N-archive/citra
1
0
Fork 0
Code Issues Pull requests Packages Projects Releases Wiki Activity

audio_core: Replace AAC decoders with single FAAD2-based decoder. (#7098)

Browse source
This commit is contained in:
Steveice10 2023-11-04 14:56:13 -07:00 committed by GitHub
parent 1570aeffcb
commit 27bad3a699
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
28 changed files with 304 additions and 2403 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
.gitmodules vendored
View file

@ -79,6 +79,9 @@
[submodule "sirit"]
path = externals/sirit
url = https://github.com/yuzu-emu/sirit
[submodule "faad2"]
path = externals/faad2/faad2
url = https://github.com/knik0/faad2
[submodule "library-headers"]
path = externals/library-headers
url = https://github.com/citra-emu/ext-library-headers.git

3
CMakeLists.txt
View file

@ -81,9 +81,6 @@ CMAKE_DEPENDENT_OPTION(ENABLE_LIBUSB "Enable libusb for GameCube Adapter support
option(USE_DISCORD_PRESENCE "Enables Discord Rich Presence" OFF)
CMAKE_DEPENDENT_OPTION(ENABLE_MF "Use Media Foundation decoder (preferred over FFmpeg)" ON "WIN32" OFF)
CMAKE_DEPENDENT_OPTION(ENABLE_AUDIOTOOLBOX "Use AudioToolbox decoder (preferred over FFmpeg)" ON "APPLE" OFF)
CMAKE_DEPENDENT_OPTION(CITRA_ENABLE_BUNDLE_TARGET "Enable the distribution bundling target." ON "NOT ANDROID AND NOT IOS" OFF)
# Compile options

18
externals/CMakeLists.txt vendored
View file

@ -156,24 +156,12 @@ endif()
# Open Source Archives
add_subdirectory(open_source_archives)
# faad2
add_subdirectory(faad2 EXCLUDE_FROM_ALL)
# Dynamic library headers
add_library(library-headers INTERFACE)
if (USE_SYSTEM_FDK_AAC_HEADERS)
find_path(SYSTEM_FDK_AAC_INCLUDES NAMES fdk-aac/aacdecoder_lib.h)
if (SYSTEM_FDK_AAC_INCLUDES STREQUAL "SYSTEM_FDK_AAC_INCLUDES-NOTFOUND")
message(WARNING "System fdk-aac headers not found. Falling back on bundled headers.")
else()
message(STATUS "Using system fdk_aac headers.")
target_include_directories(library-headers SYSTEM INTERFACE ${SYSTEM_FDK_AAC_INCLUDES})
set(FOUND_FDK_AAC_HEADERS ON)
endif()
endif()
if (NOT FOUND_FDK_AAC_HEADERS)
message(STATUS "Using bundled fdk_aac headers.")
target_include_directories(library-headers SYSTEM INTERFACE ./library-headers/fdk-aac/include)
endif()
if (USE_SYSTEM_FFMPEG_HEADERS)
find_path(SYSTEM_FFMPEG_INCLUDES NAMES libavutil/avutil.h)
if (SYSTEM_FFMPEG_INCLUDES STREQUAL "SYSTEM_FFMPEG_INCLUDES-NOTFOUND")

3
externals/cmake-modules/CitraHandleSystemLibs.cmake vendored
View file

@ -15,7 +15,6 @@ option(USE_SYSTEM_DYNARMIC "Use the system dynarmic (instead of the bundled one)
option(USE_SYSTEM_FMT "Use the system fmt (instead of the bundled one)" OFF)
option(USE_SYSTEM_XBYAK "Use the system xbyak (instead of the bundled one)" OFF)
option(USE_SYSTEM_INIH "Use the system inih (instead of the bundled one)" OFF)
option(USE_SYSTEM_FDK_AAC_HEADERS "Use the system fdk-aac headers (instead of the bundled one)" OFF)
option(USE_SYSTEM_FFMPEG_HEADERS "Use the system FFmpeg headers (instead of the bundled one)" OFF)
option(USE_SYSTEM_GLSLANG "Use the system glslang and SPIR-V libraries (instead of the bundled ones)" OFF)
option(USE_SYSTEM_ZSTD "Use the system Zstandard library (instead of the bundled one)" OFF)
@ -36,7 +35,6 @@ CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_DYNARMIC "Disable system Dynarmic" OFF "US
CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_FMT "Disable system fmt" OFF "USE_SYSTEM_LIBS" OFF)
CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_XBYAK "Disable system xbyak" OFF "USE_SYSTEM_LIBS" OFF)
CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_INIH "Disable system inih" OFF "USE_SYSTEM_LIBS" OFF)
CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_FDK_AAC_HEADERS "Disable system fdk_aac" OFF "USE_SYSTEM_LIBS" OFF)
CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_FFMPEG_HEADERS "Disable system ffmpeg" OFF "USE_SYSTEM_LIBS" OFF)
CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_GLSLANG "Disable system glslang" OFF "USE_SYSTEM_LIBS" OFF)
CMAKE_DEPENDENT_OPTION(DISABLE_SYSTEM_ZSTD "Disable system Zstandard" OFF "USE_SYSTEM_LIBS" OFF)
@ -57,7 +55,6 @@ set(LIB_VAR_LIST
FMT
XBYAK
INIH
FDK_AAC_HEADERS
FFMPEG_HEADERS
GLSLANG
ZSTD

102
externals/faad2/CMakeLists.txt vendored Normal file
View file

@ -0,0 +1,102 @@
# Copy source to build directory for some modifications.
set(FAAD2_SOURCE_DIR "${CMAKE_CURRENT_BINARY_DIR}/faad2/libfaad")
if (NOT EXISTS "${FAAD2_SOURCE_DIR}")
file(COPY faad2/libfaad/ DESTINATION "${FAAD2_SOURCE_DIR}/")
# These are fixed defines for some reason and not controllable with compile flags.
file(READ "${FAAD2_SOURCE_DIR}/common.h" FAAD2_COMMON_H)
# Disable SBR decoding since we don't want it for AAC-LC.
string(REGEX REPLACE "#define SBR_DEC" "" FAAD2_COMMON_H "${FAAD2_COMMON_H}")
# Disable PS decoding. This can cause mono to be upmixed to stereo, which we don't want.
string(REGEX REPLACE "#define PS_DEC" "" FAAD2_COMMON_H "${FAAD2_COMMON_H}")
file(WRITE "${FAAD2_SOURCE_DIR}/common.h" "${FAAD2_COMMON_H}")
endif()
# Source list from faad2/libfaad/Makefile.am, cut down to just what we need for AAC-LC.
add_library(faad2 STATIC EXCLUDE_FROM_ALL
"${FAAD2_SOURCE_DIR}/bits.c"
"${FAAD2_SOURCE_DIR}/cfft.c"
"${FAAD2_SOURCE_DIR}/common.c"
"${FAAD2_SOURCE_DIR}/decoder.c"
"${FAAD2_SOURCE_DIR}/drc.c"
"${FAAD2_SOURCE_DIR}/error.c"
"${FAAD2_SOURCE_DIR}/filtbank.c"
"${FAAD2_SOURCE_DIR}/huffman.c"
"${FAAD2_SOURCE_DIR}/is.c"
"${FAAD2_SOURCE_DIR}/mdct.c"
"${FAAD2_SOURCE_DIR}/mp4.c"
"${FAAD2_SOURCE_DIR}/ms.c"
"${FAAD2_SOURCE_DIR}/output.c"
"${FAAD2_SOURCE_DIR}/pns.c"
"${FAAD2_SOURCE_DIR}/pulse.c"
"${FAAD2_SOURCE_DIR}/specrec.c"
"${FAAD2_SOURCE_DIR}/syntax.c"
"${FAAD2_SOURCE_DIR}/tns.c"
)
target_include_directories(faad2 PUBLIC faad2/include PRIVATE "${FAAD2_SOURCE_DIR}")
# Configure compile definitions.
# Read version from autoconf script for configuring constant.
file(READ faad2/configure.ac CONFIGURE_SCRIPT)
string(REGEX MATCH "AC_INIT\\(faad2, ([0-9.]+)\\)" _ ${CONFIGURE_SCRIPT})
set(FAAD_VERSION ${CMAKE_MATCH_1})
message(STATUS "Building faad2 version ${FAAD_VERSION}")
# Check for functions and headers.
include(CheckFunctionExists)
include(CheckIncludeFiles)
check_function_exists(getpwuid HAVE_GETPWUID)
check_function_exists(lrintf HAVE_LRINTF)
check_function_exists(memcpy HAVE_MEMCPY)
check_function_exists(strchr HAVE_STRCHR)
check_function_exists(strsep HAVE_STRSEP)
check_include_files(dlfcn.h HAVE_DLFCN_H)
check_include_files(errno.h HAVE_ERRNO_H)
check_include_files(float.h HAVE_FLOAT_H)
check_include_files(inttypes.h HAVE_INTTYPES_H)
check_include_files(IOKit/IOKitLib.h HAVE_IOKIT_IOKITLIB_H)
check_include_files(limits.h HAVE_LIMITS_H)
check_include_files(mathf.h HAVE_MATHF_H)
check_include_files(stdint.h HAVE_STDINT_H)
check_include_files(stdio.h HAVE_STDIO_H)
check_include_files(stdlib.h HAVE_STDLIB_H)
check_include_files(strings.h HAVE_STRINGS_H)
check_include_files(string.h HAVE_STRING_H)
check_include_files(sysfs/libsysfs.h HAVE_SYSFS_LIBSYSFS_H)
check_include_files(sys/stat.h HAVE_SYS_STAT_H)
check_include_files(sys/time.h HAVE_SYS_TIME_H)
check_include_files(sys/types.h HAVE_SYS_TYPES_H)
check_include_files(unistd.h HAVE_UNISTD_H)
# faad2 uses a relative include for its config.h which breaks under CMake.
# We can use target_compile_definitions to pass on the configuration instead.
target_compile_definitions(faad2 PRIVATE
-DFAAD_VERSION=${FAAD_VERSION}
-DPACKAGE_VERSION=\"${FAAD_VERSION}\"
-DSTDC_HEADERS
-DHAVE_GETPWUID=${HAVE_GETPWUID}
-DHAVE_LRINTF=${HAVE_LRINTF}
-DHAVE_MEMCPY=${HAVE_MEMCPY}
-DHAVE_STRCHR=${HAVE_STRCHR}
-DHAVE_STRSEP=${HAVE_STRSEP}
-DHAVE_DLFCN_H=${HAVE_DLFCN_H}
-DHAVE_ERRNO_H=${HAVE_ERRNO_H}
-DHAVE_FLOAT_H=${HAVE_FLOAT_H}
-DHAVE_INTTYPES_H=${HAVE_INTTYPES_H}
-DHAVE_IOKIT_IOKITLIB_H=${HAVE_IOKIT_IOKITLIB_H}
-DHAVE_LIMITS_H=${HAVE_LIMITS_H}
-DHAVE_MATHF_H=${HAVE_MATHF_H}
-DHAVE_STDINT_H=${HAVE_STDINT_H}
-DHAVE_STDIO_H=${HAVE_STDIO_H}
-DHAVE_STDLIB_H=${HAVE_STDLIB_H}
-DHAVE_STRINGS_H=${HAVE_STRINGS_H}
-DHAVE_STRING_H=${HAVE_STRING_H}
-DHAVE_SYSFS_LIBSYSFS_H=${HAVE_SYSFS_LIBSYSFS_H}
-DHAVE_SYS_STAT_H=${HAVE_SYS_STAT_H}
-DHAVE_SYS_TIME_H=${HAVE_SYS_TIME_H}
-DHAVE_SYS_TYPES_H=${HAVE_SYS_TYPES_H}
-DHAVE_UNISTD_H=${HAVE_UNISTD_H}
# Only compile for AAC-LC decoding.
-DLC_ONLY_DECODER
)

1
externals/faad2/faad2 vendored Submodule

@ -0,0 +1 @@
Subproject commit 3918dee56063500d0aa23d6c3c94b211ac471a8c

39
src/audio_core/CMakeLists.txt
View file

@ -4,15 +4,11 @@ add_library(audio_core STATIC
codec.h
dsp_interface.cpp
dsp_interface.h
hle/adts.h
hle/adts_reader.cpp
hle/common.h
hle/decoder.cpp
hle/decoder.h
hle/fdk_decoder.cpp
hle/fdk_decoder.h
hle/ffmpeg_decoder.cpp
hle/ffmpeg_decoder.h
hle/faad2_decoder.cpp
hle/faad2_decoder.h
hle/filter.cpp
hle/filter.h
hle/hle.cpp
@ -48,36 +44,7 @@ add_library(audio_core STATIC
create_target_directory_groups(audio_core)
target_link_libraries(audio_core PUBLIC citra_common citra_core)
target_link_libraries(audio_core PRIVATE SoundTouch teakra)
if(ENABLE_MF)
target_sources(audio_core PRIVATE
hle/wmf_decoder.cpp
hle/wmf_decoder.h
hle/wmf_decoder_utils.cpp
hle/wmf_decoder_utils.h
)
# We dynamically load the required symbols from mf.dll and mfplat.dll but mfuuid is not a dll
# just a static library of GUIDS so include that one directly.
target_link_libraries(audio_core PRIVATE mfuuid.lib)
target_compile_definitions(audio_core PUBLIC HAVE_MF)
elseif(ENABLE_AUDIOTOOLBOX)
target_sources(audio_core PRIVATE
hle/audiotoolbox_decoder.cpp
hle/audiotoolbox_decoder.h
)
find_library(AUDIOTOOLBOX AudioToolbox)
target_link_libraries(audio_core PRIVATE ${AUDIOTOOLBOX})
target_compile_definitions(audio_core PUBLIC HAVE_AUDIOTOOLBOX)
endif()
if(ANDROID)
target_sources(audio_core PRIVATE
hle/mediandk_decoder.cpp
hle/mediandk_decoder.h
)
target_link_libraries(audio_core PRIVATE mediandk)
endif()
target_link_libraries(audio_core PRIVATE faad2 SoundTouch teakra)
if(ENABLE_SDL2)
target_link_libraries(audio_core PRIVATE SDL2::SDL2)

