf); + break; + default: + /* This case can't happen. */ + return 0; + } + val->encoding = ATTR_VAL_ADDRESS_INDEX; + val->u.uint = offset; + return 1; + } + case DW_FORM_ref_sup4: + val->encoding = ATTR_VAL_REF_SECTION; + val->u.uint = read_uint32 (buf); + return 1; + case DW_FORM_ref_sup8: + val->encoding = ATTR_VAL_REF_SECTION; + val->u.uint = read_uint64 (buf); + return 1; + case DW_FORM_implicit_const: + val->encoding = ATTR_VAL_UINT; + val->u.uint = implicit_val; + return 1; + case DW_FORM_loclistx: + /* We don't distinguish this from DW_FORM_sec_offset. It + * shouldn't matter since we don't care about loclists. */ + val->encoding = ATTR_VAL_REF_SECTION; + val->u.uint = read_uleb128 (buf); + return 1; + case DW_FORM_rnglistx: + val->encoding = ATTR_VAL_RNGLISTS_INDEX; + val->u.uint = read_uleb128 (buf); + return 1; + case DW_FORM_GNU_addr_index: + val->encoding = ATTR_VAL_REF_SECTION; + val->u.uint = read_uleb128 (buf); + return 1; + case DW_FORM_GNU_str_index: + val->encoding = ATTR_VAL_REF_SECTION; + val->u.uint = read_uleb128 (buf); + return 1; + case DW_FORM_GNU_ref_alt: + val->u.uint = read_offset (buf, is_dwarf64); + if (altlink == NULL) + { + val->encoding = ATTR_VAL_NONE; + return 1; + } + val->encoding = ATTR_VAL_REF_ALT_INFO; + return 1; + case DW_FORM_strp_sup: case DW_FORM_GNU_strp_alt: + { + uint64_t offset; + + offset = read_offset (buf, is_dwarf64); + if (altlink == NULL) + { + val->encoding = ATTR_VAL_NONE; + return 1; + } + if (offset >= altlink->dwarf_sections.size[DEBUG_STR]) + { + dwarf_buf_error (buf, "DW_FORM_strp_sup out of range", 0); + return 0; + } + val->encoding = ATTR_VAL_STRING; + val->u.string = + (const char *) altlink->dwarf_sections.data[DEBUG_STR] + offset; + return 1; + } + default: + dwarf_buf_error (buf, "unrecognized DWARF form", -1); + return 0; + } +} + +/* If we can determine the value of a string attribute, set *STRING to + point to the string. Return 1 on success, 0 on error. If we don't + know the value, we consider that a success, and we don't change + *STRING. An error is only reported for some sort of out of range + offset. */ + +static int +resolve_string (const struct dwarf_sections *dwarf_sections, int is_dwarf64, + int is_bigendian, uint64_t str_offsets_base, + const struct attr_val *val, + backtrace_error_callback error_callback, void *data, + const char **string) +{ + switch (val->encoding) + { + case ATTR_VAL_STRING: + *string = val->u.string; + return 1; + + case ATTR_VAL_STRING_INDEX: + { + uint64_t offset; + struct dwarf_buf offset_buf; + + offset = val->u.uint * (is_dwarf64 ? 8 : 4) + str_offsets_base; + if (offset + (is_dwarf64 ? 8 : 4) + > dwarf_sections->size[DEBUG_STR_OFFSETS]) + { + error_callback (data, "DW_FORM_strx value out of range", 0); + return 0; + } + + offset_buf.name = ".debug_str_offsets"; + offset_buf.start = dwarf_sections->data[DEBUG_STR_OFFSETS]; + offset_buf.buf = dwarf_sections->data[DEBUG_STR_OFFSETS] + offset; + offset_buf.left = dwarf_sections->size[DEBUG_STR_OFFSETS] - offset; + offset_buf.is_bigendian = is_bigendian; + offset_buf.error_callback = error_callback; + offset_buf.data = data; + offset_buf.reported_underflow = 0; + + offset = read_offset (&offset_buf, is_dwarf64); + if (offset >= dwarf_sections->size[DEBUG_STR]) + { + dwarf_buf_error (&offset_buf, + "DW_FORM_strx offset out of range", + 0); + return 0; + } + *string = (const char *) dwarf_sections->data[DEBUG_STR] + offset; + return 1; + } + + default: + return 1; + } +} + +/* Set *ADDRESS to the real address for a ATTR_VAL_ADDRESS_INDEX. + Return 1 on success, 0 on error. */ + +static int +resolve_addr_index (const struct dwarf_sections *dwarf_sections, + uint64_t addr_base, int addrsize, int is_bigendian, + uint64_t addr_index, + backtrace_error_callback error_callback, void *data, + uintptr_t *address) +{ + uint64_t offset; + struct dwarf_buf addr_buf; + + offset = addr_index * addrsize + addr_base; + if (offset + addrsize > dwarf_sections->size[DEBUG_ADDR]) + { + error_callback (data, "DW_FORM_addrx value out of range", 0); + return 0; + } + + addr_buf.name = ".debug_addr"; + addr_buf.start = dwarf_sections->data[DEBUG_ADDR]; + addr_buf.buf = dwarf_sections->data[DEBUG_ADDR] + offset; + addr_buf.left = dwarf_sections->size[DEBUG_ADDR] - offset; + addr_buf.is_bigendian = is_bigendian; + addr_buf.error_callback = error_callback; + addr_buf.data = data; + addr_buf.reported_underflow = 0; + + *address = (uintptr_t) read_address (&addr_buf, addrsize); + return 1; +} + +/* Compare a unit offset against a unit for bsearch. */ + +static int +units_search (const void *vkey, const void *ventry) +{ + const size_t *key = (const size_t *) vkey; + const struct unit *entry = *((const struct unit *const *) ventry); + size_t offset; + + offset = *key; + if (offset < entry->low_offset) + return -1; + else if (offset >= entry->high_offset) + return 1; + else + return 0; +} + +/* Find a unit in PU containing OFFSET. */ + +static struct unit * +find_unit (struct unit **pu, size_t units_count, size_t offset) +{ + struct unit **u; + u = (struct unit**)bsearch (&offset, pu, units_count, sizeof (struct unit *), units_search); + return u == NULL ? NULL : *u; +} + +/* Compare function_addrs for qsort. When ranges are nested, make the + smallest one sort last. */ + +static int +function_addrs_compare (const void *v1, const void *v2) +{ + const struct function_addrs *a1 = (const struct function_addrs *) v1; + const struct function_addrs *a2 = (const struct function_addrs *) v2; + + if (a1->low < a2->low) + return -1; + if (a1->low > a2->low) + return 1; + if (a1->high < a2->high) + return 1; + if (a1->high > a2->high) + return -1; + return strcmp (a1->function->name, a2->function->name); +} + +/* Compare a PC against a function_addrs for bsearch. We always + allocate an entra entry at the end of the vector, so that this + routine can safely look at the next entry. Note that if there are + multiple ranges containing PC, which one will be returned is + unpredictable. We compensate for that in dwarf_fileline. */ + +static int +function_addrs_search (const void *vkey, const void *ventry) +{ + const uintptr_t *key = (const uintptr_t *) vkey; + const struct function_addrs *entry = (const struct function_addrs *) ventry; + uintptr_t pc; + + pc = *key; + if (pc < entry->low) + return -1; + else if (pc > (entry + 1)->low) + return 1; + else + return 0; +} + +/* Add a new compilation unit address range to a vector. This is + called via add_ranges. Returns 1 on success, 0 on failure. */ + +static int +add_unit_addr (struct backtrace_state *state, void *rdata, + uintptr_t lowpc, uintptr_t highpc, + backtrace_error_callback error_callback, void *data, + void *pvec) +{ + struct unit *u = (struct unit *) rdata; + struct unit_addrs_vector *vec = (struct unit_addrs_vector *) pvec; + struct unit_addrs *p; + + /* Try to merge with the last entry. */ + if (vec->count > 0) + { + p = (struct unit_addrs *) vec->vec.base + (vec->count - 1); + if ((lowpc == p->high || lowpc == p->high + 1) + && u == p->u) + { + if (highpc > p->high) + p->high = highpc; + return 1; + } + } + + p = ((struct unit_addrs *) + backtrace_vector_grow (state, sizeof (struct unit_addrs), + error_callback, data, &vec->vec)); + if (p == NULL) + return 0; + + p->low = lowpc; + p->high = highpc; + p->u = u; + + ++vec->count; + + return 1; +} + +/* Compare unit_addrs for qsort. When ranges are nested, make the + smallest one sort last. */ + +static int +unit_addrs_compare (const void *v1, const void *v2) +{ + const struct unit_addrs *a1 = (const struct unit_addrs *) v1; + const struct unit_addrs *a2 = (const struct unit_addrs *) v2; + + if (a1->low < a2->low) + return -1; + if (a1->low > a2->low) + return 1; + if (a1->high < a2->high) + return 1; + if (a1->high > a2->high) + return -1; + if (a1->u->lineoff < a2->u->lineoff) + return -1; + if (a1->u->lineoff > a2->u->lineoff) + return 1; + return 0; +} + +/* Compare a PC against a unit_addrs for bsearch. We always allocate + an entry entry at the end of the vector, so that this routine can + safely look at the next entry. Note that if there are multiple + ranges containing PC, which one will be returned is unpredictable. + We compensate for that in dwarf_fileline. */ + +static int +unit_addrs_search (const void *vkey, const void *ventry) +{ + const uintptr_t *key = (const uintptr_t *) vkey; + const struct unit_addrs *entry = (const struct unit_addrs *) ventry; + uintptr_t pc; + + pc = *key; + if (pc < entry->low) + return -1; + else if (pc > (entry + 1)->low) + return 1; + else + return 0; +} + +/* Sort the line vector by PC. We want a stable sort here to maintain + the order of lines for the same PC values. Since the sequence is + being sorted in place, their addresses cannot be relied on to + maintain stability. That is the purpose of the index member. */ + +static int +line_compare (const void *v1, const void *v2) +{ + const struct line *ln1 = (const struct line *) v1; + const struct line *ln2 = (const struct line *) v2; + + if (ln1->pc < ln2->pc) + return -1; + else if (ln1->pc > ln2->pc) + return 1; + else if (ln1->idx < ln2->idx) + return -1; + else if (ln1->idx > ln2->idx) + return 1; + else + return 0; +} + +/* Find a PC in a line vector. We always allocate an extra entry at + the end of the lines vector, so that this routine can safely look + at the next entry. Note that when there are multiple mappings for + the same PC value, this will return the last one. */ + +static int +line_search (const void *vkey, const void *ventry) +{ + const uintptr_t *key = (const uintptr_t *) vkey; + const struct line *entry = (const struct line *) ventry; + uintptr_t pc; + + pc = *key; + if (pc < entry->pc) + return -1; + else if (pc >= (entry + 1)->pc) + return 1; + else + return 0; +} + +/* Sort the abbrevs by the abbrev code. This function is passed to + both qsort and bsearch. */ + +static int +abbrev_compare (const void *v1, const void *v2) +{ + const struct abbrev *a1 = (const struct abbrev *) v1; + const struct abbrev *a2 = (const struct abbrev *) v2; + + if (a1->code < a2->code) + return -1; + else if (a1->code > a2->code) + return 1; + else + { + /* This really shouldn't happen. It means there are two + different abbrevs with the same code, and that means we don't + know which one lookup_abbrev should return. */ + return 0; + } +} + +/* Read the abbreviation table for a compilation unit. Returns 1 on + success, 0 on failure. */ + +static int +read_abbrevs (struct backtrace_state *state, uint64_t abbrev_offset, + const unsigned char *dwarf_abbrev, size_t dwarf_abbrev_size, + int is_bigendian, backtrace_error_callback error_callback, + void *data, struct abbrevs *abbrevs) +{ + struct dwarf_buf abbrev_buf; + struct dwarf_buf count_buf; + size_t num_abbrevs; + + abbrevs->num_abbrevs = 0; + abbrevs->abbrevs = NULL; + + if (abbrev_offset >= dwarf_abbrev_size) + { + error_callback (data, "abbrev offset out of range", 0); + return 0; + } + + abbrev_buf.name = ".debug_abbrev"; + abbrev_buf.start = dwarf_abbrev; + abbrev_buf.buf = dwarf_abbrev + abbrev_offset; + abbrev_buf.left = dwarf_abbrev_size - abbrev_offset; + abbrev_buf.is_bigendian = is_bigendian; + abbrev_buf.error_callback = error_callback; + abbrev_buf.data = data; + abbrev_buf.reported_underflow = 0; + + /* Count the number of abbrevs in this list. */ + + count_buf = abbrev_buf; + num_abbrevs = 0; + while (read_uleb128 (&count_buf) != 0) + { + if (count_buf.reported_underflow) + return 0; + ++num_abbrevs; + // Skip tag. + read_uleb128 (&count_buf); + // Skip has_children. + read_byte (&count_buf); + // Skip attributes. + while (read_uleb128 (&count_buf) != 0) + { + uint64_t form; + + form = read_uleb128 (&count_buf); + if ((enum dwarf_form) form == DW_FORM_implicit_const) + read_sleb128 (&count_buf); + } + // Skip form of last attribute. + read_uleb128 (&count_buf); + } + + if (count_buf.reported_underflow) + return 0; + + if (num_abbrevs == 0) + return 1; + + abbrevs->abbrevs = ((struct abbrev *) + backtrace_alloc (state, + num_abbrevs * sizeof (struct abbrev), + error_callback, data)); + if (abbrevs->abbrevs == NULL) + return 0; + abbrevs->num_abbrevs = num_abbrevs; + memset (abbrevs->abbrevs, 0, num_abbrevs * sizeof (struct abbrev)); + + num_abbrevs = 0; + while (1) + { + uint64_t code; + struct abbrev a; + size_t num_attrs; + struct attr *attrs; + + if (abbrev_buf.reported_underflow) + goto fail; + + code = read_uleb128 (&abbrev_buf); + if (code == 0) + break; + + a.code = code; + a.tag = (enum dwarf_tag) read_uleb128 (&abbrev_buf); + a.has_children = read_byte (&abbrev_buf); + + count_buf = abbrev_buf; + num_attrs = 0; + while (read_uleb128 (&count_buf) != 0) + { + uint64_t form; + + ++num_attrs; + form = read_uleb128 (&count_buf); + if ((enum dwarf_form) form == DW_FORM_implicit_const) + read_sleb128 (&count_buf); + } + + if (num_attrs == 0) + { + attrs = NULL; + read_uleb128 (&abbrev_buf); + read_uleb128 (&abbrev_buf); + } + else + { + attrs = ((struct attr *) + backtrace_alloc (state, num_attrs * sizeof *attrs, + error_callback, data)); + if (attrs == NULL) + goto fail; + num_attrs = 0; + while (1) + { + uint64_t name; + uint64_t form; + + name = read_uleb128 (&abbrev_buf); + form = read_uleb128 (&abbrev_buf); + if (name == 0) + break; + attrs[num_attrs].name = (enum dwarf_attribute) name; + attrs[num_attrs].form = (enum dwarf_form) form; + if ((enum dwarf_form) form == DW_FORM_implicit_const) + attrs[num_attrs].val = read_sleb128 (&abbrev_buf); + else + attrs[num_attrs].val = 0; + ++num_attrs; + } + } + + a.num_attrs = num_attrs; + a.attrs = attrs; + + abbrevs->abbrevs[num_abbrevs] = a; + ++num_abbrevs; + } + + backtrace_qsort (abbrevs->abbrevs, abbrevs->num_abbrevs, + sizeof (struct abbrev), abbrev_compare); + + return 1; + + fail: + free_abbrevs (state, abbrevs, error_callback, data); + return 0; +} + +/* Return the abbrev information for an abbrev code. */ + +static const struct abbrev * +lookup_abbrev (struct abbrevs *abbrevs, uint64_t code, + backtrace_error_callback error_callback, void *data) +{ + struct abbrev key; + void *p; + + /* With GCC, where abbrevs are simply numbered in order, we should + be able to just look up the entry. */ + if (code - 1 < abbrevs->num_abbrevs + && abbrevs->abbrevs[code - 1].code == code) + return &abbrevs->abbrevs[code - 1]; + + /* Otherwise we have to search. */ + memset (&key, 0, sizeof key); + key.code = code; + p = bsearch (&key, abbrevs->abbrevs, abbrevs->num_abbrevs, + sizeof (struct abbrev), abbrev_compare); + if (p == NULL) + { + error_callback (data, "invalid abbreviation code", 0); + return NULL; + } + return (const struct abbrev *) p; +} + +/* This struct is used to gather address range information while + reading attributes. We use this while building a mapping from + address ranges to compilation units and then again while mapping + from address ranges to function entries. Normally either + lowpc/highpc is set or ranges is set. */ + +struct pcrange { + uintptr_t lowpc; /* The low PC value. */ + int have_lowpc; /* Whether a low PC value was found. */ + int lowpc_is_addr_index; /* Whether lowpc is in .debug_addr. */ + uintptr_t highpc; /* The high PC value. */ + int have_highpc; /* Whether a high PC value was found. */ + int highpc_is_relative; /* Whether highpc is relative to lowpc. */ + int highpc_is_addr_index; /* Whether highpc is in .debug_addr. */ + uint64_t ranges; /* Offset in ranges section. */ + int have_ranges; /* Whether ranges is valid. */ + int ranges_is_index; /* Whether ranges is DW_FORM_rnglistx. */ +}; + +/* Update PCRANGE from an attribute value. */ + +static void +update_pcrange (const struct attr* attr, const struct attr_val* val, + struct pcrange *pcrange) +{ + switch (attr->name) + { + case DW_AT_low_pc: + if (val->encoding == ATTR_VAL_ADDRESS) + { + pcrange->lowpc = (uintptr_t) val->u.uint; + pcrange->have_lowpc = 1; + } + else if (val->encoding == ATTR_VAL_ADDRESS_INDEX) + { + pcrange->lowpc = (uintptr_t) val->u.uint; + pcrange->have_lowpc = 1; + pcrange->lowpc_is_addr_index = 1; + } + break; + + case DW_AT_high_pc: + if (val->encoding == ATTR_VAL_ADDRESS) + { + pcrange->highpc = (uintptr_t) val->u.uint; + pcrange->have_highpc = 1; + } + else if (val->encoding == ATTR_VAL_UINT) + { + pcrange->highpc = (uintptr_t) val->u.uint; + pcrange->have_highpc = 1; + pcrange->highpc_is_relative = 1; + } + else if (val->encoding == ATTR_VAL_ADDRESS_INDEX) + { + pcrange->highpc = (uintptr_t) val->u.uint; + pcrange->have_highpc = 1; + pcrange->highpc_is_addr_index = 1; + } + break; + + case DW_AT_ranges: + if (val->encoding == ATTR_VAL_UINT + || val->encoding == ATTR_VAL_REF_SECTION) + { + pcrange->ranges = val->u.uint; + pcrange->have_ranges = 1; + } + else if (val->encoding == ATTR_VAL_RNGLISTS_INDEX) + { + pcrange->ranges = val->u.uint; + pcrange->have_ranges = 1; + pcrange->ranges_is_index = 1; + } + break; + + default: + break; + } +} + +/* Call ADD_RANGE for a low/high PC pair. Returns 1 on success, 0 on + error. */ + +static int +add_low_high_range (struct backtrace_state *state, + const struct dwarf_sections *dwarf_sections, + uintptr_t base_address, int is_bigendian, + struct unit *u, const struct pcrange *pcrange, + int (*add_range) (struct backtrace_state *state, + void *rdata, uintptr_t lowpc, + uintptr_t highpc, + backtrace_error_callback error_callback, + void *data, void *vec), + void *rdata, + backtrace_error_callback error_callback, void *data, + void *vec) +{ + uintptr_t lowpc; + uintptr_t highpc; + + lowpc = pcrange->lowpc; + if (pcrange->lowpc_is_addr_index) + { + if (!resolve_addr_index (dwarf_sections, u->addr_base, u->addrsize, + is_bigendian, lowpc, error_callback, data, + &lowpc)) + return 0; + } + + highpc = pcrange->highpc; + if (pcrange->highpc_is_addr_index) + { + if (!resolve_addr_index (dwarf_sections, u->addr_base, u->addrsize, + is_bigendian, highpc, error_callback, data, + &highpc)) + return 0; + } + if (pcrange->highpc_is_relative) + highpc += lowpc; + + /* Add in the base address of the module when recording PC values, + so that we can look up the PC directly. */ + lowpc += base_address; + highpc += base_address; + + return add_range (state, rdata, lowpc, highpc, error_callback, data, vec); +} + +/* Call ADD_RANGE for each range read from .debug_ranges, as used in + DWARF versions 2 through 4. */ + +static int +add_ranges_from_ranges ( + struct backtrace_state *state, + const struct dwarf_sections *dwarf_sections, + uintptr_t base_address, int is_bigendian, + struct unit *u, uintptr_t base, + const struct pcrange *pcrange, + int (*add_range) (struct backtrace_state *state, void *rdata, + uintptr_t lowpc, uintptr_t highpc, + backtrace_error_callback error_callback, void *data, + void *vec), + void *rdata, + backtrace_error_callback error_callback, void *data, + void *vec) +{ + struct dwarf_buf ranges_buf; + + if (pcrange->ranges >= dwarf_sections->size[DEBUG_RANGES]) + { + error_callback (data, "ranges offset out of range", 0); + return 0; + } + + ranges_buf.name = ".debug_ranges"; + ranges_buf.start = dwarf_sections->data[DEBUG_RANGES]; + ranges_buf.buf = dwarf_sections->data[DEBUG_RANGES] + pcrange->ranges; + ranges_buf.left = dwarf_sections->size[DEBUG_RANGES] - pcrange->ranges; + ranges_buf.is_bigendian = is_bigendian; + ranges_buf.error_callback = error_callback; + ranges_buf.data = data; + ranges_buf.reported_underflow = 0; + + while (1) + { + uint64_t low; + uint64_t high; + + if (ranges_buf.reported_underflow) + return 0; + + low = read_address (&ranges_buf, u->addrsize); + high = read_address (&ranges_buf, u->addrsize); + + if (low == 0 && high == 0) + break; + + if (is_highest_address (low, u->addrsize)) + base = (uintptr_t) high; + else + { + if (!add_range (state, rdata, + (uintptr_t) low + base + base_address, + (uintptr_t) high + base + base_address, + error_callback, data, vec)) + return 0; + } + } + + if (ranges_buf.reported_underflow) + return 0; + + return 1; +} + +/* Call ADD_RANGE for each range read from .debug_rnglists, as used in + DWARF version 5. */ + +static int +add_ranges_from_rnglists ( + struct backtrace_state *state, + const struct dwarf_sections *dwarf_sections, + uintptr_t base_address, int is_bigendian, + struct unit *u, uintptr_t base, + const struct pcrange *pcrange, + int (*add_range) (struct backtrace_state *state, void *rdata, + uintptr_t lowpc, uintptr_t highpc, + backtrace_error_callback error_callback, void *data, + void *vec), + void *rdata, + backtrace_error_callback error_callback, void *data, + void *vec) +{ + uint64_t offset; + struct dwarf_buf rnglists_buf; + + if (!pcrange->ranges_is_index) + offset = pcrange->ranges; + else + offset = u->rnglists_base + pcrange->ranges * (u->is_dwarf64 ? 8 : 4); + if (offset >= dwarf_sections->size[DEBUG_RNGLISTS]) + { + error_callback (data, "rnglists offset out of range", 0); + return 0; + } + + rnglists_buf.name = ".debug_rnglists"; + rnglists_buf.start = dwarf_sections->data[DEBUG_RNGLISTS]; + rnglists_buf.buf = dwarf_sections->data[DEBUG_RNGLISTS] + offset; + rnglists_buf.left = dwarf_sections->size[DEBUG_RNGLISTS] - offset; + rnglists_buf.is_bigendian = is_bigendian; + rnglists_buf.error_callback = error_callback; + rnglists_buf.data = data; + rnglists_buf.reported_underflow = 0; + + if (pcrange->ranges_is_index) + { + offset = read_offset (&rnglists_buf, u->is_dwarf64); + offset += u->rnglists_base; + if (offset >= dwarf_sections->size[DEBUG_RNGLISTS]) + { + error_callback (data, "rnglists index offset out of range", 0); + return 0; + } + rnglists_buf.buf = dwarf_sections->data[DEBUG_RNGLISTS] + offset; + rnglists_buf.left = dwarf_sections->size[DEBUG_RNGLISTS] - offset; + } + + while (1) + { + unsigned char rle; + + rle = read_byte (&rnglists_buf); + if (rle == DW_RLE_end_of_list) + break; + switch (rle) + { + case DW_RLE_base_addressx: + { + uint64_t index; + + index = read_uleb128 (&rnglists_buf); + if (!resolve_addr_index (dwarf_sections, u->addr_base, + u->addrsize, is_bigendian, index, + error_callback, data, &base)) + return 0; + } + break; + + case DW_RLE_startx_endx: + { + uint64_t index; + uintptr_t low; + uintptr_t high; + + index = read_uleb128 (&rnglists_buf); + if (!resolve_addr_index (dwarf_sections, u->addr_base, + u->addrsize, is_bigendian, index, + error_callback, data, &low)) + return 0; + index = read_uleb128 (&rnglists_buf); + if (!resolve_addr_index (dwarf_sections, u->addr_base, + u->addrsize, is_bigendian, index, + error_callback, data, &high)) + return 0; + if (!add_range (state, rdata, low + base_address, + high + base_address, error_callback, data, + vec)) + return 0; + } + break; + + case DW_RLE_startx_length: + { + uint64_t index; + uintptr_t low; + uintptr_t length; + + index = read_uleb128 (&rnglists_buf); + if (!resolve_addr_index (dwarf_sections, u->addr_base, + u->addrsize, is_bigendian, index, + error_callback, data, &low)) + return 0; + length = read_uleb128 (&rnglists_buf); + low += base_address; + if (!add_range (state, rdata, low, low + length, + error_callback, data, vec)) + return 0; + } + break; + + case DW_RLE_offset_pair: + { + uint64_t low; + uint64_t high; + + low = read_uleb128 (&rnglists_buf); + high = read_uleb128 (&rnglists_buf); + if (!add_range (state, rdata, low + base + base_address, + high + base + base_address, + error_callback, data, vec)) + return 0; + } + break; + + case DW_RLE_base_address: + base = (uintptr_t) read_address (&rnglists_buf, u->addrsize); + break; + + case DW_RLE_start_end: + { + uintptr_t low; + uintptr_t high; + + low = (uintptr_t) read_address (&rnglists_buf, u->addrsize); + high = (uintptr_t) read_address (&rnglists_buf, u->addrsize); + if (!add_range (state, rdata, low + base_address, + high + base_address, error_callback, data, + vec)) + return 0; + } + break; + + case DW_RLE_start_length: + { + uintptr_t low; + uintptr_t length; + + low = (uintptr_t) read_address (&rnglists_buf, u->addrsize); + length = (uintptr_t) read_uleb128 (&rnglists_buf); + low += base_address; + if (!add_range (state, rdata, low, low + length, + error_callback, data, vec)) + return 0; + } + break; + + default: + dwarf_buf_error (&rnglists_buf, "unrecognized DW_RLE value", -1); + return 0; + } + } + + if (rnglists_buf.reported_underflow) + return 0; + + return 1; +} + +/* Call ADD_RANGE for each lowpc/highpc pair in PCRANGE. RDATA is + passed to ADD_RANGE, and is either a struct unit * or a struct + function *. VEC is the vector we are adding ranges to, and is + either a struct unit_addrs_vector * or a struct function_vector *. + Returns 1 on success, 0 on error. */ + +static int +add_ranges (struct backtrace_state *state, + const struct dwarf_sections *dwarf_sections, + uintptr_t base_address, int is_bigendian, + struct unit *u, uintptr_t base, const struct pcrange *pcrange, + int (*add_range) (struct backtrace_state *state, void *rdata, + uintptr_t lowpc, uintptr_t highpc, + backtrace_error_callback error_callback, + void *data, void *vec), + void *rdata, + backtrace_error_callback error_callback, void *data, + void *vec) +{ + if (pcrange->have_lowpc && pcrange->have_highpc) + return add_low_high_range (state, dwarf_sections, base_address, + is_bigendian, u, pcrange, add_range, rdata, + error_callback, data, vec); + + if (!pcrange->have_ranges) + { + /* Did not find any address ranges to add. */ + return 1; + } + + if (u->version < 5) + return add_ranges_from_ranges (state, dwarf_sections, base_address, + is_bigendian, u, base, pcrange, add_range, + rdata, error_callback, data, vec); + else + return add_ranges_from_rnglists (state, dwarf_sections, base_address, + is_bigendian, u, base, pcrange, add_range, + rdata, error_callback, data, vec); +} + +/* Find the address range covered by a compilation unit, reading from + UNIT_BUF and adding values to U. Returns 1 if all data could be + read, 0 if there is some error. */ + +static int +find_address_ranges (struct backtrace_state *state, uintptr_t base_address, + struct dwarf_buf *unit_buf, + const struct dwarf_sections *dwarf_sections, + int is_bigendian, struct dwarf_data *altlink, + backtrace_error_callback error_callback, void *data, + struct unit *u, struct unit_addrs_vector *addrs, + enum dwarf_tag *unit_tag) +{ + while (unit_buf->left > 0) + { + uint64_t code; + const struct abbrev *abbrev; + struct pcrange pcrange; + struct attr_val name_val; + int have_name_val; + struct attr_val comp_dir_val; + int have_comp_dir_val; + size_t i; + + code = read_uleb128 (unit_buf); + if (code == 0) + return 1; + + abbrev = lookup_abbrev (&u->abbrevs, code, error_callback, data); + if (abbrev == NULL) + return 0; + + if (unit_tag != NULL) + *unit_tag = abbrev->tag; + + memset (&pcrange, 0, sizeof pcrange); + memset (&name_val, 0, sizeof name_val); + have_name_val = 0; + memset (&comp_dir_val, 0, sizeof comp_dir_val); + have_comp_dir_val = 0; + for (i = 0; i < abbrev->num_attrs; ++i) + { + struct attr_val val; + + if (!read_attribute (abbrev->attrs[i].form, abbrev->attrs[i].val, + unit_buf, u->is_dwarf64, u->version, + u->addrsize, dwarf_sections, altlink, &val)) + return 0; + + switch (abbrev->attrs[i].name) + { + case DW_AT_low_pc: case DW_AT_high_pc: case DW_AT_ranges: + update_pcrange (&abbrev->attrs[i], &val, &pcrange); + break; + + case DW_AT_stmt_list: + if ((abbrev->tag == DW_TAG_compile_unit + || abbrev->tag == DW_TAG_skeleton_unit) + && (val.encoding == ATTR_VAL_UINT + || val.encoding == ATTR_VAL_REF_SECTION)) + u->lineoff = val.u.uint; + break; + + case DW_AT_name: + if (abbrev->tag == DW_TAG_compile_unit + || abbrev->tag == DW_TAG_skeleton_unit) + { + name_val = val; + have_name_val = 1; + } + break; + + case DW_AT_comp_dir: + if (abbrev->tag == DW_TAG_compile_unit + || abbrev->tag == DW_TAG_skeleton_unit) + { + comp_dir_val = val; + have_comp_dir_val = 1; + } + break; + + case DW_AT_str_offsets_base: + if ((abbrev->tag == DW_TAG_compile_unit + || abbrev->tag == DW_TAG_skeleton_unit) + && val.encoding == ATTR_VAL_REF_SECTION) + u->str_offsets_base = val.u.uint; + break; + + case DW_AT_addr_base: + if ((abbrev->tag == DW_TAG_compile_unit + || abbrev->tag == DW_TAG_skeleton_unit) + && val.encoding == ATTR_VAL_REF_SECTION) + u->addr_base = val.u.uint; + break; + + case DW_AT_rnglists_base: + if ((abbrev->tag == DW_TAG_compile_unit + || abbrev->tag == DW_TAG_skeleton_unit) + && val.encoding == ATTR_VAL_REF_SECTION) + u->rnglists_base = val.u.uint; + break; + + default: + break; + } + } + + // Resolve strings after we're sure that we have seen + // DW_AT_str_offsets_base. + if (have_name_val) + { + if (!resolve_string (dwarf_sections, u->is_dwarf64, is_bigendian, + u->str_offsets_base, &name_val, + error_callback, data, &u->filename)) + return 0; + } + if (have_comp_dir_val) + { + if (!resolve_string (dwarf_sections, u->is_dwarf64, is_bigendian, + u->str_offsets_base, &comp_dir_val, + error_callback, data, &u->comp_dir)) + return 0; + } + + if (abbrev->tag == DW_TAG_compile_unit + || abbrev->tag == DW_TAG_subprogram + || abbrev->tag == DW_TAG_skeleton_unit) + { + if (!add_ranges (state, dwarf_sections, base_address, + is_bigendian, u, pcrange.lowpc, &pcrange, + add_unit_addr, (void *) u, error_callback, data, + (void *) addrs)) + return 0; + + /* If we found the PC range in the DW_TAG_compile_unit or + DW_TAG_skeleton_unit, we can stop now. */ + if ((abbrev->tag == DW_TAG_compile_unit + || abbrev->tag == DW_TAG_skeleton_unit) + && (pcrange.have_ranges + || (pcrange.have_lowpc && pcrange.have_highpc))) + return 1; + } + + if (abbrev->has_children) + { + if (!find_address_ranges (state, base_address, unit_buf, + dwarf_sections, is_bigendian, altlink, + error_callback, data, u, addrs, NULL)) + return 0; + } + } + + return 1; +} + +/* Build a mapping from address ranges to the compilation units where + the line number information for that range can be found. Returns 1 + on success, 0 on failure. */ + +static int +build_address_map (struct backtrace_state *state, uintptr_t base_address, + const struct dwarf_sections *dwarf_sections, + int is_bigendian, struct dwarf_data *altlink, + backtrace_error_callback error_callback, void *data, + struct unit_addrs_vector *addrs, + struct unit_vector *unit_vec) +{ + struct dwarf_buf info; + struct backtrace_vector units; + size_t units_count; + size_t i; + struct unit **pu; + size_t unit_offset = 0; + struct unit_addrs *pa; + + memset (&addrs->vec, 0, sizeof addrs->vec); + memset (&unit_vec->vec, 0, sizeof unit_vec->vec); + addrs->count = 0; + unit_vec->count = 0; + + /* Read through the .debug_info section. FIXME: Should we use the + .debug_aranges section? gdb and addr2line don't use it, but I'm + not sure why. */ + + info.name = ".debug_info"; + info.start = dwarf_sections->data[DEBUG_INFO]; + info.buf = info.start; + info.left = dwarf_sections->size[DEBUG_INFO]; + info.is_bigendian = is_bigendian; + info.error_callback = error_callback; + info.data = data; + info.reported_underflow = 0; + + memset (&units, 0, sizeof units); + units_count = 0; + + while (info.left > 0) + { + const unsigned char *unit_data_start; + uint64_t len; + int is_dwarf64; + struct dwarf_buf unit_buf; + int version; + int unit_type; + uint64_t abbrev_offset; + int addrsize; + struct unit *u; + enum dwarf_tag unit_tag; + + if (info.reported_underflow) + goto fail; + + unit_data_start = info.buf; + + len = read_initial_length (&info, &is_dwarf64); + unit_buf = info; + unit_buf.left = len; + + if (!advance (&info, len)) + goto fail; + + version = read_uint16 (&unit_buf); + if (version < 2 || version > 5) + { + dwarf_buf_error (&unit_buf, "unrecognized DWARF version", -1); + goto fail; + } + + if (version < 5) + unit_type = 0; + else + { + unit_type = read_byte (&unit_buf); + if (unit_type == DW_UT_type || unit_type == DW_UT_split_type) + { + /* This unit doesn't have anything we need. */ + continue; + } + } + + pu = ((struct unit **) + backtrace_vector_grow (state, sizeof (struct unit *), + error_callback, data, &units)); + if (pu == NULL) + goto fail; + + u = ((struct unit *) + backtrace_alloc (state, sizeof *u, error_callback, data)); + if (u == NULL) + goto fail; + + *pu = u; + ++units_count; + + if (version < 5) + addrsize = 0; /* Set below. */ + else + addrsize = read_byte (&unit_buf); + + memset (&u->abbrevs, 0, sizeof u->abbrevs); + abbrev_offset = read_offset (&unit_buf, is_dwarf64); + if (!read_abbrevs (state, abbrev_offset, + dwarf_sections->data[DEBUG_ABBREV], + dwarf_sections->size[DEBUG_ABBREV], + is_bigendian, error_callback, data, &u->abbrevs)) + goto fail; + + if (version < 5) + addrsize = read_byte (&unit_buf); + + switch (unit_type) + { + case 0: + break; + case DW_UT_compile: case DW_UT_partial: + break; + case DW_UT_skeleton: case DW_UT_split_compile: + read_uint64 (&unit_buf); /* dwo_id */ + break; + default: + break; + } + + u->low_offset = unit_offset; + unit_offset += len + (is_dwarf64 ? 12 : 4); + u->high_offset = unit_offset; + u->unit_data = unit_buf.buf; + u->unit_data_len = unit_buf.left; + u->unit_data_offset = unit_buf.buf - unit_data_start; + u->version = version; + u->is_dwarf64 = is_dwarf64; + u->addrsize = addrsize; + u->filename = NULL; + u->comp_dir = NULL; + u->abs_filename = NULL; + u->lineoff = 0; + u->str_offsets_base = 0; + u->addr_base = 0; + u->rnglists_base = 0; + + /* The actual line number mappings will be read as needed. */ + u->lines = NULL; + u->lines_count = 0; + u->function_addrs = NULL; + u->function_addrs_count = 0; + + if (!find_address_ranges (state, base_address, &unit_buf, dwarf_sections, + is_bigendian, altlink, error_callback, data, + u, addrs, &unit_tag)) + goto fail; + + if (unit_buf.reported_underflow) + goto fail; + } + if (info.reported_underflow) + goto fail; + + /* Add a trailing addrs entry, but don't include it in addrs->count. */ + pa = ((struct unit_addrs *) + backtrace_vector_grow (state, sizeof (struct unit_addrs), + error_callback, data, &addrs->vec)); + if (pa == NULL) + goto fail; + pa->low = 0; + --pa->low; + pa->high = pa->low; + pa->u = NULL; + + unit_vec->vec = units; + unit_vec->count = units_count; + return 1; + + fail: + if (units_count > 0) + { + pu = (struct unit **) units.base; + for (i = 0; i < units_count; i++) + { + free_abbrevs (state, &pu[i]->abbrevs, error_callback, data); + backtrace_free (state, pu[i], sizeof **pu, error_callback, data); + } + backtrace_vector_free (state, &units, error_callback, data); + } + if (addrs->count > 0) + { + backtrace_vector_free (state, &addrs->vec, error_callback, data); + addrs->count = 0; + } + return 0; +} + +/* Add a new mapping to the vector of line mappings that we are + building. Returns 1 on success, 0 on failure. */ + +static int +add_line (struct backtrace_state *state, struct dwarf_data *ddata, + uintptr_t pc, const char *filename, int lineno, + backtrace_error_callback error_callback, void *data, + struct line_vector *vec) +{ + struct line *ln; + + /* If we are adding the same mapping, ignore it. This can happen + when using discriminators. */ + if (vec->count > 0) + { + ln = (struct line *) vec->vec.base + (vec->count - 1); + if (pc == ln->pc && filename == ln->filename && lineno == ln->lineno) + return 1; + } + + ln = ((struct line *) + backtrace_vector_grow (state, sizeof (struct line), error_callback, + data, &vec->vec)); + if (ln == NULL) + return 0; + + /* Add in the base address here, so that we can look up the PC + directly. */ + ln->pc = pc + ddata->base_address; + + ln->filename = filename; + ln->lineno = lineno; + ln->idx = vec->count; + + ++vec->count; + + return 1; +} + +/* Free the line header information. */ + +static void +free_line_header (struct backtrace_state *state, struct line_header *hdr, + backtrace_error_callback error_callback, void *data) +{ + if (hdr->dirs_count != 0) + backtrace_free (state, hdr->dirs, hdr->dirs_count * sizeof (const char *), + error_callback, data); + backtrace_free (state, hdr->filenames, + hdr->filenames_count * sizeof (char *), + error_callback, data); +} + +/* Read the directories and file names for a line header for version + 2, setting fields in HDR. Return 1 on success, 0 on failure. */ + +static int +read_v2_paths (struct backtrace_state *state, struct unit *u, + struct dwarf_buf *hdr_buf, struct line_header *hdr) +{ + const unsigned char *p; + const unsigned char *pend; + size_t i; + + /* Count the number of directory entries. */ + hdr->dirs_count = 0; + p = hdr_buf->buf; + pend = p + hdr_buf->left; + while (p < pend && *p != '\0') + { + p += strnlen((const char *) p, pend - p) + 1; + ++hdr->dirs_count; + } + + /* The index of the first entry in the list of directories is 1. Index 0 is + used for the current directory of the compilation. To simplify index + handling, we set entry 0 to the compilation unit directory. */ + ++hdr->dirs_count; + hdr->dirs = ((const char **) + backtrace_alloc (state, + hdr->dirs_count * sizeof (const char *), + hdr_buf->error_callback, + hdr_buf->data)); + if (hdr->dirs == NULL) + return 0; + + hdr->dirs[0] = u->comp_dir; + i = 1; + while (*hdr_buf->buf != '\0') + { + if (hdr_buf->reported_underflow) + return 0; + + hdr->dirs[i] = read_string (hdr_buf); + if (hdr->dirs[i] == NULL) + return 0; + ++i; + } + if (!advance (hdr_buf, 1)) + return 0; + + /* Count the number of file entries. */ + hdr->filenames_count = 0; + p = hdr_buf->buf; + pend = p + hdr_buf->left; + while (p < pend && *p != '\0') + { + p += strnlen ((const char *) p, pend - p) + 1; + p += leb128_len (p); + p += leb128_len (p); + p += leb128_len (p); + ++hdr->filenames_count; + } + + /* The index of the first entry in the list of file names is 1. Index 0 is + used for the DW_AT_name of the compilation unit. To simplify index + handling, we set entry 0 to the compilation unit file name. */ + ++hdr->filenames_count; + hdr->filenames = ((const char **) + backtrace_alloc (state, + hdr->filenames_count * sizeof (char *), + hdr_buf->error_callback, + hdr_buf->data)); + if (hdr->filenames == NULL) + return 0; + hdr->filenames[0] = u->filename; + i = 1; + while (*hdr_buf->buf != '\0') + { + const char *filename; + uint64_t dir_index; + + if (hdr_buf->reported_underflow) + return 0; + + filename = read_string (hdr_buf); + if (filename == NULL) + return 0; + dir_index = read_uleb128 (hdr_buf); + if (IS_ABSOLUTE_PATH (filename) + || (dir_index < hdr->dirs_count && hdr->dirs[dir_index] == NULL)) + hdr->filenames[i] = filename; + else + { + const char *dir; + size_t dir_len; + size_t filename_len; + char *s; + + if (dir_index < hdr->dirs_count) + dir = hdr->dirs[dir_index]; + else + { + dwarf_buf_error (hdr_buf, + ("invalid directory index in " + "line number program header"), + 0); + return 0; + } + dir_len = strlen (dir); + filename_len = strlen (filename); + s = ((char *) backtrace_alloc (state, dir_len + filename_len + 2, + hdr_buf->error_callback, + hdr_buf->data)); + if (s == NULL) + return 0; + memcpy (s, dir, dir_len); + /* FIXME: If we are on a DOS-based file system, and the + directory or the file name use backslashes, then we + should use a backslash here. */ + s[dir_len] = '/'; + memcpy (s + dir_len + 1, filename, filename_len + 1); + hdr->filenames[i] = s; + } + + /* Ignore the modification time and size. */ + read_uleb128 (hdr_buf); + read_uleb128 (hdr_buf); + + ++i; + } + + return 1; +} + +/* Read a single version 5 LNCT entry for a directory or file name in a + line header. Sets *STRING to the resulting name, ignoring other + data. Return 1 on success, 0 on failure. */ + +static int +read_lnct (struct backtrace_state *state, struct dwarf_data *ddata, + struct unit *u, struct dwarf_buf *hdr_buf, + const struct line_header *hdr, size_t formats_count, + const struct line_header_format *formats, const char **string) +{ + size_t i; + const char *dir; + const char *path; + + dir = NULL; + path = NULL; + for (i = 0; i < formats_count; i++) + { + struct attr_val val; + + if (!read_attribute (formats[i].form, 0, hdr_buf, u->is_dwarf64, + u->version, hdr->addrsize, &ddata->dwarf_sections, + ddata->altlink, &val)) + return 0; + switch (formats[i].lnct) + { + case DW_LNCT_path: + if (!resolve_string (&ddata->dwarf_sections, u->is_dwarf64, + ddata->is_bigendian, u->str_offsets_base, + &val, hdr_buf->error_callback, hdr_buf->data, + &path)) + return 0; + break; + case DW_LNCT_directory_index: + if (val.encoding == ATTR_VAL_UINT) + { + if (val.u.uint >= hdr->dirs_count) + { + dwarf_buf_error (hdr_buf, + ("invalid directory index in " + "line number program header"), + 0); + return 0; + } + dir = hdr->dirs[val.u.uint]; + } + break; + default: + /* We don't care about timestamps or sizes or hashes. */ + break; + } + } + + if (path == NULL) + { + dwarf_buf_error (hdr_buf, + "missing file name in line number program header", + 0); + return 0; + } + + if (dir == NULL) + *string = path; + else + { + size_t dir_len; + size_t path_len; + char *s; + + dir_len = strlen (dir); + path_len = strlen (path); + s = (char *) backtrace_alloc (state, dir_len + path_len + 2, + hdr_buf->error_callback, hdr_buf->data); + if (s == NULL) + return 0; + memcpy (s, dir, dir_len); + /* FIXME: If we are on a DOS-based file system, and the + directory or the path name use backslashes, then we should + use a backslash here. */ + s[dir_len] = '/'; + memcpy (s + dir_len + 1, path, path_len + 1); + *string = s; + } + + return 1; +} + +/* Read a set of DWARF 5 line header format entries, setting *PCOUNT + and *PPATHS. Return 1 on success, 0 on failure. */ + +static int +read_line_header_format_entries (struct backtrace_state *state, + struct dwarf_data *ddata, + struct unit *u, + struct dwarf_buf *hdr_buf, + struct line_header *hdr, + size_t *pcount, + const char ***ppaths) +{ + size_t formats_count; + struct line_header_format *formats; + size_t paths_count; + const char **paths; + size_t i; + int ret; + + formats_count = read_byte (hdr_buf); + if (formats_count == 0) + formats = NULL; + else + { + formats = ((struct line_header_format *) + backtrace_alloc (state, + (formats_count + * sizeof (struct line_header_format)), + hdr_buf->error_callback, + hdr_buf->data)); + if (formats == NULL) + return 0; + + for (i = 0; i < formats_count; i++) + { + formats[i].lnct = (int) read_uleb128(hdr_buf); + formats[i].form = (enum dwarf_form) read_uleb128 (hdr_buf); + } + } + + paths_count = read_uleb128 (hdr_buf); + if (paths_count == 0) + { + *pcount = 0; + *ppaths = NULL; + ret = 1; + goto exit; + } + + paths = ((const char **) + backtrace_alloc (state, paths_count * sizeof (const char *), + hdr_buf->error_callback, hdr_buf->data)); + if (paths == NULL) + { + ret = 0; + goto exit; + } + for (i = 0; i < paths_count; i++) + { + if (!read_lnct (state, ddata, u, hdr_buf, hdr, formats_count, + formats, &paths[i])) + { + backtrace_free (state, paths, + paths_count * sizeof (const char *), + hdr_buf->error_callback, hdr_buf->data); + ret = 0; + goto exit; + } + } + + *pcount = paths_count; + *ppaths = paths; + + ret = 1; + + exit: + if (formats != NULL) + backtrace_free (state, formats, + formats_count * sizeof (struct line_header_format), + hdr_buf->error_callback, hdr_buf->data); + + return ret; +} + +/* Read the line header. Return 1 on success, 0 on failure. */ + +static int +read_line_header (struct backtrace_state *state, struct dwarf_data *ddata, + struct unit *u, int is_dwarf64, struct dwarf_buf *line_buf, + struct line_header *hdr) +{ + uint64_t hdrlen; + struct dwarf_buf hdr_buf; + + hdr->version = read_uint16 (line_buf); + if (hdr->version < 2 || hdr->version > 5) + { + dwarf_buf_error (line_buf, "unsupported line number version", -1); + return 0; + } + + if (hdr->version < 5) + hdr->addrsize = u->addrsize; + else + { + hdr->addrsize = read_byte (line_buf); + /* We could support a non-zero segment_selector_size but I doubt + we'll ever see it. */ + if (read_byte (line_buf) != 0) + { + dwarf_buf_error (line_buf, + "non-zero segment_selector_size not supported", + -1); + return 0; + } + } + + hdrlen = read_offset (line_buf, is_dwarf64); + + hdr_buf = *line_buf; + hdr_buf.left = hdrlen; + + if (!advance (line_buf, hdrlen)) + return 0; + + hdr->min_insn_len = read_byte (&hdr_buf); + if (hdr->version < 4) + hdr->max_ops_per_insn = 1; + else + hdr->max_ops_per_insn = read_byte (&hdr_buf); + + /* We don't care about default_is_stmt. */ + read_byte (&hdr_buf); + + hdr->line_base = read_sbyte (&hdr_buf); + hdr->line_range = read_byte (&hdr_buf); + + hdr->opcode_base = read_byte (&hdr_buf); + hdr->opcode_lengths = hdr_buf.buf; + if (!advance (&hdr_buf, hdr->opcode_base - 1)) + return 0; + + if (hdr->version < 5) + { + if (!read_v2_paths (state, u, &hdr_buf, hdr)) + return 0; + } + else + { + if (!read_line_header_format_entries (state, ddata, u, &hdr_buf, hdr, + &hdr->dirs_count, + &hdr->dirs)) + return 0; + if (!read_line_header_format_entries (state, ddata, u, &hdr_buf, hdr, + &hdr->filenames_count, + &hdr->filenames)) + return 0; + } + + if (hdr_buf.reported_underflow) + return 0; + + return 1; +} + +/* Read the line program, adding line mappings to VEC. Return 1 on + success, 0 on failure. */ + +static int +read_line_program (struct backtrace_state *state, struct dwarf_data *ddata, + const struct line_header *hdr, struct dwarf_buf *line_buf, + struct line_vector *vec) +{ + uint64_t address; + unsigned int op_index; + const char *reset_filename; + const char *filename; + int lineno; + + address = 0; + op_index = 0; + if (hdr->filenames_count > 1) + reset_filename = hdr->filenames[1]; + else + reset_filename = ""; + filename = reset_filename; + lineno = 1; + while (line_buf->left > 0) + { + unsigned int op; + + op = read_byte (line_buf); + if (op >= hdr->opcode_base) + { + unsigned int advance; + + /* Special opcode. */ + op -= hdr->opcode_base; + advance = op / hdr->line_range; + address += (hdr->min_insn_len * (op_index + advance) + / hdr->max_ops_per_insn); + op_index = (op_index + advance) % hdr->max_ops_per_insn; + lineno += hdr->line_base + (int) (op % hdr->line_range); + add_line (state, ddata, address, filename, lineno, + line_buf->error_callback, line_buf->data, vec); + } + else if (op == DW_LNS_extended_op) + { + uint64_t len; + + len = read_uleb128 (line_buf); + op = read_byte (line_buf); + switch (op) + { + case DW_LNE_end_sequence: + /* FIXME: Should we mark the high PC here? It seems + that we already have that information from the + compilation unit. */ + address = 0; + op_index = 0; + filename = reset_filename; + lineno = 1; + break; + case DW_LNE_set_address: + address = read_address (line_buf, hdr->addrsize); + break; + case DW_LNE_define_file: + { + const char *f; + unsigned int dir_index; + + f = read_string (line_buf); + if (f == NULL) + return 0; + dir_index = read_uleb128 (line_buf); + /* Ignore that time and length. */ + read_uleb128 (line_buf); + read_uleb128 (line_buf); + if (IS_ABSOLUTE_PATH (f)) + filename = f; + else + { + const char *dir; + size_t dir_len; + size_t f_len; + char *p; + + if (dir_index < hdr->dirs_count) + dir = hdr->dirs[dir_index]; + else + { + dwarf_buf_error (line_buf, + ("invalid directory index " + "in line number program"), + 0); + return 0; + } + dir_len = strlen (dir); + f_len = strlen (f); + p = ((char *) + backtrace_alloc (state, dir_len + f_len + 2, + line_buf->error_callback, + line_buf->data)); + if (p == NULL) + return 0; + memcpy (p, dir, dir_len); + /* FIXME: If we are on a DOS-based file system, + and the directory or the file name use + backslashes, then we should use a backslash + here. */ + p[dir_len] = '/'; + memcpy (p + dir_len + 1, f, f_len + 1); + filename = p; + } + } + break; + case DW_LNE_set_discriminator: + /* We don't care about discriminators. */ + read_uleb128 (line_buf); + break; + default: + if (!advance (line_buf, len - 1)) + return 0; + break; + } + } + else + { + switch (op) + { + case DW_LNS_copy: + add_line (state, ddata, address, filename, lineno, + line_buf->error_callback, line_buf->data, vec); + break; + case DW_LNS_advance_pc: + { + uint64_t advance; + + advance = read_uleb128 (line_buf); + address += (hdr->min_insn_len * (op_index + advance) + / hdr->max_ops_per_insn); + op_index = (op_index + advance) % hdr->max_ops_per_insn; + } + break; + case DW_LNS_advance_line: + lineno += (int) read_sleb128 (line_buf); + break; + case DW_LNS_set_file: + { + uint64_t fileno; + + fileno = read_uleb128 (line_buf); + if (fileno >= hdr->filenames_count) + { + dwarf_buf_error (line_buf, + ("invalid file number in " + "line number program"), + 0); + return 0; + } + filename = hdr->filenames[fileno]; + } + break; + case DW_LNS_set_column: + read_uleb128 (line_buf); + break; + case DW_LNS_negate_stmt: + break; + case DW_LNS_set_basic_block: + break; + case DW_LNS_const_add_pc: + { + unsigned int advance; + + op = 255 - hdr->opcode_base; + advance = op / hdr->line_range; + address += (hdr->min_insn_len * (op_index + advance) + / hdr->max_ops_per_insn); + op_index = (op_index + advance) % hdr->max_ops_per_insn; + } + break; + case DW_LNS_fixed_advance_pc: + address += read_uint16 (line_buf); + op_index = 0; + break; + case DW_LNS_set_prologue_end: + break; + case DW_LNS_set_epilogue_begin: + break; + case DW_LNS_set_isa: + read_uleb128 (line_buf); + break; + default: + { + unsigned int i; + + for (i = hdr->opcode_lengths[op - 1]; i > 0; --i) + read_uleb128 (line_buf); + } + break; + } + } + } + + return 1; +} + +/* Read the line number information for a compilation unit. Returns 1 + on success, 0 on failure. */ + +static int +read_line_info (struct backtrace_state *state, struct dwarf_data *ddata, + backtrace_error_callback error_callback, void *data, + struct unit *u, struct line_header *hdr, struct line **lines, + size_t *lines_count) +{ + struct line_vector vec; + struct dwarf_buf line_buf; + uint64_t len; + int is_dwarf64; + struct line *ln; + + memset (&vec.vec, 0, sizeof vec.vec); + vec.count = 0; + + memset (hdr, 0, sizeof *hdr); + + if (u->lineoff != (off_t) (size_t) u->lineoff + || (size_t) u->lineoff >= ddata->dwarf_sections.size[DEBUG_LINE]) + { + error_callback (data, "unit line offset out of range", 0); + goto fail; + } + + line_buf.name = ".debug_line"; + line_buf.start = ddata->dwarf_sections.data[DEBUG_LINE]; + line_buf.buf = ddata->dwarf_sections.data[DEBUG_LINE] + u->lineoff; + line_buf.left = ddata->dwarf_sections.size[DEBUG_LINE] - u->lineoff; + line_buf.is_bigendian = ddata->is_bigendian; + line_buf.error_callback = error_callback; + line_buf.data = data; + line_buf.reported_underflow = 0; + + len = read_initial_length (&line_buf, &is_dwarf64); + line_buf.left = len; + + if (!read_line_header (state, ddata, u, is_dwarf64, &line_buf, hdr)) + goto fail; + + if (!read_line_program (state, ddata, hdr, &line_buf, &vec)) + goto fail; + + if (line_buf.reported_underflow) + goto fail; + + if (vec.count == 0) + { + /* This is not a failure in the sense of a generating an error, + but it is a failure in that sense that we have no useful + information. */ + goto fail; + } + + /* Allocate one extra entry at the end. */ + ln = ((struct line *) + backtrace_vector_grow (state, sizeof (struct line), error_callback, + data, &vec.vec)); + if (ln == NULL) + goto fail; + ln->pc = (uintptr_t) -1; + ln->filename = NULL; + ln->lineno = 0; + ln->idx = 0; + + if (!backtrace_vector_release (state, &vec.vec, error_callback, data)) + goto fail; + + ln = (struct line *) vec.vec.base; + backtrace_qsort (ln, vec.count, sizeof (struct line), line_compare); + + *lines = ln; + *lines_count = vec.count; + + return 1; + + fail: + backtrace_vector_free (state, &vec.vec, error_callback, data); + free_line_header (state, hdr, error_callback, data); + *lines = (struct line *) (uintptr_t) -1; + *lines_count = 0; + return 0; +} + +static const char *read_referenced_name (struct dwarf_data *, struct unit *, + uint64_t, backtrace_error_callback, + void *); + +/* Read the name of a function from a DIE referenced by ATTR with VAL. */ + +static const char * +read_referenced_name_from_attr (struct dwarf_data *ddata, struct unit *u, + struct attr *attr, struct attr_val *val, + backtrace_error_callback error_callback, + void *data) +{ + switch (attr->name) + { + case DW_AT_abstract_origin: + case DW_AT_specification: + break; + default: + return NULL; + } + + if (attr->form == DW_FORM_ref_sig8) + return NULL; + + if (val->encoding == ATTR_VAL_REF_INFO) + { + struct unit *unit + = find_unit (ddata->units, ddata->units_count, + val->u.uint); + if (unit == NULL) + return NULL; + + uint64_t offset = val->u.uint - unit->low_offset; + return read_referenced_name (ddata, unit, offset, error_callback, data); + } + + if (val->encoding == ATTR_VAL_UINT + || val->encoding == ATTR_VAL_REF_UNIT) + return read_referenced_name (ddata, u, val->u.uint, error_callback, data); + + if (val->encoding == ATTR_VAL_REF_ALT_INFO) + { + struct unit *alt_unit + = find_unit (ddata->altlink->units, ddata->altlink->units_count, + val->u.uint); + if (alt_unit == NULL) + return NULL; + + uint64_t offset = val->u.uint - alt_unit->low_offset; + return read_referenced_name (ddata->altlink, alt_unit, offset, + error_callback, data); + } + + return NULL; +} + +/* Read the name of a function from a DIE referenced by a + DW_AT_abstract_origin or DW_AT_specification tag. OFFSET is within + the same compilation unit. */ + +static const char * +read_referenced_name (struct dwarf_data *ddata, struct unit *u, + uint64_t offset, backtrace_error_callback error_callback, + void *data) +{ + struct dwarf_buf unit_buf; + uint64_t code; + const struct abbrev *abbrev; + const char *ret; + size_t i; + + /* OFFSET is from the start of the data for this compilation unit. + U->unit_data is the data, but it starts U->unit_data_offset bytes + from the beginning. */ + + if (offset < u->unit_data_offset + || offset - u->unit_data_offset >= u->unit_data_len) + { + error_callback (data, + "abstract origin or specification out of range", + 0); + return NULL; + } + + offset -= u->unit_data_offset; + + unit_buf.name = ".debug_info"; + unit_buf.start = ddata->dwarf_sections.data[DEBUG_INFO]; + unit_buf.buf = u->unit_data + offset; + unit_buf.left = u->unit_data_len - offset; + unit_buf.is_bigendian = ddata->is_bigendian; + unit_buf.error_callback = error_callback; + unit_buf.data = data; + unit_buf.reported_underflow = 0; + + code = read_uleb128 (&unit_buf); + if (code == 0) + { + dwarf_buf_error (&unit_buf, + "invalid abstract origin or specification", + 0); + return NULL; + } + + abbrev = lookup_abbrev (&u->abbrevs, code, error_callback, data); + if (abbrev == NULL) + return NULL; + + ret = NULL; + for (i = 0; i < abbrev->num_attrs; ++i) + { + struct attr_val val; + + if (!read_attribute (abbrev->attrs[i].form, abbrev->attrs[i].val, + &unit_buf, u->is_dwarf64, u->version, u->addrsize, + &ddata->dwarf_sections, ddata->altlink, &val)) + return NULL; + + switch (abbrev->attrs[i].name) + { + case DW_AT_name: + /* Third name preference: don't override. A name we found in some + other way, will normally be more useful -- e.g., this name is + normally not mangled. */ + if (ret != NULL) + break; + if (!resolve_string (&ddata->dwarf_sections, u->is_dwarf64, + ddata->is_bigendian, u->str_offsets_base, + &val, error_callback, data, &ret)) + return NULL; + break; + + case DW_AT_linkage_name: + case DW_AT_MIPS_linkage_name: + /* First name preference: override all. */ + { + const char *s; + + s = NULL; + if (!resolve_string (&ddata->dwarf_sections, u->is_dwarf64, + ddata->is_bigendian, u->str_offsets_base, + &val, error_callback, data, &s)) + return NULL; + if (s != NULL) + return s; + } + break; + + case DW_AT_specification: + /* Second name preference: override DW_AT_name, don't override + DW_AT_linkage_name. */ + { + const char *name; + + name = read_referenced_name_from_attr (ddata, u, &abbrev->attrs[i], + &val, error_callback, data); + if (name != NULL) + ret = name; + } + break; + + default: + break; + } + } + + return ret; +} + +/* Add a range to a unit that maps to a function. This is called via + add_ranges. Returns 1 on success, 0 on error. */ + +static int +add_function_range (struct backtrace_state *state, void *rdata, + uintptr_t lowpc, uintptr_t highpc, + backtrace_error_callback error_callback, void *data, + void *pvec) +{ + struct function *function = (struct function *) rdata; + struct function_vector *vec = (struct function_vector *) pvec; + struct function_addrs *p; + + if (vec->count > 0) + { + p = (struct function_addrs *) vec->vec.base + (vec->count - 1); + if ((lowpc == p->high || lowpc == p->high + 1) + && function == p->function) + { + if (highpc > p->high) + p->high = highpc; + return 1; + } + } + + p = ((struct function_addrs *) + backtrace_vector_grow (state, sizeof (struct function_addrs), + error_callback, data, &vec->vec)); + if (p == NULL) + return 0; + + p->low = lowpc; + p->high = highpc; + p->function = function; + + ++vec->count; + + return 1; +} + +/* Read one entry plus all its children. Add function addresses to + VEC. Returns 1 on success, 0 on error. */ + +static int +read_function_entry (struct backtrace_state *state, struct dwarf_data *ddata, + struct unit *u, uintptr_t base, struct dwarf_buf *unit_buf, + const struct line_header *lhdr, + backtrace_error_callback error_callback, void *data, + struct function_vector *vec_function, + struct function_vector *vec_inlined) +{ + while (unit_buf->left > 0) + { + uint64_t code; + const struct abbrev *abbrev; + int is_function; + struct function *function; + struct function_vector *vec; + size_t i; + struct pcrange pcrange; + int have_linkage_name; + + code = read_uleb128 (unit_buf); + if (code == 0) + return 1; + + abbrev = lookup_abbrev (&u->abbrevs, code, error_callback, data); + if (abbrev == NULL) + return 0; + + is_function = (abbrev->tag == DW_TAG_subprogram + || abbrev->tag == DW_TAG_entry_point + || abbrev->tag == DW_TAG_inlined_subroutine); + + if (abbrev->tag == DW_TAG_inlined_subroutine) + vec = vec_inlined; + else + vec = vec_function; + + function = NULL; + if (is_function) + { + function = ((struct function *) + backtrace_alloc (state, sizeof *function, + error_callback, data)); + if (function == NULL) + return 0; + memset (function, 0, sizeof *function); + } + + memset (&pcrange, 0, sizeof pcrange); + have_linkage_name = 0; + for (i = 0; i < abbrev->num_attrs; ++i) + { + struct attr_val val; + + if (!read_attribute (abbrev->attrs[i].form, abbrev->attrs[i].val, + unit_buf, u->is_dwarf64, u->version, + u->addrsize, &ddata->dwarf_sections, + ddata->altlink, &val)) + return 0; + + /* The compile unit sets the base address for any address + ranges in the function entries. */ + if ((abbrev->tag == DW_TAG_compile_unit + || abbrev->tag == DW_TAG_skeleton_unit) + && abbrev->attrs[i].name == DW_AT_low_pc) + { + if (val.encoding == ATTR_VAL_ADDRESS) + base = (uintptr_t) val.u.uint; + else if (val.encoding == ATTR_VAL_ADDRESS_INDEX) + { + if (!resolve_addr_index (&ddata->dwarf_sections, + u->addr_base, u->addrsize, + ddata->is_bigendian, val.u.uint, + error_callback, data, &base)) + return 0; + } + } + + if (is_function) + { + switch (abbrev->attrs[i].name) + { + case DW_AT_call_file: + if (val.encoding == ATTR_VAL_UINT) + { + if (val.u.uint >= lhdr->filenames_count) + { + dwarf_buf_error (unit_buf, + ("invalid file number in " + "DW_AT_call_file attribute"), + 0); + return 0; + } + function->caller_filename = lhdr->filenames[val.u.uint]; + } + break; + + case DW_AT_call_line: + if (val.encoding == ATTR_VAL_UINT) + function->caller_lineno = val.u.uint; + break; + + case DW_AT_abstract_origin: + case DW_AT_specification: + /* Second name preference: override DW_AT_name, don't override + DW_AT_linkage_name. */ + if (have_linkage_name) + break; + { + const char *name; + + name + = read_referenced_name_from_attr (ddata, u, + &abbrev->attrs[i], &val, + error_callback, data); + if (name != NULL) + function->name = name; + } + break; + + case DW_AT_name: + /* Third name preference: don't override. */ + if (function->name != NULL) + break; + if (!resolve_string (&ddata->dwarf_sections, u->is_dwarf64, + ddata->is_bigendian, + u->str_offsets_base, &val, + error_callback, data, &function->name)) + return 0; + break; + + case DW_AT_linkage_name: + case DW_AT_MIPS_linkage_name: + /* First name preference: override all. */ + { + const char *s; + + s = NULL; + if (!resolve_string (&ddata->dwarf_sections, u->is_dwarf64, + ddata->is_bigendian, + u->str_offsets_base, &val, + error_callback, data, &s)) + return 0; + if (s != NULL) + { + function->name = s; + have_linkage_name = 1; + } + } + break; + + case DW_AT_low_pc: case DW_AT_high_pc: case DW_AT_ranges: + update_pcrange (&abbrev->attrs[i], &val, &pcrange); + break; + + default: + break; + } + } + } + + /* If we couldn't find a name for the function, we have no use + for it. */ + if (is_function && function->name == NULL) + { + backtrace_free (state, function, sizeof *function, + error_callback, data); + is_function = 0; + } + + if (is_function) + { + if (pcrange.have_ranges + || (pcrange.have_lowpc && pcrange.have_highpc)) + { + if (!add_ranges (state, &ddata->dwarf_sections, + ddata->base_address, ddata->is_bigendian, + u, base, &pcrange, add_function_range, + (void *) function, error_callback, data, + (void *) vec)) + return 0; + } + else + { + backtrace_free (state, function, sizeof *function, + error_callback, data); + is_function = 0; + } + } + + if (abbrev->has_children) + { + if (!is_function) + { + if (!read_function_entry (state, ddata, u, base, unit_buf, lhdr, + error_callback, data, vec_function, + vec_inlined)) + return 0; + } + else + { + struct function_vector fvec; + + /* Gather any information for inlined functions in + FVEC. */ + + memset (&fvec, 0, sizeof fvec); + + if (!read_function_entry (state, ddata, u, base, unit_buf, lhdr, + error_callback, data, vec_function, + &fvec)) + return 0; + + if (fvec.count > 0) + { + struct function_addrs *p; + struct function_addrs *faddrs; + + /* Allocate a trailing entry, but don't include it + in fvec.count. */ + p = ((struct function_addrs *) + backtrace_vector_grow (state, + sizeof (struct function_addrs), + error_callback, data, + &fvec.vec)); + if (p == NULL) + return 0; + p->low = 0; + --p->low; + p->high = p->low; + p->function = NULL; + + if (!backtrace_vector_release (state, &fvec.vec, + error_callback, data)) + return 0; + + faddrs = (struct function_addrs *) fvec.vec.base; + backtrace_qsort (faddrs, fvec.count, + sizeof (struct function_addrs), + function_addrs_compare); + + function->function_addrs = faddrs; + function->function_addrs_count = fvec.count; + } + } + } + } + + return 1; +} + +/* Read function name information for a compilation unit. We look + through the whole unit looking for function tags. */ + +static void +read_function_info (struct backtrace_state *state, struct dwarf_data *ddata, + const struct line_header *lhdr, + backtrace_error_callback error_callback, void *data, + struct unit *u, struct function_vector *fvec, + struct function_addrs **ret_addrs, + size_t *ret_addrs_count) +{ + struct function_vector lvec; + struct function_vector *pfvec; + struct dwarf_buf unit_buf; + struct function_addrs *p; + struct function_addrs *addrs; + size_t addrs_count; + + /* Use FVEC if it is not NULL. Otherwise use our own vector. */ + if (fvec != NULL) + pfvec = fvec; + else + { + memset (&lvec, 0, sizeof lvec); + pfvec = &lvec; + } + + unit_buf.name = ".debug_info"; + unit_buf.start = ddata->dwarf_sections.data[DEBUG_INFO]; + unit_buf.buf = u->unit_data; + unit_buf.left = u->unit_data_len; + unit_buf.is_bigendian = ddata->is_bigendian; + unit_buf.error_callback = error_callback; + unit_buf.data = data; + unit_buf.reported_underflow = 0; + + while (unit_buf.left > 0) + { + if (!read_function_entry (state, ddata, u, 0, &unit_buf, lhdr, + error_callback, data, pfvec, pfvec)) + return; + } + + if (pfvec->count == 