Query Cachge: Fully rework Vulkan's query cache

This commit is contained in:
Fernando Sahmkow 2023-08-04 03:32:30 +02:00
parent bdc01254a9
commit f1a2e36711
35 changed files with 1573 additions and 355 deletions

View file

@ -130,13 +130,17 @@ void LogSettings() {
log_path("DataStorage_SDMCDir", Common::FS::GetYuzuPath(Common::FS::YuzuPath::SDMCDir)); log_path("DataStorage_SDMCDir", Common::FS::GetYuzuPath(Common::FS::YuzuPath::SDMCDir));
} }
void UpdateGPUAccuracy() {
values.current_gpu_accuracy = values.gpu_accuracy.GetValue();
}
bool IsGPULevelExtreme() { bool IsGPULevelExtreme() {
return values.gpu_accuracy.GetValue() == GpuAccuracy::Extreme; return values.current_gpu_accuracy == GpuAccuracy::Extreme;
} }
bool IsGPULevelHigh() { bool IsGPULevelHigh() {
return values.gpu_accuracy.GetValue() == GpuAccuracy::Extreme || return values.current_gpu_accuracy == GpuAccuracy::Extreme ||
values.gpu_accuracy.GetValue() == GpuAccuracy::High; values.current_gpu_accuracy == GpuAccuracy::High;
} }
bool IsFastmemEnabled() { bool IsFastmemEnabled() {

View file

@ -307,6 +307,7 @@ struct Values {
Specialization::Default, Specialization::Default,
true, true,
true}; true};
GpuAccuracy current_gpu_accuracy{GpuAccuracy::High};
SwitchableSetting<AnisotropyMode, true> max_anisotropy{ SwitchableSetting<AnisotropyMode, true> max_anisotropy{
linkage, AnisotropyMode::Automatic, AnisotropyMode::Automatic, AnisotropyMode::X16, linkage, AnisotropyMode::Automatic, AnisotropyMode::Automatic, AnisotropyMode::X16,
"max_anisotropy", Category::RendererAdvanced}; "max_anisotropy", Category::RendererAdvanced};
@ -522,6 +523,7 @@ struct Values {
extern Values values; extern Values values;
void UpdateGPUAccuracy();
bool IsGPULevelExtreme(); bool IsGPULevelExtreme();
bool IsGPULevelHigh(); bool IsGPULevelHigh();

View file

@ -272,13 +272,20 @@ std::pair<typename P::Buffer*, u32> BufferCache<P>::ObtainBuffer(GPUVAddr gpu_ad
if (!cpu_addr) { if (!cpu_addr) {
return {&slot_buffers[NULL_BUFFER_ID], 0}; return {&slot_buffers[NULL_BUFFER_ID], 0};
} }
const BufferId buffer_id = FindBuffer(*cpu_addr, size); return ObtainCPUBuffer(*cpu_addr, size, sync_info, post_op);
}
template <class P>
std::pair<typename P::Buffer*, u32> BufferCache<P>::ObtainCPUBuffer(VAddr cpu_addr, u32 size,
ObtainBufferSynchronize sync_info,
ObtainBufferOperation post_op) {
const BufferId buffer_id = FindBuffer(cpu_addr, size);
Buffer& buffer = slot_buffers[buffer_id]; Buffer& buffer = slot_buffers[buffer_id];
// synchronize op // synchronize op
switch (sync_info) { switch (sync_info) {
case ObtainBufferSynchronize::FullSynchronize: case ObtainBufferSynchronize::FullSynchronize:
SynchronizeBuffer(buffer, *cpu_addr, size); SynchronizeBuffer(buffer, cpu_addr, size);
break; break;
default: default:
break; break;
@ -286,11 +293,11 @@ std::pair<typename P::Buffer*, u32> BufferCache<P>::ObtainBuffer(GPUVAddr gpu_ad
switch (post_op) { switch (post_op) {
case ObtainBufferOperation::MarkAsWritten: case ObtainBufferOperation::MarkAsWritten:
MarkWrittenBuffer(buffer_id, *cpu_addr, size); MarkWrittenBuffer(buffer_id, cpu_addr, size);
break; break;
case ObtainBufferOperation::DiscardWrite: { case ObtainBufferOperation::DiscardWrite: {
VAddr cpu_addr_start = Common::AlignDown(*cpu_addr, 64); VAddr cpu_addr_start = Common::AlignDown(cpu_addr, 64);
VAddr cpu_addr_end = Common::AlignUp(*cpu_addr + size, 64); VAddr cpu_addr_end = Common::AlignUp(cpu_addr + size, 64);
IntervalType interval{cpu_addr_start, cpu_addr_end}; IntervalType interval{cpu_addr_start, cpu_addr_end};
ClearDownload(interval); ClearDownload(interval);
common_ranges.subtract(interval); common_ranges.subtract(interval);
@ -300,7 +307,7 @@ std::pair<typename P::Buffer*, u32> BufferCache<P>::ObtainBuffer(GPUVAddr gpu_ad
break; break;
} }
return {&buffer, buffer.Offset(*cpu_addr)}; return {&buffer, buffer.Offset(cpu_addr)};
} }
template <class P> template <class P>

View file

@ -295,6 +295,10 @@ public:
[[nodiscard]] std::pair<Buffer*, u32> ObtainBuffer(GPUVAddr gpu_addr, u32 size, [[nodiscard]] std::pair<Buffer*, u32> ObtainBuffer(GPUVAddr gpu_addr, u32 size,
ObtainBufferSynchronize sync_info, ObtainBufferSynchronize sync_info,
ObtainBufferOperation post_op); ObtainBufferOperation post_op);
[[nodiscard]] std::pair<Buffer*, u32> ObtainCPUBuffer(VAddr gpu_addr, u32 size,
ObtainBufferSynchronize sync_info,
ObtainBufferOperation post_op);
void FlushCachedWrites(); void FlushCachedWrites();
/// Return true when there are uncommitted buffers to be downloaded /// Return true when there are uncommitted buffers to be downloaded
@ -335,6 +339,14 @@ public:
[[nodiscard]] std::pair<Buffer*, u32> GetDrawIndirectBuffer(); [[nodiscard]] std::pair<Buffer*, u32> GetDrawIndirectBuffer();
template <typename Func>
void BufferOperations(Func&& func) {
do {
channel_state->has_deleted_buffers = false;
func();
} while (channel_state->has_deleted_buffers);
}
std::recursive_mutex mutex; std::recursive_mutex mutex;
Runtime& runtime; Runtime& runtime;

View file

@ -51,7 +51,7 @@ public:
virtual void CreateChannel(Tegra::Control::ChannelState& channel); virtual void CreateChannel(Tegra::Control::ChannelState& channel);
/// Bind a channel for execution. /// Bind a channel for execution.
void BindToChannel(s32 id); virtual void BindToChannel(s32 id);
/// Erase channel's state. /// Erase channel's state.
void EraseChannel(s32 id); void EraseChannel(s32 id);