28
src/audio_core/hle/adts.h
View file

@ -1,28 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include "common/common_types.h"
namespace AudioCore {
struct ADTSData {
u8 header_length = 0;
bool mpeg2 = false;
u8 profile = 0;
u8 channels = 0;
u8 channel_idx = 0;
u8 framecount = 0;
u8 samplerate_idx = 0;
u32 length = 0;
u32 samplerate = 0;
};
ADTSData ParseADTS(const u8* buffer);
// last two bytes of MF AAC decoder user data
// see https://docs.microsoft.com/en-us/windows/desktop/medfound/aac-decoder#example-media-types
u16 MFGetAACTag(const ADTSData& input);
} // namespace AudioCore

79
src/audio_core/hle/adts_reader.cpp
View file

@ -1,79 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <array>
#include "adts.h"
#include "common/bit_field.h"
namespace AudioCore {
constexpr std::array<u32, 16> freq_table = {96000, 88200, 64000, 48000, 44100, 32000, 24000, 22050,
16000, 12000, 11025, 8000, 7350, 0, 0, 0};
constexpr std::array<u8, 8> channel_table = {0, 1, 2, 3, 4, 5, 6, 8};
struct ADTSHeader {
union {
std::array<u8, 7> raw{};
BitFieldBE<52, 12, u64> sync_word;
BitFieldBE<51, 1, u64> mpeg2;
BitFieldBE<49, 2, u64> layer;
BitFieldBE<48, 1, u64> protection_absent;
BitFieldBE<46, 2, u64> profile;
BitFieldBE<42, 4, u64> samplerate_idx;
BitFieldBE<41, 1, u64> private_bit;
BitFieldBE<38, 3, u64> channel_idx;
BitFieldBE<37, 1, u64> originality;
BitFieldBE<36, 1, u64> home;
BitFieldBE<35, 1, u64> copyright_id;
BitFieldBE<34, 1, u64> copyright_id_start;
BitFieldBE<21, 13, u64> frame_length;
BitFieldBE<10, 11, u64> buffer_fullness;
BitFieldBE<8, 2, u64> frame_count;
};
};
ADTSData ParseADTS(const u8* buffer) {
ADTSHeader header;
memcpy(header.raw.data(), buffer, sizeof(header.raw));
// sync word 0xfff
if (header.sync_word != 0xfff) {
return {};
}
ADTSData out{};
// bit 16 = no CRC
out.header_length = header.protection_absent ? 7 : 9;
out.mpeg2 = static_cast<bool>(header.mpeg2);
// bit 17 to 18
out.profile = static_cast<u8>(header.profile) + 1;
// bit 19 to 22
out.samplerate_idx = static_cast<u8>(header.samplerate_idx);
out.samplerate = header.samplerate_idx > 15 ? 0 : freq_table[header.samplerate_idx];
// bit 24 to 26
out.channel_idx = static_cast<u8>(header.channel_idx);
out.channels = (header.channel_idx > 7) ? 0 : channel_table[header.channel_idx];
// bit 55 to 56
out.framecount = static_cast<u8>(header.frame_count + 1);
// bit 31 to 43
out.length = static_cast<u32>(header.frame_length);
return out;
}
// last two bytes of MF AAC decoder user data
// Audio object type (5 bits)
// Sample rate profile (4 bits)
// Channel configuration profile (4 bits)
// Frame length flag (1 bit)
// Depends on core coder (1 bit)
// Extension flag (1 bit)
u16 MFGetAACTag(const ADTSData& input) {
u16 tag = 0;
tag |= input.profile << 11;
tag |= input.samplerate_idx << 7;
tag |= input.channel_idx << 3;
return tag;
}
} // namespace AudioCore

264
src/audio_core/hle/audiotoolbox_decoder.cpp
View file

@ -1,264 +0,0 @@
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <AudioToolbox/AudioToolbox.h>
#include "audio_core/audio_types.h"
#include "audio_core/hle/adts.h"
#include "audio_core/hle/audiotoolbox_decoder.h"
namespace AudioCore::HLE {
static constexpr auto bytes_per_sample = sizeof(s16);
static constexpr auto aac_frames_per_packet = 1024;
static constexpr auto error_out_of_data = -1932;
class AudioToolboxDecoder::Impl {
public:
explicit Impl(Memory::MemorySystem& memory);
~Impl();
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request);
private:
std::optional<BinaryMessage> Initalize(const BinaryMessage& request);
std::optional<BinaryMessage> Decode(const BinaryMessage& request);
void Clear();
bool InitializeDecoder(AudioCore::ADTSData& adts_header);
static OSStatus DataFunc(AudioConverterRef in_audio_converter, u32* io_number_data_packets,
AudioBufferList* io_data,
AudioStreamPacketDescription** out_data_packet_description,
void* in_user_data);
Memory::MemorySystem& memory;
AudioCore::ADTSData adts_config;
AudioStreamBasicDescription output_format = {};
AudioConverterRef converter = nullptr;
u8* curr_data = nullptr;
u32 curr_data_len = 0;
AudioStreamPacketDescription packet_description;
};
AudioToolboxDecoder::Impl::Impl(Memory::MemorySystem& memory_) : memory(memory_) {}
std::optional<BinaryMessage> AudioToolboxDecoder::Impl::Initalize(const BinaryMessage& request) {
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
Clear();
return response;
}
AudioToolboxDecoder::Impl::~Impl() {
Clear();
}
void AudioToolboxDecoder::Impl::Clear() {
curr_data = nullptr;
curr_data_len = 0;
adts_config = {};
output_format = {};
if (converter) {
AudioConverterDispose(converter);
converter = nullptr;
}
}
std::optional<BinaryMessage> AudioToolboxDecoder::Impl::ProcessRequest(
const BinaryMessage& request) {
if (request.header.codec != DecoderCodec::DecodeAAC) {
LOG_ERROR(Audio_DSP, "AudioToolbox AAC Decoder cannot handle such codec: {}",
static_cast<u16>(request.header.codec));
return {};
}
switch (request.header.cmd) {
case DecoderCommand::Init: {
return Initalize(request);
}
case DecoderCommand::EncodeDecode: {
return Decode(request);
}
case DecoderCommand::Shutdown:
case DecoderCommand::SaveState:
case DecoderCommand::LoadState: {
LOG_WARNING(Audio_DSP, "Got unimplemented binary request: {}",
static_cast<u16>(request.header.cmd));
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
return response;
}
default:
LOG_ERROR(Audio_DSP, "Got unknown binary request: {}",
static_cast<u16>(request.header.cmd));
return {};
}
}
bool AudioToolboxDecoder::Impl::InitializeDecoder(AudioCore::ADTSData& adts_header) {
if (converter) {
if (adts_config.channels == adts_header.channels &&
adts_config.samplerate == adts_header.samplerate) {
return true;
} else {
Clear();
}
}
AudioStreamBasicDescription input_format = {
.mSampleRate = static_cast<Float64>(adts_header.samplerate),
.mFormatID = kAudioFormatMPEG4AAC,
.mFramesPerPacket = aac_frames_per_packet,
.mChannelsPerFrame = adts_header.channels,
};
u32 bytes_per_frame = input_format.mChannelsPerFrame * bytes_per_sample;
output_format = {
.mSampleRate = input_format.mSampleRate,
.mFormatID = kAudioFormatLinearPCM,
.mFormatFlags = kLinearPCMFormatFlagIsSignedInteger | kLinearPCMFormatFlagIsPacked,
.mBytesPerPacket = bytes_per_frame,
.mFramesPerPacket = 1,
.mBytesPerFrame = bytes_per_frame,
.mChannelsPerFrame = input_format.mChannelsPerFrame,
.mBitsPerChannel = bytes_per_sample * 8,
};
auto status = AudioConverterNew(&input_format, &output_format, &converter);
if (status != noErr) {
LOG_ERROR(Audio_DSP, "Could not create AAC audio converter: {}", status);
Clear();
return false;
}
adts_config = adts_header;
return true;
}
OSStatus AudioToolboxDecoder::Impl::DataFunc(
AudioConverterRef in_audio_converter, u32* io_number_data_packets, AudioBufferList* io_data,
AudioStreamPacketDescription** out_data_packet_description, void* in_user_data) {
auto impl = reinterpret_cast<Impl*>(in_user_data);
if (!impl || !impl->curr_data || impl->curr_data_len == 0) {
*io_number_data_packets = 0;
return error_out_of_data;
}
io_data->mNumberBuffers = 1;
io_data->mBuffers[0].mNumberChannels = 0;
io_data->mBuffers[0].mDataByteSize = impl->curr_data_len;
io_data->mBuffers[0].mData = impl->curr_data;
*io_number_data_packets = 1;
if (out_data_packet_description != nullptr) {
impl->packet_description.mStartOffset = 0;
impl->packet_description.mVariableFramesInPacket = 0;
impl->packet_description.mDataByteSize = impl->curr_data_len;
*out_data_packet_description = &impl->packet_description;
}
impl->curr_data = nullptr;
impl->curr_data_len = 0;
return noErr;
}
std::optional<BinaryMessage> AudioToolboxDecoder::Impl::Decode(const BinaryMessage& request) {
BinaryMessage response{};
response.header.codec = request.header.codec;
response.header.cmd = request.header.cmd;
response.decode_aac_response.size = request.decode_aac_request.size;
if (request.decode_aac_request.src_addr < Memory::FCRAM_PADDR ||
request.decode_aac_request.src_addr + request.decode_aac_request.size >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds src_addr {:08x}",
request.decode_aac_request.src_addr);
return {};
}
const auto data =
memory.GetFCRAMPointer(request.decode_aac_request.src_addr - Memory::FCRAM_PADDR);
auto adts_header = AudioCore::ParseADTS(data);
curr_data = data + adts_header.header_length;
curr_data_len = request.decode_aac_request.size - adts_header.header_length;
if (!InitializeDecoder(adts_header)) {
return std::nullopt;
}
// Up to 2048 samples, up to 2 channels each
s16 decoder_output[4096];
AudioBufferList out_buffer{1,
{{
output_format.mChannelsPerFrame,
sizeof(decoder_output),
decoder_output,
}}};
u32 num_packets = sizeof(decoder_output) / output_format.mBytesPerPacket;
auto status = AudioConverterFillComplexBuffer(converter, DataFunc, this, &num_packets,
&out_buffer, nullptr);
if (status != noErr && status != error_out_of_data) {
LOG_ERROR(Audio_DSP, "Could not decode AAC data: {}", status);
Clear();
return std::nullopt;
}
// De-interleave samples.
std::array<std::vector<s16>, 2> out_streams;
auto num_frames = num_packets * output_format.mFramesPerPacket;
for (u32 frame = 0; frame < num_frames; frame++) {
for (u32 ch = 0; ch < output_format.mChannelsPerFrame; ch++) {
out_streams[ch].push_back(
decoder_output[(frame * output_format.mChannelsPerFrame) + ch]);
}
}
curr_data = nullptr;
curr_data_len = 0;
response.decode_aac_response.sample_rate =
GetSampleRateEnum(static_cast<u32>(output_format.mSampleRate));
response.decode_aac_response.num_channels = output_format.mChannelsPerFrame;
response.decode_aac_response.num_samples = num_frames;
// transfer the decoded buffer from vector to the FCRAM
for (std::size_t ch = 0; ch < out_streams.size(); ch++) {
if (!out_streams[ch].empty()) {
auto byte_size = out_streams[ch].size() * bytes_per_sample;
auto dst = ch == 0 ? request.decode_aac_request.dst_addr_ch0
: request.decode_aac_request.dst_addr_ch1;
if (dst < Memory::FCRAM_PADDR ||
dst + byte_size > Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds dst_addr_ch{} {:08x}", ch, dst);
return {};
}
std::memcpy(memory.GetFCRAMPointer(dst - Memory::FCRAM_PADDR), out_streams[ch].data(),
byte_size);
}
}
return response;
}
AudioToolboxDecoder::AudioToolboxDecoder(Memory::MemorySystem& memory)
: impl(std::make_unique<Impl>(memory)) {}
AudioToolboxDecoder::~AudioToolboxDecoder() = default;
std::optional<BinaryMessage> AudioToolboxDecoder::ProcessRequest(const BinaryMessage& request) {
return impl->ProcessRequest(request);
}
bool AudioToolboxDecoder::IsValid() const {
return true;
}
} // namespace AudioCore::HLE