0) + return; + + /* Allocate a trailing entry, but don't include it in + pfvec->count. */ + p = ((struct function_addrs *) + backtrace_vector_grow (state, sizeof (struct function_addrs), + error_callback, data, &pfvec->vec)); + if (p == NULL) + return; + p->low = 0; + --p->low; + p->high = p->low; + p->function = NULL; + + addrs_count = pfvec->count; + + if (fvec == NULL) + { + if (!backtrace_vector_release (state, &lvec.vec, error_callback, data)) + return; + addrs = (struct function_addrs *) pfvec->vec.base; + } + else + { + /* Finish this list of addresses, but leave the remaining space in + the vector available for the next function unit. */ + addrs = ((struct function_addrs *) + backtrace_vector_finish (state, &fvec->vec, + error_callback, data)); + if (addrs == NULL) + return; + fvec->count = 0; + } + + backtrace_qsort (addrs, addrs_count, sizeof (struct function_addrs), + function_addrs_compare); + + *ret_addrs = addrs; + *ret_addrs_count = addrs_count; +} + +/* See if PC is inlined in FUNCTION. If it is, print out the inlined + information, and update FILENAME and LINENO for the caller. + Returns whatever CALLBACK returns, or 0 to keep going. */ + +static int +report_inlined_functions (uintptr_t pc, struct function *function, const char* comp_dir, + backtrace_full_callback callback, void *data, + const char **filename, int *lineno) +{ + struct function_addrs *p; + struct function_addrs *match; + struct function *inlined; + int ret; + + if (function->function_addrs_count == 0) + return 0; + + /* Our search isn't safe if pc == -1, as that is the sentinel + value. */ + if (pc + 1 == 0) + return 0; + + p = ((struct function_addrs *) + bsearch (&pc, function->function_addrs, + function->function_addrs_count, + sizeof (struct function_addrs), + function_addrs_search)); + if (p == NULL) + return 0; + + /* Here pc >= p->low && pc < (p + 1)->low. The function_addrs are + sorted by low, so if pc > p->low we are at the end of a range of + function_addrs with the same low value. If pc == p->low walk + forward to the end of the range with that low value. Then walk + backward and use the first range that includes pc. */ + while (pc == (p + 1)->low) + ++p; + match = NULL; + while (1) + { + if (pc < p->high) + { + match = p; + break; + } + if (p == function->function_addrs) + break; + if ((p - 1)->low < p->low) + break; + --p; + } + if (match == NULL) + return 0; + + /* We found an inlined call. */ + + inlined = match->function; + + /* Report any calls inlined into this one. */ + ret = report_inlined_functions (pc, inlined, comp_dir, callback, data, + filename, lineno); + if (ret != 0) + return ret; + + /* Report this inlined call. */ + if (*filename[0] != '/' && comp_dir) + { + char buf[1024]; + snprintf (buf, 1024, "%s/%s", comp_dir, *filename); + ret = callback (data, pc, match->low, buf, *lineno, inlined->name); + } + else + { + ret = callback (data, pc, match->low, *filename, *lineno, inlined->name); + } + if (ret != 0) + return ret; + + /* Our caller will report the caller of the inlined function; tell + it the appropriate filename and line number. */ + *filename = inlined->caller_filename; + *lineno = inlined->caller_lineno; + + return 0; +} + +/* Look for a PC in the DWARF mapping for one module. On success, + call CALLBACK and return whatever it returns. On error, call + ERROR_CALLBACK and return 0. Sets *FOUND to 1 if the PC is found, + 0 if not. */ + +static int +dwarf_lookup_pc (struct backtrace_state *state, struct dwarf_data *ddata, + uintptr_t pc, backtrace_full_callback callback, + backtrace_error_callback error_callback, void *data, + int *found) +{ + struct unit_addrs *entry; + int found_entry; + struct unit *u; + int new_data; + struct line *lines; + struct line *ln; + struct function_addrs *p; + struct function_addrs *fmatch; + struct function *function; + const char *filename; + int lineno; + int ret; + + *found = 1; + + /* Find an address range that includes PC. Our search isn't safe if + PC == -1, as we use that as a sentinel value, so skip the search + in that case. */ + entry = (ddata->addrs_count == 0 || pc + 1 == 0 + ? NULL + : (struct unit_addrs*)bsearch (&pc, ddata->addrs, ddata->addrs_count, + sizeof (struct unit_addrs), unit_addrs_search)); + + if (entry == NULL) + { + *found = 0; + return 0; + } + + /* Here pc >= entry->low && pc < (entry + 1)->low. The unit_addrs + are sorted by low, so if pc > p->low we are at the end of a range + of unit_addrs with the same low value. If pc == p->low walk + forward to the end of the range with that low value. Then walk + backward and use the first range that includes pc. */ + while (pc == (entry + 1)->low) + ++entry; + found_entry = 0; + while (1) + { + if (pc < entry->high) + { + found_entry = 1; + break; + } + if (entry == ddata->addrs) + break; + if ((entry - 1)->low < entry->low) + break; + --entry; + } + if (!found_entry) + { + *found = 0; + return 0; + } + + /* We need the lines, lines_count, function_addrs, + function_addrs_count fields of u. If they are not set, we need + to set them. When running in threaded mode, we need to allow for + the possibility that some other thread is setting them + simultaneously. */ + + u = entry->u; + lines = u->lines; + + /* Skip units with no useful line number information by walking + backward. Useless line number information is marked by setting + lines == -1. */ + while (entry > ddata->addrs + && pc >= (entry - 1)->low + && pc < (entry - 1)->high) + { + if (state->threaded) + lines = (struct line *) backtrace_atomic_load_pointer (&u->lines); + + if (lines != (struct line *) (uintptr_t) -1) + break; + + --entry; + + u = entry->u; + lines = u->lines; + } + + if (state->threaded) + lines = backtrace_atomic_load_pointer (&u->lines); + + new_data = 0; + if (lines == NULL) + { + struct function_addrs *function_addrs; + size_t function_addrs_count; + struct line_header lhdr; + size_t count; + + /* We have never read the line information for this unit. Read + it now. */ + + function_addrs = NULL; + function_addrs_count = 0; + if (read_line_info (state, ddata, error_callback, data, entry->u, &lhdr, + &lines, &count)) + { + struct function_vector *pfvec; + + /* If not threaded, reuse DDATA->FVEC for better memory + consumption. */ + if (state->threaded) + pfvec = NULL; + else + pfvec = &ddata->fvec; + read_function_info (state, ddata, &lhdr, error_callback, data, + entry->u, pfvec, &function_addrs, + &function_addrs_count); + free_line_header (state, &lhdr, error_callback, data); + new_data = 1; + } + + /* Atomically store the information we just read into the unit. + If another thread is simultaneously writing, it presumably + read the same information, and we don't care which one we + wind up with; we just leak the other one. We do have to + write the lines field last, so that the acquire-loads above + ensure that the other fields are set. */ + + if (!state->threaded) + { + u->lines_count = count; + u->function_addrs = function_addrs; + u->function_addrs_count = function_addrs_count; + u->lines = lines; + } + else + { + backtrace_atomic_store_size_t (&u->lines_count, count); + backtrace_atomic_store_pointer (&u->function_addrs, function_addrs); + backtrace_atomic_store_size_t (&u->function_addrs_count, + function_addrs_count); + backtrace_atomic_store_pointer (&u->lines, lines); + } + } + + /* Now all fields of U have been initialized. */ + + if (lines == (struct line *) (uintptr_t) -1) + { + /* If reading the line number information failed in some way, + try again to see if there is a better compilation unit for + this PC. */ + if (new_data) + return dwarf_lookup_pc (state, ddata, pc, callback, error_callback, + data, found); + return callback (data, pc, 0, NULL, 0, NULL); + } + + /* Search for PC within this unit. */ + + ln = (struct line *) bsearch (&pc, lines, entry->u->lines_count, + sizeof (struct line), line_search); + if (ln == NULL) + { + /* The PC is between the low_pc and high_pc attributes of the + compilation unit, but no entry in the line table covers it. + This implies that the start of the compilation unit has no + line number information. */ + + if (entry->u->abs_filename == NULL) + { + const char *filename; + + filename = entry->u->filename; + if (filename != NULL + && !IS_ABSOLUTE_PATH (filename) + && entry->u->comp_dir != NULL) + { + size_t filename_len; + const char *dir; + size_t dir_len; + char *s; + + filename_len = strlen (filename); + dir = entry->u->comp_dir; + dir_len = strlen (dir); + s = (char *) backtrace_alloc (state, dir_len + filename_len + 2, + error_callback, data); + if (s == NULL) + { + *found = 0; + return 0; + } + memcpy (s, dir, dir_len); + /* FIXME: Should use backslash if DOS file system. */ + s[dir_len] = '/'; + memcpy (s + dir_len + 1, filename, filename_len + 1); + filename = s; + } + entry->u->abs_filename = filename; + } + + return callback (data, pc, 0, entry->u->abs_filename, 0, NULL); + } + + /* Search for function name within this unit. */ + + if (entry->u->function_addrs_count == 0) + return callback (data, pc, 0, ln->filename, ln->lineno, NULL); + + p = ((struct function_addrs *) + bsearch (&pc, entry->u->function_addrs, + entry->u->function_addrs_count, + sizeof (struct function_addrs), + function_addrs_search)); + if (p == NULL) + return callback (data, pc, 0, ln->filename, ln->lineno, NULL); + + /* Here pc >= p->low && pc < (p + 1)->low. The function_addrs are + sorted by low, so if pc > p->low we are at the end of a range of + function_addrs with the same low value. If pc == p->low walk + forward to the end of the range with that low value. Then walk + backward and use the first range that includes pc. */ + while (pc == (p + 1)->low) + ++p; + fmatch = NULL; + while (1) + { + if (pc < p->high) + { + fmatch = p; + break; + } + if (p == entry->u->function_addrs) + break; + if ((p - 1)->low < p->low) + break; + --p; + } + if (fmatch == NULL) + return callback (data, pc, 0, ln->filename, ln->lineno, NULL); + + function = fmatch->function; + + filename = ln->filename; + lineno = ln->lineno; + + ret = report_inlined_functions (pc, function, entry->u->comp_dir, callback, data, + &filename, &lineno); + if (ret != 0) + return ret; + + if (filename[0] != '/' && entry->u->comp_dir) + { + char buf[1024]; + snprintf (buf, 1024, "%s/%s", entry->u->comp_dir, filename); + return callback (data, pc, fmatch->low, buf, lineno, function->name); + } + else + { + return callback (data, pc, fmatch->low, filename, lineno, function->name); + } +} + +bool dwarf_fileline_dwarf_lookup_pc_in_all_entries(struct backtrace_state *state, uintptr_t pc, + backtrace_full_callback callback, backtrace_error_callback error_callback, void *data, + int& found, int ret) +{ + for (struct dwarf_data* ddata = (struct dwarf_data *)state->fileline_data; + ddata != NULL; + ddata = ddata->next) + { + ret = dwarf_lookup_pc(state, ddata, pc, callback, error_callback, data, &found); + if (ret != 0 || found) return true; + } + return false; +} + +/* Return the file/line information for a PC using the DWARF mapping + we built earlier. */ + +static int +dwarf_fileline (struct backtrace_state *state, uintptr_t pc, + backtrace_full_callback callback, + backtrace_error_callback error_callback, void *data) +{ + struct dwarf_data *ddata; + int found; + int ret = 0; + + if (!state->threaded) + { + if (dwarf_fileline_dwarf_lookup_pc_in_all_entries(state, pc, callback, error_callback, data, found, ret)) + { + return ret; + } + + // if we failed to obtain an entry in range, it can mean that the address map has been changed and new entries + // have been loaded in the meantime. Request a refresh and try again. + if (state->request_known_address_ranges_refresh_fn) + { + int new_range_count = state->request_known_address_ranges_refresh_fn(state, pc); + if (new_range_count > 0) + { + if (dwarf_fileline_dwarf_lookup_pc_in_all_entries(state, pc, callback, error_callback, data, found, ret)) + { + return ret; + } + } + } + + } + else + { + struct dwarf_data **pp; + + pp = (struct dwarf_data **) (void *) &state->fileline_data; + while (1) + { + ddata = backtrace_atomic_load_pointer (pp); + if (ddata == NULL) + break; + + ret = dwarf_lookup_pc (state, ddata, pc, callback, error_callback, + data, &found); + if (ret != 0 || found) + return ret; + + pp = &ddata->next; + } + } + + /* FIXME: See if any libraries have been dlopen'ed. */ + + return callback (data, pc, 0, NULL, 0, NULL); +} + +/* Initialize our data structures from the DWARF debug info for a + file. Return NULL on failure. */ + +static struct dwarf_data * +build_dwarf_data (struct backtrace_state *state, + uintptr_t base_address, + const struct dwarf_sections *dwarf_sections, + int is_bigendian, + struct dwarf_data *altlink, + backtrace_error_callback error_callback, + void *data) +{ + struct unit_addrs_vector addrs_vec; + struct unit_addrs *addrs; + size_t addrs_count; + struct unit_vector units_vec; + struct unit **units; + size_t units_count; + struct dwarf_data *fdata; + + if (!build_address_map (state, base_address, dwarf_sections, is_bigendian, + altlink, error_callback, data, &addrs_vec, + &units_vec)) + return NULL; + + if (!backtrace_vector_release (state, &addrs_vec.vec, error_callback, data)) + return NULL; + if (!backtrace_vector_release (state, &units_vec.vec, error_callback, data)) + return NULL; + addrs = (struct unit_addrs *) addrs_vec.vec.base; + units = (struct unit **) units_vec.vec.base; + addrs_count = addrs_vec.count; + units_count = units_vec.count; + backtrace_qsort (addrs, addrs_count, sizeof (struct unit_addrs), + unit_addrs_compare); + /* No qsort for units required, already sorted. */ + + fdata = ((struct dwarf_data *) + backtrace_alloc (state, sizeof (struct dwarf_data), + error_callback, data)); + if (fdata == NULL) + return NULL; + + fdata->next = NULL; + fdata->altlink = altlink; + fdata->base_address = base_address; + fdata->addrs = addrs; + fdata->addrs_count = addrs_count; + fdata->units = units; + fdata->units_count = units_count; + fdata->dwarf_sections = *dwarf_sections; + fdata->is_bigendian = is_bigendian; + memset (&fdata->fvec, 0, sizeof fdata->fvec); + + return fdata; +} + +/* Build our data structures from the DWARF sections for a module. + Set FILELINE_FN and STATE->FILELINE_DATA. Return 1 on success, 0 + on failure. */ + +int +backtrace_dwarf_add (struct backtrace_state *state, + uintptr_t base_address, + const struct dwarf_sections *dwarf_sections, + int is_bigendian, + struct dwarf_data *fileline_altlink, + backtrace_error_callback error_callback, + void *data, fileline *fileline_fn, + struct dwarf_data **fileline_entry) +{ + struct dwarf_data *fdata; + + fdata = build_dwarf_data (state, base_address, dwarf_sections, is_bigendian, + fileline_altlink, error_callback, data); + if (fdata == NULL) + return 0; + + if (fileline_entry != NULL) + *fileline_entry = fdata; + + if (!state->threaded) + { + struct dwarf_data **pp; + + for (pp = (struct dwarf_data **) (void *) &state->fileline_data; + *pp != NULL; + pp = &(*pp)->next) + ; + *pp = fdata; + } + else + { + while (1) + { + struct dwarf_data **pp; + + pp = (struct dwarf_data **) (void *) &state->fileline_data; + + while (1) + { + struct dwarf_data *p; + + p = backtrace_atomic_load_pointer (pp); + + if (p == NULL) + break; + + pp = &p->next; + } + + if (__sync_bool_compare_and_swap (pp, NULL, fdata)) + break; + } + } + + *fileline_fn = dwarf_fileline; + + return 1; +} + +} diff --git a/externals/tracy/public/libbacktrace/elf.cpp b/externals/tracy/public/libbacktrace/elf.cpp new file mode 100644 index 000000000..22a4ba202 --- /dev/null +++ b/externals/tracy/public/libbacktrace/elf.cpp @@ -0,0 +1,7582 @@ +/* elf.c -- Get debug data from an ELF file for backtraces. + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#include "config.h" + +#include +#include +#include +#include +#include +#include +#include + +#ifdef HAVE_DL_ITERATE_PHDR +#include +#endif + +#include "backtrace.hpp" +#include "internal.hpp" + +#include "../client/TracyFastVector.hpp" +#include "../common/TracyAlloc.hpp" + +#ifndef S_ISLNK + #ifndef S_IFLNK + #define S_IFLNK 0120000 + #endif + #ifndef S_IFMT + #define S_IFMT 0170000 + #endif + #define S_ISLNK(m) (((m) & S_IFMT) == S_IFLNK) +#endif + +#ifndef __GNUC__ +#define __builtin_prefetch(p, r, l) +#ifndef unlikely +#define unlikely(x) (x) +#endif +#else +#ifndef unlikely +#define unlikely(x) __builtin_expect(!!(x), 0) +#endif +#endif + +namespace tracy +{ + +#ifdef TRACY_DEBUGINFOD +int GetDebugInfoDescriptor( const char* buildid_data, size_t buildid_size ); +#endif + +#if !defined(HAVE_DECL_STRNLEN) || !HAVE_DECL_STRNLEN + +/* If strnlen is not declared, provide our own version. */ + +static size_t +xstrnlen (const char *s, size_t maxlen) +{ + size_t i; + + for (i = 0; i < maxlen; ++i) + if (s[i] == '\0') + break; + return i; +} + +#define strnlen xstrnlen + +#endif + +#ifndef HAVE_LSTAT + +/* Dummy version of lstat for systems that don't have it. */ + +static int +xlstat (const char *path ATTRIBUTE_UNUSED, struct stat *st ATTRIBUTE_UNUSED) +{ + return -1; +} + +#define lstat xlstat + +#endif + +#ifndef HAVE_READLINK + +/* Dummy version of readlink for systems that don't have it. */ + +static ssize_t +xreadlink (const char *path ATTRIBUTE_UNUSED, char *buf ATTRIBUTE_UNUSED, + size_t bufsz ATTRIBUTE_UNUSED) +{ + return -1; +} + +#define readlink xreadlink + +#endif + +#ifndef HAVE_DL_ITERATE_PHDR + +/* Dummy version of dl_iterate_phdr for systems that don't have it. */ + +#define dl_phdr_info x_dl_phdr_info +#define dl_iterate_phdr x_dl_iterate_phdr + +struct dl_phdr_info +{ + uintptr_t dlpi_addr; + const char *dlpi_name; +}; + +static int +dl_iterate_phdr (int (*callback) (struct dl_phdr_info *, + size_t, void *) ATTRIBUTE_UNUSED, + void *data ATTRIBUTE_UNUSED) +{ + return 0; +} + +#endif /* ! defined (HAVE_DL_ITERATE_PHDR) */ + +/* The configure script must tell us whether we are 32-bit or 64-bit + ELF. We could make this code test and support either possibility, + but there is no point. This code only works for the currently + running executable, which means that we know the ELF mode at + configure time. */ + +#if BACKTRACE_ELF_SIZE != 32 && BACKTRACE_ELF_SIZE != 64 +#error "Unknown BACKTRACE_ELF_SIZE" +#endif + +/* might #include which might define our constants + with slightly different values. Undefine them to be safe. */ + +#undef EI_NIDENT +#undef EI_MAG0 +#undef EI_MAG1 +#undef EI_MAG2 +#undef EI_MAG3 +#undef EI_CLASS +#undef EI_DATA +#undef EI_VERSION +#undef ELF_MAG0 +#undef ELF_MAG1 +#undef ELF_MAG2 +#undef ELF_MAG3 +#undef ELFCLASS32 +#undef ELFCLASS64 +#undef ELFDATA2LSB +#undef ELFDATA2MSB +#undef EV_CURRENT +#undef ET_DYN +#undef EM_PPC64 +#undef EF_PPC64_ABI +#undef SHN_LORESERVE +#undef SHN_XINDEX +#undef SHN_UNDEF +#undef SHT_PROGBITS +#undef SHT_SYMTAB +#undef SHT_STRTAB +#undef SHT_DYNSYM +#undef SHF_COMPRESSED +#undef STT_OBJECT +#undef STT_FUNC +#undef NT_GNU_BUILD_ID +#undef ELFCOMPRESS_ZLIB +#undef ELFCOMPRESS_ZSTD + +/* Basic types. */ + +typedef uint16_t b_elf_half; /* Elf_Half. */ +typedef uint32_t b_elf_word; /* Elf_Word. */ +typedef int32_t b_elf_sword; /* Elf_Sword. */ + +#if BACKTRACE_ELF_SIZE == 32 + +typedef uint32_t b_elf_addr; /* Elf_Addr. */ +typedef uint32_t b_elf_off; /* Elf_Off. */ + +typedef uint32_t b_elf_wxword; /* 32-bit Elf_Word, 64-bit ELF_Xword. */ + +#else + +typedef uint64_t b_elf_addr; /* Elf_Addr. */ +typedef uint64_t b_elf_off; /* Elf_Off. */ +typedef uint64_t b_elf_xword; /* Elf_Xword. */ +typedef int64_t b_elf_sxword; /* Elf_Sxword. */ + +typedef uint64_t b_elf_wxword; /* 32-bit Elf_Word, 64-bit ELF_Xword. */ + +#endif + +/* Data structures and associated constants. */ + +#define EI_NIDENT 16 + +typedef struct { + unsigned char e_ident[EI_NIDENT]; /* ELF "magic number" */ + b_elf_half e_type; /* Identifies object file type */ + b_elf_half e_machine; /* Specifies required architecture */ + b_elf_word e_version; /* Identifies object file version */ + b_elf_addr e_entry; /* Entry point virtual address */ + b_elf_off e_phoff; /* Program header table file offset */ + b_elf_off e_shoff; /* Section header table file offset */ + b_elf_word e_flags; /* Processor-specific flags */ + b_elf_half e_ehsize; /* ELF header size in bytes */ + b_elf_half e_phentsize; /* Program header table entry size */ + b_elf_half e_phnum; /* Program header table entry count */ + b_elf_half e_shentsize; /* Section header table entry size */ + b_elf_half e_shnum; /* Section header table entry count */ + b_elf_half e_shstrndx; /* Section header string table index */ +} b_elf_ehdr; /* Elf_Ehdr. */ + +#define EI_MAG0 0 +#define EI_MAG1 1 +#define EI_MAG2 2 +#define EI_MAG3 3 +#define EI_CLASS 4 +#define EI_DATA 5 +#define EI_VERSION 6 + +#define ELFMAG0 0x7f +#define ELFMAG1 'E' +#define ELFMAG2 'L' +#define ELFMAG3 'F' + +#define ELFCLASS32 1 +#define ELFCLASS64 2 + +#define ELFDATA2LSB 1 +#define ELFDATA2MSB 2 + +#define EV_CURRENT 1 + +#define ET_DYN 3 + +#define EM_PPC64 21 +#define EF_PPC64_ABI 3 + +typedef struct { + b_elf_word sh_name; /* Section name, index in string tbl */ + b_elf_word sh_type; /* Type of section */ + b_elf_wxword sh_flags; /* Miscellaneous section attributes */ + b_elf_addr sh_addr; /* Section virtual addr at execution */ + b_elf_off sh_offset; /* Section file offset */ + b_elf_wxword sh_size; /* Size of section in bytes */ + b_elf_word sh_link; /* Index of another section */ + b_elf_word sh_info; /* Additional section information */ + b_elf_wxword sh_addralign; /* Section alignment */ + b_elf_wxword sh_entsize; /* Entry size if section holds table */ +} b_elf_shdr; /* Elf_Shdr. */ + +#define SHN_UNDEF 0x0000 /* Undefined section */ +#define SHN_LORESERVE 0xFF00 /* Begin range of reserved indices */ +#define SHN_XINDEX 0xFFFF /* Section index is held elsewhere */ + +#define SHT_PROGBITS 1 +#define SHT_SYMTAB 2 +#define SHT_STRTAB 3 +#define SHT_DYNSYM 11 + +#define SHF_COMPRESSED 0x800 + +#if BACKTRACE_ELF_SIZE == 32 + +typedef struct +{ + b_elf_word st_name; /* Symbol name, index in string tbl */ + b_elf_addr st_value; /* Symbol value */ + b_elf_word st_size; /* Symbol size */ + unsigned char st_info; /* Symbol binding and type */ + unsigned char st_other; /* Visibility and other data */ + b_elf_half st_shndx; /* Symbol section index */ +} b_elf_sym; /* Elf_Sym. */ + +#else /* BACKTRACE_ELF_SIZE != 32 */ + +typedef struct +{ + b_elf_word st_name; /* Symbol name, index in string tbl */ + unsigned char st_info; /* Symbol binding and type */ + unsigned char st_other; /* Visibility and other data */ + b_elf_half st_shndx; /* Symbol section index */ + b_elf_addr st_value; /* Symbol value */ + b_elf_xword st_size; /* Symbol size */ +} b_elf_sym; /* Elf_Sym. */ + +#endif /* BACKTRACE_ELF_SIZE != 32 */ + +#define STT_OBJECT 1 +#define STT_FUNC 2 + +typedef struct +{ + uint32_t namesz; + uint32_t descsz; + uint32_t type; + char name[1]; +} b_elf_note; + +#define NT_GNU_BUILD_ID 3 + +#if BACKTRACE_ELF_SIZE == 32 + +typedef struct +{ + b_elf_word ch_type; /* Compresstion algorithm */ + b_elf_word ch_size; /* Uncompressed size */ + b_elf_word ch_addralign; /* Alignment for uncompressed data */ +} b_elf_chdr; /* Elf_Chdr */ + +#else /* BACKTRACE_ELF_SIZE != 32 */ + +typedef struct +{ + b_elf_word ch_type; /* Compression algorithm */ + b_elf_word ch_reserved; /* Reserved */ + b_elf_xword ch_size; /* Uncompressed size */ + b_elf_xword ch_addralign; /* Alignment for uncompressed data */ +} b_elf_chdr; /* Elf_Chdr */ + +#endif /* BACKTRACE_ELF_SIZE != 32 */ + +#define ELFCOMPRESS_ZLIB 1 +#define ELFCOMPRESS_ZSTD 2 + +/* Names of sections, indexed by enum dwarf_section in internal.h. */ + +static const char * const dwarf_section_names[DEBUG_MAX] = +{ + ".debug_info", + ".debug_line", + ".debug_abbrev", + ".debug_ranges", + ".debug_str", + ".debug_addr", + ".debug_str_offsets", + ".debug_line_str", + ".debug_rnglists" +}; + +/* Information we gather for the sections we care about. */ + +struct debug_section_info +{ + /* Section file offset. */ + off_t offset; + /* Section size. */ + size_t size; + /* Section contents, after read from file. */ + const unsigned char *data; + /* Whether the SHF_COMPRESSED flag is set for the section. */ + int compressed; +}; + +/* Information we keep for an ELF symbol. */ + +struct elf_symbol +{ + /* The name of the symbol. */ + const char *name; + /* The address of the symbol. */ + uintptr_t address; + /* The size of the symbol. */ + size_t size; +}; + +/* Information to pass to elf_syminfo. */ + +struct elf_syminfo_data +{ + /* Symbols for the next module. */ + struct elf_syminfo_data *next; + /* The ELF symbols, sorted by address. */ + struct elf_symbol *symbols; + /* The number of symbols. */ + size_t count; +}; + +/* A view that works for either a file or memory. */ + +struct elf_view +{ + struct backtrace_view view; + int release; /* If non-zero, must call backtrace_release_view. */ +}; + +/* Information about PowerPC64 ELFv1 .opd section. */ + +struct elf_ppc64_opd_data +{ + /* Address of the .opd section. */ + b_elf_addr addr; + /* Section data. */ + const char *data; + /* Size of the .opd section. */ + size_t size; + /* Corresponding section view. */ + struct elf_view view; +}; + +/* Create a view of SIZE bytes from DESCRIPTOR/MEMORY at OFFSET. */ + +static int +elf_get_view (struct backtrace_state *state, int descriptor, + const unsigned char *memory, size_t memory_size, off_t offset, + uint64_t size, backtrace_error_callback error_callback, + void *data, struct elf_view *view) +{ + if (memory == NULL) + { + view->release = 1; + return backtrace_get_view (state, descriptor, offset, size, + error_callback, data, &view->view); + } + else + { + if ((uint64_t) offset + size > (uint64_t) memory_size) + { + error_callback (data, "out of range for in-memory file", 0); + return 0; + } + view->view.data = (const void *) (memory + offset); + view->view.base = NULL; + view->view.len = size; + view->release = 0; + return 1; + } +} + +/* Release a view read by elf_get_view. */ + +static void +elf_release_view (struct backtrace_state *state, struct elf_view *view, + backtrace_error_callback error_callback, void *data) +{ + if (view->release) + backtrace_release_view (state, &view->view, error_callback, data); +} + +/* Compute the CRC-32 of BUF/LEN. This uses the CRC used for + .gnu_debuglink files. */ + +static uint32_t +elf_crc32 (uint32_t crc, const unsigned char *buf, size_t len) +{ + static const uint32_t crc32_table[256] = + { + 0x00000000, 0x77073096, 0xee0e612c, 0x990951ba, 0x076dc419, + 0x706af48f, 0xe963a535, 0x9e6495a3, 0x0edb8832, 0x79dcb8a4, + 0xe0d5e91e, 0x97d2d988, 0x09b64c2b, 0x7eb17cbd, 0xe7b82d07, + 0x90bf1d91, 0x1db71064, 0x6ab020f2, 0xf3b97148, 0x84be41de, + 0x1adad47d, 0x6ddde4eb, 0xf4d4b551, 0x83d385c7, 0x136c9856, + 0x646ba8c0, 0xfd62f97a, 0x8a65c9ec, 0x14015c4f, 0x63066cd9, + 0xfa0f3d63, 0x8d080df5, 0x3b6e20c8, 0x4c69105e, 0xd56041e4, + 0xa2677172, 0x3c03e4d1, 0x4b04d447, 0xd20d85fd, 0xa50ab56b, + 0x35b5a8fa, 0x42b2986c, 0xdbbbc9d6, 0xacbcf940, 0x32d86ce3, + 0x45df5c75, 0xdcd60dcf, 0xabd13d59, 0x26d930ac, 0x51de003a, + 0xc8d75180, 0xbfd06116, 0x21b4f4b5, 0x56b3c423, 0xcfba9599, + 0xb8bda50f, 0x2802b89e, 0x5f058808, 0xc60cd9b2, 0xb10be924, + 0x2f6f7c87, 0x58684c11, 0xc1611dab, 0xb6662d3d, 0x76dc4190, + 0x01db7106, 0x98d220bc, 0xefd5102a, 0x71b18589, 0x06b6b51f, + 0x9fbfe4a5, 0xe8b8d433, 0x7807c9a2, 0x0f00f934, 0x9609a88e, + 0xe10e9818, 0x7f6a0dbb, 0x086d3d2d, 0x91646c97, 0xe6635c01, + 0x6b6b51f4, 0x1c6c6162, 0x856530d8, 0xf262004e, 0x6c0695ed, + 0x1b01a57b, 0x8208f4c1, 0xf50fc457, 0x65b0d9c6, 0x12b7e950, + 0x8bbeb8ea, 0xfcb9887c, 0x62dd1ddf, 0x15da2d49, 0x8cd37cf3, + 0xfbd44c65, 0x4db26158, 0x3ab551ce, 0xa3bc0074, 0xd4bb30e2, + 0x4adfa541, 0x3dd895d7, 0xa4d1c46d, 0xd3d6f4fb, 0x4369e96a, + 0x346ed9fc, 0xad678846, 0xda60b8d0, 0x44042d73, 0x33031de5, + 0xaa0a4c5f, 0xdd0d7cc9, 0x5005713c, 0x270241aa, 0xbe0b1010, + 0xc90c2086, 0x5768b525, 0x206f85b3, 0xb966d409, 0xce61e49f, + 0x5edef90e, 0x29d9c998, 0xb0d09822, 0xc7d7a8b4, 0x59b33d17, + 0x2eb40d81, 0xb7bd5c3b, 0xc0ba6cad, 0xedb88320, 0x9abfb3b6, + 0x03b6e20c, 0x74b1d29a, 0xead54739, 0x9dd277af, 0x04db2615, + 0x73dc1683, 0xe3630b12, 0x94643b84, 0x0d6d6a3e, 0x7a6a5aa8, + 0xe40ecf0b, 0x9309ff9d, 0x0a00ae27, 0x7d079eb1, 0xf00f9344, + 0x8708a3d2, 0x1e01f268, 0x6906c2fe, 0xf762575d, 0x806567cb, + 0x196c3671, 0x6e6b06e7, 0xfed41b76, 0x89d32be0, 0x10da7a5a, + 0x67dd4acc, 0xf9b9df6f, 0x8ebeeff9, 0x17b7be43, 0x60b08ed5, + 0xd6d6a3e8, 0xa1d1937e, 0x38d8c2c4, 0x4fdff252, 0xd1bb67f1, + 0xa6bc5767, 0x3fb506dd, 0x48b2364b, 0xd80d2bda, 0xaf0a1b4c, + 0x36034af6, 0x41047a60, 0xdf60efc3, 0xa867df55, 0x316e8eef, + 0x4669be79, 0xcb61b38c, 0xbc66831a, 0x256fd2a0, 0x5268e236, + 0xcc0c7795, 0xbb0b4703, 0x220216b9, 0x5505262f, 0xc5ba3bbe, + 0xb2bd0b28, 0x2bb45a92, 0x5cb36a04, 0xc2d7ffa7, 0xb5d0cf31, + 0x2cd99e8b, 0x5bdeae1d, 0x9b64c2b0, 0xec63f226, 0x756aa39c, + 0x026d930a, 0x9c0906a9, 0xeb0e363f, 0x72076785, 0x05005713, + 0x95bf4a82, 0xe2b87a14, 0x7bb12bae, 0x0cb61b38, 0x92d28e9b, + 0xe5d5be0d, 0x7cdcefb7, 0x0bdbdf21, 0x86d3d2d4, 0xf1d4e242, + 0x68ddb3f8, 0x1fda836e, 0x81be16cd, 0xf6b9265b, 0x6fb077e1, + 0x18b74777, 0x88085ae6, 0xff0f6a70, 0x66063bca, 0x11010b5c, + 0x8f659eff, 0xf862ae69, 0x616bffd3, 0x166ccf45, 0xa00ae278, + 0xd70dd2ee, 0x4e048354, 0x3903b3c2, 0xa7672661, 0xd06016f7, + 0x4969474d, 0x3e6e77db, 0xaed16a4a, 0xd9d65adc, 0x40df0b66, + 0x37d83bf0, 0xa9bcae53, 0xdebb9ec5, 0x47b2cf7f, 0x30b5ffe9, + 0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6, 0xbad03605, + 0xcdd70693, 0x54de5729, 0x23d967bf, 0xb3667a2e, 0xc4614ab8, + 0x5d681b02, 0x2a6f2b94, 0xb40bbe37, 0xc30c8ea1, 0x5a05df1b, + 0x2d02ef8d + }; + const unsigned char *end; + + crc = ~crc; + for (end = buf + len; buf < end; ++ buf) + crc = crc32_table[(crc ^ *buf) & 0xff] ^ (crc >> 8); + return ~crc; +} + +/* Return the CRC-32 of the entire file open at DESCRIPTOR. */ + +static uint32_t +elf_crc32_file (struct backtrace_state *state, int descriptor, + backtrace_error_callback error_callback, void *data) +{ + struct stat st; + struct backtrace_view file_view; + uint32_t ret; + + if (fstat (descriptor, &st) < 0) + { + error_callback (data, "fstat", errno); + return 0; + } + + if (!backtrace_get_view (state, descriptor, 0, st.st_size, error_callback, + data, &file_view)) + return 0; + + ret = elf_crc32 (0, (const unsigned char *) file_view.data, st.st_size); + + backtrace_release_view (state, &file_view, error_callback, data); + + return ret; +} + +/* A dummy callback function used when we can't find a symbol + table. */ + +static void +elf_nosyms (struct backtrace_state *state ATTRIBUTE_UNUSED, + uintptr_t addr ATTRIBUTE_UNUSED, + backtrace_syminfo_callback callback ATTRIBUTE_UNUSED, + backtrace_error_callback error_callback, void *data) +{ + error_callback (data, "no symbol table in ELF executable", -1); +} + +/* A callback function used when we can't find any debug info. */ + +static int +elf_nodebug (struct backtrace_state *state, uintptr_t pc, + backtrace_full_callback callback, + backtrace_error_callback error_callback, void *data) +{ + if (state->syminfo_fn != NULL && state->syminfo_fn != elf_nosyms) + { + struct backtrace_call_full bdata; + + /* Fetch symbol information so that we can least get the + function name. */ + + bdata.full_callback = callback; + bdata.full_error_callback = error_callback; + bdata.full_data = data; + bdata.ret = 0; + state->syminfo_fn (state, pc, backtrace_syminfo_to_full_callback, + backtrace_syminfo_to_full_error_callback, &bdata); + return bdata.ret; + } + + error_callback (data, "no debug info in ELF executable", -1); + return 0; +} + +/* Compare struct elf_symbol for qsort. */ + +static int +elf_symbol_compare (const void *v1, const void *v2) +{ + const struct elf_symbol *e1 = (const struct elf_symbol *) v1; + const struct elf_symbol *e2 = (const struct elf_symbol *) v2; + + if (e1->address < e2->address) + return -1; + else if (e1->address > e2->address) + return 1; + else + return 0; +} + +/* Compare an ADDR against an elf_symbol for bsearch. We allocate one + extra entry in the array so that this can look safely at the next + entry. */ + +static int +elf_symbol_search (const void *vkey, const void *ventry) +{ + const uintptr_t *key = (const uintptr_t *) vkey; + const struct elf_symbol *entry = (const struct elf_symbol *) ventry; + uintptr_t addr; + + addr = *key; + if (addr < entry->address) + return -1; + else if (addr >= entry->address + entry->size) + return 1; + else + return 0; +} + +/* Initialize the symbol table info for elf_syminfo. */ + +static int +elf_initialize_syminfo (struct backtrace_state *state, + uintptr_t base_address, + const unsigned char *symtab_data, size_t symtab_size, + const unsigned char *strtab, size_t strtab_size, + backtrace_error_callback error_callback, + void *data, struct elf_syminfo_data *sdata, + struct elf_ppc64_opd_data *opd) +{ + size_t sym_count; + const b_elf_sym *sym; + size_t elf_symbol_count; + size_t elf_symbol_size; + struct elf_symbol *elf_symbols; + size_t i; + unsigned int j; + + sym_count = symtab_size / sizeof (b_elf_sym); + + /* We only care about function symbols. Count them. */ + sym = (const b_elf_sym *) symtab_data; + elf_symbol_count = 0; + for (i = 0; i < sym_count; ++i, ++sym) + { + int info; + + info = sym->st_info & 0xf; + if ((info == STT_FUNC || info == STT_OBJECT) + && sym->st_shndx != SHN_UNDEF) + ++elf_symbol_count; + } + + elf_symbol_size = elf_symbol_count * sizeof (struct elf_symbol); + elf_symbols = ((struct elf_symbol *) + backtrace_alloc (state, elf_symbol_size, error_callback, + data)); + if (elf_symbols == NULL) + return 0; + + sym = (const b_elf_sym *) symtab_data; + j = 0; + for (i = 0; i < sym_count; ++i, ++sym) + { + int info; + + info = sym->st_info & 0xf; + if (info != STT_FUNC && info != STT_OBJECT) + continue; + if (sym->st_shndx == SHN_UNDEF) + continue; + if (sym->st_name >= strtab_size) + { + error_callback (data, "symbol string index out of range", 0); + backtrace_free (state, elf_symbols, elf_symbol_size, error_callback, + data); + return 0; + } + elf_symbols[j].name = (const char *) strtab + sym->st_name; + /* Special case PowerPC64 ELFv1 symbols in .opd section, if the symbol + is a function descriptor, read the actual code address from the + descriptor. */ + if (opd + && sym->st_value >= opd->addr + && sym->st_value < opd->addr + opd->size) + elf_symbols[j].address + = *(const b_elf_addr *) (opd->data + (sym->st_value - opd->addr)); + else + elf_symbols[j].address = sym->st_value; + elf_symbols[j].address += base_address; + elf_symbols[j].size = sym->st_size; + ++j; + } + + backtrace_qsort (elf_symbols, elf_symbol_count, sizeof (struct elf_symbol), + elf_symbol_compare); + + sdata->next = NULL; + sdata->symbols = elf_symbols; + sdata->count = elf_symbol_count; + + return 1; +} + +/* Add EDATA to the list in STATE. */ + +static void +elf_add_syminfo_data (struct backtrace_state *state, + struct elf_syminfo_data *edata) +{ + if (!state->threaded) + { + struct elf_syminfo_data **pp; + + for (pp = (struct elf_syminfo_data **) (void *) &state->syminfo_data; + *pp != NULL; + pp = &(*pp)->next) + ; + *pp = edata; + } + else + { + while (1) + { + struct elf_syminfo_data **pp; + + pp = (struct elf_syminfo_data **) (void *) &state->syminfo_data; + + while (1) + { + struct elf_syminfo_data *p; + + p = backtrace_atomic_load_pointer (pp); + + if (p == NULL) + break; + + pp = &p->next; + } + + if (__sync_bool_compare_and_swap (pp, NULL, edata)) + break; + } + } +} + +/* Return the symbol name and value for an ADDR. */ + +static void +elf_syminfo (struct backtrace_state *state, uintptr_t addr, + backtrace_syminfo_callback callback, + backtrace_error_callback error_callback ATTRIBUTE_UNUSED, + void *data) +{ + struct elf_syminfo_data *edata; + struct elf_symbol *sym = NULL; + + if (!state->threaded) + { + for (edata = (struct elf_syminfo_data *) state->syminfo_data; + edata != NULL; + edata = edata->next) + { + sym = ((struct elf_symbol *) + bsearch (&addr, edata->symbols, edata->count, + sizeof (struct elf_symbol), elf_symbol_search)); + if (sym != NULL) + break; + } + } + else + { + struct elf_syminfo_data **pp; + + pp = (struct elf_syminfo_data **) (void *) &state->syminfo_data; + while (1) + { + edata = backtrace_atomic_load_pointer (pp); + if (edata == NULL) + break; + + sym = ((struct elf_symbol *) + bsearch (&addr, edata->symbols, edata->count, + sizeof (struct elf_symbol), elf_symbol_search)); + if (sym != NULL) + break; + + pp = &edata->next; + } + } + + if (sym == NULL) + callback (data, addr, NULL, 0, 0); + else + callback (data, addr, sym->name, sym->address, sym->size); +} + +/* Return whether FILENAME is a symlink. */ + +static int +elf_is_symlink (const char *filename) +{ + struct stat st; + + if (lstat (filename, &st) < 0) + return 0; + return S_ISLNK (st.st_mode); +} + +/* Return the results of reading the symlink FILENAME in a buffer + allocated by backtrace_alloc. Return the length of the buffer in + *LEN. */ + +static char * +elf_readlink (struct backtrace_state *state, const char *filename, + backtrace_error_callback error_callback, void *data, + size_t *plen) +{ + size_t len; + char *buf; + + len = 128; + while (1) + { + ssize_t rl; + + buf = (char*)backtrace_alloc (state, len, error_callback, data); + if (buf == NULL) + return NULL; + rl = readlink (filename, buf, len); + if (rl < 0) + { + backtrace_free (state, buf, len, error_callback, data); + return NULL; + } + if ((size_t) rl < len - 1) + { + buf[rl] = '\0'; + *plen = len; + return buf; + } + backtrace_free (state, buf, len, error_callback, data); + len *= 2; + } +} + +#define SYSTEM_BUILD_ID_DIR "/usr/lib/debug/.build-id/" + +/* Open a separate debug info file, using the build ID to find it. + Returns an open file descriptor, or -1. + + The GDB manual says that the only place gdb looks for a debug file + when the build ID is known is in /usr/lib/debug/.build-id. */ + +static int +elf_open_debugfile_by_buildid (struct backtrace_state *state, + const char *buildid_data, size_t buildid_size, + const char *filename, + backtrace_error_callback error_callback, + void *data) +{ + const char * const prefix = SYSTEM_BUILD_ID_DIR; + const size_t prefix_len = strlen (prefix); + const char * const suffix = ".debug"; + const size_t suffix_len = strlen (suffix); + size_t len; + char *bd_filename; + char *t; + size_t i; + int ret; + int does_not_exist; + + len = prefix_len + buildid_size * 2 + suffix_len + 2; + bd_filename = (char*)backtrace_alloc (state, len, error_callback, data); + if (bd_filename == NULL) + return -1; + + t = bd_filename; + memcpy (t, prefix, prefix_len); + t += prefix_len; + for (i = 0; i < buildid_size; i++) + { + unsigned char b; + unsigned char nib; + + b = (unsigned char) buildid_data[i]; + nib = (b & 0xf0) >> 4; + *t++ = nib < 10 ? '0' + nib : 'a' + nib - 10; + nib = b & 0x0f; + *t++ = nib < 10 ? '0' + nib : 'a' + nib - 10; + if (i == 0) + *t++ = '/'; + } + memcpy (t, suffix, suffix_len); + t[suffix_len] = '\0'; + + ret = backtrace_open (bd_filename, error_callback, data, &does_not_exist); + + backtrace_free (state, bd_filename, len, error_callback, data); + + /* gdb checks that the debuginfo file has the same build ID note. + That seems kind of pointless to me--why would it have the right + name but not the right build ID?--so skipping the check. */ + +#ifdef TRACY_DEBUGINFOD + if (ret == -1) + return GetDebugInfoDescriptor( buildid_data, buildid_size, filename ); + else + return ret; +#else + return ret; +#endif +} + +/* Try to open a file whose name is PREFIX (length PREFIX_LEN) + concatenated with PREFIX2 (length PREFIX2_LEN) concatenated with + DEBUGLINK_NAME. Returns an open file descriptor, or -1. */ + +static int +elf_try_debugfile (struct backtrace_state *state, const char *prefix, + size_t prefix_len, const char *prefix2, size_t prefix2_len, + const char *debuglink_name, + backtrace_error_callback error_callback, void *data) +{ + size_t debuglink_len; + size_t try_len; + char *Try; + int does_not_exist; + int ret; + + debuglink_len = strlen (debuglink_name); + try_len = prefix_len + prefix2_len + debuglink_len + 1; + Try = (char*)backtrace_alloc (state, try_len, error_callback, data); + if (Try == NULL) + return -1; + + memcpy (Try, prefix, prefix_len); + memcpy (Try + prefix_len, prefix2, prefix2_len); + memcpy (Try + prefix_len + prefix2_len, debuglink_name, debuglink_len); + Try[prefix_len + prefix2_len + debuglink_len] = '\0'; + + ret = backtrace_open (Try, error_callback, data, &does_not_exist); + + backtrace_free (state, Try, try_len, error_callback, data); + + return ret; +} + +/* Find a separate debug info file, using the debuglink section data + to find it. Returns an open file descriptor, or -1. */ + +static int +elf_find_debugfile_by_debuglink (struct backtrace_state *state, + const char *filename, + const char *debuglink_name, + backtrace_error_callback error_callback, + void *data) +{ + int ret; + char *alc; + size_t alc_len; + const char *slash; + int ddescriptor; + const char *prefix; + size_t prefix_len; + + /* Resolve symlinks in FILENAME. Since FILENAME is fairly likely to + be /proc/self/exe, symlinks are common. We don't try to resolve + the whole path name, just the base name. */ + ret = -1; + alc = NULL; + alc_len = 0; + while (elf_is_symlink (filename)) + { + char *new_buf; + size_t new_len; + + new_buf = elf_readlink (state, filename, error_callback, data, &new_len); + if (new_buf == NULL) + break; + + if (new_buf[0] == '/') + filename = new_buf; + else + { + slash = strrchr (filename, '/'); + if (slash == NULL) + filename = new_buf; + else + { + size_t clen; + char *c; + + slash++; + clen = slash - filename + strlen (new_buf) + 1; + c = (char*)backtrace_alloc (state, clen, error_callback, data); + if (c == NULL) + goto done; + + memcpy (c, filename, slash - filename); + memcpy (c + (slash - filename), new_buf, strlen (new_buf)); + c[slash - filename + strlen (new_buf)] = '\0'; + backtrace_free (state, new_buf, new_len, error_callback, data); + filename = c; + new_buf = c; + new_len = clen; + } + } + + if (alc != NULL) + backtrace_free (state, alc, alc_len, error_callback, data); + alc = new_buf; + alc_len = new_len; + } + + /* Look for DEBUGLINK_NAME in the same directory as FILENAME. */ + + slash = strrchr (filename, '/'); + if (slash == NULL) + { + prefix = ""; + prefix_len = 0; + } + else + { + slash++; + prefix = filename; + prefix_len = slash - filename; + } + + ddescriptor = elf_try_debugfile (state, prefix, prefix_len, "", 0, + debuglink_name, error_callback, data); + if (ddescriptor >= 0) + { + ret = ddescriptor; + goto done; + } + + /* Look for DEBUGLINK_NAME in a .debug subdirectory of FILENAME. */ + + ddescriptor = elf_try_debugfile (state, prefix, prefix_len, ".debug/", + strlen (".debug/"), debuglink_name, + error_callback, data); + if (ddescriptor >= 0) + { + ret = ddescriptor; + goto done; + } + + /* Look for DEBUGLINK_NAME in /usr/lib/debug. */ + + ddescriptor = elf_try_debugfile (state, "/usr/lib/debug/", + strlen ("/usr/lib/debug/"), prefix, + prefix_len, debuglink_name, + error_callback, data); + if (ddescriptor >= 0) + ret = ddescriptor; + + done: + if (alc != NULL && alc_len > 0) + backtrace_free (state, alc, alc_len, error_callback, data); + return ret; +} + +/* Open a separate debug info file, using the debuglink section data + to find it. Returns an open file descriptor, or -1. */ + +static int +elf_open_debugfile_by_debuglink (struct backtrace_state *state, + const char *filename, + const char *debuglink_name, + uint32_t debuglink_crc, + backtrace_error_callback error_callback, + void *data) +{ + int ddescriptor; + + ddescriptor = elf_find_debugfile_by_debuglink (state, filename, + debuglink_name, + error_callback, data); + if (ddescriptor < 0) + return -1; + + if (debuglink_crc != 0) + { + uint32_t got_crc; + + got_crc = elf_crc32_file (state, ddescriptor, error_callback, data); + if (got_crc != debuglink_crc) + { + backtrace_close (ddescriptor, error_callback, data); + return -1; + } + } + + return ddescriptor; +} + +/* A function useful for setting a breakpoint for an inflation failure + when this code is compiled with -g. */ + +static void +elf_uncompress_failed(void) +{ +} + +/* *PVAL is the current value being read from the stream, and *PBITS + is the number of valid bits. Ensure that *PVAL holds at least 15 + bits by reading additional bits from *PPIN, up to PINEND, as + needed. Updates *PPIN, *PVAL and *PBITS. Returns 1 on success, 0 + on error. */ + +static int +elf_fetch_bits (const unsigned char **ppin, const unsigned char *pinend, + uint64_t *pval, unsigned int *pbits) +{ + unsigned int bits; + const unsigned char *pin; + uint64_t val; + uint32_t next; + + bits = *pbits; + if (bits >= 15) + return 1; + pin = *ppin; + val = *pval; + + if (unlikely (pinend - pin < 4)) + { + elf_uncompress_failed (); + return 0; + } + +#if defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) \ + && defined(__ORDER_BIG_ENDIAN__) \ + && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ \ + || __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) + /* We've ensured that PIN is aligned. */ + next = *(const uint32_t *)pin; + +#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ + next = __builtin_bswap32 (next); +#endif +#else + next = pin[0] | (pin[1] << 8) | (pin[2] << 16) | (pin[3] << 24); +#endif + + val |= (uint64_t)next << bits; + bits += 32; + pin += 4; + + /* We will need the next four bytes soon. */ + __builtin_prefetch (pin, 0, 0); + + *ppin = pin; + *pval = val; + *pbits = bits; + return 1; +} + +/* This is like elf_fetch_bits, but it fetchs the bits backward, and ensures at + least 16 bits. This is for zstd. */ + +static int +elf_fetch_bits_backward (const unsigned char **ppin, + const unsigned char *pinend, + uint64_t *pval, unsigned int *pbits) +{ + unsigned int bits; + const unsigned char *pin; + uint64_t val; + uint32_t next; + + bits = *pbits; + if (bits >= 16) + return 1; + pin = *ppin; + val = *pval; + + if (unlikely (pin <= pinend)) + { + if (bits == 0) + { + elf_uncompress_failed (); + return 0; + } + return 1; + } + + pin -= 4; + +#if defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) \ + && defined(__ORDER_BIG_ENDIAN__) \ + && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ \ + || __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) + /* We've ensured that PIN is aligned. */ + next = *(const uint32_t *)pin; + +#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ + next = __builtin_bswap32 (next); +#endif +#else + next = pin[0] | (pin[1] << 8) | (pin[2] << 16) | (pin[3] << 24); +#endif + + val <<= 32; + val |= next; + bits += 32; + + if (unlikely (pin < pinend)) + { + val >>= (pinend - pin) * 8; + bits -= (pinend - pin) * 8; + } + + *ppin = pin; + *pval = val; + *pbits = bits; + return 1; +} + +/* Initialize backward fetching when the bitstream starts with a 1 bit in the + last byte in memory (which is the first one that we read). This is used by + zstd decompression. Returns 1 on success, 0 on error. */ + +static int +elf_fetch_backward_init (const unsigned char **ppin, + const unsigned char *pinend, + uint64_t *pval, unsigned int *pbits) +{ + const unsigned char *pin; + unsigned int stream_start; + uint64_t val; + unsigned int bits; + + pin = *ppin; + stream_start = (unsigned int)*pin; + if (unlikely (stream_start == 0)) + { + elf_uncompress_failed (); + return 0; + } + val = 0; + bits = 0; + + /* Align to a 32-bit boundary. */ + while ((((uintptr_t)pin) & 3) != 0) + { + val <<= 8; + val |= (uint64_t)*pin; + bits += 8; + --pin; + } + + val <<= 8; + val |= (uint64_t)*pin; + bits += 8; + + *ppin = pin; + *pval = val; + *pbits = bits; + if (!elf_fetch_bits_backward (ppin, pinend, pval, pbits)) + return 0; + + *pbits -= __builtin_clz (stream_start) - (sizeof (unsigned int) - 1) * 8 + 1; + + if (!elf_fetch_bits_backward (ppin, pinend, pval, pbits)) + return 0; + + return 1; +} + +/* Huffman code tables, like the rest of the zlib format, are defined + by RFC 1951. We store a Huffman code table as a series of tables + stored sequentially in memory. Each entry in a table is 16 bits. + The first, main, table has 256 entries. It is followed by a set of + secondary tables of length 2 to 128 entries. The maximum length of + a code sequence in the deflate format is 15 bits, so that is all we + need. Each secondary table has an index, which is the offset of + the table in the overall memory storage. + + The deflate format says that all codes of a given bit length are + lexicographically consecutive. Perhaps we could have 130 values + that require a 15-bit code, perhaps requiring three secondary + tables of size 128. I don't know if this is actually possible, but + it suggests that the maximum size required for secondary tables is + 3 * 128 + 3 * 64 ... == 768. The zlib enough program reports 660 + as the maximum. We permit 768, since in addition to the 256 for + the primary table, with two bytes per entry, and with the two + tables we need, that gives us a page. + + A single table entry needs to store a value or (for the main table + only) the index and size of a secondary table. Values range from 0 + to 285, inclusive. Secondary table indexes, per above, range from + 0 to 510. For a value we need to store the number of bits we need + to determine that value (one value may appear multiple times in the + table), which is 1 to 8. For a secondary table we need to store + the number of bits used to index into the table, which is 1 to 7. + And of course we need 1 bit to decide whether we have a value or a + secondary table index. So each entry needs 9 bits for value/table + index, 3 bits for size, 1 bit what it is. For simplicity we use 16 + bits per entry. */ + +/* Number of entries we allocate to for one code table. We get a page + for the two code tables we need. */ + +#define ZLIB_HUFFMAN_TABLE_SIZE (1024) + +/* Bit masks and shifts for the values in the table. */ + +#define ZLIB_HUFFMAN_VALUE_MASK 0x01ff +#define ZLIB_HUFFMAN_BITS_SHIFT 9 +#define ZLIB_HUFFMAN_BITS_MASK 0x7 +#define ZLIB_HUFFMAN_SECONDARY_SHIFT 12 + +/* For working memory while inflating we need two code tables, we need + an array of code lengths (max value 15, so we use unsigned char), + and an array of unsigned shorts used while building a table. The + latter two arrays must be large enough to hold the maximum number + of code lengths, which RFC 1951 defines as 286 + 30. */ + +#define ZLIB_TABLE_SIZE \ + (2 * ZLIB_HUFFMAN_TABLE_SIZE * sizeof (uint16_t) \ + + (286 + 30) * sizeof (uint16_t) \ + + (286 + 30) * sizeof (unsigned char)) + +#define ZLIB_TABLE_CODELEN_OFFSET \ + (2 * ZLIB_HUFFMAN_TABLE_SIZE * sizeof (uint16_t) \ + + (286 + 30) * sizeof (uint16_t)) + +#define ZLIB_TABLE_WORK_OFFSET \ + (2 * ZLIB_HUFFMAN_TABLE_SIZE * sizeof (uint16_t)) + +#ifdef BACKTRACE_GENERATE_FIXED_HUFFMAN_TABLE + +/* Used by the main function that generates the fixed table to learn + the table size. */ +static size_t final_next_secondary; + +#endif + +/* Build a Huffman code table from an array of lengths in CODES of + length CODES_LEN. The table is stored into *TABLE. ZDEBUG_TABLE + is the same as for elf_zlib_inflate, used to find some work space. + Returns 1 on success, 0 on error. */ + +static int +elf_zlib_inflate_table (unsigned char *codes, size_t codes_len, + uint16_t *zdebug_table, uint16_t *table) +{ + uint16_t count[16]; + uint16_t start[16]; + uint16_t prev[16]; + uint16_t firstcode[7]; + uint16_t *next; + size_t i; + size_t j; + unsigned int code; + size_t next_secondary; + + /* Count the number of code of each length. Set NEXT[val] to be the + next value after VAL with the same bit length. */ + + next = (uint16_t *) (((unsigned char *) zdebug_table) + + ZLIB_TABLE_WORK_OFFSET); + + memset (&count[0], 0, 16 * sizeof (uint16_t)); + for (i = 0; i < codes_len; ++i) + { + if (unlikely (codes[i] >= 16)) + { + elf_uncompress_failed (); + return 0; + } + + if (count[codes[i]] == 0) + { + start[codes[i]] = i; + prev[codes[i]] = i; + } + else + { + next[prev[codes[i]]] = i; + prev[codes[i]] = i; + } + + ++count[codes[i]]; + } + + /* For each length, fill in the table for the codes of that + length. */ + + memset (table, 0, ZLIB_HUFFMAN_TABLE_SIZE * sizeof (uint16_t)); + + /* Handle the values that do not require a secondary table. */ + + code = 0; + for (j = 1; j <= 8; ++j) + { + unsigned int jcnt; + unsigned int val; + + jcnt = count[j]; + if (jcnt == 0) + continue; + + if (unlikely (jcnt > (1U << j))) + { + elf_uncompress_failed (); + return 0; + } + + /* There are JCNT values that have this length, the values + starting from START[j] continuing through NEXT[VAL]. Those + values are assigned consecutive values starting at CODE. */ + + val = start[j]; + for (i = 0; i < jcnt; ++i) + { + uint16_t tval; + size_t ind; + unsigned int incr; + + /* In the compressed bit stream, the value VAL is encoded as + J bits with the value C. */ + + if (unlikely ((val & ~ZLIB_HUFFMAN_VALUE_MASK) != 0)) + { + elf_uncompress_failed (); + return 0; + } + + tval = val | ((j - 1) << ZLIB_HUFFMAN_BITS_SHIFT); + + /* The table lookup uses 8 bits. If J is less than 8, we + don't know what the other bits will be. We need to fill + in all possibilities in the table. Since the Huffman + code is unambiguous, those entries can't be used for any + other code. */ + + for (ind = code; ind < 0x100; ind += 1 << j) + { + if (unlikely (table[ind] != 0)) + { + elf_uncompress_failed (); + return 0; + } + table[ind] = tval; + } + + /* Advance to the next value with this length. */ + if (i + 1 < jcnt) + val = next[val]; + + /* The Huffman codes are stored in the bitstream with the + most significant bit first, as is required to make them + unambiguous. The effect is that when we read them from + the bitstream we see the bit sequence in reverse order: + the most significant bit of the Huffman code is the least + significant bit of the value we read from the bitstream. + That means that to make our table lookups work, we need + to reverse the bits of CODE. Since reversing bits is + tedious and in general requires using a table, we instead + increment CODE in reverse order. That is, if the number + of bits we are currently using, here named J, is 3, we + count as 000, 100, 010, 110, 001, 101, 011, 111, which is + to say the numbers from 0 to 7 but with the bits + reversed. Going to more bits, aka incrementing J, + effectively just adds more zero bits as the beginning, + and as such does not change the numeric value of CODE. + + To increment CODE of length J in reverse order, find the + most significant zero bit and set it to one while + clearing all higher bits. In other words, add 1 modulo + 2^J, only reversed. */ + + incr = 1U << (j - 1); + while ((code & incr) != 0) + incr >>= 1; + if (incr == 0) + code = 0; + else + { + code &= incr - 1; + code += incr; + } + } + } + + /* Handle the values that require a secondary table. */ + + /* Set FIRSTCODE, the number at which the codes start, for each + length. */ + + for (j = 9; j < 16; j++) + { + unsigned int jcnt; + unsigned int k; + + jcnt = count[j]; + if (jcnt == 0) + continue; + + /* There are JCNT values that have this length, the values + starting from START[j]. Those values are assigned + consecutive values starting at CODE. */ + + firstcode[j - 9] = code; + + /* Reverse add JCNT to CODE modulo 2^J. */ + for (k = 0; k < j; ++k) + { + if ((jcnt & (1U << k)) != 0) + { + unsigned int m; + unsigned int bit; + + bit = 1U << (j - k - 1); + for (m = 0; m < j - k; ++m, bit >>= 1) + { + if ((code & bit) == 0) + { + code += bit; + break; + } + code &= ~bit; + } + jcnt &= ~(1U << k); + } + } + if (unlikely (jcnt != 0)) + { + elf_uncompress_failed (); + return 0; + } + } + + /* For J from 9 to 15, inclusive, we store COUNT[J] consecutive + values starting at START[J] with consecutive codes starting at + FIRSTCODE[J - 9]. In the primary table we need to point to the + secondary table, and the secondary table will be indexed by J - 9 + bits. We count down from 15 so that we install the larger + secondary tables first, as the smaller ones may be embedded in + the larger ones. */ + + next_secondary = 0; /* Index of next secondary table (after primary). */ + for (j = 15; j >= 9; j--) + { + unsigned int jcnt; + unsigned int val; + size_t primary; /* Current primary index. */ + size_t secondary; /* Offset to current secondary table. */ + size_t secondary_bits; /* Bit size of current secondary table. */ + + jcnt = count[j]; + if (jcnt == 0) + continue; + + val = start[j]; + code = firstcode[j - 9]; + primary = 0x100; + secondary = 0; + secondary_bits = 0; + for (i = 0; i < jcnt; ++i) + { + uint16_t tval; + size_t ind; + unsigned int incr; + + if ((code & 0xff) != primary) + { + uint16_t tprimary; + + /* Fill in a new primary table entry. */ + + primary = code & 0xff; + + tprimary = table[primary]; + if (tprimary == 0) + { + /* Start a new secondary table. */ + + if (unlikely ((next_secondary & ZLIB_HUFFMAN_VALUE_MASK) + != next_secondary)) + { + elf_uncompress_failed (); + return 0; + } + + secondary = next_secondary; + secondary_bits = j - 8; + next_secondary += 1 << secondary_bits; + table[primary] = (secondary + + ((j - 8) << ZLIB_HUFFMAN_BITS_SHIFT) + + (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)); + } + else + { + /* There is an existing entry. It had better be a + secondary table with enough bits. */ + if (unlikely ((tprimary + & (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)) + == 0)) + { + elf_uncompress_failed (); + return 0; + } + secondary = tprimary & ZLIB_HUFFMAN_VALUE_MASK; + secondary_bits = ((tprimary >> ZLIB_HUFFMAN_BITS_SHIFT) + & ZLIB_HUFFMAN_BITS_MASK); + if (unlikely (secondary_bits < j - 8)) + { + elf_uncompress_failed (); + return 0; + } + } + } + + /* Fill in secondary table entries. */ + + tval = val | ((j - 8) << ZLIB_HUFFMAN_BITS_SHIFT); + + for (ind = code >> 8; + ind < (1U << secondary_bits); + ind += 1U << (j - 8)) + { + if (unlikely (table[secondary + 0x100 + ind] != 0)) + { + elf_uncompress_failed (); + return 0; + } + table[secondary + 0x100 + ind] = tval; + } + + if (i + 1 < jcnt) + val = next[val]; + + incr = 1U << (j - 1); + while ((code & incr) != 0) + incr >>= 1; + if (incr == 0) + code = 0; + else + { + code &= incr - 1; + code += incr; + } + } + } + +#ifdef BACKTRACE_GENERATE_FIXED_HUFFMAN_TABLE + final_next_secondary = next_secondary; +#endif + + return 1; +} + +#ifdef BACKTRACE_GENERATE_FIXED_HUFFMAN_TABLE + +/* Used to generate the fixed Huffman table for block type 1. */ + +#include + +static uint16_t table[ZLIB_TABLE_SIZE]; +static unsigned char codes[288]; + +int +main () +{ + size_t i; + + for (i = 0; i <= 143; ++i) + codes[i] = 8; + for (i = 144; i <= 255; ++i) + codes[i] = 9; + for (i = 256; i <= 279; ++i) + codes[i] = 7; + for (i = 280; i <= 287; ++i) + codes[i] = 8; + if (!elf_zlib_inflate_table (&codes[0], 288, &table[0], &table[0])) + { + fprintf (stderr, "elf_zlib_inflate_table failed\n"); + exit (EXIT_FAILURE); + } + + printf ("static const uint16_t elf_zlib_default_table[%#zx] =\n", + final_next_secondary + 0x100); + printf ("{\n"); + for (i = 0; i < final_next_secondary + 0x100; i += 8) + { + size_t j; + + printf (" "); + for (j = i; j < final_next_secondary + 0x100 && j < i + 8; ++j) + printf (" %#x,", table[j]); + printf ("\n"); + } + printf ("};\n"); + printf ("\n"); + + for (i = 0; i < 32; ++i) + codes[i] = 5; + if (!elf_zlib_inflate_table (&codes[0], 32, &table[0], &table[0])) + { + fprintf (stderr, "elf_zlib_inflate_table failed\n"); + exit (EXIT_FAILURE); + } + + printf ("static const uint16_t elf_zlib_default_dist_table[%#zx] =\n", + final_next_secondary + 0x100); + printf ("{\n"); + for (i = 0; i < final_next_secondary + 0x100; i += 8) + { + size_t j; + + printf (" "); + for (j = i; j < final_next_secondary + 0x100 && j < i + 8; ++j) + printf (" %#x,", table[j]); + printf ("\n"); + } + printf ("};\n"); + + return 0; +} + +#endif + +/* The fixed tables generated by the #ifdef'ed out main function + above. */ + +static const uint16_t elf_zlib_default_table[0x170] = +{ + 0xd00, 0xe50, 0xe10, 0xf18, 0xd10, 0xe70, 0xe30, 0x1230, + 0xd08, 0xe60, 0xe20, 0x1210, 0xe00, 0xe80, 0xe40, 0x1250, + 0xd04, 0xe58, 0xe18, 0x1200, 0xd14, 0xe78, 0xe38, 0x1240, + 0xd0c, 0xe68, 0xe28, 0x1220, 0xe08, 0xe88, 0xe48, 0x1260, + 0xd02, 0xe54, 0xe14, 0xf1c, 0xd12, 0xe74, 0xe34, 0x1238, + 0xd0a, 0xe64, 0xe24, 0x1218, 0xe04, 0xe84, 0xe44, 0x1258, + 0xd06, 0xe5c, 0xe1c, 0x1208, 0xd16, 0xe7c, 0xe3c, 0x1248, + 0xd0e, 0xe6c, 0xe2c, 0x1228, 0xe0c, 0xe8c, 0xe4c, 0x1268, + 0xd01, 0xe52, 0xe12, 0xf1a, 0xd11, 0xe72, 0xe32, 0x1234, + 0xd09, 0xe62, 0xe22, 0x1214, 0xe02, 0xe82, 0xe42, 0x1254, + 0xd05, 0xe5a, 0xe1a, 0x1204, 0xd15, 0xe7a, 0xe3a, 0x1244, + 0xd0d, 0xe6a, 0xe2a, 0x1224, 0xe0a, 0xe8a, 0xe4a, 0x1264, + 0xd03, 0xe56, 0xe16, 0xf1e, 0xd13, 0xe76, 0xe36, 0x123c, + 0xd0b, 0xe66, 0xe26, 0x121c, 0xe06, 0xe86, 0xe46, 0x125c, + 0xd07, 0xe5e, 0xe1e, 0x120c, 0xd17, 0xe7e, 0xe3e, 0x124c, + 0xd0f, 0xe6e, 0xe2e, 0x122c, 0xe0e, 0xe8e, 0xe4e, 0x126c, + 0xd00, 0xe51, 0xe11, 0xf19, 0xd10, 0xe71, 0xe31, 0x1232, + 0xd08, 0xe61, 0xe21, 0x1212, 0xe01, 0xe81, 0xe41, 0x1252, + 0xd04, 0xe59, 0xe19, 0x1202, 0xd14, 0xe79, 0xe39, 0x1242, + 0xd0c, 0xe69, 0xe29, 0x1222, 0xe09, 0xe89, 0xe49, 0x1262, + 0xd02, 0xe55, 0xe15, 0xf1d, 0xd12, 0xe75, 0xe35, 0x123a, + 0xd0a, 0xe65, 0xe25, 0x121a, 0xe05, 0xe85, 0xe45, 0x125a, + 0xd06, 0xe5d, 0xe1d, 0x120a, 0xd16, 0xe7d, 0xe3d, 0x124a, + 0xd0e, 0xe6d, 0xe2d, 0x122a, 0xe0d, 0xe8d, 0xe4d, 0x126a, + 0xd01, 0xe53, 0xe13, 0xf1b, 0xd11, 0xe73, 0xe33, 0x1236, + 0xd09, 0xe63, 0xe23, 0x1216, 0xe03, 0xe83, 0xe43, 0x1256, + 0xd05, 0xe5b, 0xe1b, 0x1206, 0xd15, 0xe7b, 0xe3b, 0x1246, + 0xd0d, 0xe6b, 0xe2b, 0x1226, 0xe0b, 0xe8b, 0xe4b, 0x1266, + 0xd03, 0xe57, 0xe17, 0xf1f, 0xd13, 0xe77, 0xe37, 0x123e, + 0xd0b, 0xe67, 0xe27, 0x121e, 0xe07, 0xe87, 0xe47, 0x125e, + 0xd07, 0xe5f, 0xe1f, 0x120e, 0xd17, 0xe7f, 0xe3f, 0x124e, + 0xd0f, 0xe6f, 0xe2f, 0x122e, 0xe0f, 0xe8f, 0xe4f, 0x126e, + 0x290, 0x291, 0x292, 0x293, 0x294, 0x295, 0x296, 0x297, + 0x298, 0x299, 0x29a, 0x29b, 0x29c, 0x29d, 0x29e, 0x29f, + 0x2a0, 0x2a1, 0x2a2, 0x2a3, 0x2a4, 0x2a5, 0x2a6, 0x2a7, + 0x2a8, 0x2a9, 0x2aa, 0x2ab, 0x2ac, 0x2ad, 0x2ae, 0x2af, + 0x2b0, 0x2b1, 0x2b2, 0x2b3, 0x2b4, 0x2b5, 0x2b6, 0x2b7, + 0x2b8, 0x2b9, 0x2ba, 0x2bb, 0x2bc, 0x2bd, 0x2be, 0x2bf, + 0x2c0, 0x2c1, 0x2c2, 0x2c3, 0x2c4, 0x2c5, 0x2c6, 0x2c7, + 0x2c8, 0x2c9, 0x2ca, 0x2cb, 0x2cc, 0x2cd, 0x2ce, 0x2cf, + 0x2d0, 0x2d1, 0x2d2, 0x2d3, 0x2d4, 0x2d5, 0x2d6, 0x2d7, + 0x2d8, 0x2d9, 0x2da, 0x2db, 0x2dc, 0x2dd, 0x2de, 0x2df, + 0x2e0, 0x2e1, 0x2e2, 0x2e3, 0x2e4, 0x2e5, 0x2e6, 0x2e7, + 0x2e8, 0x2e9, 0x2ea, 0x2eb, 0x2ec, 0x2ed, 0x2ee, 0x2ef, + 0x2f0, 0x2f1, 0x2f2, 0x2f3, 0x2f4, 0x2f5, 0x2f6, 0x2f7, + 0x2f8, 0x2f9, 0x2fa, 0x2fb, 0x2fc, 0x2fd, 0x2fe, 0x2ff, +}; + +static const uint16_t elf_zlib_default_dist_table[0x100] = +{ + 0x800, 0x810, 0x808, 0x818, 0x804, 0x814, 0x80c, 0x81c, + 0x802, 0x812, 0x80a, 0x81a, 0x806, 0x816, 0x80e, 0x81e, + 0x801, 0x811, 0x809, 0x819, 0x805, 0x815, 0x80d, 0x81d, + 0x803, 0x813, 0x80b, 0x81b, 0x807, 0x817, 0x80f, 0x81f, + 0x800, 0x810, 0x808, 0x818, 0x804, 0x814, 0x80c, 0x81c, + 0x802, 0x812, 0x80a, 0x81a, 0x806, 0x816, 0x80e, 0x81e, + 0x801, 0x811, 0x809, 0x819, 0x805, 0x815, 0x80d, 0x81d, + 0x803, 0x813, 0x80b, 0x81b, 0x807, 0x817, 0x80f, 0x81f, + 0x800, 0x810, 0x808, 0x818, 0x804, 0x814, 0x80c, 0x81c, + 0x802, 0x812, 0x80a, 0x81a, 0x806, 0x816, 0x80e, 0x81e, + 0x801, 0x811, 0x809, 0x819, 0x805, 0x815, 0x80d, 0x81d, + 0x803, 0x813, 0x80b, 0x81b, 0x807, 0x817, 0x80f, 0x81f, + 0x800, 0x810, 0x808, 0x818, 0x804, 0x814, 0x80c, 0x81c, + 0x802, 0x812, 0x80a, 0x81a, 0x806, 0x816, 0x80e, 0x81e, + 0x801, 0x811, 0x809, 