View file

@ -20,8 +20,6 @@
namespace Tegra::Engines { namespace Tegra::Engines {
using VideoCore::QueryType;
/// First register id that is actually a Macro call. /// First register id that is actually a Macro call.
constexpr u32 MacroRegistersStart = 0xE00; constexpr u32 MacroRegistersStart = 0xE00;
@ -500,27 +498,21 @@ void Maxwell3D::StampQueryResult(u64 payload, bool long_query) {
} }
void Maxwell3D::ProcessQueryGet() { void Maxwell3D::ProcessQueryGet() {
VideoCommon::QueryPropertiesFlags flags{};
if (regs.report_semaphore.query.short_query == 0) {
flags |= VideoCommon::QueryPropertiesFlags::HasTimeout;
}
const GPUVAddr sequence_address{regs.report_semaphore.Address()};
const VideoCommon::QueryType query_type =
static_cast<VideoCommon::QueryType>(regs.report_semaphore.query.report.Value());
const u32 payload = regs.report_semaphore.payload;
const u32 subreport = regs.report_semaphore.query.sub_report;
switch (regs.report_semaphore.query.operation) { switch (regs.report_semaphore.query.operation) {
case Regs::ReportSemaphore::Operation::Release: case Regs::ReportSemaphore::Operation::Release:
if (regs.report_semaphore.query.short_query != 0) { if (regs.report_semaphore.query.short_query != 0) {
const GPUVAddr sequence_address{regs.report_semaphore.Address()}; flags |= VideoCommon::QueryPropertiesFlags::IsAFence;
const u32 payload = regs.report_semaphore.payload;
std::function<void()> operation([this, sequence_address, payload] {
memory_manager.Write<u32>(sequence_address, payload);
});
rasterizer->SignalFence(std::move(operation));
} else {
struct LongQueryResult {
u64_le value;
u64_le timestamp;
};
const GPUVAddr sequence_address{regs.report_semaphore.Address()};
const u32 payload = regs.report_semaphore.payload;
[this, sequence_address, payload] {
memory_manager.Write<u64>(sequence_address + sizeof(u64), system.GPU().GetTicks());
memory_manager.Write<u64>(sequence_address, payload);
}();
} }
rasterizer->Query(sequence_address, query_type, flags, payload, subreport);
break; break;
case Regs::ReportSemaphore::Operation::Acquire: case Regs::ReportSemaphore::Operation::Acquire:
// TODO(Blinkhawk): Under this operation, the GPU waits for the CPU to write a value that // TODO(Blinkhawk): Under this operation, the GPU waits for the CPU to write a value that
@ -528,11 +520,7 @@ void Maxwell3D::ProcessQueryGet() {
UNIMPLEMENTED_MSG("Unimplemented query operation ACQUIRE"); UNIMPLEMENTED_MSG("Unimplemented query operation ACQUIRE");
break; break;
case Regs::ReportSemaphore::Operation::ReportOnly: case Regs::ReportSemaphore::Operation::ReportOnly:
if (const std::optional<u64> result = GetQueryResult()) { rasterizer->Query(sequence_address, query_type, flags, payload, subreport);
// If the query returns an empty optional it means it's cached and deferred.
// In this case we have a non-empty result, so we stamp it immediately.
StampQueryResult(*result, regs.report_semaphore.query.short_query == 0);
}
break; break;
case Regs::ReportSemaphore::Operation::Trap: case Regs::ReportSemaphore::Operation::Trap:
UNIMPLEMENTED_MSG("Unimplemented query operation TRAP"); UNIMPLEMENTED_MSG("Unimplemented query operation TRAP");
@ -544,6 +532,10 @@ void Maxwell3D::ProcessQueryGet() {
} }
void Maxwell3D::ProcessQueryCondition() { void Maxwell3D::ProcessQueryCondition() {
if (rasterizer->AccelerateConditionalRendering()) {
execute_on = true;
return;
}
const GPUVAddr condition_address{regs.render_enable.Address()}; const GPUVAddr condition_address{regs.render_enable.Address()};
switch (regs.render_enable_override) { switch (regs.render_enable_override) {
case Regs::RenderEnable::Override::AlwaysRender: case Regs::RenderEnable::Override::AlwaysRender:
@ -553,10 +545,6 @@ void Maxwell3D::ProcessQueryCondition() {
execute_on = false; execute_on = false;
break; break;
case Regs::RenderEnable::Override::UseRenderEnable: { case Regs::RenderEnable::Override::UseRenderEnable: {
if (rasterizer->AccelerateConditionalRendering()) {
execute_on = true;
return;
}
switch (regs.render_enable.mode) { switch (regs.render_enable.mode) {
case Regs::RenderEnable::Mode::True: { case Regs::RenderEnable::Mode::True: {
execute_on = true; execute_on = true;
@ -606,7 +594,13 @@ void Maxwell3D::ProcessCounterReset() {
#endif #endif
switch (regs.clear_report_value) { switch (regs.clear_report_value) {
case Regs::ClearReport::ZPassPixelCount: case Regs::ClearReport::ZPassPixelCount:
rasterizer->ResetCounter(QueryType::SamplesPassed); rasterizer->ResetCounter(VideoCommon::QueryType::ZPassPixelCount64);
break;
case Regs::ClearReport::PrimitivesGenerated:
rasterizer->ResetCounter(VideoCommon::QueryType::StreamingByteCount);
break;
case Regs::ClearReport::VtgPrimitivesOut:
rasterizer->ResetCounter(VideoCommon::QueryType::StreamingByteCount);
break; break;
default: default:
LOG_DEBUG(Render_OpenGL, "Unimplemented counter reset={}", regs.clear_report_value); LOG_DEBUG(Render_OpenGL, "Unimplemented counter reset={}", regs.clear_report_value);
@ -620,28 +614,6 @@ void Maxwell3D::ProcessSyncPoint() {
rasterizer->SignalSyncPoint(sync_point); rasterizer->SignalSyncPoint(sync_point);
} }
std::optional<u64> Maxwell3D::GetQueryResult() {
switch (regs.report_semaphore.query.report) {
case Regs::ReportSemaphore::Report::Payload:
return regs.report_semaphore.payload;
case Regs::ReportSemaphore::Report::ZPassPixelCount64:
#if ANDROID
if (!Settings::IsGPULevelHigh()) {
// This is problematic on Android, disable on GPU Normal.
return 120;
}
#endif
// Deferred.
rasterizer->Query(regs.report_semaphore.Address(), QueryType::SamplesPassed,
system.GPU().GetTicks());
return std::nullopt;
default:
LOG_DEBUG(HW_GPU, "Unimplemented query report type {}",
regs.report_semaphore.query.report.Value());
return 1;
}
}
void Maxwell3D::ProcessCBBind(size_t stage_index) { void Maxwell3D::ProcessCBBind(size_t stage_index) {
// Bind the buffer currently in CB_ADDRESS to the specified index in the desired shader // Bind the buffer currently in CB_ADDRESS to the specified index in the desired shader
// stage. // stage.

View file

@ -3182,9 +3182,6 @@ private:
/// Handles writes to syncing register. /// Handles writes to syncing register.
void ProcessSyncPoint(); void ProcessSyncPoint();
/// Returns a query's value or an empty object if the value will be deferred through a cache.
std::optional<u64> GetQueryResult();
void RefreshParametersImpl(); void RefreshParametersImpl();
bool IsMethodExecutable(u32 method); bool IsMethodExecutable(u32 method);

View file

@ -362,21 +362,17 @@ void MaxwellDMA::ReleaseSemaphore() {
const auto type = regs.launch_dma.semaphore_type; const auto type = regs.launch_dma.semaphore_type;
const GPUVAddr address = regs.semaphore.address; const GPUVAddr address = regs.semaphore.address;
const u32 payload = regs.semaphore.payload; const u32 payload = regs.semaphore.payload;
VideoCommon::QueryPropertiesFlags flags{VideoCommon::QueryPropertiesFlags::IsAFence};
switch (type) { switch (type) {
case LaunchDMA::SemaphoreType::NONE: case LaunchDMA::SemaphoreType::NONE:
break; break;
case LaunchDMA::SemaphoreType::RELEASE_ONE_WORD_SEMAPHORE: { case LaunchDMA::SemaphoreType::RELEASE_ONE_WORD_SEMAPHORE: {
std::function<void()> operation( rasterizer->Query(address, VideoCommon::QueryType::Payload, flags, payload, 0);
[this, address, payload] { memory_manager.Write<u32>(address, payload); });
rasterizer->SignalFence(std::move(operation));
break; break;
} }
case LaunchDMA::SemaphoreType::RELEASE_FOUR_WORD_SEMAPHORE: { case LaunchDMA::SemaphoreType::RELEASE_FOUR_WORD_SEMAPHORE: {
std::function<void()> operation([this, address, payload] { rasterizer->Query(address, VideoCommon::QueryType::Payload,
memory_manager.Write<u64>(address + sizeof(u64), system.GPU().GetTicks()); flags | VideoCommon::QueryPropertiesFlags::HasTimeout, payload, 0);
memory_manager.Write<u64>(address, payload);
});
rasterizer->SignalFence(std::move(operation));
break; break;
} }
default: default:

View file

@ -82,10 +82,7 @@ void Puller::ProcessSemaphoreTriggerMethod() {
if (op == GpuSemaphoreOperation::WriteLong) { if (op == GpuSemaphoreOperation::WriteLong) {
const GPUVAddr sequence_address{regs.semaphore_address.SemaphoreAddress()}; const GPUVAddr sequence_address{regs.semaphore_address.SemaphoreAddress()};
const u32 payload = regs.semaphore_sequence; const u32 payload = regs.semaphore_sequence;
[this, sequence_address, payload] { rasterizer->Query(sequence_address, VideoCommon::QueryType::Payload, VideoCommon::QueryPropertiesFlags::HasTimeout, payload, 0);
memory_manager.Write<u64>(sequence_address + sizeof(u64), gpu.GetTicks());
memory_manager.Write<u64>(sequence_address, payload);
}();
} else { } else {
do { do {
const u32 word{memory_manager.Read<u32>(regs.semaphore_address.SemaphoreAddress())}; const u32 word{memory_manager.Read<u32>(regs.semaphore_address.SemaphoreAddress())};
@ -120,10 +117,7 @@ void Puller::ProcessSemaphoreTriggerMethod() {
void Puller::ProcessSemaphoreRelease() { void Puller::ProcessSemaphoreRelease() {
const GPUVAddr sequence_address{regs.semaphore_address.SemaphoreAddress()}; const GPUVAddr sequence_address{regs.semaphore_address.SemaphoreAddress()};
const u32 payload = regs.semaphore_release; const u32 payload = regs.semaphore_release;
std::function<void()> operation([this, sequence_address, payload] { rasterizer->Query(sequence_address, VideoCommon::QueryType::Payload, VideoCommon::QueryPropertiesFlags::IsAFence, payload, 0);
memory_manager.Write<u32>(sequence_address, payload);
});
rasterizer->SignalFence(std::move(operation));
} }
void Puller::ProcessSemaphoreAcquire() { void Puller::ProcessSemaphoreAcquire() {
@ -132,7 +126,6 @@ void Puller::ProcessSemaphoreAcquire() {
while (word != value) { while (word != value) {
regs.acquire_active = true; regs.acquire_active = true;
regs.acquire_value = value; regs.acquire_value = value;
std::this_thread::sleep_for(std::chrono::milliseconds(1));
rasterizer->ReleaseFences(); rasterizer->ReleaseFences();
word = memory_manager.Read<u32>(regs.semaphore_address.SemaphoreAddress()); word = memory_manager.Read<u32>(regs.semaphore_address.SemaphoreAddress());
// TODO(kemathe73) figure out how to do the acquire_timeout // TODO(kemathe73) figure out how to do the acquire_timeout