186
src/audio_core/hle/faad2_decoder.cpp Normal file
View file

@ -0,0 +1,186 @@
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <neaacdec.h>
#include "audio_core/hle/faad2_decoder.h"
namespace AudioCore::HLE {
class FAAD2Decoder::Impl {
public:
explicit Impl(Memory::MemorySystem& memory);
~Impl();
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request);
bool IsValid() const {
return decoder != nullptr;
}
private:
std::optional<BinaryMessage> Initalize(const BinaryMessage& request);
std::optional<BinaryMessage> Decode(const BinaryMessage& request);
Memory::MemorySystem& memory;
NeAACDecHandle decoder = nullptr;
};
FAAD2Decoder::Impl::Impl(Memory::MemorySystem& memory) : memory(memory) {
decoder = NeAACDecOpen();
if (decoder == nullptr) {
LOG_CRITICAL(Audio_DSP, "Could not open FAAD2 decoder.");
return;
}
auto config = NeAACDecGetCurrentConfiguration(decoder);
config->defObjectType = LC;
config->outputFormat = FAAD_FMT_16BIT;
if (!NeAACDecSetConfiguration(decoder, config)) {
LOG_CRITICAL(Audio_DSP, "Could not configure FAAD2 decoder.");
NeAACDecClose(decoder);
decoder = nullptr;
return;
}
LOG_INFO(Audio_DSP, "Created FAAD2 AAC decoder.");
}
FAAD2Decoder::Impl::~Impl() {
if (decoder) {
NeAACDecClose(decoder);
decoder = nullptr;
LOG_INFO(Audio_DSP, "Destroyed FAAD2 AAC decoder.");
}
}
std::optional<BinaryMessage> FAAD2Decoder::Impl::ProcessRequest(const BinaryMessage& request) {
if (request.header.codec != DecoderCodec::DecodeAAC) {
LOG_ERROR(Audio_DSP, "FAAD2 AAC Decoder cannot handle such codec: {}",
static_cast<u16>(request.header.codec));
return {};
}
switch (request.header.cmd) {
case DecoderCommand::Init: {
return Initalize(request);
}
case DecoderCommand::EncodeDecode: {
return Decode(request);
}
case DecoderCommand::Shutdown:
case DecoderCommand::SaveState:
case DecoderCommand::LoadState: {
LOG_WARNING(Audio_DSP, "Got unimplemented binary request: {}",
static_cast<u16>(request.header.cmd));
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
return response;
}
default:
LOG_ERROR(Audio_DSP, "Got unknown binary request: {}",
static_cast<u16>(request.header.cmd));
return {};
}
}
std::optional<BinaryMessage> FAAD2Decoder::Impl::Initalize(const BinaryMessage& request) {
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
return response;
}
std::optional<BinaryMessage> FAAD2Decoder::Impl::Decode(const BinaryMessage& request) {
BinaryMessage response{};
response.header.codec = request.header.codec;
response.header.cmd = request.header.cmd;
response.decode_aac_response.size = request.decode_aac_request.size;
// This is a hack to continue games when a failure occurs.
response.decode_aac_response.sample_rate = DecoderSampleRate::Rate48000;
response.decode_aac_response.num_channels = 2;
response.decode_aac_response.num_samples = 1024;
if (request.decode_aac_request.src_addr < Memory::FCRAM_PADDR ||
request.decode_aac_request.src_addr + request.decode_aac_request.size >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds src_addr {:08x}",
request.decode_aac_request.src_addr);
return response;
}
u8* data = memory.GetFCRAMPointer(request.decode_aac_request.src_addr - Memory::FCRAM_PADDR);
u32 data_len = request.decode_aac_request.size;
unsigned long sample_rate;
u8 num_channels;
auto init_result = NeAACDecInit(decoder, data, data_len, &sample_rate, &num_channels);
if (init_result < 0) {
LOG_ERROR(Audio_DSP, "Could not initialize FAAD2 AAC decoder for request: {}", init_result);
return response;
}
// Advance past the frame header if needed.
data += init_result;
data_len -= init_result;
std::array<std::vector<s16>, 2> out_streams;
while (data_len > 0) {
NeAACDecFrameInfo frame_info;
auto curr_sample_buffer =
static_cast<s16*>(NeAACDecDecode(decoder, &frame_info, data, data_len));
if (curr_sample_buffer == nullptr || frame_info.error != 0) {
LOG_ERROR(Audio_DSP, "Failed to decode AAC buffer using FAAD2: {}", frame_info.error);
return response;
}
// Split the decode result into channels.
u32 num_samples = frame_info.samples / frame_info.channels;
for (u32 sample = 0; sample < num_samples; sample++) {
for (u32 ch = 0; ch < frame_info.channels; ch++) {
out_streams[ch].push_back(curr_sample_buffer[(sample * frame_info.channels) + ch]);
}
}
data += frame_info.bytesconsumed;
data_len -= frame_info.bytesconsumed;
}
// Transfer the decoded buffer from vector to the FCRAM.
for (std::size_t ch = 0; ch < out_streams.size(); ch++) {
if (out_streams[ch].empty()) {
continue;
}
auto byte_size = out_streams[ch].size() * sizeof(s16);
auto dst = ch == 0 ? request.decode_aac_request.dst_addr_ch0
: request.decode_aac_request.dst_addr_ch1;
if (dst < Memory::FCRAM_PADDR ||
dst + byte_size > Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds dst_addr_ch{} {:08x}", ch, dst);
return response;
}
std::memcpy(memory.GetFCRAMPointer(dst - Memory::FCRAM_PADDR), out_streams[ch].data(),
byte_size);
}
// Set the output frame info.
response.decode_aac_response.sample_rate = GetSampleRateEnum(sample_rate);
response.decode_aac_response.num_channels = num_channels;
response.decode_aac_response.num_samples = static_cast<u32_le>(out_streams[0].size());
return response;
}
FAAD2Decoder::FAAD2Decoder(Memory::MemorySystem& memory) : impl(std::make_unique<Impl>(memory)) {}
FAAD2Decoder::~FAAD2Decoder() = default;
std::optional<BinaryMessage> FAAD2Decoder::ProcessRequest(const BinaryMessage& request) {
return impl->ProcessRequest(request);
}
bool FAAD2Decoder::IsValid() const {
return impl->IsValid();
}
} // namespace AudioCore::HLE

6
src/audio_core/hle/audiotoolbox_decoder.h → src/audio_core/hle/faad2_decoder.h
View file

@ -8,10 +8,10 @@
namespace AudioCore::HLE {
class AudioToolboxDecoder final : public DecoderBase {
class FAAD2Decoder final : public DecoderBase {
public:
explicit AudioToolboxDecoder(Memory::MemorySystem& memory);
~AudioToolboxDecoder() override;
explicit FAAD2Decoder(Memory::MemorySystem& memory);
~FAAD2Decoder() override;
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request) override;
bool IsValid() const override;

236
src/audio_core/hle/fdk_decoder.cpp
View file

@ -1,236 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "audio_core/hle/fdk_decoder.h"
#include "common/dynamic_library/fdk-aac.h"
using namespace DynamicLibrary;
namespace AudioCore::HLE {
class FDKDecoder::Impl {
public:
explicit Impl(Memory::MemorySystem& memory);
~Impl();
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request);
bool IsValid() const {
return decoder != nullptr;
}
private:
std::optional<BinaryMessage> Initalize(const BinaryMessage& request);
std::optional<BinaryMessage> Decode(const BinaryMessage& request);
void Clear();
Memory::MemorySystem& memory;
HANDLE_AACDECODER decoder = nullptr;
};
FDKDecoder::Impl::Impl(Memory::MemorySystem& memory) : memory(memory) {
if (!FdkAac::LoadFdkAac()) {
return;
}
// allocate an array of LIB_INFO structures
// if we don't pre-fill the whole segment with zeros, when we call `aacDecoder_GetLibInfo`
// it will segfault, upon investigation, there is some code in fdk_aac depends on your initial
// values in this array
LIB_INFO decoder_info[FDK_MODULE_LAST] = {};
// get library information and fill the struct
if (FdkAac::aacDecoder_GetLibInfo(decoder_info) != 0) {
LOG_ERROR(Audio_DSP, "Failed to retrieve fdk_aac library information!");
return;
}
LOG_INFO(Audio_DSP, "Using fdk_aac version {} (build date: {})", decoder_info[0].versionStr,
decoder_info[0].build_date);
// choose the input format when initializing: 1 layer of ADTS
decoder = FdkAac::aacDecoder_Open(TRANSPORT_TYPE::TT_MP4_ADTS, 1);
// set maximum output channel to two (stereo)
// if the input samples have more channels, fdk_aac will perform a downmix
AAC_DECODER_ERROR ret = FdkAac::aacDecoder_SetParam(decoder, AAC_PCM_MAX_OUTPUT_CHANNELS, 2);
if (ret != AAC_DEC_OK) {
// unable to set this parameter reflects the decoder implementation might be broken
// we'd better shuts down everything
FdkAac::aacDecoder_Close(decoder);
decoder = nullptr;
LOG_ERROR(Audio_DSP, "Unable to set downmix parameter: {}", ret);
return;
}
}
std::optional<BinaryMessage> FDKDecoder::Impl::Initalize(const BinaryMessage& request) {
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
if (decoder) {
LOG_INFO(Audio_DSP, "FDK Decoder initialized");
Clear();
} else {
LOG_ERROR(Audio_DSP, "Decoder not initialized");
}
return response;
}
FDKDecoder::Impl::~Impl() {
if (decoder) {
FdkAac::aacDecoder_Close(decoder);
}
}
void FDKDecoder::Impl::Clear() {
s16 decoder_output[8192];
// flush and re-sync the decoder, discarding the internal buffer
// we actually don't care if this succeeds or not
// FLUSH - flush internal buffer
// INTR - treat the current internal buffer as discontinuous
// CONCEAL - try to interpolate and smooth out the samples
if (decoder) {
FdkAac::aacDecoder_DecodeFrame(decoder, decoder_output, 8192,
AACDEC_FLUSH & AACDEC_INTR & AACDEC_CONCEAL);
}
}
std::optional<BinaryMessage> FDKDecoder::Impl::ProcessRequest(const BinaryMessage& request) {
if (request.header.codec != DecoderCodec::DecodeAAC) {
LOG_ERROR(Audio_DSP, "FDK AAC Decoder cannot handle such codec: {}",
static_cast<u16>(request.header.codec));
return {};
}
switch (request.header.cmd) {
case DecoderCommand::Init: {
return Initalize(request);
}
case DecoderCommand::EncodeDecode: {
return Decode(request);
}
case DecoderCommand::Shutdown:
case DecoderCommand::SaveState:
case DecoderCommand::LoadState: {
LOG_WARNING(Audio_DSP, "Got unimplemented binary request: {}",
static_cast<u16>(request.header.cmd));
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
return response;
}
default:
LOG_ERROR(Audio_DSP, "Got unknown binary request: {}",
static_cast<u16>(request.header.cmd));
return {};
}
}
std::optional<BinaryMessage> FDKDecoder::Impl::Decode(const BinaryMessage& request) {
BinaryMessage response{};
response.header.codec = request.header.codec;
response.header.cmd = request.header.cmd;
response.decode_aac_response.size = request.decode_aac_request.size;
if (!decoder) {
LOG_DEBUG(Audio_DSP, "Decoder not initalized");
// This is a hack to continue games that are not compiled with the aac codec
response.decode_aac_response.num_channels = 2;
response.decode_aac_response.num_samples = 1024;
return response;
}
if (request.decode_aac_request.src_addr < Memory::FCRAM_PADDR ||
request.decode_aac_request.src_addr + request.decode_aac_request.size >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds src_addr {:08x}",
request.decode_aac_request.src_addr);
return {};
}
u8* data = memory.GetFCRAMPointer(request.decode_aac_request.src_addr - Memory::FCRAM_PADDR);
std::array<std::vector<s16>, 2> out_streams;
u32 data_size = request.decode_aac_request.size;
// decoding loops
AAC_DECODER_ERROR result = AAC_DEC_OK;
// Up to 2048 samples, up to 2 channels each
s16 decoder_output[4096];
// note that we don't free this pointer as it is automatically freed by fdk_aac
CStreamInfo* stream_info;
// how many bytes to be queued into the decoder, decrementing from the buffer size
u32 buffer_remaining = data_size;
// alias the data_size as an u32
u32 input_size = data_size;
while (buffer_remaining) {
// queue the input buffer, fdk_aac will automatically slice out the buffer it needs
// from the input buffer
result = FdkAac::aacDecoder_Fill(decoder, &data, &input_size, &buffer_remaining);
if (result != AAC_DEC_OK) {
// there are some issues when queuing the input buffer
LOG_ERROR(Audio_DSP, "Failed to enqueue the input samples");
return std::nullopt;
}
// get output from decoder
result = FdkAac::aacDecoder_DecodeFrame(decoder, decoder_output,
sizeof(decoder_output) / sizeof(s16), 0);
if (result == AAC_DEC_OK) {
// get the stream information
stream_info = FdkAac::aacDecoder_GetStreamInfo(decoder);
// fill the stream information for binary response
response.decode_aac_response.sample_rate = GetSampleRateEnum(stream_info->sampleRate);
response.decode_aac_response.num_channels = stream_info->numChannels;
response.decode_aac_response.num_samples = stream_info->frameSize;
// fill the output
// the sample size = frame_size * channel_counts
for (int sample = 0; sample < stream_info->frameSize; sample++) {
for (int ch = 0; ch < stream_info->numChannels; ch++) {
out_streams[ch].push_back(
decoder_output[(sample * stream_info->numChannels) + ch]);
}
}
} else if (result == AAC_DEC_TRANSPORT_SYNC_ERROR) {
// decoder has some synchronization problems, try again with new samples,
// using old samples might trigger this error again
continue;
} else {
LOG_ERROR(Audio_DSP, "Error decoding the sample: {}", result);
return std::nullopt;
}
}
// transfer the decoded buffer from vector to the FCRAM
for (std::size_t ch = 0; ch < out_streams.size(); ch++) {
if (!out_streams[ch].empty()) {
auto byte_size = out_streams[ch].size() * sizeof(s16);
auto dst = ch == 0 ? request.decode_aac_request.dst_addr_ch0
: request.decode_aac_request.dst_addr_ch1;
if (dst < Memory::FCRAM_PADDR ||
dst + byte_size > Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds dst_addr_ch{} {:08x}", ch, dst);
return {};
}
std::memcpy(memory.GetFCRAMPointer(dst - Memory::FCRAM_PADDR), out_streams[ch].data(),
byte_size);
}
}
return response;
}
FDKDecoder::FDKDecoder(Memory::MemorySystem& memory) : impl(std::make_unique<Impl>(memory)) {}
FDKDecoder::~FDKDecoder() = default;
std::optional<BinaryMessage> FDKDecoder::ProcessRequest(const BinaryMessage& request) {
return impl->ProcessRequest(request);
}
bool FDKDecoder::IsValid() const {
return impl->IsValid();
}
} // namespace AudioCore::HLE