0x819, 0x805, 0x815, 0x80d, 0x81d, + 0x803, 0x813, 0x80b, 0x81b, 0x807, 0x817, 0x80f, 0x81f, + 0x800, 0x810, 0x808, 0x818, 0x804, 0x814, 0x80c, 0x81c, + 0x802, 0x812, 0x80a, 0x81a, 0x806, 0x816, 0x80e, 0x81e, + 0x801, 0x811, 0x809, 0x819, 0x805, 0x815, 0x80d, 0x81d, + 0x803, 0x813, 0x80b, 0x81b, 0x807, 0x817, 0x80f, 0x81f, + 0x800, 0x810, 0x808, 0x818, 0x804, 0x814, 0x80c, 0x81c, + 0x802, 0x812, 0x80a, 0x81a, 0x806, 0x816, 0x80e, 0x81e, + 0x801, 0x811, 0x809, 0x819, 0x805, 0x815, 0x80d, 0x81d, + 0x803, 0x813, 0x80b, 0x81b, 0x807, 0x817, 0x80f, 0x81f, + 0x800, 0x810, 0x808, 0x818, 0x804, 0x814, 0x80c, 0x81c, + 0x802, 0x812, 0x80a, 0x81a, 0x806, 0x816, 0x80e, 0x81e, + 0x801, 0x811, 0x809, 0x819, 0x805, 0x815, 0x80d, 0x81d, + 0x803, 0x813, 0x80b, 0x81b, 0x807, 0x817, 0x80f, 0x81f, + 0x800, 0x810, 0x808, 0x818, 0x804, 0x814, 0x80c, 0x81c, + 0x802, 0x812, 0x80a, 0x81a, 0x806, 0x816, 0x80e, 0x81e, + 0x801, 0x811, 0x809, 0x819, 0x805, 0x815, 0x80d, 0x81d, + 0x803, 0x813, 0x80b, 0x81b, 0x807, 0x817, 0x80f, 0x81f, +}; + +/* Inflate a zlib stream from PIN/SIN to POUT/SOUT. Return 1 on + success, 0 on some error parsing the stream. */ + +static int +elf_zlib_inflate (const unsigned char *pin, size_t sin, uint16_t *zdebug_table, + unsigned char *pout, size_t sout) +{ + unsigned char *porigout; + const unsigned char *pinend; + unsigned char *poutend; + + /* We can apparently see multiple zlib streams concatenated + together, so keep going as long as there is something to read. + The last 4 bytes are the checksum. */ + porigout = pout; + pinend = pin + sin; + poutend = pout + sout; + while ((pinend - pin) > 4) + { + uint64_t val; + unsigned int bits; + int last; + + /* Read the two byte zlib header. */ + + if (unlikely ((pin[0] & 0xf) != 8)) /* 8 is zlib encoding. */ + { + /* Unknown compression method. */ + elf_uncompress_failed (); + return 0; + } + if (unlikely ((pin[0] >> 4) > 7)) + { + /* Window size too large. Other than this check, we don't + care about the window size. */ + elf_uncompress_failed (); + return 0; + } + if (unlikely ((pin[1] & 0x20) != 0)) + { + /* Stream expects a predefined dictionary, but we have no + dictionary. */ + elf_uncompress_failed (); + return 0; + } + val = (pin[0] << 8) | pin[1]; + if (unlikely (val % 31 != 0)) + { + /* Header check failure. */ + elf_uncompress_failed (); + return 0; + } + pin += 2; + + /* Align PIN to a 32-bit boundary. */ + + val = 0; + bits = 0; + while ((((uintptr_t) pin) & 3) != 0) + { + val |= (uint64_t)*pin << bits; + bits += 8; + ++pin; + } + + /* Read blocks until one is marked last. */ + + last = 0; + + while (!last) + { + unsigned int type; + const uint16_t *tlit; + const uint16_t *tdist; + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + last = val & 1; + type = (val >> 1) & 3; + val >>= 3; + bits -= 3; + + if (unlikely (type == 3)) + { + /* Invalid block type. */ + elf_uncompress_failed (); + return 0; + } + + if (type == 0) + { + uint16_t len; + uint16_t lenc; + + /* An uncompressed block. */ + + /* If we've read ahead more than a byte, back up. */ + while (bits >= 8) + { + --pin; + bits -= 8; + } + + val = 0; + bits = 0; + if (unlikely ((pinend - pin) < 4)) + { + /* Missing length. */ + elf_uncompress_failed (); + return 0; + } + len = pin[0] | (pin[1] << 8); + lenc = pin[2] | (pin[3] << 8); + pin += 4; + lenc = ~lenc; + if (unlikely (len != lenc)) + { + /* Corrupt data. */ + elf_uncompress_failed (); + return 0; + } + if (unlikely (len > (unsigned int) (pinend - pin) + || len > (unsigned int) (poutend - pout))) + { + /* Not enough space in buffers. */ + elf_uncompress_failed (); + return 0; + } + memcpy (pout, pin, len); + pout += len; + pin += len; + + /* Align PIN. */ + while ((((uintptr_t) pin) & 3) != 0) + { + val |= (uint64_t)*pin << bits; + bits += 8; + ++pin; + } + + /* Go around to read the next block. */ + continue; + } + + if (type == 1) + { + tlit = elf_zlib_default_table; + tdist = elf_zlib_default_dist_table; + } + else + { + unsigned int nlit; + unsigned int ndist; + unsigned int nclen; + unsigned char codebits[19]; + unsigned char *plenbase; + unsigned char *plen; + unsigned char *plenend; + + /* Read a Huffman encoding table. The various magic + numbers here are from RFC 1951. */ + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + nlit = (val & 0x1f) + 257; + val >>= 5; + ndist = (val & 0x1f) + 1; + val >>= 5; + nclen = (val & 0xf) + 4; + val >>= 4; + bits -= 14; + if (unlikely (nlit > 286 || ndist > 30)) + { + /* Values out of range. */ + elf_uncompress_failed (); + return 0; + } + + /* Read and build the table used to compress the + literal, length, and distance codes. */ + + memset(&codebits[0], 0, 19); + + /* There are always at least 4 elements in the + table. */ + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + codebits[16] = val & 7; + codebits[17] = (val >> 3) & 7; + codebits[18] = (val >> 6) & 7; + codebits[0] = (val >> 9) & 7; + val >>= 12; + bits -= 12; + + if (nclen == 4) + goto codebitsdone; + + codebits[8] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 5) + goto codebitsdone; + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + codebits[7] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 6) + goto codebitsdone; + + codebits[9] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 7) + goto codebitsdone; + + codebits[6] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 8) + goto codebitsdone; + + codebits[10] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 9) + goto codebitsdone; + + codebits[5] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 10) + goto codebitsdone; + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + codebits[11] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 11) + goto codebitsdone; + + codebits[4] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 12) + goto codebitsdone; + + codebits[12] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 13) + goto codebitsdone; + + codebits[3] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 14) + goto codebitsdone; + + codebits[13] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 15) + goto codebitsdone; + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + codebits[2] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 16) + goto codebitsdone; + + codebits[14] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 17) + goto codebitsdone; + + codebits[1] = val & 7; + val >>= 3; + bits -= 3; + + if (nclen == 18) + goto codebitsdone; + + codebits[15] = val & 7; + val >>= 3; + bits -= 3; + + codebitsdone: + + if (!elf_zlib_inflate_table (codebits, 19, zdebug_table, + zdebug_table)) + return 0; + + /* Read the compressed bit lengths of the literal, + length, and distance codes. We have allocated space + at the end of zdebug_table to hold them. */ + + plenbase = (((unsigned char *) zdebug_table) + + ZLIB_TABLE_CODELEN_OFFSET); + plen = plenbase; + plenend = plen + nlit + ndist; + while (plen < plenend) + { + uint16_t t; + unsigned int b; + uint16_t v; + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + t = zdebug_table[val & 0xff]; + + /* The compression here uses bit lengths up to 7, so + a secondary table is never necessary. */ + if (unlikely ((t & (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)) + != 0)) + { + elf_uncompress_failed (); + return 0; + } + + b = (t >> ZLIB_HUFFMAN_BITS_SHIFT) & ZLIB_HUFFMAN_BITS_MASK; + val >>= b + 1; + bits -= b + 1; + + v = t & ZLIB_HUFFMAN_VALUE_MASK; + if (v < 16) + *plen++ = v; + else if (v == 16) + { + unsigned int c; + unsigned int prev; + + /* Copy previous entry 3 to 6 times. */ + + if (unlikely (plen == plenbase)) + { + elf_uncompress_failed (); + return 0; + } + + /* We used up to 7 bits since the last + elf_fetch_bits, so we have at least 8 bits + available here. */ + + c = 3 + (val & 0x3); + val >>= 2; + bits -= 2; + if (unlikely ((unsigned int) (plenend - plen) < c)) + { + elf_uncompress_failed (); + return 0; + } + + prev = plen[-1]; + switch (c) + { + case 6: + *plen++ = prev; + ATTRIBUTE_FALLTHROUGH; + case 5: + *plen++ = prev; + ATTRIBUTE_FALLTHROUGH; + case 4: + *plen++ = prev; + } + *plen++ = prev; + *plen++ = prev; + *plen++ = prev; + } + else if (v == 17) + { + unsigned int c; + + /* Store zero 3 to 10 times. */ + + /* We used up to 7 bits since the last + elf_fetch_bits, so we have at least 8 bits + available here. */ + + c = 3 + (val & 0x7); + val >>= 3; + bits -= 3; + if (unlikely ((unsigned int) (plenend - plen) < c)) + { + elf_uncompress_failed (); + return 0; + } + + switch (c) + { + case 10: + *plen++ = 0; + ATTRIBUTE_FALLTHROUGH; + case 9: + *plen++ = 0; + ATTRIBUTE_FALLTHROUGH; + case 8: + *plen++ = 0; + ATTRIBUTE_FALLTHROUGH; + case 7: + *plen++ = 0; + ATTRIBUTE_FALLTHROUGH; + case 6: + *plen++ = 0; + ATTRIBUTE_FALLTHROUGH; + case 5: + *plen++ = 0; + ATTRIBUTE_FALLTHROUGH; + case 4: + *plen++ = 0; + } + *plen++ = 0; + *plen++ = 0; + *plen++ = 0; + } + else if (v == 18) + { + unsigned int c; + + /* Store zero 11 to 138 times. */ + + /* We used up to 7 bits since the last + elf_fetch_bits, so we have at least 8 bits + available here. */ + + c = 11 + (val & 0x7f); + val >>= 7; + bits -= 7; + if (unlikely ((unsigned int) (plenend - plen) < c)) + { + elf_uncompress_failed (); + return 0; + } + + memset (plen, 0, c); + plen += c; + } + else + { + elf_uncompress_failed (); + return 0; + } + } + + /* Make sure that the stop code can appear. */ + + plen = plenbase; + if (unlikely (plen[256] == 0)) + { + elf_uncompress_failed (); + return 0; + } + + /* Build the decompression tables. */ + + if (!elf_zlib_inflate_table (plen, nlit, zdebug_table, + zdebug_table)) + return 0; + if (!elf_zlib_inflate_table (plen + nlit, ndist, zdebug_table, + (zdebug_table + + ZLIB_HUFFMAN_TABLE_SIZE))) + return 0; + tlit = zdebug_table; + tdist = zdebug_table + ZLIB_HUFFMAN_TABLE_SIZE; + } + + /* Inflate values until the end of the block. This is the + main loop of the inflation code. */ + + while (1) + { + uint16_t t; + unsigned int b; + uint16_t v; + unsigned int lit; + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + t = tlit[val & 0xff]; + b = (t >> ZLIB_HUFFMAN_BITS_SHIFT) & ZLIB_HUFFMAN_BITS_MASK; + v = t & ZLIB_HUFFMAN_VALUE_MASK; + + if ((t & (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)) == 0) + { + lit = v; + val >>= b + 1; + bits -= b + 1; + } + else + { + t = tlit[v + 0x100 + ((val >> 8) & ((1U << b) - 1))]; + b = (t >> ZLIB_HUFFMAN_BITS_SHIFT) & ZLIB_HUFFMAN_BITS_MASK; + lit = t & ZLIB_HUFFMAN_VALUE_MASK; + val >>= b + 8; + bits -= b + 8; + } + + if (lit < 256) + { + if (unlikely (pout == poutend)) + { + elf_uncompress_failed (); + return 0; + } + + *pout++ = lit; + + /* We will need to write the next byte soon. We ask + for high temporal locality because we will write + to the whole cache line soon. */ + __builtin_prefetch (pout, 1, 3); + } + else if (lit == 256) + { + /* The end of the block. */ + break; + } + else + { + unsigned int dist; + unsigned int len; + + /* Convert lit into a length. */ + + if (lit < 265) + len = lit - 257 + 3; + else if (lit == 285) + len = 258; + else if (unlikely (lit > 285)) + { + elf_uncompress_failed (); + return 0; + } + else + { + unsigned int extra; + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + /* This is an expression for the table of length + codes in RFC 1951 3.2.5. */ + lit -= 265; + extra = (lit >> 2) + 1; + len = (lit & 3) << extra; + len += 11; + len += ((1U << (extra - 1)) - 1) << 3; + len += val & ((1U << extra) - 1); + val >>= extra; + bits -= extra; + } + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + t = tdist[val & 0xff]; + b = (t >> ZLIB_HUFFMAN_BITS_SHIFT) & ZLIB_HUFFMAN_BITS_MASK; + v = t & ZLIB_HUFFMAN_VALUE_MASK; + + if ((t & (1U << ZLIB_HUFFMAN_SECONDARY_SHIFT)) == 0) + { + dist = v; + val >>= b + 1; + bits -= b + 1; + } + else + { + t = tdist[v + 0x100 + ((val >> 8) & ((1U << b) - 1))]; + b = ((t >> ZLIB_HUFFMAN_BITS_SHIFT) + & ZLIB_HUFFMAN_BITS_MASK); + dist = t & ZLIB_HUFFMAN_VALUE_MASK; + val >>= b + 8; + bits -= b + 8; + } + + /* Convert dist to a distance. */ + + if (dist == 0) + { + /* A distance of 1. A common case, meaning + repeat the last character LEN times. */ + + if (unlikely (pout == porigout)) + { + elf_uncompress_failed (); + return 0; + } + + if (unlikely ((unsigned int) (poutend - pout) < len)) + { + elf_uncompress_failed (); + return 0; + } + + memset (pout, pout[-1], len); + pout += len; + } + else if (unlikely (dist > 29)) + { + elf_uncompress_failed (); + return 0; + } + else + { + if (dist < 4) + dist = dist + 1; + else + { + unsigned int extra; + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + /* This is an expression for the table of + distance codes in RFC 1951 3.2.5. */ + dist -= 4; + extra = (dist >> 1) + 1; + dist = (dist & 1) << extra; + dist += 5; + dist += ((1U << (extra - 1)) - 1) << 2; + dist += val & ((1U << extra) - 1); + val >>= extra; + bits -= extra; + } + + /* Go back dist bytes, and copy len bytes from + there. */ + + if (unlikely ((unsigned int) (pout - porigout) < dist)) + { + elf_uncompress_failed (); + return 0; + } + + if (unlikely ((unsigned int) (poutend - pout) < len)) + { + elf_uncompress_failed (); + return 0; + } + + if (dist >= len) + { + memcpy (pout, pout - dist, len); + pout += len; + } + else + { + while (len > 0) + { + unsigned int copy; + + copy = len < dist ? len : dist; + memcpy (pout, pout - dist, copy); + len -= copy; + pout += copy; + } + } + } + } + } + } + } + + /* We should have filled the output buffer. */ + if (unlikely (pout != poutend)) + { + elf_uncompress_failed (); + return 0; + } + + return 1; +} + +/* Verify the zlib checksum. The checksum is in the 4 bytes at + CHECKBYTES, and the uncompressed data is at UNCOMPRESSED / + UNCOMPRESSED_SIZE. Returns 1 on success, 0 on failure. */ + +static int +elf_zlib_verify_checksum (const unsigned char *checkbytes, + const unsigned char *uncompressed, + size_t uncompressed_size) +{ + unsigned int i; + unsigned int cksum; + const unsigned char *p; + uint32_t s1; + uint32_t s2; + size_t hsz; + + cksum = 0; + for (i = 0; i < 4; i++) + cksum = (cksum << 8) | checkbytes[i]; + + s1 = 1; + s2 = 0; + + /* Minimize modulo operations. */ + + p = uncompressed; + hsz = uncompressed_size; + while (hsz >= 5552) + { + for (i = 0; i < 5552; i += 16) + { + /* Manually unroll loop 16 times. */ + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + } + hsz -= 5552; + s1 %= 65521; + s2 %= 65521; + } + + while (hsz >= 16) + { + /* Manually unroll loop 16 times. */ + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + s1 = s1 + *p++; + s2 = s2 + s1; + + hsz -= 16; + } + + for (i = 0; i < hsz; ++i) + { + s1 = s1 + *p++; + s2 = s2 + s1; + } + + s1 %= 65521; + s2 %= 65521; + + if (unlikely ((s2 << 16) + s1 != cksum)) + { + elf_uncompress_failed (); + return 0; + } + + return 1; +} + +/* Inflate a zlib stream from PIN/SIN to POUT/SOUT, and verify the + checksum. Return 1 on success, 0 on error. */ + +static int +elf_zlib_inflate_and_verify (const unsigned char *pin, size_t sin, + uint16_t *zdebug_table, unsigned char *pout, + size_t sout) +{ + if (!elf_zlib_inflate (pin, sin, zdebug_table, pout, sout)) + return 0; + if (!elf_zlib_verify_checksum (pin + sin - 4, pout, sout)) + return 0; + return 1; +} + +/* For working memory during zstd compression, we need + - a literal length FSE table: 512 64-bit values == 4096 bytes + - a match length FSE table: 512 64-bit values == 4096 bytes + - a offset FSE table: 256 64-bit values == 2048 bytes + - a Huffman tree: 2048 uint16_t values == 4096 bytes + - scratch space, one of + - to build an FSE table: 512 uint16_t values == 1024 bytes + - to build a Huffman tree: 512 uint16_t + 256 uint32_t == 2048 bytes +*/ + +#define ZSTD_TABLE_SIZE \ + (2 * 512 * sizeof (struct elf_zstd_fse_baseline_entry) \ + + 256 * sizeof (struct elf_zstd_fse_baseline_entry) \ + + 2048 * sizeof (uint16_t) \ + + 512 * sizeof (uint16_t) + 256 * sizeof (uint32_t)) + +#define ZSTD_TABLE_LITERAL_FSE_OFFSET (0) + +#define ZSTD_TABLE_MATCH_FSE_OFFSET \ + (512 * sizeof (struct elf_zstd_fse_baseline_entry)) + +#define ZSTD_TABLE_OFFSET_FSE_OFFSET \ + (ZSTD_TABLE_MATCH_FSE_OFFSET \ + + 512 * sizeof (struct elf_zstd_fse_baseline_entry)) + +#define ZSTD_TABLE_HUFFMAN_OFFSET \ + (ZSTD_TABLE_OFFSET_FSE_OFFSET \ + + 256 * sizeof (struct elf_zstd_fse_baseline_entry)) + +#define ZSTD_TABLE_WORK_OFFSET \ + (ZSTD_TABLE_HUFFMAN_OFFSET + 2048 * sizeof (uint16_t)) + +/* An entry in a zstd FSE table. */ + +struct elf_zstd_fse_entry +{ + /* The value that this FSE entry represents. */ + unsigned char symbol; + /* The number of bits to read to determine the next state. */ + unsigned char bits; + /* Add the bits to this base to get the next state. */ + uint16_t base; +}; + +static int +elf_zstd_build_fse (const int16_t *, int, uint16_t *, int, + struct elf_zstd_fse_entry *); + +/* Read a zstd FSE table and build the decoding table in *TABLE, updating *PPIN + as it reads. ZDEBUG_TABLE is scratch space; it must be enough for 512 + uint16_t values (1024 bytes). MAXIDX is the maximum number of symbols + permitted. *TABLE_BITS is the maximum number of bits for symbols in the + table: the size of *TABLE is at least 1 << *TABLE_BITS. This updates + *TABLE_BITS to the actual number of bits. Returns 1 on success, 0 on + error. */ + +static int +elf_zstd_read_fse (const unsigned char **ppin, const unsigned char *pinend, + uint16_t *zdebug_table, int maxidx, + struct elf_zstd_fse_entry *table, int *table_bits) +{ + const unsigned char *pin; + int16_t *norm; + uint16_t *next; + uint64_t val; + unsigned int bits; + int accuracy_log; + uint32_t remaining; + uint32_t threshold; + int bits_needed; + int idx; + int prev0; + + pin = *ppin; + + norm = (int16_t *) zdebug_table; + next = zdebug_table + 256; + + if (unlikely (pin + 3 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + + /* Align PIN to a 32-bit boundary. */ + + val = 0; + bits = 0; + while ((((uintptr_t) pin) & 3) != 0) + { + val |= (uint64_t)*pin << bits; + bits += 8; + ++pin; + } + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + accuracy_log = (val & 0xf) + 5; + if (accuracy_log > *table_bits) + { + elf_uncompress_failed (); + return 0; + } + *table_bits = accuracy_log; + val >>= 4; + bits -= 4; + + /* This code is mostly copied from the reference implementation. */ + + /* The number of remaining probabilities, plus 1. This sets the number of + bits that need to be read for the next value. */ + remaining = (1 << accuracy_log) + 1; + + /* The current difference between small and large values, which depends on + the number of remaining values. Small values use one less bit. */ + threshold = 1 << accuracy_log; + + /* The number of bits used to compute threshold. */ + bits_needed = accuracy_log + 1; + + /* The next character value. */ + idx = 0; + + /* Whether the last count was 0. */ + prev0 = 0; + + while (remaining > 1 && idx <= maxidx) + { + uint32_t max; + int32_t count; + + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + + if (prev0) + { + int zidx; + + /* Previous count was 0, so there is a 2-bit repeat flag. If the + 2-bit flag is 0b11, it adds 3 and then there is another repeat + flag. */ + zidx = idx; + while ((val & 0xfff) == 0xfff) + { + zidx += 3 * 6; + val >>= 12; + bits -= 12; + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + } + while ((val & 3) == 3) + { + zidx += 3; + val >>= 2; + bits -= 2; + if (!elf_fetch_bits (&pin, pinend, &val, &bits)) + return 0; + } + /* We have at least 13 bits here, don't need to fetch. */ + zidx += val & 3; + val >>= 2; + bits -= 2; + + if (unlikely (zidx > maxidx)) + { + elf_uncompress_failed (); + return 0; + } + + for (; idx < zidx; idx++) + norm[idx] = 0; + + prev0 = 0; + continue; + } + + max = (2 * threshold - 1) - remaining; + if ((val & (threshold - 1)) < max) + { + /* A small value. */ + count = (int32_t) ((uint32_t) val & (threshold - 1)); + val >>= bits_needed - 1; + bits -= bits_needed - 1; + } + else + { + /* A large value. */ + count = (int32_t) ((uint32_t) val & (2 * threshold - 1)); + if (count >= (int32_t) threshold) + count -= (int32_t) max; + val >>= bits_needed; + bits -= bits_needed; + } + + count--; + if (count >= 0) + remaining -= count; + else + remaining--; + if (unlikely (idx >= 256)) + { + elf_uncompress_failed (); + return 0; + } + norm[idx] = (int16_t) count; + ++idx; + + prev0 = count == 0; + + while (remaining < threshold) + { + bits_needed--; + threshold >>= 1; + } + } + + if (unlikely (remaining != 1)) + { + elf_uncompress_failed (); + return 0; + } + + /* If we've read ahead more than a byte, back up. */ + while (bits >= 8) + { + --pin; + bits -= 8; + } + + *ppin = pin; + + for (; idx <= maxidx; idx++) + norm[idx] = 0; + + return elf_zstd_build_fse (norm, idx, next, *table_bits, table); +} + +/* Build the FSE decoding table from a list of probabilities. This reads from + NORM of length IDX, uses NEXT as scratch space, and writes to *TABLE, whose + size is TABLE_BITS. */ + +static int +elf_zstd_build_fse (const int16_t *norm, int idx, uint16_t *next, + int table_bits, struct elf_zstd_fse_entry *table) +{ + int table_size; + int high_threshold; + int i; + int pos; + int step; + int mask; + + table_size = 1 << table_bits; + high_threshold = table_size - 1; + for (i = 0; i < idx; i++) + { + int16_t n; + + n = norm[i]; + if (n >= 0) + next[i] = (uint16_t) n; + else + { + table[high_threshold].symbol = (unsigned char) i; + high_threshold--; + next[i] = 1; + } + } + + pos = 0; + step = (table_size >> 1) + (table_size >> 3) + 3; + mask = table_size - 1; + for (i = 0; i < idx; i++) + { + int n; + int j; + + n = (int) norm[i]; + for (j = 0; j < n; j++) + { + table[pos].symbol = (unsigned char) i; + pos = (pos + step) & mask; + while (unlikely (pos > high_threshold)) + pos = (pos + step) & mask; + } + } + if (unlikely (pos != 0)) + { + elf_uncompress_failed (); + return 0; + } + + for (i = 0; i < table_size; i++) + { + unsigned char sym; + uint16_t next_state; + int high_bit; + int bits; + + sym = table[i].symbol; + next_state = next[sym]; + ++next[sym]; + + if (next_state == 0) + { + elf_uncompress_failed (); + return 0; + } + high_bit = 31 - __builtin_clz (next_state); + + bits = table_bits - high_bit; + table[i].bits = (unsigned char) bits; + table[i].base = (uint16_t) ((next_state << bits) - table_size); + } + + return 1; +} + +/* Encode the baseline and bits into a single 32-bit value. */ + +#define ZSTD_ENCODE_BASELINE_BITS(baseline, basebits) \ + ((uint32_t)(baseline) | ((uint32_t)(basebits) << 24)) + +#define ZSTD_DECODE_BASELINE(baseline_basebits) \ + ((uint32_t)(baseline_basebits) & 0xffffff) + +#define ZSTD_DECODE_BASEBITS(baseline_basebits) \ + ((uint32_t)(baseline_basebits) >> 24) + +/* Given a literal length code, we need to read a number of bits and add that + to a baseline. For states 0 to 15 the baseline is the state and the number + of bits is zero. */ + +#define ZSTD_LITERAL_LENGTH_BASELINE_OFFSET (16) + +static const uint32_t elf_zstd_literal_length_base[] = +{ + ZSTD_ENCODE_BASELINE_BITS(16, 1), + ZSTD_ENCODE_BASELINE_BITS(18, 1), + ZSTD_ENCODE_BASELINE_BITS(20, 1), + ZSTD_ENCODE_BASELINE_BITS(22, 1), + ZSTD_ENCODE_BASELINE_BITS(24, 2), + ZSTD_ENCODE_BASELINE_BITS(28, 2), + ZSTD_ENCODE_BASELINE_BITS(32, 3), + ZSTD_ENCODE_BASELINE_BITS(40, 3), + ZSTD_ENCODE_BASELINE_BITS(48, 4), + ZSTD_ENCODE_BASELINE_BITS(64, 6), + ZSTD_ENCODE_BASELINE_BITS(128, 7), + ZSTD_ENCODE_BASELINE_BITS(256, 8), + ZSTD_ENCODE_BASELINE_BITS(512, 9), + ZSTD_ENCODE_BASELINE_BITS(1024, 10), + ZSTD_ENCODE_BASELINE_BITS(2048, 11), + ZSTD_ENCODE_BASELINE_BITS(4096, 12), + ZSTD_ENCODE_BASELINE_BITS(8192, 13), + ZSTD_ENCODE_BASELINE_BITS(16384, 14), + ZSTD_ENCODE_BASELINE_BITS(32768, 15), + ZSTD_ENCODE_BASELINE_BITS(65536, 16) +}; + +/* The same applies to match length codes. For states 0 to 31 the baseline is + the state + 3 and the number of bits is zero. */ + +#define ZSTD_MATCH_LENGTH_BASELINE_OFFSET (32) + +static const uint32_t elf_zstd_match_length_base[] = +{ + ZSTD_ENCODE_BASELINE_BITS(35, 1), + ZSTD_ENCODE_BASELINE_BITS(37, 1), + ZSTD_ENCODE_BASELINE_BITS(39, 1), + ZSTD_ENCODE_BASELINE_BITS(41, 1), + ZSTD_ENCODE_BASELINE_BITS(43, 2), + ZSTD_ENCODE_BASELINE_BITS(47, 2), + ZSTD_ENCODE_BASELINE_BITS(51, 3), + ZSTD_ENCODE_BASELINE_BITS(59, 3), + ZSTD_ENCODE_BASELINE_BITS(67, 4), + ZSTD_ENCODE_BASELINE_BITS(83, 4), + ZSTD_ENCODE_BASELINE_BITS(99, 5), + ZSTD_ENCODE_BASELINE_BITS(131, 7), + ZSTD_ENCODE_BASELINE_BITS(259, 8), + ZSTD_ENCODE_BASELINE_BITS(515, 9), + ZSTD_ENCODE_BASELINE_BITS(1027, 10), + ZSTD_ENCODE_BASELINE_BITS(2051, 11), + ZSTD_ENCODE_BASELINE_BITS(4099, 12), + ZSTD_ENCODE_BASELINE_BITS(8195, 13), + ZSTD_ENCODE_BASELINE_BITS(16387, 14), + ZSTD_ENCODE_BASELINE_BITS(32771, 15), + ZSTD_ENCODE_BASELINE_BITS(65539, 16) +}; + +/* An entry in an FSE table used for literal/match/length values. For these we + have to map the symbol to a baseline value, and we have to read zero or more + bits and add that value to the baseline value. Rather than look the values + up in a separate table, we grow the FSE table so that we get better memory + caching. */ + +struct elf_zstd_fse_baseline_entry +{ + /* The baseline for the value that this FSE entry represents.. */ + uint32_t baseline; + /* The number of bits to read to add to the baseline. */ + unsigned char basebits; + /* The number of bits to read to determine the next state. */ + unsigned char bits; + /* Add the bits to this base to get the next state. */ + uint16_t base; +}; + +/* Convert the literal length FSE table FSE_TABLE to an FSE baseline table at + BASELINE_TABLE. Note that FSE_TABLE and BASELINE_TABLE will overlap. */ + +static int +elf_zstd_make_literal_baseline_fse ( + const struct elf_zstd_fse_entry *fse_table, + int table_bits, + struct elf_zstd_fse_baseline_entry *baseline_table) +{ + size_t count; + const struct elf_zstd_fse_entry *pfse; + struct elf_zstd_fse_baseline_entry *pbaseline; + + /* Convert backward to avoid overlap. */ + + count = 1U << table_bits; + pfse = fse_table + count; + pbaseline = baseline_table + count; + while (pfse > fse_table) + { + unsigned char symbol; + unsigned char bits; + uint16_t base; + + --pfse; + --pbaseline; + symbol = pfse->symbol; + bits = pfse->bits; + base = pfse->base; + if (symbol < ZSTD_LITERAL_LENGTH_BASELINE_OFFSET) + { + pbaseline->baseline = (uint32_t)symbol; + pbaseline->basebits = 0; + } + else + { + unsigned int idx; + uint32_t basebits; + + if (unlikely (symbol > 35)) + { + elf_uncompress_failed (); + return 0; + } + idx = symbol - ZSTD_LITERAL_LENGTH_BASELINE_OFFSET; + basebits = elf_zstd_literal_length_base[idx]; + pbaseline->baseline = ZSTD_DECODE_BASELINE(basebits); + pbaseline->basebits = ZSTD_DECODE_BASEBITS(basebits); + } + pbaseline->bits = bits; + pbaseline->base = base; + } + + return 1; +} + +/* Convert the offset length FSE table FSE_TABLE to an FSE baseline table at + BASELINE_TABLE. Note that FSE_TABLE and BASELINE_TABLE will overlap. */ + +static int +elf_zstd_make_offset_baseline_fse ( + const struct elf_zstd_fse_entry *fse_table, + int table_bits, + struct elf_zstd_fse_baseline_entry *baseline_table) +{ + size_t count; + const struct elf_zstd_fse_entry *pfse; + struct elf_zstd_fse_baseline_entry *pbaseline; + + /* Convert backward to avoid overlap. */ + + count = 1U << table_bits; + pfse = fse_table + count; + pbaseline = baseline_table + count; + while (pfse > fse_table) + { + unsigned char symbol; + unsigned char bits; + uint16_t base; + + --pfse; + --pbaseline; + symbol = pfse->symbol; + bits = pfse->bits; + base = pfse->base; + if (unlikely (symbol > 31)) + { + elf_uncompress_failed (); + return 0; + } + + /* The simple way to write this is + + pbaseline->baseline = (uint32_t)1 << symbol; + pbaseline->basebits = symbol; + + That will give us an offset value that corresponds to the one + described in the RFC. However, for offset values > 3, we have to + subtract 3. And for offset values 1, 2, 3 we use a repeated offset. + The baseline is always a power of 2, and is never 0, so for these low + values we will see one entry that is baseline 1, basebits 0, and one + entry that is baseline 2, basebits 1. All other entries will have + baseline >= 4 and basebits >= 2. + + So we can check for RFC offset <= 3 by checking for basebits <= 1. + And that means that we can subtract 3 here and not worry about doing + it in the hot loop. */ + + pbaseline->baseline = (uint32_t)1 << symbol; + if (symbol >= 2) + pbaseline->baseline -= 3; + pbaseline->basebits = symbol; + pbaseline->bits = bits; + pbaseline->base = base; + } + + return 1; +} + +/* Convert the match length FSE table FSE_TABLE to an FSE baseline table at + BASELINE_TABLE. Note that FSE_TABLE and BASELINE_TABLE will overlap. */ + +static int +elf_zstd_make_match_baseline_fse ( + const struct elf_zstd_fse_entry *fse_table, + int table_bits, + struct elf_zstd_fse_baseline_entry *baseline_table) +{ + size_t count; + const struct elf_zstd_fse_entry *pfse; + struct elf_zstd_fse_baseline_entry *pbaseline; + + /* Convert backward to avoid overlap. */ + + count = 1U << table_bits; + pfse = fse_table + count; + pbaseline = baseline_table + count; + while (pfse > fse_table) + { + unsigned char symbol; + unsigned char bits; + uint16_t base; + + --pfse; + --pbaseline; + symbol = pfse->symbol; + bits = pfse->bits; + base = pfse->base; + if (symbol < ZSTD_MATCH_LENGTH_BASELINE_OFFSET) + { + pbaseline->baseline = (uint32_t)symbol + 3; + pbaseline->basebits = 0; + } + else + { + unsigned int idx; + uint32_t basebits; + + if (unlikely (symbol > 52)) + { + elf_uncompress_failed (); + return 0; + } + idx = symbol - ZSTD_MATCH_LENGTH_BASELINE_OFFSET; + basebits = elf_zstd_match_length_base[idx]; + pbaseline->baseline = ZSTD_DECODE_BASELINE(basebits); + pbaseline->basebits = ZSTD_DECODE_BASEBITS(basebits); + } + pbaseline->bits = bits; + pbaseline->base = base; + } + + return 1; +} + +#ifdef BACKTRACE_GENERATE_ZSTD_FSE_TABLES + +/* Used to generate the predefined FSE decoding tables for zstd. */ + +#include + +/* These values are straight from RFC 8878. */ + +static int16_t lit[36] = +{ + 4, 3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1, + 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 2, 1, 1, 1, 1, 1, + -1,-1,-1,-1 +}; + +static int16_t match[53] = +{ + 1, 4, 3, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,-1,-1, + -1,-1,-1,-1,-1 +}; + +static int16_t offset[29] = +{ + 1, 1, 1, 1, 1, 1, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1,-1,-1,-1,-1,-1 +}; + +static uint16_t next[256]; + +static void +print_table (const struct elf_zstd_fse_baseline_entry *table, size_t size) +{ + size_t i; + + printf ("{\n"); + for (i = 0; i < size; i += 3) + { + int j; + + printf (" "); + for (j = 0; j < 3 && i + j < size; ++j) + printf (" { %u, %d, %d, %d },", table[i + j].baseline, + table[i + j].basebits, table[i + j].bits, + table[i + j].base); + printf ("\n"); + } + printf ("};\n"); +} + +int +main () +{ + struct elf_zstd_fse_entry lit_table[64]; + struct elf_zstd_fse_baseline_entry lit_baseline[64]; + struct elf_zstd_fse_entry match_table[64]; + struct elf_zstd_fse_baseline_entry match_baseline[64]; + struct elf_zstd_fse_entry offset_table[32]; + struct elf_zstd_fse_baseline_entry offset_baseline[32]; + + if (!elf_zstd_build_fse (lit, sizeof lit / sizeof lit[0], next, + 6, lit_table)) + { + fprintf (stderr, "elf_zstd_build_fse failed\n"); + exit (EXIT_FAILURE); + } + + if (!elf_zstd_make_literal_baseline_fse (lit_table, 6, lit_baseline)) + { + fprintf (stderr, "elf_zstd_make_literal_baseline_fse failed\n"); + exit (EXIT_FAILURE); + } + + printf ("static const struct elf_zstd_fse_baseline_entry " + "elf_zstd_lit_table[64] =\n"); + print_table (lit_baseline, + sizeof lit_baseline / sizeof lit_baseline[0]); + printf ("\n"); + + if (!elf_zstd_build_fse (match, sizeof match / sizeof match[0], next, + 6, match_table)) + { + fprintf (stderr, "elf_zstd_build_fse failed\n"); + exit (EXIT_FAILURE); + } + + if (!elf_zstd_make_match_baseline_fse (match_table, 6, match_baseline)) + { + fprintf (stderr, "elf_zstd_make_match_baseline_fse failed\n"); + exit (EXIT_FAILURE); + } + + printf ("static const struct elf_zstd_fse_baseline_entry " + "elf_zstd_match_table[64] =\n"); + print_table (match_baseline, + sizeof match_baseline / sizeof match_baseline[0]); + printf ("\n"); + + if (!elf_zstd_build_fse (offset, sizeof offset / sizeof offset[0], next, + 5, offset_table)) + { + fprintf (stderr, "elf_zstd_build_fse failed\n"); + exit (EXIT_FAILURE); + } + + if (!elf_zstd_make_offset_baseline_fse (offset_table, 5, offset_baseline)) + { + fprintf (stderr, "elf_zstd_make_offset_baseline_fse failed\n"); + exit (EXIT_FAILURE); + } + + printf ("static const struct elf_zstd_fse_baseline_entry " + "elf_zstd_offset_table[32] =\n"); + print_table (offset_baseline, + sizeof offset_baseline / sizeof offset_baseline[0]); + printf ("\n"); + + return 0; +} + +#endif + +/* The fixed tables generated by the #ifdef'ed out main function + above. */ + +static const struct elf_zstd_fse_baseline_entry elf_zstd_lit_table[64] = +{ + { 0, 0, 4, 0 }, { 0, 0, 4, 16 }, { 1, 0, 5, 32 }, + { 3, 0, 5, 0 }, { 4, 0, 5, 0 }, { 6, 0, 5, 0 }, + { 7, 0, 5, 0 }, { 9, 0, 5, 0 }, { 10, 0, 5, 0 }, + { 12, 0, 5, 0 }, { 14, 0, 6, 0 }, { 16, 1, 5, 0 }, + { 20, 1, 5, 0 }, { 22, 1, 5, 0 }, { 28, 2, 5, 0 }, + { 32, 3, 5, 0 }, { 48, 4, 5, 0 }, { 64, 6, 5, 32 }, + { 128, 7, 5, 0 }, { 256, 8, 6, 0 }, { 1024, 10, 6, 0 }, + { 4096, 12, 6, 0 }, { 0, 0, 4, 32 }, { 1, 0, 4, 0 }, + { 2, 0, 5, 0 }, { 4, 0, 5, 32 }, { 5, 0, 5, 0 }, + { 7, 0, 5, 32 }, { 8, 0, 5, 0 }, { 10, 0, 5, 32 }, + { 11, 0, 5, 0 }, { 13, 0, 6, 0 }, { 16, 1, 5, 32 }, + { 18, 1, 5, 0 }, { 22, 1, 5, 32 }, { 24, 2, 5, 0 }, + { 32, 3, 5, 32 }, { 40, 3, 5, 0 }, { 64, 6, 4, 0 }, + { 64, 6, 4, 16 }, { 128, 7, 5, 32 }, { 512, 9, 6, 0 }, + { 2048, 11, 6, 0 }, { 0, 0, 4, 48 }, { 1, 0, 4, 16 }, + { 2, 0, 5, 32 }, { 3, 0, 5, 32 }, { 5, 0, 5, 32 }, + { 6, 0, 5, 32 }, { 8, 0, 5, 32 }, { 9, 0, 5, 32 }, + { 11, 0, 5, 32 }, { 12, 0, 5, 32 }, { 15, 0, 6, 0 }, + { 18, 1, 5, 32 }, { 20, 1, 5, 32 }, { 24, 2, 5, 32 }, + { 28, 2, 5, 32 }, { 40, 3, 5, 32 }, { 48, 4, 5, 32 }, + { 65536, 16, 6, 0 }, { 32768, 15, 6, 0 }, { 16384, 14, 6, 0 }, + { 8192, 13, 6, 0 }, +}; + +static const struct elf_zstd_fse_baseline_entry elf_zstd_match_table[64] = +{ + { 3, 0, 6, 0 }, { 4, 0, 4, 0 }, { 5, 0, 5, 32 }, + { 6, 0, 5, 0 }, { 8, 0, 5, 0 }, { 9, 0, 5, 0 }, + { 11, 0, 5, 0 }, { 13, 0, 6, 0 }, { 16, 0, 6, 0 }, + { 19, 0, 6, 0 }, { 22, 0, 6, 0 }, { 25, 0, 6, 0 }, + { 28, 0, 6, 0 }, { 31, 0, 6, 0 }, { 34, 0, 6, 0 }, + { 37, 1, 6, 0 }, { 41, 1, 6, 0 }, { 47, 2, 6, 0 }, + { 59, 3, 6, 0 }, { 83, 4, 6, 0 }, { 131, 7, 6, 0 }, + { 515, 9, 6, 0 }, { 4, 0, 4, 16 }, { 5, 0, 4, 0 }, + { 6, 0, 5, 32 }, { 7, 0, 5, 0 }, { 9, 0, 5, 32 }, + { 10, 0, 5, 0 }, { 12, 0, 6, 0 }, { 15, 0, 6, 0 }, + { 18, 0, 6, 0 }, { 21, 0, 6, 0 }, { 24, 0, 6, 0 }, + { 27, 0, 6, 0 }, { 30, 0, 6, 0 }, { 33, 0, 6, 0 }, + { 35, 1, 6, 0 }, { 39, 1, 6, 0 }, { 43, 2, 6, 0 }, + { 51, 3, 6, 0 }, { 67, 4, 6, 0 }, { 99, 5, 6, 0 }, + { 259, 8, 6, 0 }, { 4, 0, 4, 32 }, { 4, 0, 4, 48 }, + { 5, 0, 4, 16 }, { 7, 0, 5, 32 }, { 8, 0, 5, 32 }, + { 10, 0, 5, 32 }, { 11, 0, 5, 32 }, { 14, 0, 6, 0 }, + { 17, 0, 6, 0 }, { 20, 0, 6, 0 }, { 23, 0, 6, 0 }, + { 26, 0, 6, 0 }, { 29, 0, 6, 0 }, { 32, 0, 6, 0 }, + { 65539, 16, 6, 0 }, { 32771, 15, 6, 0 }, { 16387, 14, 6, 0 }, + { 8195, 13, 6, 0 }, { 4099, 12, 6, 0 }, { 2051, 11, 6, 0 }, + { 1027, 10, 6, 0 }, +}; + +static const struct elf_zstd_fse_baseline_entry elf_zstd_offset_table[32] = +{ + { 1, 0, 5, 0 }, { 61, 6, 4, 0 }, { 509, 9, 5, 0 }, + { 32765, 15, 5, 0 }, { 2097149, 21, 5, 0 }, { 5, 3, 5, 0 }, + { 125, 7, 4, 0 }, { 4093, 12, 5, 0 }, { 262141, 18, 5, 0 }, + { 8388605, 23, 5, 0 }, { 29, 5, 5, 0 }, { 253, 8, 4, 0 }, + { 16381, 14, 5, 0 }, { 1048573, 20, 5, 0 }, { 1, 2, 5, 0 }, + { 125, 7, 4, 16 }, { 2045, 11, 5, 0 }, { 131069, 17, 5, 0 }, + { 4194301, 22, 5, 0 }, { 13, 4, 5, 0 }, { 253, 8, 4, 16 }, + { 8189, 13, 5, 0 }, { 524285, 19, 5, 0 }, { 2, 1, 5, 0 }, + { 61, 6, 4, 16 }, { 1021, 10, 5, 0 }, { 65533, 16, 5, 0 }, + { 268435453, 28, 5, 0 }, { 134217725, 27, 5, 0 }, { 67108861, 26, 5, 0 }, + { 33554429, 25, 5, 0 }, { 16777213, 24, 5, 0 }, +}; + +/* Read a zstd Huffman table and build the decoding table in *TABLE, reading + and updating *PPIN. This sets *PTABLE_BITS to the number of bits of the + table, such that the table length is 1 << *TABLE_BITS. ZDEBUG_TABLE is + scratch space; it must be enough for 512 uint16_t values + 256 32-bit values + (2048 bytes). Returns 1 on success, 0 on error. */ + +static int +elf_zstd_read_huff (const unsigned char **ppin, const unsigned char *pinend, + uint16_t *zdebug_table, uint16_t *table, int *ptable_bits) +{ + const unsigned char *pin; + unsigned char hdr; + unsigned char *weights; + size_t count; + uint32_t *weight_mark; + size_t i; + uint32_t weight_mask; + size_t table_bits; + + pin = *ppin; + if (unlikely (pin >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + hdr = *pin; + ++pin; + + weights = (unsigned char *) zdebug_table; + + if (hdr < 128) + { + /* Table is compressed using FSE. */ + + struct elf_zstd_fse_entry *fse_table; + int fse_table_bits; + uint16_t *scratch; + const unsigned char *pfse; + const unsigned char *pback; + uint64_t val; + unsigned int bits; + unsigned int state1, state2; + + /* SCRATCH is used temporarily by elf_zstd_read_fse. It overlaps + WEIGHTS. */ + scratch = zdebug_table; + fse_table = (struct elf_zstd_fse_entry *) (scratch + 512); + fse_table_bits = 6; + + pfse = pin; + if (!elf_zstd_read_fse (&pfse, pinend, scratch, 255, fse_table, + &fse_table_bits)) + return 0; + + if (unlikely (pin + hdr > pinend)) + { + elf_uncompress_failed (); + return 0; + } + + /* We no longer need SCRATCH. Start recording weights. We need up to + 256 bytes of weights and 64 bytes of rank counts, so it won't overlap + FSE_TABLE. */ + + pback = pin + hdr - 1; + + if (!elf_fetch_backward_init (&pback, pfse, &val, &bits)) + return 0; + + bits -= fse_table_bits; + state1 = (val >> bits) & ((1U << fse_table_bits) - 1); + bits -= fse_table_bits; + state2 = (val >> bits) & ((1U << fse_table_bits) - 1); + + /* There are two independent FSE streams, tracked by STATE1 and STATE2. + We decode them alternately. */ + + count = 0; + while (1) + { + struct elf_zstd_fse_entry *pt; + uint64_t v; + + pt = &fse_table[state1]; + + if (unlikely (pin < pinend) && bits < pt->bits) + { + if (unlikely (count >= 254)) + { + elf_uncompress_failed (); + return 0; + } + weights[count] = (unsigned char) pt->symbol; + weights[count + 1] = (unsigned char) fse_table[state2].symbol; + count += 2; + break; + } + + if (unlikely (pt->bits == 0)) + v = 0; + else + { + if (!elf_fetch_bits_backward (&pback, pfse, &val, &bits)) + return 0; + + bits -= pt->bits; + v = (val >> bits) & (((uint64_t)1 << pt->bits) - 1); + } + + state1 = pt->base + v; + + if (unlikely (count >= 255)) + { + elf_uncompress_failed (); + return 0; + } + + weights[count] = pt->symbol; + ++count; + + pt = &fse_table[state2]; + + if (unlikely (pin < pinend && bits < pt->bits)) + { + if (unlikely (count >= 254)) + { + elf_uncompress_failed (); + return 0; + } + weights[count] = (unsigned char) pt->symbol; + weights[count + 1] = (unsigned char) fse_table[state1].symbol; + count += 2; + break; + } + + if (unlikely (pt->bits == 0)) + v = 0; + else + { + if (!elf_fetch_bits_backward (&pback, pfse, &val, &bits)) + return 0; + + bits -= pt->bits; + v = (val >> bits) & (((uint64_t)1 << pt->bits) - 1); + } + + state2 = pt->base + v; + + if (unlikely (count >= 255)) + { + elf_uncompress_failed (); + return 0; + } + + weights[count] = pt->symbol; + ++count; + } + + pin += hdr; + } + else + { + /* Table is not compressed. Each weight is 4 bits. */ + + count = hdr - 127; + if (unlikely (pin + ((count + 1) / 2) >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + for (i = 0; i < count; i += 2) + { + unsigned char b; + + b = *pin; + ++pin; + weights[i] = b >> 4; + weights[i + 1] = b & 0xf; + } + } + + weight_mark = (uint32_t *) (weights + 256); + memset (weight_mark, 0, 13 * sizeof (uint32_t)); + weight_mask = 0; + for (i = 0; i < count; ++i) + { + unsigned char w; + + w = weights[i]; + if (unlikely (w > 12)) + { + elf_uncompress_failed (); + return 0; + } + ++weight_mark[w]; + if (w > 0) + weight_mask += 1U << (w - 1); + } + if (unlikely (weight_mask == 0)) + { + elf_uncompress_failed (); + return 0; + } + + table_bits = 32 - __builtin_clz (weight_mask); + if (unlikely (table_bits > 11)) + { + elf_uncompress_failed (); + return 0; + } + + /* Work out the last weight value, which is omitted because the weights must + sum to a power of two. */ + { + uint32_t left; + uint32_t high_bit; + + left = ((uint32_t)1 << table_bits) - weight_mask; + if (left == 0) + { + elf_uncompress_failed (); + return 0; + } + high_bit = 31 - __builtin_clz (left); + if (((uint32_t)1 << high_bit) != left) + { + elf_uncompress_failed (); + return 0; + } + + if (unlikely (count >= 256)) + { + elf_uncompress_failed (); + return 0; + } + + weights[count] = high_bit + 1; + ++count; + ++weight_mark[high_bit + 1]; + } + + if (weight_mark[1] < 2 || (weight_mark[1] & 1) != 0) + { + elf_uncompress_failed (); + return 0; + } + + /* Change WEIGHT_MARK from a count of weights to the index of the first + symbol for that weight. We shift the indexes to also store how many we + have seen so far, below. */ + { + uint32_t next; + + next = 0; + for (i = 0; i < table_bits; ++i) + { + uint32_t cur; + + cur = next; + next += weight_mark[i + 1] << i; + weight_mark[i + 1] = cur; + } + } + + for (i = 0; i < count; ++i) + { + unsigned char weight; + uint32_t length; + uint16_t tval; + size_t start; + uint32_t j; + + weight = weights[i]; + if (weight == 0) + continue; + + length = 1U << (weight - 1); + tval = (i << 8) | (table_bits + 1 - weight); + start = weight_mark[weight]; + for (j = 0; j < length; ++j) + table[start + j] = tval; + weight_mark[weight] += length; + } + + *ppin = pin; + *ptable_bits = (int)table_bits; + + return 1; +} + +/* Read and decompress the literals and store them ending at POUTEND. This + works because we are going to use all the literals in the output, so they + must fit into the output buffer. HUFFMAN_TABLE, and PHUFFMAN_TABLE_BITS + store the Huffman table across calls. SCRATCH is used to read a Huffman + table. Store the start of the decompressed literals in *PPLIT. Update + *PPIN. Return 1 on success, 0 on error. */ + +static int +elf_zstd_read_literals (const unsigned char **ppin, + const unsigned char *pinend, + unsigned char *pout, + unsigned char *poutend, + uint16_t *scratch, + uint16_t *huffman_table, + int *phuffman_table_bits, + unsigned char **pplit) +{ + const unsigned char *pin; + unsigned char *plit; + unsigned char hdr; + uint32_t regenerated_size; + uint32_t compressed_size; + int streams; + uint32_t total_streams_size; + unsigned int huffman_table_bits; + uint64_t huffman_mask; + + pin = *ppin; + if (unlikely (pin >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + hdr = *pin; + ++pin; + + if ((hdr & 3) == 0 || (hdr & 3) == 1) + { + int raw; + + /* Raw_Literals_Block or RLE_Literals_Block */ + + raw = (hdr & 3) == 0; + + switch ((hdr >> 2) & 3) + { + case 0: case 2: + regenerated_size = hdr >> 3; + break; + case 1: + if (unlikely (pin >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + regenerated_size = (hdr >> 4) + ((uint32_t)(*pin) << 4); + ++pin; + break; + case 3: + if (unlikely (pin + 1 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + regenerated_size = ((hdr >> 4) + + ((uint32_t)*pin << 4) + + ((uint32_t)pin[1] << 12)); + pin += 2; + break; + default: + elf_uncompress_failed (); + return 0; + } + + if (unlikely ((size_t)(poutend - pout) < regenerated_size)) + { + elf_uncompress_failed (); + return 0; + } + + plit = poutend - regenerated_size; + + if (raw) + { + if (unlikely (pin + regenerated_size >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + memcpy (plit, pin, regenerated_size); + pin += regenerated_size; + } + else + { + if (pin >= pinend) + { + elf_uncompress_failed (); + return 0; + } + memset (plit, *pin, regenerated_size); + ++pin; + } + + *ppin = pin; + *pplit = plit; + + return 1; + } + + /* Compressed_Literals_Block or Treeless_Literals_Block */ + + switch ((hdr >> 2) & 3) + { + case 0: case 1: + if (unlikely (pin + 1 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + regenerated_size = (hdr >> 4) | ((uint32_t)(*pin & 0x3f) << 4); + compressed_size = (uint32_t)*pin >> 6 | ((uint32_t)pin[1] << 2); + pin += 2; + streams = ((hdr >> 2) & 3) == 0 ? 1 : 4; + break; + case 2: + if (unlikely (pin + 2 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + regenerated_size = (((uint32_t)hdr >> 4) + | ((uint32_t)*pin << 4) + | (((uint32_t)pin[1] & 3) << 12)); + compressed_size = (((uint32_t)pin[1] >> 2) + | ((uint32_t)pin[2] << 6)); + pin += 3; + streams = 4; + break; + case 3: + if (unlikely (pin + 3 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + regenerated_size = (((uint32_t)hdr >> 4) + | ((uint32_t)*pin << 4) + | (((uint32_t)pin[1] & 0x3f) << 12)); + compressed_size = (((uint32_t)pin[1] >> 6) + | ((uint32_t)pin[2] << 2) + | ((uint32_t)pin[3] << 10)); + pin += 4; + streams = 4; + break; + default: + elf_uncompress_failed (); + return 0; + } + + if (unlikely (pin + compressed_size > pinend)) + { + elf_uncompress_failed (); + return 0; + } + + pinend = pin + compressed_size; + *ppin = pinend; + + if (unlikely ((size_t)(poutend - pout) < regenerated_size)) + { + elf_uncompress_failed (); + return 0; + } + + plit = poutend - regenerated_size; + + *pplit = plit; + + total_streams_size = compressed_size; + if ((hdr & 3) == 2) + { + const unsigned char *ptable; + + /* Compressed_Literals_Block. Read Huffman tree. */ + + ptable = pin; + if (!elf_zstd_read_huff (&ptable, pinend, scratch, huffman_table, + phuffman_table_bits)) + return 0; + + if (unlikely (total_streams_size < (size_t)(ptable - pin))) + { + elf_uncompress_failed (); + return 0; + } + + total_streams_size -= ptable - pin; + pin = ptable; + } + else + { + /* Treeless_Literals_Block. Reuse previous Huffman tree. */ + if (unlikely (*phuffman_table_bits == 0)) + { + elf_uncompress_failed (); + return 0; + } + } + + /* Decompress COMPRESSED_SIZE bytes of data at PIN using the huffman table, + storing REGENERATED_SIZE bytes of decompressed data at PLIT. */ + + huffman_table_bits = (unsigned int)*phuffman_table_bits; + huffman_mask = ((uint64_t)1 << huffman_table_bits) - 1; + + if (streams == 1) + { + const unsigned char *pback; + const unsigned char *pbackend; + uint64_t val; + unsigned int bits; + uint32_t i; + + pback = pin + total_streams_size - 1; + pbackend = pin; + if (!elf_fetch_backward_init (&pback, pbackend, &val, &bits)) + return 0; + + /* This is one of the inner loops of the decompression algorithm, so we + put some effort into optimization. We can't get more than 64 bytes + from a single call to elf_fetch_bits_backward, and we can't subtract + more than 11 bits at a time. */ + + if (regenerated_size >= 64) + { + unsigned char *plitstart; + unsigned char *plitstop; + + plitstart = plit; + plitstop = plit + regenerated_size - 64; + while (plit < plitstop) + { + uint16_t t; + + if (!elf_fetch_bits_backward (&pback, pbackend, &val, &bits)) + return 0; + + if (bits < 16) + break; + + while (bits >= 33) + { + t = huffman_table[(val >> (bits - huffman_table_bits)) + & huffman_mask]; + *plit = t >> 8; + ++plit; + bits -= t & 0xff; + + t = huffman_table[(val >> (bits - huffman_table_bits)) + & huffman_mask]; + *plit = t >> 8; + ++plit; + bits -= t & 0xff; + + t = huffman_table[(val >> (bits - huffman_table_bits)) + & huffman_mask]; + *plit = t >> 8; + ++plit; + bits -= t & 0xff; + } + + while (bits > 11) + { + t = huffman_table[(val >> (bits - huffman_table_bits)) + & huffman_mask]; + *plit = t >> 8; + ++plit; + bits -= t & 0xff; + } + } + + regenerated_size -= plit - plitstart; + } + + for (i = 0; i < regenerated_size; ++i) + { + uint16_t t; + + if (!elf_fetch_bits_backward (&pback, pbackend, &val, &bits)) + return 0; + + if (unlikely (bits < huffman_table_bits)) + { + t = huffman_table[(val << (huffman_table_bits - bits)) + & huffman_mask]; + if (unlikely (bits < (t & 0xff))) + { + elf_uncompress_failed (); + return 0; + } + } + else + t = huffman_table[(val >> (bits - huffman_table_bits)) + & huffman_mask]; + + *plit = t >> 8; + ++plit; + bits -= t & 0xff; + } + + return 1; + } + + { + uint32_t stream_size1, stream_size2, stream_size3, stream_size4; + uint32_t tot; + const unsigned char *pback1, *pback2, *pback3, *pback4; + const unsigned char *pbackend1, *pbackend2, *pbackend3, *pbackend4; + uint64_t val1, val2, val3, val4; + unsigned int bits1, bits2, bits3, bits4; + unsigned char *plit1, *plit2, *plit3, *plit4; + uint32_t regenerated_stream_size; + uint32_t regenerated_stream_size4; + uint16_t t1, t2, t3, t4; + uint32_t i; + uint32_t limit; + + /* Read jump table. */ + if (unlikely (pin + 5 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + stream_size1 = (uint32_t)*pin | ((uint32_t)pin[1] << 8); + pin += 2; + stream_size2 = (uint32_t)*pin | ((uint32_t)pin[1] << 8); + pin += 2; + stream_size3 = (uint32_t)*pin | ((uint32_t)pin[1] << 8); + pin += 2; + tot = stream_size1 + stream_size2 + stream_size3; + if (unlikely (tot > total_streams_size - 6)) + { + elf_uncompress_failed (); + return 0; + } + stream_size4 = total_streams_size - 6 - tot; + + pback1 = pin + stream_size1 - 1; + pbackend1 = pin; + + pback2 = pback1 + stream_size2; + pbackend2 = pback1 + 1; + + pback3 = pback2 + stream_size3; + pbackend3 = pback2 + 1; + + pback4 = pback3 + stream_size4; + pbackend4 = pback3 + 1; + + if (!elf_fetch_backward_init (&pback1, pbackend1, &val1, &bits1)) + return 0; + if (!elf_fetch_backward_init (&pback2, pbackend2, &val2, &bits2)) + return 0; + if (!elf_fetch_backward_init (&pback3, pbackend3, &val3, &bits3)) + return 0; + if (!elf_fetch_backward_init (&pback4, pbackend4, &val4, &bits4)) + return 0; + + regenerated_stream_size = (regenerated_size + 3) / 4; + + plit1 = plit; + plit2 = plit1 + regenerated_stream_size; + plit3 = plit2 + regenerated_stream_size; + plit4 = plit3 + regenerated_stream_size; + + regenerated_stream_size4 = regenerated_size - regenerated_stream_size * 3; + + /* We can't get more than 64 literal bytes from a single call to + elf_fetch_bits_backward. The fourth stream can be up to 3 bytes less, + so use as the limit. */ + + limit = regenerated_stream_size4 <= 64 ? 0 : regenerated_stream_size4 - 64; + i = 0; + while (i < limit) + { + if (!elf_fetch_bits_backward (&pback1, pbackend1, &val1, &bits1)) + return 0; + if (!elf_fetch_bits_backward (&pback2, pbackend2, &val2, &bits2)) + return 0; + if (!elf_fetch_bits_backward (&pback3, pbackend3, &val3, &bits3)) + return 0; + if (!elf_fetch_bits_backward (&pback4, pbackend4, &val4, &bits4)) + return 0; + + /* We can't subtract more than 11 bits at a time. */ + + do + { + t1 = huffman_table[(val1 >> (bits1 - huffman_table_bits)) + & huffman_mask]; + t2 = huffman_table[(val2 >> (bits2 - huffman_table_bits)) + & huffman_mask]; + t3 = huffman_table[(val3 >> (bits3 - huffman_table_bits)) + & huffman_mask]; + t4 = huffman_table[(val4 >> (bits4 - huffman_table_bits)) + & huffman_mask]; + + *plit1 = t1 >> 8; + ++plit1; + bits1 -= t1 & 0xff; + + *plit2 = t2 >> 8; + ++plit2; + bits2 -= t2 & 0xff; + + *plit3 = t3 >> 8; + ++plit3; + bits3 -= t3 & 0xff; + + *plit4 = t4 >> 8; + ++plit4; + bits4 -= t4 & 0xff; + + ++i; + } + while (bits1 > 11 && bits2 > 11 && bits3 > 11 && bits4 > 11); + } + + while (i < regenerated_stream_size) + { + int use4; + + use4 = i < regenerated_stream_size4; + + if (!elf_fetch_bits_backward (&pback1, pbackend1, &val1, &bits1)) + return 0; + if (!elf_fetch_bits_backward (&pback2, pbackend2, &val2, &bits2)) + return 0; + if (!elf_fetch_bits_backward (&pback3, pbackend3, &val3, &bits3)) + return 0; + if (use4) + { + if (!elf_fetch_bits_backward (&pback4, pbackend4, &val4, &bits4)) + return 0; + } + + if (unlikely (bits1 < huffman_table_bits)) + { + t1 = huffman_table[(val1 << (huffman_table_bits - bits1)) + & huffman_mask]; + if (unlikely (bits1 < (t1 & 0xff))) + { + elf_uncompress_failed (); + return 0; + } + } + else + t1 = huffman_table[(val1 >> (bits1 - huffman_table_bits)) + & huffman_mask]; + + if (unlikely (bits2 < huffman_table_bits)) + { + t2 = huffman_table[(val2 << (huffman_table_bits - bits2)) + & huffman_mask]; + if (unlikely (bits2 < (t2 & 0xff))) + { + elf_uncompress_failed (); + return 0; + } + } + else + t2 = huffman_table[(val2 >> (bits2 - huffman_table_bits)) + & huffman_mask]; + + if (unlikely (bits3 < huffman_table_bits)) + { + t3 = huffman_table[(val3 << (huffman_table_bits - bits3)) + & huffman_mask]; + if (unlikely (bits3 < (t3 & 0xff))) + { + elf_uncompress_failed (); + return 0; + } + } + else + t3 = huffman_table[(val3 >> (bits3 - huffman_table_bits)) + & huffman_mask]; + + if (use4) + { + if (unlikely (bits4 < huffman_table_bits)) + { + t4 = huffman_table[(val4 << (huffman_table_bits - bits4)) + & huffman_mask]; + if (unlikely (bits4 < (t4 & 0xff))) + { + elf_uncompress_failed (); + return 0; + } + } + else + t4 = huffman_table[(val4 >> (bits4 - huffman_table_bits)) + & huffman_mask]; + + *plit4 = t4 >> 8; + ++plit4; + bits4 -= t4 & 0xff; + } + + *plit1 = t1 >> 8; + ++plit1; + bits1 -= t1 & 0xff; + + *plit2 = t2 >> 8; + ++plit2; + bits2 -= t2 & 0xff; + + *plit3 = t3 >> 8; + ++plit3; + bits3 -= t3 & 0xff; + + ++i; + } + } + + return 1; +} + +/* The information used to decompress a sequence code, which can be a literal + length, an offset, or a match length. */ + +struct elf_zstd_seq_decode +{ + const struct elf_zstd_fse_baseline_entry *table; + int table_bits; +}; + +/* Unpack a sequence code compression mode. */ + +static int +elf_zstd_unpack_seq_decode (int mode, + const unsigned char **ppin, + const unsigned char *pinend, + const struct elf_zstd_fse_baseline_entry *predef, + int predef_bits, + uint16_t *scratch, + int maxidx, + struct elf_zstd_fse_baseline_entry *table, + int table_bits, + int (*conv)(const struct elf_zstd_fse_entry *, + int, + struct elf_zstd_fse_baseline_entry *), + struct elf_zstd_seq_decode *decode) +{ + switch (mode) + { + case 0: + decode->table = predef; + decode->table_bits = predef_bits; + break; + + case 1: + { + struct elf_zstd_fse_entry entry; + + if (unlikely (*ppin >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + entry.symbol = **ppin; + ++*ppin; + entry.bits = 0; + entry.base = 0; + decode->table_bits = 0; + if (!conv (&entry, 0, table)) + return 0; + } + break; + + case 2: + { + struct elf_zstd_fse_entry *fse_table; + + /* We use the same space for the simple FSE table and the baseline + table. */ + fse_table = (struct elf_zstd_fse_entry *)table; + decode->table_bits = table_bits; + if (!elf_zstd_read_fse (ppin, pinend, scratch, maxidx, fse_table, + &decode->table_bits)) + return 0; + if (!conv (fse_table, decode->table_bits, table)) + return 0; + decode->table = table; + } + break; + + case 3: + if (unlikely (decode->table_bits == -1)) + { + elf_uncompress_failed (); + return 0; + } + break; + + default: + elf_uncompress_failed (); + return 0; + } + + return 1; +} + +/* Decompress a zstd stream from PIN/SIN to POUT/SOUT. Code based on RFC 8878. + Return 1 on success, 0 on error. */ + +static int +elf_zstd_decompress (const unsigned char *pin, size_t sin, + unsigned char *zdebug_table, unsigned char *pout, + size_t sout) +{ + const unsigned char *pinend; + unsigned char *poutstart; + unsigned char *poutend; + struct elf_zstd_seq_decode literal_decode; + struct elf_zstd_fse_baseline_entry *literal_fse_table; + struct elf_zstd_seq_decode match_decode; + struct elf_zstd_fse_baseline_entry *match_fse_table; + struct elf_zstd_seq_decode offset_decode; + struct elf_zstd_fse_baseline_entry *offset_fse_table; + uint16_t *huffman_table; + int huffman_table_bits; + uint32_t repeated_offset1; + uint32_t repeated_offset2; + uint32_t repeated_offset3; + uint16_t *scratch; + unsigned char hdr; + int has_checksum; + uint64_t content_size; + int last_block; + + pinend = pin + sin; + poutstart = pout; + poutend = pout + sout; + + literal_decode.table = NULL; + literal_decode.table_bits = -1; + literal_fse_table = ((struct elf_zstd_fse_baseline_entry *) + (zdebug_table + ZSTD_TABLE_LITERAL_FSE_OFFSET)); + + match_decode.table = NULL; + match_decode.table_bits = -1; + match_fse_table = ((struct elf_zstd_fse_baseline_entry *) + (zdebug_table + ZSTD_TABLE_MATCH_FSE_OFFSET)); + + offset_decode.table = NULL; + offset_decode.table_bits = -1; + offset_fse_table = ((struct elf_zstd_fse_baseline_entry *) + (zdebug_table + ZSTD_TABLE_OFFSET_FSE_OFFSET)); + huffman_table = ((uint16_t *) + (zdebug_table + ZSTD_TABLE_HUFFMAN_OFFSET)); + huffman_table_bits = 0; + scratch = ((uint16_t *) + (zdebug_table + ZSTD_TABLE_WORK_OFFSET)); + + repeated_offset1 = 1; + repeated_offset2 = 4; + repeated_offset3 = 8; + + if (unlikely (sin < 4)) + { + elf_uncompress_failed (); + return 0; + } + + /* These values are the zstd magic number. */ + if (unlikely (pin[0] != 0x28 + || pin[1] != 0xb5 + || pin[2] != 0x2f + || pin[3] != 0xfd)) + { + elf_uncompress_failed (); + return 0; + } + + pin += 4; + + if (unlikely (pin >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + + hdr = *pin++; + + /* We expect a single frame. */ + if (unlikely ((hdr & (1 << 5)) == 0)) + { + elf_uncompress_failed (); + return 0; + } + /* Reserved bit must be zero. */ + if (unlikely ((hdr & (1 << 3)) != 0)) + { + elf_uncompress_failed (); + return 0; + } + /* We do not expect a dictionary. */ + if (unlikely ((hdr & 3) != 0)) + { + elf_uncompress_failed (); + return 0; + } + has_checksum = (hdr & (1 << 2)) != 0; + switch (hdr >> 6) + { + case 0: + if (unlikely (pin >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + content_size = (uint64_t) *pin++; + break; + case 1: + if (unlikely (pin + 1 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + content_size = (((uint64_t) pin[0]) | (((uint64_t) pin[1]) << 8)) + 256; + pin += 2; + break; + case 2: + if (unlikely (pin + 3 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + content_size = ((uint64_t) pin[0] + | (((uint64_t) pin[1]) << 8) + | (((uint64_t) pin[2]) << 16) + | (((uint64_t) pin[3]) << 24)); + pin += 4; + break; + case 3: + if (unlikely (pin + 7 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + content_size = ((uint64_t) pin[0] + | (((uint64_t) pin[1]) << 8) + | (((uint64_t) pin[2]) << 16) + | (((uint64_t) pin[3]) << 24) + | (((uint64_t) pin[4]) << 32) + | (((uint64_t) pin[5]) << 40) + | (((uint64_t) pin[6]) << 48) + | (((uint64_t) pin[7]) << 56)); + pin += 8; + break; + default: + elf_uncompress_failed (); + return 0; + } + + if (unlikely (content_size != (size_t) content_size + || (size_t) content_size != sout)) + { + elf_uncompress_failed (); + return 0; + } + + last_block = 0; + while (!last_block) + { + uint32_t block_hdr; + int block_type; + uint32_t block_size; + + if (unlikely (pin + 2 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + block_hdr = ((uint32_t) pin[0] + | (((uint32_t) pin[1]) << 8) + | (((uint32_t) pin[2]) << 16)); + pin += 3; + + last_block = block_hdr & 1; + block_type = (block_hdr >> 1) & 3; + block_size = block_hdr >> 3; + + switch (block_type) + { + case 0: + /* Raw_Block */ + if (unlikely ((size_t) block_size > (size_t) (pinend - pin))) + { + elf_uncompress_failed (); + return 0; + } + if (unlikely ((size_t) block_size > (size_t) (poutend - pout))) + { + elf_uncompress_failed (); + return 0; + } + memcpy (pout, pin, block_size); + pout += block_size; + pin += block_size; + break; + + case 1: + /* RLE_Block */ + if (unlikely (pin >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + if (unlikely ((size_t) block_size > (size_t) (poutend - pout))) + { + elf_uncompress_failed (); + return 0; + } + memset (pout, *pin, block_size); + pout += block_size; + pin++; + break; + + case 2: + { + const unsigned char *pblockend; + unsigned char *plitstack; + unsigned char *plit; + uint32_t literal_count; + unsigned char seq_hdr; + size_t seq_count; + size_t seq; + const unsigned char *pback; + uint64_t val; + unsigned int bits; + unsigned int literal_state; + unsigned int offset_state; + unsigned int match_state; + + /* Compressed_Block */ + if (unlikely ((size_t) block_size > (size_t) (pinend - pin))) + { + elf_uncompress_failed (); + return 0; + } + + pblockend = pin + block_size; + + /* Read the literals into the end of the output space, and leave + PLIT pointing at them. */ + + if (!elf_zstd_read_literals (&pin, pblockend, pout, poutend, + scratch, huffman_table, + &huffman_table_bits, + &plitstack)) + return 0; + plit = plitstack; + literal_count = poutend - plit; + + seq_hdr = *pin; + pin++; + if (seq_hdr < 128) + seq_count = seq_hdr; + else if (seq_hdr < 255) + { + if (unlikely (pin >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + seq_count = ((seq_hdr - 128) << 8) + *pin; + pin++; + } + else + { + if (unlikely (pin + 1 >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + seq_count = *pin + (pin[1] << 8) + 0x7f00; + pin += 2; + } + + if (seq_count > 0) + { + int (*pfn)(const struct elf_zstd_fse_entry *, + int, struct elf_zstd_fse_baseline_entry *); + + if (unlikely (pin >= pinend)) + { + elf_uncompress_failed (); + return 0; + } + seq_hdr = *pin; + ++pin; + + pfn = elf_zstd_make_literal_baseline_fse; + if (!elf_zstd_unpack_seq_decode ((seq_hdr >> 6) & 3, + &pin, pinend, + &elf_zstd_lit_table[0], 6, + scratch, 35, + literal_fse_table, 9, pfn, + &literal_decode)) + return 0; + + pfn = elf_zstd_make_offset_baseline_fse; + if (!elf_zstd_unpack_seq_decode ((seq_hdr >> 4) & 3, + &pin, pinend, + &elf_zstd_offset_table[0], 5, + scratch, 31, + offset_fse_table, 8, pfn, + &offset_decode)) + return 0; + + pfn = elf_zstd_make_match_baseline_fse; + if (!elf_zstd_unpack_seq_decode ((seq_hdr >> 2) & 3, + &pin, pinend, + &elf_zstd_match_table[0], 6, + scratch, 52, + match_fse_table, 9, pfn, + &match_decode)) + return 0; + } + + pback = pblockend - 1; + if (!elf_fetch_backward_init (&pback, pin, &val, &bits)) + return 0; + + bits -= literal_decode.table_bits; + literal_state = ((val >> bits) + & ((1U << literal_decode.table_bits) - 1)); + + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) + return 0; + bits -= offset_decode.table_bits; + offset_state = ((val >> bits) + & ((1U << offset_decode.table_bits) - 1)); + + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) + return 0; + bits -= match_decode.table_bits; + match_state = ((val >> bits) + & ((1U << match_decode.table_bits) - 1)); + + seq = 0; + while (1) + { + const struct elf_zstd_fse_baseline_entry *pt; + uint32_t offset_basebits; + uint32_t offset_baseline; + uint32_t offset_bits; + uint32_t offset_base; + uint32_t offset; + uint32_t match_baseline; + uint32_t match_bits; + uint32_t match_base; + uint32_t match; + uint32_t literal_baseline; + uint32_t literal_bits; + uint32_t literal_base; + uint32_t literal; + uint32_t need; + uint32_t add; + + pt = &offset_decode.table[offset_state]; + offset_basebits = pt->basebits; + offset_baseline = pt->baseline; + offset_bits = pt->bits; + offset_base = pt->base; + + /* This case can be more than 16 bits, which is all that + elf_fetch_bits_backward promises. */ + need = offset_basebits; + add = 0; + if (unlikely (need > 16)) + { + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) + return 0; + bits -= 16; + add = (val >> bits) & ((1U << 16) - 1); + need -= 16; + add <<= need; + } + if (need > 0) + { + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) + return 0; + bits -= need; + add += (val >> bits) & ((1U << need) - 1); + } + + offset = offset_baseline + add; + + pt = &match_decode.table[match_state]; + need = pt->basebits; + match_baseline = pt->baseline; + match_bits = pt->bits; + match_base = pt->base; + + add = 0; + if (need > 0) + { + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) + return 0; + bits -= need; + add = (val >> bits) & ((1U << need) - 1); + } + + match = match_baseline + add; + + pt = &literal_decode.table[literal_state]; + need = pt->basebits; + literal_baseline = pt->baseline; + literal_bits = pt->bits; + literal_base = pt->base; + + add = 0; + if (need > 0) + { + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) + return 0; + bits -= need; + add = (val >> bits) & ((1U << need) - 1); + } + + literal = literal_baseline + add; + + /* See the comment in elf_zstd_make_offset_baseline_fse. */ + if (offset_basebits > 1) + { + repeated_offset3 = repeated_offset2; + repeated_offset2 = repeated_offset1; + repeated_offset1 = offset; + } + else + { + if (unlikely (literal == 0)) + ++offset; + switch (offset) + { + case 1: + offset = repeated_offset1; + break; + case 2: + offset = repeated_offset2; + repeated_offset2 = repeated_offset1; + repeated_offset1 = offset; + break; + case 3: + offset = repeated_offset3; + repeated_offset3 = repeated_offset2; + repeated_offset2 = repeated_offset1; + repeated_offset1 = offset; + break; + case 4: + offset = repeated_offset1 - 1; + repeated_offset3 = repeated_offset2; + repeated_offset2 = repeated_offset1; + repeated_offset1 = offset; + break; + } + } + + ++seq; + if (seq < seq_count) + { + uint32_t v; + + /* Update the three states. */ + + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) + return 0; + + need = literal_bits; + bits -= need; + v = (val >> bits) & (((uint32_t)1 << need) - 1); + + literal_state = literal_base + v; + + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) + return 0; + + need = match_bits; + bits -= need; + v = (val >> bits) & (((uint32_t)1 << need) - 1); + + match_state = match_base + v; + + if (!elf_fetch_bits_backward (&pback, pin, &val, &bits)) + return 0; + + need = offset_bits; + bits -= need; + v = (val >> bits) & (((uint32_t)1 << need) - 1); + + offset_state = offset_base + v; + } + + /* The next sequence is now in LITERAL, OFFSET, MATCH. */ + + /* Copy LITERAL bytes from the literals. */ + + if (unlikely ((size_t)(poutend - pout) < literal)) + { + elf_uncompress_failed (); + return 0; + } + + if (unlikely (literal_count < literal)) + { + elf_uncompress_failed (); + return 0; + } + + literal_count -= literal; + + /* Often LITERAL is small, so handle small cases quickly. */ + switch (literal) + { + case 8: + *pout++ = *plit++; + /* FALLTHROUGH */ + case 7: + *pout++ = *plit++; + /* FALLTHROUGH */ + case 6: + *pout++ = *plit++; + /* FALLTHROUGH */ + case 5: + *pout++ = *plit++; + /* FALLTHROUGH */ + case 4: + *pout++ = *plit++; + /* FALLTHROUGH */ + case 3: + *pout++ = *plit++; + /* FALLTHROUGH */ + case 2: + *pout++ = *plit++; + /* FALLTHROUGH */ + case 1: + *pout++ = *plit++; + break; + + case 0: + break; + + default: + if (unlikely ((size_t)(plit - pout) < literal)) + { + uint32_t move; + + move = plit - pout; + while (literal > move) + { + memcpy (pout, plit, move); + pout += move; + plit += move; + literal -= move; + } + } + + memcpy (pout, plit, literal); + pout += literal; + plit += literal; + } + + if (match > 0) + { + /* Copy MATCH bytes from the decoded output at OFFSET. */ + + if (unlikely ((size_t)(poutend - pout) < match)) + { + elf_uncompress_failed (); + return 0; + } + + if (unlikely ((size_t)(pout - poutstart) < offset)) + { + elf_uncompress_failed (); + return 0; + } + + if (offset >= match) + { + memcpy (pout, pout - offset, match); + pout += match; + } + else + { + while (match > 0) + { + uint32_t copy; + + copy = match < offset ? match : offset; + memcpy (pout, pout - offset, copy); + match -= copy; + pout += copy; + } + } + } + + if (unlikely (seq >= seq_count)) + { + /* Copy remaining literals. */ + if (literal_count > 0 && plit != pout) + { + if (unlikely ((size_t)(poutend - pout) + < literal_count)) + { + elf_uncompress_failed (); + return 0; + } + + if ((size_t)(plit - pout) < literal_count) + { + uint32_t move; + + move = plit - pout; + while (literal_count > move) + { + memcpy (pout, plit, move); + pout += move; + plit += move; + literal_count -= move; + } + } + + memcpy (pout, plit, literal_count); + } + + pout += literal_count; + + break; + } + } + + pin = pblockend; + } + break; + + case 3: + default: + elf_uncompress_failed (); + return 0; + } + } + + if (has_checksum) + { + if (unlikely (pin + 4 > pinend)) + { + elf_uncompress_failed (); + return 0; + } + + /* We don't currently verify the checksum. Currently running GNU ld with + --compress-debug-sections=zstd does not seem to generate a + checksum. */ + + pin += 4; + } + + if (pin != pinend) + { + elf_uncompress_failed (); + return 0; + } + + return 1; +} + +#define ZDEBUG_TABLE_SIZE \ + (ZLIB_TABLE_SIZE > ZSTD_TABLE_SIZE ? ZLIB_TABLE_SIZE : ZSTD_TABLE_SIZE) + +/* Uncompress the old compressed debug format, the one emitted by + --compress-debug-sections=zlib-gnu. The compressed data is in + COMPRESSED / COMPRESSED_SIZE, and the function writes to + *UNCOMPRESSED / *UNCOMPRESSED_SIZE. ZDEBUG_TABLE is work space to + hold Huffman tables. Returns 0 on error, 1 on successful + decompression or if something goes wrong. In general we try to + carry on, by returning 1, even if we can't decompress. */ + +static int +elf_uncompress_zdebug (struct backtrace_state *state, + const unsigned char *compressed, size_t compressed_size, + uint16_t *zdebug_table, + backtrace_error_callback error_callback, void *data, + unsigned char **uncompressed, size_t *uncompressed_size) +{ + size_t sz; + size_t i; + unsigned char *po; + + *uncompressed = NULL; + *uncompressed_size = 0; + + /* The format starts with the four bytes ZLIB, followed by the 8 + byte length of the uncompressed data in big-endian order, + followed by a zlib stream. */ + + if (compressed_size < 12 || memcmp (compressed, "ZLIB", 4) != 0) + return 1; + + sz = 0; + for (i = 0; i < 8; i++) + sz = (sz << 8) | compressed[i + 4]; + + if (*uncompressed != NULL && *uncompressed_size >= sz) + po = *uncompressed; + else + { + po = (unsigned char *) backtrace_alloc (state, sz, error_callback, data); + if (po == NULL) + return 0; + } + + if (!elf_zlib_inflate_and_verify (compressed + 12, compressed_size - 12, + zdebug_table, po, sz)) + return 1; + + *uncompressed = po; + *uncompressed_size = sz; + + return 1; +} + +/* Uncompress the new compressed debug format, the official standard + ELF approach emitted by --compress-debug-sections=zlib-gabi. The + compressed data is in COMPRESSED / COMPRESSED_SIZE, and the + function writes to *UNCOMPRESSED / *UNCOMPRESSED_SIZE. + ZDEBUG_TABLE is work space as for elf_uncompress_zdebug. Returns 0 + on error, 1 on successful decompression or if something goes wrong. + In general we try to carry on, by returning 1, even if we can't + decompress. */ + +static int +elf_uncompress_chdr (struct backtrace_state *state, + const unsigned char *compressed, size_t compressed_size, + uint16_t *zdebug_table, + backtrace_error_callback error_callback, void *data, + unsigned char **uncompressed, size_t *uncompressed_size) +{ + const b_elf_chdr *chdr; + char *alc; + size_t alc_len; + unsigned char *po; + + *uncompressed = NULL; + *uncompressed_size = 0; + + /* The format starts with an ELF compression header. */ + if (compressed_size < sizeof (b_elf_chdr)) + return 1; + + chdr = (const b_elf_chdr *) compressed; + + alc = NULL; + alc_len = 0; + if (*uncompressed != NULL && *uncompressed_size >= chdr->ch_size) + po = *uncompressed; + else + { + alc_len = chdr->ch_size; + alc = (char*)backtrace_alloc (state, alc_len, error_callback, data); + if (alc == NULL) + return 0; + po = (unsigned char *) alc; + } + + switch (chdr->ch_type) + { + case ELFCOMPRESS_ZLIB: + if (!elf_zlib_inflate_and_verify (compressed + sizeof (b_elf_chdr), + compressed_size - sizeof (b_elf_chdr), + zdebug_table, po, chdr->ch_size)) + goto skip; + break; + + case ELFCOMPRESS_ZSTD: + if (!elf_zstd_decompress (compressed + sizeof (b_elf_chdr), + compressed_size - sizeof (b_elf_chdr), + (unsigned char *)zdebug_table, po, + chdr->ch_size)) + goto skip; + break; + + default: + /* Unsupported compression algorithm. */ + goto skip; + } + + *uncompressed = po; + *uncompressed_size = chdr->ch_size; + + return 1; + + skip: + if (alc != NULL && alc_len > 0) + backtrace_free (state, alc, alc_len, error_callback, data); + return 1; +} + +/* This function is a hook for testing the zlib support. It is only + used by tests. */ + +int +backtrace_uncompress_zdebug (struct backtrace_state *state, + const unsigned char *compressed, + size_t compressed_size, + backtrace_error_callback error_callback, + void *data, unsigned char **uncompressed, + size_t *uncompressed_size) +{ + uint16_t *zdebug_table; + int ret; + + zdebug_table = ((uint16_t *) backtrace_alloc (state, ZDEBUG_TABLE_SIZE, + error_callback, data)); + if (zdebug_table == NULL) + return 0; + ret = elf_uncompress_zdebug (state, compressed, compressed_size, + zdebug_table, error_callback, data, + uncompressed, uncompressed_size); + backtrace_free (state, zdebug_table, ZDEBUG_TABLE_SIZE, + error_callback, data); + return ret; +} + +/* This function is a hook for testing the zstd support. It is only used by + tests. */ + +int +backtrace_uncompress_zstd (struct backtrace_state *state, + const unsigned char *compressed, + size_t compressed_size, + backtrace_error_callback error_callback, + void *data, unsigned char *uncompressed, + size_t uncompressed_size) +{ + unsigned char *zdebug_table; + int ret; + + zdebug_table = ((unsigned char *) backtrace_alloc (state, ZDEBUG_TABLE_SIZE, + error_callback, data)); + if (zdebug_table == NULL) + return 0; + ret = elf_zstd_decompress (compressed, compressed_size, + zdebug_table, uncompressed, uncompressed_size); + backtrace_free (state, zdebug_table, ZDEBUG_TABLE_SIZE, + error_callback, data); + return ret; +} + +/* Number of LZMA states. */ +#define LZMA_STATES (12) + +/* Number of LZMA position states. The pb value of the property byte + is the number of bits to include in these states, and the maximum + value of pb is 4. */ +#define LZMA_POS_STATES (16) + +/* Number of LZMA distance states. These are used match distances + with a short match length: up to 4 bytes. */ +#define LZMA_DIST_STATES (4) + +/* Number of LZMA distance slots. LZMA uses six bits to encode larger + match lengths, so 1 << 6 possible probabilities. */ +#define LZMA_DIST_SLOTS (64) + +/* LZMA distances 0 to 3 are encoded directly, larger values use a + probability model. */ +#define LZMA_DIST_MODEL_START (4) + +/* The LZMA probability model ends at 14. */ +#define LZMA_DIST_MODEL_END (14) + +/* LZMA distance slots for distances less than 127. */ +#define LZMA_FULL_DISTANCES (128) + +/* LZMA uses four alignment bits. */ +#define LZMA_ALIGN_SIZE (16) + +/* LZMA match length is encoded with 4, 5, or 10 bits, some of which + are already known. */ +#define LZMA_LEN_LOW_SYMBOLS (8) +#define LZMA_LEN_MID_SYMBOLS (8) +#define LZMA_LEN_HIGH_SYMBOLS (256) + +/* LZMA literal encoding. */ +#define LZMA_LITERAL_CODERS_MAX (16) +#define LZMA_LITERAL_CODER_SIZE (0x300) + +/* LZMA is based on a large set of probabilities, each managed + independently. Each probability is an 11 bit number that we store + in a uint16_t. We use a single large array of probabilities. */ + +/* Lengths of entries in the LZMA probabilities array. The names used + here are copied from the Linux kernel implementation. */ + +#define LZMA_PROB_IS_MATCH_LEN (LZMA_STATES * LZMA_POS_STATES) +#define LZMA_PROB_IS_REP_LEN LZMA_STATES +#define LZMA_PROB_IS_REP0_LEN LZMA_STATES +#define LZMA_PROB_IS_REP1_LEN LZMA_STATES +#define LZMA_PROB_IS_REP2_LEN LZMA_STATES +#define LZMA_PROB_IS_REP0_LONG_LEN (LZMA_STATES * LZMA_POS_STATES) +#define LZMA_PROB_DIST_SLOT_LEN (LZMA_DIST_STATES * LZMA_DIST_SLOTS) +#define LZMA_PROB_DIST_SPECIAL_LEN (LZMA_FULL_DISTANCES - LZMA_DIST_MODEL_END) +#define LZMA_PROB_DIST_ALIGN_LEN LZMA_ALIGN_SIZE +#define LZMA_PROB_MATCH_LEN_CHOICE_LEN 1 +#define LZMA_PROB_MATCH_LEN_CHOICE2_LEN 1 +#define LZMA_PROB_MATCH_LEN_LOW_LEN (LZMA_POS_STATES * LZMA_LEN_LOW_SYMBOLS) +#define LZMA_PROB_MATCH_LEN_MID_LEN (LZMA_POS_STATES * LZMA_LEN_MID_SYMBOLS) +#define LZMA_PROB_MATCH_LEN_HIGH_LEN LZMA_LEN_HIGH_SYMBOLS +#define LZMA_PROB_REP_LEN_CHOICE_LEN 1 +#define LZMA_PROB_REP_LEN_CHOICE2_LEN 1 +#define LZMA_PROB_REP_LEN_LOW_LEN (LZMA_POS_STATES * LZMA_LEN_LOW_SYMBOLS) +#define LZMA_PROB_REP_LEN_MID_LEN (LZMA_POS_STATES * LZMA_LEN_MID_SYMBOLS) +#define LZMA_PROB_REP_LEN_HIGH_LEN LZMA_LEN_HIGH_SYMBOLS +#define LZMA_PROB_LITERAL_LEN \ + (LZMA_LITERAL_CODERS_MAX * LZMA_LITERAL_CODER_SIZE) + +/* Offsets into the LZMA probabilities array. This is mechanically + generated from the above lengths. */ + +#define LZMA_PROB_IS_MATCH_OFFSET 0 +#define LZMA_PROB_IS_REP_OFFSET \ + (LZMA_PROB_IS_MATCH_OFFSET + LZMA_PROB_IS_MATCH_LEN) +#define LZMA_PROB_IS_REP0_OFFSET \ + (LZMA_PROB_IS_REP_OFFSET + LZMA_PROB_IS_REP_LEN) +#define LZMA_PROB_IS_REP1_OFFSET \ + (LZMA_PROB_IS_REP0_OFFSET + LZMA_PROB_IS_REP0_LEN) +#define LZMA_PROB_IS_REP2_OFFSET \ + (LZMA_PROB_IS_REP1_OFFSET + LZMA_PROB_IS_REP1_LEN) +#define LZMA_PROB_IS_REP0_LONG_OFFSET \ + (LZMA_PROB_IS_REP2_OFFSET + LZMA_PROB_IS_REP2_LEN) +#define LZMA_PROB_DIST_SLOT_OFFSET \ + (LZMA_PROB_IS_REP0_LONG_OFFSET + LZMA_PROB_IS_REP0_LONG_LEN) +#define LZMA_PROB_DIST_SPECIAL_OFFSET \ + (LZMA_PROB_DIST_SLOT_OFFSET + LZMA_PROB_DIST_SLOT_LEN) +#define LZMA_PROB_DIST_ALIGN_OFFSET \ + (LZMA_PROB_DIST_SPECIAL_OFFSET + LZMA_PROB_DIST_SPECIAL_LEN) +#define LZMA_PROB_MATCH_LEN_CHOICE_OFFSET \ + (LZMA_PROB_DIST_ALIGN_OFFSET + LZMA_PROB_DIST_ALIGN_LEN) +#define LZMA_PROB_MATCH_LEN_CHOICE2_OFFSET \ + (LZMA_PROB_MATCH_LEN_CHOICE_OFFSET + LZMA_PROB_MATCH_LEN_CHOICE_LEN) +#define LZMA_PROB_MATCH_LEN_LOW_OFFSET \ + (LZMA_PROB_MATCH_LEN_CHOICE2_OFFSET + LZMA_PROB_MATCH_LEN_CHOICE2_LEN) +#define LZMA_PROB_MATCH_LEN_MID_OFFSET \ + (LZMA_PROB_MATCH_LEN_LOW_OFFSET + LZMA_PROB_MATCH_LEN_LOW_LEN) +#define LZMA_PROB_MATCH_LEN_HIGH_OFFSET \ + (LZMA_PROB_MATCH_LEN_MID_OFFSET + LZMA_PROB_MATCH_LEN_MID_LEN) +#define LZMA_PROB_REP_LEN_CHOICE_OFFSET \ + (LZMA_PROB_MATCH_LEN_HIGH_OFFSET + LZMA_PROB_MATCH_LEN_HIGH_LEN) +#define LZMA_PROB_REP_LEN_CHOICE2_OFFSET \ + (LZMA_PROB_REP_LEN_CHOICE_OFFSET + LZMA_PROB_REP_LEN_CHOICE_LEN) +#define LZMA_PROB_REP_LEN_LOW_OFFSET \ + (LZMA_PROB_REP_LEN_CHOICE2_OFFSET + LZMA_PROB_REP_LEN_CHOICE2_LEN) +#define LZMA_PROB_REP_LEN_MID_OFFSET \ + (LZMA_PROB_REP_LEN_LOW_OFFSET + LZMA_PROB_REP_LEN_LOW_LEN) +#define LZMA_PROB_REP_LEN_HIGH_OFFSET \ + (LZMA_PROB_REP_LEN_MID_OFFSET + LZMA_PROB_REP_LEN_MID_LEN) +#define LZMA_PROB_LITERAL_OFFSET \ + (LZMA_PROB_REP_LEN_HIGH_OFFSET + LZMA_PROB_REP_LEN_HIGH_LEN) + +#define LZMA_PROB_TOTAL_COUNT \ + (LZMA_PROB_LITERAL_OFFSET + LZMA_PROB_LITERAL_LEN) + +/* Check that the number of LZMA probabilities is the same as the + Linux kernel implementation. */ + +#if LZMA_PROB_TOTAL_COUNT != 1846 + (1 << 4) * 0x300 + #error Wrong number of LZMA probabilities +#endif + +/* Expressions for the offset in the LZMA probabilities array of a + specific probability. */ + +#define LZMA_IS_MATCH(state, pos) \ + (LZMA_PROB_IS_MATCH_OFFSET + (state) * LZMA_POS_STATES + (pos)) +#define LZMA_IS_REP(state) \ + (LZMA_PROB_IS_REP_OFFSET + (state)) +#define LZMA_IS_REP0(state) \ + (LZMA_PROB_IS_REP0_OFFSET + (state)) +#define LZMA_IS_REP1(state) \ + (LZMA_PROB_IS_REP1_OFFSET + (state)) +#define LZMA_IS_REP2(state) \ + (LZMA_PROB_IS_REP2_OFFSET + (state)) +#define LZMA_IS_REP0_LONG(state, pos) \ + (LZMA_PROB_IS_REP0_LONG_OFFSET + (state) * LZMA_POS_STATES + (pos)) +#define LZMA_DIST_SLOT(dist, slot) \ + (LZMA_PROB_DIST_SLOT_OFFSET + (dist) * LZMA_DIST_SLOTS + (slot)) +#define LZMA_DIST_SPECIAL(dist) \ + (LZMA_PROB_DIST_SPECIAL_OFFSET + (dist)) +#define LZMA_DIST_ALIGN(dist) \ + (LZMA_PROB_DIST_ALIGN_OFFSET + (dist)) +#define LZMA_MATCH_LEN_CHOICE \ + LZMA_PROB_MATCH_LEN_CHOICE_OFFSET +#define LZMA_MATCH_LEN_CHOICE2 \ + LZMA_PROB_MATCH_LEN_CHOICE2_OFFSET +#define LZMA_MATCH_LEN_LOW(pos, sym) \ + (LZMA_PROB_MATCH_LEN_LOW_OFFSET + (pos) * LZMA_LEN_LOW_SYMBOLS + (sym)) +#define LZMA_MATCH_LEN_MID(pos, sym) \ + (LZMA_PROB_MATCH_LEN_MID_OFFSET + (pos) * LZMA_LEN_MID_SYMBOLS + (sym)) +#define LZMA_MATCH_LEN_HIGH(sym) \ + (LZMA_PROB_MATCH_LEN_HIGH_OFFSET + (sym)) +#define LZMA_REP_LEN_CHOICE \ + LZMA_PROB_REP_LEN_CHOICE_OFFSET +#define LZMA_REP_LEN_CHOICE2 \ + LZMA_PROB_REP_LEN_CHOICE2_OFFSET +#define LZMA_REP_LEN_LOW(pos, sym) \ + (LZMA_PROB_REP_LEN_LOW_OFFSET + (pos) * LZMA_LEN_LOW_SYMBOLS + (sym)) +#define LZMA_REP_LEN_MID(pos, sym) \ + (LZMA_PROB_REP_LEN_MID_OFFSET + (pos) * LZMA_LEN_MID_SYMBOLS + (sym)) +#define LZMA_REP_LEN_HIGH(sym) \ + (LZMA_PROB_REP_LEN_HIGH_OFFSET + (sym)) +#define LZMA_LITERAL(code, size) \ + (LZMA_PROB_LITERAL_OFFSET + (code) * LZMA_LITERAL_CODER_SIZE + (size)) + +/* Read an LZMA varint from BUF, reading and updating *POFFSET, + setting *VAL. Returns 0 on error, 1 on success. */ + +static int +elf_lzma_varint (const unsigned char *compressed, size_t compressed_size, + size_t *poffset, uint64_t *val) +{ + size_t off; + int i; + uint64_t v; + unsigned char b; + + off = *poffset; + i = 0; + v = 0; + while (1) + { + if (unlikely (off >= compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + b = compressed[off]; + v |= (b & 0x7f) << (i * 7); + ++off; + if ((b & 0x80) == 0) + { + *poffset = off; + *val = v; + return 1; + } + ++i; + if (unlikely (i >= 9)) + { + elf_uncompress_failed (); + return 0; + } + } +} + +/* Normalize the LZMA range decoder, pulling in an extra input byte if + needed. */ + +static void +elf_lzma_range_normalize (const unsigned char *compressed, + size_t compressed_size, size_t *poffset, + uint32_t *prange, uint32_t *pcode) +{ + if (*prange < (1U << 24)) + { + if (unlikely (*poffset >= compressed_size)) + { + /* We assume this will be caught elsewhere. */ + elf_uncompress_failed (); + return; + } + *prange <<= 8; + *pcode <<= 8; + *pcode += compressed[*poffset]; + ++*poffset; + } +} + +/* Read and return a single bit from the LZMA stream, reading and + updating *PROB. Each bit comes from the range coder. */ + +static int +elf_lzma_bit (const unsigned char *compressed, size_t compressed_size, + uint16_t *prob, size_t *poffset, uint32_t *prange, + uint32_t *pcode) +{ + uint32_t bound; + + elf_lzma_range_normalize (compressed, compressed_size, poffset, + prange, pcode); + bound = (*prange >> 11) * (uint32_t) *prob; + if (*pcode < bound) + { + *prange = bound; + *prob += ((1U << 11) - *prob) >> 5; + return 0; + } + else + { + *prange -= bound; + *pcode -= bound; + *prob -= *prob >> 5; + return 1; + } +} + +/* Read an integer of size BITS from the LZMA stream, most significant + bit first. The bits are predicted using PROBS. */ + +static uint32_t +elf_lzma_integer (const unsigned char *compressed, size_t compressed_size, + uint16_t *probs, uint32_t bits, size_t *poffset, + uint32_t *prange, uint32_t *pcode) +{ + uint32_t sym; + uint32_t i; + + sym = 1; + for (i = 0; i < bits; i++) + { + int bit; + + bit = elf_lzma_bit (compressed, compressed_size, probs + sym, poffset, + prange, pcode); + sym <<= 1; + sym += bit; + } + return sym - (1 << bits); +} + +/* Read an integer of size BITS from the LZMA stream, least + significant bit first. The bits are predicted using PROBS. */ + +static uint32_t +elf_lzma_reverse_integer (const unsigned char *compressed, + size_t compressed_size, uint16_t *probs, + uint32_t bits, size_t *poffset, uint32_t *prange, + uint32_t *pcode) +{ + uint32_t sym; + uint32_t val; + uint32_t i; + + sym = 1; + val = 0; + for (i = 0; i < bits; i++) + { + int bit; + + bit = elf_lzma_bit (compressed, compressed_size, probs + sym, poffset, + prange, pcode); + sym <<= 1; + sym += bit; + val += bit << i; + } + return val; +} + +/* Read a length from the LZMA stream. IS_REP picks either LZMA_MATCH + or LZMA_REP probabilities. */ + +static uint32_t +elf_lzma_len (const unsigned char *compressed, size_t compressed_size, + uint16_t *probs, int is_rep, unsigned int pos_state, + size_t *poffset, uint32_t *prange, uint32_t *pcode) +{ + uint16_t *probs_choice; + uint16_t *probs_sym; + uint32_t bits; + uint32_t len; + + probs_choice = probs + (is_rep + ? LZMA_REP_LEN_CHOICE + : LZMA_MATCH_LEN_CHOICE); + if (elf_lzma_bit (compressed, compressed_size, probs_choice, poffset, + prange, pcode)) + { + probs_choice = probs + (is_rep + ? LZMA_REP_LEN_CHOICE2 + : LZMA_MATCH_LEN_CHOICE2); + if (elf_lzma_bit (compressed, compressed_size, probs_choice, + poffset, prange, pcode)) + { + probs_sym = probs + (is_rep + ? LZMA_REP_LEN_HIGH (0) + : LZMA_MATCH_LEN_HIGH (0)); + bits = 8; + len = 2 + 8 + 8; + } + else + { + probs_sym = probs + (is_rep + ? LZMA_REP_LEN_MID (pos_state, 0) + : LZMA_MATCH_LEN_MID (pos_state, 0)); + bits = 3; + len = 2 + 8; + } + } + else + { + probs_sym = probs + (is_rep + ? LZMA_REP_LEN_LOW (pos_state, 0) + : LZMA_MATCH_LEN_LOW (pos_state, 0)); + bits = 3; + len = 2; + } + + len += elf_lzma_integer (compressed, compressed_size, probs_sym, bits, + poffset, prange, pcode); + return len; +} + +/* Uncompress one LZMA block from a minidebug file. The compressed + data is at COMPRESSED + *POFFSET. Update *POFFSET. Store the data + into the memory at UNCOMPRESSED, size UNCOMPRESSED_SIZE. CHECK is + the stream flag from the xz header. Return 1 on successful + decompression. */ + +static int +elf_uncompress_lzma_block (const unsigned char *compressed, + size_t compressed_size, unsigned char check, + uint16_t *probs, unsigned char *uncompressed, + size_t uncompressed_size, size_t *poffset) +{ + size_t off; + size_t block_header_offset; + size_t block_header_size; + unsigned char block_flags; + uint64_t header_compressed_size; + uint64_t header_uncompressed_size; + unsigned char lzma2_properties; + uint32_t computed_crc; + uint32_t stream_crc; + size_t uncompressed_offset; + size_t dict_start_offset; + unsigned int lc; + unsigned int lp; + unsigned int pb; + uint32_t range; + uint32_t code; + uint32_t lstate; + uint32_t dist[4]; + + off = *poffset; + block_header_offset = off; + + /* Block header size is a single byte. */ + if (unlikely (off >= compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + block_header_size = (compressed[off] + 1) * 4; + if (unlikely (off + block_header_size > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + /* Block flags. */ + block_flags = compressed[off + 1]; + if (unlikely ((block_flags & 0x3c) != 0)) + { + elf_uncompress_failed (); + return 0; + } + + off += 2; + + /* Optional compressed size. */ + header_compressed_size = 0; + if ((block_flags & 0x40) != 0) + { + *poffset = off; + if (!elf_lzma_varint (compressed, compressed_size, poffset, + &header_compressed_size)) + return 0; + off = *poffset; + } + + /* Optional uncompressed size. */ + header_uncompressed_size = 0; + if ((block_flags & 0x80) != 0) + { + *poffset = off; + if (!elf_lzma_varint (compressed, compressed_size, poffset, + &header_uncompressed_size)) + return 0; + off = *poffset; + } + + /* The recipe for creating a minidebug file is to run the xz program + with no arguments, so we expect exactly one filter: lzma2. */ + + if (unlikely ((block_flags & 0x3) != 0)) + { + elf_uncompress_failed (); + return 0; + } + + if (unlikely (off + 2 >= block_header_offset + block_header_size)) + { + elf_uncompress_failed (); + return 0; + } + + /* The filter ID for LZMA2 is 0x21. */ + if (unlikely (compressed[off] != 0x21)) + { + elf_uncompress_failed (); + return 0; + } + ++off; + + /* The size of the filter properties for LZMA2 is 1. */ + if (unlikely (compressed[off] != 1)) + { + elf_uncompress_failed (); + return 0; + } + ++off; + + lzma2_properties = compressed[off]; + ++off; + + if (unlikely (lzma2_properties > 40)) + { + elf_uncompress_failed (); + return 0; + } + + /* The properties describe the dictionary size, but we don't care + what that is. */ + + /* Block header padding. */ + if (unlikely (off + 4 > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + off = (off + 3) &~ (size_t) 3; + + if (unlikely (off + 4 > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + /* Block header CRC. */ + computed_crc = elf_crc32 (0, compressed + block_header_offset, + block_header_size - 4); + stream_crc = (compressed[off] + | (compressed[off + 1] << 8) + | (compressed[off + 2] << 16) + | (compressed[off + 3] << 24)); + if (unlikely (computed_crc != stream_crc)) + { + elf_uncompress_failed (); + return 0; + } + off += 4; + + /* Read a sequence of LZMA2 packets. */ + + uncompressed_offset = 0; + dict_start_offset = 0; + lc = 0; + lp = 0; + pb = 0; + lstate = 0; + while (off < compressed_size) + { + unsigned char control; + + range = 0xffffffff; + code = 0; + + control = compressed[off]; + ++off; + if (unlikely (control == 0)) + { + /* End of packets. */ + break; + } + + if (control == 1 || control >= 0xe0) + { + /* Reset dictionary to empty. */ + dict_start_offset = uncompressed_offset; + } + + if (control < 0x80) + { + size_t chunk_size; + + /* The only valid values here are 1 or 2. A 1 means to + reset the dictionary (done above). Then we see an + uncompressed chunk. */ + + if (unlikely (control > 2)) + { + elf_uncompress_failed (); + return 0; + } + + /* An uncompressed chunk is a two byte size followed by + data. */ + + if (unlikely (off + 2 > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + chunk_size = compressed[off] << 8; + chunk_size += compressed[off + 1]; + ++chunk_size; + + off += 2; + + if (unlikely (off + chunk_size > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + if (unlikely (uncompressed_offset + chunk_size > uncompressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + memcpy (uncompressed + uncompressed_offset, compressed + off, + chunk_size); + uncompressed_offset += chunk_size; + off += chunk_size; + } + else + { + size_t uncompressed_chunk_start; + size_t uncompressed_chunk_size; + size_t compressed_chunk_size; + size_t limit; + + /* An LZMA chunk. This starts with an uncompressed size and + a compressed size. */ + + if (unlikely (off + 4 >= compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + uncompressed_chunk_start = uncompressed_offset; + + uncompressed_chunk_size = (control & 0x1f) << 16; + uncompressed_chunk_size += compressed[off] << 8; + uncompressed_chunk_size += compressed[off + 1]; + ++uncompressed_chunk_size; + + compressed_chunk_size = compressed[off + 2] << 8; + compressed_chunk_size += compressed[off + 3]; + ++compressed_chunk_size; + + off += 4; + + /* Bit 7 (0x80) is set. + Bits 6 and 5 (0x40 and 0x20) are as follows: + 0: don't reset anything + 1: reset state + 2: reset state, read properties + 3: reset state, read properties, reset dictionary (done above) */ + + if (control >= 0xc0) + { + unsigned char props; + + /* Bit 6 is set, read properties. */ + + if (unlikely (off >= compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + props = compressed[off]; + ++off; + if (unlikely (props > (4 * 5 + 4) * 9 + 8)) + { + elf_uncompress_failed (); + return 0; + } + pb = 0; + while (props >= 9 * 5) + { + props -= 9 * 5; + ++pb; + } + lp = 0; + while (props > 9) + { + props -= 9; + ++lp; + } + lc = props; + if (unlikely (lc + lp > 4)) + { + elf_uncompress_failed (); + return 0; + } + } + + if (control >= 0xa0) + { + size_t i; + + /* Bit 5 or 6 is set, reset LZMA state. */ + + lstate = 0; + memset (&dist, 0, sizeof dist); + for (i = 0; i < LZMA_PROB_TOTAL_COUNT; i++) + probs[i] = 1 << 10; + range = 0xffffffff; + code = 0; + } + + /* Read the range code. */ + + if (unlikely (off + 5 > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + /* The byte at compressed[off] is ignored for some + reason. */ + + code = ((compressed[off + 1] << 24) + + (compressed[off + 2] << 16) + + (compressed[off + 3] << 8) + + compressed[off + 4]); + off += 5; + + /* This is the main LZMA decode loop. */ + + limit = off + compressed_chunk_size; + *poffset = off; + while (*poffset < limit) + { + unsigned int pos_state; + + if (unlikely (uncompressed_offset + == (uncompressed_chunk_start + + uncompressed_chunk_size))) + { + /* We've decompressed all the expected bytes. */ + break; + } + + pos_state = ((uncompressed_offset - dict_start_offset) + & ((1 << pb) - 1)); + + if (elf_lzma_bit (compressed, compressed_size, + probs + LZMA_IS_MATCH (lstate, pos_state), + poffset, &range, &code)) + { + uint32_t len; + + if (elf_lzma_bit (compressed, compressed_size, + probs + LZMA_IS_REP (lstate), + poffset, &range, &code)) + { + int short_rep; + uint32_t next_dist; + + /* Repeated match. */ + + short_rep = 0; + if (elf_lzma_bit (compressed, compressed_size, + probs + LZMA_IS_REP0 (lstate), + poffset, &range, &code)) + { + if (elf_lzma_bit (compressed, compressed_size, + probs + LZMA_IS_REP1 (lstate), + poffset, &range, &code)) + { + if (elf_lzma_bit (compressed, compressed_size, + probs + LZMA_IS_REP2 (lstate), + poffset, &range, &code)) + { + next_dist = dist[3]; + dist[3] = dist[2]; + } + else + { + next_dist = dist[2]; + } + dist[2] = dist[1]; + } + else + { + next_dist = dist[1]; + } + + dist[1] = dist[0]; + dist[0] = next_dist; + } + else + { + if (!elf_lzma_bit (compressed, compressed_size, + (probs + + LZMA_IS_REP0_LONG (lstate, + pos_state)), + poffset, &range, &code)) + short_rep = 1; + } + + if (lstate < 7) + lstate = short_rep ? 9 : 8; + else + lstate = 11; + + if (short_rep) + len = 1; + else + len = elf_lzma_len (compressed, compressed_size, + probs, 1, pos_state, poffset, + &range, &code); + } + else + { + uint32_t dist_state; + uint32_t dist_slot; + uint16_t *probs_dist; + + /* Match. */ + + if (lstate < 7) + lstate = 7; + else + lstate = 10; + dist[3] = dist[2]; + dist[2] = dist[1]; + dist[1] = dist[0]; + len = elf_lzma_len (compressed, compressed_size, + probs, 0, pos_state, poffset, + &range, &code); + + if (len < 4 + 2) + dist_state = len - 2; + else + dist_state = 3; + probs_dist = probs + LZMA_DIST_SLOT (dist_state, 0); + dist_slot = elf_lzma_integer (compressed, + compressed_size, + probs_dist, 6, + poffset, &range, + &code); + if (dist_slot < LZMA_DIST_MODEL_START) + dist[0] = dist_slot; + else + { + uint32_t limit; + + limit = (dist_slot >> 1) - 1; + dist[0] = 2 + (dist_slot & 1); + if (dist_slot < LZMA_DIST_MODEL_END) + { + dist[0] <<= limit; + probs_dist = (probs + + LZMA_DIST_SPECIAL(dist[0] + - dist_slot + - 1)); + dist[0] += + elf_lzma_reverse_integer (compressed, + compressed_size, + probs_dist, + limit, poffset, + &range, &code); + } + else + { + uint32_t dist0; + uint32_t i; + + dist0 = dist[0]; + for (i = 0; i < limit - 4; i++) + { + uint32_t mask; + + elf_lzma_range_normalize (compressed, + compressed_size, + poffset, + &range, &code); + range >>= 1; + code -= range; + mask = -(code >> 31); + code += range & mask; + dist0 <<= 1; + dist0 += mask + 1; + } + dist0 <<= 4; + probs_dist = probs + LZMA_DIST_ALIGN (0); + dist0 += + elf_lzma_reverse_integer (compressed, + compressed_size, + probs_dist, 4, + poffset, + &range, &code); + dist[0] = dist0; + } + } + } + + if (unlikely (uncompressed_offset + - dict_start_offset < dist[0] + 1)) + { + elf_uncompress_failed (); + return 0; + } + if (unlikely (uncompressed_offset + len > uncompressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + if (dist[0] == 0) + { + /* A common case, meaning repeat the last + character LEN times. */ + memset (uncompressed + uncompressed_offset, + uncompressed[uncompressed_offset - 1], + len); + uncompressed_offset += len; + } + else if (dist[0] + 1 >= len) + { + memcpy (uncompressed + uncompressed_offset, + uncompressed + uncompressed_offset - dist[0] - 1, + len); + uncompressed_offset += len; + } + else + { + while (len > 0) + { + uint32_t copy; + + copy = len < dist[0] + 1 ? len : dist[0] + 1; + memcpy (uncompressed + uncompressed_offset, + (uncompressed + uncompressed_offset + - dist[0] - 1), + copy); + len -= copy; + uncompressed_offset += copy; + } + } + } + else + { + unsigned char prev; + unsigned char low; + size_t high; + uint16_t *lit_probs; + unsigned int sym; + + /* Literal value. */ + + if (uncompressed_offset > 0) + prev = uncompressed[uncompressed_offset - 1]; + else + prev = 0; + low = prev >> (8 - lc); + high = (((uncompressed_offset - dict_start_offset) + & ((1 << lp) - 1)) + << lc); + lit_probs = probs + LZMA_LITERAL (low + high, 0); + if (lstate < 7) + sym = elf_lzma_integer (compressed, compressed_size, + lit_probs, 8, poffset, &range, + &code); + else + { + unsigned int match; + unsigned int bit; + unsigned int match_bit; + unsigned int idx; + + sym = 1; + if (uncompressed_offset >= dist[0] + 1) + match = uncompressed[uncompressed_offset - dist[0] - 1]; + else + match = 0; + match <<= 1; + bit = 0x100; + do + { + match_bit = match & bit; + match <<= 1; + idx = bit + match_bit + sym; + sym <<= 1; + if (elf_lzma_bit (compressed, compressed_size, + lit_probs + idx, poffset, + &range, &code)) + { + ++sym; + bit &= match_bit; + } + else + { + bit &= ~ match_bit; + } + } + while (sym < 0x100); + } + + if (unlikely (uncompressed_offset >= uncompressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + uncompressed[uncompressed_offset] = (unsigned char) sym; + ++uncompressed_offset; + if (lstate <= 3) + lstate = 0; + else if (lstate <= 9) + lstate -= 3; + else + lstate -= 6; + } + } + + elf_lzma_range_normalize (compressed, compressed_size, poffset, + &range, &code); + + off = *poffset; + } + } + + /* We have reached the end of the block. Pad to four byte + boundary. */ + off = (off + 3) &~ (size_t) 3; + if (unlikely (off > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + + switch (check) + { + case 0: + /* No check. */ + break; + + case 1: + /* CRC32 */ + if (unlikely (off + 4 > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + computed_crc = elf_crc32 (0, uncompressed, uncompressed_offset); + stream_crc = (compressed[off] + | (compressed[off + 1] << 8) + | (compressed[off + 2] << 16) + | (compressed[off + 3] << 24)); + if (computed_crc != stream_crc) + { + elf_uncompress_failed (); + return 0; + } + off += 4; + break; + + case 4: + /* CRC64. We don't bother computing a CRC64 checksum. */ + if (unlikely (off + 8 > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + off += 8; + break; + + case 10: + /* SHA. We don't bother computing a SHA checksum. */ + if (unlikely (off + 32 > compressed_size)) + { + elf_uncompress_failed (); + return 0; + } + off += 32; + break; + + default: + elf_uncompress_failed (); + return 0; + } + + *poffset = off; + + return 1; +} + +/* Uncompress LZMA data found in a minidebug file. The minidebug + format is described at + https://sourceware.org/gdb/current/onlinedocs/gdb/MiniDebugInfo.html. + Returns 0 on error, 1 on successful decompression. For this + function we return 0 on failure to decompress, as the calling code + will carry on in that case. */ + +static int +elf_uncompress_lzma (struct backtrace_state *state, + const unsigned char *compressed, size_t compressed_size, + backtrace_error_callback error_callback, void *data, + unsigned char **uncompressed, size_t *uncompressed_size) +{ + size_t header_size; + size_t footer_size; + unsigned char check; + uint32_t computed_crc; + uint32_t stream_crc; + size_t offset; + size_t index_size; + size_t footer_offset; + size_t index_offset; + uint64_t index_compressed_size; + uint64_t index_uncompressed_size; + unsigned char *mem; + uint16_t *probs; + size_t compressed_block_size; + + /* The format starts with a stream header and ends with a stream + footer. */ + header_size = 12; + footer_size = 12; + if (unlikely (compressed_size < header_size + footer_size)) + { + elf_uncompress_failed (); + return 0; + } + + /* The stream header starts with a magic string. */ + if (unlikely (memcmp (compressed, "\375" "7zXZ\0", 6) != 0)) + { + elf_uncompress_failed (); + return 0; + } + + /* Next come stream flags. The first byte is zero, the second byte + is the check. */ + if (unlikely (compressed[6] != 0)) + { + elf_uncompress_failed (); + return 0; + } + check = compressed[7]; + if (unlikely ((check & 0xf8) != 0)) + { + elf_uncompress_failed (); + return 0; + } + + /* Next comes a CRC of the stream flags. */ + computed_crc = elf_crc32 (0, compressed + 6, 2); + stream_crc = (compressed[8] + | (compressed[9] << 8) + | (compressed[10] << 16) + | (compressed[11] << 24)); + if (unlikely (computed_crc != stream_crc)) + { + elf_uncompress_failed (); + return 0; + } + + /* Now that we've parsed the header, parse the footer, so that we + can get the uncompressed size. */ + + /* The footer ends with two magic bytes. */ + + offset = compressed_size; + if (unlikely (memcmp (compressed + offset - 2, "YZ", 2) != 0)) + { + elf_uncompress_failed (); + return 0; + } + offset -= 2; + + /* Before that are the stream flags, which should be the same as the + flags in the header. */ + if (unlikely (compressed[offset - 2] != 0 + || compressed[offset - 1] != check)) + { + elf_uncompress_failed (); + return 0; + } + offset -= 2; + + /* Before that is the size of the index field, which precedes the + footer. */ + index_size = (compressed[offset - 4] + | (compressed[offset - 3] << 8) + | (compressed[offset - 2] << 16) + | (compressed[offset - 1] << 24)); + index_size = (index_size + 1) * 4; + offset -= 4; + + /* Before that is a footer CRC. */ + computed_crc = elf_crc32 (0, compressed + offset, 6); + stream_crc = (compressed[offset - 4] + | (compressed[offset - 3] << 8) + | (compressed[offset - 2] << 16) + | (compressed[offset - 1] << 24)); + if (unlikely (computed_crc != stream_crc)) + { + elf_uncompress_failed (); + return 0; + } + offset -= 4; + + /* The index comes just before the footer. */ + if (unlikely (offset < index_size + header_size)) + { + elf_uncompress_failed (); + return 0; + } + + footer_offset = offset; + offset -= index_size; + index_offset = offset; + + /* The index starts with a zero byte. */ + if (unlikely (compressed[offset] != 0)) + { + elf_uncompress_failed (); + return 0; + } + ++offset; + + /* Next is the number of blocks. We expect zero blocks for an empty + stream, and otherwise a single block. */ + if (unlikely (compressed[offset] == 0)) + { + *uncompressed = NULL; + *uncompressed_size = 0; + return 1; + } + if (unlikely (compressed[offset] != 1)) + { + elf_uncompress_failed (); + return 0; + } + ++offset; + + /* Next is the compressed size and the uncompressed size. */ + if (!elf_lzma_varint (compressed, compressed_size, &offset, + &index_compressed_size)) + return 0; + if (!elf_lzma_varint (compressed, compressed_size, &offset, + &index_uncompressed_size)) + return 0; + + /* Pad to a four byte boundary. */ + offset = (offset + 3) &~ (size_t) 3; + + /* Next is a CRC of the index. */ + computed_crc = elf_crc32 (0, compressed + index_offset, + offset - index_offset); + stream_crc = (compressed[offset] + | (compressed[offset + 1] << 8) + | (compressed[offset + 2] << 16) + | (compressed[offset + 3] << 24)); + if (unlikely (computed_crc != stream_crc)) + { + elf_uncompress_failed (); + return 0; + } + offset += 4; + + /* We should now be back at the footer. */ + if (unlikely (offset != footer_offset)) + { + elf_uncompress_failed (); + return 0; + } + + /* Allocate space to hold the uncompressed data. If we succeed in + uncompressing the LZMA data, we never free this memory. */ + mem = (unsigned char *) backtrace_alloc (state, index_uncompressed_size, + error_callback, data); + if (unlikely (mem == NULL)) + return 0; + *uncompressed = mem; + *uncompressed_size = index_uncompressed_size; + + /* Allocate space for probabilities. */ + probs = ((uint16_t *) + backtrace_alloc (state, + LZMA_PROB_TOTAL_COUNT * sizeof (uint16_t), + error_callback, data)); + if (unlikely (probs == NULL)) + { + backtrace_free (state, mem, index_uncompressed_size, error_callback, + data); + return 0; + } + + /* Uncompress the block, which follows the header. */ + offset = 12; + if (!elf_uncompress_lzma_block (compressed, compressed_size, check, probs, + mem, index_uncompressed_size, &offset)) + { + backtrace_free (state, mem, index_uncompressed_size, error_callback, + data); + return 0; + } + + compressed_block_size = offset - 12; + if (unlikely (compressed_block_size + != ((index_compressed_size + 3) &~ (size_t) 3))) + { + elf_uncompress_failed (); + backtrace_free (state, mem, index_uncompressed_size, error_callback, + data); + return 0; + } + + offset = (offset + 3) &~ (size_t) 3; + if (unlikely (offset != index_offset)) + { + elf_uncompress_failed (); + backtrace_free (state, mem, index_uncompressed_size, error_callback, + data); + return 0; + } + + return 1; +} + +/* This function is a hook for testing the LZMA support. It is only + used by tests. */ + +int +backtrace_uncompress_lzma (struct backtrace_state *state, + const unsigned char *compressed, + size_t compressed_size, + backtrace_error_callback error_callback, + void *data, unsigned char **uncompressed, + size_t *uncompressed_size) +{ + return elf_uncompress_lzma (state, compressed, compressed_size, + error_callback, data, uncompressed, + uncompressed_size); +} + +/* Add the backtrace data for one ELF file. Returns 1 on success, + 0 on failure (in both cases descriptor is closed) or -1 if exe + is non-zero and the ELF file is ET_DYN, which tells the caller that + elf_add will need to be called on the descriptor again after + base_address is determined. */ + +static int +elf_add (struct backtrace_state *state, const char *filename, int descriptor, + const unsigned char *memory, size_t memory_size, + uintptr_t base_address, backtrace_error_callback error_callback, + void *data, fileline *fileline_fn, int *found_sym, int *found_dwarf, + struct dwarf_data **fileline_entry, int exe, int debuginfo, + const char *with_buildid_data, uint32_t with_buildid_size) +{ + struct elf_view ehdr_view; + b_elf_ehdr ehdr; + off_t shoff; + unsigned int shnum; + unsigned int shstrndx; + struct elf_view shdrs_view; + int shdrs_view_valid; + const b_elf_shdr *shdrs; + const b_elf_shdr *shstrhdr; + size_t shstr_size; + off_t shstr_off; + struct elf_view names_view; + int names_view_valid; + const char *names; + unsigned int symtab_shndx; + unsigned int dynsym_shndx; + unsigned int i; + struct debug_section_info sections[DEBUG_MAX]; + struct debug_section_info zsections[DEBUG_MAX]; + struct elf_view symtab_view; + int symtab_view_valid; + struct elf_view strtab_view; + int strtab_view_valid; + struct elf_view buildid_view; + int buildid_view_valid; + const char *buildid_data; + uint32_t buildid_size; + struct elf_view debuglink_view; + int debuglink_view_valid; + const char *debuglink_name; + uint32_t debuglink_crc; + struct elf_view debugaltlink_view; + int debugaltlink_view_valid; + const char *debugaltlink_name; + const char *debugaltlink_buildid_data; + uint32_t debugaltlink_buildid_size; + struct elf_view gnu_debugdata_view; + int gnu_debugdata_view_valid; + size_t gnu_debugdata_size; + unsigned char *gnu_debugdata_uncompressed; + size_t gnu_debugdata_uncompressed_size; + off_t min_offset; + off_t max_offset; + off_t debug_size; + struct elf_view debug_view; + int debug_view_valid; + unsigned int using_debug_view; + uint16_t *zdebug_table; + struct elf_view split_debug_view[DEBUG_MAX]; + unsigned char split_debug_view_valid[DEBUG_MAX]; + struct elf_ppc64_opd_data opd_data, *opd; + struct dwarf_sections dwarf_sections; + struct dwarf_data *fileline_altlink = NULL; + + if (!debuginfo) + { + *found_sym = 0; + *found_dwarf = 0; + } + + shdrs_view_valid = 0; + names_view_valid = 0; + symtab_view_valid = 0; + strtab_view_valid = 0; + buildid_view_valid = 0; + buildid_data = NULL; + buildid_size = 0; + debuglink_view_valid = 0; + debuglink_name = NULL; + debuglink_crc = 0; + debugaltlink_view_valid = 0; + debugaltlink_name = NULL; + debugaltlink_buildid_data = NULL; + debugaltlink_buildid_size = 0; + gnu_debugdata_view_valid = 0; + gnu_debugdata_size = 0; + debug_view_valid = 0; + memset (&split_debug_view_valid[0], 0, sizeof split_debug_view_valid); + opd = NULL; + + if (!elf_get_view (state, descriptor, memory, memory_size, 0, sizeof ehdr, + error_callback, data, &ehdr_view)) + goto fail; + + memcpy (&ehdr, ehdr_view.view.data, sizeof ehdr); + + elf_release_view (state, &ehdr_view, error_callback, data); + + if (ehdr.e_ident[EI_MAG0] != ELFMAG0 + || ehdr.e_ident[EI_MAG1] != ELFMAG1 + || ehdr.e_ident[EI_MAG2] != ELFMAG2 + || ehdr.e_ident[EI_MAG3] != ELFMAG3) + { + error_callback (data, "executable file is not ELF", 0); + goto fail; + } + if (ehdr.e_ident[EI_VERSION] != EV_CURRENT) + { + error_callback (data, "executable file is unrecognized ELF version", 0); + goto fail; + } + +#if BACKTRACE_ELF_SIZE == 32 +#define BACKTRACE_ELFCLASS ELFCLASS32 +#else +#define BACKTRACE_ELFCLASS ELFCLASS64 +#endif + + if (ehdr.e_ident[EI_CLASS] != BACKTRACE_ELFCLASS) + { + error_callback (data, "executable file is unexpected ELF class", 0); + goto fail; + } + + if (ehdr.e_ident[EI_DATA] != ELFDATA2LSB + && ehdr.e_ident[EI_DATA] != ELFDATA2MSB) + { + error_callback (data, "executable file has unknown endianness", 0); + goto fail; + } + + /* If the executable is ET_DYN, it is either a PIE, or we are running + directly a shared library with .interp. We need to wait for + dl_iterate_phdr in that case to determine the actual base_address. */ + if (exe && ehdr.e_type == ET_DYN) + return -1; + + shoff = ehdr.e_shoff; + shnum = ehdr.e_shnum; + shstrndx = ehdr.e_shstrndx; + + if ((shnum == 0 || shstrndx == SHN_XINDEX) + && shoff != 0) + { + struct elf_view shdr_view; + const b_elf_shdr *shdr; + + if (!elf_get_view (state, descriptor, memory, memory_size, shoff, + sizeof shdr, error_callback, data, &shdr_view)) + goto fail; + + shdr = (const b_elf_shdr *) shdr_view.view.data; + + if (shnum == 0) + shnum = shdr->sh_size; + + if (shstrndx == SHN_XINDEX) + { + shstrndx = shdr->sh_link; + + /* Versions of the GNU binutils between 2.12 and 2.18 did + not handle objects with more than SHN_LORESERVE sections + correctly. All large section indexes were offset by + 0x100. There is more information at + http://sourceware.org/bugzilla/show_bug.cgi?id-5900 . + Fortunately these object files are easy to detect, as the + GNU binutils always put the section header string table + near the end of the list of sections. Thus if the + section header string table index is larger than the + number of sections, then we know we have to subtract + 0x100 to get the real section index. */ + if (shstrndx >= shnum && shstrndx >= SHN_LORESERVE + 0x100) + shstrndx -= 0x100; + } + + elf_release_view (state, &shdr_view, error_callback, data); + } + + if (shnum == 0 || shstrndx == 0) + goto fail; + + /* To translate PC to file/line when using DWARF, we need to find + the .debug_info and .debug_line sections. */ + + /* Read the section headers, skipping the first one. */ + + if (!elf_get_view (state, descriptor, memory, memory_size, + shoff + sizeof (b_elf_shdr), + (shnum - 1) * sizeof (b_elf_shdr), + error_callback, data, &shdrs_view)) + goto fail; + shdrs_view_valid = 1; + shdrs = (const b_elf_shdr *) shdrs_view.view.data; + + /* Read the section names. */ + + shstrhdr = &shdrs[shstrndx - 1]; + shstr_size = shstrhdr->sh_size; + shstr_off = shstrhdr->sh_offset; + + if (!elf_get_view (state, descriptor, memory, memory_size, shstr_off, + shstrhdr->sh_size, error_callback, data, &names_view)) + goto fail; + names_view_valid = 1; + names = (const char *) names_view.view.data; + + symtab_shndx = 0; + dynsym_shndx = 0; + + memset (sections, 0, sizeof sections); + memset (zsections, 0, sizeof zsections); + + /* Look for the symbol table. */ + for (i = 1; i < shnum; ++i) + { + const b_elf_shdr *shdr; + unsigned int sh_name; + const char *name; + int j; + + shdr = &shdrs[i - 1]; + + if (shdr->sh_type == SHT_SYMTAB) + symtab_shndx = i; + else if (shdr->sh_type == SHT_DYNSYM) + dynsym_shndx = i; + + sh_name = shdr->sh_name; + if (sh_name >= shstr_size) + { + error_callback (data, "ELF section name out of range", 0); + goto fail; + } + + name = names + sh_name; + + for (j = 0; j < (int) DEBUG_MAX; ++j) + { + if (strcmp (name, dwarf_section_names[j]) == 0) + { + sections[j].offset = shdr->sh_offset; + sections[j].size = shdr->sh_size; + sections[j].compressed = (shdr->sh_flags & SHF_COMPRESSED) != 0; + break; + } + } + + if (name[0] == '.' && name[1] == 'z') + { + for (j = 0; j < (int) DEBUG_MAX; ++j) + { + if (strcmp (name + 2, dwarf_section_names[j] + 1) == 0) + { + zsections[j].offset = shdr->sh_offset; + zsections[j].size = shdr->sh_size; + break; + } + } + } + + /* Read the build ID if present. This could check for any + SHT_NOTE section with the right note name and type, but gdb + looks for a specific section name. */ + if ((!debuginfo || with_buildid_data != NULL) + && !buildid_view_valid + && strcmp (name, ".note.gnu.build-id") == 0) + { + const b_elf_note *note; + + if (!elf_get_view (state, descriptor, memory, memory_size, + shdr->sh_offset, shdr->sh_size, error_callback, + data, &buildid_view)) + goto fail; + + buildid_view_valid = 1; + note = (const b_elf_note *) buildid_view.view.data; + if (note->type == NT_GNU_BUILD_ID + && note->namesz == 4 + && strncmp (note->name, "GNU", 4) == 0 + && shdr->sh_size <= 12 + ((note->namesz + 3) & ~ 3) + note->descsz) + { + buildid_data = ¬e->name[0] + ((note->namesz + 3) & ~ 3); + buildid_size = note->descsz; + } + + if (with_buildid_size != 0) + { + if (buildid_size != with_buildid_size) + goto fail; + + if (memcmp (buildid_data, with_buildid_data, buildid_size) != 0) + goto fail; + } + } + + /* Read the debuglink file if present. */ + if (!debuginfo + && !debuglink_view_valid + && strcmp (name, ".gnu_debuglink") == 0) + { + const char *debuglink_data; + size_t crc_offset; + + if (!elf_get_view (state, descriptor, memory, memory_size, + shdr->sh_offset, shdr->sh_size, error_callback, + data, &debuglink_view)) + goto fail; + + debuglink_view_valid = 1; + debuglink_data = (const char *) debuglink_view.view.data; + crc_offset = strnlen (debuglink_data, shdr->sh_size); + crc_offset = (crc_offset + 3) & ~3; + if (crc_offset + 4 <= shdr->sh_size) + { + debuglink_name = debuglink_data; + debuglink_crc = *(const uint32_t*)(debuglink_data + crc_offset); + } + } + + if (!debugaltlink_view_valid + && strcmp (name, ".gnu_debugaltlink") == 0) + { + const char *debugaltlink_data; + size_t debugaltlink_name_len; + + if (!elf_get_view (state, descriptor, memory, memory_size, + shdr->sh_offset, shdr->sh_size, error_callback, + data, &debugaltlink_view)) + goto fail; + + debugaltlink_view_valid = 1; + debugaltlink_data = (const char *) debugaltlink_view.view.data; + debugaltlink_name = debugaltlink_data; + debugaltlink_name_len = strnlen (debugaltlink_data, shdr->sh_size); + if (debugaltlink_name_len < shdr->sh_size) + { + /* Include terminating zero. */ + debugaltlink_name_len += 1; + + debugaltlink_buildid_data + = debugaltlink_data + debugaltlink_name_len; + debugaltlink_buildid_size = shdr->sh_size - debugaltlink_name_len; + } + } + + if (!gnu_debugdata_view_valid + && strcmp (name, ".gnu_debugdata") == 0) + { + if (!elf_get_view (state, descriptor, memory, memory_size, + shdr->sh_offset, shdr->sh_size, error_callback, + data, &gnu_debugdata_view)) + goto fail; + + gnu_debugdata_size = shdr->sh_size; + gnu_debugdata_view_valid = 1; + } + + /* Read the .opd section on PowerPC64 ELFv1. */ + if (ehdr.e_machine == EM_PPC64 + && (ehdr.e_flags & EF_PPC64_ABI) < 2 + && shdr->sh_type == SHT_PROGBITS + && strcmp (name, ".opd") == 0) + { + if (!elf_get_view (state, descriptor, memory, memory_size, + shdr->sh_offset, shdr->sh_size, error_callback, + data, &opd_data.view)) + goto fail; + + opd = &opd_data; + opd->addr = shdr->sh_addr; + opd->data = (const char *) opd_data.view.view.data; + opd->size = shdr->sh_size; + } + } + + if (symtab_shndx == 0) + symtab_shndx = dynsym_shndx; + if (symtab_shndx != 0 && !debuginfo) + { + const b_elf_shdr *symtab_shdr; + unsigned int strtab_shndx; + const b_elf_shdr *strtab_shdr; + struct elf_syminfo_data *sdata; + + symtab_shdr = &shdrs[symtab_shndx - 1]; + strtab_shndx = symtab_shdr->sh_link; + if (strtab_shndx >= shnum) + { + error_callback (data, + "ELF symbol table strtab link out of range", 0); + goto fail; + } + strtab_shdr = &shdrs[strtab_shndx - 1]; + + if (!elf_get_view (state, descriptor, memory, memory_size, + symtab_shdr->sh_offset, symtab_shdr->sh_size, + error_callback, data, &symtab_view)) + goto fail; + symtab_view_valid = 1; + + if (!elf_get_view (state, descriptor, memory, memory_size, + strtab_shdr->sh_offset, strtab_shdr->sh_size, + error_callback, data, &strtab_view)) + goto fail; + strtab_view_valid = 1; + + sdata = ((struct elf_syminfo_data *) + backtrace_alloc (state, sizeof *sdata, error_callback, data)); + if (sdata == NULL) + goto fail; + + if (!elf_initialize_syminfo (state, base_address, + (const unsigned char*)symtab_view.view.data, symtab_shdr->sh_size, + (const unsigned char*)strtab_view.view.data, strtab_shdr->sh_size, + error_callback, data, sdata, opd)) + { + backtrace_free (state, sdata, sizeof *sdata, error_callback, data); + goto fail; + } + + /* We no longer need the symbol table, but we hold on to the + string table permanently. */ + elf_release_view (state, &symtab_view, error_callback, data); + symtab_view_valid = 0; + strtab_view_valid = 0; + + *found_sym = 1; + + elf_add_syminfo_data (state, sdata); + } + + elf_release_view (state, &shdrs_view, error_callback, data); + shdrs_view_valid = 0; + elf_release_view (state, &names_view, error_callback, data); + names_view_valid = 0; + + /* If the debug info is in a separate file, read that one instead. */ + + if (buildid_data != NULL) + { + int d; + + d = elf_open_debugfile_by_buildid (state, buildid_data, buildid_size, + filename, error_callback, data); + if (d >= 0) + { + int ret; + + elf_release_view (state, &buildid_view, error_callback, data); + if (debuglink_view_valid) + elf_release_view (state, &debuglink_view, error_callback, data); + if (debugaltlink_view_valid) + elf_release_view (state, &debugaltlink_view, error_callback, data); + ret = elf_add (state, "", d, NULL, 0, base_address, error_callback, + data, fileline_fn, found_sym, found_dwarf, NULL, 0, + 1, NULL, 0); + if (ret < 0) + backtrace_close (d, error_callback, data); + else if (descriptor >= 0) + backtrace_close (descriptor, error_callback, data); + return ret; + } + } + + if (buildid_view_valid) + { + elf_release_view (state, &buildid_view, error_callback, data); + buildid_view_valid = 0; + } + + if (opd) + { + elf_release_view (state, &opd->view, error_callback, data); + opd = NULL; + } + + if (debuglink_name != NULL) + { + int d; + + d = elf_open_debugfile_by_debuglink (state, filename, debuglink_name, + debuglink_crc, error_callback, + data); + if (d >= 0) + { + int ret; + + elf_release_view (state, &debuglink_view, error_callback, data); + if (debugaltlink_view_valid) + elf_release_view (state, &debugaltlink_view, error_callback, data); + ret = elf_add (state, "", d, NULL, 0, base_address, error_callback, + data, fileline_fn, found_sym, found_dwarf, NULL, 0, + 1, NULL, 0); + if (ret < 0) + backtrace_close (d, error_callback, data); + else if (descriptor >= 0) + backtrace_close(descriptor, error_callback, data); + return ret; + } + } + + if (debuglink_view_valid) + { + elf_release_view (state, &debuglink_view, error_callback, data); + debuglink_view_valid = 0; + } + + if (debugaltlink_name != NULL) + { + int d; + + d = elf_open_debugfile_by_debuglink (state, filename, debugaltlink_name, + 0, error_callback, data); + if (d >= 0) + { + int ret; + + ret = elf_add (state, filename, d, NULL, 0, base_address, + error_callback, data, fileline_fn, found_sym, + found_dwarf, &fileline_altlink, 0, 1, + debugaltlink_buildid_data, debugaltlink_buildid_size); + elf_release_view (state, &debugaltlink_view, error_callback, data); + debugaltlink_view_valid = 0; + if (ret < 0) + { + backtrace_close (d, error_callback, data); + return ret; + } + } + } + + if (debugaltlink_view_valid) + { + elf_release_view (state, &debugaltlink_view, error_callback, data); + debugaltlink_view_valid = 0; + } + + if (gnu_debugdata_view_valid) + { + int ret; + + ret = elf_uncompress_lzma (state, + ((const unsigned char *) + gnu_debugdata_view.view.data), + gnu_debugdata_size, error_callback, data, + &gnu_debugdata_uncompressed, + &gnu_debugdata_uncompressed_size); + + elf_release_view (state, &gnu_debugdata_view, error_callback, data); + gnu_debugdata_view_valid = 0; + + if (ret) + { + ret = elf_add (state, filename, -1, gnu_debugdata_uncompressed, + gnu_debugdata_uncompressed_size, base_address, + error_callback, data, fileline_fn, found_sym, + found_dwarf, NULL, 0, 0, NULL, 0); + if (ret >= 0 && descriptor >= 0) + backtrace_close(descriptor, error_callback, data); + return ret; + } + } + + /* Read all the debug sections in a single view, since they are + probably adjacent in the file. If any of sections are + uncompressed, we never release this view. */ + + min_offset = 0; + max_offset = 0; + debug_size = 0; + for (i = 0; i < (int) DEBUG_MAX; ++i) + { + off_t end; + + if (sections[i].size != 0) + { + if (min_offset == 0 || sections[i].offset < min_offset) + min_offset = sections[i].offset; + end = sections[i].offset + sections[i].size; + if (end > max_offset) + max_offset = end; + debug_size += sections[i].size; + } + if (zsections[i].size != 0) + { + if (min_offset == 0 || zsections[i].offset < min_offset) + min_offset = zsections[i].offset; + end = zsections[i].offset + zsections[i].size; + if (end > max_offset) + max_offset = end; + debug_size += zsections[i].size; + } + } + if (min_offset == 0 || max_offset == 0) + { + if (descriptor >= 0) + { + if (!backtrace_close (descriptor, error_callback, data)) + goto fail; + } + return 1; + } + + /* If the total debug section size is large, assume that there are + gaps between the sections, and read them individually. */ + + if (max_offset - min_offset < 0x20000000 + || max_offset - min_offset < debug_size + 0x10000) + { + if (!elf_get_view (state, descriptor, memory, memory_size, min_offset, + max_offset - min_offset, error_callback, data, + &debug_view)) + goto fail; + debug_view_valid = 1; + } + else + { + memset (&split_debug_view[0], 0, sizeof split_debug_view); + for (i = 0; i < (int) DEBUG_MAX; ++i) + { + struct debug_section_info *dsec; + + if (sections[i].size != 0) + dsec = §ions[i]; + else if (zsections[i].size != 0) + dsec = &zsections[i]; + else + continue; + + if (!elf_get_view (state, descriptor, memory, memory_size, + dsec->offset, dsec->size, error_callback, data, + &split_debug_view[i])) + goto fail; + split_debug_view_valid[i] = 1; + + if (sections[i].size != 0) + sections[i].data = ((const unsigned char *) + split_debug_view[i].view.data); + else + zsections[i].data = ((const unsigned char *) + split_debug_view[i].view.data); + } + } + + /* We've read all we need from the executable. */ + if (descriptor >= 0) + { + if (!backtrace_close (descriptor, error_callback, data)) + goto fail; + descriptor = -1; + } + + using_debug_view = 0; + if (debug_view_valid) + { + for (i = 0; i < (int) DEBUG_MAX; ++i) + { + if (sections[i].size == 0) + sections[i].data = NULL; + else + { + sections[i].data = ((const unsigned char *) debug_view.view.data + + (sections[i].offset - min_offset)); + ++using_debug_view; + } + + if (zsections[i].size == 0) + zsections[i].data = NULL; + else + zsections[i].data = ((const unsigned char *) debug_view.view.data + + (zsections[i].offset - min_offset)); + } + } + + /* Uncompress the old format (--compress-debug-sections=zlib-gnu). */ + + zdebug_table = NULL; + for (i = 0; i < (int) DEBUG_MAX; ++i) + { + if (sections[i].size == 0 && zsections[i].size > 0) + { + unsigned char *uncompressed_data; + size_t uncompressed_size; + + if (zdebug_table == NULL) + { + zdebug_table = ((uint16_t *) + backtrace_alloc (state, ZLIB_TABLE_SIZE, + error_callback, data)); + if (zdebug_table == NULL) + goto fail; + } + + uncompressed_data = NULL; + uncompressed_size = 0; + if (!elf_uncompress_zdebug (state, zsections[i].data, + zsections[i].size, zdebug_table, + error_callback, data, + &uncompressed_data, &uncompressed_size)) + goto fail; + sections[i].data = uncompressed_data; + sections[i].size = uncompressed_size; + sections[i].compressed = 0; + + if (split_debug_view_valid[i]) + { + elf_release_view (state, &split_debug_view[i], + error_callback, data); + split_debug_view_valid[i] = 0; + } + } + } + + if (zdebug_table != NULL) + { + backtrace_free (state, zdebug_table, ZLIB_TABLE_SIZE, + error_callback, data); + zdebug_table = NULL; + } + + /* Uncompress the official ELF format + (--compress-debug-sections=zlib-gabi, --compress-debug-sections=zstd). */ + for (i = 0; i < (int) DEBUG_MAX; ++i) + { + unsigned char *uncompressed_data; + size_t uncompressed_size; + + if (sections[i].size == 0 || !sections[i].compressed) + continue; + + if (zdebug_table == NULL) + { + zdebug_table = ((uint16_t *) + backtrace_alloc (state, ZDEBUG_TABLE_SIZE, + error_callback, data)); + if (zdebug_table == NULL) + goto fail; + } + + uncompressed_data = NULL; + uncompressed_size = 0; + if (!elf_uncompress_chdr (state, sections[i].data, sections[i].size, + zdebug_table, error_callback, data, + &uncompressed_data, &uncompressed_size)) + goto fail; + sections[i].data = uncompressed_data; + sections[i].size = uncompressed_size; + sections[i].compressed = 0; + + if (debug_view_valid) + --using_debug_view; + else if (split_debug_view_valid[i]) + { + elf_release_view (state, &split_debug_view[i], error_callback, data); + split_debug_view_valid[i] = 0; + } + } + + if (zdebug_table != NULL) + backtrace_free (state, zdebug_table, ZDEBUG_TABLE_SIZE, + error_callback, data); + + if (debug_view_valid && using_debug_view == 0) + { + elf_release_view (state, &debug_view, error_callback, data); + debug_view_valid = 0; + } + + for (i = 0; i < (int) DEBUG_MAX; ++i) + { + dwarf_sections.data[i] = sections[i].data; + dwarf_sections.size[i] = sections[i].size; + } + + if (!backtrace_dwarf_add (state, base_address, &dwarf_sections, + ehdr.e_ident[EI_DATA] == ELFDATA2MSB, + fileline_altlink, + error_callback, data, fileline_fn, + fileline_entry)) + goto fail; + + *found_dwarf = 1; + + return 1; + + fail: + if (shdrs_view_valid) + elf_release_view (state, &shdrs_view, error_callback, data); + if (names_view_valid) + elf_release_view (state, &names_view, error_callback, data); + if (symtab_view_valid) + elf_release_view (state, &symtab_view, error_callback, data); + if (strtab_view_valid) + elf_release_view (state, &strtab_view, error_callback, data); + if (debuglink_view_valid) + elf_release_view (state, &debuglink_view, error_callback, data); + if (debugaltlink_view_valid) + elf_release_view (state, &debugaltlink_view, error_callback, data); + if (gnu_debugdata_view_valid) + elf_release_view (state, &gnu_debugdata_view, error_callback, data); + if (buildid_view_valid) + elf_release_view (state, &buildid_view, error_callback, data); + if (debug_view_valid) + elf_release_view (state, &debug_view, error_callback, data); + for (i = 0; i < (int) DEBUG_MAX; ++i) + { + if (split_debug_view_valid[i]) + elf_release_view (state, &split_debug_view[i], error_callback, data); + } + if (opd) + elf_release_view (state, &opd->view, error_callback, data); + if (descriptor >= 0) + backtrace_close (descriptor, error_callback, data); + return 0; +} + +/* Data passed to phdr_callback. */ + +struct phdr_data +{ + struct backtrace_state *state; + backtrace_error_callback error_callback; + void *data; + fileline *fileline_fn; + int *found_sym; + int *found_dwarf; + const char *exe_filename; + int exe_descriptor; +}; + +/* Callback