View file

@ -104,9 +104,28 @@ public:
SignalFence(std::move(func)); SignalFence(std::move(func));
} }
void WaitPendingFences() { void WaitPendingFences(bool force) {
if constexpr (!can_async_check) { if constexpr (!can_async_check) {
TryReleasePendingFences<true>(); if (force) {
TryReleasePendingFences<true>();
} else {
TryReleasePendingFences<false>();
}
} else {
if (!force) {
return;
}
std::mutex wait_mutex;
std::condition_variable wait_cv;
std::atomic<bool> wait_finished{};
std::function<void()> func([&] {
std::scoped_lock lk(wait_mutex);
wait_finished.store(true, std::memory_order_relaxed);
wait_cv.notify_all();
});
SignalFence(std::move(func));
std::unique_lock lk(wait_mutex);
wait_cv.wait(lk, [&wait_finished] { return wait_finished.load(std::memory_order_relaxed); });
} }
} }

View file

@ -102,7 +102,8 @@ struct GPU::Impl {
/// Signal the ending of command list. /// Signal the ending of command list.
void OnCommandListEnd() { void OnCommandListEnd() {
rasterizer->ReleaseFences(); rasterizer->ReleaseFences(false);
Settings::UpdateGPUAccuracy();
} }
/// Request a host GPU memory flush from the CPU. /// Request a host GPU memory flush from the CPU.
@ -220,6 +221,7 @@ struct GPU::Impl {
/// This can be used to launch any necessary threads and register any necessary /// This can be used to launch any necessary threads and register any necessary
/// core timing events. /// core timing events.
void Start() { void Start() {
Settings::UpdateGPUAccuracy();
gpu_thread.StartThread(*renderer, renderer->Context(), *scheduler); gpu_thread.StartThread(*renderer, renderer->Context(), *scheduler);
} }

View file

@ -41,6 +41,7 @@ set(SHADER_FILES
pitch_unswizzle.comp pitch_unswizzle.comp
present_bicubic.frag present_bicubic.frag
present_gaussian.frag present_gaussian.frag
resolve_conditional_render.comp
smaa_edge_detection.vert smaa_edge_detection.vert
smaa_edge_detection.frag smaa_edge_detection.frag
smaa_blending_weight_calculation.vert smaa_blending_weight_calculation.vert

View file

@ -0,0 +1,20 @@
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#version 450
layout(local_size_x = 1) in;
layout(std430, binding = 0) buffer Query {
uvec2 initial;
uvec2 unknown;
uvec2 current;
};
layout(std430, binding = 1) buffer Result {
uint result;
};
void main() {
result = all(equal(initial, current)) ? 1 : 0;
}

View file

@ -319,6 +319,25 @@ private:
} }
}; };
class HLE_DrawIndirectByteCount final : public HLEMacroImpl {
public:
explicit HLE_DrawIndirectByteCount(Maxwell3D& maxwell3d_) : HLEMacroImpl(maxwell3d_) {}
void Execute(const std::vector<u32>& parameters, [[maybe_unused]] u32 method) override {
maxwell3d.RefreshParameters();
maxwell3d.regs.draw.begin = parameters[0];
maxwell3d.regs.draw_auto_stride = parameters[1];
maxwell3d.regs.draw_auto_byte_count = parameters[2];
if (maxwell3d.ShouldExecute()) {
maxwell3d.draw_manager->DrawArray(
maxwell3d.regs.draw.topology, 0,
maxwell3d.regs.draw_auto_byte_count / maxwell3d.regs.draw_auto_stride, 0, 1);
}
}
};
class HLE_C713C83D8F63CCF3 final : public HLEMacroImpl { class HLE_C713C83D8F63CCF3 final : public HLEMacroImpl {
public: public:
explicit HLE_C713C83D8F63CCF3(Maxwell3D& maxwell3d_) : HLEMacroImpl(maxwell3d_) {} explicit HLE_C713C83D8F63CCF3(Maxwell3D& maxwell3d_) : HLEMacroImpl(maxwell3d_) {}
@ -536,6 +555,11 @@ HLEMacro::HLEMacro(Maxwell3D& maxwell3d_) : maxwell3d{maxwell3d_} {
[](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> { [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
return std::make_unique<HLE_TransformFeedbackSetup>(maxwell3d__); return std::make_unique<HLE_TransformFeedbackSetup>(maxwell3d__);
})); }));
builders.emplace(0xB5F74EDB717278ECULL,
std::function<std::unique_ptr<CachedMacro>(Maxwell3D&)>(
[](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
return std::make_unique<HLE_DrawIndirectByteCount>(maxwell3d__);
}));
} }
HLEMacro::~HLEMacro() = default; HLEMacro::~HLEMacro() = default;

View file

@ -25,6 +25,13 @@
#include "video_core/rasterizer_interface.h" #include "video_core/rasterizer_interface.h"
#include "video_core/texture_cache/slot_vector.h" #include "video_core/texture_cache/slot_vector.h"
namespace VideoCore {
enum class QueryType {
SamplesPassed,
};
constexpr std::size_t NumQueryTypes = 1;
} // namespace VideoCore
namespace VideoCommon { namespace VideoCommon {
using AsyncJobId = SlotId; using AsyncJobId = SlotId;
@ -98,10 +105,10 @@ private:
}; };
template <class QueryCache, class CachedQuery, class CounterStream, class HostCounter> template <class QueryCache, class CachedQuery, class CounterStream, class HostCounter>
class QueryCacheBase : public VideoCommon::ChannelSetupCaches<VideoCommon::ChannelInfo> { class QueryCacheLegacy : public VideoCommon::ChannelSetupCaches<VideoCommon::ChannelInfo> {
public: public:
explicit QueryCacheBase(VideoCore::RasterizerInterface& rasterizer_, explicit QueryCacheLegacy(VideoCore::RasterizerInterface& rasterizer_,
Core::Memory::Memory& cpu_memory_) Core::Memory::Memory& cpu_memory_)
: rasterizer{rasterizer_}, : rasterizer{rasterizer_},
// Use reinterpret_cast instead of static_cast as workaround for // Use reinterpret_cast instead of static_cast as workaround for
// UBSan bug (https://github.com/llvm/llvm-project/issues/59060) // UBSan bug (https://github.com/llvm/llvm-project/issues/59060)

View file

@ -9,6 +9,7 @@
#include <utility> #include <utility>
#include "common/common_types.h" #include "common/common_types.h"
#include "common/polyfill_thread.h" #include "common/polyfill_thread.h"
#include "video_core/query_cache/types.h"
#include "video_core/cache_types.h" #include "video_core/cache_types.h"
#include "video_core/engines/fermi_2d.h" #include "video_core/engines/fermi_2d.h"
#include "video_core/gpu.h" #include "video_core/gpu.h"
@ -26,11 +27,6 @@ struct ChannelState;
namespace VideoCore { namespace VideoCore {
enum class QueryType {
SamplesPassed,
};
constexpr std::size_t NumQueryTypes = 1;
enum class LoadCallbackStage { enum class LoadCallbackStage {
Prepare, Prepare,
Build, Build,
@ -58,10 +54,10 @@ public:
virtual void DispatchCompute() = 0; virtual void DispatchCompute() = 0;
/// Resets the counter of a query /// Resets the counter of a query
virtual void ResetCounter(QueryType type) = 0; virtual void ResetCounter(VideoCommon::QueryType type) = 0;
/// Records a GPU query and caches it /// Records a GPU query and caches it
virtual void Query(GPUVAddr gpu_addr, QueryType type, std::optional<u64> timestamp) = 0; virtual void Query(GPUVAddr gpu_addr, VideoCommon::QueryType type, VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) = 0;
/// Signal an uniform buffer binding /// Signal an uniform buffer binding
virtual void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, virtual void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr,
@ -83,7 +79,7 @@ public:
virtual void SignalReference() = 0; virtual void SignalReference() = 0;
/// Release all pending fences. /// Release all pending fences.
virtual void ReleaseFences() = 0; virtual void ReleaseFences(bool force = true) = 0;
/// Notify rasterizer that all caches should be flushed to Switch memory /// Notify rasterizer that all caches should be flushed to Switch memory
virtual void FlushAll() = 0; virtual void FlushAll() = 0;