23
src/audio_core/hle/fdk_decoder.h
View file

@ -1,23 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include "audio_core/hle/decoder.h"
namespace AudioCore::HLE {
class FDKDecoder final : public DecoderBase {
public:
explicit FDKDecoder(Memory::MemorySystem& memory);
~FDKDecoder() override;
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request) override;
bool IsValid() const override;
private:
class Impl;
std::unique_ptr<Impl> impl;
};
} // namespace AudioCore::HLE

290
src/audio_core/hle/ffmpeg_decoder.cpp
View file

@ -1,290 +0,0 @@
// Copyright 2018 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "audio_core/hle/ffmpeg_decoder.h"
#include "common/dynamic_library/ffmpeg.h"
using namespace DynamicLibrary;
namespace AudioCore::HLE {
class FFMPEGDecoder::Impl {
public:
explicit Impl(Memory::MemorySystem& memory);
~Impl();
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request);
bool IsValid() const {
return have_ffmpeg_dl;
}
private:
std::optional<BinaryMessage> Initalize(const BinaryMessage& request);
void Clear();
std::optional<BinaryMessage> Decode(const BinaryMessage& request);
struct AVPacketDeleter {
void operator()(AVPacket* packet) const {
FFmpeg::av_packet_free(&packet);
}
};
struct AVCodecContextDeleter {
void operator()(AVCodecContext* context) const {
FFmpeg::avcodec_free_context(&context);
}
};
struct AVCodecParserContextDeleter {
void operator()(AVCodecParserContext* parser) const {
FFmpeg::av_parser_close(parser);
}
};
struct AVFrameDeleter {
void operator()(AVFrame* frame) const {
FFmpeg::av_frame_free(&frame);
}
};
bool initalized = false;
bool have_ffmpeg_dl;
Memory::MemorySystem& memory;
const AVCodec* codec;
std::unique_ptr<AVCodecContext, AVCodecContextDeleter> av_context;
std::unique_ptr<AVCodecParserContext, AVCodecParserContextDeleter> parser;
std::unique_ptr<AVPacket, AVPacketDeleter> av_packet;
std::unique_ptr<AVFrame, AVFrameDeleter> decoded_frame;
};
FFMPEGDecoder::Impl::Impl(Memory::MemorySystem& memory) : memory(memory) {
have_ffmpeg_dl = FFmpeg::LoadFFmpeg();
}
FFMPEGDecoder::Impl::~Impl() = default;
std::optional<BinaryMessage> FFMPEGDecoder::Impl::ProcessRequest(const BinaryMessage& request) {
if (request.header.codec != DecoderCodec::DecodeAAC) {
LOG_ERROR(Audio_DSP, "Got wrong codec {}", static_cast<u16>(request.header.codec));
return {};
}
switch (request.header.cmd) {
case DecoderCommand::Init: {
return Initalize(request);
}
case DecoderCommand::EncodeDecode: {
return Decode(request);
}
case DecoderCommand::Shutdown:
case DecoderCommand::SaveState:
case DecoderCommand::LoadState: {
LOG_WARNING(Audio_DSP, "Got unimplemented binary request: {}",
static_cast<u16>(request.header.cmd));
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
return response;
}
default:
LOG_ERROR(Audio_DSP, "Got unknown binary request: {}",
static_cast<u16>(request.header.cmd));
return {};
}
}
std::optional<BinaryMessage> FFMPEGDecoder::Impl::Initalize(const BinaryMessage& request) {
if (initalized) {
Clear();
}
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
if (!have_ffmpeg_dl) {
return response;
}
av_packet.reset(FFmpeg::av_packet_alloc());
codec = FFmpeg::avcodec_find_decoder(AV_CODEC_ID_AAC);
if (!codec) {
LOG_ERROR(Audio_DSP, "Codec not found\n");
return response;
}
parser.reset(FFmpeg::av_parser_init(codec->id));
if (!parser) {
LOG_ERROR(Audio_DSP, "Parser not found\n");
return response;
}
av_context.reset(FFmpeg::avcodec_alloc_context3(codec));
if (!av_context) {
LOG_ERROR(Audio_DSP, "Could not allocate audio codec context\n");
return response;
}
if (FFmpeg::avcodec_open2(av_context.get(), codec, nullptr) < 0) {
LOG_ERROR(Audio_DSP, "Could not open codec\n");
return response;
}
initalized = true;
return response;
}
void FFMPEGDecoder::Impl::Clear() {
if (!have_ffmpeg_dl) {
return;
}
av_context.reset();
parser.reset();
decoded_frame.reset();
av_packet.reset();
}
std::optional<BinaryMessage> FFMPEGDecoder::Impl::Decode(const BinaryMessage& request) {
BinaryMessage response{};
response.header.codec = request.header.codec;
response.header.cmd = request.header.cmd;
response.decode_aac_response.size = request.decode_aac_request.size;
if (!initalized) {
LOG_DEBUG(Audio_DSP, "Decoder not initalized");
// This is a hack to continue games that are not compiled with the aac codec
response.decode_aac_response.num_channels = 2;
response.decode_aac_response.num_samples = 1024;
return response;
}
if (request.decode_aac_request.src_addr < Memory::FCRAM_PADDR ||
request.decode_aac_request.src_addr + request.decode_aac_request.size >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds src_addr {:08x}",
request.decode_aac_request.src_addr);
return {};
}
u8* data = memory.GetFCRAMPointer(request.decode_aac_request.src_addr - Memory::FCRAM_PADDR);
std::array<std::vector<u8>, 2> out_streams;
std::size_t data_size = request.decode_aac_request.size;
while (data_size > 0) {
if (!decoded_frame) {
decoded_frame.reset(FFmpeg::av_frame_alloc());
if (!decoded_frame) {
LOG_ERROR(Audio_DSP, "Could not allocate audio frame");
return {};
}
}
int ret = FFmpeg::av_parser_parse2(parser.get(), av_context.get(), &av_packet->data,
&av_packet->size, data, static_cast<int>(data_size),
AV_NOPTS_VALUE, AV_NOPTS_VALUE, 0);
if (ret < 0) {
LOG_ERROR(Audio_DSP, "Error while parsing");
return {};
}
data += ret;
data_size -= ret;
ret = FFmpeg::avcodec_send_packet(av_context.get(), av_packet.get());
if (ret < 0) {
LOG_ERROR(Audio_DSP, "Error submitting the packet to the decoder");
return {};
}
if (av_packet->size) {
while (ret >= 0) {
ret = FFmpeg::avcodec_receive_frame(av_context.get(), decoded_frame.get());
if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF)
break;
else if (ret < 0) {
LOG_ERROR(Audio_DSP, "Error during decoding");
return {};
}
int bytes_per_sample = FFmpeg::av_get_bytes_per_sample(av_context->sample_fmt);
if (bytes_per_sample < 0) {
LOG_ERROR(Audio_DSP, "Failed to calculate data size");
return {};
}
#if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(59, 24, 100)
auto num_channels = static_cast<u32>(decoded_frame->ch_layout.nb_channels);
#else
auto num_channels = static_cast<u32>(decoded_frame->channels);
#endif
ASSERT(num_channels <= out_streams.size());
std::size_t size = bytes_per_sample * (decoded_frame->nb_samples);
response.decode_aac_response.sample_rate =
GetSampleRateEnum(decoded_frame->sample_rate);
response.decode_aac_response.num_channels = num_channels;
response.decode_aac_response.num_samples += decoded_frame->nb_samples;
// FFmpeg converts to 32 signed floating point PCM, we need s16 PCM so we need to
// convert it
f32 val_float;
for (std::size_t current_pos(0); current_pos < size;) {
for (std::size_t channel(0); channel < num_channels; channel++) {
std::memcpy(&val_float, decoded_frame->data[channel] + current_pos,
sizeof(val_float));
val_float = std::clamp(val_float, -1.0f, 1.0f);
s16 val = static_cast<s16>(0x7FFF * val_float);
out_streams[channel].push_back(val & 0xFF);
out_streams[channel].push_back(val >> 8);
}
current_pos += sizeof(val_float);
}
}
}
}
if (out_streams[0].size() != 0) {
if (request.decode_aac_request.dst_addr_ch0 < Memory::FCRAM_PADDR ||
request.decode_aac_request.dst_addr_ch0 + out_streams[0].size() >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds dst_addr_ch0 {:08x}",
request.decode_aac_request.dst_addr_ch0);
return {};
}
std::memcpy(
memory.GetFCRAMPointer(request.decode_aac_request.dst_addr_ch0 - Memory::FCRAM_PADDR),
out_streams[0].data(), out_streams[0].size());
}
if (out_streams[1].size() != 0) {
if (request.decode_aac_request.dst_addr_ch1 < Memory::FCRAM_PADDR ||
request.decode_aac_request.dst_addr_ch1 + out_streams[1].size() >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds dst_addr_ch1 {:08x}",
request.decode_aac_request.dst_addr_ch1);
return {};
}
std::memcpy(
memory.GetFCRAMPointer(request.decode_aac_request.dst_addr_ch1 - Memory::FCRAM_PADDR),
out_streams[1].data(), out_streams[1].size());
}
return response;
}
FFMPEGDecoder::FFMPEGDecoder(Memory::MemorySystem& memory) : impl(std::make_unique<Impl>(memory)) {}
FFMPEGDecoder::~FFMPEGDecoder() = default;
std::optional<BinaryMessage> FFMPEGDecoder::ProcessRequest(const BinaryMessage& request) {
return impl->ProcessRequest(request);
}
bool FFMPEGDecoder::IsValid() const {
return impl->IsValid();
}
} // namespace AudioCore::HLE

23
src/audio_core/hle/ffmpeg_decoder.h
View file

@ -1,23 +0,0 @@
// Copyright 2018 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include "audio_core/hle/decoder.h"
namespace AudioCore::HLE {
class FFMPEGDecoder final : public DecoderBase {
public:
explicit FFMPEGDecoder(Memory::MemorySystem& memory);
~FFMPEGDecoder() override;
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request) override;
bool IsValid() const override;
private:
class Impl;
std::unique_ptr<Impl> impl;
};
} // namespace AudioCore::HLE