passed to dl_iterate_phdr. Load debug info from shared + libraries. */ + +struct PhdrIterate +{ + char* dlpi_name; + ElfW(Addr) dlpi_addr; + ElfW(Addr) dlpi_end_addr; +}; +FastVector s_phdrData(16); + +struct ElfAddrRange +{ + ElfW(Addr) dlpi_addr; + ElfW(Addr) dlpi_end_addr; +}; +FastVector s_sortedKnownElfRanges(16); + +static int address_in_known_elf_ranges(uintptr_t pc) +{ + auto it = std::lower_bound( s_sortedKnownElfRanges.begin(), s_sortedKnownElfRanges.end(), pc, + []( const ElfAddrRange& lhs, const uintptr_t rhs ) { return uintptr_t(lhs.dlpi_addr) > rhs; } ); + if( it != s_sortedKnownElfRanges.end() && pc <= it->dlpi_end_addr ) + { + return true; + } + return false; +} + +static int +phdr_callback_mock (struct dl_phdr_info *info, size_t size ATTRIBUTE_UNUSED, + void *pdata) +{ + if( address_in_known_elf_ranges(info->dlpi_addr) ) + { + return 0; + } + + auto ptr = s_phdrData.push_next(); + if (info->dlpi_name) + { + size_t sz = strlen (info->dlpi_name) + 1; + ptr->dlpi_name = (char*)tracy_malloc (sz); + memcpy (ptr->dlpi_name, info->dlpi_name, sz); + } + else ptr->dlpi_name = nullptr; + ptr->dlpi_addr = info->dlpi_addr; + + // calculate the end address as well, so we can quickly determine if a PC is within the range of this image + ptr->dlpi_end_addr = uintptr_t(info->dlpi_addr) + (info->dlpi_phnum ? uintptr_t( + info->dlpi_phdr[info->dlpi_phnum - 1].p_vaddr + + info->dlpi_phdr[info->dlpi_phnum - 1].p_memsz) : 0); + + return 0; +} + +static int +#ifdef __i386__ +__attribute__ ((__force_align_arg_pointer__)) +#endif +phdr_callback (struct PhdrIterate *info, void *pdata) +{ + struct phdr_data *pd = (struct phdr_data *) pdata; + const char *filename; + int descriptor; + int does_not_exist; + fileline elf_fileline_fn; + int found_dwarf; + + /* There is not much we can do if we don't have the module name, + unless executable is ET_DYN, where we expect the very first + phdr_callback to be for the PIE. */ + if (info->dlpi_name == NULL || info->dlpi_name[0] == '\0') + { + if (pd->exe_descriptor == -1) + return 0; + filename = pd->exe_filename; + descriptor = pd->exe_descriptor; + pd->exe_descriptor = -1; + } + else + { + if (pd->exe_descriptor != -1) + { + backtrace_close (pd->exe_descriptor, pd->error_callback, pd->data); + pd->exe_descriptor = -1; + } + + filename = info->dlpi_name; + descriptor = backtrace_open (info->dlpi_name, pd->error_callback, + pd->data, &does_not_exist); + if (descriptor < 0) + return 0; + } + + if (elf_add (pd->state, filename, descriptor, NULL, 0, info->dlpi_addr, + pd->error_callback, pd->data, &elf_fileline_fn, pd->found_sym, + &found_dwarf, NULL, 0, 0, NULL, 0)) + { + if (found_dwarf) + { + *pd->found_dwarf = 1; + *pd->fileline_fn = elf_fileline_fn; + } + } + + return 0; +} + +static int elf_iterate_phdr_and_add_new_files(phdr_data *pd) +{ + assert(s_phdrData.empty()); + // dl_iterate_phdr, will only add entries for elf files loaded in a previouly unseen range + dl_iterate_phdr(phdr_callback_mock, nullptr); + + if(s_phdrData.size() == 0) + { + return 0; + } + + uint32_t headersAdded = 0; + for (auto &v : s_phdrData) + { + phdr_callback(&v, (void *)pd); + + auto newEntry = s_sortedKnownElfRanges.push_next(); + newEntry->dlpi_addr = v.dlpi_addr; + newEntry->dlpi_end_addr = v.dlpi_end_addr; + + tracy_free(v.dlpi_name); + + headersAdded++; + } + + s_phdrData.clear(); + + std::sort( s_sortedKnownElfRanges.begin(), s_sortedKnownElfRanges.end(), + []( const ElfAddrRange& lhs, const ElfAddrRange& rhs ) { return lhs.dlpi_addr > rhs.dlpi_addr; } ); + + return headersAdded; +} + +#ifdef TRACY_LIBBACKTRACE_ELF_DYNLOAD_SUPPORT +/* Request an elf entry update if the pc passed in is not in any of the known elf ranges. +This could mean that new images were dlopened and we need to add those new elf entries */ +static int elf_refresh_address_ranges_if_needed(struct backtrace_state *state, uintptr_t pc) +{ + if ( address_in_known_elf_ranges(pc) ) + { + return 0; + } + + struct phdr_data pd; + int found_sym = 0; + int found_dwarf = 0; + fileline fileline_fn = nullptr; + pd.state = state; + pd.error_callback = nullptr; + pd.data = nullptr; + pd.fileline_fn = &fileline_fn; + pd.found_sym = &found_sym; + pd.found_dwarf = &found_dwarf; + pd.exe_filename = nullptr; + pd.exe_descriptor = -1; + + return elf_iterate_phdr_and_add_new_files(&pd); +} +#endif //#ifdef TRACY_LIBBACKTRACE_ELF_DYNLOAD_SUPPORT + +/* Initialize the backtrace data we need from an ELF executable. At + the ELF level, all we need to do is find the debug info + sections. */ + +int +backtrace_initialize (struct backtrace_state *state, const char *filename, + int descriptor, backtrace_error_callback error_callback, + void *data, fileline *fileline_fn) +{ + int ret; + int found_sym; + int found_dwarf; + fileline elf_fileline_fn = elf_nodebug; + struct phdr_data pd; + + ret = elf_add (state, filename, descriptor, NULL, 0, 0, error_callback, data, + &elf_fileline_fn, &found_sym, &found_dwarf, NULL, 1, 0, NULL, + 0); + if (!ret) + return 0; + + pd.state = state; + pd.error_callback = error_callback; + pd.data = data; + pd.fileline_fn = &elf_fileline_fn; + pd.found_sym = &found_sym; + pd.found_dwarf = &found_dwarf; + pd.exe_filename = filename; + pd.exe_descriptor = ret < 0 ? descriptor : -1; + + elf_iterate_phdr_and_add_new_files(&pd); + + if (!state->threaded) + { + if (found_sym) + state->syminfo_fn = elf_syminfo; + else if (state->syminfo_fn == NULL) + state->syminfo_fn = elf_nosyms; + } + else + { + if (found_sym) + backtrace_atomic_store_pointer (&state->syminfo_fn, &elf_syminfo); + else + (void) __sync_bool_compare_and_swap (&state->syminfo_fn, NULL, + elf_nosyms); + } + + if (!state->threaded) + *fileline_fn = state->fileline_fn; + else + *fileline_fn = backtrace_atomic_load_pointer (&state->fileline_fn); + + if (*fileline_fn == NULL || *fileline_fn == elf_nodebug) + *fileline_fn = elf_fileline_fn; + + // install an address range refresh callback so we can cope with dynamically loaded elf files +#ifdef TRACY_LIBBACKTRACE_ELF_DYNLOAD_SUPPORT + state->request_known_address_ranges_refresh_fn = elf_refresh_address_ranges_if_needed; +#else + state->request_known_address_ranges_refresh_fn = NULL; +#endif + + return 1; +} + +} diff --git a/externals/tracy/public/libbacktrace/fileline.cpp b/externals/tracy/public/libbacktrace/fileline.cpp new file mode 100644 index 000000000..8645d754a --- /dev/null +++ b/externals/tracy/public/libbacktrace/fileline.cpp @@ -0,0 +1,351 @@ +/* fileline.c -- Get file and line number information in a backtrace. + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#include "config.h" + +#include +#include +#include +#include +#include +#include + +#if defined (HAVE_KERN_PROC_ARGS) || defined (HAVE_KERN_PROC) +#include +#endif + +#ifdef HAVE_MACH_O_DYLD_H +#include +#endif + +#include "backtrace.hpp" +#include "internal.hpp" + +#ifndef HAVE_GETEXECNAME +#define getexecname() NULL +#endif + +namespace tracy +{ + +#if !defined (HAVE_KERN_PROC_ARGS) && !defined (HAVE_KERN_PROC) + +#define sysctl_exec_name1(state, error_callback, data) NULL +#define sysctl_exec_name2(state, error_callback, data) NULL + +#else /* defined (HAVE_KERN_PROC_ARGS) || |defined (HAVE_KERN_PROC) */ + +static char * +sysctl_exec_name (struct backtrace_state *state, + int mib0, int mib1, int mib2, int mib3, + backtrace_error_callback error_callback, void *data) +{ + int mib[4]; + size_t len; + char *name; + size_t rlen; + + mib[0] = mib0; + mib[1] = mib1; + mib[2] = mib2; + mib[3] = mib3; + + if (sysctl (mib, 4, NULL, &len, NULL, 0) < 0) + return NULL; + name = (char *) backtrace_alloc (state, len, error_callback, data); + if (name == NULL) + return NULL; + rlen = len; + if (sysctl (mib, 4, name, &rlen, NULL, 0) < 0) + { + backtrace_free (state, name, len, error_callback, data); + return NULL; + } + return name; +} + +#ifdef HAVE_KERN_PROC_ARGS + +static char * +sysctl_exec_name1 (struct backtrace_state *state, + backtrace_error_callback error_callback, void *data) +{ + /* This variant is used on NetBSD. */ + return sysctl_exec_name (state, CTL_KERN, KERN_PROC_ARGS, -1, + KERN_PROC_PATHNAME, error_callback, data); +} + +#else + +#define sysctl_exec_name1(state, error_callback, data) NULL + +#endif + +#ifdef HAVE_KERN_PROC + +static char * +sysctl_exec_name2 (struct backtrace_state *state, + backtrace_error_callback error_callback, void *data) +{ + /* This variant is used on FreeBSD. */ + return sysctl_exec_name (state, CTL_KERN, KERN_PROC, KERN_PROC_PATHNAME, -1, + error_callback, data); +} + +#else + +#define sysctl_exec_name2(state, error_callback, data) NULL + +#endif + +#endif /* defined (HAVE_KERN_PROC_ARGS) || |defined (HAVE_KERN_PROC) */ + +#ifdef HAVE_MACH_O_DYLD_H + +static char * +macho_get_executable_path (struct backtrace_state *state, + backtrace_error_callback error_callback, void *data) +{ + uint32_t len; + char *name; + + len = 0; + if (_NSGetExecutablePath (NULL, &len) == 0) + return NULL; + name = (char *) backtrace_alloc (state, len, error_callback, data); + if (name == NULL) + return NULL; + if (_NSGetExecutablePath (name, &len) != 0) + { + backtrace_free (state, name, len, error_callback, data); + return NULL; + } + return name; +} + +#else /* !defined (HAVE_MACH_O_DYLD_H) */ + +#define macho_get_executable_path(state, error_callback, data) NULL + +#endif /* !defined (HAVE_MACH_O_DYLD_H) */ + +/* Initialize the fileline information from the executable. Returns 1 + on success, 0 on failure. */ + +static int +fileline_initialize (struct backtrace_state *state, + backtrace_error_callback error_callback, void *data) +{ + int failed; + fileline fileline_fn; + int pass; + int called_error_callback; + int descriptor; + const char *filename; + char buf[64]; + + if (!state->threaded) + failed = state->fileline_initialization_failed; + else + failed = backtrace_atomic_load_int (&state->fileline_initialization_failed); + + if (failed) + { + error_callback (data, "failed to read executable information", -1); + return 0; + } + + if (!state->threaded) + fileline_fn = state->fileline_fn; + else + fileline_fn = backtrace_atomic_load_pointer (&state->fileline_fn); + if (fileline_fn != NULL) + return 1; + + /* We have not initialized the information. Do it now. */ + + descriptor = -1; + called_error_callback = 0; + for (pass = 0; pass < 8; ++pass) + { + int does_not_exist; + + switch (pass) + { + case 0: + filename = state->filename; + break; + case 1: + filename = getexecname (); + break; + case 2: + filename = "/proc/self/exe"; + break; + case 3: + filename = "/proc/curproc/file"; + break; + case 4: + snprintf (buf, sizeof (buf), "/proc/%ld/object/a.out", + (long) getpid ()); + filename = buf; + break; + case 5: + filename = sysctl_exec_name1 (state, error_callback, data); + break; + case 6: + filename = sysctl_exec_name2 (state, error_callback, data); + break; + case 7: + filename = macho_get_executable_path (state, error_callback, data); + break; + default: + abort (); + } + + if (filename == NULL) + continue; + + descriptor = backtrace_open (filename, error_callback, data, + &does_not_exist); + if (descriptor < 0 && !does_not_exist) + { + called_error_callback = 1; + break; + } + if (descriptor >= 0) + break; + } + + if (descriptor < 0) + { + if (!called_error_callback) + { + if (state->filename != NULL) + error_callback (data, state->filename, ENOENT); + else + error_callback (data, + "libbacktrace could not find executable to open", + 0); + } + failed = 1; + } + + if (!failed) + { + if (!backtrace_initialize (state, filename, descriptor, error_callback, + data, &fileline_fn)) + failed = 1; + } + + if (failed) + { + if (!state->threaded) + state->fileline_initialization_failed = 1; + else + backtrace_atomic_store_int (&state->fileline_initialization_failed, 1); + return 0; + } + + if (!state->threaded) + state->fileline_fn = fileline_fn; + else + { + backtrace_atomic_store_pointer (&state->fileline_fn, fileline_fn); + + /* Note that if two threads initialize at once, one of the data + sets may be leaked. */ + } + + return 1; +} + +/* Given a PC, find the file name, line number, and function name. */ + +int +backtrace_pcinfo (struct backtrace_state *state, uintptr_t pc, + backtrace_full_callback callback, + backtrace_error_callback error_callback, void *data) +{ + if (!fileline_initialize (state, error_callback, data)) + return 0; + + if (state->fileline_initialization_failed) + return 0; + + return state->fileline_fn (state, pc, callback, error_callback, data); +} + +/* Given a PC, find the symbol for it, and its value. */ + +int +backtrace_syminfo (struct backtrace_state *state, uintptr_t pc, + backtrace_syminfo_callback callback, + backtrace_error_callback error_callback, void *data) +{ + if (!fileline_initialize (state, error_callback, data)) + return 0; + + if (state->fileline_initialization_failed) + return 0; + + state->syminfo_fn (state, pc, callback, error_callback, data); + return 1; +} + +/* A backtrace_syminfo_callback that can call into a + backtrace_full_callback, used when we have a symbol table but no + debug info. */ + +void +backtrace_syminfo_to_full_callback (void *data, uintptr_t pc, + const char *symname, + uintptr_t symval ATTRIBUTE_UNUSED, + uintptr_t symsize ATTRIBUTE_UNUSED) +{ + struct backtrace_call_full *bdata = (struct backtrace_call_full *) data; + + bdata->ret = bdata->full_callback (bdata->full_data, pc, 0, NULL, 0, symname); +} + +/* An error callback that corresponds to + backtrace_syminfo_to_full_callback. */ + +void +backtrace_syminfo_to_full_error_callback (void *data, const char *msg, + int errnum) +{ + struct backtrace_call_full *bdata = (struct backtrace_call_full *) data; + + bdata->full_error_callback (bdata->full_data, msg, errnum); +} + +} diff --git a/externals/tracy/public/libbacktrace/filenames.hpp b/externals/tracy/public/libbacktrace/filenames.hpp new file mode 100644 index 000000000..aa7bd7adf --- /dev/null +++ b/externals/tracy/public/libbacktrace/filenames.hpp @@ -0,0 +1,52 @@ +/* btest.c -- Filename header for libbacktrace library + Copyright (C) 2012-2018 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#ifndef GCC_VERSION +# define GCC_VERSION (__GNUC__ * 1000 + __GNUC_MINOR__) +#endif + +#if (GCC_VERSION < 2007) +# define __attribute__(x) +#endif + +#ifndef ATTRIBUTE_UNUSED +# define ATTRIBUTE_UNUSED __attribute__ ((__unused__)) +#endif + +#if defined(__MSDOS__) || defined(_WIN32) || defined(__OS2__) || defined (__CYGWIN__) +# define IS_DIR_SEPARATOR(c) ((c) == '/' || (c) == '\\') +# define HAS_DRIVE_SPEC(f) ((f)[0] != '\0' && (f)[1] == ':') +# define IS_ABSOLUTE_PATH(f) (IS_DIR_SEPARATOR((f)[0]) || HAS_DRIVE_SPEC(f)) +#else +# define IS_DIR_SEPARATOR(c) ((c) == '/') +# define IS_ABSOLUTE_PATH(f) (IS_DIR_SEPARATOR((f)[0])) +#endif diff --git a/externals/tracy/public/libbacktrace/internal.hpp b/externals/tracy/public/libbacktrace/internal.hpp new file mode 100644 index 000000000..fea298fa2 --- /dev/null +++ b/externals/tracy/public/libbacktrace/internal.hpp @@ -0,0 +1,401 @@ +/* internal.h -- Internal header file for stack backtrace library. + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#ifndef BACKTRACE_INTERNAL_H +#define BACKTRACE_INTERNAL_H + +/* We assume that and "backtrace.h" have already been + included. */ + +#ifndef GCC_VERSION +# define GCC_VERSION (__GNUC__ * 1000 + __GNUC_MINOR__) +#endif + +#if (GCC_VERSION < 2007) +# define __attribute__(x) +#endif + +#ifndef ATTRIBUTE_UNUSED +# define ATTRIBUTE_UNUSED __attribute__ ((__unused__)) +#endif + +#ifndef ATTRIBUTE_MALLOC +# if (GCC_VERSION >= 2096) +# define ATTRIBUTE_MALLOC __attribute__ ((__malloc__)) +# else +# define ATTRIBUTE_MALLOC +# endif +#endif + +#ifndef ATTRIBUTE_FALLTHROUGH +# if (GCC_VERSION >= 7000) +# define ATTRIBUTE_FALLTHROUGH __attribute__ ((__fallthrough__)) +# else +# define ATTRIBUTE_FALLTHROUGH +# endif +#endif + +#ifndef HAVE_SYNC_FUNCTIONS + +/* Define out the sync functions. These should never be called if + they are not available. */ + +#define __sync_bool_compare_and_swap(A, B, C) (abort(), 1) +#define __sync_lock_test_and_set(A, B) (abort(), 0) +#define __sync_lock_release(A) abort() + +#endif /* !defined (HAVE_SYNC_FUNCTIONS) */ + +#ifdef HAVE_ATOMIC_FUNCTIONS + +/* We have the atomic builtin functions. */ + +#define backtrace_atomic_load_pointer(p) \ + __atomic_load_n ((p), __ATOMIC_ACQUIRE) +#define backtrace_atomic_load_int(p) \ + __atomic_load_n ((p), __ATOMIC_ACQUIRE) +#define backtrace_atomic_store_pointer(p, v) \ + __atomic_store_n ((p), (v), __ATOMIC_RELEASE) +#define backtrace_atomic_store_size_t(p, v) \ + __atomic_store_n ((p), (v), __ATOMIC_RELEASE) +#define backtrace_atomic_store_int(p, v) \ + __atomic_store_n ((p), (v), __ATOMIC_RELEASE) + +#else /* !defined (HAVE_ATOMIC_FUNCTIONS) */ +#ifdef HAVE_SYNC_FUNCTIONS + +/* We have the sync functions but not the atomic functions. Define + the atomic ones in terms of the sync ones. */ + +extern void *backtrace_atomic_load_pointer (void *); +extern int backtrace_atomic_load_int (int *); +extern void backtrace_atomic_store_pointer (void *, void *); +extern void backtrace_atomic_store_size_t (size_t *, size_t); +extern void backtrace_atomic_store_int (int *, int); + +#else /* !defined (HAVE_SYNC_FUNCTIONS) */ + +/* We have neither the sync nor the atomic functions. These will + never be called. */ + +#define backtrace_atomic_load_pointer(p) (abort(), (void *) NULL) +#define backtrace_atomic_load_int(p) (abort(), 0) +#define backtrace_atomic_store_pointer(p, v) abort() +#define backtrace_atomic_store_size_t(p, v) abort() +#define backtrace_atomic_store_int(p, v) abort() + +#endif /* !defined (HAVE_SYNC_FUNCTIONS) */ +#endif /* !defined (HAVE_ATOMIC_FUNCTIONS) */ + +namespace tracy +{ + +/* The type of the function that collects file/line information. This + is like backtrace_pcinfo. */ + +typedef int (*fileline) (struct backtrace_state *state, uintptr_t pc, + backtrace_full_callback callback, + backtrace_error_callback error_callback, void *data); + +/* The type of the function that collects symbol information. This is + like backtrace_syminfo. */ + +typedef void (*syminfo) (struct backtrace_state *state, uintptr_t pc, + backtrace_syminfo_callback callback, + backtrace_error_callback error_callback, void *data); + +/* The type of the function that will trigger an known address range refresh + (if pc passed in is for an address whichs lies ourtisde of known ranges) */ +typedef int (*request_known_address_ranges_refresh)(struct backtrace_state *state, + uintptr_t pc); + +/* What the backtrace state pointer points to. */ + +struct backtrace_state +{ + /* The name of the executable. */ + const char *filename; + /* Non-zero if threaded. */ + int threaded; + /* The master lock for fileline_fn, fileline_data, syminfo_fn, + syminfo_data, fileline_initialization_failed and everything the + data pointers point to. */ + void *lock; + /* The function that returns file/line information. */ + fileline fileline_fn; + /* The data to pass to FILELINE_FN. */ + void *fileline_data; + /* The function that returns symbol information. */ + syminfo syminfo_fn; + /* The data to pass to SYMINFO_FN. */ + void *syminfo_data; + /* Whether initializing the file/line information failed. */ + int fileline_initialization_failed; + /* The lock for the freelist. */ + int lock_alloc; + /* The freelist when using mmap. */ + struct backtrace_freelist_struct *freelist; + /* Trigger an known address range refresh */ + request_known_address_ranges_refresh request_known_address_ranges_refresh_fn; +}; + +/* Open a file for reading. Returns -1 on error. If DOES_NOT_EXIST + is not NULL, *DOES_NOT_EXIST will be set to 0 normally and set to 1 + if the file does not exist. If the file does not exist and + DOES_NOT_EXIST is not NULL, the function will return -1 and will + not call ERROR_CALLBACK. On other errors, or if DOES_NOT_EXIST is + NULL, the function will call ERROR_CALLBACK before returning. */ +extern int backtrace_open (const char *filename, + backtrace_error_callback error_callback, + void *data, + int *does_not_exist); + +/* A view of the contents of a file. This supports mmap when + available. A view will remain in memory even after backtrace_close + is called on the file descriptor from which the view was + obtained. */ + +struct backtrace_view +{ + /* The data that the caller requested. */ + const void *data; + /* The base of the view. */ + void *base; + /* The total length of the view. */ + size_t len; +}; + +/* Create a view of SIZE bytes from DESCRIPTOR at OFFSET. Store the + result in *VIEW. Returns 1 on success, 0 on error. */ +extern int backtrace_get_view (struct backtrace_state *state, int descriptor, + off_t offset, uint64_t size, + backtrace_error_callback error_callback, + void *data, struct backtrace_view *view); + +/* Release a view created by backtrace_get_view. */ +extern void backtrace_release_view (struct backtrace_state *state, + struct backtrace_view *view, + backtrace_error_callback error_callback, + void *data); + +/* Close a file opened by backtrace_open. Returns 1 on success, 0 on + error. */ + +extern int backtrace_close (int descriptor, + backtrace_error_callback error_callback, + void *data); + +/* Sort without using memory. */ + +extern void backtrace_qsort (void *base, size_t count, size_t size, + int (*compar) (const void *, const void *)); + +/* Allocate memory. This is like malloc. If ERROR_CALLBACK is NULL, + this does not report an error, it just returns NULL. */ + +extern void *backtrace_alloc (struct backtrace_state *state, size_t size, + backtrace_error_callback error_callback, + void *data) ATTRIBUTE_MALLOC; + +/* Free memory allocated by backtrace_alloc. If ERROR_CALLBACK is + NULL, this does not report an error. */ + +extern void backtrace_free (struct backtrace_state *state, void *mem, + size_t size, + backtrace_error_callback error_callback, + void *data); + +/* A growable vector of some struct. This is used for more efficient + allocation when we don't know the final size of some group of data + that we want to represent as an array. */ + +struct backtrace_vector +{ + /* The base of the vector. */ + void *base; + /* The number of bytes in the vector. */ + size_t size; + /* The number of bytes available at the current allocation. */ + size_t alc; +}; + +/* Grow VEC by SIZE bytes. Return a pointer to the newly allocated + bytes. Note that this may move the entire vector to a new memory + location. Returns NULL on failure. */ + +extern void *backtrace_vector_grow (struct backtrace_state *state, size_t size, + backtrace_error_callback error_callback, + void *data, + struct backtrace_vector *vec); + +/* Finish the current allocation on VEC. Prepare to start a new + allocation. The finished allocation will never be freed. Returns + a pointer to the base of the finished entries, or NULL on + failure. */ + +extern void* backtrace_vector_finish (struct backtrace_state *state, + struct backtrace_vector *vec, + backtrace_error_callback error_callback, + void *data); + +/* Release any extra space allocated for VEC. This may change + VEC->base. Returns 1 on success, 0 on failure. */ + +extern int backtrace_vector_release (struct backtrace_state *state, + struct backtrace_vector *vec, + backtrace_error_callback error_callback, + void *data); + +/* Free the space managed by VEC. This will reset VEC. */ + +static inline void +backtrace_vector_free (struct backtrace_state *state, + struct backtrace_vector *vec, + backtrace_error_callback error_callback, void *data) +{ + vec->alc += vec->size; + vec->size = 0; + backtrace_vector_release (state, vec, error_callback, data); +} + +/* Read initial debug data from a descriptor, and set the + fileline_data, syminfo_fn, and syminfo_data fields of STATE. + Return the fileln_fn field in *FILELN_FN--this is done this way so + that the synchronization code is only implemented once. This is + called after the descriptor has first been opened. It will close + the descriptor if it is no longer needed. Returns 1 on success, 0 + on error. There will be multiple implementations of this function, + for different file formats. Each system will compile the + appropriate one. */ + +extern int backtrace_initialize (struct backtrace_state *state, + const char *filename, + int descriptor, + backtrace_error_callback error_callback, + void *data, + fileline *fileline_fn); + +/* An enum for the DWARF sections we care about. */ + +enum dwarf_section +{ + DEBUG_INFO, + DEBUG_LINE, + DEBUG_ABBREV, + DEBUG_RANGES, + DEBUG_STR, + DEBUG_ADDR, + DEBUG_STR_OFFSETS, + DEBUG_LINE_STR, + DEBUG_RNGLISTS, + + DEBUG_MAX +}; + +/* Data for the DWARF sections we care about. */ + +struct dwarf_sections +{ + const unsigned char *data[DEBUG_MAX]; + size_t size[DEBUG_MAX]; +}; + +/* DWARF data read from a file, used for .gnu_debugaltlink. */ + +struct dwarf_data; + +/* Add file/line information for a DWARF module. */ + +extern int backtrace_dwarf_add (struct backtrace_state *state, + uintptr_t base_address, + const struct dwarf_sections *dwarf_sections, + int is_bigendian, + struct dwarf_data *fileline_altlink, + backtrace_error_callback error_callback, + void *data, fileline *fileline_fn, + struct dwarf_data **fileline_entry); + +/* A data structure to pass to backtrace_syminfo_to_full. */ + +struct backtrace_call_full +{ + backtrace_full_callback full_callback; + backtrace_error_callback full_error_callback; + void *full_data; + int ret; +}; + +/* A backtrace_syminfo_callback that can call into a + backtrace_full_callback, used when we have a symbol table but no + debug info. */ + +extern void backtrace_syminfo_to_full_callback (void *data, uintptr_t pc, + const char *symname, + uintptr_t symval, + uintptr_t symsize); + +/* An error callback that corresponds to + backtrace_syminfo_to_full_callback. */ + +extern void backtrace_syminfo_to_full_error_callback (void *, const char *, + int); + +/* A test-only hook for elf_uncompress_zdebug. */ + +extern int backtrace_uncompress_zdebug (struct backtrace_state *, + const unsigned char *compressed, + size_t compressed_size, + backtrace_error_callback, void *data, + unsigned char **uncompressed, + size_t *uncompressed_size); + +/* A test-only hook for elf_zstd_decompress. */ + +extern int backtrace_uncompress_zstd (struct backtrace_state *, + const unsigned char *compressed, + size_t compressed_size, + backtrace_error_callback, void *data, + unsigned char *uncompressed, + size_t uncompressed_size); + +/* A test-only hook for elf_uncompress_lzma. */ + +extern int backtrace_uncompress_lzma (struct backtrace_state *, + const unsigned char *compressed, + size_t compressed_size, + backtrace_error_callback, void *data, + unsigned char **uncompressed, + size_t *uncompressed_size); + +} + +#endif diff --git a/externals/tracy/public/libbacktrace/macho.cpp b/externals/tracy/public/libbacktrace/macho.cpp new file mode 100644 index 000000000..6cccdabaa --- /dev/null +++ b/externals/tracy/public/libbacktrace/macho.cpp @@ -0,0 +1,1360 @@ +/* elf.c -- Get debug data from a Mach-O file for backtraces. + Copyright (C) 2020-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#include "config.h" + +#include +#include +#include +#include + +#ifdef HAVE_MACH_O_DYLD_H +#include +#endif + +#include "backtrace.hpp" +#include "internal.hpp" + +namespace tracy +{ + +/* Mach-O file header for a 32-bit executable. */ + +struct macho_header_32 +{ + uint32_t magic; /* Magic number (MACH_O_MAGIC_32) */ + uint32_t cputype; /* CPU type */ + uint32_t cpusubtype; /* CPU subtype */ + uint32_t filetype; /* Type of file (object, executable) */ + uint32_t ncmds; /* Number of load commands */ + uint32_t sizeofcmds; /* Total size of load commands */ + uint32_t flags; /* Flags for special features */ +}; + +/* Mach-O file header for a 64-bit executable. */ + +struct macho_header_64 +{ + uint32_t magic; /* Magic number (MACH_O_MAGIC_64) */ + uint32_t cputype; /* CPU type */ + uint32_t cpusubtype; /* CPU subtype */ + uint32_t filetype; /* Type of file (object, executable) */ + uint32_t ncmds; /* Number of load commands */ + uint32_t sizeofcmds; /* Total size of load commands */ + uint32_t flags; /* Flags for special features */ + uint32_t reserved; /* Reserved */ +}; + +/* Mach-O file header for a fat executable. */ + +struct macho_header_fat +{ + uint32_t magic; /* Magic number (MACH_O_MH_(MAGIC|CIGAM)_FAT(_64)?) */ + uint32_t nfat_arch; /* Number of components */ +}; + +/* Values for the header magic field. */ + +#define MACH_O_MH_MAGIC_32 0xfeedface +#define MACH_O_MH_MAGIC_64 0xfeedfacf +#define MACH_O_MH_MAGIC_FAT 0xcafebabe +#define MACH_O_MH_CIGAM_FAT 0xbebafeca +#define MACH_O_MH_MAGIC_FAT_64 0xcafebabf +#define MACH_O_MH_CIGAM_FAT_64 0xbfbafeca + +/* Value for the header filetype field. */ + +#define MACH_O_MH_EXECUTE 0x02 +#define MACH_O_MH_DYLIB 0x06 +#define MACH_O_MH_DSYM 0x0a + +/* A component of a fat file. A fat file starts with a + macho_header_fat followed by nfat_arch instances of this + struct. */ + +struct macho_fat_arch +{ + uint32_t cputype; /* CPU type */ + uint32_t cpusubtype; /* CPU subtype */ + uint32_t offset; /* File offset of this entry */ + uint32_t size; /* Size of this entry */ + uint32_t align; /* Alignment of this entry */ +}; + +/* A component of a 64-bit fat file. This is used if the magic field + is MAGIC_FAT_64. This is only used when some file size or file + offset is too large to represent in the 32-bit format. */ + +struct macho_fat_arch_64 +{ + uint32_t cputype; /* CPU type */ + uint32_t cpusubtype; /* CPU subtype */ + uint64_t offset; /* File offset of this entry */ + uint64_t size; /* Size of this entry */ + uint32_t align; /* Alignment of this entry */ + uint32_t reserved; /* Reserved */ +}; + +/* Values for the fat_arch cputype field (and the header cputype + field). */ + +#define MACH_O_CPU_ARCH_ABI64 0x01000000 + +#define MACH_O_CPU_TYPE_X86 7 +#define MACH_O_CPU_TYPE_ARM 12 +#define MACH_O_CPU_TYPE_PPC 18 + +#define MACH_O_CPU_TYPE_X86_64 (MACH_O_CPU_TYPE_X86 | MACH_O_CPU_ARCH_ABI64) +#define MACH_O_CPU_TYPE_ARM64 (MACH_O_CPU_TYPE_ARM | MACH_O_CPU_ARCH_ABI64) +#define MACH_O_CPU_TYPE_PPC64 (MACH_O_CPU_TYPE_PPC | MACH_O_CPU_ARCH_ABI64) + +/* The header of a load command. */ + +struct macho_load_command +{ + uint32_t cmd; /* The type of load command */ + uint32_t cmdsize; /* Size in bytes of the entire command */ +}; + +/* Values for the load_command cmd field. */ + +#define MACH_O_LC_SEGMENT 0x01 +#define MACH_O_LC_SYMTAB 0x02 +#define MACH_O_LC_SEGMENT_64 0x19 +#define MACH_O_LC_UUID 0x1b + +/* The length of a section of segment name. */ + +#define MACH_O_NAMELEN (16) + +/* LC_SEGMENT load command. */ + +struct macho_segment_command +{ + uint32_t cmd; /* The type of load command (LC_SEGMENT) */ + uint32_t cmdsize; /* Size in bytes of the entire command */ + char segname[MACH_O_NAMELEN]; /* Segment name */ + uint32_t vmaddr; /* Virtual memory address */ + uint32_t vmsize; /* Virtual memory size */ + uint32_t fileoff; /* Offset of data to be mapped */ + uint32_t filesize; /* Size of data in file */ + uint32_t maxprot; /* Maximum permitted virtual protection */ + uint32_t initprot; /* Initial virtual memory protection */ + uint32_t nsects; /* Number of sections in this segment */ + uint32_t flags; /* Flags */ +}; + +/* LC_SEGMENT_64 load command. */ + +struct macho_segment_64_command +{ + uint32_t cmd; /* The type of load command (LC_SEGMENT) */ + uint32_t cmdsize; /* Size in bytes of the entire command */ + char segname[MACH_O_NAMELEN]; /* Segment name */ + uint64_t vmaddr; /* Virtual memory address */ + uint64_t vmsize; /* Virtual memory size */ + uint64_t fileoff; /* Offset of data to be mapped */ + uint64_t filesize; /* Size of data in file */ + uint32_t maxprot; /* Maximum permitted virtual protection */ + uint32_t initprot; /* Initial virtual memory protection */ + uint32_t nsects; /* Number of sections in this segment */ + uint32_t flags; /* Flags */ +}; + +/* LC_SYMTAB load command. */ + +struct macho_symtab_command +{ + uint32_t cmd; /* The type of load command (LC_SEGMENT) */ + uint32_t cmdsize; /* Size in bytes of the entire command */ + uint32_t symoff; /* File offset of symbol table */ + uint32_t nsyms; /* Number of symbols */ + uint32_t stroff; /* File offset of string table */ + uint32_t strsize; /* String table size */ +}; + +/* The length of a Mach-O uuid. */ + +#define MACH_O_UUID_LEN (16) + +/* LC_UUID load command. */ + +struct macho_uuid_command +{ + uint32_t cmd; /* Type of load command (LC_UUID) */ + uint32_t cmdsize; /* Size in bytes of command */ + unsigned char uuid[MACH_O_UUID_LEN]; /* UUID */ +}; + +/* 32-bit section header within a LC_SEGMENT segment. */ + +struct macho_section +{ + char sectname[MACH_O_NAMELEN]; /* Section name */ + char segment[MACH_O_NAMELEN]; /* Segment of this section */ + uint32_t addr; /* Address in memory */ + uint32_t size; /* Section size */ + uint32_t offset; /* File offset */ + uint32_t align; /* Log2 of section alignment */ + uint32_t reloff; /* File offset of relocations */ + uint32_t nreloc; /* Number of relocs for this section */ + uint32_t flags; /* Flags */ + uint32_t reserved1; + uint32_t reserved2; +}; + +/* 64-bit section header within a LC_SEGMENT_64 segment. */ + +struct macho_section_64 +{ + char sectname[MACH_O_NAMELEN]; /* Section name */ + char segment[MACH_O_NAMELEN]; /* Segment of this section */ + uint64_t addr; /* Address in memory */ + uint64_t size; /* Section size */ + uint32_t offset; /* File offset */ + uint32_t align; /* Log2 of section alignment */ + uint32_t reloff; /* File offset of section relocations */ + uint32_t nreloc; /* Number of relocs for this section */ + uint32_t flags; /* Flags */ + uint32_t reserved1; + uint32_t reserved2; + uint32_t reserved3; +}; + +/* 32-bit symbol data. */ + +struct macho_nlist +{ + uint32_t n_strx; /* Index of name in string table */ + uint8_t n_type; /* Type flag */ + uint8_t n_sect; /* Section number */ + uint16_t n_desc; /* Stabs description field */ + uint32_t n_value; /* Value */ +}; + +/* 64-bit symbol data. */ + +struct macho_nlist_64 +{ + uint32_t n_strx; /* Index of name in string table */ + uint8_t n_type; /* Type flag */ + uint8_t n_sect; /* Section number */ + uint16_t n_desc; /* Stabs description field */ + uint64_t n_value; /* Value */ +}; + +/* Value found in nlist n_type field. */ + +#define MACH_O_N_EXT 0x01 /* Extern symbol */ +#define MACH_O_N_ABS 0x02 /* Absolute symbol */ +#define MACH_O_N_SECT 0x0e /* Defined in section */ + +#define MACH_O_N_TYPE 0x0e /* Mask for type bits */ +#define MACH_O_N_STAB 0xe0 /* Stabs debugging symbol */ + +/* Information we keep for a Mach-O symbol. */ + +struct macho_symbol +{ + const char *name; /* Symbol name */ + uintptr_t address; /* Symbol address */ +}; + +/* Information to pass to macho_syminfo. */ + +struct macho_syminfo_data +{ + struct macho_syminfo_data *next; /* Next module */ + struct macho_symbol *symbols; /* Symbols sorted by address */ + size_t count; /* Number of symbols */ +}; + +/* Names of sections, indexed by enum dwarf_section in internal.h. */ + +static const char * const dwarf_section_names[DEBUG_MAX] = +{ + "__debug_info", + "__debug_line", + "__debug_abbrev", + "__debug_ranges", + "__debug_str", + "", /* DEBUG_ADDR */ + "__debug_str_offs", + "", /* DEBUG_LINE_STR */ + "__debug_rnglists" +}; + +/* Forward declaration. */ + +static int macho_add (struct backtrace_state *, const char *, int, off_t, + const unsigned char *, uintptr_t, int, + backtrace_error_callback, void *, fileline *, int *); + +/* A dummy callback function used when we can't find any debug info. */ + +static int +macho_nodebug (struct backtrace_state *state ATTRIBUTE_UNUSED, + uintptr_t pc ATTRIBUTE_UNUSED, + backtrace_full_callback callback ATTRIBUTE_UNUSED, + backtrace_error_callback error_callback, void *data) +{ + error_callback (data, "no debug info in Mach-O executable", -1); + return 0; +} + +/* A dummy callback function used when we can't find a symbol + table. */ + +static void +macho_nosyms (struct backtrace_state *state ATTRIBUTE_UNUSED, + uintptr_t addr ATTRIBUTE_UNUSED, + backtrace_syminfo_callback callback ATTRIBUTE_UNUSED, + backtrace_error_callback error_callback, void *data) +{ + error_callback (data, "no symbol table in Mach-O executable", -1); +} + +/* Add a single DWARF section to DWARF_SECTIONS, if we need the + section. Returns 1 on success, 0 on failure. */ + +static int +macho_add_dwarf_section (struct backtrace_state *state, int descriptor, + const char *sectname, uint32_t offset, uint64_t size, + backtrace_error_callback error_callback, void *data, + struct dwarf_sections *dwarf_sections) +{ + int i; + + for (i = 0; i < (int) DEBUG_MAX; ++i) + { + if (dwarf_section_names[i][0] != '\0' + && strncmp (sectname, dwarf_section_names[i], MACH_O_NAMELEN) == 0) + { + struct backtrace_view section_view; + + /* FIXME: Perhaps it would be better to try to use a single + view to read all the DWARF data, as we try to do for + ELF. */ + + if (!backtrace_get_view (state, descriptor, offset, size, + error_callback, data, §ion_view)) + return 0; + dwarf_sections->data[i] = (const unsigned char *) section_view.data; + dwarf_sections->size[i] = size; + break; + } + } + return 1; +} + +/* Collect DWARF sections from a DWARF segment. Returns 1 on success, + 0 on failure. */ + +static int +macho_add_dwarf_segment (struct backtrace_state *state, int descriptor, + off_t offset, unsigned int cmd, const char *psecs, + size_t sizesecs, unsigned int nsects, + backtrace_error_callback error_callback, void *data, + struct dwarf_sections *dwarf_sections) +{ + size_t sec_header_size; + size_t secoffset; + unsigned int i; + + switch (cmd) + { + case MACH_O_LC_SEGMENT: + sec_header_size = sizeof (struct macho_section); + break; + case MACH_O_LC_SEGMENT_64: + sec_header_size = sizeof (struct macho_section_64); + break; + default: + abort (); + } + + secoffset = 0; + for (i = 0; i < nsects; ++i) + { + if (secoffset + sec_header_size > sizesecs) + { + error_callback (data, "section overflow withing segment", 0); + return 0; + } + + switch (cmd) + { + case MACH_O_LC_SEGMENT: + { + struct macho_section section; + + memcpy (§ion, psecs + secoffset, sizeof section); + macho_add_dwarf_section (state, descriptor, section.sectname, + offset + section.offset, section.size, + error_callback, data, dwarf_sections); + } + break; + + case MACH_O_LC_SEGMENT_64: + { + struct macho_section_64 section; + + memcpy (§ion, psecs + secoffset, sizeof section); + macho_add_dwarf_section (state, descriptor, section.sectname, + offset + section.offset, section.size, + error_callback, data, dwarf_sections); + } + break; + + default: + abort (); + } + + secoffset += sec_header_size; + } + + return 1; +} + +/* Compare struct macho_symbol for qsort. */ + +static int +macho_symbol_compare (const void *v1, const void *v2) +{ + const struct macho_symbol *m1 = (const struct macho_symbol *) v1; + const struct macho_symbol *m2 = (const struct macho_symbol *) v2; + + if (m1->address < m2->address) + return -1; + else if (m1->address > m2->address) + return 1; + else + return 0; +} + +/* Compare an address against a macho_symbol for bsearch. We allocate + one extra entry in the array so that this can safely look at the + next entry. */ + +static int +macho_symbol_search (const void *vkey, const void *ventry) +{ + const uintptr_t *key = (const uintptr_t *) vkey; + const struct macho_symbol *entry = (const struct macho_symbol *) ventry; + uintptr_t addr; + + addr = *key; + if (addr < entry->address) + return -1; + else if (entry->name[0] == '\0' + && entry->address == ~(uintptr_t) 0) + return -1; + else if ((entry + 1)->name[0] == '\0' + && (entry + 1)->address == ~(uintptr_t) 0) + return -1; + else if (addr >= (entry + 1)->address) + return 1; + else + return 0; +} + +/* Return whether the symbol type field indicates a symbol table entry + that we care about: a function or data symbol. */ + +static int +macho_defined_symbol (uint8_t type) +{ + if ((type & MACH_O_N_STAB) != 0) + return 0; + if ((type & MACH_O_N_EXT) != 0) + return 0; + switch (type & MACH_O_N_TYPE) + { + case MACH_O_N_ABS: + return 1; + case MACH_O_N_SECT: + return 1; + default: + return 0; + } +} + +/* Add symbol table information for a Mach-O file. */ + +static int +macho_add_symtab (struct backtrace_state *state, int descriptor, + uintptr_t base_address, int is_64, + off_t symoff, unsigned int nsyms, off_t stroff, + unsigned int strsize, + backtrace_error_callback error_callback, void *data) +{ + size_t symsize; + struct backtrace_view sym_view; + int sym_view_valid; + struct backtrace_view str_view; + int str_view_valid; + size_t ndefs; + size_t symtaboff; + unsigned int i; + size_t macho_symbol_size; + struct macho_symbol *macho_symbols; + unsigned int j; + struct macho_syminfo_data *sdata; + + sym_view_valid = 0; + str_view_valid = 0; + macho_symbol_size = 0; + macho_symbols = NULL; + + if (is_64) + symsize = sizeof (struct macho_nlist_64); + else + symsize = sizeof (struct macho_nlist); + + if (!backtrace_get_view (state, descriptor, symoff, nsyms * symsize, + error_callback, data, &sym_view)) + goto fail; + sym_view_valid = 1; + + if (!backtrace_get_view (state, descriptor, stroff, strsize, + error_callback, data, &str_view)) + return 0; + str_view_valid = 1; + + ndefs = 0; + symtaboff = 0; + for (i = 0; i < nsyms; ++i, symtaboff += symsize) + { + if (is_64) + { + struct macho_nlist_64 nlist; + + memcpy (&nlist, (const char *) sym_view.data + symtaboff, + sizeof nlist); + if (macho_defined_symbol (nlist.n_type)) + ++ndefs; + } + else + { + struct macho_nlist nlist; + + memcpy (&nlist, (const char *) sym_view.data + symtaboff, + sizeof nlist); + if (macho_defined_symbol (nlist.n_type)) + ++ndefs; + } + } + + /* Add 1 to ndefs to make room for a sentinel. */ + macho_symbol_size = (ndefs + 1) * sizeof (struct macho_symbol); + macho_symbols = ((struct macho_symbol *) + backtrace_alloc (state, macho_symbol_size, error_callback, + data)); + if (macho_symbols == NULL) + goto fail; + + j = 0; + symtaboff = 0; + for (i = 0; i < nsyms; ++i, symtaboff += symsize) + { + uint32_t strx; + uint64_t value; + const char *name; + + strx = 0; + value = 0; + if (is_64) + { + struct macho_nlist_64 nlist; + + memcpy (&nlist, (const char *) sym_view.data + symtaboff, + sizeof nlist); + if (!macho_defined_symbol (nlist.n_type)) + continue; + + strx = nlist.n_strx; + value = nlist.n_value; + } + else + { + struct macho_nlist nlist; + + memcpy (&nlist, (const char *) sym_view.data + symtaboff, + sizeof nlist); + if (!macho_defined_symbol (nlist.n_type)) + continue; + + strx = nlist.n_strx; + value = nlist.n_value; + } + + if (strx >= strsize) + { + error_callback (data, "symbol string index out of range", 0); + goto fail; + } + + name = (const char *) str_view.data + strx; + if (name[0] == '_') + ++name; + macho_symbols[j].name = name; + macho_symbols[j].address = value + base_address; + ++j; + } + + sdata = ((struct macho_syminfo_data *) + backtrace_alloc (state, sizeof *sdata, error_callback, data)); + if (sdata == NULL) + goto fail; + + /* We need to keep the string table since it holds the names, but we + can release the symbol table. */ + + backtrace_release_view (state, &sym_view, error_callback, data); + sym_view_valid = 0; + str_view_valid = 0; + + /* Add a trailing sentinel symbol. */ + macho_symbols[j].name = ""; + macho_symbols[j].address = ~(uintptr_t) 0; + + backtrace_qsort (macho_symbols, ndefs + 1, sizeof (struct macho_symbol), + macho_symbol_compare); + + sdata->next = NULL; + sdata->symbols = macho_symbols; + sdata->count = ndefs; + + if (!state->threaded) + { + struct macho_syminfo_data **pp; + + for (pp = (struct macho_syminfo_data **) (void *) &state->syminfo_data; + *pp != NULL; + pp = &(*pp)->next) + ; + *pp = sdata; + } + else + { + while (1) + { + struct macho_syminfo_data **pp; + + pp = (struct macho_syminfo_data **) (void *) &state->syminfo_data; + + while (1) + { + struct macho_syminfo_data *p; + + p = backtrace_atomic_load_pointer (pp); + + if (p == NULL) + break; + + pp = &p->next; + } + + if (__sync_bool_compare_and_swap (pp, NULL, sdata)) + break; + } + } + + return 1; + + fail: + if (macho_symbols != NULL) + backtrace_free (state, macho_symbols, macho_symbol_size, + error_callback, data); + if (sym_view_valid) + backtrace_release_view (state, &sym_view, error_callback, data); + if (str_view_valid) + backtrace_release_view (state, &str_view, error_callback, data); + return 0; +} + +/* Return the symbol name and value for an ADDR. */ + +static void +macho_syminfo (struct backtrace_state *state, uintptr_t addr, + backtrace_syminfo_callback callback, + backtrace_error_callback error_callback ATTRIBUTE_UNUSED, + void *data) +{ + struct macho_syminfo_data *sdata; + struct macho_symbol *sym; + + sym = NULL; + if (!state->threaded) + { + for (sdata = (struct macho_syminfo_data *) state->syminfo_data; + sdata != NULL; + sdata = sdata->next) + { + sym = ((struct macho_symbol *) + bsearch (&addr, sdata->symbols, sdata->count, + sizeof (struct macho_symbol), macho_symbol_search)); + if (sym != NULL) + break; + } + } + else + { + struct macho_syminfo_data **pp; + + pp = (struct macho_syminfo_data **) (void *) &state->syminfo_data; + while (1) + { + sdata = backtrace_atomic_load_pointer (pp); + if (sdata == NULL) + break; + + sym = ((struct macho_symbol *) + bsearch (&addr, sdata->symbols, sdata->count, + sizeof (struct macho_symbol), macho_symbol_search)); + if (sym != NULL) + break; + + pp = &sdata->next; + } + } + + if (sym == NULL) + callback (data, addr, NULL, 0, 0); + else + callback (data, addr, sym->name, sym->address, 0); +} + +/* Look through a fat file to find the relevant executable. Returns 1 + on success, 0 on failure (in both cases descriptor is closed). */ + +static int +macho_add_fat (struct backtrace_state *state, const char *filename, + int descriptor, int swapped, off_t offset, + const unsigned char *match_uuid, uintptr_t base_address, + int skip_symtab, uint32_t nfat_arch, int is_64, + backtrace_error_callback error_callback, void *data, + fileline *fileline_fn, int *found_sym) +{ + int arch_view_valid; + unsigned int cputype; + size_t arch_size; + struct backtrace_view arch_view; + unsigned int i; + + arch_view_valid = 0; + +#if defined (__x86_64__) + cputype = MACH_O_CPU_TYPE_X86_64; +#elif defined (__i386__) + cputype = MACH_O_CPU_TYPE_X86; +#elif defined (__aarch64__) + cputype = MACH_O_CPU_TYPE_ARM64; +#elif defined (__arm__) + cputype = MACH_O_CPU_TYPE_ARM; +#elif defined (__ppc__) + cputype = MACH_O_CPU_TYPE_PPC; +#elif defined (__ppc64__) + cputype = MACH_O_CPU_TYPE_PPC64; +#else + error_callback (data, "unknown Mach-O architecture", 0); + goto fail; +#endif + + if (is_64) + arch_size = sizeof (struct macho_fat_arch_64); + else + arch_size = sizeof (struct macho_fat_arch); + + if (!backtrace_get_view (state, descriptor, offset, + nfat_arch * arch_size, + error_callback, data, &arch_view)) + goto fail; + + for (i = 0; i < nfat_arch; ++i) + { + uint32_t fcputype; + uint64_t foffset; + + if (is_64) + { + struct macho_fat_arch_64 fat_arch_64; + + memcpy (&fat_arch_64, + (const char *) arch_view.data + i * arch_size, + arch_size); + fcputype = fat_arch_64.cputype; + foffset = fat_arch_64.offset; + if (swapped) + { + fcputype = __builtin_bswap32 (fcputype); + foffset = __builtin_bswap64 (foffset); + } + } + else + { + struct macho_fat_arch fat_arch_32; + + memcpy (&fat_arch_32, + (const char *) arch_view.data + i * arch_size, + arch_size); + fcputype = fat_arch_32.cputype; + foffset = (uint64_t) fat_arch_32.offset; + if (swapped) + { + fcputype = __builtin_bswap32 (fcputype); + foffset = (uint64_t) __builtin_bswap32 ((uint32_t) foffset); + } + } + + if (fcputype == cputype) + { + /* FIXME: What about cpusubtype? */ + backtrace_release_view (state, &arch_view, error_callback, data); + return macho_add (state, filename, descriptor, foffset, match_uuid, + base_address, skip_symtab, error_callback, data, + fileline_fn, found_sym); + } + } + + error_callback (data, "could not find executable in fat file", 0); + + fail: + if (arch_view_valid) + backtrace_release_view (state, &arch_view, error_callback, data); + if (descriptor != -1) + backtrace_close (descriptor, error_callback, data); + return 0; +} + +/* Look for the dsym file for FILENAME. This is called if FILENAME + does not have debug info or a symbol table. Returns 1 on success, + 0 on failure. */ + +static int +macho_add_dsym (struct backtrace_state *state, const char *filename, + uintptr_t base_address, const unsigned char *uuid, + backtrace_error_callback error_callback, void *data, + fileline* fileline_fn) +{ + const char *p; + const char *dirname; + char *diralc; + size_t dirnamelen; + const char *basename; + size_t basenamelen; + const char *dsymsuffixdir; + size_t dsymsuffixdirlen; + size_t dsymlen; + char *dsym; + char *ps; + int d; + int does_not_exist; + int dummy_found_sym; + + diralc = NULL; + dirnamelen = 0; + dsym = NULL; + dsymlen = 0; + + p = strrchr (filename, '/'); + if (p == NULL) + { + dirname = "."; + dirnamelen = 1; + basename = filename; + basenamelen = strlen (basename); + diralc = NULL; + } + else + { + dirnamelen = p - filename; + diralc = (char*)backtrace_alloc (state, dirnamelen + 1, error_callback, data); + if (diralc == NULL) + goto fail; + memcpy (diralc, filename, dirnamelen); + diralc[dirnamelen] = '\0'; + dirname = diralc; + basename = p + 1; + basenamelen = strlen (basename); + } + + dsymsuffixdir = ".dSYM/Contents/Resources/DWARF/"; + dsymsuffixdirlen = strlen (dsymsuffixdir); + + dsymlen = (dirnamelen + + 1 + + basenamelen + + dsymsuffixdirlen + + basenamelen + + 1); + dsym = (char*)backtrace_alloc (state, dsymlen, error_callback, data); + if (dsym == NULL) + goto fail; + + ps = dsym; + memcpy (ps, dirname, dirnamelen); + ps += dirnamelen; + *ps++ = '/'; + memcpy (ps, basename, basenamelen); + ps += basenamelen; + memcpy (ps, dsymsuffixdir, dsymsuffixdirlen); + ps += dsymsuffixdirlen; + memcpy (ps, basename, basenamelen); + ps += basenamelen; + *ps = '\0'; + + if (diralc != NULL) + { + backtrace_free (state, diralc, dirnamelen + 1, error_callback, data); + diralc = NULL; + } + + d = backtrace_open (dsym, error_callback, data, &does_not_exist); + if (d < 0) + { + /* The file does not exist, so we can't read the debug info. + Just return success. */ + backtrace_free (state, dsym, dsymlen, error_callback, data); + return 1; + } + + if (!macho_add (state, dsym, d, 0, uuid, base_address, 1, + error_callback, data, fileline_fn, &dummy_found_sym)) + goto fail; + + backtrace_free (state, dsym, dsymlen, error_callback, data); + + return 1; + + fail: + if (dsym != NULL) + backtrace_free (state, dsym, dsymlen, error_callback, data); + if (diralc != NULL) + backtrace_free (state, diralc, dirnamelen, error_callback, data); + return 0; +} + +/* Add the backtrace data for a Macho-O file. Returns 1 on success, 0 + on failure (in both cases descriptor is closed). + + FILENAME: the name of the executable. + DESCRIPTOR: an open descriptor for the executable, closed here. + OFFSET: the offset within the file of this executable, for fat files. + MATCH_UUID: if not NULL, UUID that must match. + BASE_ADDRESS: the load address of the executable. + SKIP_SYMTAB: if non-zero, ignore the symbol table; used for dSYM files. + FILELINE_FN: set to the fileline function, by backtrace_dwarf_add. + FOUND_SYM: set to non-zero if we found the symbol table. +*/ + +static int +macho_add (struct backtrace_state *state, const char *filename, int descriptor, + off_t offset, const unsigned char *match_uuid, + uintptr_t base_address, int skip_symtab, + backtrace_error_callback error_callback, void *data, + fileline *fileline_fn, int *found_sym) +{ + struct backtrace_view header_view; + struct macho_header_32 header; + off_t hdroffset; + int is_64; + struct backtrace_view cmds_view; + int cmds_view_valid; + struct dwarf_sections dwarf_sections; + int have_dwarf; + unsigned char uuid[MACH_O_UUID_LEN]; + int have_uuid; + size_t cmdoffset; + unsigned int i; + + *found_sym = 0; + + cmds_view_valid = 0; + + /* The 32-bit and 64-bit file headers start out the same, so we can + just always read the 32-bit version. A fat header is shorter but + it will always be followed by data, so it's OK to read extra. */ + + if (!backtrace_get_view (state, descriptor, offset, + sizeof (struct macho_header_32), + error_callback, data, &header_view)) + goto fail; + + memcpy (&header, header_view.data, sizeof header); + + backtrace_release_view (state, &header_view, error_callback, data); + + switch (header.magic) + { + case MACH_O_MH_MAGIC_32: + is_64 = 0; + hdroffset = offset + sizeof (struct macho_header_32); + break; + case MACH_O_MH_MAGIC_64: + is_64 = 1; + hdroffset = offset + sizeof (struct macho_header_64); + break; + case MACH_O_MH_MAGIC_FAT: + case MACH_O_MH_MAGIC_FAT_64: + { + struct macho_header_fat fat_header; + + hdroffset = offset + sizeof (struct macho_header_fat); + memcpy (&fat_header, &header, sizeof fat_header); + return macho_add_fat (state, filename, descriptor, 0, hdroffset, + match_uuid, base_address, skip_symtab, + fat_header.nfat_arch, + header.magic == MACH_O_MH_MAGIC_FAT_64, + error_callback, data, fileline_fn, found_sym); + } + case MACH_O_MH_CIGAM_FAT: + case MACH_O_MH_CIGAM_FAT_64: + { + struct macho_header_fat fat_header; + uint32_t nfat_arch; + + hdroffset = offset + sizeof (struct macho_header_fat); + memcpy (&fat_header, &header, sizeof fat_header); + nfat_arch = __builtin_bswap32 (fat_header.nfat_arch); + return macho_add_fat (state, filename, descriptor, 1, hdroffset, + match_uuid, base_address, skip_symtab, + nfat_arch, + header.magic == MACH_O_MH_CIGAM_FAT_64, + error_callback, data, fileline_fn, found_sym); + } + default: + error_callback (data, "executable file is not in Mach-O format", 0); + goto fail; + } + + switch (header.filetype) + { + case MACH_O_MH_EXECUTE: + case MACH_O_MH_DYLIB: + case MACH_O_MH_DSYM: + break; + default: + error_callback (data, "executable file is not an executable", 0); + goto fail; + } + + if (!backtrace_get_view (state, descriptor, hdroffset, header.sizeofcmds, + error_callback, data, &cmds_view)) + goto fail; + cmds_view_valid = 1; + + memset (&dwarf_sections, 0, sizeof dwarf_sections); + have_dwarf = 0; + memset (&uuid, 0, sizeof uuid); + have_uuid = 0; + + cmdoffset = 0; + for (i = 0; i < header.ncmds; ++i) + { + const char *pcmd; + struct macho_load_command load_command; + + if (cmdoffset + sizeof load_command > header.sizeofcmds) + break; + + pcmd = (const char *) cmds_view.data + cmdoffset; + memcpy (&load_command, pcmd, sizeof load_command); + + switch (load_command.cmd) + { + case MACH_O_LC_SEGMENT: + { + struct macho_segment_command segcmd; + + memcpy (&segcmd, pcmd, sizeof segcmd); + if (memcmp (segcmd.segname, + "__DWARF\0\0\0\0\0\0\0\0\0", + MACH_O_NAMELEN) == 0) + { + if (!macho_add_dwarf_segment (state, descriptor, offset, + load_command.cmd, + pcmd + sizeof segcmd, + (load_command.cmdsize + - sizeof segcmd), + segcmd.nsects, error_callback, + data, &dwarf_sections)) + goto fail; + have_dwarf = 1; + } + } + break; + + case MACH_O_LC_SEGMENT_64: + { + struct macho_segment_64_command segcmd; + + memcpy (&segcmd, pcmd, sizeof segcmd); + if (memcmp (segcmd.segname, + "__DWARF\0\0\0\0\0\0\0\0\0", + MACH_O_NAMELEN) == 0) + { + if (!macho_add_dwarf_segment (state, descriptor, offset, + load_command.cmd, + pcmd + sizeof segcmd, + (load_command.cmdsize + - sizeof segcmd), + segcmd.nsects, error_callback, + data, &dwarf_sections)) + goto fail; + have_dwarf = 1; + } + } + break; + + case MACH_O_LC_SYMTAB: + if (!skip_symtab) + { + struct macho_symtab_command symcmd; + + memcpy (&symcmd, pcmd, sizeof symcmd); + if (!macho_add_symtab (state, descriptor, base_address, is_64, + offset + symcmd.symoff, symcmd.nsyms, + offset + symcmd.stroff, symcmd.strsize, + error_callback, data)) + goto fail; + + *found_sym = 1; + } + break; + + case MACH_O_LC_UUID: + { + struct macho_uuid_command uuidcmd; + + memcpy (&uuidcmd, pcmd, sizeof uuidcmd); + memcpy (&uuid[0], &uuidcmd.uuid[0], MACH_O_UUID_LEN); + have_uuid = 1; + } + break; + + default: + break; + } + + cmdoffset += load_command.cmdsize; + } + + if (!backtrace_close (descriptor, error_callback, data)) + goto fail; + descriptor = -1; + + backtrace_release_view (state, &cmds_view, error_callback, data); + cmds_view_valid = 0; + + if (match_uuid != NULL) + { + /* If we don't have a UUID, or it doesn't match, just ignore + this file. */ + if (!have_uuid + || memcmp (match_uuid, &uuid[0], MACH_O_UUID_LEN) != 0) + return 1; + } + + if (have_dwarf) + { + int is_big_endian; + + is_big_endian = 0; +#if defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) +#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ + is_big_endian = 1; +#endif +#endif + + if (!backtrace_dwarf_add (state, base_address, &dwarf_sections, + is_big_endian, NULL, error_callback, data, + fileline_fn, NULL)) + goto fail; + } + + if (!have_dwarf && have_uuid) + { + if (!macho_add_dsym (state, filename, base_address, &uuid[0], + error_callback, data, fileline_fn)) + goto fail; + } + + return 1; + + fail: + if (cmds_view_valid) + backtrace_release_view (state, &cmds_view, error_callback, data); + if (descriptor != -1) + backtrace_close (descriptor, error_callback, data); + return 0; +} + +#ifdef HAVE_MACH_O_DYLD_H + +/* Initialize the backtrace data we need from a Mach-O executable + using the dyld support functions. This closes descriptor. */ + +int +backtrace_initialize (struct backtrace_state *state, const char *filename, + int descriptor, backtrace_error_callback error_callback, + void *data, fileline *fileline_fn) +{ + uint32_t c; + uint32_t i; + int closed_descriptor; + int found_sym; + fileline macho_fileline_fn; + + closed_descriptor = 0; + found_sym = 0; + macho_fileline_fn = macho_nodebug; + + c = _dyld_image_count (); + for (i = 0; i < c; ++i) + { + uintptr_t base_address; + const char *name; + int d; + fileline mff; + int mfs; + + name = _dyld_get_image_name (i); + if (name == NULL) + continue; + + if (strcmp (name, filename) == 0 && !closed_descriptor) + { + d = descriptor; + closed_descriptor = 1; + } + else + { + int does_not_exist; + + d = backtrace_open (name, error_callback, data, &does_not_exist); + if (d < 0) + continue; + } + + base_address = _dyld_get_image_vmaddr_slide (i); + + mff = macho_nodebug; + if (!macho_add (state, name, d, 0, NULL, base_address, 0, + error_callback, data, &mff, &mfs)) + continue; + + if (mff != macho_nodebug) + macho_fileline_fn = mff; + if (mfs) + found_sym = 1; + } + + if (!closed_descriptor) + backtrace_close (descriptor, error_callback, data); + + if (!state->threaded) + { + if (found_sym) + state->syminfo_fn = macho_syminfo; + else if (state->syminfo_fn == NULL) + state->syminfo_fn = macho_nosyms; + } + else + { + if (found_sym) + backtrace_atomic_store_pointer (&state->syminfo_fn, &macho_syminfo); + else + (void) __sync_bool_compare_and_swap (&state->syminfo_fn, NULL, + macho_nosyms); + } + + if (!state->threaded) + *fileline_fn = state->fileline_fn; + else + *fileline_fn = backtrace_atomic_load_pointer (&state->fileline_fn); + + if (*fileline_fn == NULL || *fileline_fn == macho_nodebug) + *fileline_fn = macho_fileline_fn; + + return 1; +} + +#else /* !defined (HAVE_MACH_O_DYLD_H) */ + +/* Initialize the backtrace data we need from a Mach-O executable + without using the dyld support functions. This closes + descriptor. */ + +int +backtrace_initialize (struct backtrace_state *state, const char *filename, + int descriptor, backtrace_error_callback error_callback, + void *data, fileline *fileline_fn) +{ + fileline macho_fileline_fn; + int found_sym; + + macho_fileline_fn = macho_nodebug; + if (!macho_add (state, filename, descriptor, 0, NULL, 0, 0, + error_callback, data, &macho_fileline_fn, &found_sym)) + return 0; + + if (!state->threaded) + { + if (found_sym) + state->syminfo_fn = macho_syminfo; + else if (state->syminfo_fn == NULL) + state->syminfo_fn = macho_nosyms; + } + else + { + if (found_sym) + backtrace_atomic_store_pointer (&state->syminfo_fn, &macho_syminfo); + else + (void) __sync_bool_compare_and_swap (&state->syminfo_fn, NULL, + macho_nosyms); + } + + if (!state->threaded) + *fileline_fn = state->fileline_fn; + else + *fileline_fn = backtrace_atomic_load_pointer (&state->fileline_fn); + + if (*fileline_fn == NULL || *fileline_fn == macho_nodebug) + *fileline_fn = macho_fileline_fn; + + return 1; +} + +#endif /* !defined (HAVE_MACH_O_DYLD_H) */ + +} diff --git a/externals/tracy/public/libbacktrace/mmapio.cpp b/externals/tracy/public/libbacktrace/mmapio.cpp new file mode 100644 index 000000000..0e8f599bb --- /dev/null +++ b/externals/tracy/public/libbacktrace/mmapio.cpp @@ -0,0 +1,115 @@ +/* mmapio.c -- File views using mmap. + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#include "config.h" + +#include +#include +#include +#include + +#include "backtrace.hpp" +#include "internal.hpp" + +#ifndef HAVE_DECL_GETPAGESIZE +extern int getpagesize (void); +#endif + +#ifndef MAP_FAILED +#define MAP_FAILED ((void *)-1) +#endif + +namespace tracy +{ + +/* This file implements file views and memory allocation when mmap is + available. */ + +/* Create a view of SIZE bytes from DESCRIPTOR at OFFSET. */ + +int +backtrace_get_view (struct backtrace_state *state ATTRIBUTE_UNUSED, + int descriptor, off_t offset, uint64_t size, + backtrace_error_callback error_callback, + void *data, struct backtrace_view *view) +{ + size_t pagesize; + unsigned int inpage; + off_t pageoff; + void *map; + + if ((uint64_t) (size_t) size != size) + { + error_callback (data, "file size too large", 0); + return 0; + } + + pagesize = getpagesize (); + inpage = offset % pagesize; + pageoff = offset - inpage; + + size += inpage; + size = (size + (pagesize - 1)) & ~ (pagesize - 1); + + map = mmap (NULL, size, PROT_READ, MAP_PRIVATE, descriptor, pageoff); + if (map == MAP_FAILED) + { + error_callback (data, "mmap", errno); + return 0; + } + + view->data = (char *) map + inpage; + view->base = map; + view->len = size; + + return 1; +} + +/* Release a view read by backtrace_get_view. */ + +void +backtrace_release_view (struct backtrace_state *state ATTRIBUTE_UNUSED, + struct backtrace_view *view, + backtrace_error_callback error_callback, + void *data) +{ + union { + const void *cv; + void *v; + } cc; + + cc.cv = view->base; + if (munmap (cc.v, view->len) < 0) + error_callback (data, "munmap", errno); +} + +} diff --git a/externals/tracy/public/libbacktrace/posix.cpp b/externals/tracy/public/libbacktrace/posix.cpp new file mode 100644 index 000000000..8233a8ea3 --- /dev/null +++ b/externals/tracy/public/libbacktrace/posix.cpp @@ -0,0 +1,109 @@ +/* posix.c -- POSIX file I/O routines for the backtrace library. + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#include "config.h" + +#include +#include +#include +#include +#include + +#include "backtrace.hpp" +#include "internal.hpp" + +#ifndef O_BINARY +#define O_BINARY 0 +#endif + +#ifndef O_CLOEXEC +#define O_CLOEXEC 0 +#endif + +#ifndef FD_CLOEXEC +#define FD_CLOEXEC 1 +#endif + +namespace tracy +{ + +/* Open a file for reading. */ + +int +backtrace_open (const char *filename, backtrace_error_callback error_callback, + void *data, int *does_not_exist) +{ + int descriptor; + + if (does_not_exist != NULL) + *does_not_exist = 0; + + descriptor = open (filename, (int) (O_RDONLY | O_BINARY | O_CLOEXEC)); + if (descriptor < 0) + { + /* If DOES_NOT_EXIST is not NULL, then don't call ERROR_CALLBACK + if the file does not exist. We treat lacking permission to + open the file as the file not existing; this case arises when + running the libgo syscall package tests as root. */ + if (does_not_exist != NULL && (errno == ENOENT || errno == EACCES)) + *does_not_exist = 1; + else + error_callback (data, filename, errno); + return -1; + } + +#ifdef HAVE_FCNTL + /* Set FD_CLOEXEC just in case the kernel does not support + O_CLOEXEC. It doesn't matter if this fails for some reason. + FIXME: At some point it should be safe to only do this if + O_CLOEXEC == 0. */ + fcntl (descriptor, F_SETFD, FD_CLOEXEC); +#endif + + return descriptor; +} + +/* Close DESCRIPTOR. */ + +int +backtrace_close (int descriptor, backtrace_error_callback error_callback, + void *data) +{ + if (close (descriptor) < 0) + { + error_callback (data, "close", errno); + return 0; + } + return 1; +} + +} diff --git a/externals/tracy/public/libbacktrace/sort.cpp b/externals/tracy/public/libbacktrace/sort.cpp new file mode 100644 index 000000000..6daee0a64 --- /dev/null +++ b/externals/tracy/public/libbacktrace/sort.cpp @@ -0,0 +1,113 @@ +/* sort.c -- Sort without allocating memory + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#include "config.h" + +#include +#include + +#include "backtrace.hpp" +#include "internal.hpp" + +namespace tracy +{ + +/* The GNU glibc version of qsort allocates memory, which we must not + do if we are invoked by a signal handler. So provide our own + sort. */ + +static void +swap (char *a, char *b, size_t size) +{ + size_t i; + + for (i = 0; i < size; i++, a++, b++) + { + char t; + + t = *a; + *a = *b; + *b = t; + } +} + +void +backtrace_qsort (void *basearg, size_t count, size_t size, + int (*compar) (const void *, const void *)) +{ + char *base = (char *) basearg; + size_t i; + size_t mid; + + tail_recurse: + if (count < 2) + return; + + /* The symbol table and DWARF tables, which is all we use this + routine for, tend to be roughly sorted. Pick the middle element + in the array as our pivot point, so that we are more likely to + cut the array in half for each recursion step. */ + swap (base, base + (count / 2) * size, size); + + mid = 0; + for (i = 1; i < count; i++) + { + if ((*compar) (base, base + i * size) > 0) + { + ++mid; + if (i != mid) + swap (base + mid * size, base + i * size, size); + } + } + + if (mid > 0) + swap (base, base + mid * size, size); + + /* Recurse with the smaller array, loop with the larger one. That + ensures that our maximum stack depth is log count. */ + if (2 * mid < count) + { + backtrace_qsort (base, mid, size, compar); + base += (mid + 1) * size; + count -= mid + 1; + goto tail_recurse; + } + else + { + backtrace_qsort (base + (mid + 1) * size, count - (mid + 1), + size, compar); + count = mid; + goto tail_recurse; + } +} + +} diff --git a/externals/tracy/public/libbacktrace/state.cpp b/externals/tracy/public/libbacktrace/state.cpp new file mode 100644 index 000000000..ea3c137c5 --- /dev/null +++ b/externals/tracy/public/libbacktrace/state.cpp @@ -0,0 +1,76 @@ +/* state.c -- Create the backtrace state. + Copyright (C) 2012-2021 Free Software Foundation, Inc. + Written by Ian Lance Taylor, Google. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + + (1) Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + (2) Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in + the documentation and/or other materials provided with the + distribution. + + (3) The name of the author may not be used to + endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR +IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE +DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES +(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, +STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING +IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. */ + +#include "config.h" + +#include +#include + +#include "backtrace.hpp" +#include "internal.hpp" + +namespace tracy +{ + +/* Create the backtrace state. This will then be passed to all the + other routines. */ + +struct backtrace_state * +backtrace_create_state (const char *filename, int threaded, + backtrace_error_callback error_callback, + void *data) +{ + struct backtrace_state init_state; + struct backtrace_state *state; + +#ifndef HAVE_SYNC_FUNCTIONS + if (threaded) + { + error_callback (data, "backtrace library does not support threads", 0); + return NULL; + } +#endif + + memset (&init_state, 0, sizeof init_state); + init_state.filename = filename; + init_state.threaded = threaded; + + state = ((struct backtrace_state *) + backtrace_alloc (&init_state, sizeof *state, error_callback, data)); + if (state == NULL) + return NULL; + *state = init_state; + + return state; +} + +} diff --git a/externals/tracy/public/tracy/Tracy.hpp b/externals/tracy/public/tracy/Tracy.hpp new file mode 100644 index 000000000..c557a2c82 --- /dev/null +++ b/externals/tracy/public/tracy/Tracy.hpp @@ -0,0 +1,284 @@ +#ifndef __TRACY_HPP__ +#define __TRACY_HPP__ + +#include "../common/TracyColor.hpp" +#include "../common/TracySystem.hpp" + +#ifndef TracyFunction +# define TracyFunction __FUNCTION__ +#endif + +#ifndef TracyFile +# define TracyFile __FILE__ +#endif + +#ifndef TracyLine +# define TracyLine __LINE__ +#endif + +#ifndef TRACY_ENABLE + +#define ZoneNamed(x,y) +#define ZoneNamedN(x,y,z) +#define ZoneNamedC(x,y,z) +#define ZoneNamedNC(x,y,z,w) + +#define ZoneTransient(x,y) +#define ZoneTransientN(x,y,z) + +#define ZoneScoped +#define ZoneScopedN(x) +#define ZoneScopedC(x) +#define ZoneScopedNC(x,y) + +#define ZoneText(x,y) +#define ZoneTextV(x,y,z) +#define ZoneName(x,y) +#define ZoneNameV(x,y,z) +#define ZoneColor(x) +#define ZoneColorV(x,y) +#define ZoneValue(x) +#define ZoneValueV(x,y) +#define ZoneIsActive false +#define ZoneIsActiveV(x) false + +#define FrameMark +#define FrameMarkNamed(x) +#define FrameMarkStart(x) +#define FrameMarkEnd(x) + +#define FrameImage(x,y,z,w,a) + +#define TracyLockable( type, varname ) type varname +#define TracyLockableN( type, varname, desc ) type varname +#define TracySharedLockable( type, varname ) type varname +#define TracySharedLockableN( type, varname, desc ) type varname +#define LockableBase( type ) type +#define SharedLockableBase( type ) type +#define LockMark(x) (void)x +#define LockableName(x,y,z) + +#define TracyPlot(x,y) +#define TracyPlotConfig(x,y,z,w,a) + +#define TracyMessage(x,y) +#define TracyMessageL(x) +#define TracyMessageC(x,y,z) +#define TracyMessageLC(x,y) +#define TracyAppInfo(x,y) + +#define TracyAlloc(x,y) +#define TracyFree(x) +#define TracySecureAlloc(x,y) +#define TracySecureFree(x) + +#define TracyAllocN(x,y,z) +#define TracyFreeN(x,y) +#define TracySecureAllocN(x,y,z) +#define TracySecureFreeN(x,y) + +#define ZoneNamedS(x,y,z) +#define ZoneNamedNS(x,y,z,w) +#define ZoneNamedCS(x,y,z,w) +#define ZoneNamedNCS(x,y,z,w,a) + +#define ZoneTransientS(x,y,z) +#define ZoneTransientNS(x,y,z,w) + +#define ZoneScopedS(x) +#define ZoneScopedNS(x,y) +#define ZoneScopedCS(x,y) +#define ZoneScopedNCS(x,y,z) + +#define TracyAllocS(x,y,z) +#define TracyFreeS(x,y) +#define TracySecureAllocS(x,y,z) +#define TracySecureFreeS(x,y) + +#define TracyAllocNS(x,y,z,w) +#define TracyFreeNS(x,y,z) +#define TracySecureAllocNS(x,y,z,w) +#define TracySecureFreeNS(x,y,z) + +#define TracyMessageS(x,y,z) +#define TracyMessageLS(x,y) +#define TracyMessageCS(x,y,z,w) +#define TracyMessageLCS(x,y,z) + +#define TracySourceCallbackRegister(x,y) +#define TracyParameterRegister(x,y) +#define TracyParameterSetup(x,y,z,w) +#define TracyIsConnected false +#define TracyIsStarted false +#define TracySetProgramName(x) + +#define TracyFiberEnter(x) +#define TracyFiberLeave + +#else + +#include + +#include "../client/TracyLock.hpp" +#include "../client/TracyProfiler.hpp" +#include "../client/TracyScoped.hpp" + +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK +# define ZoneNamed( varname, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ) +# define ZoneNamedN( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ) +# define ZoneNamedC( varname, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ) +# define ZoneNamedNC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ) + +# define ZoneTransient( varname, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), nullptr, 0, TRACY_CALLSTACK, active ) +# define ZoneTransientN( varname, name, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), TRACY_CALLSTACK, active ) +#else +# define ZoneNamed( varname, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), active ) +# define ZoneNamedN( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), active ) +# define ZoneNamedC( varname, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), active ) +# define ZoneNamedNC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), active ) + +# define ZoneTransient( varname, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), nullptr, 0, active ) +# define ZoneTransientN( varname, name, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), active ) +#endif + +#define ZoneScoped ZoneNamed( ___tracy_scoped_zone, true ) +#define ZoneScopedN( name ) ZoneNamedN( ___tracy_scoped_zone, name, true ) +#define ZoneScopedC( color ) ZoneNamedC( ___tracy_scoped_zone, color, true ) +#define ZoneScopedNC( name, color ) ZoneNamedNC( ___tracy_scoped_zone, name, color, true ) + +#define ZoneText( txt, size ) ___tracy_scoped_zone.Text( txt, size ) +#define ZoneTextV( varname, txt, size ) varname.Text( txt, size ) +#define ZoneName( txt, size ) ___tracy_scoped_zone.Name( txt, size ) +#define ZoneNameV( varname, txt, size ) varname.Name( txt, size ) +#define ZoneColor( color ) ___tracy_scoped_zone.Color( color ) +#define ZoneColorV( varname, color ) varname.Color( color ) +#define ZoneValue( value ) ___tracy_scoped_zone.Value( value ) +#define ZoneValueV( varname, value ) varname.Value( value ) +#define ZoneIsActive ___tracy_scoped_zone.IsActive() +#define ZoneIsActiveV( varname ) varname.IsActive() + +#define FrameMark tracy::Profiler::SendFrameMark( nullptr ) +#define FrameMarkNamed( name ) tracy::Profiler::SendFrameMark( name ) +#define FrameMarkStart( name ) tracy::Profiler::SendFrameMark( name, tracy::QueueType::FrameMarkMsgStart ) +#define FrameMarkEnd( name ) tracy::Profiler::SendFrameMark( name, tracy::QueueType::FrameMarkMsgEnd ) + +#define FrameImage( image, width, height, offset, flip ) tracy::Profiler::SendFrameImage( image, width, height, offset, flip ) + +#define TracyLockable( type, varname ) tracy::Lockable varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, #type " " #varname, TracyFile, TracyLine, 0 }; return &srcloc; }() } +#define TracyLockableN( type, varname, desc ) tracy::Lockable varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, desc, TracyFile, TracyLine, 0 }; return &srcloc; }() } +#define TracySharedLockable( type, varname ) tracy::SharedLockable varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, #type " " #varname, TracyFile, TracyLine, 0 }; return &srcloc; }() } +#define TracySharedLockableN( type, varname, desc ) tracy::SharedLockable varname { [] () -> const tracy::SourceLocationData* { static constexpr tracy::SourceLocationData srcloc { nullptr, desc, TracyFile, TracyLine, 0 }; return &srcloc; }() } +#define LockableBase( type ) tracy::Lockable +#define SharedLockableBase( type ) tracy::SharedLockable +#define LockMark( varname ) static constexpr tracy::SourceLocationData __tracy_lock_location_##varname { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; varname.Mark( &__tracy_lock_location_##varname ) +#define LockableName( varname, txt, size ) varname.CustomName( txt, size ) + +#define TracyPlot( name, val ) tracy::Profiler::PlotData( name, val ) +#define TracyPlotConfig( name, type, step, fill, color ) tracy::Profiler::ConfigurePlot( name, type, step, fill, color ) + +#define TracyAppInfo( txt, size ) tracy::Profiler::MessageAppInfo( txt, size ) + +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK +# define TracyMessage( txt, size ) tracy::Profiler::Message( txt, size, TRACY_CALLSTACK ) +# define TracyMessageL( txt ) tracy::Profiler::Message( txt, TRACY_CALLSTACK ) +# define TracyMessageC( txt, size, color ) tracy::Profiler::MessageColor( txt, size, color, TRACY_CALLSTACK ) +# define TracyMessageLC( txt, color ) tracy::Profiler::MessageColor( txt, color, TRACY_CALLSTACK ) + +# define TracyAlloc( ptr, size ) tracy::Profiler::MemAllocCallstack( ptr, size, TRACY_CALLSTACK, false ) +# define TracyFree( ptr ) tracy::Profiler::MemFreeCallstack( ptr, TRACY_CALLSTACK, false ) +# define TracySecureAlloc( ptr, size ) tracy::Profiler::MemAllocCallstack( ptr, size, TRACY_CALLSTACK, true ) +# define TracySecureFree( ptr ) tracy::Profiler::MemFreeCallstack( ptr, TRACY_CALLSTACK, true ) + +# define TracyAllocN( ptr, size, name ) tracy::Profiler::MemAllocCallstackNamed( ptr, size, TRACY_CALLSTACK, false, name ) +# define TracyFreeN( ptr, name ) tracy::Profiler::MemFreeCallstackNamed( ptr, TRACY_CALLSTACK, false, name ) +# define TracySecureAllocN( ptr, size, name ) tracy::Profiler::MemAllocCallstackNamed( ptr, size, TRACY_CALLSTACK, true, name ) +# define TracySecureFreeN( ptr, name ) tracy::Profiler::MemFreeCallstackNamed( ptr, TRACY_CALLSTACK, true, name ) +#else +# define TracyMessage( txt, size ) tracy::Profiler::Message( txt, size, 0 ) +# define TracyMessageL( txt ) tracy::Profiler::Message( txt, 0 ) +# define TracyMessageC( txt, size, color ) tracy::Profiler::MessageColor( txt, size, color, 0 ) +# define TracyMessageLC( txt, color ) tracy::Profiler::MessageColor( txt, color, 0 ) + +# define TracyAlloc( ptr, size ) tracy::Profiler::MemAlloc( ptr, size, false ) +# define TracyFree( ptr ) tracy::Profiler::MemFree( ptr, false ) +# define TracySecureAlloc( ptr, size ) tracy::Profiler::MemAlloc( ptr, size, true ) +# define TracySecureFree( ptr ) tracy::Profiler::MemFree( ptr, true ) + +# define TracyAllocN( ptr, size, name ) tracy::Profiler::MemAllocNamed( ptr, size, false, name ) +# define TracyFreeN( ptr, name ) tracy::Profiler::MemFreeNamed( ptr, false, name ) +# define TracySecureAllocN( ptr, size, name ) tracy::Profiler::MemAllocNamed( ptr, size, true, name ) +# define TracySecureFreeN( ptr, name ) tracy::Profiler::MemFreeNamed( ptr, true, name ) +#endif + +#ifdef TRACY_HAS_CALLSTACK +# define ZoneNamedS( varname, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), depth, active ) +# define ZoneNamedNS( varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), depth, active ) +# define ZoneNamedCS( varname, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { nullptr, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), depth, active ) +# define ZoneNamedNCS( varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::ScopedZone varname( &TracyConcat(__tracy_source_location,TracyLine), depth, active ) + +# define ZoneTransientS( varname, depth, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), nullptr, 0, depth, active ) +# define ZoneTransientNS( varname, name, depth, active ) tracy::ScopedZone varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), depth, active ) + +# define ZoneScopedS( depth ) ZoneNamedS( ___tracy_scoped_zone, depth, true ) +# define ZoneScopedNS( name, depth ) ZoneNamedNS( ___tracy_scoped_zone, name, depth, true ) +# define ZoneScopedCS( color, depth ) ZoneNamedCS( ___tracy_scoped_zone, color, depth, true ) +# define ZoneScopedNCS( name, color, depth ) ZoneNamedNCS( ___tracy_scoped_zone, name, color, depth, true ) + +# define TracyAllocS( ptr, size, depth ) tracy::Profiler::MemAllocCallstack( ptr, size, depth, false ) +# define TracyFreeS( ptr, depth ) tracy::Profiler::MemFreeCallstack( ptr, depth, false ) +# define TracySecureAllocS( ptr, size, depth ) tracy::Profiler::MemAllocCallstack( ptr, size, depth, true ) +# define TracySecureFreeS( ptr, depth ) tracy::Profiler::MemFreeCallstack( ptr, depth, true ) + +# define TracyAllocNS( ptr, size, depth, name ) tracy::Profiler::MemAllocCallstackNamed( ptr, size, depth, false, name ) +# define TracyFreeNS( ptr, depth, name ) tracy::Profiler::MemFreeCallstackNamed( ptr, depth, false, name ) +# define TracySecureAllocNS( ptr, size, depth, name ) tracy::Profiler::MemAllocCallstackNamed( ptr, size, depth, true, name ) +# define TracySecureFreeNS( ptr, depth, name ) tracy::Profiler::MemFreeCallstackNamed( ptr, depth, true, name ) + +# define TracyMessageS( txt, size, depth ) tracy::Profiler::Message( txt, size, depth ) +# define TracyMessageLS( txt, depth ) tracy::Profiler::Message( txt, depth ) +# define TracyMessageCS( txt, size, color, depth ) tracy::Profiler::MessageColor( txt, size, color, depth ) +# define TracyMessageLCS( txt, color, depth ) tracy::Profiler::MessageColor( txt, color, depth ) +#else +# define ZoneNamedS( varname, depth, active ) ZoneNamed( varname, active ) +# define ZoneNamedNS( varname, name, depth, active ) ZoneNamedN( varname, name, active ) +# define ZoneNamedCS( varname, color, depth, active ) ZoneNamedC( varname, color, active ) +# define ZoneNamedNCS( varname, name, color, depth, active ) ZoneNamedNC( varname, name, color, active ) + +# define ZoneTransientS( varname, depth, active ) ZoneTransient( varname, active ) +# define ZoneTransientNS( varname, name, depth, active ) ZoneTransientN( varname, name, active ) + +# define ZoneScopedS( depth ) ZoneScoped +# define ZoneScopedNS( name, depth ) ZoneScopedN( name ) +# define ZoneScopedCS( color, depth ) ZoneScopedC( color ) +# define ZoneScopedNCS( name, color, depth ) ZoneScopedNC( name, color ) + +# define TracyAllocS( ptr, size, depth ) TracyAlloc( ptr, size ) +# define TracyFreeS( ptr, depth ) TracyFree( ptr ) +# define TracySecureAllocS( ptr, size, depth ) TracySecureAlloc( ptr, size ) +# define TracySecureFreeS( ptr, depth ) TracySecureFree( ptr ) + +# define TracyAllocNS( ptr, size, depth, name ) TracyAllocN( ptr, size, name ) +# define TracyFreeNS( ptr, depth, name ) TracyFreeN( ptr, name ) +# define TracySecureAllocNS( ptr, size, depth, name ) TracySecureAllocN( ptr, size, name ) +# define TracySecureFreeNS( ptr, depth, name ) TracySecureFreeN( ptr, name ) + +# define TracyMessageS( txt, size, depth ) TracyMessage( txt, size ) +# define TracyMessageLS( txt, depth ) TracyMessageL( txt ) +# define TracyMessageCS( txt, size, color, depth ) TracyMessageC( txt, size, color ) +# define TracyMessageLCS( txt, color, depth ) TracyMessageLC( txt, color ) +#endif + +#define TracySourceCallbackRegister( cb, data ) tracy::Profiler::SourceCallbackRegister( cb, data ) +#define TracyParameterRegister( cb, data ) tracy::Profiler::ParameterRegister( cb, data ) +#define TracyParameterSetup( idx, name, isBool, val ) tracy::Profiler::ParameterSetup( idx, name, isBool, val ) +#define TracyIsConnected tracy::GetProfiler().IsConnected() +#define TracySetProgramName( name ) tracy::GetProfiler().SetProgramName( name ); + +#ifdef TRACY_FIBERS +# define TracyFiberEnter( fiber ) tracy::Profiler::EnterFiber( fiber ) +# define TracyFiberLeave tracy::Profiler::LeaveFiber() +#endif + +#endif + +#endif diff --git a/externals/tracy/public/tracy/TracyC.h b/externals/tracy/public/tracy/TracyC.h new file mode 100644 index 000000000..6e62d56ab --- /dev/null +++ b/externals/tracy/public/tracy/TracyC.h @@ -0,0 +1,383 @@ +#ifndef __TRACYC_HPP__ +#define __TRACYC_HPP__ + +#include +#include + +#include "../client/TracyCallstack.h" +#include "../common/TracyApi.h" + +#ifdef __cplusplus +extern "C" { +#endif + +enum TracyPlotFormatEnum +{ + TracyPlotFormatNumber, + TracyPlotFormatMemory, + TracyPlotFormatPercentage, + TracyPlotFormatWatt +}; + +TRACY_API void ___tracy_set_thread_name( const char* name ); + +#define TracyCSetThreadName( name ) ___tracy_set_thread_name( name ); + +#ifndef TracyFunction +# define TracyFunction __FUNCTION__ +#endif + +#ifndef TracyFile +# define TracyFile __FILE__ +#endif + +#ifndef TracyLine +# define TracyLine __LINE__ +#endif + +#ifndef TRACY_ENABLE + +typedef const void* TracyCZoneCtx; + +#define TracyCZone(c,x) +#define TracyCZoneN(c,x,y) +#define TracyCZoneC(c,x,y) +#define TracyCZoneNC(c,x,y,z) +#define TracyCZoneEnd(c) +#define TracyCZoneText(c,x,y) +#define TracyCZoneName(c,x,y) +#define TracyCZoneColor(c,x) +#define TracyCZoneValue(c,x) + +#define TracyCAlloc(x,y) +#define TracyCFree(x) +#define TracyCSecureAlloc(x,y) +#define TracyCSecureFree(x) + +#define TracyCAllocN(x,y,z) +#define TracyCFreeN(x,y) +#define TracyCSecureAllocN(x,y,z) +#define TracyCSecureFreeN(x,y) + +#define TracyCFrameMark +#define TracyCFrameMarkNamed(x) +#define TracyCFrameMarkStart(x) +#define TracyCFrameMarkEnd(x) +#define TracyCFrameImage(x,y,z,w,a) + +#define TracyCPlot(x,y) +#define TracyCPlotF(x,y) +#define TracyCPlotI(x,y) +#define TracyCPlotConfig(x,y,z,w,a) + +#define TracyCMessage(x,y) +#define TracyCMessageL(x) +#define TracyCMessageC(x,y,z) +#define TracyCMessageLC(x,y) +#define TracyCAppInfo(x,y) + +#define TracyCZoneS(x,y,z) +#define TracyCZoneNS(x,y,z,w) +#define TracyCZoneCS(x,y,z,w) +#define TracyCZoneNCS(x,y,z,w,a) + +#define TracyCAllocS(x,y,z) +#define TracyCFreeS(x,y) +#define TracyCSecureAllocS(x,y,z) +#define TracyCSecureFreeS(x,y) + +#define TracyCAllocNS(x,y,z,w) +#define TracyCFreeNS(x,y,z) +#define TracyCSecureAllocNS(x,y,z,w) +#define TracyCSecureFreeNS(x,y,z) + +#define TracyCMessageS(x,y,z) +#define TracyCMessageLS(x,y) +#define TracyCMessageCS(x,y,z,w) +#define TracyCMessageLCS(x,y,z) + +#define TracyCIsConnected 0 +#define TracyCIsStarted 0 + +#ifdef TRACY_FIBERS +# define TracyCFiberEnter(fiber) +# define TracyCFiberLeave +#endif + +#else + +#ifndef TracyConcat +# define TracyConcat(x,y) TracyConcatIndirect(x,y) +#endif +#ifndef TracyConcatIndirect +# define TracyConcatIndirect(x,y) x##y +#endif + +struct ___tracy_source_location_data +{ + const char* name; + const char* function; + const char* file; + uint32_t line; + uint32_t color; +}; + +struct ___tracy_c_zone_context +{ + uint32_t id; + int active; +}; + +struct ___tracy_gpu_time_data +{ + int64_t gpuTime; + uint16_t queryId; + uint8_t context; +}; + +struct ___tracy_gpu_zone_begin_data { + uint64_t srcloc; + uint16_t queryId; + uint8_t context; +}; + +struct ___tracy_gpu_zone_begin_callstack_data { + uint64_t srcloc; + int depth; + uint16_t queryId; + uint8_t context; +}; + +struct ___tracy_gpu_zone_end_data { + uint16_t queryId; + uint8_t context; +}; + +struct ___tracy_gpu_new_context_data { + int64_t gpuTime; + float period; + uint8_t context; + uint8_t flags; + uint8_t type; +}; + +struct ___tracy_gpu_context_name_data { + uint8_t context; + const char* name; + uint16_t len; +}; + +struct ___tracy_gpu_calibration_data { + int64_t gpuTime; + int64_t cpuDelta; + uint8_t context; +}; + +struct ___tracy_gpu_time_sync_data { + int64_t gpuTime; + uint8_t context; +}; + +// Some containers don't support storing const types. +// This struct, as visible to user, is immutable, so treat it as if const was declared here. +typedef /*const*/ struct ___tracy_c_zone_context TracyCZoneCtx; + + +#ifdef TRACY_MANUAL_LIFETIME +TRACY_API void ___tracy_startup_profiler(void); +TRACY_API void ___tracy_shutdown_profiler(void); +TRACY_API int ___tracy_profiler_started(void); + +# define TracyCIsStarted ___tracy_profiler_started() +#else +# define TracyCIsStarted 1 +#endif + +TRACY_API uint64_t ___tracy_alloc_srcloc( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz ); +TRACY_API uint64_t ___tracy_alloc_srcloc_name( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz ); + +TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin( const struct ___tracy_source_location_data* srcloc, int active ); +TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin_callstack( const struct ___tracy_source_location_data* srcloc, int depth, int active ); +TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin_alloc( uint64_t srcloc, int active ); +TRACY_API TracyCZoneCtx ___tracy_emit_zone_begin_alloc_callstack( uint64_t srcloc, int depth, int active ); +TRACY_API void ___tracy_emit_zone_end( TracyCZoneCtx ctx ); +TRACY_API void ___tracy_emit_zone_text( TracyCZoneCtx ctx, const char* txt, size_t size ); +TRACY_API void ___tracy_emit_zone_name( TracyCZoneCtx ctx, const char* txt, size_t size ); +TRACY_API void ___tracy_emit_zone_color( TracyCZoneCtx ctx, uint32_t color ); +TRACY_API void ___tracy_emit_zone_value( TracyCZoneCtx ctx, uint64_t value ); + +TRACY_API void ___tracy_emit_gpu_zone_begin( const struct ___tracy_gpu_zone_begin_data ); +TRACY_API void ___tracy_emit_gpu_zone_begin_callstack( const struct ___tracy_gpu_zone_begin_callstack_data ); +TRACY_API void ___tracy_emit_gpu_zone_begin_alloc( const struct ___tracy_gpu_zone_begin_data ); +TRACY_API void ___tracy_emit_gpu_zone_begin_alloc_callstack( const struct ___tracy_gpu_zone_begin_callstack_data ); +TRACY_API void ___tracy_emit_gpu_zone_end( const struct ___tracy_gpu_zone_end_data data ); +TRACY_API void ___tracy_emit_gpu_time( const struct ___tracy_gpu_time_data ); +TRACY_API void ___tracy_emit_gpu_new_context( const struct ___tracy_gpu_new_context_data ); +TRACY_API void ___tracy_emit_gpu_context_name( const struct ___tracy_gpu_context_name_data ); +TRACY_API void ___tracy_emit_gpu_calibration( const struct ___tracy_gpu_calibration_data ); +TRACY_API void ___tracy_emit_gpu_time_sync( const struct ___tracy_gpu_time_sync_data ); + +TRACY_API void ___tracy_emit_gpu_zone_begin_serial( const struct ___tracy_gpu_zone_begin_data ); +TRACY_API void ___tracy_emit_gpu_zone_begin_callstack_serial( const struct ___tracy_gpu_zone_begin_callstack_data ); +TRACY_API void ___tracy_emit_gpu_zone_begin_alloc_serial( const struct ___tracy_gpu_zone_begin_data ); +TRACY_API void ___tracy_emit_gpu_zone_begin_alloc_callstack_serial( const struct ___tracy_gpu_zone_begin_callstack_data ); +TRACY_API void ___tracy_emit_gpu_zone_end_serial( const struct ___tracy_gpu_zone_end_data data ); +TRACY_API void ___tracy_emit_gpu_time_serial( const struct ___tracy_gpu_time_data ); +TRACY_API void ___tracy_emit_gpu_new_context_serial( const struct ___tracy_gpu_new_context_data ); +TRACY_API void ___tracy_emit_gpu_context_name_serial( const struct ___tracy_gpu_context_name_data ); +TRACY_API void ___tracy_emit_gpu_calibration_serial( const struct ___tracy_gpu_calibration_data ); +TRACY_API void ___tracy_emit_gpu_time_sync_serial( const struct ___tracy_gpu_time_sync_data ); + +TRACY_API int ___tracy_connected(void); + +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK +# define TracyCZone( ctx, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ); +# define TracyCZoneN( ctx, name, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ); +# define TracyCZoneC( ctx, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ); +# define TracyCZoneNC( ctx, name, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), TRACY_CALLSTACK, active ); +#else +# define TracyCZone( ctx, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,TracyLine), active ); +# define TracyCZoneN( ctx, name, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,TracyLine), active ); +# define TracyCZoneC( ctx, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,TracyLine), active ); +# define TracyCZoneNC( ctx, name, color, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin( &TracyConcat(__tracy_source_location,TracyLine), active ); +#endif + +#define TracyCZoneEnd( ctx ) ___tracy_emit_zone_end( ctx ); + +#define TracyCZoneText( ctx, txt, size ) ___tracy_emit_zone_text( ctx, txt, size ); +#define TracyCZoneName( ctx, txt, size ) ___tracy_emit_zone_name( ctx, txt, size ); +#define TracyCZoneColor( ctx, color ) ___tracy_emit_zone_color( ctx, color ); +#define TracyCZoneValue( ctx, value ) ___tracy_emit_zone_value( ctx, value ); + + +TRACY_API void ___tracy_emit_memory_alloc( const void* ptr, size_t size, int secure ); +TRACY_API void ___tracy_emit_memory_alloc_callstack( const void* ptr, size_t size, int depth, int secure ); +TRACY_API void ___tracy_emit_memory_free( const void* ptr, int secure ); +TRACY_API void ___tracy_emit_memory_free_callstack( const void* ptr, int depth, int secure ); +TRACY_API void ___tracy_emit_memory_alloc_named( const void* ptr, size_t size, int secure, const char* name ); +TRACY_API void ___tracy_emit_memory_alloc_callstack_named( const void* ptr, size_t size, int depth, int secure, const char* name ); +TRACY_API void ___tracy_emit_memory_free_named( const void* ptr, int secure, const char* name ); +TRACY_API void ___tracy_emit_memory_free_callstack_named( const void* ptr, int depth, int secure, const char* name ); + +TRACY_API void ___tracy_emit_message( const char* txt, size_t size, int callstack ); +TRACY_API void ___tracy_emit_messageL( const char* txt, int callstack ); +TRACY_API void ___tracy_emit_messageC( const char* txt, size_t size, uint32_t color, int callstack ); +TRACY_API void ___tracy_emit_messageLC( const char* txt, uint32_t color, int callstack ); + +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK +# define TracyCAlloc( ptr, size ) ___tracy_emit_memory_alloc_callstack( ptr, size, TRACY_CALLSTACK, 0 ) +# define TracyCFree( ptr ) ___tracy_emit_memory_free_callstack( ptr, TRACY_CALLSTACK, 0 ) +# define TracyCSecureAlloc( ptr, size ) ___tracy_emit_memory_alloc_callstack( ptr, size, TRACY_CALLSTACK, 1 ) +# define TracyCSecureFree( ptr ) ___tracy_emit_memory_free_callstack( ptr, TRACY_CALLSTACK, 1 ) + +# define TracyCAllocN( ptr, size, name ) ___tracy_emit_memory_alloc_callstack_named( ptr, size, TRACY_CALLSTACK, 0, name ) +# define TracyCFreeN( ptr, name ) ___tracy_emit_memory_free_callstack_named( ptr, TRACY_CALLSTACK, 0, name ) +# define TracyCSecureAllocN( ptr, size, name ) ___tracy_emit_memory_alloc_callstack_named( ptr, size, TRACY_CALLSTACK, 1, name ) +# define TracyCSecureFreeN( ptr, name ) ___tracy_emit_memory_free_callstack_named( ptr, TRACY_CALLSTACK, 1, name ) + +# define TracyCMessage( txt, size ) ___tracy_emit_message( txt, size, TRACY_CALLSTACK ); +# define TracyCMessageL( txt ) ___tracy_emit_messageL( txt, TRACY_CALLSTACK ); +# define TracyCMessageC( txt, size, color ) ___tracy_emit_messageC( txt, size, color, TRACY_CALLSTACK ); +# define TracyCMessageLC( txt, color ) ___tracy_emit_messageLC( txt, color, TRACY_CALLSTACK ); +#else +# define TracyCAlloc( ptr, size ) ___tracy_emit_memory_alloc( ptr, size, 0 ); +# define TracyCFree( ptr ) ___tracy_emit_memory_free( ptr, 0 ); +# define TracyCSecureAlloc( ptr, size ) ___tracy_emit_memory_alloc( ptr, size, 1 ); +# define TracyCSecureFree( ptr ) ___tracy_emit_memory_free( ptr, 1 ); + +# define TracyCAllocN( ptr, size, name ) ___tracy_emit_memory_alloc_named( ptr, size, 0, name ); +# define TracyCFreeN( ptr, name ) ___tracy_emit_memory_free_named( ptr, 0, name ); +# define TracyCSecureAllocN( ptr, size, name ) ___tracy_emit_memory_alloc_named( ptr, size, 1, name ); +# define TracyCSecureFreeN( ptr, name ) ___tracy_emit_memory_free_named( ptr, 1, name ); + +# define TracyCMessage( txt, size ) ___tracy_emit_message( txt, size, 0 ); +# define TracyCMessageL( txt ) ___tracy_emit_messageL( txt, 0 ); +# define TracyCMessageC( txt, size, color ) ___tracy_emit_messageC( txt, size, color, 0 ); +# define TracyCMessageLC( txt, color ) ___tracy_emit_messageLC( txt, color, 0 ); +#endif + + +TRACY_API void ___tracy_emit_frame_mark( const char* name ); +TRACY_API void ___tracy_emit_frame_mark_start( const char* name ); +TRACY_API void ___tracy_emit_frame_mark_end( const char* name ); +TRACY_API void ___tracy_emit_frame_image( const void* image, uint16_t w, uint16_t h, uint8_t offset, int flip ); + +#define TracyCFrameMark ___tracy_emit_frame_mark( 0 ); +#define TracyCFrameMarkNamed( name ) ___tracy_emit_frame_mark( name ); +#define TracyCFrameMarkStart( name ) ___tracy_emit_frame_mark_start( name ); +#define TracyCFrameMarkEnd( name ) ___tracy_emit_frame_mark_end( name ); +#define TracyCFrameImage( image, width, height, offset, flip ) ___tracy_emit_frame_image( image, width, height, offset, flip ); + + +TRACY_API void ___tracy_emit_plot( const char* name, double val ); +TRACY_API void ___tracy_emit_plot_float( const char* name, float val ); +TRACY_API void ___tracy_emit_plot_int( const char* name, int64_t val ); +TRACY_API void ___tracy_emit_plot_config( const char* name, int type, int step, int fill, uint32_t color ); +TRACY_API void ___tracy_emit_message_appinfo( const char* txt, size_t size ); + +#define TracyCPlot( name, val ) ___tracy_emit_plot( name, val ); +#define TracyCPlotF( name, val ) ___tracy_emit_plot_float( name, val ); +#define TracyCPlotI( name, val ) ___tracy_emit_plot_int( name, val ); +#define TracyCPlotConfig( name, type, step, fill, color ) ___tracy_emit_plot_config( name, type, step, fill, color ); +#define TracyCAppInfo( txt, size ) ___tracy_emit_message_appinfo( txt, size ); + + +#ifdef TRACY_HAS_CALLSTACK +# define TracyCZoneS( ctx, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), depth, active ); +# define TracyCZoneNS( ctx, name, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, 0 }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), depth, active ); +# define TracyCZoneCS( ctx, color, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { NULL, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), depth, active ); +# define TracyCZoneNCS( ctx, name, color, depth, active ) static const struct ___tracy_source_location_data TracyConcat(__tracy_source_location,TracyLine) = { name, __func__, TracyFile, (uint32_t)TracyLine, color }; TracyCZoneCtx ctx = ___tracy_emit_zone_begin_callstack( &TracyConcat(__tracy_source_location,TracyLine), depth, active ); + +# define TracyCAllocS( ptr, size, depth ) ___tracy_emit_memory_alloc_callstack( ptr, size, depth, 0 ) +# define TracyCFreeS( ptr, depth ) ___tracy_emit_memory_free_callstack( ptr, depth, 0 ) +# define TracyCSecureAllocS( ptr, size, depth ) ___tracy_emit_memory_alloc_callstack( ptr, size, depth, 1 ) +# define TracyCSecureFreeS( ptr, depth ) ___tracy_emit_memory_free_callstack( ptr, depth, 1 ) + +# define TracyCAllocNS( ptr, size, depth, name ) ___tracy_emit_memory_alloc_callstack_named( ptr, size, depth, 0, name ) +# define TracyCFreeNS( ptr, depth, name ) ___tracy_emit_memory_free_callstack_named( ptr, depth, 0, name ) +# define TracyCSecureAllocNS( ptr, size, depth, name ) ___tracy_emit_memory_alloc_callstack_named( ptr, size, depth, 1, name ) +# define TracyCSecureFreeNS( ptr, depth, name ) ___tracy_emit_memory_free_callstack_named( ptr, depth, 1, name ) + +# define TracyCMessageS( txt, size, depth ) ___tracy_emit_message( txt, size, depth ); +# define TracyCMessageLS( txt, depth ) ___tracy_emit_messageL( txt, depth ); +# define TracyCMessageCS( txt, size, color, depth ) ___tracy_emit_messageC( txt, size, color, depth ); +# define TracyCMessageLCS( txt, color, depth ) ___tracy_emit_messageLC( txt, color, depth ); +#else +# define TracyCZoneS( ctx, depth, active ) TracyCZone( ctx, active ) +# define TracyCZoneNS( ctx, name, depth, active ) TracyCZoneN( ctx, name, active ) +# define TracyCZoneCS( ctx, color, depth, active ) TracyCZoneC( ctx, color, active ) +# define TracyCZoneNCS( ctx, name, color, depth, active ) TracyCZoneNC( ctx, name, color, active ) + +# define TracyCAllocS( ptr, size, depth ) TracyCAlloc( ptr, size ) +# define TracyCFreeS( ptr, depth ) TracyCFree( ptr ) +# define TracyCSecureAllocS( ptr, size, depth ) TracyCSecureAlloc( ptr, size ) +# define TracyCSecureFreeS( ptr, depth ) TracyCSecureFree( ptr ) + +# define TracyCAllocNS( ptr, size, depth, name ) TracyCAllocN( ptr, size, name ) +# define TracyCFreeNS( ptr, depth, name ) TracyCFreeN( ptr, name ) +# define TracyCSecureAllocNS( ptr, size, depth, name ) TracyCSecureAllocN( ptr, size, name ) +# define TracyCSecureFreeNS( ptr, depth, name ) TracyCSecureFreeN( ptr, name ) + +# define TracyCMessageS( txt, size, depth ) TracyCMessage( txt, size ) +# define TracyCMessageLS( txt, depth ) TracyCMessageL( txt ) +# define TracyCMessageCS( txt, size, color, depth ) TracyCMessageC( txt, size, color ) +# define TracyCMessageLCS( txt, color, depth ) TracyCMessageLC( txt, color ) +#endif + +#define TracyCIsConnected ___tracy_connected() + +#ifdef TRACY_FIBERS +TRACY_API void ___tracy_fiber_enter( const char* fiber ); +TRACY_API void ___tracy_fiber_leave( void ); + +# define TracyCFiberEnter( fiber ) ___tracy_fiber_enter( fiber ); +# define TracyCFiberLeave ___tracy_fiber_leave(); +#endif + +#endif + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/externals/tracy/public/tracy/TracyD3D11.hpp b/externals/tracy/public/tracy/TracyD3D11.hpp new file mode 100644 index 000000000..8aebdb265 --- /dev/null +++ b/externals/tracy/public/tracy/TracyD3D11.hpp @@ -0,0 +1,446 @@ +#ifndef __TRACYD3D11_HPP__ +#define __TRACYD3D11_HPP__ + +#ifndef TRACY_ENABLE + +#define TracyD3D11Context(device,queue) nullptr +#define TracyD3D11Destroy(ctx) +#define TracyD3D11ContextName(ctx, name, size) + +#define TracyD3D11NewFrame(ctx) + +#define TracyD3D11Zone(ctx, name) +#define TracyD3D11ZoneC(ctx, name, color) +#define TracyD3D11NamedZone(ctx, varname, name, active) +#define TracyD3D11NamedZoneC(ctx, varname, name, color, active) +#define TracyD3D11ZoneTransient(ctx, varname, name, active) + +#define TracyD3D11ZoneS(ctx, name, depth) +#define TracyD3D11ZoneCS(ctx, name, color, depth) +#define TracyD3D11NamedZoneS(ctx, varname, name, depth, active) +#define TracyD3D11NamedZoneCS(ctx, varname, name, color, depth, active) +#define TracyD3D11ZoneTransientS(ctx, varname, name, depth, active) + +#define TracyD3D11Collect(ctx) + +namespace tracy +{ +class D3D11ZoneScope {}; +} + +using TracyD3D11Ctx = void*; + +#else + +#include +#include +#include + +#include "Tracy.hpp" +#include "../client/TracyProfiler.hpp" +#include "../client/TracyCallstack.hpp" +#include "../common/TracyYield.hpp" + +#include + +#define TracyD3D11Panic(msg, ...) do { assert(false && "TracyD3D11: " msg); TracyMessageLC("TracyD3D11: " msg, tracy::Color::Red4); __VA_ARGS__; } while(false); + +namespace tracy +{ + +class D3D11Ctx +{ + friend class D3D11ZoneScope; + + static constexpr uint32_t MaxQueries = 64 * 1024; + + enum CollectMode { POLL, BLOCK }; + +public: + D3D11Ctx( ID3D11Device* device, ID3D11DeviceContext* devicectx ) + { + // TODO: consider calling ID3D11Device::GetImmediateContext() instead of passing it as an argument + m_device = device; + device->AddRef(); + m_immediateDevCtx = devicectx; + devicectx->AddRef(); + + { + D3D11_QUERY_DESC desc = { }; + desc.Query = D3D11_QUERY_TIMESTAMP_DISJOINT; + if (FAILED(m_device->CreateQuery(&desc, &m_disjointQuery))) + { + TracyD3D11Panic("unable to create disjoint timestamp query.", return); + } + } + + for (ID3D11Query*& query : m_queries) + { + D3D11_QUERY_DESC desc = { }; + desc.Query = D3D11_QUERY_TIMESTAMP; + if (FAILED(m_device->CreateQuery(&desc, &query))) + { + TracyD3D11Panic("unable to create timestamp query.", return); + } + } + + // Calibrate CPU and GPU timestamps + int64_t tcpu = 0; + int64_t tgpu = 0; + for (int attempts = 0; attempts < 50; attempts++) + { + m_immediateDevCtx->Begin(m_disjointQuery); + m_immediateDevCtx->End(m_queries[0]); + m_immediateDevCtx->End(m_disjointQuery); + + int64_t tcpu0 = Profiler::GetTime(); + WaitForQuery(m_disjointQuery); + int64_t tcpu1 = Profiler::GetTime(); + + D3D11_QUERY_DATA_TIMESTAMP_DISJOINT disjoint = { }; + if (m_immediateDevCtx->GetData(m_disjointQuery, &disjoint, sizeof(disjoint), 0) != S_OK) + { + TracyMessageLC("TracyD3D11: unable to query GPU timestamp; retrying...", tracy::Color::Tomato); + continue; + } + + if (disjoint.Disjoint) + continue; + + UINT64 timestamp = 0; + if (m_immediateDevCtx->GetData(m_queries[0], ×tamp, sizeof(timestamp), 0) != S_OK) + continue; // this should never happen, since the enclosing disjoint query succeeded + + tcpu = tcpu0 + (tcpu1 - tcpu0) * 1 / 2; + tgpu = timestamp * (1000000000 / disjoint.Frequency); + break; + } + + // ready to roll + m_contextId = GetGpuCtxCounter().fetch_add(1); + m_immediateDevCtx->Begin(m_disjointQuery); + m_previousCheckpoint = m_nextCheckpoint = 0; + + auto* item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuNewContext ); + MemWrite( &item->gpuNewContext.cpuTime, tcpu ); + MemWrite( &item->gpuNewContext.gpuTime, tgpu ); + MemWrite( &item->gpuNewContext.thread, uint32_t(0) ); // #TODO: why not GetThreadHandle()? + MemWrite( &item->gpuNewContext.period, 1.0f ); + MemWrite( &item->gpuNewContext.context, m_contextId); + MemWrite( &item->gpuNewContext.flags, uint8_t(0) ); + MemWrite( &item->gpuNewContext.type, GpuContextType::Direct3D11 ); + +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + + Profiler::QueueSerialFinish(); + } + + ~D3D11Ctx() + { + // collect all pending timestamps before destroying everything + do + { + Collect(BLOCK); + } while (m_previousCheckpoint != m_queryCounter); + + for (ID3D11Query* query : m_queries) + { + query->Release(); + } + m_immediateDevCtx->End(m_disjointQuery); + m_disjointQuery->Release(); + m_immediateDevCtx->Release(); + m_device->Release(); + } + + void Name( const char* name, uint16_t len ) + { + auto ptr = (char*)tracy_malloc( len ); + memcpy( ptr, name, len ); + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuContextName ); + MemWrite( &item->gpuContextNameFat.context, m_contextId ); + MemWrite( &item->gpuContextNameFat.ptr, (uint64_t)ptr ); + MemWrite( &item->gpuContextNameFat.size, len ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + + void Collect(CollectMode mode = POLL) + { + ZoneScopedC( Color::Red4 ); + +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) + { + m_previousCheckpoint = m_nextCheckpoint = m_queryCounter; + return; + } +#endif + + if (m_previousCheckpoint == m_nextCheckpoint) + { + uintptr_t nextCheckpoint = m_queryCounter; + if (nextCheckpoint == m_nextCheckpoint) + { + return; + } + m_nextCheckpoint = nextCheckpoint; + m_immediateDevCtx->End(m_disjointQuery); + } + + if (mode == CollectMode::BLOCK) + { + WaitForQuery(m_disjointQuery); + } + + D3D11_QUERY_DATA_TIMESTAMP_DISJOINT disjoint = { }; + if (m_immediateDevCtx->GetData(m_disjointQuery, &disjoint, sizeof(disjoint), D3D11_ASYNC_GETDATA_DONOTFLUSH) != S_OK) + { + return; + } + + if (disjoint.Disjoint == TRUE) + { + m_previousCheckpoint = m_nextCheckpoint; + TracyD3D11Panic("disjoint timestamps detected; dropping."); + return; + } + + auto begin = m_previousCheckpoint; + auto end = m_nextCheckpoint; + for (auto i = begin; i != end; ++i) + { + uint32_t k = RingIndex(i); + UINT64 timestamp = 0; + if (m_immediateDevCtx->GetData(m_queries[k], ×tamp, sizeof(timestamp), 0) != S_OK) + { + TracyD3D11Panic("timestamp expected to be ready, but it was not!"); + break; + } + timestamp *= (1000000000ull / disjoint.Frequency); + auto* item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuTime); + MemWrite(&item->gpuTime.gpuTime, static_cast(timestamp)); + MemWrite(&item->gpuTime.queryId, static_cast(k)); + MemWrite(&item->gpuTime.context, m_contextId); + Profiler::QueueSerialFinish(); + } + + // disjoint timestamp queries should only be invoked once per frame or less + // https://learn.microsoft.com/en-us/windows/win32/api/d3d11/ne-d3d11-d3d11_query + m_immediateDevCtx->Begin(m_disjointQuery); + m_previousCheckpoint = m_nextCheckpoint; + } + +private: + tracy_force_inline uint32_t RingIndex(uintptr_t index) + { + index %= MaxQueries; + return static_cast(index); + } + + tracy_force_inline uint32_t RingCount(uintptr_t begin, uintptr_t end) + { + // wrap-around safe: all unsigned + uintptr_t count = end - begin; + return static_cast(count); + } + + tracy_force_inline uint32_t NextQueryId() + { + auto id = m_queryCounter++; + if (RingCount(m_previousCheckpoint, id) >= MaxQueries) + { + TracyD3D11Panic("too many pending timestamp queries."); + // #TODO: return some sentinel value; ideally a "hidden" query index + } + return RingIndex(id); + } + + tracy_force_inline ID3D11Query* GetQueryObjectFromId(uint32_t id) + { + return m_queries[id]; + } + + tracy_force_inline void WaitForQuery(ID3D11Query* query) + { + m_immediateDevCtx->Flush(); + while (m_immediateDevCtx->GetData(query, nullptr, 0, 0) != S_OK) + YieldThread(); // busy-wait :-( attempt to reduce power usage with _mm_pause() & friends... + } + + tracy_force_inline uint8_t GetContextId() const + { + return m_contextId; + } + + ID3D11Device* m_device = nullptr; + ID3D11DeviceContext* m_immediateDevCtx = nullptr; + + ID3D11Query* m_queries[MaxQueries]; + ID3D11Query* m_disjointQuery = nullptr; + + uint8_t m_contextId = 255; // NOTE: apparently, 255 means invalid id; is this documented anywhere? + + uintptr_t m_queryCounter = 0; + + uintptr_t m_previousCheckpoint = 0; + uintptr_t m_nextCheckpoint = 0; +}; + +class D3D11ZoneScope +{ +public: + tracy_force_inline D3D11ZoneScope( D3D11Ctx* ctx, const SourceLocationData* srcloc, bool active ) + : D3D11ZoneScope(ctx, active) + { + if( !m_active ) return; + + auto* item = Profiler::QueueSerial(); + WriteQueueItem(item, QueueType::GpuZoneBeginSerial, reinterpret_cast(srcloc)); + } + + tracy_force_inline D3D11ZoneScope( D3D11Ctx* ctx, const SourceLocationData* srcloc, int depth, bool active ) + : D3D11ZoneScope(ctx, active) + { + if( !m_active ) return; + + auto* item = Profiler::QueueSerialCallstack(Callstack(depth)); + WriteQueueItem(item, QueueType::GpuZoneBeginCallstackSerial, reinterpret_cast(srcloc)); + } + + tracy_force_inline D3D11ZoneScope(D3D11Ctx* ctx, uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, bool active) + : D3D11ZoneScope(ctx, active) + { + if( !m_active ) return; + + const auto sourceLocation = Profiler::AllocSourceLocation(line, source, sourceSz, function, functionSz, name, nameSz); + + auto* item = Profiler::QueueSerial(); + WriteQueueItem(item, QueueType::GpuZoneBeginAllocSrcLocSerial, sourceLocation); + } + + tracy_force_inline D3D11ZoneScope(D3D11Ctx* ctx, uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, int depth, bool active) + : D3D11ZoneScope(ctx, active) + { + if( !m_active ) return; + + const auto sourceLocation = Profiler::AllocSourceLocation(line, source, sourceSz, function, functionSz, name, nameSz); + + auto* item = Profiler::QueueSerialCallstack(Callstack(depth)); + WriteQueueItem(item, QueueType::GpuZoneBeginAllocSrcLocCallstackSerial, sourceLocation); + } + + tracy_force_inline ~D3D11ZoneScope() + { + if( !m_active ) return; + + const auto queryId = m_ctx->NextQueryId(); + m_ctx->m_immediateDevCtx->End(m_ctx->GetQueryObjectFromId(queryId)); + + auto* item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuZoneEndSerial ); + MemWrite( &item->gpuZoneEnd.cpuTime, Profiler::GetTime() ); + MemWrite( &item->gpuZoneEnd.thread, GetThreadHandle() ); + MemWrite( &item->gpuZoneEnd.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneEnd.context, m_ctx->GetContextId() ); + Profiler::QueueSerialFinish(); + } + +private: + tracy_force_inline D3D11ZoneScope( D3D11Ctx* ctx, bool active ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( active ) +#endif + { + if( !m_active ) return; + m_ctx = ctx; + } + + void WriteQueueItem(tracy::QueueItem* item, tracy::QueueType queueItemType, uint64_t sourceLocation) + { + const auto queryId = m_ctx->NextQueryId(); + m_ctx->m_immediateDevCtx->End(m_ctx->GetQueryObjectFromId(queryId)); + + MemWrite( &item->hdr.type, queueItemType); + MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() ); + MemWrite( &item->gpuZoneBegin.srcloc, sourceLocation ); + MemWrite( &item->gpuZoneBegin.thread, GetThreadHandle() ); + MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneBegin.context, m_ctx->GetContextId() ); + Profiler::QueueSerialFinish(); + } + + const bool m_active; + + D3D11Ctx* m_ctx; +}; + +static inline D3D11Ctx* CreateD3D11Context( ID3D11Device* device, ID3D11DeviceContext* devicectx ) +{ + auto ctx = (D3D11Ctx*)tracy_malloc( sizeof( D3D11Ctx ) ); + new(ctx) D3D11Ctx( device, devicectx ); + return ctx; +} + +static inline void DestroyD3D11Context( D3D11Ctx* ctx ) +{ + ctx->~D3D11Ctx(); + tracy_free( ctx ); +} +} + +#undef TracyD3D11Panic + +using TracyD3D11Ctx = tracy::D3D11Ctx*; + +#define TracyD3D11Context( device, devicectx ) tracy::CreateD3D11Context( device, devicectx ); +#define TracyD3D11Destroy(ctx) tracy::DestroyD3D11Context(ctx); +#define TracyD3D11ContextName(ctx, name, size) ctx->Name(name, size); + +#define TracyD3D11UnnamedZone ___tracy_gpu_d3d11_zone +#define TracyD3D11SrcLocSymbol TracyConcat(__tracy_gpu_d3d11_source_location,TracyLine) +#define TracyD3D11SrcLocObject(name, color) static constexpr tracy::SourceLocationData TracyD3D11SrcLocSymbol { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; + +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK +# define TracyD3D11Zone( ctx, name ) TracyD3D11NamedZoneS( ctx, TracyD3D11UnnamedZone, name, TRACY_CALLSTACK, true ) +# define TracyD3D11ZoneC( ctx, name, color ) TracyD3D11NamedZoneCS( ctx, TracyD3D11UnnamedZone, name, color, TRACY_CALLSTACK, true ) +# define TracyD3D11NamedZone( ctx, varname, name, active ) TracyD3D11SrcLocObject(name, 0); tracy::D3D11ZoneScope varname( ctx, &TracyD3D11SrcLocSymbol, TRACY_CALLSTACK, active ); +# define TracyD3D11NamedZoneC( ctx, varname, name, color, active ) TracyD3D11SrcLocObject(name, color); tracy::D3D11ZoneScope varname( ctx, &TracyD3D11SrcLocSymbol, TRACY_CALLSTACK, active ); +# define TracyD3D11ZoneTransient(ctx, varname, name, active) TracyD3D11ZoneTransientS(ctx, varname, cmdList, name, TRACY_CALLSTACK, active) +#else +# define TracyD3D11Zone( ctx, name ) TracyD3D11NamedZone( ctx, TracyD3D11UnnamedZone, name, true ) +# define TracyD3D11ZoneC( ctx, name, color ) TracyD3D11NamedZoneC( ctx, TracyD3D11UnnamedZone, name, color, true ) +# define TracyD3D11NamedZone( ctx, varname, name, active ) TracyD3D11SrcLocObject(name, 0); tracy::D3D11ZoneScope varname( ctx, &TracyD3D11SrcLocSymbol, active ); +# define TracyD3D11NamedZoneC( ctx, varname, name, color, active ) TracyD3D11SrcLocObject(name, color); tracy::D3D11ZoneScope varname( ctx, &TracyD3D11SrcLocSymbol, active ); +# define TracyD3D11ZoneTransient(ctx, varname, name, active) tracy::D3D11ZoneScope varname{ ctx, TracyLine, TracyFile, strlen(TracyFile), TracyFunction, strlen(TracyFunction), name, strlen(name), active }; +#endif + +#ifdef TRACY_HAS_CALLSTACK +# define TracyD3D11ZoneS( ctx, name, depth ) TracyD3D11NamedZoneS( ctx, TracyD3D11UnnamedZone, name, depth, true ) +# define TracyD3D11ZoneCS( ctx, name, color, depth ) TracyD3D11NamedZoneCS( ctx, TracyD3D11UnnamedZone, name, color, depth, true ) +# define TracyD3D11NamedZoneS( ctx, varname, name, depth, active ) TracyD3D11SrcLocObject(name, 0); tracy::D3D11ZoneScope varname( ctx, &TracyD3D11SrcLocSymbol, depth, active ); +# define TracyD3D11NamedZoneCS( ctx, varname, name, color, depth, active ) TracyD3D11SrcLocObject(name, color); tracy::D3D11ZoneScope varname( ctx, &TracyD3D11SrcLocSymbol, depth, active ); +# define TracyD3D11ZoneTransientS(ctx, varname, name, depth, active) tracy::D3D11ZoneScope varname{ ctx, TracyLine, TracyFile, strlen(TracyFile), TracyFunction, strlen(TracyFunction), name, strlen(name), depth, active }; +#else +# define TracyD3D11ZoneS( ctx, name, depth, active ) TracyD3D11Zone( ctx, name ) +# define TracyD3D11ZoneCS( ctx, name, color, depth, active ) TracyD3D11ZoneC( name, color ) +# define TracyD3D11NamedZoneS( ctx, varname, name, depth, active ) TracyD3D11NamedZone( ctx, varname, name, active ) +# define TracyD3D11NamedZoneCS( ctx, varname, name, color, depth, active ) TracyD3D11NamedZoneC( ctx, varname, name, color, active ) +# define TracyD3D11ZoneTransientS(ctx, varname, name, depth, active) TracyD3D11ZoneTransient(ctx, varname, name, active) +#endif + +#define TracyD3D11Collect( ctx ) ctx->Collect(); + +#endif + +#endif diff --git a/externals/tracy/public/tracy/TracyD3D12.hpp b/externals/tracy/public/tracy/TracyD3D12.hpp new file mode 100644 index 000000000..41567937e --- /dev/null +++ b/externals/tracy/public/tracy/TracyD3D12.hpp @@ -0,0 +1,500 @@ +#ifndef __TRACYD3D12_HPP__ +#define __TRACYD3D12_HPP__ + +#ifndef TRACY_ENABLE + +#define TracyD3D12Context(device, queue) nullptr +#define TracyD3D12Destroy(ctx) +#define TracyD3D12ContextName(ctx, name, size) + +#define TracyD3D12NewFrame(ctx) + +#define TracyD3D12Zone(ctx, cmdList, name) +#define TracyD3D12ZoneC(ctx, cmdList, name, color) +#define TracyD3D12NamedZone(ctx, varname, cmdList, name, active) +#define TracyD3D12NamedZoneC(ctx, varname, cmdList, name, color, active) +#define TracyD3D12ZoneTransient(ctx, varname, cmdList, name, active) + +#define TracyD3D12ZoneS(ctx, cmdList, name, depth) +#define TracyD3D12ZoneCS(ctx, cmdList, name, color, depth) +#define TracyD3D12NamedZoneS(ctx, varname, cmdList, name, depth, active) +#define TracyD3D12NamedZoneCS(ctx, varname, cmdList, name, color, depth, active) +#define TracyD3D12ZoneTransientS(ctx, varname, cmdList, name, depth, active) + +#define TracyD3D12Collect(ctx) + +namespace tracy +{ + class D3D12ZoneScope {}; +} + +using TracyD3D12Ctx = void*; + +#else + +#include "Tracy.hpp" +#include "../client/TracyProfiler.hpp" +#include "../client/TracyCallstack.hpp" + +#include +#include +#include +#include +#include + +#define TracyD3D12Panic(msg, ...) do { assert(false && "TracyD3D12: " msg); TracyMessageLC("TracyD3D12: " msg, tracy::Color::Red4); __VA_ARGS__; } while(false); + +namespace tracy +{ + + struct D3D12QueryPayload + { + uint32_t m_queryIdStart = 0; + uint32_t m_queryCount = 0; + }; + + // Command queue context. + class D3D12QueueCtx + { + friend class D3D12ZoneScope; + + ID3D12Device* m_device = nullptr; + ID3D12CommandQueue* m_queue = nullptr; + uint8_t m_contextId = 255; // TODO: apparently, 255 means "invalid id"; is this documented somewhere? + ID3D12QueryHeap* m_queryHeap = nullptr; + ID3D12Resource* m_readbackBuffer = nullptr; + + // In-progress payload. + uint32_t m_queryLimit = 0; + std::atomic m_queryCounter = 0; + uint32_t m_previousQueryCounter = 0; + + uint32_t m_activePayload = 0; + ID3D12Fence* m_payloadFence = nullptr; + std::queue m_payloadQueue; + + UINT64 m_prevCalibrationTicksCPU = 0; + + void RecalibrateClocks() + { + UINT64 cpuTimestamp; + UINT64 gpuTimestamp; + if (FAILED(m_queue->GetClockCalibration(&gpuTimestamp, &cpuTimestamp))) + { + TracyD3D12Panic("failed to obtain queue clock calibration counters.", return); + } + + int64_t cpuDeltaTicks = cpuTimestamp - m_prevCalibrationTicksCPU; + if (cpuDeltaTicks > 0) + { + static const int64_t nanosecodsPerTick = int64_t(1000000000) / GetFrequencyQpc(); + int64_t cpuDeltaNS = cpuDeltaTicks * nanosecodsPerTick; + // Save the device cpu timestamp, not the Tracy profiler timestamp: + m_prevCalibrationTicksCPU = cpuTimestamp; + + cpuTimestamp = Profiler::GetTime(); + + auto* item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuCalibration); + MemWrite(&item->gpuCalibration.gpuTime, gpuTimestamp); + MemWrite(&item->gpuCalibration.cpuTime, cpuTimestamp); + MemWrite(&item->gpuCalibration.cpuDelta, cpuDeltaNS); + MemWrite(&item->gpuCalibration.context, GetId()); + SubmitQueueItem(item); + } + } + + tracy_force_inline void SubmitQueueItem(tracy::QueueItem* item) + { +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem(*item); +#endif + Profiler::QueueSerialFinish(); + } + + public: + D3D12QueueCtx(ID3D12Device* device, ID3D12CommandQueue* queue) + : m_device(device) + , m_queue(queue) + { + // Verify we support timestamp queries on this queue. + + if (queue->GetDesc().Type == D3D12_COMMAND_LIST_TYPE_COPY) + { + D3D12_FEATURE_DATA_D3D12_OPTIONS3 featureData{}; + + HRESULT hr = device->CheckFeatureSupport(D3D12_FEATURE_D3D12_OPTIONS3, &featureData, sizeof(featureData)); + if (FAILED(hr) || (featureData.CopyQueueTimestampQueriesSupported == FALSE)) + { + TracyD3D12Panic("Platform does not support profiling of copy queues.", return); + } + } + + static constexpr uint32_t MaxQueries = 64 * 1024; // Must be even, because queries are (begin, end) pairs + m_queryLimit = MaxQueries; + + D3D12_QUERY_HEAP_DESC heapDesc{}; + heapDesc.Type = queue->GetDesc().Type == D3D12_COMMAND_LIST_TYPE_COPY ? D3D12_QUERY_HEAP_TYPE_COPY_QUEUE_TIMESTAMP : D3D12_QUERY_HEAP_TYPE_TIMESTAMP; + heapDesc.Count = m_queryLimit; + heapDesc.NodeMask = 0; // #TODO: Support multiple adapters. + + while (FAILED(device->CreateQueryHeap(&heapDesc, IID_PPV_ARGS(&m_queryHeap)))) + { + m_queryLimit /= 2; + heapDesc.Count = m_queryLimit; + } + + // Create a readback buffer, which will be used as a destination for the query data. + + D3D12_RESOURCE_DESC readbackBufferDesc{}; + readbackBufferDesc.Alignment = 0; + readbackBufferDesc.Dimension = D3D12_RESOURCE_DIMENSION_BUFFER; + readbackBufferDesc.Width = m_queryLimit * sizeof(uint64_t); + readbackBufferDesc.Height = 1; + readbackBufferDesc.DepthOrArraySize = 1; + readbackBufferDesc.Format = DXGI_FORMAT_UNKNOWN; + readbackBufferDesc.Layout = D3D12_TEXTURE_LAYOUT_ROW_MAJOR; // Buffers are always row major. + readbackBufferDesc.MipLevels = 1; + readbackBufferDesc.SampleDesc.Count = 1; + readbackBufferDesc.SampleDesc.Quality = 0; + readbackBufferDesc.Flags = D3D12_RESOURCE_FLAG_NONE; + + D3D12_HEAP_PROPERTIES readbackHeapProps{}; + readbackHeapProps.Type = D3D12_HEAP_TYPE_READBACK; + readbackHeapProps.CPUPageProperty = D3D12_CPU_PAGE_PROPERTY_UNKNOWN; + readbackHeapProps.MemoryPoolPreference = D3D12_MEMORY_POOL_UNKNOWN; + readbackHeapProps.CreationNodeMask = 0; + readbackHeapProps.VisibleNodeMask = 0; // #TODO: Support multiple adapters. + + if (FAILED(device->CreateCommittedResource(&readbackHeapProps, D3D12_HEAP_FLAG_NONE, &readbackBufferDesc, D3D12_RESOURCE_STATE_COPY_DEST, nullptr, IID_PPV_ARGS(&m_readbackBuffer)))) + { + TracyD3D12Panic("Failed to create query readback buffer.", return); + } + + if (FAILED(device->CreateFence(0, D3D12_FENCE_FLAG_NONE, IID_PPV_ARGS(&m_payloadFence)))) + { + TracyD3D12Panic("Failed to create payload fence.", return); + } + + float period = [queue]() + { + uint64_t timestampFrequency; + if (FAILED(queue->GetTimestampFrequency(×tampFrequency))) + { + return 0.0f; + } + return static_cast( 1E+09 / static_cast(timestampFrequency) ); + }(); + + if (period == 0.0f) + { + TracyD3D12Panic("Failed to get timestamp frequency.", return); + } + + uint64_t cpuTimestamp; + uint64_t gpuTimestamp; + if (FAILED(queue->GetClockCalibration(&gpuTimestamp, &cpuTimestamp))) + { + TracyD3D12Panic("Failed to get queue clock calibration.", return); + } + + // Save the device cpu timestamp, not the profiler's timestamp. + m_prevCalibrationTicksCPU = cpuTimestamp; + + cpuTimestamp = Profiler::GetTime(); + + // all checked: ready to roll + m_contextId = GetGpuCtxCounter().fetch_add(1); + + auto* item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuNewContext); + MemWrite(&item->gpuNewContext.cpuTime, cpuTimestamp); + MemWrite(&item->gpuNewContext.gpuTime, gpuTimestamp); + MemWrite(&item->gpuNewContext.thread, decltype(item->gpuNewContext.thread)(0)); // #TODO: why 0 instead of GetThreadHandle()? + MemWrite(&item->gpuNewContext.period, period); + MemWrite(&item->gpuNewContext.context, GetId()); + MemWrite(&item->gpuNewContext.flags, GpuContextCalibration); + MemWrite(&item->gpuNewContext.type, GpuContextType::Direct3D12); + SubmitQueueItem(item); + } + + ~D3D12QueueCtx() + { + ZoneScopedC(Color::Red4); + // collect all pending timestamps + while (m_payloadFence->GetCompletedValue() != m_activePayload) + /* busy-wait ... */; + Collect(); + m_payloadFence->Release(); + m_readbackBuffer->Release(); + m_queryHeap->Release(); + } + + + void NewFrame() + { + uint32_t queryCounter = m_queryCounter.exchange(0); + m_payloadQueue.emplace(D3D12QueryPayload{ m_previousQueryCounter, queryCounter }); + m_previousQueryCounter += queryCounter; + + if (m_previousQueryCounter >= m_queryLimit) + { + m_previousQueryCounter -= m_queryLimit; + } + + m_queue->Signal(m_payloadFence, ++m_activePayload); + } + + void Name( const char* name, uint16_t len ) + { + auto ptr = (char*)tracy_malloc( len ); + memcpy( ptr, name, len ); + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuContextName ); + MemWrite( &item->gpuContextNameFat.context, GetId()); + MemWrite( &item->gpuContextNameFat.ptr, (uint64_t)ptr ); + MemWrite( &item->gpuContextNameFat.size, len ); + SubmitQueueItem(item); + } + + void Collect() + { + ZoneScopedC(Color::Red4); + +#ifdef TRACY_ON_DEMAND + if (!GetProfiler().IsConnected()) + { + m_queryCounter = 0; + + return; + } +#endif + + // Find out what payloads are available. + const auto newestReadyPayload = m_payloadFence->GetCompletedValue(); + const auto payloadCount = m_payloadQueue.size() - (m_activePayload - newestReadyPayload); + + if (!payloadCount) + { + return; // No payloads are available yet, exit out. + } + + D3D12_RANGE mapRange{ 0, m_queryLimit * sizeof(uint64_t) }; + + // Map the readback buffer so we can fetch the query data from the GPU. + void* readbackBufferMapping = nullptr; + + if (FAILED(m_readbackBuffer->Map(0, &mapRange, &readbackBufferMapping))) + { + TracyD3D12Panic("Failed to map readback buffer.", return); + } + + auto* timestampData = static_cast(readbackBufferMapping); + + for (uint32_t i = 0; i < payloadCount; ++i) + { + const auto& payload = m_payloadQueue.front(); + + for (uint32_t j = 0; j < payload.m_queryCount; ++j) + { + const auto counter = (payload.m_queryIdStart + j) % m_queryLimit; + const auto timestamp = timestampData[counter]; + const auto queryId = counter; + + auto* item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuTime); + MemWrite(&item->gpuTime.gpuTime, timestamp); + MemWrite(&item->gpuTime.queryId, static_cast(queryId)); + MemWrite(&item->gpuTime.context, GetId()); + + Profiler::QueueSerialFinish(); + } + + m_payloadQueue.pop(); + } + + m_readbackBuffer->Unmap(0, nullptr); + + // Recalibrate to account for drift. + RecalibrateClocks(); + } + + private: + tracy_force_inline uint32_t NextQueryId() + { + uint32_t queryCounter = m_queryCounter.fetch_add(2); + if (queryCounter >= m_queryLimit) + { + TracyD3D12Panic("Submitted too many GPU queries! Consider increasing MaxQueries."); + // #TODO: consider returning an invalid id or sentinel value here + } + + const uint32_t id = (m_previousQueryCounter + queryCounter) % m_queryLimit; + + return id; + } + + tracy_force_inline uint8_t GetId() const + { + return m_contextId; + } + }; + + class D3D12ZoneScope + { + const bool m_active; + D3D12QueueCtx* m_ctx = nullptr; + ID3D12GraphicsCommandList* m_cmdList = nullptr; + uint32_t m_queryId = 0; // Used for tracking in nested zones. + + tracy_force_inline void WriteQueueItem(QueueItem* item, QueueType type, uint64_t srcLocation) + { + MemWrite(&item->hdr.type, type); + MemWrite(&item->gpuZoneBegin.cpuTime, Profiler::GetTime()); + MemWrite(&item->gpuZoneBegin.srcloc, srcLocation); + MemWrite(&item->gpuZoneBegin.thread, GetThreadHandle()); + MemWrite(&item->gpuZoneBegin.queryId, static_cast(m_queryId)); + MemWrite(&item->gpuZoneBegin.context, m_ctx->GetId()); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline D3D12ZoneScope(D3D12QueueCtx* ctx, ID3D12GraphicsCommandList* cmdList, bool active) +#ifdef TRACY_ON_DEMAND + : m_active(active&& GetProfiler().IsConnected()) +#else + : m_active(active) +#endif + { + if (!m_active) return; + + m_ctx = ctx; + m_cmdList = cmdList; + + m_queryId = m_ctx->NextQueryId(); + m_cmdList->EndQuery(m_ctx->m_queryHeap, D3D12_QUERY_TYPE_TIMESTAMP, m_queryId); + } + + public: + tracy_force_inline D3D12ZoneScope(D3D12QueueCtx* ctx, ID3D12GraphicsCommandList* cmdList, const SourceLocationData* srcLocation, bool active) + : D3D12ZoneScope(ctx, cmdList, active) + { + if (!m_active) return; + + auto* item = Profiler::QueueSerial(); + WriteQueueItem(item, QueueType::GpuZoneBeginSerial, reinterpret_cast(srcLocation)); + } + + tracy_force_inline D3D12ZoneScope(D3D12QueueCtx* ctx, ID3D12GraphicsCommandList* cmdList, const SourceLocationData* srcLocation, int depth, bool active) + : D3D12ZoneScope(ctx, cmdList, active) + { + if (!m_active) return; + + auto* item = Profiler::QueueSerialCallstack(Callstack(depth)); + WriteQueueItem(item, QueueType::GpuZoneBeginCallstackSerial, reinterpret_cast(srcLocation)); + } + + tracy_force_inline D3D12ZoneScope(D3D12QueueCtx* ctx, uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, ID3D12GraphicsCommandList* cmdList, bool active) + : D3D12ZoneScope(ctx, cmdList, active) + { + if (!m_active) return; + + const auto sourceLocation = Profiler::AllocSourceLocation(line, source, sourceSz, function, functionSz, name, nameSz); + + auto* item = Profiler::QueueSerial(); + WriteQueueItem(item, QueueType::GpuZoneBeginAllocSrcLocSerial, sourceLocation); + } + + tracy_force_inline D3D12ZoneScope(D3D12QueueCtx* ctx, uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, ID3D12GraphicsCommandList* cmdList, int depth, bool active) + : D3D12ZoneScope(ctx, cmdList, active) + { + if (!m_active) return; + + const auto sourceLocation = Profiler::AllocSourceLocation(line, source, sourceSz, function, functionSz, name, nameSz); + + auto* item = Profiler::QueueSerialCallstack(Callstack(depth)); + WriteQueueItem(item, QueueType::GpuZoneBeginAllocSrcLocCallstackSerial, sourceLocation); + } + + tracy_force_inline ~D3D12ZoneScope() + { + if (!m_active) return; + + const auto queryId = m_queryId + 1; // Our end query slot is immediately after the begin slot. + m_cmdList->EndQuery(m_ctx->m_queryHeap, D3D12_QUERY_TYPE_TIMESTAMP, queryId); + + auto* item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuZoneEndSerial); + MemWrite(&item->gpuZoneEnd.cpuTime, Profiler::GetTime()); + MemWrite(&item->gpuZoneEnd.thread, GetThreadHandle()); + MemWrite(&item->gpuZoneEnd.queryId, static_cast(queryId)); + MemWrite(&item->gpuZoneEnd.context, m_ctx->GetId()); + Profiler::QueueSerialFinish(); + + m_cmdList->ResolveQueryData(m_ctx->m_queryHeap, D3D12_QUERY_TYPE_TIMESTAMP, m_queryId, 2, m_ctx->m_readbackBuffer, m_queryId * sizeof(uint64_t)); + } + }; + + static inline D3D12QueueCtx* CreateD3D12Context(ID3D12Device* device, ID3D12CommandQueue* queue) + { + auto* ctx = static_cast(tracy_malloc(sizeof(D3D12QueueCtx))); + new (ctx) D3D12QueueCtx{ device, queue }; + + return ctx; + } + + static inline void DestroyD3D12Context(D3D12QueueCtx* ctx) + { + ctx->~D3D12QueueCtx(); + tracy_free(ctx); + } + +} + +#undef TracyD3D12Panic + +using TracyD3D12Ctx = tracy::D3D12QueueCtx*; + +#define TracyD3D12Context(device, queue) tracy::CreateD3D12Context(device, queue); +#define TracyD3D12Destroy(ctx) tracy::DestroyD3D12Context(ctx); +#define TracyD3D12ContextName(ctx, name, size) ctx->Name(name, size); + +#define TracyD3D12NewFrame(ctx) ctx->NewFrame(); + +#define TracyD3D12UnnamedZone ___tracy_gpu_d3d12_zone +#define TracyD3D12SrcLocSymbol TracyConcat(__tracy_d3d12_source_location,TracyLine) +#define TracyD3D12SrcLocObject(name, color) static constexpr tracy::SourceLocationData TracyD3D12SrcLocSymbol { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; + +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK +# define TracyD3D12Zone(ctx, cmdList, name) TracyD3D12NamedZoneS(ctx, TracyD3D12UnnamedZone, cmdList, name, TRACY_CALLSTACK, true) +# define TracyD3D12ZoneC(ctx, cmdList, name, color) TracyD3D12NamedZoneCS(ctx, TracyD3D12UnnamedZone, cmdList, name, color, TRACY_CALLSTACK, true) +# define TracyD3D12NamedZone(ctx, varname, cmdList, name, active) TracyD3D12SrcLocObject(name, 0); tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyD3D12SrcLocSymbol, TRACY_CALLSTACK, active }; +# define TracyD3D12NamedZoneC(ctx, varname, cmdList, name, color, active) TracyD3D12SrcLocObject(name, color); tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyD3D12SrcLocSymbol, TRACY_CALLSTACK, active }; +# define TracyD3D12ZoneTransient(ctx, varname, cmdList, name, active) TracyD3D12ZoneTransientS(ctx, varname, cmdList, name, TRACY_CALLSTACK, active) +#else +# define TracyD3D12Zone(ctx, cmdList, name) TracyD3D12NamedZone(ctx, TracyD3D12UnnamedZone, cmdList, name, true) +# define TracyD3D12ZoneC(ctx, cmdList, name, color) TracyD3D12NamedZoneC(ctx, TracyD3D12UnnamedZone, cmdList, name, color, true) +# define TracyD3D12NamedZone(ctx, varname, cmdList, name, active) TracyD3D12SrcLocObject(name, 0); tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyD3D12SrcLocSymbol, active }; +# define TracyD3D12NamedZoneC(ctx, varname, cmdList, name, color, active) TracyD3D12SrcLocObject(name, color); tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyD3D12SrcLocSymbol, active }; +# define TracyD3D12ZoneTransient(ctx, varname, cmdList, name, active) tracy::D3D12ZoneScope varname{ ctx, TracyLine, TracyFile, strlen(TracyFile), TracyFunction, strlen(TracyFunction), name, strlen(name), cmdList, active }; +#endif + +#ifdef TRACY_HAS_CALLSTACK +# define TracyD3D12ZoneS(ctx, cmdList, name, depth) TracyD3D12NamedZoneS(ctx, TracyD3D12UnnamedZone, cmdList, name, depth, true) +# define TracyD3D12ZoneCS(ctx, cmdList, name, color, depth) TracyD3D12NamedZoneCS(ctx, TracyD3D12UnnamedZone, cmdList, name, color, depth, true) +# define TracyD3D12NamedZoneS(ctx, varname, cmdList, name, depth, active) TracyD3D12SrcLocObject(name, 0); tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyD3D12SrcLocSymbol, depth, active }; +# define TracyD3D12NamedZoneCS(ctx, varname, cmdList, name, color, depth, active) TracyD3D12SrcLocObject(name, color); tracy::D3D12ZoneScope varname{ ctx, cmdList, &TracyD3D12SrcLocSymbol, depth, active }; +# define TracyD3D12ZoneTransientS(ctx, varname, cmdList, name, depth, active) tracy::D3D12ZoneScope varname{ ctx, TracyLine, TracyFile, strlen(TracyFile), TracyFunction, strlen(TracyFunction), name, strlen(name), cmdList, depth, active }; +#else +# define TracyD3D12ZoneS(ctx, cmdList, name, depth) TracyD3D12Zone(ctx, cmdList, name) +# define TracyD3D12ZoneCS(ctx, cmdList, name, color, depth) TracyD3D12Zone(ctx, cmdList, name, color) +# define TracyD3D12NamedZoneS(ctx, varname, cmdList, name, depth, active) TracyD3D12NamedZone(ctx, varname, cmdList, name, active) +# define TracyD3D12NamedZoneCS(ctx, varname, cmdList, name, color, depth, active) TracyD3D12NamedZoneC(ctx, varname, cmdList, name, color, active) +# define TracyD3D12ZoneTransientS(ctx, varname, cmdList, name, depth, active) TracyD3D12ZoneTransient(ctx, varname, cmdList, name, active) +#endif + +#define TracyD3D12Collect(ctx) ctx->Collect(); + +#endif + +#endif diff --git a/externals/tracy/public/tracy/TracyLua.hpp b/externals/tracy/public/tracy/TracyLua.hpp new file mode 100644 index 000000000..c972ffb26 --- /dev/null +++ b/externals/tracy/public/tracy/TracyLua.hpp @@ -0,0 +1,431 @@ +#ifndef __TRACYLUA_HPP__ +#define __TRACYLUA_HPP__ + +// Include this file after you include lua headers. + +#ifndef TRACY_ENABLE + +#include + +namespace tracy +{ + +namespace detail +{ +static inline int noop( lua_State* L ) { return 0; } +} + +static inline void LuaRegister( lua_State* L ) +{ + lua_newtable( L ); + lua_pushcfunction( L, detail::noop ); + lua_setfield( L, -2, "ZoneBegin" ); + lua_pushcfunction( L, detail::noop ); + lua_setfield( L, -2, "ZoneBeginN" ); + lua_pushcfunction( L, detail::noop ); + lua_setfield( L, -2, "ZoneBeginS" ); + lua_pushcfunction( L, detail::noop ); + lua_setfield( L, -2, "ZoneBeginNS" ); + lua_pushcfunction( L, detail::noop ); + lua_setfield( L, -2, "ZoneEnd" ); + lua_pushcfunction( L, detail::noop ); + lua_setfield( L, -2, "ZoneText" ); + lua_pushcfunction( L, detail::noop ); + lua_setfield( L, -2, "ZoneName" ); + lua_pushcfunction( L, detail::noop ); + lua_setfield( L, -2, "Message" ); + lua_setglobal( L, "tracy" ); +} + +static inline char* FindEnd( char* ptr ) +{ + unsigned int cnt = 1; + while( cnt != 0 ) + { + if( *ptr == '(' ) cnt++; + else if( *ptr == ')' ) cnt--; + ptr++; + } + return ptr; +} + +static inline void LuaRemove( char* script ) +{ + while( *script ) + { + if( strncmp( script, "tracy.", 6 ) == 0 ) + { + if( strncmp( script + 6, "Zone", 4 ) == 0 ) + { + if( strncmp( script + 10, "End()", 5 ) == 0 ) + { + memset( script, ' ', 15 ); + script += 15; + } + else if( strncmp( script + 10, "Begin()", 7 ) == 0 ) + { + memset( script, ' ', 17 ); + script += 17; + } + else if( strncmp( script + 10, "Text(", 5 ) == 0 ) + { + auto end = FindEnd( script + 15 ); + memset( script, ' ', end - script ); + script = end; + } + else if( strncmp( script + 10, "Name(", 5 ) == 0 ) + { + auto end = FindEnd( script + 15 ); + memset( script, ' ', end - script ); + script = end; + } + else if( strncmp( script + 10, "BeginN(", 7 ) == 0 ) + { + auto end = FindEnd( script + 17 ); + memset( script, ' ', end - script ); + script = end; + } + else if( strncmp( script + 10, "BeginS(", 7 ) == 0 ) + { + auto end = FindEnd( script + 17 ); + memset( script, ' ', end - script ); + script = end; + } + else if( strncmp( script + 10, "BeginNS(", 8 ) == 0 ) + { + auto end = FindEnd( script + 18 ); + memset( script, ' ', end - script ); + script = end; + } + else + { + script += 10; + } + } + else if( strncmp( script + 6, "Message(", 8 ) == 0 ) + { + auto end = FindEnd( script + 14 ); + memset( script, ' ', end - script ); + script = end; + } + else + { + script += 6; + } + } + else + { + script++; + } + } +} + +} + +#else + +#include +#include + +#include "../common/TracyColor.hpp" +#include "../common/TracyAlign.hpp" +#include "../common/TracyForceInline.hpp" +#include "../common/TracySystem.hpp" +#include "../client/TracyProfiler.hpp" + +namespace tracy +{ + +#ifdef TRACY_ON_DEMAND +TRACY_API LuaZoneState& GetLuaZoneState(); +#endif + +namespace detail +{ + +#ifdef TRACY_HAS_CALLSTACK +static tracy_force_inline void SendLuaCallstack( lua_State* L, uint32_t depth ) +{ + assert( depth <= 64 ); + lua_Debug dbg[64]; + const char* func[64]; + uint32_t fsz[64]; + uint32_t ssz[64]; + + uint8_t cnt; + uint16_t spaceNeeded = sizeof( cnt ); + for( cnt=0; cnt::max)() ); + memcpy( dst, fsz+i, 2 ); dst += 2; + memcpy( dst, func[i], fsz[i] ); dst += fsz[i]; + assert( ssz[i] <= (std::numeric_limits::max)() ); + memcpy( dst, ssz+i, 2 ); dst += 2; + memcpy( dst, dbg[i].source, ssz[i] ), dst += ssz[i]; + } + assert( dst - ptr == spaceNeeded + 2 ); + + TracyQueuePrepare( QueueType::CallstackAlloc ); + MemWrite( &item->callstackAllocFat.ptr, (uint64_t)ptr ); + MemWrite( &item->callstackAllocFat.nativePtr, (uint64_t)Callstack( depth ) ); + TracyQueueCommit( callstackAllocFatThread ); +} + +static inline int LuaZoneBeginS( lua_State* L ) +{ +#ifdef TRACY_ON_DEMAND + const auto zoneCnt = GetLuaZoneState().counter++; + if( zoneCnt != 0 && !GetLuaZoneState().active ) return 0; + GetLuaZoneState().active = GetProfiler().IsConnected(); + if( !GetLuaZoneState().active ) return 0; +#endif + +#ifdef TRACY_CALLSTACK + const uint32_t depth = TRACY_CALLSTACK; +#else + const auto depth = uint32_t( lua_tointeger( L, 1 ) ); +#endif + SendLuaCallstack( L, depth ); + + lua_Debug dbg; + lua_getstack( L, 1, &dbg ); + lua_getinfo( L, "Snl", &dbg ); + const auto srcloc = Profiler::AllocSourceLocation( dbg.currentline, dbg.source, dbg.name ? dbg.name : dbg.short_src ); + + TracyQueuePrepare( QueueType::ZoneBeginAllocSrcLocCallstack ); + MemWrite( &item->zoneBegin.time, Profiler::GetTime() ); + MemWrite( &item->zoneBegin.srcloc, srcloc ); + TracyQueueCommit( zoneBeginThread ); + + return 0; +} + +static inline int LuaZoneBeginNS( lua_State* L ) +{ +#ifdef TRACY_ON_DEMAND + const auto zoneCnt = GetLuaZoneState().counter++; + if( zoneCnt != 0 && !GetLuaZoneState().active ) return 0; + GetLuaZoneState().active = GetProfiler().IsConnected(); + if( !GetLuaZoneState().active ) return 0; +#endif + +#ifdef TRACY_CALLSTACK + const uint32_t depth = TRACY_CALLSTACK; +#else + const auto depth = uint32_t( lua_tointeger( L, 2 ) ); +#endif + SendLuaCallstack( L, depth ); + + lua_Debug dbg; + lua_getstack( L, 1, &dbg ); + lua_getinfo( L, "Snl", &dbg ); + size_t nsz; + const auto name = lua_tolstring( L, 1, &nsz ); + const auto srcloc = Profiler::AllocSourceLocation( dbg.currentline, dbg.source, dbg.name ? dbg.name : dbg.short_src, name, nsz ); + + TracyQueuePrepare( QueueType::ZoneBeginAllocSrcLocCallstack ); + MemWrite( &item->zoneBegin.time, Profiler::GetTime() ); + MemWrite( &item->zoneBegin.srcloc, srcloc ); + TracyQueueCommit( zoneBeginThread ); + + return 0; +} +#endif + +static inline int LuaZoneBegin( lua_State* L ) +{ +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK + return LuaZoneBeginS( L ); +#else +#ifdef TRACY_ON_DEMAND + const auto zoneCnt = GetLuaZoneState().counter++; + if( zoneCnt != 0 && !GetLuaZoneState().active ) return 0; + GetLuaZoneState().active = GetProfiler().IsConnected(); + if( !GetLuaZoneState().active ) return 0; +#endif + + lua_Debug dbg; + lua_getstack( L, 1, &dbg ); + lua_getinfo( L, "Snl", &dbg ); + const auto srcloc = Profiler::AllocSourceLocation( dbg.currentline, dbg.source, dbg.name ? dbg.name : dbg.short_src ); + + TracyQueuePrepare( QueueType::ZoneBeginAllocSrcLoc ); + MemWrite( &item->zoneBegin.time, Profiler::GetTime() ); + MemWrite( &item->zoneBegin.srcloc, srcloc ); + TracyQueueCommit( zoneBeginThread ); + return 0; +#endif +} + +static inline int LuaZoneBeginN( lua_State* L ) +{ +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK + return LuaZoneBeginNS( L ); +#else +#ifdef TRACY_ON_DEMAND + const auto zoneCnt = GetLuaZoneState().counter++; + if( zoneCnt != 0 && !GetLuaZoneState().active ) return 0; + GetLuaZoneState().active = GetProfiler().IsConnected(); + if( !GetLuaZoneState().active ) return 0; +#endif + + lua_Debug dbg; + lua_getstack( L, 1, &dbg ); + lua_getinfo( L, "Snl", &dbg ); + size_t nsz; + const auto name = lua_tolstring( L, 1, &nsz ); + const auto srcloc = Profiler::AllocSourceLocation( dbg.currentline, dbg.source, dbg.name ? dbg.name : dbg.short_src, name, nsz ); + + TracyQueuePrepare( QueueType::ZoneBeginAllocSrcLoc ); + MemWrite( &item->zoneBegin.time, Profiler::GetTime() ); + MemWrite( &item->zoneBegin.srcloc, srcloc ); + TracyQueueCommit( zoneBeginThread ); + return 0; +#endif +} + +static inline int LuaZoneEnd( lua_State* L ) +{ +#ifdef TRACY_ON_DEMAND + assert( GetLuaZoneState().counter != 0 ); + GetLuaZoneState().counter--; + if( !GetLuaZoneState().active ) return 0; + if( !GetProfiler().IsConnected() ) + { + GetLuaZoneState().active = false; + return 0; + } +#endif + + TracyQueuePrepare( QueueType::ZoneEnd ); + MemWrite( &item->zoneEnd.time, Profiler::GetTime() ); + TracyQueueCommit( zoneEndThread ); + return 0; +} + +static inline int LuaZoneText( lua_State* L ) +{ +#ifdef TRACY_ON_DEMAND + if( !GetLuaZoneState().active ) return 0; + if( !GetProfiler().IsConnected() ) + { + GetLuaZoneState().active = false; + return 0; + } +#endif + + auto txt = lua_tostring( L, 1 ); + const auto size = strlen( txt ); + assert( size < (std::numeric_limits::max)() ); + + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, txt, size ); + + TracyQueuePrepare( QueueType::ZoneText ); + MemWrite( &item->zoneTextFat.text, (uint64_t)ptr ); + MemWrite( &item->zoneTextFat.size, (uint16_t)size ); + TracyQueueCommit( zoneTextFatThread ); + return 0; +} + +static inline int LuaZoneName( lua_State* L ) +{ +#ifdef TRACY_ON_DEMAND + if( !GetLuaZoneState().active ) return 0; + if( !GetProfiler().IsConnected() ) + { + GetLuaZoneState().active = false; + return 0; + } +#endif + + auto txt = lua_tostring( L, 1 ); + const auto size = strlen( txt ); + assert( size < (std::numeric_limits::max)() ); + + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, txt, size ); + + TracyQueuePrepare( QueueType::ZoneName ); + MemWrite( &item->zoneTextFat.text, (uint64_t)ptr ); + MemWrite( &item->zoneTextFat.size, (uint16_t)size ); + TracyQueueCommit( zoneTextFatThread ); + return 0; +} + +static inline int LuaMessage( lua_State* L ) +{ +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) return 0; +#endif + + auto txt = lua_tostring( L, 1 ); + const auto size = strlen( txt ); + assert( size < (std::numeric_limits::max)() ); + + auto ptr = (char*)tracy_malloc( size ); + memcpy( ptr, txt, size ); + + TracyQueuePrepare( QueueType::Message ); + MemWrite( &item->messageFat.time, Profiler::GetTime() ); + MemWrite( &item->messageFat.text, (uint64_t)ptr ); + MemWrite( &item->messageFat.size, (uint16_t)size ); + TracyQueueCommit( messageFatThread ); + return 0; +} + +} + +static inline void LuaRegister( lua_State* L ) +{ + lua_newtable( L ); + lua_pushcfunction( L, detail::LuaZoneBegin ); + lua_setfield( L, -2, "ZoneBegin" ); + lua_pushcfunction( L, detail::LuaZoneBeginN ); + lua_setfield( L, -2, "ZoneBeginN" ); +#ifdef TRACY_HAS_CALLSTACK + lua_pushcfunction( L, detail::LuaZoneBeginS ); + lua_setfield( L, -2, "ZoneBeginS" ); + lua_pushcfunction( L, detail::LuaZoneBeginNS ); + lua_setfield( L, -2, "ZoneBeginNS" ); +#else + lua_pushcfunction( L, detail::LuaZoneBegin ); + lua_setfield( L, -2, "ZoneBeginS" ); + lua_pushcfunction( L, detail::LuaZoneBeginN ); + lua_setfield( L, -2, "ZoneBeginNS" ); +#endif + lua_pushcfunction( L, detail::LuaZoneEnd ); + lua_setfield( L, -2, "ZoneEnd" ); + lua_pushcfunction( L, detail::LuaZoneText ); + lua_setfield( L, -2, "ZoneText" ); + lua_pushcfunction( L, detail::LuaZoneName ); + lua_setfield( L, -2, "ZoneName" ); + lua_pushcfunction( L, detail::LuaMessage ); + lua_setfield( L, -2, "Message" ); + lua_setglobal( L, "tracy" ); +} + +static inline void LuaRemove( char* script ) {} + +} + +#endif + +#endif diff --git a/externals/tracy/public/tracy/TracyOpenCL.hpp b/externals/tracy/public/tracy/TracyOpenCL.hpp new file mode 100644 index 000000000..34466ccc9 --- /dev/null +++ b/externals/tracy/public/tracy/TracyOpenCL.hpp @@ -0,0 +1,414 @@ +#ifndef __TRACYOPENCL_HPP__ +#define __TRACYOPENCL_HPP__ + +#if !defined TRACY_ENABLE + +#define TracyCLContext(c, x) nullptr +#define TracyCLDestroy(c) +#define TracyCLContextName(c, x, y) + +#define TracyCLNamedZone(c, x, y, z) +#define TracyCLNamedZoneC(c, x, y, z, w) +#define TracyCLZone(c, x) +#define TracyCLZoneC(c, x, y) +#define TracyCLZoneTransient(c,x,y,z) + +#define TracyCLNamedZoneS(c, x, y, z, w) +#define TracyCLNamedZoneCS(c, x, y, z, w, v) +#define TracyCLZoneS(c, x, y) +#define TracyCLZoneCS(c, x, y, z) +#define TracyCLZoneTransientS(c,x,y,z,w) + +#define TracyCLNamedZoneSetEvent(x, e) +#define TracyCLZoneSetEvent(e) + +#define TracyCLCollect(c) + +namespace tracy +{ + class OpenCLCtxScope {}; +} + +using TracyCLCtx = void*; + +#else + +#include + +#include +#include +#include + +#include "Tracy.hpp" +#include "../client/TracyCallstack.hpp" +#include "../client/TracyProfiler.hpp" +#include "../common/TracyAlloc.hpp" + +#define TRACY_CL_TO_STRING_INDIRECT(T) #T +#define TRACY_CL_TO_STRING(T) TRACY_CL_TO_STRING_INDIRECT(T) +#define TRACY_CL_ASSERT(p) if(!(p)) { \ + TracyMessageL( "TRACY_CL_ASSERT failed on " TracyFile ":" TRACY_CL_TO_STRING(TracyLine) ); \ + assert(false && "TRACY_CL_ASSERT failed"); \ +} +#define TRACY_CL_CHECK_ERROR(err) if(err != CL_SUCCESS) { \ + std::ostringstream oss; \ + oss << "TRACY_CL_CHECK_ERROR failed on " << TracyFile << ":" << TracyLine \ + << ": error code " << err; \ + auto msg = oss.str(); \ + TracyMessage(msg.data(), msg.size()); \ + assert(false && "TRACY_CL_CHECK_ERROR failed"); \ +} + +namespace tracy { + + enum class EventPhase : uint8_t + { + Begin, + End + }; + + struct EventInfo + { + cl_event event; + EventPhase phase; + }; + + class OpenCLCtx + { + public: + enum { QueryCount = 64 * 1024 }; + + OpenCLCtx(cl_context context, cl_device_id device) + : m_contextId(GetGpuCtxCounter().fetch_add(1, std::memory_order_relaxed)) + , m_head(0) + , m_tail(0) + { + int64_t tcpu, tgpu; + TRACY_CL_ASSERT(m_contextId != 255); + + cl_int err = CL_SUCCESS; + cl_command_queue queue = clCreateCommandQueue(context, device, CL_QUEUE_PROFILING_ENABLE, &err); + TRACY_CL_CHECK_ERROR(err) + uint32_t dummyValue = 42; + cl_mem dummyBuffer = clCreateBuffer(context, CL_MEM_WRITE_ONLY, sizeof(uint32_t), nullptr, &err); + TRACY_CL_CHECK_ERROR(err) + cl_event writeBufferEvent; + TRACY_CL_CHECK_ERROR(clEnqueueWriteBuffer(queue, dummyBuffer, CL_FALSE, 0, sizeof(uint32_t), &dummyValue, 0, nullptr, &writeBufferEvent)); + TRACY_CL_CHECK_ERROR(clWaitForEvents(1, &writeBufferEvent)); + + tcpu = Profiler::GetTime(); + + cl_int eventStatus; + TRACY_CL_CHECK_ERROR(clGetEventInfo(writeBufferEvent, CL_EVENT_COMMAND_EXECUTION_STATUS, sizeof(cl_int), &eventStatus, nullptr)); + TRACY_CL_ASSERT(eventStatus == CL_COMPLETE); + TRACY_CL_CHECK_ERROR(clGetEventProfilingInfo(writeBufferEvent, CL_PROFILING_COMMAND_END, sizeof(cl_ulong), &tgpu, nullptr)); + TRACY_CL_CHECK_ERROR(clReleaseEvent(writeBufferEvent)); + TRACY_CL_CHECK_ERROR(clReleaseMemObject(dummyBuffer)); + TRACY_CL_CHECK_ERROR(clReleaseCommandQueue(queue)); + + auto item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuNewContext); + MemWrite(&item->gpuNewContext.cpuTime, tcpu); + MemWrite(&item->gpuNewContext.gpuTime, tgpu); + memset(&item->gpuNewContext.thread, 0, sizeof(item->gpuNewContext.thread)); + MemWrite(&item->gpuNewContext.period, 1.0f); + MemWrite(&item->gpuNewContext.type, GpuContextType::OpenCL); + MemWrite(&item->gpuNewContext.context, (uint8_t) m_contextId); + MemWrite(&item->gpuNewContext.flags, (uint8_t)0); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem(*item); +#endif + Profiler::QueueSerialFinish(); + } + + void Name( const char* name, uint16_t len ) + { + auto ptr = (char*)tracy_malloc( len ); + memcpy( ptr, name, len ); + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuContextName ); + MemWrite( &item->gpuContextNameFat.context, (uint8_t)m_contextId ); + MemWrite( &item->gpuContextNameFat.ptr, (uint64_t)ptr ); + MemWrite( &item->gpuContextNameFat.size, len ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + + void Collect() + { + ZoneScopedC(Color::Red4); + + if (m_tail == m_head) return; + +#ifdef TRACY_ON_DEMAND + if (!GetProfiler().IsConnected()) + { + m_head = m_tail = 0; + } +#endif + + for (; m_tail != m_head; m_tail = (m_tail + 1) % QueryCount) + { + EventInfo eventInfo = GetQuery(m_tail); + cl_int eventStatus; + cl_int err = clGetEventInfo(eventInfo.event, CL_EVENT_COMMAND_EXECUTION_STATUS, sizeof(cl_int), &eventStatus, nullptr); + if (err != CL_SUCCESS) + { + std::ostringstream oss; + oss << "clGetEventInfo falied with error code " << err << ", on event " << eventInfo.event << ", skipping..."; + auto msg = oss.str(); + TracyMessage(msg.data(), msg.size()); + if (eventInfo.event == nullptr) { + TracyMessageL("A TracyCLZone must be paird with a TracyCLZoneSetEvent, check your code!"); + } + assert(false && "clGetEventInfo failed, maybe a TracyCLZone is not paired with TracyCLZoneSetEvent"); + continue; + } + if (eventStatus != CL_COMPLETE) return; + + cl_int eventInfoQuery = (eventInfo.phase == EventPhase::Begin) + ? CL_PROFILING_COMMAND_START + : CL_PROFILING_COMMAND_END; + + cl_ulong eventTimeStamp = 0; + err = clGetEventProfilingInfo(eventInfo.event, eventInfoQuery, sizeof(cl_ulong), &eventTimeStamp, nullptr); + if (err == CL_PROFILING_INFO_NOT_AVAILABLE) + { + TracyMessageL("command queue is not created with CL_QUEUE_PROFILING_ENABLE flag, check your code!"); + assert(false && "command queue is not created with CL_QUEUE_PROFILING_ENABLE flag"); + } + else + TRACY_CL_CHECK_ERROR(err); + + TRACY_CL_ASSERT(eventTimeStamp != 0); + + auto item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuTime); + MemWrite(&item->gpuTime.gpuTime, (int64_t)eventTimeStamp); + MemWrite(&item->gpuTime.queryId, (uint16_t)m_tail); + MemWrite(&item->gpuTime.context, m_contextId); + Profiler::QueueSerialFinish(); + + if (eventInfo.phase == EventPhase::End) + { + // Done with the event, so release it + TRACY_CL_CHECK_ERROR(clReleaseEvent(eventInfo.event)); + } + } + } + + tracy_force_inline uint8_t GetId() const + { + return m_contextId; + } + + tracy_force_inline unsigned int NextQueryId(EventInfo eventInfo) + { + const auto id = m_head; + m_head = (m_head + 1) % QueryCount; + TRACY_CL_ASSERT(m_head != m_tail); + m_query[id] = eventInfo; + return id; + } + + tracy_force_inline EventInfo& GetQuery(unsigned int id) + { + TRACY_CL_ASSERT(id < QueryCount); + return m_query[id]; + } + + private: + + unsigned int m_contextId; + + EventInfo m_query[QueryCount]; + unsigned int m_head; // index at which a new event should be inserted + unsigned int m_tail; // oldest event + + }; + + class OpenCLCtxScope { + public: + tracy_force_inline OpenCLCtxScope(OpenCLCtx* ctx, const SourceLocationData* srcLoc, bool is_active) +#ifdef TRACY_ON_DEMAND + : m_active(is_active&& GetProfiler().IsConnected()) +#else + : m_active(is_active) +#endif + , m_ctx(ctx) + , m_event(nullptr) + { + if (!m_active) return; + + m_beginQueryId = ctx->NextQueryId(EventInfo{ nullptr, EventPhase::Begin }); + + auto item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuZoneBeginSerial); + MemWrite(&item->gpuZoneBegin.cpuTime, Profiler::GetTime()); + MemWrite(&item->gpuZoneBegin.srcloc, (uint64_t)srcLoc); + MemWrite(&item->gpuZoneBegin.thread, GetThreadHandle()); + MemWrite(&item->gpuZoneBegin.queryId, (uint16_t)m_beginQueryId); + MemWrite(&item->gpuZoneBegin.context, ctx->GetId()); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline OpenCLCtxScope(OpenCLCtx* ctx, const SourceLocationData* srcLoc, int depth, bool is_active) +#ifdef TRACY_ON_DEMAND + : m_active(is_active&& GetProfiler().IsConnected()) +#else + : m_active(is_active) +#endif + , m_ctx(ctx) + , m_event(nullptr) + { + if (!m_active) return; + + m_beginQueryId = ctx->NextQueryId(EventInfo{ nullptr, EventPhase::Begin }); + + GetProfiler().SendCallstack(depth); + + auto item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuZoneBeginCallstackSerial); + MemWrite(&item->gpuZoneBegin.cpuTime, Profiler::GetTime()); + MemWrite(&item->gpuZoneBegin.srcloc, (uint64_t)srcLoc); + MemWrite(&item->gpuZoneBegin.thread, GetThreadHandle()); + MemWrite(&item->gpuZoneBegin.queryId, (uint16_t)m_beginQueryId); + MemWrite(&item->gpuZoneBegin.context, ctx->GetId()); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline OpenCLCtxScope(OpenCLCtx* ctx, uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, bool is_active) +#ifdef TRACY_ON_DEMAND + : m_active(is_active && GetProfiler().IsConnected()) +#else + : m_active(is_active) +#endif + , m_ctx(ctx) + , m_event(nullptr) + { + if (!m_active) return; + + m_beginQueryId = ctx->NextQueryId(EventInfo{ nullptr, EventPhase::Begin }); + + const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz ); + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuZoneBeginAllocSrcLocSerial ); + MemWrite(&item->gpuZoneBegin.cpuTime, Profiler::GetTime()); + MemWrite(&item->gpuZoneBegin.srcloc, srcloc); + MemWrite(&item->gpuZoneBegin.thread, GetThreadHandle()); + MemWrite(&item->gpuZoneBegin.queryId, (uint16_t)m_beginQueryId); + MemWrite(&item->gpuZoneBegin.context, ctx->GetId()); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline OpenCLCtxScope(OpenCLCtx* ctx, uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, int depth, bool is_active) +#ifdef TRACY_ON_DEMAND + : m_active(is_active && GetProfiler().IsConnected()) +#else + : m_active(is_active) +#endif + , m_ctx(ctx) + , m_event(nullptr) + { + if (!m_active) return; + + m_beginQueryId = ctx->NextQueryId(EventInfo{ nullptr, EventPhase::Begin }); + + const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz ); + auto item = Profiler::QueueSerialCallstack( Callstack( depth ) ); + MemWrite(&item->hdr.type, QueueType::GpuZoneBeginAllocSrcLocCallstackSerial); + MemWrite(&item->gpuZoneBegin.cpuTime, Profiler::GetTime()); + MemWrite(&item->gpuZoneBegin.srcloc, srcloc); + MemWrite(&item->gpuZoneBegin.thread, GetThreadHandle()); + MemWrite(&item->gpuZoneBegin.queryId, (uint16_t)m_beginQueryId); + MemWrite(&item->gpuZoneBegin.context, ctx->GetId()); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline void SetEvent(cl_event event) + { + if (!m_active) return; + m_event = event; + TRACY_CL_CHECK_ERROR(clRetainEvent(m_event)); + m_ctx->GetQuery(m_beginQueryId).event = m_event; + } + + tracy_force_inline ~OpenCLCtxScope() + { + if (!m_active) return; + const auto queryId = m_ctx->NextQueryId(EventInfo{ m_event, EventPhase::End }); + + auto item = Profiler::QueueSerial(); + MemWrite(&item->hdr.type, QueueType::GpuZoneEndSerial); + MemWrite(&item->gpuZoneEnd.cpuTime, Profiler::GetTime()); + MemWrite(&item->gpuZoneEnd.thread, GetThreadHandle()); + MemWrite(&item->gpuZoneEnd.queryId, (uint16_t)queryId); + MemWrite(&item->gpuZoneEnd.context, m_ctx->GetId()); + Profiler::QueueSerialFinish(); + } + + const bool m_active; + OpenCLCtx* m_ctx; + cl_event m_event; + unsigned int m_beginQueryId; + }; + + static inline OpenCLCtx* CreateCLContext(cl_context context, cl_device_id device) + { + auto ctx = (OpenCLCtx*)tracy_malloc(sizeof(OpenCLCtx)); + new (ctx) OpenCLCtx(context, device); + return ctx; + } + + static inline void DestroyCLContext(OpenCLCtx* ctx) + { + ctx->~OpenCLCtx(); + tracy_free(ctx); + } + +} // namespace tracy + +using TracyCLCtx = tracy::OpenCLCtx*; + +#define TracyCLContext(context, device) tracy::CreateCLContext(context, device); +#define TracyCLDestroy(ctx) tracy::DestroyCLContext(ctx); +#define TracyCLContextName(context, name, size) ctx->Name(name, size); +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK +# define TracyCLNamedZone(ctx, varname, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); +# define TracyCLNamedZoneC(ctx, varname, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); +# define TracyCLZone(ctx, name) TracyCLNamedZoneS(ctx, __tracy_gpu_zone, name, TRACY_CALLSTACK, true) +# define TracyCLZoneC(ctx, name, color) TracyCLNamedZoneCS(ctx, __tracy_gpu_zone, name, color, TRACY_CALLSTACK, true) +# define TracyCLZoneTransient( ctx, varname, name, active ) tracy::OpenCLCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), TRACY_CALLSTACK, active ); +#else +# define TracyCLNamedZone(ctx, varname, name, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine){ name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), active); +# define TracyCLNamedZoneC(ctx, varname, name, color, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine){ name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), active); +# define TracyCLZone(ctx, name) TracyCLNamedZone(ctx, __tracy_gpu_zone, name, true) +# define TracyCLZoneC(ctx, name, color) TracyCLNamedZoneC(ctx, __tracy_gpu_zone, name, color, true ) +# define TracyCLZoneTransient( ctx, varname, name, active ) tracy::OpenCLCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), active ); +#endif + +#ifdef TRACY_HAS_CALLSTACK +# define TracyCLNamedZoneS(ctx, varname, name, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine){ name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active); +# define TracyCLNamedZoneCS(ctx, varname, name, color, depth, active) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine){ name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::OpenCLCtxScope varname(ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active); +# define TracyCLZoneS(ctx, name, depth) TracyCLNamedZoneS(ctx, __tracy_gpu_zone, name, depth, true) +# define TracyCLZoneCS(ctx, name, color, depth) TracyCLNamedZoneCS(ctx, __tracy_gpu_zone, name, color, depth, true) +# define TracyCLZoneTransientS( ctx, varname, name, depth, active ) tracy::OpenCLCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), depth, active ); +#else +# define TracyCLNamedZoneS(ctx, varname, name, depth, active) TracyCLNamedZone(ctx, varname, name, active) +# define TracyCLNamedZoneCS(ctx, varname, name, color, depth, active) TracyCLNamedZoneC(ctx, varname, name, color, active) +# define TracyCLZoneS(ctx, name, depth) TracyCLZone(ctx, name) +# define TracyCLZoneCS(ctx, name, color, depth) TracyCLZoneC(ctx, name, color) +# define TracyCLZoneTransientS( ctx, varname, name, depth, active ) TracyCLZoneTransient( ctx, varname, name, active ) +#endif + +#define TracyCLNamedZoneSetEvent(varname, event) varname.SetEvent(event) +#define TracyCLZoneSetEvent(event) __tracy_gpu_zone.SetEvent(event) + +#define TracyCLCollect(ctx) ctx->Collect() + +#endif + +#endif diff --git a/externals/tracy/public/tracy/TracyOpenGL.hpp b/externals/tracy/public/tracy/TracyOpenGL.hpp new file mode 100644 index 000000000..3bdadccee --- /dev/null +++ b/externals/tracy/public/tracy/TracyOpenGL.hpp @@ -0,0 +1,325 @@ +#ifndef __TRACYOPENGL_HPP__ +#define __TRACYOPENGL_HPP__ + +#if !defined TRACY_ENABLE || defined __APPLE__ + +#define TracyGpuContext +#define TracyGpuContextName(x,y) +#define TracyGpuNamedZone(x,y,z) +#define TracyGpuNamedZoneC(x,y,z,w) +#define TracyGpuZone(x) +#define TracyGpuZoneC(x,y) +#define TracyGpuZoneTransient(x,y,z) +#define TracyGpuCollect + +#define TracyGpuNamedZoneS(x,y,z,w) +#define TracyGpuNamedZoneCS(x,y,z,w,a) +#define TracyGpuZoneS(x,y) +#define TracyGpuZoneCS(x,y,z) +#define TracyGpuZoneTransientS(x,y,z,w) + +namespace tracy +{ +struct SourceLocationData; +class GpuCtxScope +{ +public: + GpuCtxScope( const SourceLocationData*, bool ) {} + GpuCtxScope( const SourceLocationData*, int, bool ) {} +}; +} + +#else + +#include +#include +#include + +#include "Tracy.hpp" +#include "../client/TracyProfiler.hpp" +#include "../client/TracyCallstack.hpp" +#include "../common/TracyAlign.hpp" +#include "../common/TracyAlloc.hpp" + +#if !defined GL_TIMESTAMP && defined GL_TIMESTAMP_EXT +# define GL_TIMESTAMP GL_TIMESTAMP_EXT +# define GL_QUERY_COUNTER_BITS GL_QUERY_COUNTER_BITS_EXT +# define glGetQueryObjectiv glGetQueryObjectivEXT +# define glGetQueryObjectui64v glGetQueryObjectui64vEXT +# define glQueryCounter glQueryCounterEXT +#endif + +#define TracyGpuContext tracy::GetGpuCtx().ptr = (tracy::GpuCtx*)tracy::tracy_malloc( sizeof( tracy::GpuCtx ) ); new(tracy::GetGpuCtx().ptr) tracy::GpuCtx; +#define TracyGpuContextName( name, size ) tracy::GetGpuCtx().ptr->Name( name, size ); +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK +# define TracyGpuNamedZone( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); +# define TracyGpuNamedZoneC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), TRACY_CALLSTACK, active ); +# define TracyGpuZone( name ) TracyGpuNamedZoneS( ___tracy_gpu_zone, name, TRACY_CALLSTACK, true ) +# define TracyGpuZoneC( name, color ) TracyGpuNamedZoneCS( ___tracy_gpu_zone, name, color, TRACY_CALLSTACK, true ) +# define TracyGpuZoneTransient( varname, name, active ) tracy::GpuCtxScope varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), TRACY_CALLSTACK, active ); +#else +# define TracyGpuNamedZone( varname, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), active ); +# define TracyGpuNamedZoneC( varname, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), active ); +# define TracyGpuZone( name ) TracyGpuNamedZone( ___tracy_gpu_zone, name, true ) +# define TracyGpuZoneC( name, color ) TracyGpuNamedZoneC( ___tracy_gpu_zone, name, color, true ) +# define TracyGpuZoneTransient( varname, name, active ) tracy::GpuCtxScope varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), active ); +#endif +#define TracyGpuCollect tracy::GetGpuCtx().ptr->Collect(); + +#ifdef TRACY_HAS_CALLSTACK +# define TracyGpuNamedZoneS( varname, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active ); +# define TracyGpuNamedZoneCS( varname, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::GpuCtxScope varname( &TracyConcat(__tracy_gpu_source_location,TracyLine), depth, active ); +# define TracyGpuZoneS( name, depth ) TracyGpuNamedZoneS( ___tracy_gpu_zone, name, depth, true ) +# define TracyGpuZoneCS( name, color, depth ) TracyGpuNamedZoneCS( ___tracy_gpu_zone, name, color, depth, true ) +# define TracyGpuZoneTransientS( varname, name, depth, active ) tracy::GpuCtxScope varname( TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), depth, active ); +#else +# define TracyGpuNamedZoneS( varname, name, depth, active ) TracyGpuNamedZone( varname, name, active ) +# define TracyGpuNamedZoneCS( varname, name, color, depth, active ) TracyGpuNamedZoneC( varname, name, color, active ) +# define TracyGpuZoneS( name, depth ) TracyGpuZone( name ) +# define TracyGpuZoneCS( name, color, depth ) TracyGpuZoneC( name, color ) +# define TracyGpuZoneTransientS( varname, name, depth, active ) TracyGpuZoneTransient( varname, name, active ) +#endif + +namespace tracy +{ + +class GpuCtx +{ + friend class GpuCtxScope; + + enum { QueryCount = 64 * 1024 }; + +public: + GpuCtx() + : m_context( GetGpuCtxCounter().fetch_add( 1, std::memory_order_relaxed ) ) + , m_head( 0 ) + , m_tail( 0 ) + { + assert( m_context != 255 ); + + glGenQueries( QueryCount, m_query ); + + int64_t tgpu; + glGetInteger64v( GL_TIMESTAMP, &tgpu ); + int64_t tcpu = Profiler::GetTime(); + + GLint bits; + glGetQueryiv( GL_TIMESTAMP, GL_QUERY_COUNTER_BITS, &bits ); + + const float period = 1.f; + const auto thread = GetThreadHandle(); + TracyLfqPrepare( QueueType::GpuNewContext ); + MemWrite( &item->gpuNewContext.cpuTime, tcpu ); + MemWrite( &item->gpuNewContext.gpuTime, tgpu ); + MemWrite( &item->gpuNewContext.thread, thread ); + MemWrite( &item->gpuNewContext.period, period ); + MemWrite( &item->gpuNewContext.context, m_context ); + MemWrite( &item->gpuNewContext.flags, uint8_t( 0 ) ); + MemWrite( &item->gpuNewContext.type, GpuContextType::OpenGl ); + +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + + TracyLfqCommit; + } + + void Name( const char* name, uint16_t len ) + { + auto ptr = (char*)tracy_malloc( len ); + memcpy( ptr, name, len ); + + TracyLfqPrepare( QueueType::GpuContextName ); + MemWrite( &item->gpuContextNameFat.context, m_context ); + MemWrite( &item->gpuContextNameFat.ptr, (uint64_t)ptr ); + MemWrite( &item->gpuContextNameFat.size, len ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + TracyLfqCommit; + } + + void Collect() + { + ZoneScopedC( Color::Red4 ); + + if( m_tail == m_head ) return; + +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) + { + m_head = m_tail = 0; + return; + } +#endif + + while( m_tail != m_head ) + { + GLint available; + glGetQueryObjectiv( m_query[m_tail], GL_QUERY_RESULT_AVAILABLE, &available ); + if( !available ) return; + + uint64_t time; + glGetQueryObjectui64v( m_query[m_tail], GL_QUERY_RESULT, &time ); + + TracyLfqPrepare( QueueType::GpuTime ); + MemWrite( &item->gpuTime.gpuTime, (int64_t)time ); + MemWrite( &item->gpuTime.queryId, (uint16_t)m_tail ); + MemWrite( &item->gpuTime.context, m_context ); + TracyLfqCommit; + + m_tail = ( m_tail + 1 ) % QueryCount; + } + } + +private: + tracy_force_inline unsigned int NextQueryId() + { + const auto id = m_head; + m_head = ( m_head + 1 ) % QueryCount; + assert( m_head != m_tail ); + return id; + } + + tracy_force_inline unsigned int TranslateOpenGlQueryId( unsigned int id ) + { + return m_query[id]; + } + + tracy_force_inline uint8_t GetId() const + { + return m_context; + } + + unsigned int m_query[QueryCount]; + uint8_t m_context; + + unsigned int m_head; + unsigned int m_tail; +}; + +class GpuCtxScope +{ +public: + tracy_force_inline GpuCtxScope( const SourceLocationData* srcloc, bool is_active ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; + + const auto queryId = GetGpuCtx().ptr->NextQueryId(); + glQueryCounter( GetGpuCtx().ptr->TranslateOpenGlQueryId( queryId ), GL_TIMESTAMP ); + + TracyLfqPrepare( QueueType::GpuZoneBegin ); + MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() ); + memset( &item->gpuZoneBegin.thread, 0, sizeof( item->gpuZoneBegin.thread ) ); + MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneBegin.context, GetGpuCtx().ptr->GetId() ); + MemWrite( &item->gpuZoneBegin.srcloc, (uint64_t)srcloc ); + TracyLfqCommit; + } + + tracy_force_inline GpuCtxScope( const SourceLocationData* srcloc, int depth, bool is_active ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; + + const auto queryId = GetGpuCtx().ptr->NextQueryId(); + glQueryCounter( GetGpuCtx().ptr->TranslateOpenGlQueryId( queryId ), GL_TIMESTAMP ); + +#ifdef TRACY_FIBERS + TracyLfqPrepare( QueueType::GpuZoneBegin ); + memset( &item->gpuZoneBegin.thread, 0, sizeof( item->gpuZoneBegin.thread ) ); +#else + GetProfiler().SendCallstack( depth ); + TracyLfqPrepare( QueueType::GpuZoneBeginCallstack ); + MemWrite( &item->gpuZoneBegin.thread, GetThreadHandle() ); +#endif + MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() ); + MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneBegin.context, GetGpuCtx().ptr->GetId() ); + MemWrite( &item->gpuZoneBegin.srcloc, (uint64_t)srcloc ); + TracyLfqCommit; + } + + tracy_force_inline GpuCtxScope( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, bool is_active ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; + + const auto queryId = GetGpuCtx().ptr->NextQueryId(); + glQueryCounter( GetGpuCtx().ptr->TranslateOpenGlQueryId( queryId ), GL_TIMESTAMP ); + + TracyLfqPrepare( QueueType::GpuZoneBeginAllocSrcLoc ); + const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz ); + MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() ); + memset( &item->gpuZoneBegin.thread, 0, sizeof( item->gpuZoneBegin.thread ) ); + MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneBegin.context, GetGpuCtx().ptr->GetId() ); + MemWrite( &item->gpuZoneBegin.srcloc, (uint64_t)srcloc ); + TracyLfqCommit; + } + + tracy_force_inline GpuCtxScope( uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, int depth, bool is_active ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; + + const auto queryId = GetGpuCtx().ptr->NextQueryId(); + glQueryCounter( GetGpuCtx().ptr->TranslateOpenGlQueryId( queryId ), GL_TIMESTAMP ); + +#ifdef TRACY_FIBERS + TracyLfqPrepare( QueueType::GpuZoneBeginAllocSrcLoc ); + memset( &item->gpuZoneBegin.thread, 0, sizeof( item->gpuZoneBegin.thread ) ); +#else + GetProfiler().SendCallstack( depth ); + TracyLfqPrepare( QueueType::GpuZoneBeginAllocSrcLocCallstack ); + MemWrite( &item->gpuZoneBegin.thread, GetThreadHandle() ); +#endif + const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz ); + MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() ); + MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneBegin.context, GetGpuCtx().ptr->GetId() ); + MemWrite( &item->gpuZoneBegin.srcloc, (uint64_t)srcloc ); + TracyLfqCommit; + } + + tracy_force_inline ~GpuCtxScope() + { + if( !m_active ) return; + + const auto queryId = GetGpuCtx().ptr->NextQueryId(); + glQueryCounter( GetGpuCtx().ptr->TranslateOpenGlQueryId( queryId ), GL_TIMESTAMP ); + + TracyLfqPrepare( QueueType::GpuZoneEnd ); + MemWrite( &item->gpuZoneEnd.cpuTime, Profiler::GetTime() ); + memset( &item->gpuZoneEnd.thread, 0, sizeof( item->gpuZoneEnd.thread ) ); + MemWrite( &item->gpuZoneEnd.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneEnd.context, GetGpuCtx().ptr->GetId() ); + TracyLfqCommit; + } + +private: + const bool m_active; +}; + +} + +#endif + +#endif diff --git a/externals/tracy/public/tracy/TracyVulkan.hpp b/externals/tracy/public/tracy/TracyVulkan.hpp new file mode 100644 index 000000000..0ece01d39 --- /dev/null +++ b/externals/tracy/public/tracy/TracyVulkan.hpp @@ -0,0 +1,714 @@ +#ifndef __TRACYVULKAN_HPP__ +#define __TRACYVULKAN_HPP__ + +#if !defined TRACY_ENABLE + +#define TracyVkContext(x,y,z,w) nullptr +#define TracyVkContextCalibrated(x,y,z,w,a,b) nullptr +#if defined VK_EXT_host_query_reset +#define TracyVkContextHostCalibrated(x,y,z,w,a) nullptr +#endif +#define TracyVkDestroy(x) +#define TracyVkContextName(c,x,y) +#define TracyVkNamedZone(c,x,y,z,w) +#define TracyVkNamedZoneC(c,x,y,z,w,a) +#define TracyVkZone(c,x,y) +#define TracyVkZoneC(c,x,y,z) +#define TracyVkZoneTransient(c,x,y,z,w) +#define TracyVkCollect(c,x) + +#define TracyVkNamedZoneS(c,x,y,z,w,a) +#define TracyVkNamedZoneCS(c,x,y,z,w,v,a) +#define TracyVkZoneS(c,x,y,z) +#define TracyVkZoneCS(c,x,y,z,w) +#define TracyVkZoneTransientS(c,x,y,z,w,a) + +namespace tracy +{ +class VkCtxScope {}; +} + +using TracyVkCtx = void*; + +#else + +#if !defined VK_NULL_HANDLE +# error "You must include Vulkan headers before including TracyVulkan.hpp" +#endif + +#include +#include +#include "Tracy.hpp" +#include "../client/TracyProfiler.hpp" +#include "../client/TracyCallstack.hpp" + +#include + +namespace tracy +{ + +#if defined TRACY_VK_USE_SYMBOL_TABLE +#define LoadVkDeviceCoreSymbols(Operation) \ + Operation(vkBeginCommandBuffer) \ + Operation(vkCmdResetQueryPool) \ + Operation(vkCmdWriteTimestamp) \ + Operation(vkCreateQueryPool) \ + Operation(vkDestroyQueryPool) \ + Operation(vkEndCommandBuffer) \ + Operation(vkGetQueryPoolResults) \ + Operation(vkQueueSubmit) \ + Operation(vkQueueWaitIdle) \ + Operation(vkResetQueryPool) + +#define LoadVkDeviceExtensionSymbols(Operation) \ + Operation(vkGetCalibratedTimestampsEXT) + +#define LoadVkInstanceExtensionSymbols(Operation) \ + Operation(vkGetPhysicalDeviceCalibrateableTimeDomainsEXT) + +#define LoadVkInstanceCoreSymbols(Operation) \ + Operation(vkGetPhysicalDeviceProperties) + +struct VkSymbolTable +{ +#define MAKE_PFN(name) PFN_##name name; + LoadVkDeviceCoreSymbols(MAKE_PFN) + LoadVkDeviceExtensionSymbols(MAKE_PFN) + LoadVkInstanceExtensionSymbols(MAKE_PFN) + LoadVkInstanceCoreSymbols(MAKE_PFN) +#undef MAKE_PFN +}; + +#define VK_FUNCTION_WRAPPER(callSignature) m_symbols.callSignature +#define CONTEXT_VK_FUNCTION_WRAPPER(callSignature) m_ctx->m_symbols.callSignature +#else +#define VK_FUNCTION_WRAPPER(callSignature) callSignature +#define CONTEXT_VK_FUNCTION_WRAPPER(callSignature) callSignature +#endif + +class VkCtx +{ + friend class VkCtxScope; + + enum { QueryCount = 64 * 1024 }; + +public: +#if defined TRACY_VK_USE_SYMBOL_TABLE + VkCtx( VkInstance instance, VkPhysicalDevice physdev, VkDevice device, VkQueue queue, VkCommandBuffer cmdbuf, PFN_vkGetInstanceProcAddr instanceProcAddr, PFN_vkGetDeviceProcAddr deviceProcAddr, bool calibrated ) +#else + VkCtx( VkPhysicalDevice physdev, VkDevice device, VkQueue queue, VkCommandBuffer cmdbuf, PFN_vkGetPhysicalDeviceCalibrateableTimeDomainsEXT vkGetPhysicalDeviceCalibrateableTimeDomainsEXT, PFN_vkGetCalibratedTimestampsEXT vkGetCalibratedTimestampsEXT) +#endif + : m_device( device ) + , m_timeDomain( VK_TIME_DOMAIN_DEVICE_EXT ) + , m_context( GetGpuCtxCounter().fetch_add( 1, std::memory_order_relaxed ) ) + , m_head( 0 ) + , m_tail( 0 ) + , m_oldCnt( 0 ) + , m_queryCount( QueryCount ) +#if !defined TRACY_VK_USE_SYMBOL_TABLE + , m_vkGetCalibratedTimestampsEXT( vkGetCalibratedTimestampsEXT ) +#endif + { + assert( m_context != 255 ); + +#if defined TRACY_VK_USE_SYMBOL_TABLE + PopulateSymbolTable(instance, instanceProcAddr, deviceProcAddr); + if ( calibrated ) + { + m_vkGetCalibratedTimestampsEXT = m_symbols.vkGetCalibratedTimestampsEXT; + } + +#endif + + if( VK_FUNCTION_WRAPPER( vkGetPhysicalDeviceCalibrateableTimeDomainsEXT ) && m_vkGetCalibratedTimestampsEXT ) + { + FindAvailableTimeDomains( physdev, VK_FUNCTION_WRAPPER( vkGetPhysicalDeviceCalibrateableTimeDomainsEXT ) ); + } + + CreateQueryPool(); + + VkCommandBufferBeginInfo beginInfo = {}; + beginInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO; + beginInfo.flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT; + + VkSubmitInfo submitInfo = {}; + submitInfo.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO; + submitInfo.commandBufferCount = 1; + submitInfo.pCommandBuffers = &cmdbuf; + + VK_FUNCTION_WRAPPER( vkBeginCommandBuffer( cmdbuf, &beginInfo ) ); + VK_FUNCTION_WRAPPER( vkCmdResetQueryPool( cmdbuf, m_query, 0, m_queryCount ) ); + VK_FUNCTION_WRAPPER( vkEndCommandBuffer( cmdbuf ) ); + VK_FUNCTION_WRAPPER( vkQueueSubmit( queue, 1, &submitInfo, VK_NULL_HANDLE ) ); + VK_FUNCTION_WRAPPER( vkQueueWaitIdle( queue ) ); + + int64_t tcpu, tgpu; + if( m_timeDomain == VK_TIME_DOMAIN_DEVICE_EXT ) + { + VK_FUNCTION_WRAPPER( vkBeginCommandBuffer( cmdbuf, &beginInfo ) ); + VK_FUNCTION_WRAPPER( vkCmdWriteTimestamp( cmdbuf, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, m_query, 0 ) ); + VK_FUNCTION_WRAPPER( vkEndCommandBuffer( cmdbuf ) ); + VK_FUNCTION_WRAPPER( vkQueueSubmit( queue, 1, &submitInfo, VK_NULL_HANDLE ) ); + VK_FUNCTION_WRAPPER( vkQueueWaitIdle( queue ) ); + + tcpu = Profiler::GetTime(); + VK_FUNCTION_WRAPPER( vkGetQueryPoolResults( device, m_query, 0, 1, sizeof( tgpu ), &tgpu, sizeof( tgpu ), VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT ) ); + + VK_FUNCTION_WRAPPER( vkBeginCommandBuffer( cmdbuf, &beginInfo ) ); + VK_FUNCTION_WRAPPER( vkCmdResetQueryPool( cmdbuf, m_query, 0, 1 ) ); + VK_FUNCTION_WRAPPER( vkEndCommandBuffer( cmdbuf ) ); + VK_FUNCTION_WRAPPER( vkQueueSubmit( queue, 1, &submitInfo, VK_NULL_HANDLE ) ); + VK_FUNCTION_WRAPPER( vkQueueWaitIdle( queue ) ); + } + else + { + FindCalibratedTimestampDeviation(); + Calibrate( device, m_prevCalibration, tgpu ); + tcpu = Profiler::GetTime(); + } + + WriteInitialItem( physdev, tcpu, tgpu ); + + m_res = (int64_t*)tracy_malloc( sizeof( int64_t ) * m_queryCount ); + } + +#if defined VK_EXT_host_query_reset + /** + * This alternative constructor does not use command buffers and instead uses functionality from + * VK_EXT_host_query_reset (core with 1.2 and non-optional) and VK_EXT_calibrated_timestamps. This requires + * the physical device to have another time domain apart from DEVICE to be calibrateable. + */ +#if defined TRACY_VK_USE_SYMBOL_TABLE + VkCtx( VkInstance instance, VkPhysicalDevice physdev, VkDevice device, PFN_vkGetInstanceProcAddr instanceProcAddr, PFN_vkGetDeviceProcAddr deviceProcAddr ) +#else + VkCtx( VkPhysicalDevice physdev, VkDevice device, PFN_vkResetQueryPoolEXT vkResetQueryPool, PFN_vkGetPhysicalDeviceCalibrateableTimeDomainsEXT vkGetPhysicalDeviceCalibrateableTimeDomainsEXT, PFN_vkGetCalibratedTimestampsEXT vkGetCalibratedTimestampsEXT ) +#endif + : m_device( device ) + , m_timeDomain( VK_TIME_DOMAIN_DEVICE_EXT ) + , m_context( GetGpuCtxCounter().fetch_add(1, std::memory_order_relaxed) ) + , m_head( 0 ) + , m_tail( 0 ) + , m_oldCnt( 0 ) + , m_queryCount( QueryCount ) +#if !defined TRACY_VK_USE_SYMBOL_TABLE + , m_vkGetCalibratedTimestampsEXT( vkGetCalibratedTimestampsEXT ) +#endif + { + assert( m_context != 255); + +#if defined TRACY_VK_USE_SYMBOL_TABLE + PopulateSymbolTable(instance, instanceProcAddr, deviceProcAddr); + m_vkGetCalibratedTimestampsEXT = m_symbols.vkGetCalibratedTimestampsEXT; +#endif + + assert( VK_FUNCTION_WRAPPER( vkResetQueryPool ) != nullptr ); + assert( VK_FUNCTION_WRAPPER( vkGetPhysicalDeviceCalibrateableTimeDomainsEXT ) != nullptr ); + assert( VK_FUNCTION_WRAPPER( vkGetCalibratedTimestampsEXT ) != nullptr ); + + FindAvailableTimeDomains( physdev, VK_FUNCTION_WRAPPER( vkGetPhysicalDeviceCalibrateableTimeDomainsEXT ) ); + + // We require a host time domain to be available to properly calibrate. + FindCalibratedTimestampDeviation(); + int64_t tgpu; + Calibrate( device, m_prevCalibration, tgpu ); + int64_t tcpu = Profiler::GetTime(); + + CreateQueryPool(); + VK_FUNCTION_WRAPPER( vkResetQueryPool( device, m_query, 0, m_queryCount ) ); + + WriteInitialItem( physdev, tcpu, tgpu ); + + m_res = (int64_t*)tracy_malloc( sizeof( int64_t ) * m_queryCount ); + } +#endif + + ~VkCtx() + { + tracy_free( m_res ); + VK_FUNCTION_WRAPPER( vkDestroyQueryPool( m_device, m_query, nullptr ) ); + } + + void Name( const char* name, uint16_t len ) + { + auto ptr = (char*)tracy_malloc( len ); + memcpy( ptr, name, len ); + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuContextName ); + MemWrite( &item->gpuContextNameFat.context, m_context ); + MemWrite( &item->gpuContextNameFat.ptr, (uint64_t)ptr ); + MemWrite( &item->gpuContextNameFat.size, len ); +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + + void Collect( VkCommandBuffer cmdbuf ) + { + ZoneScopedC( Color::Red4 ); + + const uint64_t head = m_head.load(std::memory_order_relaxed); + if( m_tail == head ) return; + +#ifdef TRACY_ON_DEMAND + if( !GetProfiler().IsConnected() ) + { + VK_FUNCTION_WRAPPER( vkCmdResetQueryPool( cmdbuf, m_query, 0, m_queryCount ) ); + m_tail = head; + m_oldCnt = 0; + int64_t tgpu; + if( m_timeDomain != VK_TIME_DOMAIN_DEVICE_EXT ) Calibrate( m_device, m_prevCalibration, tgpu ); + return; + } +#endif + assert( head > m_tail ); + + const unsigned int wrappedTail = (unsigned int)( m_tail % m_queryCount ); + + unsigned int cnt; + if( m_oldCnt != 0 ) + { + cnt = m_oldCnt; + m_oldCnt = 0; + } + else + { + cnt = (unsigned int)( head - m_tail ); + assert( cnt <= m_queryCount ); + if( wrappedTail + cnt > m_queryCount ) + { + cnt = m_queryCount - wrappedTail; + } + } + + + if( VK_FUNCTION_WRAPPER( vkGetQueryPoolResults( m_device, m_query, wrappedTail, cnt, sizeof( int64_t ) * m_queryCount, m_res, sizeof( int64_t ), VK_QUERY_RESULT_64_BIT ) == VK_NOT_READY ) ) + { + m_oldCnt = cnt; + return; + } + + for( unsigned int idx=0; idxhdr.type, QueueType::GpuTime ); + MemWrite( &item->gpuTime.gpuTime, m_res[idx] ); + MemWrite( &item->gpuTime.queryId, uint16_t( wrappedTail + idx ) ); + MemWrite( &item->gpuTime.context, m_context ); + Profiler::QueueSerialFinish(); + } + + if( m_timeDomain != VK_TIME_DOMAIN_DEVICE_EXT ) + { + int64_t tgpu, tcpu; + Calibrate( m_device, tcpu, tgpu ); + const auto refCpu = Profiler::GetTime(); + const auto delta = tcpu - m_prevCalibration; + if( delta > 0 ) + { + m_prevCalibration = tcpu; + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuCalibration ); + MemWrite( &item->gpuCalibration.gpuTime, tgpu ); + MemWrite( &item->gpuCalibration.cpuTime, refCpu ); + MemWrite( &item->gpuCalibration.cpuDelta, delta ); + MemWrite( &item->gpuCalibration.context, m_context ); + Profiler::QueueSerialFinish(); + } + } + + VK_FUNCTION_WRAPPER( vkCmdResetQueryPool( cmdbuf, m_query, wrappedTail, cnt ) ); + + m_tail += cnt; + } + + tracy_force_inline unsigned int NextQueryId() + { + const uint64_t id = m_head.fetch_add(1, std::memory_order_relaxed); + return id % m_queryCount; + } + + tracy_force_inline uint8_t GetId() const + { + return m_context; + } + + tracy_force_inline VkQueryPool GetQueryPool() const + { + return m_query; + } + +private: + tracy_force_inline void Calibrate( VkDevice device, int64_t& tCpu, int64_t& tGpu ) + { + assert( m_timeDomain != VK_TIME_DOMAIN_DEVICE_EXT ); + VkCalibratedTimestampInfoEXT spec[2] = { + { VK_STRUCTURE_TYPE_CALIBRATED_TIMESTAMP_INFO_EXT, nullptr, VK_TIME_DOMAIN_DEVICE_EXT }, + { VK_STRUCTURE_TYPE_CALIBRATED_TIMESTAMP_INFO_EXT, nullptr, m_timeDomain }, + }; + uint64_t ts[2]; + uint64_t deviation; + do + { + m_vkGetCalibratedTimestampsEXT( device, 2, spec, ts, &deviation ); + } + while( deviation > m_deviation ); + +#if defined _WIN32 + tGpu = ts[0]; + tCpu = ts[1] * m_qpcToNs; +#elif defined __linux__ && defined CLOCK_MONOTONIC_RAW + tGpu = ts[0]; + tCpu = ts[1]; +#else + assert( false ); +#endif + } + + tracy_force_inline void CreateQueryPool() + { + VkQueryPoolCreateInfo poolInfo = {}; + poolInfo.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO; + poolInfo.queryCount = m_queryCount; + poolInfo.queryType = VK_QUERY_TYPE_TIMESTAMP; + while ( VK_FUNCTION_WRAPPER( vkCreateQueryPool( m_device, &poolInfo, nullptr, &m_query ) != VK_SUCCESS ) ) + { + m_queryCount /= 2; + poolInfo.queryCount = m_queryCount; + } + } + + tracy_force_inline void FindAvailableTimeDomains( VkPhysicalDevice physicalDevice, PFN_vkGetPhysicalDeviceCalibrateableTimeDomainsEXT _vkGetPhysicalDeviceCalibrateableTimeDomainsEXT ) + { + uint32_t num; + _vkGetPhysicalDeviceCalibrateableTimeDomainsEXT( physicalDevice, &num, nullptr ); + if(num > 4) num = 4; + VkTimeDomainEXT data[4]; + _vkGetPhysicalDeviceCalibrateableTimeDomainsEXT( physicalDevice, &num, data ); + VkTimeDomainEXT supportedDomain = (VkTimeDomainEXT)-1; +#if defined _WIN32 + supportedDomain = VK_TIME_DOMAIN_QUERY_PERFORMANCE_COUNTER_EXT; +#elif defined __linux__ && defined CLOCK_MONOTONIC_RAW + supportedDomain = VK_TIME_DOMAIN_CLOCK_MONOTONIC_RAW_EXT; +#endif + for( uint32_t i=0; i deviation[i] ) { + minDeviation = deviation[i]; + } + } + m_deviation = minDeviation * 3 / 2; + +#if defined _WIN32 + m_qpcToNs = int64_t( 1000000000. / GetFrequencyQpc() ); +#endif + } + + tracy_force_inline void WriteInitialItem( VkPhysicalDevice physdev, int64_t tcpu, int64_t tgpu ) + { + uint8_t flags = 0; + if( m_timeDomain != VK_TIME_DOMAIN_DEVICE_EXT ) flags |= GpuContextCalibration; + + VkPhysicalDeviceProperties prop; + VK_FUNCTION_WRAPPER( vkGetPhysicalDeviceProperties( physdev, &prop ) ); + const float period = prop.limits.timestampPeriod; + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuNewContext ); + MemWrite( &item->gpuNewContext.cpuTime, tcpu ); + MemWrite( &item->gpuNewContext.gpuTime, tgpu ); + memset( &item->gpuNewContext.thread, 0, sizeof( item->gpuNewContext.thread ) ); + MemWrite( &item->gpuNewContext.period, period ); + MemWrite( &item->gpuNewContext.context, m_context ); + MemWrite( &item->gpuNewContext.flags, flags ); + MemWrite( &item->gpuNewContext.type, GpuContextType::Vulkan ); + +#ifdef TRACY_ON_DEMAND + GetProfiler().DeferItem( *item ); +#endif + Profiler::QueueSerialFinish(); + } + +#if defined TRACY_VK_USE_SYMBOL_TABLE + void PopulateSymbolTable( VkInstance instance, PFN_vkGetInstanceProcAddr instanceProcAddr, PFN_vkGetDeviceProcAddr deviceProcAddr ) + { +#define VK_GET_DEVICE_SYMBOL( name ) \ + (PFN_##name)deviceProcAddr( m_device, #name ); +#define VK_LOAD_DEVICE_SYMBOL( name ) \ + m_symbols.name = VK_GET_DEVICE_SYMBOL( name ); +#define VK_GET_INSTANCE_SYMBOL( name ) \ + (PFN_##name)instanceProcAddr( instance, #name ); +#define VK_LOAD_INSTANCE_SYMBOL( name ) \ + m_symbols.name = VK_GET_INSTANCE_SYMBOL( name ); + + LoadVkDeviceCoreSymbols( VK_LOAD_DEVICE_SYMBOL ) + LoadVkDeviceExtensionSymbols( VK_LOAD_DEVICE_SYMBOL ) + LoadVkInstanceExtensionSymbols( VK_LOAD_INSTANCE_SYMBOL ) + LoadVkInstanceCoreSymbols( VK_LOAD_INSTANCE_SYMBOL ) +#undef VK_GET_DEVICE_SYMBOL +#undef VK_LOAD_DEVICE_SYMBOL +#undef VK_GET_INSTANCE_SYMBOL +#undef VK_LOAD_INSTANCE_SYMBOL + } +#endif + + VkDevice m_device; + VkQueryPool m_query; + VkTimeDomainEXT m_timeDomain; +#if defined TRACY_VK_USE_SYMBOL_TABLE + VkSymbolTable m_symbols; +#endif + uint64_t m_deviation; + int64_t m_qpcToNs; + int64_t m_prevCalibration; + uint8_t m_context; + + std::atomic m_head; + uint64_t m_tail; + unsigned int m_oldCnt; + unsigned int m_queryCount; + + int64_t* m_res; + + PFN_vkGetCalibratedTimestampsEXT m_vkGetCalibratedTimestampsEXT; +}; + +class VkCtxScope +{ +public: + tracy_force_inline VkCtxScope( VkCtx* ctx, const SourceLocationData* srcloc, VkCommandBuffer cmdbuf, bool is_active ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; + m_cmdbuf = cmdbuf; + m_ctx = ctx; + + const auto queryId = ctx->NextQueryId(); + CONTEXT_VK_FUNCTION_WRAPPER( vkCmdWriteTimestamp( cmdbuf, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, ctx->m_query, queryId ) ); + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuZoneBeginSerial ); + MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() ); + MemWrite( &item->gpuZoneBegin.srcloc, (uint64_t)srcloc ); + MemWrite( &item->gpuZoneBegin.thread, GetThreadHandle() ); + MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneBegin.context, ctx->GetId() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline VkCtxScope( VkCtx* ctx, const SourceLocationData* srcloc, VkCommandBuffer cmdbuf, int depth, bool is_active ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; + m_cmdbuf = cmdbuf; + m_ctx = ctx; + + const auto queryId = ctx->NextQueryId(); + CONTEXT_VK_FUNCTION_WRAPPER( vkCmdWriteTimestamp( cmdbuf, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, ctx->m_query, queryId ) ); + + auto item = Profiler::QueueSerialCallstack( Callstack( depth ) ); + MemWrite( &item->hdr.type, QueueType::GpuZoneBeginCallstackSerial ); + MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() ); + MemWrite( &item->gpuZoneBegin.srcloc, (uint64_t)srcloc ); + MemWrite( &item->gpuZoneBegin.thread, GetThreadHandle() ); + MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneBegin.context, ctx->GetId() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline VkCtxScope( VkCtx* ctx, uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, VkCommandBuffer cmdbuf, bool is_active ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; + m_cmdbuf = cmdbuf; + m_ctx = ctx; + + const auto queryId = ctx->NextQueryId(); + CONTEXT_VK_FUNCTION_WRAPPER( vkCmdWriteTimestamp( cmdbuf, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, ctx->m_query, queryId ) ); + + const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz ); + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuZoneBeginAllocSrcLocSerial ); + MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() ); + MemWrite( &item->gpuZoneBegin.srcloc, srcloc ); + MemWrite( &item->gpuZoneBegin.thread, GetThreadHandle() ); + MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneBegin.context, ctx->GetId() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline VkCtxScope( VkCtx* ctx, uint32_t line, const char* source, size_t sourceSz, const char* function, size_t functionSz, const char* name, size_t nameSz, VkCommandBuffer cmdbuf, int depth, bool is_active ) +#ifdef TRACY_ON_DEMAND + : m_active( is_active && GetProfiler().IsConnected() ) +#else + : m_active( is_active ) +#endif + { + if( !m_active ) return; + m_cmdbuf = cmdbuf; + m_ctx = ctx; + + const auto queryId = ctx->NextQueryId(); + CONTEXT_VK_FUNCTION_WRAPPER( vkCmdWriteTimestamp( cmdbuf, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, ctx->m_query, queryId ) ); + + const auto srcloc = Profiler::AllocSourceLocation( line, source, sourceSz, function, functionSz, name, nameSz ); + auto item = Profiler::QueueSerialCallstack( Callstack( depth ) ); + MemWrite( &item->hdr.type, QueueType::GpuZoneBeginAllocSrcLocCallstackSerial ); + MemWrite( &item->gpuZoneBegin.cpuTime, Profiler::GetTime() ); + MemWrite( &item->gpuZoneBegin.srcloc, srcloc ); + MemWrite( &item->gpuZoneBegin.thread, GetThreadHandle() ); + MemWrite( &item->gpuZoneBegin.