View file

@ -26,16 +26,18 @@ void RasterizerNull::Draw(bool is_indexed, u32 instance_count) {}
void RasterizerNull::DrawTexture() {} void RasterizerNull::DrawTexture() {}
void RasterizerNull::Clear(u32 layer_count) {} void RasterizerNull::Clear(u32 layer_count) {}
void RasterizerNull::DispatchCompute() {} void RasterizerNull::DispatchCompute() {}
void RasterizerNull::ResetCounter(VideoCore::QueryType type) {} void RasterizerNull::ResetCounter(VideoCommon::QueryType type) {}
void RasterizerNull::Query(GPUVAddr gpu_addr, VideoCore::QueryType type, void RasterizerNull::Query(GPUVAddr gpu_addr, VideoCommon::QueryType type,
std::optional<u64> timestamp) { VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) {
if (!gpu_memory) { if (!gpu_memory) {
return; return;
} }
if (True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout)) {
gpu_memory->Write(gpu_addr, u64{0}); u64 ticks = m_gpu.GetTicks();
if (timestamp) { gpu_memory->Write<u64>(gpu_addr + 8, ticks);
gpu_memory->Write(gpu_addr + 8, *timestamp); gpu_memory->Write<u64>(gpu_addr, static_cast<u64>(payload));
} else {
gpu_memory->Write<u32>(gpu_addr, payload);
} }
} }
void RasterizerNull::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, void RasterizerNull::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr,
@ -74,7 +76,7 @@ void RasterizerNull::SignalSyncPoint(u32 value) {
syncpoint_manager.IncrementHost(value); syncpoint_manager.IncrementHost(value);
} }
void RasterizerNull::SignalReference() {} void RasterizerNull::SignalReference() {}
void RasterizerNull::ReleaseFences() {} void RasterizerNull::ReleaseFences(bool) {}
void RasterizerNull::FlushAndInvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType) {} void RasterizerNull::FlushAndInvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType) {}
void RasterizerNull::WaitForIdle() {} void RasterizerNull::WaitForIdle() {}
void RasterizerNull::FragmentBarrier() {} void RasterizerNull::FragmentBarrier() {}

View file

@ -42,8 +42,8 @@ public:
void DrawTexture() override; void DrawTexture() override;
void Clear(u32 layer_count) override; void Clear(u32 layer_count) override;
void DispatchCompute() override; void DispatchCompute() override;
void ResetCounter(VideoCore::QueryType type) override; void ResetCounter(VideoCommon::QueryType type) override;
void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override; void Query(GPUVAddr gpu_addr, VideoCommon::QueryType type, VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) override;
void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override; void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override;
void DisableGraphicsUniformBuffer(size_t stage, u32 index) override; void DisableGraphicsUniformBuffer(size_t stage, u32 index) override;
void FlushAll() override; void FlushAll() override;
@ -63,7 +63,7 @@ public:
void SyncOperation(std::function<void()>&& func) override; void SyncOperation(std::function<void()>&& func) override;
void SignalSyncPoint(u32 value) override; void SignalSyncPoint(u32 value) override;
void SignalReference() override; void SignalReference() override;
void ReleaseFences() override; void ReleaseFences(bool force) override;
void FlushAndInvalidateRegion( void FlushAndInvalidateRegion(
VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override; VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
void WaitForIdle() override; void WaitForIdle() override;

View file

@ -27,7 +27,7 @@ constexpr GLenum GetTarget(VideoCore::QueryType type) {
} // Anonymous namespace } // Anonymous namespace
QueryCache::QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_) QueryCache::QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_)
: QueryCacheBase(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {} : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {}
QueryCache::~QueryCache() = default; QueryCache::~QueryCache() = default;

View file

@ -26,7 +26,7 @@ class RasterizerOpenGL;
using CounterStream = VideoCommon::CounterStreamBase<QueryCache, HostCounter>; using CounterStream = VideoCommon::CounterStreamBase<QueryCache, HostCounter>;
class QueryCache final class QueryCache final
: public VideoCommon::QueryCacheBase<QueryCache, CachedQuery, CounterStream, HostCounter> { : public VideoCommon::QueryCacheLegacy<QueryCache, CachedQuery, CounterStream, HostCounter> {
public: public:
explicit QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_); explicit QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_);
~QueryCache(); ~QueryCache();

View file

@ -396,13 +396,31 @@ void RasterizerOpenGL::DispatchCompute() {
has_written_global_memory |= pipeline->WritesGlobalMemory(); has_written_global_memory |= pipeline->WritesGlobalMemory();
} }
void RasterizerOpenGL::ResetCounter(VideoCore::QueryType type) { void RasterizerOpenGL::ResetCounter(VideoCommon::QueryType type) {
query_cache.ResetCounter(type); if (type == VideoCommon::QueryType::ZPassPixelCount64) {
query_cache.ResetCounter(VideoCore::QueryType::SamplesPassed);
}
} }
void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCore::QueryType type, void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCommon::QueryType type,
std::optional<u64> timestamp) { VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) {
query_cache.Query(gpu_addr, type, timestamp); if (type == VideoCommon::QueryType::ZPassPixelCount64) {
std::optional<u64> timestamp{True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout)
? std::make_optional<u64>(gpu.GetTicks()) : std:: nullopt };
if (True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout)) {
query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, {gpu.GetTicks()});
} else {
query_cache.Query(gpu_addr, VideoCore::QueryType::SamplesPassed, std::nullopt);
}
return;
}
if (True(flags & VideoCommon::QueryPropertiesFlags::HasTimeout)) {
u64 ticks = gpu.GetTicks();
gpu_memory->Write<u64>(gpu_addr + 8, ticks);
gpu_memory->Write<u64>(gpu_addr, static_cast<u64>(payload));
} else {
gpu_memory->Write<u32>(gpu_addr, payload);
}
} }
void RasterizerOpenGL::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, void RasterizerOpenGL::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr,
@ -573,8 +591,8 @@ void RasterizerOpenGL::SignalReference() {
fence_manager.SignalOrdering(); fence_manager.SignalOrdering();
} }
void RasterizerOpenGL::ReleaseFences() { void RasterizerOpenGL::ReleaseFences(bool force) {
fence_manager.WaitPendingFences(); fence_manager.WaitPendingFences(force);
} }
void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size, void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size,

View file

@ -86,8 +86,8 @@ public:
void DrawTexture() override; void DrawTexture() override;
void Clear(u32 layer_count) override; void Clear(u32 layer_count) override;
void DispatchCompute() override; void DispatchCompute() override;
void ResetCounter(VideoCore::QueryType type) override; void ResetCounter(VideoCommon::QueryType type) override;
void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override; void Query(GPUVAddr gpu_addr, VideoCommon::QueryType type, VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) override;
void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override; void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override;
void DisableGraphicsUniformBuffer(size_t stage, u32 index) override; void DisableGraphicsUniformBuffer(size_t stage, u32 index) override;
void FlushAll() override; void FlushAll() override;
@ -107,7 +107,7 @@ public:
void SyncOperation(std::function<void()>&& func) override; void SyncOperation(std::function<void()>&& func) override;
void SignalSyncPoint(u32 value) override; void SignalSyncPoint(u32 value) override;
void SignalReference() override; void SignalReference() override;
void ReleaseFences() override; void ReleaseFences(bool force = true) override;
void FlushAndInvalidateRegion( void FlushAndInvalidateRegion(
VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override; VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
void WaitForIdle() override; void WaitForIdle() override;

View file

@ -61,6 +61,9 @@ vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allo
if (device.IsExtTransformFeedbackSupported()) { if (device.IsExtTransformFeedbackSupported()) {
flags |= VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT; flags |= VK_BUFFER_USAGE_TRANSFORM_FEEDBACK_BUFFER_BIT_EXT;
} }
if (device.IsExtConditionalRendering()) {
flags |= VK_BUFFER_USAGE_CONDITIONAL_RENDERING_BIT_EXT;
}
const VkBufferCreateInfo buffer_ci = { const VkBufferCreateInfo buffer_ci = {
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO, .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr, .pNext = nullptr,