32
src/audio_core/hle/hle.cpp
View file

@ -8,23 +8,15 @@
#include <boost/serialization/vector.hpp>
#include <boost/serialization/weak_ptr.hpp>
#include "audio_core/audio_types.h"
#include "common/archives.h"
#ifdef HAVE_MF
#include "audio_core/hle/wmf_decoder.h"
#elif HAVE_AUDIOTOOLBOX
#include "audio_core/hle/audiotoolbox_decoder.h"
#elif ANDROID
#include "audio_core/hle/mediandk_decoder.h"
#endif
#include "audio_core/hle/common.h"
#include "audio_core/hle/decoder.h"
#include "audio_core/hle/fdk_decoder.h"
#include "audio_core/hle/ffmpeg_decoder.h"
#include "audio_core/hle/faad2_decoder.h"
#include "audio_core/hle/hle.h"
#include "audio_core/hle/mixers.h"
#include "audio_core/hle/shared_memory.h"
#include "audio_core/hle/source.h"
#include "audio_core/sink.h"
#include "common/archives.h"
#include "common/assert.h"
#include "common/common_types.h"
#include "common/hash.h"
@ -121,26 +113,8 @@ private:
static std::vector<std::function<std::unique_ptr<HLE::DecoderBase>(Memory::MemorySystem&)>>
decoder_backends = {
#if defined(HAVE_MF)
[](Memory::MemorySystem& memory) -> std::unique_ptr<HLE::DecoderBase> {
return std::make_unique<HLE::WMFDecoder>(memory);
},
#endif
#if defined(HAVE_AUDIOTOOLBOX)
[](Memory::MemorySystem& memory) -> std::unique_ptr<HLE::DecoderBase> {
return std::make_unique<HLE::AudioToolboxDecoder>(memory);
},
#endif
#if ANDROID
[](Memory::MemorySystem& memory) -> std::unique_ptr<HLE::DecoderBase> {
return std::make_unique<HLE::MediaNDKDecoder>(memory);
},
#endif
[](Memory::MemorySystem& memory) -> std::unique_ptr<HLE::DecoderBase> {
return std::make_unique<HLE::FDKDecoder>(memory);
},
[](Memory::MemorySystem& memory) -> std::unique_ptr<HLE::DecoderBase> {
return std::make_unique<HLE::FFMPEGDecoder>(memory);
return std::make_unique<HLE::FAAD2Decoder>(memory);
},
};

253
src/audio_core/hle/mediandk_decoder.cpp
View file

@ -1,253 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <media/NdkMediaCodec.h>
#include <media/NdkMediaError.h>
#include <media/NdkMediaFormat.h>
#include <memory>
#include <vector>
#include "audio_core/hle/adts.h"
#include "audio_core/hle/mediandk_decoder.h"
namespace AudioCore::HLE {
struct AMediaCodecRelease {
void operator()(AMediaCodec* codec) const {
AMediaCodec_stop(codec);
AMediaCodec_delete(codec);
};
};
class MediaNDKDecoder::Impl {
public:
explicit Impl(Memory::MemorySystem& memory);
~Impl();
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request);
bool SetMediaType(const AudioCore::ADTSData& adts_data);
private:
std::optional<BinaryMessage> Initalize(const BinaryMessage& request);
std::optional<BinaryMessage> Decode(const BinaryMessage& request);
Memory::MemorySystem& memory;
std::unique_ptr<AMediaCodec, AMediaCodecRelease> decoder;
// default: 2 channles, 48000 samplerate
AudioCore::ADTSData mADTSData{
/*header_length*/ 7, /*mpeg2*/ false, /*profile*/ 2,
/*channels*/ 2, /*channel_idx*/ 2, /*framecount*/ 0,
/*samplerate_idx*/ 3, /*length*/ 0, /*samplerate*/ 48000};
};
MediaNDKDecoder::Impl::Impl(Memory::MemorySystem& memory_) : memory(memory_) {
SetMediaType(mADTSData);
}
MediaNDKDecoder::Impl::~Impl() = default;
std::optional<BinaryMessage> MediaNDKDecoder::Impl::Initalize(const BinaryMessage& request) {
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
return response;
}
bool MediaNDKDecoder::Impl::SetMediaType(const AudioCore::ADTSData& adts_data) {
const char* mime = "audio/mp4a-latm";
if (decoder && mADTSData.profile == adts_data.profile &&
mADTSData.channel_idx == adts_data.channel_idx &&
mADTSData.samplerate_idx == adts_data.samplerate_idx) {
return true;
}
decoder.reset(AMediaCodec_createDecoderByType(mime));
if (decoder == nullptr) {
return false;
}
u8 csd_0[2];
csd_0[0] = static_cast<u8>((adts_data.profile << 3) | (adts_data.samplerate_idx >> 1));
csd_0[1] =
static_cast<u8>(((adts_data.samplerate_idx << 7) & 0x80) | (adts_data.channel_idx << 3));
AMediaFormat* format = AMediaFormat_new();
AMediaFormat_setString(format, AMEDIAFORMAT_KEY_MIME, mime);
AMediaFormat_setInt32(format, AMEDIAFORMAT_KEY_SAMPLE_RATE, adts_data.samplerate);
AMediaFormat_setInt32(format, AMEDIAFORMAT_KEY_CHANNEL_COUNT, adts_data.channels);
AMediaFormat_setInt32(format, AMEDIAFORMAT_KEY_IS_ADTS, 1);
AMediaFormat_setBuffer(format, "csd-0", csd_0, sizeof(csd_0));
media_status_t status = AMediaCodec_configure(decoder.get(), format, NULL, NULL, 0);
if (status != AMEDIA_OK) {
AMediaFormat_delete(format);
decoder.reset();
return false;
}
status = AMediaCodec_start(decoder.get());
if (status != AMEDIA_OK) {
AMediaFormat_delete(format);
decoder.reset();
return false;
}
AMediaFormat_delete(format);
mADTSData = adts_data;
return true;
}
std::optional<BinaryMessage> MediaNDKDecoder::Impl::ProcessRequest(const BinaryMessage& request) {
if (request.header.codec != DecoderCodec::DecodeAAC) {
LOG_ERROR(Audio_DSP, "AAC Decoder cannot handle such codec: {}",
static_cast<u16>(request.header.codec));
return {};
}
switch (request.header.cmd) {
case DecoderCommand::Init: {
return Initalize(request);
}
case DecoderCommand::EncodeDecode: {
return Decode(request);
}
case DecoderCommand::Shutdown:
case DecoderCommand::SaveState:
case DecoderCommand::LoadState: {
LOG_WARNING(Audio_DSP, "Got unimplemented binary request: {}",
static_cast<u16>(request.header.cmd));
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
return response;
}
default:
LOG_ERROR(Audio_DSP, "Got unknown binary request: {}",
static_cast<u16>(request.header.cmd));
return {};
}
}
std::optional<BinaryMessage> MediaNDKDecoder::Impl::Decode(const BinaryMessage& request) {
BinaryMessage response{};
response.header.codec = request.header.codec;
response.header.cmd = request.header.cmd;
response.decode_aac_response.size = request.decode_aac_request.size;
response.decode_aac_response.num_samples = 1024;
if (request.decode_aac_request.src_addr < Memory::FCRAM_PADDR ||
request.decode_aac_request.src_addr + request.decode_aac_request.size >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds src_addr {:08x}",
request.decode_aac_request.src_addr);
return response;
}
const u8* data =
memory.GetFCRAMPointer(request.decode_aac_request.src_addr - Memory::FCRAM_PADDR);
ADTSData adts_data = AudioCore::ParseADTS(data);
SetMediaType(adts_data);
response.decode_aac_response.sample_rate = GetSampleRateEnum(adts_data.samplerate);
response.decode_aac_response.num_channels = adts_data.channels;
if (!decoder) {
LOG_ERROR(Audio_DSP, "Missing decoder for profile: {}, channels: {}, samplerate: {}",
adts_data.profile, adts_data.channels, adts_data.samplerate);
return {};
}
// input
constexpr int timeout = 160;
std::size_t buffer_size = 0;
u8* buffer = nullptr;
ssize_t buffer_index = AMediaCodec_dequeueInputBuffer(decoder.get(), timeout);
if (buffer_index < 0) {
LOG_ERROR(Audio_DSP, "Failed to enqueue the input samples: {}", buffer_index);
return response;
}
buffer = AMediaCodec_getInputBuffer(decoder.get(), buffer_index, &buffer_size);
if (buffer_size < request.decode_aac_request.size) {
return response;
}
std::memcpy(buffer, data, request.decode_aac_request.size);
media_status_t status = AMediaCodec_queueInputBuffer(decoder.get(), buffer_index, 0,
request.decode_aac_request.size, 0, 0);
if (status != AMEDIA_OK) {
LOG_WARNING(Audio_DSP, "Try queue input buffer again later!");
return response;
}
// output
AMediaCodecBufferInfo info;
std::array<std::vector<u16>, 2> out_streams;
buffer_index = AMediaCodec_dequeueOutputBuffer(decoder.get(), &info, timeout);
switch (buffer_index) {
case AMEDIACODEC_INFO_TRY_AGAIN_LATER:
LOG_WARNING(Audio_DSP, "Failed to dequeue output buffer: timeout!");
break;
case AMEDIACODEC_INFO_OUTPUT_BUFFERS_CHANGED:
LOG_WARNING(Audio_DSP, "Failed to dequeue output buffer: buffers changed!");
break;
case AMEDIACODEC_INFO_OUTPUT_FORMAT_CHANGED: {
AMediaFormat* format = AMediaCodec_getOutputFormat(decoder.get());
LOG_WARNING(Audio_DSP, "output format: {}", AMediaFormat_toString(format));
AMediaFormat_delete(format);
buffer_index = AMediaCodec_dequeueOutputBuffer(decoder.get(), &info, timeout);
}
default: {
int offset = info.offset;
buffer = AMediaCodec_getOutputBuffer(decoder.get(), buffer_index, &buffer_size);
while (offset < info.size) {
for (int channel = 0; channel < response.decode_aac_response.num_channels; channel++) {
u16 pcm_data;
std::memcpy(&pcm_data, buffer + offset, sizeof(pcm_data));
out_streams[channel].push_back(pcm_data);
offset += sizeof(pcm_data);
}
}
AMediaCodec_releaseOutputBuffer(decoder.get(), buffer_index, info.size != 0);
}
}
// transfer the decoded buffer from vector to the FCRAM
size_t stream0_size = out_streams[0].size() * sizeof(u16);
if (stream0_size != 0) {
if (request.decode_aac_request.dst_addr_ch0 < Memory::FCRAM_PADDR ||
request.decode_aac_request.dst_addr_ch0 + stream0_size >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds dst_addr_ch0 {:08x}",
request.decode_aac_request.dst_addr_ch0);
return response;
}
std::memcpy(
memory.GetFCRAMPointer(request.decode_aac_request.dst_addr_ch0 - Memory::FCRAM_PADDR),
out_streams[0].data(), stream0_size);
}
size_t stream1_size = out_streams[1].size() * sizeof(u16);
if (stream1_size != 0) {
if (request.decode_aac_request.dst_addr_ch1 < Memory::FCRAM_PADDR ||
request.decode_aac_request.dst_addr_ch1 + stream1_size >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds dst_addr_ch1 {:08x}",
request.decode_aac_request.dst_addr_ch1);
return response;
}
std::memcpy(
memory.GetFCRAMPointer(request.decode_aac_request.dst_addr_ch1 - Memory::FCRAM_PADDR),
out_streams[1].data(), stream1_size);
}
return response;
}
MediaNDKDecoder::MediaNDKDecoder(Memory::MemorySystem& memory)
: impl(std::make_unique<Impl>(memory)) {}
MediaNDKDecoder::~MediaNDKDecoder() = default;
std::optional<BinaryMessage> MediaNDKDecoder::ProcessRequest(const BinaryMessage& request) {
return impl->ProcessRequest(request);
}
bool MediaNDKDecoder::IsValid() const {
return true;
}
} // namespace AudioCore::HLE

22
src/audio_core/hle/mediandk_decoder.h
View file

@ -1,22 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include "audio_core/hle/decoder.h"
namespace AudioCore::HLE {
class MediaNDKDecoder final : public DecoderBase {
public:
explicit MediaNDKDecoder(Memory::MemorySystem& memory);
~MediaNDKDecoder() override;
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request) override;
bool IsValid() const override;
private:
class Impl;
std::unique_ptr<Impl> impl;
};
} // namespace AudioCore::HLE