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneBegin.context, ctx->GetId() ); + Profiler::QueueSerialFinish(); + } + + tracy_force_inline ~VkCtxScope() + { + if( !m_active ) return; + + const auto queryId = m_ctx->NextQueryId(); + CONTEXT_VK_FUNCTION_WRAPPER( vkCmdWriteTimestamp( m_cmdbuf, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, m_ctx->m_query, queryId ) ); + + auto item = Profiler::QueueSerial(); + MemWrite( &item->hdr.type, QueueType::GpuZoneEndSerial ); + MemWrite( &item->gpuZoneEnd.cpuTime, Profiler::GetTime() ); + MemWrite( &item->gpuZoneEnd.thread, GetThreadHandle() ); + MemWrite( &item->gpuZoneEnd.queryId, uint16_t( queryId ) ); + MemWrite( &item->gpuZoneEnd.context, m_ctx->GetId() ); + Profiler::QueueSerialFinish(); + } + +private: + const bool m_active; + + VkCommandBuffer m_cmdbuf; + VkCtx* m_ctx; +}; + +#if defined TRACY_VK_USE_SYMBOL_TABLE +static inline VkCtx* CreateVkContext( VkInstance instance, VkPhysicalDevice physdev, VkDevice device, VkQueue queue, VkCommandBuffer cmdbuf, PFN_vkGetInstanceProcAddr instanceProcAddr, PFN_vkGetDeviceProcAddr getDeviceProcAddr, bool calibrated = false ) +#else +static inline VkCtx* CreateVkContext( VkPhysicalDevice physdev, VkDevice device, VkQueue queue, VkCommandBuffer cmdbuf, PFN_vkGetPhysicalDeviceCalibrateableTimeDomainsEXT gpdctd, PFN_vkGetCalibratedTimestampsEXT gct ) +#endif +{ + auto ctx = (VkCtx*)tracy_malloc( sizeof( VkCtx ) ); +#if defined TRACY_VK_USE_SYMBOL_TABLE + new(ctx) VkCtx( instance, physdev, device, queue, cmdbuf, instanceProcAddr, getDeviceProcAddr, calibrated ); +#else + new(ctx) VkCtx( physdev, device, queue, cmdbuf, gpdctd, gct ); +#endif + return ctx; +} + +#if defined VK_EXT_host_query_reset +#if defined TRACY_VK_USE_SYMBOL_TABLE +static inline VkCtx* CreateVkContext( VkInstance instance, VkPhysicalDevice physdev, VkDevice device, PFN_vkGetInstanceProcAddr instanceProcAddr, PFN_vkGetDeviceProcAddr getDeviceProcAddr ) +#else +static inline VkCtx* CreateVkContext( VkPhysicalDevice physdev, VkDevice device, PFN_vkResetQueryPoolEXT qpreset, PFN_vkGetPhysicalDeviceCalibrateableTimeDomainsEXT gpdctd, PFN_vkGetCalibratedTimestampsEXT gct ) +#endif +{ + auto ctx = (VkCtx*)tracy_malloc( sizeof( VkCtx ) ); +#if defined TRACY_VK_USE_SYMBOL_TABLE + new(ctx) VkCtx( instance, physdev, device, instanceProcAddr, getDeviceProcAddr ); +#else + new(ctx) VkCtx( physdev, device, qpreset, gpdctd, gct ); +#endif + return ctx; +} +#endif + +static inline void DestroyVkContext( VkCtx* ctx ) +{ + ctx->~VkCtx(); + tracy_free( ctx ); +} + +} + +using TracyVkCtx = tracy::VkCtx*; + +#if defined TRACY_VK_USE_SYMBOL_TABLE +#define TracyVkContext( instance, physdev, device, queue, cmdbuf, instanceProcAddr, deviceProcAddr ) tracy::CreateVkContext( instance, physdev, device, queue, cmdbuf, instanceProcAddr, deviceProcAddr ); +#else +#define TracyVkContext( physdev, device, queue, cmdbuf ) tracy::CreateVkContext( physdev, device, queue, cmdbuf, nullptr, nullptr ); +#endif +#if defined TRACY_VK_USE_SYMBOL_TABLE +#define TracyVkContextCalibrated( instance, physdev, device, queue, cmdbuf, instanceProcAddr, deviceProcAddr ) tracy::CreateVkContext( instance, physdev, device, queue, cmdbuf, instanceProcAddr, deviceProcAddr, true ); +#else +#define TracyVkContextCalibrated( physdev, device, queue, cmdbuf, gpdctd, gct ) tracy::CreateVkContext( physdev, device, queue, cmdbuf, gpdctd, gct ); +#endif +#if defined VK_EXT_host_query_reset +#if defined TRACY_VK_USE_SYMBOL_TABLE +#define TracyVkContextHostCalibrated( instance, physdev, device, instanceProcAddr, deviceProcAddr ) tracy::CreateVkContext( instance, physdev, device, instanceProcAddr, deviceProcAddr ); +#else +#define TracyVkContextHostCalibrated( physdev, device, qpreset, gpdctd, gct ) tracy::CreateVkContext( physdev, device, qpreset, gpdctd, gct ); +#endif +#endif +#define TracyVkDestroy( ctx ) tracy::DestroyVkContext( ctx ); +#define TracyVkContextName( ctx, name, size ) ctx->Name( name, size ); +#if defined TRACY_HAS_CALLSTACK && defined TRACY_CALLSTACK +# define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, TRACY_CALLSTACK, active ); +# define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, TRACY_CALLSTACK, active ); +# define TracyVkZone( ctx, cmdbuf, name ) TracyVkNamedZoneS( ctx, ___tracy_gpu_zone, cmdbuf, name, TRACY_CALLSTACK, true ) +# define TracyVkZoneC( ctx, cmdbuf, name, color ) TracyVkNamedZoneCS( ctx, ___tracy_gpu_zone, cmdbuf, name, color, TRACY_CALLSTACK, true ) +# define TracyVkZoneTransient( ctx, varname, cmdbuf, name, active ) TracyVkZoneTransientS( ctx, varname, cmdbuf, name, TRACY_CALLSTACK, active ) +#else +# define TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, active ); +# define TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, active ); +# define TracyVkZone( ctx, cmdbuf, name ) TracyVkNamedZone( ctx, ___tracy_gpu_zone, cmdbuf, name, true ) +# define TracyVkZoneC( ctx, cmdbuf, name, color ) TracyVkNamedZoneC( ctx, ___tracy_gpu_zone, cmdbuf, name, color, true ) +# define TracyVkZoneTransient( ctx, varname, cmdbuf, name, active ) tracy::VkCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), cmdbuf, active ); +#endif +#define TracyVkCollect( ctx, cmdbuf ) ctx->Collect( cmdbuf ); + +#ifdef TRACY_HAS_CALLSTACK +# define TracyVkNamedZoneS( ctx, varname, cmdbuf, name, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, 0 }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, depth, active ); +# define TracyVkNamedZoneCS( ctx, varname, cmdbuf, name, color, depth, active ) static constexpr tracy::SourceLocationData TracyConcat(__tracy_gpu_source_location,TracyLine) { name, TracyFunction, TracyFile, (uint32_t)TracyLine, color }; tracy::VkCtxScope varname( ctx, &TracyConcat(__tracy_gpu_source_location,TracyLine), cmdbuf, depth, active ); +# define TracyVkZoneS( ctx, cmdbuf, name, depth ) TracyVkNamedZoneS( ctx, ___tracy_gpu_zone, cmdbuf, name, depth, true ) +# define TracyVkZoneCS( ctx, cmdbuf, name, color, depth ) TracyVkNamedZoneCS( ctx, ___tracy_gpu_zone, cmdbuf, name, color, depth, true ) +# define TracyVkZoneTransientS( ctx, varname, cmdbuf, name, depth, active ) tracy::VkCtxScope varname( ctx, TracyLine, TracyFile, strlen( TracyFile ), TracyFunction, strlen( TracyFunction ), name, strlen( name ), cmdbuf, depth, active ); +#else +# define TracyVkNamedZoneS( ctx, varname, cmdbuf, name, depth, active ) TracyVkNamedZone( ctx, varname, cmdbuf, name, active ) +# define TracyVkNamedZoneCS( ctx, varname, cmdbuf, name, color, depth, active ) TracyVkNamedZoneC( ctx, varname, cmdbuf, name, color, active ) +# define TracyVkZoneS( ctx, cmdbuf, name, depth ) TracyVkZone( ctx, cmdbuf, name ) +# define TracyVkZoneCS( ctx, cmdbuf, name, color, depth ) TracyVkZoneC( ctx, cmdbuf, name, color ) +# define TracyVkZoneTransientS( ctx, varname, cmdbuf, name, depth, active ) TracyVkZoneTransient( ctx, varname, cmdbuf, name, active ) +#endif + +#endif + +#endif diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt index 461cd57e6..7f220e000 100644 --- a/src/CMakeLists.txt +++ b/src/CMakeLists.txt @@ -9,6 +9,10 @@ if (ENABLE_LTO) set(CMAKE_INTERPROCEDURAL_OPTIMIZATION TRUE) endif() +if (ENABLE_PROFILING) + add_definitions(-DTRACY_ENABLE -DTRACY_ON_DEMAND) +endif() + # Set compilation flags if (MSVC) set(CMAKE_CONFIGURATION_TYPES Debug Release CACHE STRING "" FORCE) diff --git a/src/android/app/src/main/jni/native.cpp b/src/android/app/src/main/jni/native.cpp index c670f7d19..f6ce58000 100644 --- a/src/android/app/src/main/jni/native.cpp +++ b/src/android/app/src/main/jni/native.cpp @@ -26,7 +26,6 @@ #include "common/file_util.h" #include "common/logging/backend.h" #include "common/logging/log.h" -#include "common/microprofile.h" #include "common/scm_rev.h" #include "common/scope_exit.h" #include "common/settings.h" @@ -121,7 +120,6 @@ static void TryShutdown() { Core::System::GetInstance().Shutdown(); window.reset(); InputManager::Shutdown(); - MicroProfileShutdown(); } static bool CheckMicPermission() { @@ -135,8 +133,6 @@ static Core::System::ResultStatus RunCitra(const std::string& filepath) { LOG_INFO(Frontend, "Citra starting..."); - MicroProfileOnThreadCreate("EmuThread"); - if (filepath.empty()) { LOG_CRITICAL(Frontend, "Failed to load ROM: No ROM specified"); return Core::System::ResultStatus::ErrorLoader; diff --git a/src/audio_core/lle/lle.cpp b/src/audio_core/lle/lle.cpp index 91ac5dc73..27f92f509 100644 --- a/src/audio_core/lle/lle.cpp +++ b/src/audio_core/lle/lle.cpp @@ -9,6 +9,7 @@ #include "audio_core/lle/lle.h" #include "common/assert.h" #include "common/bit_field.h" +#include "common/profiling.h" #include "common/swap.h" #include "common/thread.h" #include "core/core.h" @@ -151,6 +152,7 @@ struct DspLle::Impl final { void TeakraThread() { while (true) { + CITRA_PROFILE("Audio", "Teakra"); teakra.Run(TeakraSlice); teakra_slice_barrier.Sync(); if (stop_signal) { @@ -174,6 +176,7 @@ struct DspLle::Impl final { if (multithread) { teakra_slice_barrier.Sync(); } else { + CITRA_PROFILE("Audio", "Teakra"); teakra.Run(TeakraSlice); } } diff --git a/src/citra/citra.cpp b/src/citra/citra.cpp index 8488b64df..22f1cb8a9 100644 --- a/src/citra/citra.cpp +++ b/src/citra/citra.cpp @@ -7,10 +7,6 @@ #include #include #include - -// This needs to be included before getopt.h because the latter #defines symbols used by it -#include "common/microprofile.h" - #include "citra/config.h" #include "citra/emu_window/emu_window_sdl2.h" #ifdef ENABLE_OPENGL @@ -325,9 +321,6 @@ int main(int argc, char** argv) { LocalFree(argv_w); #endif - MicroProfileOnThreadCreate("EmuThread"); - SCOPE_EXIT({ MicroProfileShutdown(); }); - if (filepath.empty()) { LOG_CRITICAL(Frontend, "Failed to load ROM: No ROM specified"); return -1; diff --git a/src/citra_qt/CMakeLists.txt b/src/citra_qt/CMakeLists.txt index b17a143d3..6281c1352 100644 --- a/src/citra_qt/CMakeLists.txt +++ b/src/citra_qt/CMakeLists.txt @@ -112,8 +112,6 @@ add_executable(citra-qt debugger/ipc/recorder.ui debugger/lle_service_modules.cpp debugger/lle_service_modules.h - debugger/profiler.cpp - debugger/profiler.h debugger/registers.cpp debugger/registers.h debugger/registers.ui diff --git a/src/citra_qt/bootmanager.cpp b/src/citra_qt/bootmanager.cpp index 8f63f6713..6d13c599c 100644 --- a/src/citra_qt/bootmanager.cpp +++ b/src/citra_qt/bootmanager.cpp @@ -11,7 +11,6 @@ #include "citra_qt/bootmanager.h" #include "citra_qt/main.h" #include "common/color.h" -#include "common/microprofile.h" #include "common/scm_rev.h" #include "common/settings.h" #include "core/3ds.h" @@ -61,7 +60,6 @@ static GMainWindow* GetMainWindow() { } void EmuThread::run() { - MicroProfileOnThreadCreate("EmuThread"); const auto scope = core_context.Acquire(); if (Settings::values.preload_textures) { @@ -131,10 +129,6 @@ void EmuThread::run() { // Shutdown the core emulation system.Shutdown(); - -#if MICROPROFILE_ENABLED - MicroProfileOnThreadExit(); -#endif } #ifdef ENABLE_OPENGL diff --git a/src/citra_qt/configuration/config.cpp b/src/citra_qt/configuration/config.cpp index 389a3becf..decb86d78 100644 --- a/src/citra_qt/configuration/config.cpp +++ b/src/citra_qt/configuration/config.cpp @@ -803,9 +803,6 @@ void Config::ReadUILayoutValues() { ReadSetting(QStringLiteral("geometryRenderWindow")).toByteArray(); UISettings::values.gamelist_header_state = ReadSetting(QStringLiteral("gameListHeaderState")).toByteArray(); - UISettings::values.microprofile_geometry = - ReadSetting(QStringLiteral("microProfileDialogGeometry")).toByteArray(); - ReadBasicSetting(UISettings::values.microprofile_visible); qt_config->endGroup(); } @@ -1275,9 +1272,6 @@ void Config::SaveUILayoutValues() { WriteSetting(QStringLiteral("state"), UISettings::values.state); WriteSetting(QStringLiteral("geometryRenderWindow"), UISettings::values.renderwindow_geometry); WriteSetting(QStringLiteral("gameListHeaderState"), UISettings::values.gamelist_header_state); - WriteSetting(QStringLiteral("microProfileDialogGeometry"), - UISettings::values.microprofile_geometry); - WriteBasicSetting(UISettings::values.microprofile_visible); qt_config->endGroup(); } diff --git a/src/citra_qt/debugger/profiler.cpp b/src/citra_qt/debugger/profiler.cpp deleted file mode 100644 index dbeade3cd..000000000 --- a/src/citra_qt/debugger/profiler.cpp +++ /dev/null @@ -1,231 +0,0 @@ -// Copyright 2015 Citra Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -#include -#include -#include -#include -#include -#include -#include "citra_qt/debugger/profiler.h" -#include "citra_qt/util/util.h" -#include "common/common_types.h" -#include "common/microprofile.h" - -// Include the implementation of the UI in this file. This isn't in microprofile.cpp because the -// non-Qt frontends don't need it (and don't implement the UI drawing hooks either). -#if MICROPROFILE_ENABLED -#define MICROPROFILEUI_IMPL 1 -#include "common/microprofileui.h" - -class MicroProfileWidget : public QWidget { -public: - MicroProfileWidget(QWidget* parent = nullptr); - -protected: - void paintEvent(QPaintEvent* event) override; - void showEvent(QShowEvent* event) override; - void hideEvent(QHideEvent* event) override; - - void mouseMoveEvent(QMouseEvent* event) override; - void mousePressEvent(QMouseEvent* event) override; - void mouseReleaseEvent(QMouseEvent* event) override; - void wheelEvent(QWheelEvent* event) override; - - void keyPressEvent(QKeyEvent* event) override; - void keyReleaseEvent(QKeyEvent* event) override; - -private: - /// This timer is used to redraw the widget's contents continuously. To save resources, it only - /// runs while the widget is visible. - QTimer update_timer; - /// Scale the coordinate system appropriately when dpi != 96. - qreal x_scale = 1.0, y_scale = 1.0; -}; - -#endif - -MicroProfileDialog::MicroProfileDialog(QWidget* parent) : QWidget(parent, Qt::Dialog) { - setObjectName(QStringLiteral("MicroProfile")); - setWindowTitle(tr("MicroProfile")); - resize(1000, 600); - // Enable the maximize button - setWindowFlags(windowFlags() | Qt::WindowMaximizeButtonHint); - -#if MICROPROFILE_ENABLED - - MicroProfileWidget* widget = new MicroProfileWidget(this); - - QLayout* layout = new QVBoxLayout(this); - layout->setContentsMargins(0, 0, 0, 0); - layout->addWidget(widget); - setLayout(layout); - - // Configure focus so that widget is focusable and the dialog automatically forwards focus to - // it. - setFocusProxy(widget); - widget->setFocusPolicy(Qt::StrongFocus); - widget->setFocus(); -#endif -} - -QAction* MicroProfileDialog::toggleViewAction() { - if (toggle_view_action == nullptr) { - toggle_view_action = new QAction(windowTitle(), this); - toggle_view_action->setCheckable(true); - toggle_view_action->setChecked(isVisible()); - connect(toggle_view_action, &QAction::toggled, this, &MicroProfileDialog::setVisible); - } - - return toggle_view_action; -} - -void MicroProfileDialog::showEvent(QShowEvent* event) { - if (toggle_view_action) { - toggle_view_action->setChecked(isVisible()); - } - QWidget::showEvent(event); -} - -void MicroProfileDialog::hideEvent(QHideEvent* event) { - if (toggle_view_action) { - toggle_view_action->setChecked(isVisible()); - } - QWidget::hideEvent(event); -} - -#if MICROPROFILE_ENABLED - -/// There's no way to pass a user pointer to MicroProfile, so this variable is used to make the -/// QPainter available inside the drawing callbacks. -static QPainter* mp_painter = nullptr; - -MicroProfileWidget::MicroProfileWidget(QWidget* parent) : QWidget(parent) { - // Send mouse motion events even when not dragging. - setMouseTracking(true); - - MicroProfileSetDisplayMode(1); // Timers screen - MicroProfileInitUI(); - - connect(&update_timer, &QTimer::timeout, this, qOverload<>(&MicroProfileWidget::update)); -} - -void MicroProfileWidget::paintEvent([[maybe_unused]] QPaintEvent* event) { - QPainter painter(this); - - // The units used by Microprofile for drawing are based in pixels on a 96 dpi display. - x_scale = qreal(painter.device()->logicalDpiX()) / 96.0; - y_scale = qreal(painter.device()->logicalDpiY()) / 96.0; - painter.scale(x_scale, y_scale); - - painter.setBackground(Qt::black); - painter.eraseRect(rect()); - - QFont font = GetMonospaceFont(); - font.setPixelSize(MICROPROFILE_TEXT_HEIGHT); - painter.setFont(font); - - mp_painter = &painter; - MicroProfileDraw(rect().width() / x_scale, rect().height() / y_scale); - mp_painter = nullptr; -} - -void MicroProfileWidget::showEvent(QShowEvent* event) { - update_timer.start(15); // ~60 Hz - QWidget::showEvent(event); -} - -void MicroProfileWidget::hideEvent(QHideEvent* event) { - update_timer.stop(); - QWidget::hideEvent(event); -} - -void MicroProfileWidget::mouseMoveEvent(QMouseEvent* event) { - const auto point = event->position().toPoint(); - MicroProfileMousePosition(point.x() / x_scale, point.y() / y_scale, 0); - event->accept(); -} - -void MicroProfileWidget::mousePressEvent(QMouseEvent* event) { - const auto point = event->position().toPoint(); - MicroProfileMousePosition(point.x() / x_scale, point.y() / y_scale, 0); - MicroProfileMouseButton(event->buttons() & Qt::LeftButton, event->buttons() & Qt::RightButton); - event->accept(); -} - -void MicroProfileWidget::mouseReleaseEvent(QMouseEvent* event) { - const auto point = event->position().toPoint(); - MicroProfileMousePosition(point.x() / x_scale, point.y() / y_scale, 0); - MicroProfileMouseButton(event->buttons() & Qt::LeftButton, event->buttons() & Qt::RightButton); - event->accept(); -} - -void MicroProfileWidget::wheelEvent(QWheelEvent* event) { - const auto point = event->position().toPoint(); - MicroProfileMousePosition(point.x() / x_scale, point.y() / y_scale, - event->angleDelta().y() / 120); - event->accept(); -} - -void MicroProfileWidget::keyPressEvent(QKeyEvent* event) { - if (event->key() == Qt::Key_Control) { - // Inform MicroProfile that the user is holding Ctrl. - MicroProfileModKey(1); - } - QWidget::keyPressEvent(event); -} - -void MicroProfileWidget::keyReleaseEvent(QKeyEvent* event) { - if (event->key() == Qt::Key_Control) { - MicroProfileModKey(0); - } - QWidget::keyReleaseEvent(event); -} - -// These functions are called by MicroProfileDraw to draw the interface elements on the screen. - -void MicroProfileDrawText(int x, int y, u32 hex_color, const char* text, u32 text_length) { - // hex_color does not include an alpha, so it must be assumed to be 255 - mp_painter->setPen(QColor::fromRgb(hex_color)); - - // It's impossible to draw a string using a monospaced font with a fixed width per cell in a - // way that's reliable across different platforms and fonts as far as I (yuriks) can tell, so - // draw each character individually in order to precisely control the text advance. - for (u32 i = 0; i < text_length; ++i) { - // Position the text baseline 1 pixel above the bottom of the text cell, this gives nice - // vertical alignment of text for a wide range of tested fonts. - mp_painter->drawText(x, y + MICROPROFILE_TEXT_HEIGHT - 2, QString{QLatin1Char{text[i]}}); - x += MICROPROFILE_TEXT_WIDTH + 1; - } -} - -void MicroProfileDrawBox(int left, int top, int right, int bottom, u32 hex_color, - MicroProfileBoxType type) { - QColor color = QColor::fromRgba(hex_color); - QBrush brush = color; - if (type == MicroProfileBoxTypeBar) { - QLinearGradient gradient(left, top, left, bottom); - gradient.setColorAt(0.f, color.lighter(125)); - gradient.setColorAt(1.f, color.darker(125)); - brush = gradient; - } - mp_painter->fillRect(left, top, right - left, bottom - top, brush); -} - -void MicroProfileDrawLine2D(u32 vertices_length, float* vertices, u32 hex_color) { - // Temporary vector used to convert between the float array and QPointF. Marked static to reuse - // the allocation across calls. - static std::vector point_buf; - - point_buf.reserve(vertices_length); - for (u32 i = 0; i < vertices_length; ++i) { - point_buf.emplace_back(vertices[i * 2 + 0], vertices[i * 2 + 1]); - } - - // hex_color does not include an alpha, so it must be assumed to be 255 - mp_painter->setPen(QColor::fromRgb(hex_color)); - mp_painter->drawPolyline(point_buf.data(), vertices_length); - point_buf.clear(); -} -#endif diff --git a/src/citra_qt/debugger/profiler.h b/src/citra_qt/debugger/profiler.h deleted file mode 100644 index 43d470aee..000000000 --- a/src/citra_qt/debugger/profiler.h +++ /dev/null @@ -1,25 +0,0 @@ -// Copyright 2015 Citra Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -#pragma once - -#include -#include "common/microprofile.h" - -class MicroProfileDialog : public QWidget { - Q_OBJECT - -public: - explicit MicroProfileDialog(QWidget* parent = nullptr); - - /// Returns a QAction that can be used to toggle visibility of this dialog. - QAction* toggleViewAction(); - -protected: - void showEvent(QShowEvent* event) override; - void hideEvent(QHideEvent* event) override; - -private: - QAction* toggle_view_action = nullptr; -}; diff --git a/src/citra_qt/main.cpp b/src/citra_qt/main.cpp index 807da30e3..10fe16aa3 100644 --- a/src/citra_qt/main.cpp +++ b/src/citra_qt/main.cpp @@ -48,7 +48,6 @@ #include "citra_qt/debugger/graphics/graphics_vertex_shader.h" #include "citra_qt/debugger/ipc/recorder.h" #include "citra_qt/debugger/lle_service_modules.h" -#include "citra_qt/debugger/profiler.h" #include "citra_qt/debugger/registers.h" #include "citra_qt/debugger/wait_tree.h" #include "citra_qt/discord.h" @@ -489,12 +488,6 @@ void GMainWindow::InitializeDebugWidgets() { QMenu* debug_menu = ui->menu_View_Debugging; -#if MICROPROFILE_ENABLED - microProfileDialog = new MicroProfileDialog(this); - microProfileDialog->hide(); - debug_menu->addAction(microProfileDialog->toggleViewAction()); -#endif - registersWidget = new RegistersWidget(system, this); addDockWidget(Qt::RightDockWidgetArea, registersWidget); registersWidget->hide(); @@ -767,10 +760,6 @@ void GMainWindow::RestoreUIState() { restoreGeometry(UISettings::values.geometry); restoreState(UISettings::values.state); render_window->restoreGeometry(UISettings::values.renderwindow_geometry); -#if MICROPROFILE_ENABLED - microProfileDialog->restoreGeometry(UISettings::values.microprofile_geometry); - microProfileDialog->setVisible(UISettings::values.microprofile_visible.GetValue()); -#endif game_list->LoadInterfaceLayout(); @@ -3064,10 +3053,6 @@ void GMainWindow::UpdateUISettings() { UISettings::values.renderwindow_geometry = render_window->saveGeometry(); } UISettings::values.state = saveState(); -#if MICROPROFILE_ENABLED - UISettings::values.microprofile_geometry = microProfileDialog->saveGeometry(); - UISettings::values.microprofile_visible = microProfileDialog->isVisible(); -#endif UISettings::values.single_window_mode = ui->action_Single_Window_Mode->isChecked(); UISettings::values.fullscreen = ui->action_Fullscreen->isChecked(); UISettings::values.display_titlebar = ui->action_Display_Dock_Widget_Headers->isChecked(); @@ -3179,8 +3164,6 @@ static Qt::HighDpiScaleFactorRoundingPolicy GetHighDpiRoundingPolicy() { int main(int argc, char* argv[]) { Common::DetachedTasks detached_tasks; - MicroProfileOnThreadCreate("Frontend"); - SCOPE_EXIT({ MicroProfileShutdown(); }); // Init settings params QCoreApplication::setOrganizationName(QStringLiteral("Citra team")); diff --git a/src/citra_qt/main.h b/src/citra_qt/main.h index b64bf1b32..299b21435 100644 --- a/src/citra_qt/main.h +++ b/src/citra_qt/main.h @@ -38,7 +38,6 @@ class GRenderWindow; class IPCRecorderWidget; class LLEServiceModulesWidget; class LoadingScreen; -class MicroProfileDialog; class MultiplayerState; class ProfilerWidget; class QFileOpenEvent; @@ -349,7 +348,6 @@ private: // Debugger panes ProfilerWidget* profilerWidget; - MicroProfileDialog* microProfileDialog; RegistersWidget* registersWidget; GPUCommandStreamWidget* graphicsWidget; GPUCommandListWidget* graphicsCommandsWidget; diff --git a/src/citra_qt/uisettings.h b/src/citra_qt/uisettings.h index 8e671f49f..3869f5c5f 100644 --- a/src/citra_qt/uisettings.h +++ b/src/citra_qt/uisettings.h @@ -67,9 +67,6 @@ struct Values { QByteArray gamelist_header_state; - QByteArray microprofile_geometry; - Settings::Setting microprofile_visible{false, "microProfileDialogVisible"}; - Settings::Setting single_window_mode{true, "singleWindowMode"}; Settings::Setting fullscreen{false, "fullscreen"}; Settings::Setting display_titlebar{true, "displayTitleBars"}; diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt index f8c953a60..5b5e159fe 100644 --- a/src/common/CMakeLists.txt +++ b/src/common/CMakeLists.txt @@ -104,9 +104,6 @@ add_library(citra_common STATIC memory_detect.h memory_ref.h memory_ref.cpp - microprofile.cpp - microprofile.h - microprofileui.h param_package.cpp param_package.h polyfill_thread.h @@ -150,6 +147,7 @@ add_library(citra_common STATIC x64/xbyak_util.h zstd_compression.cpp zstd_compression.h + profiling.h ) if (UNIX AND NOT APPLE) @@ -184,9 +182,14 @@ endif() create_target_directory_groups(citra_common) -target_link_libraries(citra_common PUBLIC fmt library-headers microprofile Boost::boost Boost::serialization Boost::iostreams) +target_link_libraries(citra_common PUBLIC fmt library-headers Boost::boost Boost::serialization Boost::iostreams) target_link_libraries(citra_common PRIVATE zstd) +if (ENABLE_PROFILING) + target_link_libraries(citra_common PUBLIC tracy) + target_compile_definitions(citra_common PUBLIC -DENABLE_PROFILING) +endif() + if ("x86_64" IN_LIST ARCHITECTURE) target_link_libraries(citra_common PRIVATE xbyak) endif() diff --git a/src/common/microprofile.cpp b/src/common/microprofile.cpp deleted file mode 100644 index ee25dd37f..000000000 --- a/src/common/microprofile.cpp +++ /dev/null @@ -1,7 +0,0 @@ -// Copyright 2015 Citra Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -// Includes the MicroProfile implementation in this file for compilation -#define MICROPROFILE_IMPL 1 -#include "common/microprofile.h" diff --git a/src/common/microprofile.h b/src/common/microprofile.h deleted file mode 100644 index 027a3ad94..000000000 --- a/src/common/microprofile.h +++ /dev/null @@ -1,25 +0,0 @@ -// Copyright 2015 Citra Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -#pragma once - -// Uncomment this to disable microprofile. This will get you cleaner profiles when using -// external sampling profilers like "Very Sleepy", and will improve performance somewhat. -// #define MICROPROFILE_ENABLED 0 - -// Customized Citra settings. -// This file wraps the MicroProfile header so that these are consistent everywhere. -#define MICROPROFILE_WEBSERVER 0 -#define MICROPROFILE_GPU_TIMERS 0 // TODO: Implement timer queries when we upgrade to OpenGL 3.3 -#define MICROPROFILE_CONTEXT_SWITCH_TRACE 0 -#define MICROPROFILE_PER_THREAD_BUFFER_SIZE (2048 << 13) // 16 MB - -#ifdef _WIN32 -// This isn't defined by the standard library in MSVC2015 -typedef void* HANDLE; -#endif - -#include - -#define MP_RGB(r, g, b) ((r) << 16 | (g) << 8 | (b) << 0) diff --git a/src/common/microprofileui.h b/src/common/microprofileui.h deleted file mode 100644 index 41abe6b75..000000000 --- a/src/common/microprofileui.h +++ /dev/null @@ -1,19 +0,0 @@ -// Copyright 2015 Citra Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -#pragma once - -#include "common/microprofile.h" - -// Customized Citra settings. -// This file wraps the MicroProfile header so that these are consistent everywhere. -#define MICROPROFILE_TEXT_WIDTH 6 -#define MICROPROFILE_TEXT_HEIGHT 12 -#define MICROPROFILE_HELP_ALT "Right-Click" -#define MICROPROFILE_HELP_MOD "Ctrl" - -// This isn't included by microprofileui.h :( -#include // For std::abs - -#include diff --git a/src/common/profiling.h b/src/common/profiling.h new file mode 100644 index 000000000..01530cefc --- /dev/null +++ b/src/common/profiling.h @@ -0,0 +1,30 @@ +// Copyright 2024 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#ifdef ENABLE_PROFILING +#include +#include "common/scope_exit.h" + +#define CITRA_PROFILE(scope, text) \ + ZoneScopedN(text); \ + ZoneText(scope, std::string::traits_type::length(scope)); + +#define CITRA_SCOPED_FRAME(text) \ + constexpr const char* CONCAT2(FrameTitle, __LINE__) = text; \ + detail::ScopeHelper([&]() { FrameMarkStart(CONCAT2(FrameTitle, __LINE__)); }, \ + [&]() { FrameMarkEnd(CONCAT2(FrameTitle, __LINE__)); }) + +#define CITRA_FRAME_BEGIN(text) FrameMarkStart(text) + +#define CITRA_FRAME_END(text) FrameMarkEnd(text) +#else + +#define CITRA_PROFILE(scope, text) +#define CITRA_SCOPED_FRAME(text) +#define CITRA_FRAME_BEGIN(text) +#define CITRA_FRAME_END(text) + +#endif diff --git a/src/common/scope_exit.h b/src/common/scope_exit.h index 1176a72b1..7bbd7cc94 100644 --- a/src/common/scope_exit.h +++ b/src/common/scope_exit.h @@ -9,19 +9,16 @@ namespace detail { template -struct ScopeExitHelper { - explicit ScopeExitHelper(Func&& func) : func(std::move(func)) {} - ~ScopeExitHelper() { - func(); +struct ScopeHelper { + explicit ScopeHelper(auto&& enter_func, Func&& exit_func) : exit_func(std::move(exit_func)) { + enter_func(); + } + ~ScopeHelper() { + exit_func(); } - Func func; + Func exit_func; }; - -template -ScopeExitHelper ScopeExit(Func&& func) { - return ScopeExitHelper(std::forward(func)); -} } // namespace detail /** @@ -41,4 +38,5 @@ ScopeExitHelper ScopeExit(Func&& func) { * } * \endcode */ -#define SCOPE_EXIT(body) auto CONCAT2(scope_exit_helper_, __LINE__) = detail::ScopeExit([&]() body) +#define SCOPE_EXIT(body) \ + auto CONCAT2(scope_exit_helper_, __LINE__) = detail::ScopeHelper([]() {}, [&]() body) diff --git a/src/core/arm/dynarmic/arm_dynarmic.cpp b/src/core/arm/dynarmic/arm_dynarmic.cpp index dbf6ea71c..7f55da913 100644 --- a/src/core/arm/dynarmic/arm_dynarmic.cpp +++ b/src/core/arm/dynarmic/arm_dynarmic.cpp @@ -6,7 +6,7 @@ #include #include #include "common/assert.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "core/arm/dynarmic/arm_dynarmic.h" #include "core/arm/dynarmic/arm_dynarmic_cp15.h" #include "core/arm/dynarmic/arm_exclusive_monitor.h" @@ -130,11 +130,9 @@ ARM_Dynarmic::ARM_Dynarmic(Core::System& system_, Memory::MemorySystem& memory_, ARM_Dynarmic::~ARM_Dynarmic() = default; -MICROPROFILE_DEFINE(ARM_Jit, "ARM JIT", "ARM JIT", MP_RGB(255, 64, 64)); - void ARM_Dynarmic::Run() { ASSERT(memory.GetCurrentPageTable() == current_page_table); - MICROPROFILE_SCOPE(ARM_Jit); + CITRA_PROFILE("Dynarmic", "ARM JIT"); jit->Run(); } diff --git a/src/core/arm/dyncom/arm_dyncom_interpreter.cpp b/src/core/arm/dyncom/arm_dyncom_interpreter.cpp index 2d333cdaf..aecde5a44 100644 --- a/src/core/arm/dyncom/arm_dyncom_interpreter.cpp +++ b/src/core/arm/dyncom/arm_dyncom_interpreter.cpp @@ -8,7 +8,7 @@ #include #include "common/common_types.h" #include "common/logging/log.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "core/arm/dyncom/arm_dyncom_dec.h" #include "core/arm/dyncom/arm_dyncom_interpreter.h" #include "core/arm/dyncom/arm_dyncom_run.h" @@ -805,8 +805,6 @@ static ThumbDecodeStatus DecodeThumbInstruction(u32 inst, u32 addr, u32* arm_ins enum { KEEP_GOING, FETCH_EXCEPTION }; -MICROPROFILE_DEFINE(DynCom_Decode, "DynCom", "Decode", MP_RGB(255, 64, 64)); - static unsigned int InterpreterTranslateInstruction(const ARMul_State* cpu, const u32 phys_addr, ARM_INST_PTR& inst_base) { u32 inst_size = 4; @@ -840,7 +838,7 @@ static unsigned int InterpreterTranslateInstruction(const ARMul_State* cpu, cons } static int InterpreterTranslateBlock(ARMul_State* cpu, std::size_t& bb_start, u32 addr) { - MICROPROFILE_SCOPE(DynCom_Decode); + CITRA_PROFILE("DynCom", "Decode"); // Decode instruction, get index // Allocate memory and init InsCream @@ -869,7 +867,7 @@ static int InterpreterTranslateBlock(ARMul_State* cpu, std::size_t& bb_start, u3 } static int InterpreterTranslateSingle(ARMul_State* cpu, std::size_t& bb_start, u32 addr) { - MICROPROFILE_SCOPE(DynCom_Decode); + CITRA_PROFILE("DynCom", "Decode"); ARM_INST_PTR inst_base = nullptr; bb_start = trans_cache_buf_top; @@ -913,10 +911,8 @@ static int clz(unsigned int x) { return n; } -MICROPROFILE_DEFINE(DynCom_Execute, "DynCom", "Execute", MP_RGB(255, 0, 0)); - unsigned InterpreterMainLoop(ARMul_State* cpu) { - MICROPROFILE_SCOPE(DynCom_Execute); + CITRA_PROFILE("DynCom", "Execute"); /// Nearest upcoming GDB code execution breakpoint, relative to the last dispatch's address. GDBStub::BreakpointAddress breakpoint_data; diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp index 4cc2c0ad0..710a471ed 100644 --- a/src/core/hle/kernel/svc.cpp +++ b/src/core/hle/kernel/svc.cpp @@ -7,7 +7,7 @@ #include #include "common/archives.h" #include "common/logging/log.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/scm_rev.h" #include "core/arm/arm_interface.h" #include "core/core.h" @@ -2233,10 +2233,8 @@ const SVC::FunctionDef* SVC::GetSVCInfo(u32 func_num) { return &SVC_Table[func_num]; } -MICROPROFILE_DEFINE(Kernel_SVC, "Kernel", "SVC", MP_RGB(70, 200, 70)); - void SVC::CallSVC(u32 immediate) { - MICROPROFILE_SCOPE(Kernel_SVC); + CITRA_PROFILE("Kernel", "SVC"); // Lock the kernel mutex when we enter the kernel HLE. std::scoped_lock lock{kernel.GetHLELock()}; diff --git a/src/core/hw/y2r.cpp b/src/core/hw/y2r.cpp index dab2bdcca..33d2cbce0 100644 --- a/src/core/hw/y2r.cpp +++ b/src/core/hw/y2r.cpp @@ -9,7 +9,7 @@ #include "common/assert.h" #include "common/color.h" #include "common/common_types.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/vector_math.h" #include "core/core.h" #include "core/hle/service/cam/y2r_u.h" @@ -209,8 +209,6 @@ static void WriteTileToOutput(u32* output, const ImageTile& tile, int height, in } } -MICROPROFILE_DEFINE(Y2R_PerformConversion, "Y2R", "PerformConversion", MP_RGB(185, 66, 245)); - /** * Performs a Y2R colorspace conversion. * @@ -262,7 +260,7 @@ MICROPROFILE_DEFINE(Y2R_PerformConversion, "Y2R", "PerformConversion", MP_RGB(18 * so they are believed to be invalid configurations anyway. */ void PerformConversion(Memory::MemorySystem& memory, ConversionConfiguration cvt) { - MICROPROFILE_SCOPE(Y2R_PerformConversion); + CITRA_PROFILE("Y2R", "Perform Conversion"); ASSERT(cvt.input_line_width % 8 == 0); ASSERT(cvt.block_alignment != BlockAlignment::Block8x8 || cvt.input_lines % 8 == 0); diff --git a/src/video_core/custom_textures/custom_tex_manager.cpp b/src/video_core/custom_textures/custom_tex_manager.cpp index df9720941..8d834fee9 100644 --- a/src/video_core/custom_textures/custom_tex_manager.cpp +++ b/src/video_core/custom_textures/custom_tex_manager.cpp @@ -6,7 +6,7 @@ #include "common/file_util.h" #include "common/literals.h" #include "common/memory_detect.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/settings.h" #include "common/string_util.h" #include "common/texture.h" @@ -22,9 +22,6 @@ namespace VideoCore { namespace { -MICROPROFILE_DEFINE(CustomTexManager_TickFrame, "CustomTexManager", "TickFrame", - MP_RGB(54, 16, 32)); - constexpr std::size_t MAX_UPLOADS_PER_TICK = 8; using namespace Common::Literals; @@ -61,7 +58,7 @@ CustomTexManager::CustomTexManager(Core::System& system_) CustomTexManager::~CustomTexManager() = default; void CustomTexManager::TickFrame() { - MICROPROFILE_SCOPE(CustomTexManager_TickFrame); + CITRA_PROFILE("CustomTexManager", "Tick Frame"); if (!textures_loaded) { return; } diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index cc06078b9..18acae138 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -3,7 +3,7 @@ // Refer to the license.txt file included. #include "common/archives.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "core/core.h" #include "core/core_timing.h" #include "core/hle/service/gsp/gsp_gpu.h" @@ -22,9 +22,6 @@ namespace VideoCore { constexpr VAddr VADDR_LCD = 0x1ED02000; constexpr VAddr VADDR_GPU = 0x1EF00000; -MICROPROFILE_DEFINE(GPU_DisplayTransfer, "GPU", "DisplayTransfer", MP_RGB(100, 100, 255)); -MICROPROFILE_DEFINE(GPU_CmdlistProcessing, "GPU", "Cmdlist Processing", MP_RGB(100, 255, 100)); - struct GPU::Impl { Core::Timing& timing; Core::System& system; @@ -230,7 +227,6 @@ void GPU::SetBufferSwap(u32 screen_id, const Service::GSP::FrameBufferInfo& info } if (screen_id == 0) { - MicroProfileFlip(); impl->system.perf_stats->EndGameFrame(); } } @@ -339,7 +335,7 @@ void GPU::SubmitCmdList(u32 index) { return; } - MICROPROFILE_SCOPE(GPU_CmdlistProcessing); + CITRA_PROFILE("GPU", "Command List Processing"); // Forward command list processing to the PICA core. const PAddr addr = config.GetPhysicalAddress(index); @@ -383,8 +379,6 @@ void GPU::MemoryTransfer() { return; } - MICROPROFILE_SCOPE(GPU_DisplayTransfer); - // Notify debugger about the display transfer. if (impl->debug_context) { impl->debug_context->OnEvent(Pica::DebugContext::Event::IncomingDisplayTransfer, nullptr); @@ -392,10 +386,12 @@ void GPU::MemoryTransfer() { // Perform memory transfer if (config.is_texture_copy) { + CITRA_PROFILE("GPU", "Texture Copy"); if (!impl->rasterizer->AccelerateTextureCopy(config)) { impl->sw_blitter->TextureCopy(config); } } else { + CITRA_PROFILE("GPU", "Display Transfer"); if (!impl->rasterizer->AccelerateDisplayTransfer(config)) { impl->sw_blitter->DisplayTransfer(config); } diff --git a/src/video_core/pica/pica_core.cpp b/src/video_core/pica/pica_core.cpp index 6450e669b..2af4a363b 100644 --- a/src/video_core/pica/pica_core.cpp +++ b/src/video_core/pica/pica_core.cpp @@ -4,7 +4,7 @@ #include "common/arch.h" #include "common/archives.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/scope_exit.h" #include "common/settings.h" #include "core/core.h" @@ -17,8 +17,6 @@ namespace Pica { -MICROPROFILE_DEFINE(GPU_Drawing, "GPU", "Drawing", MP_RGB(50, 50, 240)); - using namespace DebugUtils; union CommandHeader { @@ -450,6 +448,8 @@ void PicaCore::SubmitImmediate(u32 value) { } void PicaCore::DrawImmediate() { + CITRA_PROFILE("PicaCore", "Draw Immediate"); + // Compile the vertex shader. shader_engine->SetupBatch(vs_setup, regs.internal.vs.main_offset); @@ -486,7 +486,7 @@ void PicaCore::DrawImmediate() { } void PicaCore::DrawArrays(bool is_indexed) { - MICROPROFILE_SCOPE(GPU_Drawing); + CITRA_PROFILE("PicaCore", "Draw Arrays"); // Track vertex in the debug recorder. if (debug_context) { diff --git a/src/video_core/rasterizer_cache/rasterizer_cache.cpp b/src/video_core/rasterizer_cache/rasterizer_cache.cpp index 3f83d1a3a..1b825e8aa 100644 --- a/src/video_core/rasterizer_cache/rasterizer_cache.cpp +++ b/src/video_core/rasterizer_cache/rasterizer_cache.cpp @@ -2,17 +2,4 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include "common/microprofile.h" - -namespace VideoCore { - -MICROPROFILE_DEFINE(RasterizerCache_CopySurface, "RasterizerCache", "CopySurface", - MP_RGB(128, 192, 64)); -MICROPROFILE_DEFINE(RasterizerCache_UploadSurface, "RasterizerCache", "UploadSurface", - MP_RGB(128, 192, 64)); -MICROPROFILE_DEFINE(RasterizerCache_DownloadSurface, "RasterizerCache", "DownloadSurface", - MP_RGB(128, 192, 64)); -MICROPROFILE_DEFINE(RasterizerCache_Invalidation, "RasterizerCache", "Invalidation", - MP_RGB(128, 64, 192)); - -} // namespace VideoCore +namespace VideoCore {} // namespace VideoCore diff --git a/src/video_core/rasterizer_cache/rasterizer_cache.h b/src/video_core/rasterizer_cache/rasterizer_cache.h index 16027fd93..d3682c6aa 100644 --- a/src/video_core/rasterizer_cache/rasterizer_cache.h +++ b/src/video_core/rasterizer_cache/rasterizer_cache.h @@ -9,7 +9,7 @@ #include #include "common/alignment.h" #include "common/logging/log.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/scope_exit.h" #include "common/settings.h" #include "core/memory.h" @@ -23,11 +23,6 @@ namespace VideoCore { -MICROPROFILE_DECLARE(RasterizerCache_CopySurface); -MICROPROFILE_DECLARE(RasterizerCache_UploadSurface); -MICROPROFILE_DECLARE(RasterizerCache_DownloadSurface); -MICROPROFILE_DECLARE(RasterizerCache_Invalidation); - constexpr auto RangeFromInterval(const auto& map, const auto& interval) { return boost::make_iterator_range(map.equal_range(interval)); } @@ -409,7 +404,7 @@ typename T::Sampler& RasterizerCache::GetSampler( template void RasterizerCache::CopySurface(Surface& src_surface, Surface& dst_surface, SurfaceInterval copy_interval) { - MICROPROFILE_SCOPE(RasterizerCache_CopySurface); + CITRA_PROFILE("RasterizerCache", "Copy Surface"); const PAddr copy_addr = copy_interval.lower(); const SurfaceParams subrect_params = dst_surface.FromInterval(copy_interval); ASSERT(subrect_params.GetInterval() == copy_interval); @@ -992,7 +987,7 @@ void RasterizerCache::ValidateSurface(SurfaceId surface_id, PAddr addr, u32 s template void RasterizerCache::UploadSurface(Surface& surface, SurfaceInterval interval) { - MICROPROFILE_SCOPE(RasterizerCache_UploadSurface); + CITRA_PROFILE("RasterizerCache", "Upload Surface"); const SurfaceParams load_info = surface.FromInterval(interval); ASSERT(load_info.addr >= surface.addr && load_info.end <= surface.end); @@ -1042,7 +1037,7 @@ u64 RasterizerCache::ComputeHash(const SurfaceParams& load_info, std::span bool RasterizerCache::UploadCustomSurface(SurfaceId surface_id, SurfaceInterval interval) { - MICROPROFILE_SCOPE(RasterizerCache_UploadSurface); + CITRA_PROFILE("RasterizerCache", "Upload Custom Surface"); Surface& surface = slot_surfaces[surface_id]; const SurfaceParams load_info = surface.FromInterval(interval); @@ -1090,7 +1085,7 @@ bool RasterizerCache::UploadCustomSurface(SurfaceId surface_id, SurfaceInterv template void RasterizerCache::DownloadSurface(Surface& surface, SurfaceInterval interval) { - MICROPROFILE_SCOPE(RasterizerCache_DownloadSurface); + CITRA_PROFILE("RasterizerCache", "Download Surface"); const SurfaceParams flush_info = surface.FromInterval(interval); const u32 flush_start = boost::icl::first(interval); @@ -1295,6 +1290,8 @@ void RasterizerCache::InvalidateRegion(PAddr addr, u32 size, SurfaceId region region_owner.MarkValid(invalid_interval); } + CITRA_PROFILE("RasterizerCache", "Invalidate Region"); + boost::container::small_vector remove_surfaces; ForEachSurfaceInRegion(addr, size, [&](SurfaceId surface_id, Surface& surface) { if (surface_id == region_owner_id) { diff --git a/src/video_core/renderer_base.cpp b/src/video_core/renderer_base.cpp index 16b59107a..d427aea91 100644 --- a/src/video_core/renderer_base.cpp +++ b/src/video_core/renderer_base.cpp @@ -2,6 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include "common/profiling.h" #include "common/settings.h" #include "core/core.h" #include "core/frontend/emu_window.h" @@ -11,16 +12,20 @@ namespace VideoCore { +[[maybe_unused]] static constexpr const char* EmuThreadFrame = "EmuThread"; + RendererBase::RendererBase(Core::System& system_, Frontend::EmuWindow& window, Frontend::EmuWindow* secondary_window_) - : system{system_}, render_window{window}, secondary_window{secondary_window_} {} + : system{system_}, render_window{window}, secondary_window{secondary_window_} { + CITRA_FRAME_BEGIN(EmuThreadFrame); +} RendererBase::~RendererBase() = default; u32 RendererBase::GetResolutionScaleFactor() { const auto graphics_api = Settings::values.graphics_api.GetValue(); if (graphics_api == Settings::GraphicsAPI::Software) { - // Software renderer always render at native resolution + // Software renderer always renders at native resolution return 1; } @@ -44,11 +49,12 @@ void RendererBase::EndFrame() { current_frame++; system.perf_stats->EndSystemFrame(); - render_window.PollEvents(); + CITRA_FRAME_END(EmuThreadFrame); system.frame_limiter.DoFrameLimiting(system.CoreTiming().GetGlobalTimeUs()); system.perf_stats->BeginSystemFrame(); + CITRA_FRAME_BEGIN(EmuThreadFrame); } bool RendererBase::IsScreenshotPending() const { diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 0487905de..93b91e10b 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -7,7 +7,7 @@ #include "common/literals.h" #include "common/logging/log.h" #include "common/math_util.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "video_core/pica/pica_core.h" #include "video_core/renderer_opengl/gl_rasterizer.h" #include "video_core/renderer_opengl/pica_to_gl.h" @@ -19,12 +19,6 @@ namespace OpenGL { namespace { -MICROPROFILE_DEFINE(OpenGL_VAO, "OpenGL", "Vertex Array Setup", MP_RGB(255, 128, 0)); -MICROPROFILE_DEFINE(OpenGL_VS, "OpenGL", "Vertex Shader Setup", MP_RGB(192, 128, 128)); -MICROPROFILE_DEFINE(OpenGL_GS, "OpenGL", "Geometry Shader Setup", MP_RGB(128, 192, 128)); -MICROPROFILE_DEFINE(OpenGL_Drawing, "OpenGL", "Drawing", MP_RGB(128, 128, 192)); -MICROPROFILE_DEFINE(OpenGL_Display, "OpenGL", "Display", MP_RGB(128, 128, 192)); - using VideoCore::SurfaceType; using namespace Common::Literals; using namespace Pica::Shader::Generator; @@ -189,7 +183,7 @@ void RasterizerOpenGL::SyncFixedState() { void RasterizerOpenGL::SetupVertexArray(u8* array_ptr, GLintptr buffer_offset, GLuint vs_input_index_min, GLuint vs_input_index_max) { - MICROPROFILE_SCOPE(OpenGL_VAO); + CITRA_PROFILE("OpenGL", "Vertex Array Setup"); const auto& vertex_attributes = regs.pipeline.vertex_attributes; PAddr base_address = vertex_attributes.GetPhysicalBaseAddress(); @@ -265,12 +259,12 @@ void RasterizerOpenGL::SetupVertexArray(u8* array_ptr, GLintptr buffer_offset, } bool RasterizerOpenGL::SetupVertexShader() { - MICROPROFILE_SCOPE(OpenGL_VS); + CITRA_PROFILE("OpenGL", "Vertex Shader Setup"); return shader_manager.UseProgrammableVertexShader(regs, pica.vs_setup); } bool RasterizerOpenGL::SetupGeometryShader() { - MICROPROFILE_SCOPE(OpenGL_GS); + CITRA_PROFILE("OpenGL", "Geometry Shader Setup"); if (regs.pipeline.use_gs != Pica::PipelineRegs::UseGS::No) { LOG_ERROR(Render_OpenGL, "Accelerate draw doesn't support geometry shader"); @@ -364,7 +358,7 @@ void RasterizerOpenGL::DrawTriangles() { } bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) { - MICROPROFILE_SCOPE(OpenGL_Drawing); + CITRA_PROFILE("OpenGL", "Drawing"); const bool shadow_rendering = regs.framebuffer.IsShadowRendering(); const bool has_stencil = regs.framebuffer.HasStencil(); @@ -725,7 +719,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const Pica::FramebufferConfig& config, if (framebuffer_addr == 0) { return false; } - MICROPROFILE_SCOPE(OpenGL_Display); + CITRA_PROFILE("OpenGL", "Display"); VideoCore::SurfaceParams src_params; src_params.addr = framebuffer_addr; diff --git a/src/video_core/renderer_opengl/gl_resource_manager.cpp b/src/video_core/renderer_opengl/gl_resource_manager.cpp index 87a53ccdf..fe038e89e 100644 --- a/src/video_core/renderer_opengl/gl_resource_manager.cpp +++ b/src/video_core/renderer_opengl/gl_resource_manager.cpp @@ -2,14 +2,11 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include "common/microprofile.h" +#include "common/profiling.h" #include "video_core/renderer_opengl/gl_resource_manager.h" #include "video_core/renderer_opengl/gl_shader_util.h" #include "video_core/renderer_opengl/gl_state.h" -MICROPROFILE_DEFINE(OpenGL_ResourceCreation, "OpenGL", "Resource Creation", MP_RGB(128, 128, 192)); -MICROPROFILE_DEFINE(OpenGL_ResourceDeletion, "OpenGL", "Resource Deletion", MP_RGB(128, 128, 192)); - namespace OpenGL { void OGLRenderbuffer::Create() { @@ -17,7 +14,6 @@ void OGLRenderbuffer::Create() { return; } - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); glGenRenderbuffers(1, &handle); } @@ -26,7 +22,6 @@ void OGLRenderbuffer::Release() { return; } - MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); glDeleteRenderbuffers(1, &handle); OpenGLState::GetCurState().ResetRenderbuffer(handle).Apply(); handle = 0; @@ -37,7 +32,6 @@ void OGLTexture::Create() { return; } - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); glGenTextures(1, &handle); } @@ -46,7 +40,6 @@ void OGLTexture::Release() { return; } - MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); glDeleteTextures(1, &handle); OpenGLState::GetCurState().ResetTexture(handle).Apply(); handle = 0; @@ -88,7 +81,6 @@ void OGLSampler::Create() { return; } - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); glGenSamplers(1, &handle); } @@ -97,7 +89,6 @@ void OGLSampler::Release() { return; } - MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); glDeleteSamplers(1, &handle); OpenGLState::GetCurState().ResetSampler(handle).Apply(); handle = 0; @@ -109,7 +100,6 @@ void OGLShader::Create(std::string_view source, GLenum type) { if (source.empty()) return; - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); handle = LoadShader(source, type); } @@ -117,7 +107,6 @@ void OGLShader::Release() { if (handle == 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); glDeleteShader(handle); handle = 0; } @@ -126,7 +115,6 @@ void OGLProgram::Create(bool separable_program, std::span shaders) if (handle != 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); handle = LoadProgram(separable_program, shaders); } @@ -135,7 +123,6 @@ void OGLProgram::Create(std::string_view vert_shader, std::string_view frag_shad vert.Create(vert_shader, GL_VERTEX_SHADER); frag.Create(frag_shader, GL_FRAGMENT_SHADER); - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); const std::array shaders{vert.handle, frag.handle}; Create(false, shaders); } @@ -144,7 +131,6 @@ void OGLProgram::Release() { if (handle == 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); glDeleteProgram(handle); OpenGLState::GetCurState().ResetProgram(handle).Apply(); handle = 0; @@ -154,7 +140,6 @@ void OGLPipeline::Create() { if (handle != 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); glGenProgramPipelines(1, &handle); } @@ -162,7 +147,6 @@ void OGLPipeline::Release() { if (handle == 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); glDeleteProgramPipelines(1, &handle); OpenGLState::GetCurState().ResetPipeline(handle).Apply(); handle = 0; @@ -172,7 +156,6 @@ void OGLBuffer::Create() { if (handle != 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); glGenBuffers(1, &handle); } @@ -180,7 +163,6 @@ void OGLBuffer::Release() { if (handle == 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); glDeleteBuffers(1, &handle); OpenGLState::GetCurState().ResetBuffer(handle).Apply(); handle = 0; @@ -190,7 +172,6 @@ void OGLVertexArray::Create() { if (handle != 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); glGenVertexArrays(1, &handle); } @@ -198,7 +179,6 @@ void OGLVertexArray::Release() { if (handle == 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); glDeleteVertexArrays(1, &handle); OpenGLState::GetCurState().ResetVertexArray(handle).Apply(); handle = 0; @@ -208,7 +188,6 @@ void OGLFramebuffer::Create() { if (handle != 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceCreation); glGenFramebuffers(1, &handle); } @@ -216,7 +195,6 @@ void OGLFramebuffer::Release() { if (handle == 0) return; - MICROPROFILE_SCOPE(OpenGL_ResourceDeletion); glDeleteFramebuffers(1, &handle); OpenGLState::GetCurState().ResetFramebuffer(handle).Apply(); handle = 0; diff --git a/src/video_core/renderer_opengl/gl_stream_buffer.cpp b/src/video_core/renderer_opengl/gl_stream_buffer.cpp index b2c781f24..f22d34a9b 100644 --- a/src/video_core/renderer_opengl/gl_stream_buffer.cpp +++ b/src/video_core/renderer_opengl/gl_stream_buffer.cpp @@ -4,13 +4,10 @@ #include "common/alignment.h" #include "common/assert.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "video_core/renderer_opengl/gl_driver.h" #include "video_core/renderer_opengl/gl_stream_buffer.h" -MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning", - MP_RGB(128, 128, 192)); - namespace OpenGL { OGLStreamBuffer::OGLStreamBuffer(Driver& driver, GLenum target, GLsizeiptr size, @@ -73,7 +70,7 @@ std::tuple OGLStreamBuffer::Map(GLsizeiptr size, GLintptr a } if (invalidate || !persistent) { - MICROPROFILE_SCOPE(OpenGL_StreamBuffer); + CITRA_PROFILE("OpenGL", "Stream Buffer Orphaning"); GLbitfield flags = GL_MAP_WRITE_BIT | (persistent ? GL_MAP_PERSISTENT_BIT : 0) | (coherent ? GL_MAP_COHERENT_BIT : GL_MAP_FLUSH_EXPLICIT_BIT) | (invalidate ? GL_MAP_INVALIDATE_BUFFER_BIT : GL_MAP_UNSYNCHRONIZED_BIT); diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp index e9423fc6d..18440efe9 100644 --- a/src/video_core/renderer_opengl/renderer_opengl.cpp +++ b/src/video_core/renderer_opengl/renderer_opengl.cpp @@ -3,7 +3,7 @@ // Refer to the license.txt file included. #include "common/logging/log.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/settings.h" #include "core/core.h" #include "core/frontend/emu_window.h" @@ -23,9 +23,6 @@ namespace OpenGL { -MICROPROFILE_DEFINE(OpenGL_RenderFrame, "OpenGL", "Render Frame", MP_RGB(128, 128, 64)); -MICROPROFILE_DEFINE(OpenGL_WaitPresent, "OpenGL", "Wait For Present", MP_RGB(128, 128, 128)); - /** * Vertex structure that the drawn screen rectangles are composed of. */ @@ -182,7 +179,7 @@ void RendererOpenGL::RenderToMailbox(const Layout::FramebufferLayout& layout, Frontend::Frame* frame; { - MICROPROFILE_SCOPE(OpenGL_WaitPresent); + CITRA_PROFILE("OpenGL", "Wait For Present"); frame = mailbox->GetRenderFrame(); @@ -209,7 +206,7 @@ void RendererOpenGL::RenderToMailbox(const Layout::FramebufferLayout& layout, } { - MICROPROFILE_SCOPE(OpenGL_RenderFrame); + CITRA_PROFILE("OpenGL", "Render Frame"); // Recreate the frame if the size of the window has changed if (layout.width != frame->width || layout.height != frame->height) { LOG_DEBUG(Render_OpenGL, "Reloading render frame"); diff --git a/src/video_core/renderer_software/sw_rasterizer.cpp b/src/video_core/renderer_software/sw_rasterizer.cpp index f0f290ebe..26df21482 100644 --- a/src/video_core/renderer_software/sw_rasterizer.cpp +++ b/src/video_core/renderer_software/sw_rasterizer.cpp @@ -4,7 +4,7 @@ #include #include "common/logging/log.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/quaternion.h" #include "common/vector_math.h" #include "core/memory.h" @@ -68,8 +68,6 @@ struct Vertex : Pica::OutputVertex { namespace { -MICROPROFILE_DEFINE(GPU_Rasterization, "GPU", "Rasterization", MP_RGB(50, 50, 240)); - struct ClippingEdge { public: constexpr ClippingEdge(Common::Vec4 coeffs, @@ -229,7 +227,7 @@ void RasterizerSoftware::MakeScreenCoords(Vertex& vtx) { void RasterizerSoftware::ProcessTriangle(const Vertex& v0, const Vertex& v1, const Vertex& v2, bool reversed) { - MICROPROFILE_SCOPE(GPU_Rasterization); + CITRA_PROFILE("Software", "Rasterization"); // Vertex positions in rasterizer coordinates static auto screen_to_rasterizer_coords = [](const Common::Vec3& vec) { diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp index 74a231bff..459b11731 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp @@ -5,7 +5,7 @@ #include "common/assert.h" #include "common/logging/log.h" #include "common/memory_detect.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/settings.h" #include "core/core.h" #include "core/frontend/emu_window.h" @@ -22,8 +22,6 @@ #include -MICROPROFILE_DEFINE(Vulkan_RenderFrame, "Vulkan", "Render Frame", MP_RGB(128, 128, 64)); - namespace Vulkan { struct ScreenRectVertex { diff --git a/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp b/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp index 3909da237..c999ae946 100644 --- a/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp +++ b/src/video_core/renderer_vulkan/vk_descriptor_pool.cpp @@ -2,15 +2,11 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include "common/microprofile.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_instance.h" namespace Vulkan { -MICROPROFILE_DEFINE(Vulkan_DescriptorSetAcquire, "Vulkan", "Descriptor Set Acquire", - MP_RGB(64, 128, 256)); - constexpr u32 MAX_BATCH_SIZE = 8; DescriptorPool::DescriptorPool(const Instance& instance_) : instance{instance_} { @@ -107,7 +103,6 @@ DescriptorSetProvider::DescriptorSetProvider( DescriptorSetProvider::~DescriptorSetProvider() = default; vk::DescriptorSet DescriptorSetProvider::Acquire(std::span data) { - MICROPROFILE_SCOPE(Vulkan_DescriptorSetAcquire); DescriptorSetData key{}; std::memcpy(key.data(), data.data(), data.size_bytes()); const auto [it, new_set] = descriptor_set_map.try_emplace(key); diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp index 5ef3eb513..3a8df7b33 100644 --- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp +++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp @@ -5,7 +5,7 @@ #include #include "common/hash.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "video_core/renderer_vulkan/pica_to_vk.h" #include "video_core/renderer_vulkan/vk_graphics_pipeline.h" #include "video_core/renderer_vulkan/vk_instance.h" @@ -14,8 +14,6 @@ namespace Vulkan { -MICROPROFILE_DEFINE(Vulkan_Pipeline, "Vulkan", "Pipeline Building", MP_RGB(0, 192, 32)); - vk::ShaderStageFlagBits MakeShaderStage(std::size_t index) { switch (index) { case 0: @@ -99,7 +97,7 @@ bool GraphicsPipeline::TryBuild(bool wait_built) { } bool GraphicsPipeline::Build(bool fail_on_compile_required) { - MICROPROFILE_SCOPE(Vulkan_Pipeline); + CITRA_PROFILE("Vulkan", "Pipeline Building"); const vk::Device device = instance.GetDevice(); std::array bindings; diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp index 9668aed69..d0c57af75 100644 --- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp +++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp @@ -7,7 +7,7 @@ #include "common/common_paths.h" #include "common/file_util.h" #include "common/logging/log.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/scope_exit.h" #include "common/settings.h" #include "video_core/renderer_vulkan/pica_to_vk.h" @@ -23,8 +23,6 @@ using namespace Pica::Shader::Generator; using Pica::Shader::FSConfig; -MICROPROFILE_DEFINE(Vulkan_Bind, "Vulkan", "Pipeline Bind", MP_RGB(192, 32, 32)); - namespace Vulkan { u32 AttribBytes(Pica::PipelineRegs::VertexAttributeFormat format, u32 size) { @@ -192,7 +190,7 @@ void PipelineCache::SaveDiskCache() { } bool PipelineCache::BindPipeline(const PipelineInfo& info, bool wait_built) { - MICROPROFILE_SCOPE(Vulkan_Bind); + CITRA_PROFILE("Vulkan", "Pipeline Bind"); u64 shader_hash = 0; for (u32 i = 0; i < MAX_SHADER_STAGES; i++) { diff --git a/src/video_core/renderer_vulkan/vk_present_window.cpp b/src/video_core/renderer_vulkan/vk_present_window.cpp index 273b20e8e..ac679c77a 100644 --- a/src/video_core/renderer_vulkan/vk_present_window.cpp +++ b/src/video_core/renderer_vulkan/vk_present_window.cpp @@ -2,7 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/settings.h" #include "common/thread.h" #include "core/frontend/emu_window.h" @@ -15,8 +15,6 @@ #include -MICROPROFILE_DEFINE(Vulkan_WaitPresent, "Vulkan", "Wait For Present", MP_RGB(128, 128, 128)); - namespace Vulkan { namespace { @@ -237,7 +235,7 @@ void PresentWindow::RecreateFrame(Frame* frame, u32 width, u32 height) { } Frame* PresentWindow::GetRenderFrame() { - MICROPROFILE_SCOPE(Vulkan_WaitPresent); + CITRA_PROFILE("Vulkan", "Wait For Present"); // Wait for free presentation frames std::unique_lock lock{free_mutex}; diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp index 0208c8cde..1fa3c8720 100644 --- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp +++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp @@ -6,7 +6,7 @@ #include "common/literals.h" #include "common/logging/log.h" #include "common/math_util.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/settings.h" #include "core/memory.h" #include "video_core/pica/pica_core.h" @@ -20,10 +20,6 @@ namespace Vulkan { namespace { -MICROPROFILE_DEFINE(Vulkan_VS, "Vulkan", "Vertex Shader Setup", MP_RGB(192, 128, 128)); -MICROPROFILE_DEFINE(Vulkan_GS, "Vulkan", "Geometry Shader Setup", MP_RGB(128, 192, 128)); -MICROPROFILE_DEFINE(Vulkan_Drawing, "Vulkan", "Drawing", MP_RGB(128, 128, 192)); - using TriangleTopology = Pica::PipelineRegs::TriangleTopology; using VideoCore::SurfaceType; @@ -320,13 +316,13 @@ void RasterizerVulkan::SetupFixedAttribs() { } bool RasterizerVulkan::SetupVertexShader() { - MICROPROFILE_SCOPE(Vulkan_VS); + CITRA_PROFILE("Vulkan", "Vertex Shader Setup"); return pipeline_cache.UseProgrammableVertexShader(regs, pica.vs_setup, pipeline_info.vertex_layout); } bool RasterizerVulkan::SetupGeometryShader() { - MICROPROFILE_SCOPE(Vulkan_GS); + CITRA_PROFILE("Vulkan", "Geometry Shader Setup"); if (regs.pipeline.use_gs != Pica::PipelineRegs::UseGS::No) { LOG_ERROR(Render_Vulkan, "Accelerate draw doesn't support geometry shader"); @@ -454,7 +450,7 @@ void RasterizerVulkan::DrawTriangles() { } bool RasterizerVulkan::Draw(bool accelerate, bool is_indexed) { - MICROPROFILE_SCOPE(Vulkan_Drawing); + CITRA_PROFILE("Vulkan", "Drawing"); const bool shadow_rendering = regs.framebuffer.IsShadowRendering(); const bool has_stencil = regs.framebuffer.HasStencil(); diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp index 315fe0ea7..638fc9fe4 100644 --- a/src/video_core/renderer_vulkan/vk_scheduler.cpp +++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp @@ -4,16 +4,13 @@ #include #include -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/settings.h" #include "common/thread.h" #include "video_core/renderer_vulkan/vk_instance.h" #include "video_core/renderer_vulkan/vk_renderpass_cache.h" #include "video_core/renderer_vulkan/vk_scheduler.h" -MICROPROFILE_DEFINE(Vulkan_WaitForWorker, "Vulkan", "Wait for worker", MP_RGB(255, 192, 192)); -MICROPROFILE_DEFINE(Vulkan_Submit, "Vulkan", "Submit Exectution", MP_RGB(255, 192, 255)); - namespace Vulkan { namespace { @@ -71,7 +68,7 @@ void Scheduler::WaitWorker() { return; } - MICROPROFILE_SCOPE(Vulkan_WaitForWorker); + CITRA_PROFILE("Vulkan", "Vulkan WaitWorker"); DispatchWork(); // Ensure the queue is drained. @@ -174,7 +171,7 @@ void Scheduler::SubmitExecution(vk::Semaphore signal_semaphore, vk::Semaphore wa const u64 signal_value = master_semaphore->NextTick(); Record([signal_semaphore, wait_semaphore, signal_value, this](vk::CommandBuffer cmdbuf) { - MICROPROFILE_SCOPE(Vulkan_Submit); + CITRA_PROFILE("Vulkan", "Vulkan Submit"); std::scoped_lock lock{submit_mutex}; master_semaphore->SubmitWork(cmdbuf, wait_semaphore, signal_semaphore, signal_value); }); diff --git a/src/video_core/renderer_vulkan/vk_swapchain.cpp b/src/video_core/renderer_vulkan/vk_swapchain.cpp index 5e725be1e..361ffd5e8 100644 --- a/src/video_core/renderer_vulkan/vk_swapchain.cpp +++ b/src/video_core/renderer_vulkan/vk_swapchain.cpp @@ -5,14 +5,11 @@ #include #include #include "common/logging/log.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/settings.h" #include "video_core/renderer_vulkan/vk_instance.h" #include "video_core/renderer_vulkan/vk_swapchain.h" -MICROPROFILE_DEFINE(Vulkan_Acquire, "Vulkan", "Swapchain Acquire", MP_RGB(185, 66, 245)); -MICROPROFILE_DEFINE(Vulkan_Present, "Vulkan", "Swapchain Present", MP_RGB(66, 185, 245)); - namespace Vulkan { Swapchain::Swapchain(const Instance& instance_, u32 width, u32 height, vk::SurfaceKHR surface_) @@ -78,7 +75,7 @@ void Swapchain::Create(u32 width_, u32 height_, vk::SurfaceKHR surface_) { } bool Swapchain::AcquireNextImage() { - MICROPROFILE_SCOPE(Vulkan_Acquire); + CITRA_PROFILE("Vulkan", "Swapchain Acquire"); vk::Device device = instance.GetDevice(); vk::Result result = device.acquireNextImageKHR(swapchain, std::numeric_limits::max(), @@ -114,7 +111,7 @@ void Swapchain::Present() { .pImageIndices = &image_index, }; - MICROPROFILE_SCOPE(Vulkan_Present); + CITRA_PROFILE("Vulkan", "Swapchain Present"); try { [[maybe_unused]] vk::Result result = instance.GetPresentQueue().presentKHR(present_info); } catch (vk::OutOfDateKHRError&) { diff --git a/src/video_core/renderer_vulkan/vk_texture_runtime.cpp b/src/video_core/renderer_vulkan/vk_texture_runtime.cpp index 4ba7e6052..dc4858367 100644 --- a/src/video_core/renderer_vulkan/vk_texture_runtime.cpp +++ b/src/video_core/renderer_vulkan/vk_texture_runtime.cpp @@ -5,7 +5,7 @@ #include #include "common/literals.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/scope_exit.h" #include "video_core/custom_textures/material.h" #include "video_core/rasterizer_cache/texture_codec.h" @@ -26,8 +26,6 @@ #pragma GCC diagnostic ignored "-Wclass-memaccess" #endif -MICROPROFILE_DEFINE(Vulkan_ImageAlloc, "Vulkan", "Texture Allocation", MP_RGB(192, 52, 235)); - namespace Vulkan { namespace { diff --git a/src/video_core/shader/shader_interpreter.cpp b/src/video_core/shader/shader_interpreter.cpp index 04d9446f1..7e36a97e0 100644 --- a/src/video_core/shader/shader_interpreter.cpp +++ b/src/video_core/shader/shader_interpreter.cpp @@ -11,7 +11,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "common/logging/log.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "common/vector_math.h" #include "video_core/pica/shader_setup.h" #include "video_core/pica/shader_unit.h" @@ -728,11 +728,8 @@ void InterpreterEngine::SetupBatch(ShaderSetup& setup, unsigned int entry_point) setup.entry_point = entry_point; } -MICROPROFILE_DEFINE(GPU_Shader, "GPU", "Shader", MP_RGB(50, 50, 240)); - void InterpreterEngine::Run(const ShaderSetup& setup, ShaderUnit& state) const { - - MICROPROFILE_SCOPE(GPU_Shader); + CITRA_PROFILE("Shader", "Shader Interpreter"); DebugData dummy_debug_data; RunInterpreter(setup, state, dummy_debug_data, setup.entry_point); diff --git a/src/video_core/shader/shader_jit.cpp b/src/video_core/shader/shader_jit.cpp index 644586226..bc45d04ae 100644 --- a/src/video_core/shader/shader_jit.cpp +++ b/src/video_core/shader/shader_jit.cpp @@ -7,7 +7,7 @@ #include "common/assert.h" #include "common/hash.h" -#include "common/microprofile.h" +#include "common/profiling.h" #include "video_core/shader/shader.h" #include "video_core/shader/shader_jit.h" #if CITRA_ARCH(arm64) @@ -41,12 +41,10 @@ void JitEngine::SetupBatch(ShaderSetup& setup, u32 entry_point) { } } -MICROPROFILE_DECLARE(GPU_Shader); - void JitEngine::Run(const ShaderSetup& setup, ShaderUnit& state) const { ASSERT(setup.cached_shader != nullptr); - MICROPROFILE_SCOPE(GPU_Shader); + CITRA_PROFILE("Shader", "Shader JIT"); const JitShader* shader = static_cast(setup.cached_shader); shader->Run(setup, state, setup.entry_point);