View file

@ -12,6 +12,7 @@
#include "common/common_types.h" #include "common/common_types.h"
#include "common/div_ceil.h" #include "common/div_ceil.h"
#include "video_core/host_shaders/astc_decoder_comp_spv.h" #include "video_core/host_shaders/astc_decoder_comp_spv.h"
#include "video_core/host_shaders/resolve_conditional_render_comp_spv.h"
#include "video_core/host_shaders/vulkan_quad_indexed_comp_spv.h" #include "video_core/host_shaders/vulkan_quad_indexed_comp_spv.h"
#include "video_core/host_shaders/vulkan_uint8_comp_spv.h" #include "video_core/host_shaders/vulkan_uint8_comp_spv.h"
#include "video_core/renderer_vulkan/vk_compute_pass.h" #include "video_core/renderer_vulkan/vk_compute_pass.h"
@ -302,6 +303,52 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
return {staging.buffer, staging.offset}; return {staging.buffer, staging.offset};
} }
ConditionalRenderingResolvePass::ConditionalRenderingResolvePass(const Device& device_,
Scheduler& scheduler_,
DescriptorPool& descriptor_pool_, ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
: ComputePass(device_, descriptor_pool_, INPUT_OUTPUT_DESCRIPTOR_SET_BINDINGS,
INPUT_OUTPUT_DESCRIPTOR_UPDATE_TEMPLATE, INPUT_OUTPUT_BANK_INFO, nullptr,
RESOLVE_CONDITIONAL_RENDER_COMP_SPV),
scheduler{scheduler_}, compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
void ConditionalRenderingResolvePass::Resolve(VkBuffer dst_buffer, VkBuffer src_buffer,
u32 src_offset, bool compare_to_zero) {
scheduler.RequestOutsideRenderPassOperationContext();
const size_t compare_size = compare_to_zero ? 8 : 24;
compute_pass_descriptor_queue.Acquire();
compute_pass_descriptor_queue.AddBuffer(src_buffer, src_offset, compare_size);
compute_pass_descriptor_queue.AddBuffer(dst_buffer, 0, sizeof(u32));
const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([this, descriptor_data](vk::CommandBuffer cmdbuf) {
static constexpr VkMemoryBarrier read_barrier{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT,
};
static constexpr VkMemoryBarrier write_barrier{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_CONDITIONAL_RENDERING_READ_BIT_EXT,
};
const VkDescriptorSet set = descriptor_allocator.Commit();
device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, read_barrier);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
cmdbuf.Dispatch(1, 1, 1);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
VK_PIPELINE_STAGE_CONDITIONAL_RENDERING_BIT_EXT, 0, write_barrier);
});
}
ASTCDecoderPass::ASTCDecoderPass(const Device& device_, Scheduler& scheduler_, ASTCDecoderPass::ASTCDecoderPass(const Device& device_, Scheduler& scheduler_,
DescriptorPool& descriptor_pool_, DescriptorPool& descriptor_pool_,
StagingBufferPool& staging_buffer_pool_, StagingBufferPool& staging_buffer_pool_,

View file

@ -82,6 +82,19 @@ private:
ComputePassDescriptorQueue& compute_pass_descriptor_queue; ComputePassDescriptorQueue& compute_pass_descriptor_queue;
}; };
class ConditionalRenderingResolvePass final : public ComputePass {
public:
explicit ConditionalRenderingResolvePass(
const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
void Resolve(VkBuffer dst_buffer, VkBuffer src_buffer, u32 src_offset, bool compare_to_zero);
private:
Scheduler& scheduler;
ComputePassDescriptorQueue& compute_pass_descriptor_queue;
};
class ASTCDecoderPass final : public ComputePass { class ASTCDecoderPass final : public ComputePass {
public: public:
explicit ASTCDecoderPass(const Device& device_, Scheduler& scheduler_, explicit ASTCDecoderPass(const Device& device_, Scheduler& scheduler_,

View file

@ -8,6 +8,7 @@
#include "video_core/fence_manager.h" #include "video_core/fence_manager.h"
#include "video_core/renderer_vulkan/vk_buffer_cache.h" #include "video_core/renderer_vulkan/vk_buffer_cache.h"
#include "video_core/renderer_vulkan/vk_texture_cache.h" #include "video_core/renderer_vulkan/vk_texture_cache.h"
#include "video_core/renderer_vulkan/vk_query_cache.h"
namespace Core { namespace Core {
class System; class System;
@ -20,7 +21,6 @@ class RasterizerInterface;
namespace Vulkan { namespace Vulkan {
class Device; class Device;
class QueryCache;
class Scheduler; class Scheduler;
class InnerFence : public VideoCommon::FenceBase { class InnerFence : public VideoCommon::FenceBase {

File diff suppressed because it is too large Load diff

View file

@ -1,101 +1,74 @@
// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later // SPDX-License-Identifier: GPL-3.0-or-later
#pragma once #pragma once
#include <cstddef>
#include <memory> #include <memory>
#include <utility>
#include <vector>
#include "common/common_types.h" #include "video_core/query_cache/query_cache_base.h"
#include "video_core/query_cache.h" #include "video_core/renderer_vulkan/vk_buffer_cache.h"
#include "video_core/renderer_vulkan/vk_resource_pool.h"
#include "video_core/vulkan_common/vulkan_wrapper.h"
namespace VideoCore { namespace VideoCore {
class RasterizerInterface; class RasterizerInterface;
} }
namespace VideoCommon {
class StreamerInterface;
}
namespace Vulkan { namespace Vulkan {
class CachedQuery;
class Device; class Device;
class HostCounter;
class QueryCache;
class Scheduler; class Scheduler;
class StagingBufferPool;
using CounterStream = VideoCommon::CounterStreamBase<QueryCache, HostCounter>; struct QueryCacheRuntimeImpl;
class QueryPool final : public ResourcePool { class QueryCacheRuntime {
public: public:
explicit QueryPool(const Device& device, Scheduler& scheduler, VideoCore::QueryType type); explicit QueryCacheRuntime(VideoCore::RasterizerInterface* rasterizer,
~QueryPool() override; Core::Memory::Memory& cpu_memory_,
Vulkan::BufferCache& buffer_cache_, const Device& device_,
const MemoryAllocator& memory_allocator_, Scheduler& scheduler_,
StagingBufferPool& staging_pool_,
ComputePassDescriptorQueue& compute_pass_descriptor_queue,
DescriptorPool& descriptor_pool);
~QueryCacheRuntime();
std::pair<VkQueryPool, u32> Commit(); template <typename SyncValuesType>
void SyncValues(std::span<SyncValuesType> values, VkBuffer base_src_buffer = nullptr);
void Reserve(std::pair<VkQueryPool, u32> query); void Barriers(bool is_prebarrier);
protected: void EndHostConditionalRendering();
void Allocate(std::size_t begin, std::size_t end) override;
void PauseHostConditionalRendering();
void ResumeHostConditionalRendering();
bool HostConditionalRenderingCompareValue(VideoCommon::LookupData object_1, bool qc_dirty);
bool HostConditionalRenderingCompareValues(VideoCommon::LookupData object_1,
VideoCommon::LookupData object_2, bool qc_dirty, bool equal_check);
VideoCommon::StreamerInterface* GetStreamerInterface(VideoCommon::QueryType query_type);
void Bind3DEngine(Tegra::Engines::Maxwell3D* maxwell3d);
template <typename Func>
void View3DRegs(Func&& func);
private: private:
static constexpr std::size_t GROW_STEP = 512; void HostConditionalRenderingCompareValueImpl(VideoCommon::LookupData object, bool is_equal);
void HostConditionalRenderingCompareBCImpl(VAddr address, bool is_equal);
const Device& device; friend struct QueryCacheRuntimeImpl;
const VideoCore::QueryType type; std::unique_ptr<QueryCacheRuntimeImpl> impl;
std::vector<vk::QueryPool> pools;
std::vector<bool> usage;
}; };
class QueryCache final struct QueryCacheParams {
: public VideoCommon::QueryCacheBase<QueryCache, CachedQuery, CounterStream, HostCounter> { using RuntimeType = Vulkan::QueryCacheRuntime;
public:
explicit QueryCache(VideoCore::RasterizerInterface& rasterizer_,
Core::Memory::Memory& cpu_memory_, const Device& device_,
Scheduler& scheduler_);
~QueryCache();
std::pair<VkQueryPool, u32> AllocateQuery(VideoCore::QueryType type);
void Reserve(VideoCore::QueryType type, std::pair<VkQueryPool, u32> query);
const Device& GetDevice() const noexcept {
return device;
}
Scheduler& GetScheduler() const noexcept {
return scheduler;
}
private:
const Device& device;
Scheduler& scheduler;
std::array<QueryPool, VideoCore::NumQueryTypes> query_pools;
}; };
class HostCounter final : public VideoCommon::HostCounterBase<QueryCache, HostCounter> { using QueryCache = VideoCommon::QueryCacheBase<QueryCacheParams>;
public:
explicit HostCounter(QueryCache& cache_, std::shared_ptr<HostCounter> dependency_,
VideoCore::QueryType type_);
~HostCounter();
void EndQuery();
private:
u64 BlockingQuery(bool async = false) const override;
QueryCache& cache;
const VideoCore::QueryType type;
const std::pair<VkQueryPool, u32> query;
const u64 tick;
};
class CachedQuery : public VideoCommon::CachedQueryBase<HostCounter> {
public:
explicit CachedQuery(QueryCache&, VideoCore::QueryType, VAddr cpu_addr_, u8* host_ptr_)
: CachedQueryBase{cpu_addr_, host_ptr_} {}
};
} // namespace Vulkan } // namespace Vulkan