313
src/audio_core/hle/wmf_decoder.cpp
View file

@ -1,313 +0,0 @@
// Copyright 2018 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "audio_core/hle/wmf_decoder.h"
#include "audio_core/hle/wmf_decoder_utils.h"
namespace AudioCore::HLE {
using namespace MFDecoder;
class WMFDecoder::Impl {
public:
explicit Impl(Memory::MemorySystem& memory);
~Impl();
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request);
bool IsValid() const {
return is_valid;
}
private:
std::optional<BinaryMessage> Initalize(const BinaryMessage& request);
std::optional<BinaryMessage> Decode(const BinaryMessage& request);
MFOutputState DecodingLoop(AudioCore::ADTSData adts_header,
std::array<std::vector<u8>, 2>& out_streams);
bool transform_initialized = false;
bool format_selected = false;
Memory::MemorySystem& memory;
unique_mfptr<IMFTransform> transform;
DWORD in_stream_id = 0;
DWORD out_stream_id = 0;
bool is_valid = false;
bool mf_started = false;
bool coinited = false;
};
WMFDecoder::Impl::Impl(Memory::MemorySystem& memory) : memory(memory) {
// Attempt to load the symbols for mf.dll
if (!InitMFDLL()) {
LOG_CRITICAL(Audio_DSP,
"Unable to load mf.dll. AAC audio through media foundation unavailable");
return;
}
HRESULT hr = S_OK;
hr = CoInitialize(NULL);
// S_FALSE will be returned when COM has already been initialized
if (hr != S_OK && hr != S_FALSE) {
ReportError("Failed to start COM components", hr);
} else {
coinited = true;
}
// lite startup is faster and all what we need is included
hr = MFDecoder::MFStartup(MF_VERSION, MFSTARTUP_LITE);
if (hr != S_OK) {
// Do you know you can't initialize MF in test mode or safe mode?
ReportError("Failed to initialize Media Foundation", hr);
} else {
mf_started = true;
}
LOG_INFO(Audio_DSP, "Media Foundation activated");
// initialize transform
transform = MFDecoderInit();
if (transform == nullptr) {
LOG_CRITICAL(Audio_DSP, "Can't initialize decoder");
return;
}
hr = transform->GetStreamIDs(1, &in_stream_id, 1, &out_stream_id);
if (hr == E_NOTIMPL) {
// if not implemented, it means this MFT does not assign stream ID for you
in_stream_id = 0;
out_stream_id = 0;
} else if (FAILED(hr)) {
ReportError("Decoder failed to initialize the stream ID", hr);
return;
}
transform_initialized = true;
is_valid = true;
}
WMFDecoder::Impl::~Impl() {
if (transform_initialized) {
MFFlush(transform.get());
// delete the transform object before shutting down MF
// otherwise access violation will occur
transform.reset();
}
if (mf_started) {
MFDecoder::MFShutdown();
}
if (coinited) {
CoUninitialize();
}
}
std::optional<BinaryMessage> WMFDecoder::Impl::ProcessRequest(const BinaryMessage& request) {
if (request.header.codec != DecoderCodec::DecodeAAC) {
LOG_ERROR(Audio_DSP, "Got unknown codec {}", static_cast<u16>(request.header.codec));
return std::nullopt;
}
switch (request.header.cmd) {
case DecoderCommand::Init: {
LOG_INFO(Audio_DSP, "WMFDecoder initializing");
return Initalize(request);
}
case DecoderCommand::EncodeDecode: {
return Decode(request);
}
case DecoderCommand::Shutdown:
case DecoderCommand::SaveState:
case DecoderCommand::LoadState: {
LOG_WARNING(Audio_DSP, "Got unimplemented binary request: {}",
static_cast<u16>(request.header.cmd));
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
return response;
}
default:
LOG_ERROR(Audio_DSP, "Got unknown binary request: {}",
static_cast<u16>(request.header.cmd));
return std::nullopt;
}
}
std::optional<BinaryMessage> WMFDecoder::Impl::Initalize(const BinaryMessage& request) {
BinaryMessage response = request;
response.header.result = ResultStatus::Success;
format_selected = false; // select format again if application request initialize the DSP
return response;
}
MFOutputState WMFDecoder::Impl::DecodingLoop(AudioCore::ADTSData adts_header,
std::array<std::vector<u8>, 2>& out_streams) {
std::optional<std::vector<f32>> output_buffer;
while (true) {
auto [output_status, output] = ReceiveSample(transform.get(), out_stream_id);
// 0 -> okay; 3 -> okay but more data available (buffer too small)
if (output_status == MFOutputState::OK || output_status == MFOutputState::HaveMoreData) {
output_buffer = CopySampleToBuffer(output.get());
// the following was taken from ffmpeg version of the decoder
f32 val_f32;
for (std::size_t i = 0; i < output_buffer->size();) {
for (std::size_t channel = 0; channel < adts_header.channels; channel++) {
val_f32 = std::clamp(output_buffer->at(i), -1.0f, 1.0f);
s16 val = static_cast<s16>(0x7FFF * val_f32);
out_streams[channel].push_back(val & 0xFF);
out_streams[channel].push_back(val >> 8);
// i is incremented on per channel basis
i++;
}
}
}
// If we return OK here, the decoder won't be in a state to receive new data and will fail
// on the next call; instead treat it like the HaveMoreData case
if (output_status == MFOutputState::OK)
continue;
// for status = 2, reset MF
if (output_status == MFOutputState::NeedReconfig) {
format_selected = false;
return MFOutputState::NeedReconfig;
}
// for status = 3, try again with new buffer
if (output_status == MFOutputState::HaveMoreData)
continue;
// according to MS document, this is not an error (?!)
if (output_status == MFOutputState::NeedMoreInput)
return MFOutputState::NeedMoreInput;
return MFOutputState::FatalError; // return on other status
}
return MFOutputState::FatalError;
}
std::optional<BinaryMessage> WMFDecoder::Impl::Decode(const BinaryMessage& request) {
BinaryMessage response{};
response.header.codec = request.header.codec;
response.header.cmd = request.header.cmd;
response.decode_aac_response.size = request.decode_aac_request.size;
response.decode_aac_response.num_channels = 2;
response.decode_aac_response.num_samples = 1024;
if (!transform_initialized) {
LOG_DEBUG(Audio_DSP, "Decoder not initialized");
// This is a hack to continue games when decoder failed to initialize
return response;
}
if (request.decode_aac_request.src_addr < Memory::FCRAM_PADDR ||
request.decode_aac_request.src_addr + request.decode_aac_request.size >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds src_addr {:08x}",
request.decode_aac_request.src_addr);
return std::nullopt;
}
const u8* data =
memory.GetFCRAMPointer(request.decode_aac_request.src_addr - Memory::FCRAM_PADDR);
std::array<std::vector<u8>, 2> out_streams;
unique_mfptr<IMFSample> sample;
MFInputState input_status = MFInputState::OK;
MFOutputState output_status = MFOutputState::OK;
std::optional<ADTSMeta> adts_meta = DetectMediaType(data, request.decode_aac_request.size);
if (!adts_meta) {
LOG_ERROR(Audio_DSP, "Unable to deduce decoding parameters from ADTS stream");
return response;
}
response.decode_aac_response.sample_rate = GetSampleRateEnum(adts_meta->ADTSHeader.samplerate);
response.decode_aac_response.num_channels = adts_meta->ADTSHeader.channels;
if (!format_selected) {
LOG_DEBUG(Audio_DSP, "New ADTS stream: channels = {}, sample rate = {}",
adts_meta->ADTSHeader.channels, adts_meta->ADTSHeader.samplerate);
SelectInputMediaType(transform.get(), in_stream_id, adts_meta->ADTSHeader,
adts_meta->AACTag, 14);
SelectOutputMediaType(transform.get(), out_stream_id);
SendSample(transform.get(), in_stream_id, nullptr);
// cache the result from detect_mediatype and call select_*_mediatype only once
// This could increase performance very slightly
transform->ProcessMessage(MFT_MESSAGE_NOTIFY_BEGIN_STREAMING, 0);
format_selected = true;
}
sample = CreateSample(data, request.decode_aac_request.size, 1, 0);
sample->SetUINT32(MFSampleExtension_CleanPoint, 1);
while (true) {
input_status = SendSample(transform.get(), in_stream_id, sample.get());
output_status = DecodingLoop(adts_meta->ADTSHeader, out_streams);
if (output_status == MFOutputState::FatalError) {
// if the decode issues are caused by MFT not accepting new samples, try again
// NOTICE: you are required to check the output even if you already knew/guessed
// MFT didn't accept the input sample
if (input_status == MFInputState::NotAccepted) {
// try again
continue;
}
LOG_ERROR(Audio_DSP, "Errors occurred when receiving output");
return response;
} else if (output_status == MFOutputState::NeedReconfig) {
// flush the transform
MFFlush(transform.get());
// decode again
return this->Decode(request);
}
break; // jump out of the loop if at least we don't have obvious issues
}
if (out_streams[0].size() != 0) {
if (request.decode_aac_request.dst_addr_ch0 < Memory::FCRAM_PADDR ||
request.decode_aac_request.dst_addr_ch0 + out_streams[0].size() >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds dst_addr_ch0 {:08x}",
request.decode_aac_request.dst_addr_ch0);
return std::nullopt;
}
std::memcpy(
memory.GetFCRAMPointer(request.decode_aac_request.dst_addr_ch0 - Memory::FCRAM_PADDR),
out_streams[0].data(), out_streams[0].size());
}
if (out_streams[1].size() != 0) {
if (request.decode_aac_request.dst_addr_ch1 < Memory::FCRAM_PADDR ||
request.decode_aac_request.dst_addr_ch1 + out_streams[1].size() >
Memory::FCRAM_PADDR + Memory::FCRAM_SIZE) {
LOG_ERROR(Audio_DSP, "Got out of bounds dst_addr_ch1 {:08x}",
request.decode_aac_request.dst_addr_ch1);
return std::nullopt;
}
std::memcpy(
memory.GetFCRAMPointer(request.decode_aac_request.dst_addr_ch1 - Memory::FCRAM_PADDR),
out_streams[1].data(), out_streams[1].size());
}
return response;
}
WMFDecoder::WMFDecoder(Memory::MemorySystem& memory) : impl(std::make_unique<Impl>(memory)) {}
WMFDecoder::~WMFDecoder() = default;
std::optional<BinaryMessage> WMFDecoder::ProcessRequest(const BinaryMessage& request) {
return impl->ProcessRequest(request);
}
bool WMFDecoder::IsValid() const {
return impl->IsValid();
}
} // namespace AudioCore::HLE

23
src/audio_core/hle/wmf_decoder.h
View file

@ -1,23 +0,0 @@
// Copyright 2018 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include "audio_core/hle/decoder.h"
namespace AudioCore::HLE {
class WMFDecoder final : public DecoderBase {
public:
explicit WMFDecoder(Memory::MemorySystem& memory);
~WMFDecoder() override;
std::optional<BinaryMessage> ProcessRequest(const BinaryMessage& request) override;
bool IsValid() const override;
private:
class Impl;
std::unique_ptr<Impl> impl;
};
} // namespace AudioCore::HLE

464
src/audio_core/hle/wmf_decoder_utils.cpp
View file

@ -1,464 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "common/logging/log.h"
#include "common/string_util.h"
#include "wmf_decoder_utils.h"
namespace MFDecoder {
// utility functions
void ReportError(std::string msg, HRESULT hr) {
if (SUCCEEDED(hr)) {
return;
}
LPWSTR err;
FormatMessageW(FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_ALLOCATE_BUFFER |
FORMAT_MESSAGE_IGNORE_INSERTS,
nullptr, hr,
// hardcode to use en_US because if any user had problems with this
// we can help them w/o translating anything
// default is to use the language currently active on the operating system
MAKELANGID(LANG_ENGLISH, SUBLANG_ENGLISH_US), (LPWSTR)&err, 0, nullptr);
if (err != nullptr) {
LOG_CRITICAL(Audio_DSP, "{}: {}", msg, Common::UTF16ToUTF8(err));
LocalFree(err);
}
LOG_CRITICAL(Audio_DSP, "{}: {:08x}", msg, hr);
}
unique_mfptr<IMFTransform> MFDecoderInit(GUID audio_format) {
HRESULT hr = S_OK;
MFT_REGISTER_TYPE_INFO reg{};
GUID category = MFT_CATEGORY_AUDIO_DECODER;
IMFActivate** activate;
unique_mfptr<IMFTransform> transform;
UINT32 num_activate;
reg.guidMajorType = MFMediaType_Audio;
reg.guidSubtype = audio_format;
hr = MFTEnumEx(category,
MFT_ENUM_FLAG_SYNCMFT | MFT_ENUM_FLAG_LOCALMFT | MFT_ENUM_FLAG_SORTANDFILTER,
&reg, nullptr, &activate, &num_activate);
if (FAILED(hr) || num_activate < 1) {
ReportError("Failed to enumerate decoders", hr);
CoTaskMemFree(activate);
return nullptr;
}
LOG_INFO(Audio_DSP, "Windows(R) Media Foundation found {} suitable decoder(s)", num_activate);
for (unsigned int n = 0; n < num_activate; n++) {
hr = activate[n]->ActivateObject(
IID_IMFTransform,
reinterpret_cast<void**>(static_cast<IMFTransform**>(Amp(transform))));
if (FAILED(hr))
transform = nullptr;
activate[n]->Release();
if (SUCCEEDED(hr))
break;
}
if (transform == nullptr) {
ReportError("Failed to initialize MFT", hr);
CoTaskMemFree(activate);
return nullptr;
}
CoTaskMemFree(activate);
return transform;
}
unique_mfptr<IMFSample> CreateSample(const void* data, DWORD len, DWORD alignment,
LONGLONG duration) {
HRESULT hr = S_OK;
unique_mfptr<IMFMediaBuffer> buf;
unique_mfptr<IMFSample> sample;
hr = MFCreateSample(Amp(sample));
if (FAILED(hr)) {
ReportError("Unable to allocate a sample", hr);
return nullptr;
}
// Yes, the argument for alignment is the actual alignment - 1
hr = MFCreateAlignedMemoryBuffer(len, alignment - 1, Amp(buf));
if (FAILED(hr)) {
ReportError("Unable to allocate a memory buffer for sample", hr);
return nullptr;
}
if (data) {
BYTE* buffer;
// lock the MediaBuffer
// this is actually not a thread-safe lock
hr = buf->Lock(&buffer, nullptr, nullptr);
if (FAILED(hr)) {
ReportError("Unable to lock down MediaBuffer", hr);
return nullptr;
}
std::memcpy(buffer, data, len);
buf->SetCurrentLength(len);
buf->Unlock();
}
sample->AddBuffer(buf.get());
hr = sample->SetSampleDuration(duration);
if (FAILED(hr)) {
// MFT will take a guess for you in this case
ReportError("Unable to set sample duration, but continuing anyway", hr);
}
return sample;
}
bool SelectInputMediaType(IMFTransform* transform, int in_stream_id,
const AudioCore::ADTSData& adts, const UINT8* user_data,
UINT32 user_data_len, GUID audio_format) {
HRESULT hr = S_OK;
unique_mfptr<IMFMediaType> t;
// actually you can get rid of the whole block of searching and filtering mess
// if you know the exact parameters of your media stream
hr = MFCreateMediaType(Amp(t));
if (FAILED(hr)) {
ReportError("Unable to create an empty MediaType", hr);
return false;
}
// basic definition
t->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Audio);
t->SetGUID(MF_MT_SUBTYPE, audio_format);
t->SetUINT32(MF_MT_AAC_PAYLOAD_TYPE, 1);
t->SetUINT32(MF_MT_AUDIO_NUM_CHANNELS, adts.channels);
t->SetUINT32(MF_MT_AUDIO_SAMPLES_PER_SECOND, adts.samplerate);
// 0xfe = 254 = "unspecified"
t->SetUINT32(MF_MT_AAC_AUDIO_PROFILE_LEVEL_INDICATION, 254);
t->SetUINT32(MF_MT_AUDIO_BLOCK_ALIGNMENT, 1);
t->SetBlob(MF_MT_USER_DATA, user_data, user_data_len);
hr = transform->SetInputType(in_stream_id, t.get(), 0);
if (FAILED(hr)) {
ReportError("failed to select input types for MFT", hr);
return false;
}
return true;
}
bool SelectOutputMediaType(IMFTransform* transform, int out_stream_id, GUID audio_format) {
HRESULT hr = S_OK;
UINT32 tmp;
unique_mfptr<IMFMediaType> type;
// If you know what you need and what you are doing, you can specify the conditions instead of
// searching but it's better to use search since MFT may or may not support your output
// parameters
for (DWORD i = 0;; i++) {
hr = transform->GetOutputAvailableType(out_stream_id, i, Amp(type));
if (hr == MF_E_NO_MORE_TYPES || hr == E_NOTIMPL) {
return true;
}
if (FAILED(hr)) {
ReportError("failed to get output types for MFT", hr);
return false;
}
hr = type->GetUINT32(MF_MT_AUDIO_BITS_PER_SAMPLE, &tmp);
if (FAILED(hr))
continue;
// select PCM-16 format
if (tmp == 32) {
hr = type->SetUINT32(MF_MT_AUDIO_BLOCK_ALIGNMENT, 1);
if (FAILED(hr)) {
ReportError("failed to set MF_MT_AUDIO_BLOCK_ALIGNMENT for MFT on output stream",
hr);
return false;
}
hr = transform->SetOutputType(out_stream_id, type.get(), 0);
if (FAILED(hr)) {
ReportError("failed to select output types for MFT", hr);
return false;
}
return true;
} else {
continue;
}
return false;
}
ReportError("MFT: Unable to find preferred output format", E_NOTIMPL);
return false;
}
std::optional<ADTSMeta> DetectMediaType(const u8* buffer, std::size_t len) {
if (len < 7) {
return std::nullopt;
}
AudioCore::ADTSData tmp;
ADTSMeta result;
// see https://docs.microsoft.com/en-us/windows/desktop/api/mmreg/ns-mmreg-heaacwaveinfo_tag
// for the meaning of the byte array below
// it might be a good idea to wrap the parameters into a struct
// and pass that struct into the function but doing that will lead to messier code
// const UINT8 aac_data[] = { 0x01, 0x00, 0xfe, 00, 00, 00, 00, 00, 00, 00, 00, 00, 0x11, 0x90
// }; first byte: 0: raw aac 1: adts 2: adif 3: latm/laos
UINT8 aac_tmp[] = {0x01, 0x00, 0xfe, 00, 00, 00, 00, 00, 00, 00, 00, 00, 0x00, 0x00};
uint16_t tag = 0;
tmp = AudioCore::ParseADTS(buffer);
if (tmp.length == 0) {
return std::nullopt;
}
tag = MFGetAACTag(tmp);
aac_tmp[12] |= (tag & 0xff00) >> 8;
aac_tmp[13] |= (tag & 0x00ff);
std::memcpy(&(result.ADTSHeader), &tmp, sizeof(AudioCore::ADTSData));
std::memcpy(&(result.AACTag), aac_tmp, 14);
return result;
}
void MFFlush(IMFTransform* transform) {
HRESULT hr = transform->ProcessMessage(MFT_MESSAGE_COMMAND_FLUSH, 0);
if (FAILED(hr)) {
ReportError("MFT: Flush command failed", hr);
}
hr = transform->ProcessMessage(MFT_MESSAGE_NOTIFY_END_OF_STREAM, 0);
if (FAILED(hr)) {
ReportError("Failed to end streaming for MFT", hr);
}
}
MFInputState SendSample(IMFTransform* transform, DWORD in_stream_id, IMFSample* in_sample) {
HRESULT hr = S_OK;
if (in_sample) {
hr = transform->ProcessInput(in_stream_id, in_sample, 0);
if (hr == MF_E_NOTACCEPTING) {
return MFInputState::NotAccepted; // try again
} else if (FAILED(hr)) {
ReportError("MFT: Failed to process input", hr);
return MFInputState::FatalError;
} // FAILED(hr)
} else {
hr = transform->ProcessMessage(MFT_MESSAGE_COMMAND_DRAIN, 0);
if (FAILED(hr)) {
ReportError("MFT: Failed to drain when processing input", hr);
}
}
return MFInputState::OK;
}
std::tuple<MFOutputState, unique_mfptr<IMFSample>> ReceiveSample(IMFTransform* transform,
DWORD out_stream_id) {
HRESULT hr;
MFT_OUTPUT_DATA_BUFFER out_buffers;
MFT_OUTPUT_STREAM_INFO out_info;
DWORD status = 0;
unique_mfptr<IMFSample> sample;
bool mft_create_sample = false;
hr = transform->GetOutputStreamInfo(out_stream_id, &out_info);
if (FAILED(hr)) {
ReportError("MFT: Failed to get stream info", hr);
return std::make_tuple(MFOutputState::FatalError, std::move(sample));
}
mft_create_sample = (out_info.dwFlags & MFT_OUTPUT_STREAM_PROVIDES_SAMPLES) ||
(out_info.dwFlags & MFT_OUTPUT_STREAM_CAN_PROVIDE_SAMPLES);
while (true) {
status = 0;
if (!mft_create_sample) {
sample = CreateSample(nullptr, out_info.cbSize, out_info.cbAlignment);
if (!sample.get()) {
ReportError("MFT: Unable to allocate memory for samples", hr);
return std::make_tuple(MFOutputState::FatalError, std::move(sample));
}
}
out_buffers.dwStreamID = out_stream_id;
out_buffers.pSample = sample.get();
hr = transform->ProcessOutput(0, 1, &out_buffers, &status);
if (!FAILED(hr)) {
break;
}
if (hr == MF_E_TRANSFORM_NEED_MORE_INPUT) {
// Most likely reasons: data corrupted; your actions not expected by MFT
return std::make_tuple(MFOutputState::NeedMoreInput, std::move(sample));
}
if (hr == MF_E_TRANSFORM_STREAM_CHANGE) {
ReportError("MFT: stream format changed, re-configuration required", hr);
return std::make_tuple(MFOutputState::NeedReconfig, std::move(sample));
}
break;
}
if (out_buffers.dwStatus & MFT_OUTPUT_DATA_BUFFER_INCOMPLETE) {
// this status is also unreliable but whatever
return std::make_tuple(MFOutputState::HaveMoreData, std::move(sample));
}
if (out_buffers.pSample == nullptr) {
ReportError("MFT: decoding failure", hr);
return std::make_tuple(MFOutputState::FatalError, std::move(sample));
}
return std::make_tuple(MFOutputState::OK, std::move(sample));
}
std::optional<std::vector<f32>> CopySampleToBuffer(IMFSample* sample) {
unique_mfptr<IMFMediaBuffer> buffer;
HRESULT hr = S_OK;
std::optional<std::vector<f32>> output;
std::vector<f32> output_buffer;
BYTE* data;
DWORD len = 0;
hr = sample->GetTotalLength(&len);
if (FAILED(hr)) {
ReportError("Failed to get the length of sample buffer", hr);
return std::nullopt;
}
hr = sample->ConvertToContiguousBuffer(Amp(buffer));
if (FAILED(hr)) {
ReportError("Failed to get sample buffer", hr);
return std::nullopt;
}
hr = buffer->Lock(&data, nullptr, nullptr);
if (FAILED(hr)) {
ReportError("Failed to lock the buffer", hr);
return std::nullopt;
}
output_buffer.resize(len / sizeof(f32));
std::memcpy(output_buffer.data(), data, len);
output = output_buffer;
// if buffer unlock fails, then... whatever, we have already got data
buffer->Unlock();
return output;
}
namespace {
struct LibraryDeleter {
using pointer = HMODULE;
void operator()(HMODULE h) const {
if (h != nullptr)
FreeLibrary(h);
}
};
std::unique_ptr<HMODULE, LibraryDeleter> mf_dll{nullptr};
std::unique_ptr<HMODULE, LibraryDeleter> mfplat_dll{nullptr};
} // namespace
bool InitMFDLL() {
mf_dll.reset(LoadLibrary(TEXT("mf.dll")));
if (!mf_dll) {
DWORD error_message_id = GetLastError();
LPSTR message_buffer = nullptr;
size_t size =
FormatMessageA(FORMAT_MESSAGE_ALLOCATE_BUFFER | FORMAT_MESSAGE_FROM_SYSTEM |
FORMAT_MESSAGE_IGNORE_INSERTS,
nullptr, error_message_id, MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT),
reinterpret_cast<LPSTR>(&message_buffer), 0, nullptr);
std::string message(message_buffer, size);
LocalFree(message_buffer);
LOG_ERROR(Audio_DSP, "Could not load mf.dll: {}", message);
return false;
}
mfplat_dll.reset(LoadLibrary(TEXT("mfplat.dll")));
if (!mfplat_dll) {
DWORD error_message_id = GetLastError();
LPSTR message_buffer = nullptr;
size_t size =
FormatMessageA(FORMAT_MESSAGE_ALLOCATE_BUFFER | FORMAT_MESSAGE_FROM_SYSTEM |
FORMAT_MESSAGE_IGNORE_INSERTS,
nullptr, error_message_id, MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT),
reinterpret_cast<LPSTR>(&message_buffer), 0, nullptr);
std::string message(message_buffer, size);
LocalFree(message_buffer);
LOG_ERROR(Audio_DSP, "Could not load mfplat.dll: {}", message);
return false;
}
MFStartup = Symbol<HRESULT(ULONG, DWORD)>(mfplat_dll.get(), "MFStartup");
if (!MFStartup) {
LOG_ERROR(Audio_DSP, "Cannot load function MFStartup");
return false;
}
MFShutdown = Symbol<HRESULT(void)>(mfplat_dll.get(), "MFShutdown");
if (!MFShutdown) {
LOG_ERROR(Audio_DSP, "Cannot load function MFShutdown");
return false;
}
MFShutdownObject = Symbol<HRESULT(IUnknown*)>(mf_dll.get(), "MFShutdownObject");
if (!MFShutdownObject) {
LOG_ERROR(Audio_DSP, "Cannot load function MFShutdownObject");
return false;
}
MFCreateAlignedMemoryBuffer = Symbol<HRESULT(DWORD, DWORD, IMFMediaBuffer**)>(
mfplat_dll.get(), "MFCreateAlignedMemoryBuffer");
if (!MFCreateAlignedMemoryBuffer) {
LOG_ERROR(Audio_DSP, "Cannot load function MFCreateAlignedMemoryBuffer");
return false;
}
MFCreateSample = Symbol<HRESULT(IMFSample**)>(mfplat_dll.get(), "MFCreateSample");
if (!MFCreateSample) {
LOG_ERROR(Audio_DSP, "Cannot load function MFCreateSample");
return false;
}
MFTEnumEx =
Symbol<HRESULT(GUID, UINT32, const MFT_REGISTER_TYPE_INFO*, const MFT_REGISTER_TYPE_INFO*,
IMFActivate***, UINT32*)>(mfplat_dll.get(), "MFTEnumEx");
if (!MFTEnumEx) {
LOG_ERROR(Audio_DSP, "Cannot load function MFTEnumEx");
return false;
}
MFCreateMediaType = Symbol<HRESULT(IMFMediaType**)>(mfplat_dll.get(), "MFCreateMediaType");
if (!MFCreateMediaType) {
LOG_ERROR(Audio_DSP, "Cannot load function MFCreateMediaType");
return false;
}
return true;
}
Symbol<HRESULT(ULONG, DWORD)> MFStartup;
Symbol<HRESULT(void)> MFShutdown;
Symbol<HRESULT(IUnknown*)> MFShutdownObject;
Symbol<HRESULT(DWORD, DWORD, IMFMediaBuffer**)> MFCreateAlignedMemoryBuffer;
Symbol<HRESULT(IMFSample**)> MFCreateSample;
Symbol<HRESULT(GUID, UINT32, const MFT_REGISTER_TYPE_INFO*, const MFT_REGISTER_TYPE_INFO*,
IMFActivate***, UINT32*)>
MFTEnumEx;
Symbol<HRESULT(IMFMediaType**)> MFCreateMediaType;
} // namespace MFDecoder

125
src/audio_core/hle/wmf_decoder_utils.h
View file

@ -1,125 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <memory>
#include <optional>
#include <string>
#include <tuple>
#include <vector>
#include <comdef.h>
#include <mfapi.h>
#include <mferror.h>
#include <mfidl.h>
#include <mftransform.h>
#include "adts.h"
namespace MFDecoder {
template <typename T>
struct Symbol {
Symbol() = default;
Symbol(HMODULE dll, const char* name) {
if (dll) {
ptr_symbol = reinterpret_cast<T*>(GetProcAddress(dll, name));
}
}
operator T*() const {
return ptr_symbol;
}
explicit operator bool() const {
return ptr_symbol != nullptr;
}
T* ptr_symbol = nullptr;
};
// Runtime load the MF symbols to prevent mf.dll not found errors on citra load
extern Symbol<HRESULT(ULONG, DWORD)> MFStartup;
extern Symbol<HRESULT(void)> MFShutdown;
extern Symbol<HRESULT(IUnknown*)> MFShutdownObject;
extern Symbol<HRESULT(DWORD, DWORD, IMFMediaBuffer**)> MFCreateAlignedMemoryBuffer;
extern Symbol<HRESULT(IMFSample**)> MFCreateSample;
extern Symbol<HRESULT(GUID, UINT32, const MFT_REGISTER_TYPE_INFO*, const MFT_REGISTER_TYPE_INFO*,
IMFActivate***, UINT32*)>
MFTEnumEx;
extern Symbol<HRESULT(IMFMediaType**)> MFCreateMediaType;
enum class MFOutputState { FatalError, OK, NeedMoreInput, NeedReconfig, HaveMoreData };
enum class MFInputState { FatalError, OK, NotAccepted };
// utility functions / templates
template <class T>
struct MFRelease {
void operator()(T* pointer) const {
pointer->Release();
};
};
template <>
struct MFRelease<IMFTransform> {
void operator()(IMFTransform* pointer) const {
MFShutdownObject(pointer);
pointer->Release();
};
};
// wrapper facilities for dealing with pointers
template <typename T>
using unique_mfptr = std::unique_ptr<T, MFRelease<T>>;
template <typename SmartPtr, typename RawPtr>
class AmpImpl {
public:
AmpImpl(SmartPtr& smart_ptr) : smart_ptr(smart_ptr) {}
~AmpImpl() {
smart_ptr.reset(raw_ptr);
}
operator RawPtr*() {
return &raw_ptr;
}
private:
SmartPtr& smart_ptr;
RawPtr raw_ptr = nullptr;
};
template <typename SmartPtr>
auto Amp(SmartPtr& smart_ptr) {
return AmpImpl<SmartPtr, decltype(smart_ptr.get())>(smart_ptr);
}
// convient function for formatting error messages
void ReportError(std::string msg, HRESULT hr);
// data type for transferring ADTS metadata between functions
struct ADTSMeta {
AudioCore::ADTSData ADTSHeader;
u8 AACTag[14];
};
// exported functions
/// Loads the symbols from mf.dll at runtime. Returns false if the symbols can't be loaded
bool InitMFDLL();
unique_mfptr<IMFTransform> MFDecoderInit(GUID audio_format = MFAudioFormat_AAC);
unique_mfptr<IMFSample> CreateSample(const void* data, DWORD len, DWORD alignment = 1,
LONGLONG duration = 0);
bool SelectInputMediaType(IMFTransform* transform, int in_stream_id,
const AudioCore::ADTSData& adts, const UINT8* user_data,
UINT32 user_data_len, GUID audio_format = MFAudioFormat_AAC);
std::optional<ADTSMeta> DetectMediaType(const u8* buffer, std::size_t len);
bool SelectOutputMediaType(IMFTransform* transform, int out_stream_id,
GUID audio_format = MFAudioFormat_PCM);
void MFFlush(IMFTransform* transform);
MFInputState SendSample(IMFTransform* transform, DWORD in_stream_id, IMFSample* in_sample);
std::tuple<MFOutputState, unique_mfptr<IMFSample>> ReceiveSample(IMFTransform* transform,
DWORD out_stream_id);
std::optional<std::vector<f32>> CopySampleToBuffer(IMFSample* sample);
} // namespace MFDecoder

2
src/common/CMakeLists.txt
View file

@ -76,8 +76,6 @@ add_library(citra_common STATIC
construct.h
dynamic_library/dynamic_library.cpp
dynamic_library/dynamic_library.h
dynamic_library/fdk-aac.cpp
dynamic_library/fdk-aac.h
dynamic_library/ffmpeg.cpp
dynamic_library/ffmpeg.h
error.cpp

57
src/common/dynamic_library/fdk-aac.cpp
View file

@ -1,57 +0,0 @@
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <memory>
#include "common/dynamic_library/dynamic_library.h"
#include "common/dynamic_library/fdk-aac.h"
#include "common/logging/log.h"
namespace DynamicLibrary::FdkAac {
aacDecoder_GetLibInfo_func aacDecoder_GetLibInfo;
aacDecoder_Open_func aacDecoder_Open;
aacDecoder_Close_func aacDecoder_Close;
aacDecoder_SetParam_func aacDecoder_SetParam;
aacDecoder_GetStreamInfo_func aacDecoder_GetStreamInfo;
aacDecoder_DecodeFrame_func aacDecoder_DecodeFrame;
aacDecoder_Fill_func aacDecoder_Fill;
static std::unique_ptr<Common::DynamicLibrary> fdk_aac;
#define LOAD_SYMBOL(library, name) \
any_failed = any_failed || (name = library->GetSymbol<name##_func>(#name)) == nullptr
bool LoadFdkAac() {
if (fdk_aac) {
return true;
}
fdk_aac = std::make_unique<Common::DynamicLibrary>("fdk-aac", 2);
if (!fdk_aac->IsLoaded()) {
LOG_WARNING(Common, "Could not dynamically load libfdk-aac: {}", fdk_aac->GetLoadError());
fdk_aac.reset();
return false;
}
auto any_failed = false;
LOAD_SYMBOL(fdk_aac, aacDecoder_GetLibInfo);
LOAD_SYMBOL(fdk_aac, aacDecoder_Open);
LOAD_SYMBOL(fdk_aac, aacDecoder_Close);
LOAD_SYMBOL(fdk_aac, aacDecoder_SetParam);
LOAD_SYMBOL(fdk_aac, aacDecoder_GetStreamInfo);
LOAD_SYMBOL(fdk_aac, aacDecoder_DecodeFrame);
LOAD_SYMBOL(fdk_aac, aacDecoder_Fill);
if (any_failed) {
LOG_WARNING(Common, "Could not find all required functions in libfdk-aac.");
fdk_aac.reset();
return false;
}
LOG_INFO(Common, "Successfully loaded libfdk-aac.");
return true;
}
} // namespace DynamicLibrary::FdkAac

34
src/common/dynamic_library/fdk-aac.h
View file

@ -1,34 +0,0 @@
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
extern "C" {
#include <fdk-aac/aacdecoder_lib.h>
}
namespace DynamicLibrary::FdkAac {
typedef INT (*aacDecoder_GetLibInfo_func)(LIB_INFO* info);
typedef HANDLE_AACDECODER (*aacDecoder_Open_func)(TRANSPORT_TYPE transportFmt, UINT nrOfLayers);
typedef void (*aacDecoder_Close_func)(HANDLE_AACDECODER self);
typedef AAC_DECODER_ERROR (*aacDecoder_SetParam_func)(const HANDLE_AACDECODER self,
const AACDEC_PARAM param, const INT value);
typedef CStreamInfo* (*aacDecoder_GetStreamInfo_func)(HANDLE_AACDECODER self);
typedef AAC_DECODER_ERROR (*aacDecoder_DecodeFrame_func)(HANDLE_AACDECODER self, INT_PCM* pTimeData,
const INT timeDataSize, const UINT flags);
typedef AAC_DECODER_ERROR (*aacDecoder_Fill_func)(HANDLE_AACDECODER self, UCHAR* pBuffer[],
const UINT bufferSize[], UINT* bytesValid);
extern aacDecoder_GetLibInfo_func aacDecoder_GetLibInfo;
extern aacDecoder_Open_func aacDecoder_Open;
extern aacDecoder_Close_func aacDecoder_Close;
extern aacDecoder_SetParam_func aacDecoder_SetParam;
extern aacDecoder_GetStreamInfo_func aacDecoder_GetStreamInfo;
extern aacDecoder_DecodeFrame_func aacDecoder_DecodeFrame;
extern aacDecoder_Fill_func aacDecoder_Fill;
bool LoadFdkAac();
} // namespace DynamicLibrary::FdkAac

1
src/tests/CMakeLists.txt
View file

@ -12,7 +12,6 @@ add_executable(tests
core/memory/vm_manager.cpp
precompiled_headers.h
audio_core/hle/hle.cpp
audio_core/hle/adts_reader.cpp
audio_core/lle/lle.cpp
audio_core/audio_fixures.h
audio_core/decoder_tests.cpp

77
src/tests/audio_core/hle/adts_reader.cpp
View file

@ -1,77 +0,0 @@
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <catch2/catch_test_macros.hpp>
#include <fmt/format.h>
#include "audio_core/hle/adts.h"
namespace {
constexpr std::array<u32, 16> freq_table = {96000, 88200, 64000, 48000, 44100, 32000, 24000, 22050,
16000, 12000, 11025, 8000, 7350, 0, 0, 0};
constexpr std::array<u8, 8> channel_table = {0, 1, 2, 3, 4, 5, 6, 8};
AudioCore::ADTSData ParseADTS_Old(const unsigned char* buffer) {
u32 tmp = 0;
AudioCore::ADTSData out{};
// sync word 0xfff
tmp = (buffer[0] << 8) | (buffer[1] & 0xf0);
if ((tmp & 0xffff) != 0xfff0) {
out.length = 0;
return out;
}
// bit 16 = no CRC
out.header_length = (buffer[1] & 0x1) ? 7 : 9;
out.mpeg2 = (buffer[1] >> 3) & 0x1;
// bit 17 to 18
out.profile = (buffer[2] >> 6) + 1;
// bit 19 to 22
tmp = (buffer[2] >> 2) & 0xf;
out.samplerate_idx = tmp;
out.samplerate = (tmp > 15) ? 0 : freq_table[tmp];
// bit 24 to 26
tmp = ((buffer[2] & 0x1) << 2) | ((buffer[3] >> 6) & 0x3);
out.channel_idx = tmp;
out.channels = (tmp > 7) ? 0 : channel_table[tmp];
// bit 55 to 56
out.framecount = (buffer[6] & 0x3) + 1;
// bit 31 to 43
tmp = (buffer[3] & 0x3) << 11;
tmp |= (buffer[4] << 3) & 0x7f8;
tmp |= (buffer[5] >> 5) & 0x7;
out.length = tmp;
return out;
}
} // namespace
TEST_CASE("ParseADTS fuzz", "[audio_core][hle]") {
for (u32 i = 0; i < 0x10000; i++) {
std::array<u8, 7> adts_header;
std::string adts_header_string = "ADTS Header: ";
for (auto& it : adts_header) {
it = static_cast<u8>(rand());
adts_header_string.append(fmt::format("{:2X} ", it));
}
INFO(adts_header_string);
AudioCore::ADTSData out_old_impl =
ParseADTS_Old(reinterpret_cast<const unsigned char*>(adts_header.data()));
AudioCore::ADTSData out = AudioCore::ParseADTS(adts_header.data());
REQUIRE(out_old_impl.length == out.length);
REQUIRE(out_old_impl.channels == out.channels);
REQUIRE(out_old_impl.channel_idx == out.channel_idx);
REQUIRE(out_old_impl.framecount == out.framecount);
REQUIRE(out_old_impl.header_length == out.header_length);
REQUIRE(out_old_impl.mpeg2 == out.mpeg2);
REQUIRE(out_old_impl.profile == out.profile);
REQUIRE(out_old_impl.samplerate == out.samplerate);
REQUIRE(out_old_impl.samplerate_idx == out.samplerate_idx);
}
}