View file

@ -24,6 +24,7 @@
#include "video_core/renderer_vulkan/vk_compute_pipeline.h" #include "video_core/renderer_vulkan/vk_compute_pipeline.h"
#include "video_core/renderer_vulkan/vk_descriptor_pool.h" #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
#include "video_core/renderer_vulkan/vk_pipeline_cache.h" #include "video_core/renderer_vulkan/vk_pipeline_cache.h"
#include "video_core/renderer_vulkan/vk_query_cache.h"
#include "video_core/renderer_vulkan/vk_rasterizer.h" #include "video_core/renderer_vulkan/vk_rasterizer.h"
#include "video_core/renderer_vulkan/vk_scheduler.h" #include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h" #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
@ -170,9 +171,11 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
buffer_cache_runtime(device, memory_allocator, scheduler, staging_pool, buffer_cache_runtime(device, memory_allocator, scheduler, staging_pool,
guest_descriptor_queue, compute_pass_descriptor_queue, descriptor_pool), guest_descriptor_queue, compute_pass_descriptor_queue, descriptor_pool),
buffer_cache(*this, cpu_memory_, buffer_cache_runtime), buffer_cache(*this, cpu_memory_, buffer_cache_runtime),
query_cache_runtime(this, cpu_memory_, buffer_cache, device, memory_allocator, scheduler,
staging_pool, compute_pass_descriptor_queue, descriptor_pool),
query_cache(gpu, *this, cpu_memory_, query_cache_runtime),
pipeline_cache(*this, device, scheduler, descriptor_pool, guest_descriptor_queue, pipeline_cache(*this, device, scheduler, descriptor_pool, guest_descriptor_queue,
render_pass_cache, buffer_cache, texture_cache, gpu.ShaderNotify()), render_pass_cache, buffer_cache, texture_cache, gpu.ShaderNotify()),
query_cache{*this, cpu_memory_, device, scheduler},
accelerate_dma(buffer_cache, texture_cache, scheduler), accelerate_dma(buffer_cache, texture_cache, scheduler),
fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache, device, scheduler), fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache, device, scheduler),
wfi_event(device.GetLogical().CreateEvent()) { wfi_event(device.GetLogical().CreateEvent()) {
@ -189,13 +192,15 @@ void RasterizerVulkan::PrepareDraw(bool is_indexed, Func&& draw_func) {
FlushWork(); FlushWork();
gpu_memory->FlushCaching(); gpu_memory->FlushCaching();
query_cache.NotifySegment(true);
#if ANDROID #if ANDROID
if (Settings::IsGPULevelHigh()) { if (Settings::IsGPULevelHigh()) {
// This is problematic on Android, disable on GPU Normal. // This is problematic on Android, disable on GPU Normal.
query_cache.UpdateCounters(); // query_cache.UpdateCounters();
} }
#else #else
query_cache.UpdateCounters(); // query_cache.UpdateCounters();
#endif #endif
GraphicsPipeline* const pipeline{pipeline_cache.CurrentGraphicsPipeline()}; GraphicsPipeline* const pipeline{pipeline_cache.CurrentGraphicsPipeline()};
@ -207,13 +212,12 @@ void RasterizerVulkan::PrepareDraw(bool is_indexed, Func&& draw_func) {
pipeline->SetEngine(maxwell3d, gpu_memory); pipeline->SetEngine(maxwell3d, gpu_memory);
pipeline->Configure(is_indexed); pipeline->Configure(is_indexed);
BeginTransformFeedback();
UpdateDynamicStates(); UpdateDynamicStates();
HandleTransformFeedback();
query_cache.CounterEnable(VideoCommon::QueryType::ZPassPixelCount64,
maxwell3d->regs.zpass_pixel_count_enable);
draw_func(); draw_func();
EndTransformFeedback();
} }
void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) { void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
@ -241,6 +245,14 @@ void RasterizerVulkan::DrawIndirect() {
const auto indirect_buffer = buffer_cache.GetDrawIndirectBuffer(); const auto indirect_buffer = buffer_cache.GetDrawIndirectBuffer();
const auto& buffer = indirect_buffer.first; const auto& buffer = indirect_buffer.first;
const auto& offset = indirect_buffer.second; const auto& offset = indirect_buffer.second;
if (params.is_byte_count) {
scheduler.Record([buffer_obj = buffer->Handle(), offset,
stride = params.stride](vk::CommandBuffer cmdbuf) {
cmdbuf.DrawIndirectByteCountEXT(1, 0, buffer_obj, offset, 0,
static_cast<u32>(stride));
});
return;
}
if (params.include_count) { if (params.include_count) {
const auto count = buffer_cache.GetDrawIndirectCount(); const auto count = buffer_cache.GetDrawIndirectCount();
const auto& draw_buffer = count.first; const auto& draw_buffer = count.first;
@ -280,13 +292,15 @@ void RasterizerVulkan::DrawTexture() {
SCOPE_EXIT({ gpu.TickWork(); }); SCOPE_EXIT({ gpu.TickWork(); });
FlushWork(); FlushWork();
query_cache.NotifySegment(true);
#if ANDROID #if ANDROID
if (Settings::IsGPULevelHigh()) { if (Settings::IsGPULevelHigh()) {
// This is problematic on Android, disable on GPU Normal. // This is problematic on Android, disable on GPU Normal.
query_cache.UpdateCounters(); // query_cache.UpdateCounters();
} }
#else #else
query_cache.UpdateCounters(); // query_cache.UpdateCounters();
#endif #endif
texture_cache.SynchronizeGraphicsDescriptors(); texture_cache.SynchronizeGraphicsDescriptors();
@ -294,6 +308,8 @@ void RasterizerVulkan::DrawTexture() {
UpdateDynamicStates(); UpdateDynamicStates();
query_cache.CounterEnable(VideoCommon::QueryType::ZPassPixelCount64,
maxwell3d->regs.zpass_pixel_count_enable);
const auto& draw_texture_state = maxwell3d->draw_manager->GetDrawTextureState(); const auto& draw_texture_state = maxwell3d->draw_manager->GetDrawTextureState();
const auto& sampler = texture_cache.GetGraphicsSampler(draw_texture_state.src_sampler); const auto& sampler = texture_cache.GetGraphicsSampler(draw_texture_state.src_sampler);
const auto& texture = texture_cache.GetImageView(draw_texture_state.src_texture); const auto& texture = texture_cache.GetImageView(draw_texture_state.src_texture);
@ -319,12 +335,16 @@ void RasterizerVulkan::Clear(u32 layer_count) {
#if ANDROID #if ANDROID
if (Settings::IsGPULevelHigh()) { if (Settings::IsGPULevelHigh()) {
// This is problematic on Android, disable on GPU Normal. // This is problematic on Android, disable on GPU Normal.
query_cache.UpdateCounters(); // query_cache.UpdateCounters();
} }
#else #else
query_cache.UpdateCounters(); // query_cache.UpdateCounters();
#endif #endif
query_cache.NotifySegment(true);
query_cache.CounterEnable(VideoCommon::QueryType::ZPassPixelCount64,
maxwell3d->regs.zpass_pixel_count_enable);
auto& regs = maxwell3d->regs; auto& regs = maxwell3d->regs;
const bool use_color = regs.clear_surface.R || regs.clear_surface.G || regs.clear_surface.B || const bool use_color = regs.clear_surface.R || regs.clear_surface.G || regs.clear_surface.B ||
regs.clear_surface.A; regs.clear_surface.A;
@ -482,13 +502,13 @@ void RasterizerVulkan::DispatchCompute() {
scheduler.Record([dim](vk::CommandBuffer cmdbuf) { cmdbuf.Dispatch(dim[0], dim[1], dim[2]); }); scheduler.Record([dim](vk::CommandBuffer cmdbuf) { cmdbuf.Dispatch(dim[0], dim[1], dim[2]); });
} }
void RasterizerVulkan::ResetCounter(VideoCore::QueryType type) { void RasterizerVulkan::ResetCounter(VideoCommon::QueryType type) {
query_cache.ResetCounter(type); query_cache.CounterReset(type);
} }
void RasterizerVulkan::Query(GPUVAddr gpu_addr, VideoCore::QueryType type, void RasterizerVulkan::Query(GPUVAddr gpu_addr, VideoCommon::QueryType type,
std::optional<u64> timestamp) { VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) {
query_cache.Query(gpu_addr, type, timestamp); query_cache.CounterReport(gpu_addr, type, flags, payload, subreport);
} }
void RasterizerVulkan::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, void RasterizerVulkan::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr,
@ -669,8 +689,8 @@ void RasterizerVulkan::SignalReference() {
fence_manager.SignalReference(); fence_manager.SignalReference();
} }
void RasterizerVulkan::ReleaseFences() { void RasterizerVulkan::ReleaseFences(bool force) {
fence_manager.WaitPendingFences(); fence_manager.WaitPendingFences(force);
} }
void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size, void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size,
@ -694,6 +714,8 @@ void RasterizerVulkan::WaitForIdle() {
flags |= VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT; flags |= VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT;
} }
query_cache.NotifyWFI();
scheduler.RequestOutsideRenderPassOperationContext(); scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([event = *wfi_event, flags](vk::CommandBuffer cmdbuf) { scheduler.Record([event = *wfi_event, flags](vk::CommandBuffer cmdbuf) {
cmdbuf.SetEvent(event, flags); cmdbuf.SetEvent(event, flags);
@ -737,19 +759,7 @@ void RasterizerVulkan::TickFrame() {
bool RasterizerVulkan::AccelerateConditionalRendering() { bool RasterizerVulkan::AccelerateConditionalRendering() {
gpu_memory->FlushCaching(); gpu_memory->FlushCaching();
if (Settings::IsGPULevelHigh()) { return query_cache.AccelerateHostConditionalRendering();
// TODO(Blinkhawk): Reimplement Host conditional rendering.
return false;
}
// Medium / Low Hack: stub any checks on queries written into the buffer cache.
const GPUVAddr condition_address{maxwell3d->regs.render_enable.Address()};
Maxwell::ReportSemaphore::Compare cmp;
if (gpu_memory->IsMemoryDirty(condition_address, sizeof(cmp),
VideoCommon::CacheType::BufferCache |
VideoCommon::CacheType::QueryCache)) {
return true;
}
return false;
} }
bool RasterizerVulkan::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surface& src, bool RasterizerVulkan::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surface& src,
@ -795,6 +805,7 @@ bool RasterizerVulkan::AccelerateDisplay(const Tegra::FramebufferConfig& config,
if (!image_view) { if (!image_view) {
return false; return false;
} }
query_cache.NotifySegment(false);
screen_info.image = image_view->ImageHandle(); screen_info.image = image_view->ImageHandle();
screen_info.image_view = image_view->Handle(Shader::TextureType::Color2D); screen_info.image_view = image_view->Handle(Shader::TextureType::Color2D);
screen_info.width = image_view->size.width; screen_info.width = image_view->size.width;
@ -933,31 +944,18 @@ void RasterizerVulkan::UpdateDynamicStates() {
} }
} }
void RasterizerVulkan::BeginTransformFeedback() { void RasterizerVulkan::HandleTransformFeedback() {
const auto& regs = maxwell3d->regs; const auto& regs = maxwell3d->regs;
if (regs.transform_feedback_enabled == 0) {
return;
}
if (!device.IsExtTransformFeedbackSupported()) { if (!device.IsExtTransformFeedbackSupported()) {
LOG_ERROR(Render_Vulkan, "Transform feedbacks used but not supported"); LOG_ERROR(Render_Vulkan, "Transform feedbacks used but not supported");
return; return;
} }
UNIMPLEMENTED_IF(regs.IsShaderConfigEnabled(Maxwell::ShaderType::TessellationInit) || query_cache.CounterEnable(VideoCommon::QueryType::StreamingByteCount,
regs.IsShaderConfigEnabled(Maxwell::ShaderType::Tessellation)); regs.transform_feedback_enabled);
scheduler.Record( if (regs.transform_feedback_enabled != 0) {
[](vk::CommandBuffer cmdbuf) { cmdbuf.BeginTransformFeedbackEXT(0, 0, nullptr, nullptr); }); UNIMPLEMENTED_IF(regs.IsShaderConfigEnabled(Maxwell::ShaderType::TessellationInit) ||
} regs.IsShaderConfigEnabled(Maxwell::ShaderType::Tessellation));
void RasterizerVulkan::EndTransformFeedback() {
const auto& regs = maxwell3d->regs;
if (regs.transform_feedback_enabled == 0) {
return;
} }
if (!device.IsExtTransformFeedbackSupported()) {
return;
}
scheduler.Record(
[](vk::CommandBuffer cmdbuf) { cmdbuf.EndTransformFeedbackEXT(0, 0, nullptr, nullptr); });
} }
void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs) { void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs) {

View file

@ -84,8 +84,8 @@ public:
void DrawTexture() override; void DrawTexture() override;
void Clear(u32 layer_count) override; void Clear(u32 layer_count) override;
void DispatchCompute() override; void DispatchCompute() override;
void ResetCounter(VideoCore::QueryType type) override; void ResetCounter(VideoCommon::QueryType type) override;
void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override; void Query(GPUVAddr gpu_addr, VideoCommon::QueryType type, VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport) override;
void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override; void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override;
void DisableGraphicsUniformBuffer(size_t stage, u32 index) override; void DisableGraphicsUniformBuffer(size_t stage, u32 index) override;
void FlushAll() override; void FlushAll() override;
@ -106,7 +106,7 @@ public:
void SyncOperation(std::function<void()>&& func) override; void SyncOperation(std::function<void()>&& func) override;
void SignalSyncPoint(u32 value) override; void SignalSyncPoint(u32 value) override;
void SignalReference() override; void SignalReference() override;
void ReleaseFences() override; void ReleaseFences(bool force = true) override;
void FlushAndInvalidateRegion( void FlushAndInvalidateRegion(
VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override; VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
void WaitForIdle() override; void WaitForIdle() override;
@ -146,9 +146,7 @@ private:
void UpdateDynamicStates(); void UpdateDynamicStates();
void BeginTransformFeedback(); void HandleTransformFeedback();
void EndTransformFeedback();
void UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs); void UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs); void UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs);
@ -195,8 +193,9 @@ private:
TextureCache texture_cache; TextureCache texture_cache;
BufferCacheRuntime buffer_cache_runtime; BufferCacheRuntime buffer_cache_runtime;
BufferCache buffer_cache; BufferCache buffer_cache;
PipelineCache pipeline_cache; QueryCacheRuntime query_cache_runtime;
QueryCache query_cache; QueryCache query_cache;
PipelineCache pipeline_cache;
AccelerateDMA accelerate_dma; AccelerateDMA accelerate_dma;
FenceManager fence_manager; FenceManager fence_manager;

View file

@ -243,10 +243,10 @@ void Scheduler::AllocateNewContext() {
#if ANDROID #if ANDROID
if (Settings::IsGPULevelHigh()) { if (Settings::IsGPULevelHigh()) {
// This is problematic on Android, disable on GPU Normal. // This is problematic on Android, disable on GPU Normal.
query_cache->UpdateCounters(); query_cache->NotifySegment(true);
} }
#else #else
query_cache->UpdateCounters(); query_cache->NotifySegment(true);
#endif #endif
} }
} }
@ -261,11 +261,12 @@ void Scheduler::EndPendingOperations() {
#if ANDROID #if ANDROID
if (Settings::IsGPULevelHigh()) { if (Settings::IsGPULevelHigh()) {
// This is problematic on Android, disable on GPU Normal. // This is problematic on Android, disable on GPU Normal.
query_cache->DisableStreams(); // query_cache->DisableStreams();
} }
#else #else
query_cache->DisableStreams(); // query_cache->DisableStreams();
#endif #endif
query_cache->NotifySegment(false);
EndRenderPass(); EndRenderPass();
} }

View file

@ -15,6 +15,7 @@
#include "common/common_types.h" #include "common/common_types.h"
#include "common/polyfill_thread.h" #include "common/polyfill_thread.h"
#include "video_core/renderer_vulkan/vk_master_semaphore.h" #include "video_core/renderer_vulkan/vk_master_semaphore.h"
#include "video_core/renderer_vulkan/vk_query_cache.h"
#include "video_core/vulkan_common/vulkan_wrapper.h" #include "video_core/vulkan_common/vulkan_wrapper.h"
namespace Vulkan { namespace Vulkan {
@ -24,7 +25,6 @@ class Device;
class Framebuffer; class Framebuffer;
class GraphicsPipeline; class GraphicsPipeline;
class StateTracker; class StateTracker;
class QueryCache;
/// The scheduler abstracts command buffer and fence management with an interface that's able to do /// The scheduler abstracts command buffer and fence management with an interface that's able to do
/// OpenGL-like operations on Vulkan command buffers. /// OpenGL-like operations on Vulkan command buffers.

View file

@ -61,6 +61,7 @@ VK_DEFINE_HANDLE(VmaAllocator)
// Define miscellaneous extensions which may be used by the implementation here. // Define miscellaneous extensions which may be used by the implementation here.
#define FOR_EACH_VK_EXTENSION(EXTENSION) \ #define FOR_EACH_VK_EXTENSION(EXTENSION) \
EXTENSION(EXT, CONDITIONAL_RENDERING, conditional_rendering) \
EXTENSION(EXT, CONSERVATIVE_RASTERIZATION, conservative_rasterization) \ EXTENSION(EXT, CONSERVATIVE_RASTERIZATION, conservative_rasterization) \
EXTENSION(EXT, DEPTH_RANGE_UNRESTRICTED, depth_range_unrestricted) \ EXTENSION(EXT, DEPTH_RANGE_UNRESTRICTED, depth_range_unrestricted) \
EXTENSION(EXT, MEMORY_BUDGET, memory_budget) \ EXTENSION(EXT, MEMORY_BUDGET, memory_budget) \
@ -93,6 +94,7 @@ VK_DEFINE_HANDLE(VmaAllocator)
// Define extensions where the absence of the extension may result in a degraded experience. // Define extensions where the absence of the extension may result in a degraded experience.
#define FOR_EACH_VK_RECOMMENDED_EXTENSION(EXTENSION_NAME) \ #define FOR_EACH_VK_RECOMMENDED_EXTENSION(EXTENSION_NAME) \
EXTENSION_NAME(VK_EXT_CONDITIONAL_RENDERING_EXTENSION_NAME) \
EXTENSION_NAME(VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME) \ EXTENSION_NAME(VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME) \
EXTENSION_NAME(VK_EXT_DEPTH_RANGE_UNRESTRICTED_EXTENSION_NAME) \ EXTENSION_NAME(VK_EXT_DEPTH_RANGE_UNRESTRICTED_EXTENSION_NAME) \
EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME) \ EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME) \
@ -536,6 +538,10 @@ public:
return extensions.shader_atomic_int64; return extensions.shader_atomic_int64;
} }
bool IsExtConditionalRendering() const {
return extensions.conditional_rendering;
}
bool HasTimelineSemaphore() const; bool HasTimelineSemaphore() const;
/// Returns the minimum supported version of SPIR-V. /// Returns the minimum supported version of SPIR-V.

View file

@ -75,6 +75,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkBeginCommandBuffer); X(vkBeginCommandBuffer);
X(vkBindBufferMemory); X(vkBindBufferMemory);
X(vkBindImageMemory); X(vkBindImageMemory);
X(vkCmdBeginConditionalRenderingEXT);
X(vkCmdBeginQuery); X(vkCmdBeginQuery);
X(vkCmdBeginRenderPass); X(vkCmdBeginRenderPass);
X(vkCmdBeginTransformFeedbackEXT); X(vkCmdBeginTransformFeedbackEXT);
@ -91,6 +92,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkCmdCopyBufferToImage); X(vkCmdCopyBufferToImage);
X(vkCmdCopyImage); X(vkCmdCopyImage);
X(vkCmdCopyImageToBuffer); X(vkCmdCopyImageToBuffer);
X(vkCmdCopyQueryPoolResults);
X(vkCmdDispatch); X(vkCmdDispatch);
X(vkCmdDispatchIndirect); X(vkCmdDispatchIndirect);
X(vkCmdDraw); X(vkCmdDraw);
@ -99,6 +101,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkCmdDrawIndexedIndirect); X(vkCmdDrawIndexedIndirect);
X(vkCmdDrawIndirectCount); X(vkCmdDrawIndirectCount);
X(vkCmdDrawIndexedIndirectCount); X(vkCmdDrawIndexedIndirectCount);
X(vkCmdEndConditionalRenderingEXT);
X(vkCmdEndQuery); X(vkCmdEndQuery);
X(vkCmdEndRenderPass); X(vkCmdEndRenderPass);
X(vkCmdEndTransformFeedbackEXT); X(vkCmdEndTransformFeedbackEXT);

View file

@ -185,6 +185,7 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkBeginCommandBuffer vkBeginCommandBuffer{}; PFN_vkBeginCommandBuffer vkBeginCommandBuffer{};
PFN_vkBindBufferMemory vkBindBufferMemory{}; PFN_vkBindBufferMemory vkBindBufferMemory{};
PFN_vkBindImageMemory vkBindImageMemory{}; PFN_vkBindImageMemory vkBindImageMemory{};
PFN_vkCmdBeginConditionalRenderingEXT vkCmdBeginConditionalRenderingEXT{};
PFN_vkCmdBeginDebugUtilsLabelEXT vkCmdBeginDebugUtilsLabelEXT{}; PFN_vkCmdBeginDebugUtilsLabelEXT vkCmdBeginDebugUtilsLabelEXT{};
PFN_vkCmdBeginQuery vkCmdBeginQuery{}; PFN_vkCmdBeginQuery vkCmdBeginQuery{};
PFN_vkCmdBeginRenderPass vkCmdBeginRenderPass{}; PFN_vkCmdBeginRenderPass vkCmdBeginRenderPass{};
@ -202,6 +203,7 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkCmdCopyBufferToImage vkCmdCopyBufferToImage{}; PFN_vkCmdCopyBufferToImage vkCmdCopyBufferToImage{};
PFN_vkCmdCopyImage vkCmdCopyImage{}; PFN_vkCmdCopyImage vkCmdCopyImage{};
PFN_vkCmdCopyImageToBuffer vkCmdCopyImageToBuffer{}; PFN_vkCmdCopyImageToBuffer vkCmdCopyImageToBuffer{};
PFN_vkCmdCopyQueryPoolResults vkCmdCopyQueryPoolResults{};
PFN_vkCmdDispatch vkCmdDispatch{}; PFN_vkCmdDispatch vkCmdDispatch{};
PFN_vkCmdDispatchIndirect vkCmdDispatchIndirect{}; PFN_vkCmdDispatchIndirect vkCmdDispatchIndirect{};
PFN_vkCmdDraw vkCmdDraw{}; PFN_vkCmdDraw vkCmdDraw{};
@ -210,6 +212,7 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkCmdDrawIndexedIndirect vkCmdDrawIndexedIndirect{}; PFN_vkCmdDrawIndexedIndirect vkCmdDrawIndexedIndirect{};
PFN_vkCmdDrawIndirectCount vkCmdDrawIndirectCount{}; PFN_vkCmdDrawIndirectCount vkCmdDrawIndirectCount{};
PFN_vkCmdDrawIndexedIndirectCount vkCmdDrawIndexedIndirectCount{}; PFN_vkCmdDrawIndexedIndirectCount vkCmdDrawIndexedIndirectCount{};
PFN_vkCmdEndConditionalRenderingEXT vkCmdEndConditionalRenderingEXT{};
PFN_vkCmdEndDebugUtilsLabelEXT vkCmdEndDebugUtilsLabelEXT{}; PFN_vkCmdEndDebugUtilsLabelEXT vkCmdEndDebugUtilsLabelEXT{};
PFN_vkCmdEndQuery vkCmdEndQuery{}; PFN_vkCmdEndQuery vkCmdEndQuery{};
PFN_vkCmdEndRenderPass vkCmdEndRenderPass{}; PFN_vkCmdEndRenderPass vkCmdEndRenderPass{};
@ -1270,6 +1273,13 @@ public:
regions.data()); regions.data());
} }
void CopyQueryPoolResults(VkQueryPool query_pool, u32 first_query, u32 query_count,
VkBuffer dst_buffer, VkDeviceSize dst_offset, VkDeviceSize stride,
VkQueryResultFlags flags) const noexcept {
dld->vkCmdCopyQueryPoolResults(handle, query_pool, first_query, query_count, dst_buffer,
dst_offset, stride, flags);
}
void FillBuffer(VkBuffer dst_buffer, VkDeviceSize dst_offset, VkDeviceSize size, void FillBuffer(VkBuffer dst_buffer, VkDeviceSize dst_offset, VkDeviceSize size,
u32 data) const noexcept { u32 data) const noexcept {
dld->vkCmdFillBuffer(handle, dst_buffer, dst_offset, size, data); dld->vkCmdFillBuffer(handle, dst_buffer, dst_offset, size, data);
@ -1448,6 +1458,15 @@ public:
counter_buffers, counter_buffer_offsets); counter_buffers, counter_buffer_offsets);
} }
void BeginConditionalRenderingEXT(
const VkConditionalRenderingBeginInfoEXT& info) const noexcept {
dld->vkCmdBeginConditionalRenderingEXT(handle, &info);
}
void EndConditionalRenderingEXT() const noexcept {
dld->vkCmdEndConditionalRenderingEXT(handle);
}
void BeginDebugUtilsLabelEXT(const char* label, std::span<float, 4> color) const noexcept { void BeginDebugUtilsLabelEXT(const char* label, std::span<float, 4> color) const noexcept {
const VkDebugUtilsLabelEXT label_info{ const VkDebugUtilsLabelEXT label_info{
.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT, .sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT,