custom_tex_manager: Implement hot-reloading

* video_core: Refactor GLSL fragment emitter

* shader: Add back custom normal maps

.ci/ios.sh

@@ -1,10 +1,5 @@
 #!/bin/bash -ex
 
-# TODO: Work around pip install issues with Python 3.12 in the GitHub runner image.
-# See: https://github.com/actions/runner-images/issues/8709
-PYTHON_PATH=$(brew --prefix python@3.11)
-export PATH=$PYTHON_PATH/bin:$PYTHON_PATH/libexec/bin:$PATH
-
 mkdir build && cd build
 cmake .. -GNinja \
     -DCMAKE_BUILD_TYPE=Release \

.ci/macos.sh

@@ -1,10 +1,5 @@
 #!/bin/bash -ex
 
-# TODO: Work around pip install issues with Python 3.12 in the GitHub runner image.
-# See: https://github.com/actions/runner-images/issues/8709
-PYTHON_PATH=$(brew --prefix python@3.11)
-export PATH=$PYTHON_PATH/bin:$PYTHON_PATH/libexec/bin:$PATH
-
 mkdir build && cd build
 cmake .. -GNinja \
     -DCMAKE_BUILD_TYPE=Release \

.gitmodules

@@ -79,9 +79,15 @@
 [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
 [submodule "libadrenotools"]
     path = externals/libadrenotools
     url = https://github.com/bylaws/libadrenotools
+[submodule "oaknut"]
+	path = externals/oaknut
+	url = https://github.com/merryhime/oaknut.git

CMakeLists.txt

@@ -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

CMakeModules/DownloadExternals.cmake

@@ -67,13 +67,22 @@ function(download_qt target)
 
     if (NOT EXISTS "${prefix}")
         message(STATUS "Downloading binaries for Qt...")
+        set(AQT_PREBUILD_BASE_URL "https://github.com/miurahr/aqtinstall/releases/download/v3.1.9")
         if (WIN32)
             set(aqt_path "${base_path}/aqt.exe")
             file(DOWNLOAD
-                https://github.com/miurahr/aqtinstall/releases/download/v3.1.7/aqt.exe
+                ${AQT_PREBUILD_BASE_URL}/aqt.exe
                 ${aqt_path} SHOW_PROGRESS)
             execute_process(COMMAND ${aqt_path} ${install_args}
                     WORKING_DIRECTORY ${base_path})
+        elseif (APPLE)
+            set(aqt_path "${base_path}/aqt-macos")
+            file(DOWNLOAD
+                ${AQT_PREBUILD_BASE_URL}/aqt-macos
+                ${aqt_path} SHOW_PROGRESS)
+            execute_process(COMMAND chmod +x ${aqt_path})
+            execute_process(COMMAND ${aqt_path} ${install_args}
+                    WORKING_DIRECTORY ${base_path})
         else()
             # aqt does not offer binary releases for other platforms, so download and run from pip.
             set(aqt_install_path "${base_path}/aqt")

CMakeModules/GenerateSCMRev.cmake

@@ -10,16 +10,20 @@ set(HASH_FILES
     "${VIDEO_CORE}/renderer_opengl/gl_shader_util.h"
     "${VIDEO_CORE}/renderer_vulkan/vk_shader_util.cpp"
     "${VIDEO_CORE}/renderer_vulkan/vk_shader_util.h"
+    "${VIDEO_CORE}/shader/generator/glsl_fs_shader_gen.cpp"
+    "${VIDEO_CORE}/shader/generator/glsl_fs_shader_gen.h"
     "${VIDEO_CORE}/shader/generator/glsl_shader_decompiler.cpp"
     "${VIDEO_CORE}/shader/generator/glsl_shader_decompiler.h"
     "${VIDEO_CORE}/shader/generator/glsl_shader_gen.cpp"
     "${VIDEO_CORE}/shader/generator/glsl_shader_gen.h"
+    "${VIDEO_CORE}/shader/generator/pica_fs_config.cpp"
+    "${VIDEO_CORE}/shader/generator/pica_fs_config.h"
     "${VIDEO_CORE}/shader/generator/shader_gen.cpp"
     "${VIDEO_CORE}/shader/generator/shader_gen.h"
     "${VIDEO_CORE}/shader/generator/shader_uniforms.cpp"
     "${VIDEO_CORE}/shader/generator/shader_uniforms.h"
-    "${VIDEO_CORE}/shader/generator/spv_shader_gen.cpp"
+    "${VIDEO_CORE}/shader/generator/spv_fs_shader_gen.cpp"
-    "${VIDEO_CORE}/shader/generator/spv_shader_gen.h"
+    "${VIDEO_CORE}/shader/generator/spv_fs_shader_gen.h"
     "${VIDEO_CORE}/shader/shader.cpp"
     "${VIDEO_CORE}/shader/shader.h"
     "${VIDEO_CORE}/pica.cpp"

externals/CMakeLists.txt

@@ -85,6 +85,11 @@ if ("x86_64" IN_LIST ARCHITECTURE)
     endif()
 endif()
 
+# Oaknut
+if ("arm64" IN_LIST ARCHITECTURE)
+    add_subdirectory(oaknut EXCLUDE_FROM_ALL)
+endif()
+
 # Dynarmic
 if ("x86_64" IN_LIST ARCHITECTURE OR "arm64" IN_LIST ARCHITECTURE)
     if(USE_SYSTEM_DYNARMIC)
@@ -156,24 +161,12 @@ 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")

externals/cmake-modules/CitraHandleSystemLibs.cmake

@@ -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

externals/faad2/CMakeLists.txt

@@ -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
+)

src/audio_core/CMakeLists.txt

@@ -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/faad2_decoder.cpp
-    hle/fdk_decoder.h
+    hle/faad2_decoder.h
-    hle/ffmpeg_decoder.cpp
-    hle/ffmpeg_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)
+target_link_libraries(audio_core PRIVATE faad2 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()
 
 if(ENABLE_SDL2)
     target_link_libraries(audio_core PRIVATE SDL2::SDL2)

src/audio_core/hle/adts.h

@@ -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

src/audio_core/hle/adts_reader.cpp

@@ -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

src/audio_core/hle/audiotoolbox_decoder.cpp

@@ -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

src/audio_core/hle/faad2_decoder.cpp

@@ -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

src/audio_core/hle/faad2_decoder.h

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

src/audio_core/hle/fdk_decoder.cpp

@@ -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

src/audio_core/hle/fdk_decoder.h

@@ -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

src/audio_core/hle/ffmpeg_decoder.cpp

@@ -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

src/audio_core/hle/ffmpeg_decoder.h

@@ -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

src/audio_core/hle/hle.cpp

@@ -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/faad2_decoder.h"
-#include "audio_core/hle/ffmpeg_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);
+            return std::make_unique<HLE::FAAD2Decoder>(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);
         },
 };
 

src/audio_core/hle/mediandk_decoder.cpp

@@ -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

src/audio_core/hle/mediandk_decoder.h

@@ -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

src/audio_core/hle/wmf_decoder.cpp

@@ -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

src/audio_core/hle/wmf_decoder.h

@@ -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

src/audio_core/hle/wmf_decoder_utils.cpp

@@ -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

src/audio_core/hle/wmf_decoder_utils.h

@@ -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

src/citra/emu_window/emu_window_sdl2_gl.cpp

@@ -42,10 +42,8 @@ private:
     SDL_GLContext context;
 };
 
-EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(Core::System& system_, bool fullscreen, bool is_secondary)
+static SDL_Window* CreateGLWindow(const std::string& window_title, bool gles) {
-    : EmuWindow_SDL2{system_, is_secondary} {
+    if (gles) {
-    // Initialize the window
-    if (Settings::values.use_gles) {
         SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 3);
         SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 2);
         SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_ES);
@@ -54,7 +52,16 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(Core::System& system_, bool fullscreen, boo
         SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 3);
         SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_CORE);
     }
+    return SDL_CreateWindow(window_title.c_str(),
+                            SDL_WINDOWPOS_UNDEFINED, // x position
+                            SDL_WINDOWPOS_UNDEFINED, // y position
+                            Core::kScreenTopWidth,
+                            Core::kScreenTopHeight + Core::kScreenBottomHeight,
+                            SDL_WINDOW_OPENGL | SDL_WINDOW_RESIZABLE | SDL_WINDOW_ALLOW_HIGHDPI);
+}
 
+EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(Core::System& system_, bool fullscreen, bool is_secondary)
+    : EmuWindow_SDL2{system_, is_secondary} {
     SDL_GL_SetAttribute(SDL_GL_DOUBLEBUFFER, 1);
     SDL_GL_SetAttribute(SDL_GL_RED_SIZE, 8);
     SDL_GL_SetAttribute(SDL_GL_GREEN_SIZE, 8);
@@ -71,16 +78,16 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(Core::System& system_, bool fullscreen, boo
 
     std::string window_title = fmt::format("Citra {} | {}-{}", Common::g_build_fullname,
                                            Common::g_scm_branch, Common::g_scm_desc);
-    render_window =
-        SDL_CreateWindow(window_title.c_str(),
-                         SDL_WINDOWPOS_UNDEFINED, // x position
-                         SDL_WINDOWPOS_UNDEFINED, // y position
-                         Core::kScreenTopWidth, Core::kScreenTopHeight + Core::kScreenBottomHeight,
-                         SDL_WINDOW_OPENGL | SDL_WINDOW_RESIZABLE | SDL_WINDOW_ALLOW_HIGHDPI);
 
+    // First, try to create a context with the requested type.
+    render_window = CreateGLWindow(window_title, Settings::values.use_gles.GetValue());
     if (render_window == nullptr) {
-        LOG_CRITICAL(Frontend, "Failed to create SDL2 window: {}", SDL_GetError());
+        // On failure, fall back to context with flipped type.
-        exit(1);
+        render_window = CreateGLWindow(window_title, !Settings::values.use_gles.GetValue());
+        if (render_window == nullptr) {
+            LOG_CRITICAL(Frontend, "Failed to create SDL2 window: {}", SDL_GetError());
+            exit(1);
+        }
     }
 
     strict_context_required = std::strcmp(SDL_GetCurrentVideoDriver(), "wayland") == 0;
@@ -106,7 +113,11 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(Core::System& system_, bool fullscreen, boo
     }
 
     render_window_id = SDL_GetWindowID(render_window);
-    auto gl_load_func = Settings::values.use_gles ? gladLoadGLES2Loader : gladLoadGLLoader;
+
+    int profile_mask = 0;
+    SDL_GL_GetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, &profile_mask);
+    auto gl_load_func =
+        profile_mask == SDL_GL_CONTEXT_PROFILE_ES ? gladLoadGLES2Loader : gladLoadGLLoader;
 
     if (!gl_load_func(static_cast<GLADloadproc>(SDL_GL_GetProcAddress))) {
         LOG_CRITICAL(Frontend, "Failed to initialize GL functions: {}", SDL_GetError());

src/citra_qt/bootmanager.cpp

@@ -138,37 +138,50 @@ void EmuThread::run() {
 }
 
 #ifdef HAS_OPENGL
+static std::unique_ptr<QOpenGLContext> CreateQOpenGLContext(bool gles) {
+    QSurfaceFormat format;
+    if (gles) {
+        format.setRenderableType(QSurfaceFormat::RenderableType::OpenGLES);
+        format.setVersion(3, 2);
+    } else {
+        format.setRenderableType(QSurfaceFormat::RenderableType::OpenGL);
+        format.setVersion(4, 3);
+    }
+    format.setProfile(QSurfaceFormat::CoreProfile);
+
+    if (Settings::values.renderer_debug) {
+        format.setOption(QSurfaceFormat::FormatOption::DebugContext);
+    }
+
+    // TODO: expose a setting for buffer value (ie default/single/double/triple)
+    format.setSwapBehavior(QSurfaceFormat::DefaultSwapBehavior);
+    format.setSwapInterval(0);
+
+    auto context = std::make_unique<QOpenGLContext>();
+    context->setFormat(format);
+    if (!context->create()) {
+        LOG_ERROR(Frontend, "Unable to create OpenGL context with GLES = {}", gles);
+        return nullptr;
+    }
+    return context;
+}
+
 class OpenGLSharedContext : public Frontend::GraphicsContext {
 public:
     /// Create the original context that should be shared from
     explicit OpenGLSharedContext() {
-        QSurfaceFormat format;
+        // First, try to create a context with the requested type.
-
+        context = CreateQOpenGLContext(Settings::values.use_gles.GetValue());
-        if (Settings::values.use_gles) {
+        if (context == nullptr) {
-            format.setRenderableType(QSurfaceFormat::RenderableType::OpenGLES);
+            // On failure, fall back to context with flipped type.
-            format.setVersion(3, 2);
+            context = CreateQOpenGLContext(!Settings::values.use_gles.GetValue());
-        } else {
+            if (context == nullptr) {
-            format.setRenderableType(QSurfaceFormat::RenderableType::OpenGL);
+                LOG_ERROR(Frontend, "Unable to create any OpenGL context.");
-            format.setVersion(4, 3);
+            }
-        }
-        format.setProfile(QSurfaceFormat::CoreProfile);
-
-        if (Settings::values.renderer_debug) {
-            format.setOption(QSurfaceFormat::FormatOption::DebugContext);
-        }
-
-        // TODO: expose a setting for buffer value (ie default/single/double/triple)
-        format.setSwapBehavior(QSurfaceFormat::DefaultSwapBehavior);
-        format.setSwapInterval(0);
-
-        context = std::make_unique<QOpenGLContext>();
-        context->setFormat(format);
-        if (!context->create()) {
-            LOG_ERROR(Frontend, "Unable to create main openGL context");
         }
 
         offscreen_surface = std::make_unique<QOffscreenSurface>(nullptr);
-        offscreen_surface->setFormat(format);
+        offscreen_surface->setFormat(context->format());
         offscreen_surface->create();
         surface = offscreen_surface.get();
     }
@@ -184,7 +197,7 @@ public:
         context->setShareContext(share_context);
         context->setFormat(format);
         if (!context->create()) {
-            LOG_ERROR(Frontend, "Unable to create shared openGL context");
+            LOG_ERROR(Frontend, "Unable to create shared OpenGL context");
         }
 
         surface = main_surface;
@@ -194,6 +207,10 @@ public:
         OpenGLSharedContext::DoneCurrent();
     }
 
+    bool IsGLES() override {
+        return context->format().renderableType() == QSurfaceFormat::RenderableType::OpenGLES;
+    }
+
     void SwapBuffers() override {
         context->swapBuffers(surface);
     }
@@ -739,8 +756,9 @@ bool GRenderWindow::LoadOpenGL() {
 #ifdef HAS_OPENGL
     auto context = CreateSharedContext();
     auto scope = context->Acquire();
+    const auto gles = context->IsGLES();
 
-    auto gl_load_func = Settings::values.use_gles ? gladLoadGLES2Loader : gladLoadGLLoader;
+    auto gl_load_func = gles ? gladLoadGLES2Loader : gladLoadGLLoader;
     if (!gl_load_func(GetProcAddressGL)) {
         QMessageBox::warning(
             this, tr("Error while initializing OpenGL!"),
@@ -751,14 +769,14 @@ bool GRenderWindow::LoadOpenGL() {
     const QString renderer =
         QString::fromUtf8(reinterpret_cast<const char*>(glGetString(GL_RENDERER)));
 
-    if (!Settings::values.use_gles && !GLAD_GL_VERSION_4_3) {
+    if (!gles && !GLAD_GL_VERSION_4_3) {
         LOG_ERROR(Frontend, "GPU does not support OpenGL 4.3: {}", renderer.toStdString());
         QMessageBox::warning(this, tr("Error while initializing OpenGL 4.3!"),
                              tr("Your GPU may not support OpenGL 4.3, or you do not have the "
                                 "latest graphics driver.<br><br>GL Renderer:<br>%1")
                                  .arg(renderer));
         return false;
-    } else if (Settings::values.use_gles && !GLAD_GL_ES_VERSION_3_2) {
+    } else if (gles && !GLAD_GL_ES_VERSION_3_2) {
         LOG_ERROR(Frontend, "GPU does not support OpenGL ES 3.2: {}", renderer.toStdString());
         QMessageBox::warning(this, tr("Error while initializing OpenGL ES 3.2!"),
                              tr("Your GPU may not support OpenGL ES 3.2, or you do not have the "

src/citra_qt/game_list.cpp

@@ -560,7 +560,9 @@ void GameList::AddGamePopup(QMenu& context_menu, const QString& path, const QStr
     QAction* properties = context_menu.addAction(tr("Properties"));
 
     const u32 program_id_high = (program_id >> 32) & 0xFFFFFFFF;
-    const bool is_application = program_id_high == 0x00040000 || program_id_high == 0x00040010;
+    // TODO: Use proper bitmasks for these kinds of checks.
+    const bool is_application = program_id_high == 0x00040000 || program_id_high == 0x00040002 ||
+                                program_id_high == 0x00040010;
 
     bool opengl_cache_exists = false;
     ForEachOpenGLCacheFile(

src/common/CMakeLists.txt

@@ -21,16 +21,20 @@ add_custom_command(OUTPUT scm_rev.cpp
       "${VIDEO_CORE}/renderer_opengl/gl_shader_util.h"
       "${VIDEO_CORE}/renderer_vulkan/vk_shader_util.cpp"
       "${VIDEO_CORE}/renderer_vulkan/vk_shader_util.h"
+      "${VIDEO_CORE}/shader/generator/glsl_fs_shader_gen.cpp"
+      "${VIDEO_CORE}/shader/generator/glsl_fs_shader_gen.h"
       "${VIDEO_CORE}/shader/generator/glsl_shader_decompiler.cpp"
       "${VIDEO_CORE}/shader/generator/glsl_shader_decompiler.h"
       "${VIDEO_CORE}/shader/generator/glsl_shader_gen.cpp"
       "${VIDEO_CORE}/shader/generator/glsl_shader_gen.h"
+      "${VIDEO_CORE}/shader/generator/pica_fs_config.cpp"
+      "${VIDEO_CORE}/shader/generator/pica_fs_config.h"
       "${VIDEO_CORE}/shader/generator/shader_gen.cpp"
       "${VIDEO_CORE}/shader/generator/shader_gen.h"
       "${VIDEO_CORE}/shader/generator/shader_uniforms.cpp"
       "${VIDEO_CORE}/shader/generator/shader_uniforms.h"
-      "${VIDEO_CORE}/shader/generator/spv_shader_gen.cpp"
+      "${VIDEO_CORE}/shader/generator/spv_fs_shader_gen.cpp"
-      "${VIDEO_CORE}/shader/generator/spv_shader_gen.h"
+      "${VIDEO_CORE}/shader/generator/spv_fs_shader_gen.h"
       "${VIDEO_CORE}/shader/shader.cpp"
       "${VIDEO_CORE}/shader/shader.h"
       "${VIDEO_CORE}/pica.cpp"
@@ -53,6 +57,8 @@ add_custom_command(OUTPUT scm_rev.cpp
 add_library(citra_common STATIC
     aarch64/cpu_detect.cpp
     aarch64/cpu_detect.h
+    aarch64/oaknut_abi.h
+    aarch64/oaknut_util.h
     alignment.h
     android_storage.h
     android_storage.cpp
@@ -76,8 +82,6 @@ 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
@@ -85,6 +89,8 @@ add_library(citra_common STATIC
     expected.h
     file_util.cpp
     file_util.h
+    file_watcher.cpp
+    file_watcher.h
     hash.h
     linear_disk_cache.h
     literals.h
@@ -181,6 +187,10 @@ if ("x86_64" IN_LIST ARCHITECTURE)
     target_link_libraries(citra_common PRIVATE xbyak)
 endif()
 
+if ("arm64" IN_LIST ARCHITECTURE)
+    target_link_libraries(citra_common PRIVATE oaknut)
+endif()
+
 if (CITRA_USE_PRECOMPILED_HEADERS)
     target_precompile_headers(citra_common PRIVATE precompiled_headers.h)
 endif()

src/common/aarch64/oaknut_abi.h

@@ -0,0 +1,155 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/arch.h"
+#if CITRA_ARCH(arm64)
+
+#include <bitset>
+#include <initializer_list>
+#include <oaknut/oaknut.hpp>
+#include "common/assert.h"
+
+namespace Common::A64 {
+
+constexpr std::size_t RegToIndex(const oaknut::Reg& reg) {
+    ASSERT(reg.index() != 31); // ZR not allowed
+    return reg.index() + (reg.is_vector() ? 32 : 0);
+}
+
+constexpr oaknut::XReg IndexToXReg(std::size_t reg_index) {
+    ASSERT(reg_index <= 30);
+    return oaknut::XReg(static_cast<int>(reg_index));
+}
+
+constexpr oaknut::VReg IndexToVReg(std::size_t reg_index) {
+    ASSERT(reg_index >= 32 && reg_index < 64);
+    return oaknut::QReg(static_cast<int>(reg_index - 32));
+}
+
+constexpr oaknut::Reg IndexToReg(std::size_t reg_index) {
+    if (reg_index < 32) {
+        return IndexToXReg(reg_index);
+    } else {
+        return IndexToVReg(reg_index);
+    }
+}
+
+inline constexpr std::bitset<64> BuildRegSet(std::initializer_list<oaknut::Reg> regs) {
+    std::bitset<64> bits;
+    for (const oaknut::Reg& reg : regs) {
+        bits.set(RegToIndex(reg));
+    }
+    return bits;
+}
+
+constexpr inline std::bitset<64> ABI_ALL_GPRS(0x00000000'7FFFFFFF);
+constexpr inline std::bitset<64> ABI_ALL_FPRS(0xFFFFFFFF'00000000);
+
+constexpr inline oaknut::XReg ABI_RETURN = oaknut::util::X0;
+constexpr inline oaknut::XReg ABI_PARAM1 = oaknut::util::X0;
+constexpr inline oaknut::XReg ABI_PARAM2 = oaknut::util::X1;
+constexpr inline oaknut::XReg ABI_PARAM3 = oaknut::util::X2;
+constexpr inline oaknut::XReg ABI_PARAM4 = oaknut::util::X3;
+
+constexpr std::bitset<64> ABI_ALL_CALLER_SAVED = 0xffffffff'4000ffff;
+constexpr std::bitset<64> ABI_ALL_CALLEE_SAVED = 0x0000ff00'7ff80000;
+
+struct ABIFrameInfo {
+    u32 subtraction;
+    u32 fprs_offset;
+};
+
+inline ABIFrameInfo ABI_CalculateFrameSize(std::bitset<64> regs, std::size_t frame_size) {
+    const size_t gprs_count = (regs & ABI_ALL_GPRS).count();
+    const size_t fprs_count = (regs & ABI_ALL_FPRS).count();
+
+    const size_t gprs_size = (gprs_count + 1) / 2 * 16;
+    const size_t fprs_size = fprs_count * 16;
+
+    size_t total_size = 0;
+    total_size += gprs_size;
+    const size_t fprs_base_subtraction = total_size;
+    total_size += fprs_size;
+    total_size += frame_size;
+
+    return ABIFrameInfo{static_cast<u32>(total_size), static_cast<u32>(fprs_base_subtraction)};
+}
+
+inline void ABI_PushRegisters(oaknut::CodeGenerator& code, std::bitset<64> regs,
+                              std::size_t frame_size = 0) {
+    using namespace oaknut;
+    using namespace oaknut::util;
+    auto frame_info = ABI_CalculateFrameSize(regs, frame_size);
+
+    // Allocate stack-space
+    if (frame_info.subtraction != 0) {
+        code.SUB(SP, SP, frame_info.subtraction);
+    }
+
+    // TODO(wunk): Push pairs of registers at a time with STP
+    std::size_t offset = 0;
+    for (std::size_t i = 0; i < 32; ++i) {
+        if (regs[i] && ABI_ALL_GPRS[i]) {
+            const XReg reg = IndexToXReg(i);
+            code.STR(reg, SP, offset);
+            offset += 8;
+        }
+    }
+
+    offset = 0;
+    for (std::size_t i = 32; i < 64; ++i) {
+        if (regs[i] && ABI_ALL_FPRS[i]) {
+            const VReg reg = IndexToVReg(i);
+            code.STR(reg.toQ(), SP, u16(frame_info.fprs_offset + offset));
+            offset += 16;
+        }
+    }
+
+    // Allocate frame-space
+    if (frame_size != 0) {
+        code.SUB(SP, SP, frame_size);
+    }
+}
+
+inline void ABI_PopRegisters(oaknut::CodeGenerator& code, std::bitset<64> regs,
+                             std::size_t frame_size = 0) {
+    using namespace oaknut;
+    using namespace oaknut::util;
+    auto frame_info = ABI_CalculateFrameSize(regs, frame_size);
+
+    // Free frame-space
+    if (frame_size != 0) {
+        code.ADD(SP, SP, frame_size);
+    }
+
+    // TODO(wunk): Pop pairs of registers at a time with LDP
+    std::size_t offset = 0;
+    for (std::size_t i = 0; i < 32; ++i) {
+        if (regs[i] && ABI_ALL_GPRS[i]) {
+            const XReg reg = IndexToXReg(i);
+            code.LDR(reg, SP, offset);
+            offset += 8;
+        }
+    }
+
+    offset = 0;
+    for (std::size_t i = 32; i < 64; ++i) {
+        if (regs[i] && ABI_ALL_FPRS[i]) {
+            const VReg reg = IndexToVReg(i);
+            code.LDR(reg.toQ(), SP, frame_info.fprs_offset + offset);
+            offset += 16;
+        }
+    }
+
+    // Free stack-space
+    if (frame_info.subtraction != 0) {
+        code.ADD(SP, SP, frame_info.subtraction);
+    }
+}
+
+} // namespace Common::A64
+
+#endif // CITRA_ARCH(arm64)

src/common/aarch64/oaknut_util.h

@@ -0,0 +1,43 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/arch.h"
+#if CITRA_ARCH(arm64)
+
+#include <type_traits>
+#include <oaknut/oaknut.hpp>
+#include "common/aarch64/oaknut_abi.h"
+
+namespace Common::A64 {
+
+// BL can only reach targets within +-128MiB(24 bits)
+inline bool IsWithin128M(uintptr_t ref, uintptr_t target) {
+    const u64 distance = target - (ref + 4);
+    return !(distance >= 0x800'0000ULL && distance <= ~0x800'0000ULL);
+}
+
+inline bool IsWithin128M(const oaknut::CodeGenerator& code, uintptr_t target) {
+    return IsWithin128M(code.ptr<uintptr_t>(), target);
+}
+
+template <typename T>
+inline void CallFarFunction(oaknut::CodeGenerator& code, const T f) {
+    static_assert(std::is_pointer_v<T>, "Argument must be a (function) pointer.");
+    const std::uintptr_t addr = reinterpret_cast<std::uintptr_t>(f);
+    if (IsWithin128M(code, addr)) {
+        code.BL(reinterpret_cast<const void*>(f));
+    } else {
+        // X16(IP0) and X17(IP1) is the standard veneer register
+        // LR is also available as an intermediate register
+        // https://developer.arm.com/documentation/102374/0101/Procedure-Call-Standard
+        code.MOVP2R(oaknut::util::X16, reinterpret_cast<const void*>(f));
+        code.BLR(oaknut::util::X16);
+    }
+}
+
+} // namespace Common::A64
+
+#endif // CITRA_ARCH(arm64)

src/common/dynamic_library/fdk-aac.cpp

@@ -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

src/common/dynamic_library/fdk-aac.h

@@ -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

src/common/file_util.cpp

@@ -534,7 +534,6 @@ u64 ScanDirectoryTree(const std::string& directory, FSTEntry& parent_entry,
 }
 
 void GetAllFilesFromNestedEntries(FSTEntry& directory, std::vector<FSTEntry>& output) {
-    std::vector<FSTEntry> files;
     for (auto& entry : directory.children) {
         if (entry.isDirectory) {
             GetAllFilesFromNestedEntries(entry, output);

src/common/file_watcher.cpp

@@ -0,0 +1,148 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <windows.h>
+#include <thread>
+
+#include "common/assert.h"
+#include "common/file_watcher.h"
+
+namespace Common {
+
+static FileAction Win32ActionToFileAction(DWORD action) {
+    switch (action) {
+    case FILE_ACTION_ADDED:
+        return FileAction::Added;
+    case FILE_ACTION_REMOVED:
+        return FileAction::Removed;
+    case FILE_ACTION_MODIFIED:
+        return FileAction::Modified;
+    case FILE_ACTION_RENAMED_OLD_NAME:
+        return FileAction::RenamedOldName;
+    case FILE_ACTION_RENAMED_NEW_NAME:
+        return FileAction::RenamedNewName;
+    default:
+        UNREACHABLE_MSG("Unknown action {}", action);
+        return FileAction::Invalid;
+    }
+}
+
+struct FileWatcher::Impl {
+    explicit Impl(const std::string& path, FileWatcher::Callback&& callback_)
+        : callback{callback_} {
+        // Create file handle for the directory we are watching.
+        dir_handle =
+            CreateFile(path.c_str(), FILE_LIST_DIRECTORY | GENERIC_READ,
+                        FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE, NULL, OPEN_EXISTING,
+                        FILE_FLAG_BACKUP_SEMANTICS | FILE_FLAG_OVERLAPPED, NULL);
+        ASSERT_MSG(dir_handle != INVALID_HANDLE_VALUE, "Unable to create watch file");
+
+        // Create an event that will terminate the thread when fired.
+        termination_event = CreateEvent(NULL, TRUE, FALSE, NULL);
+        ASSERT_MSG(termination_event != INVALID_HANDLE_VALUE, "Unable to create watch event");
+
+        // Create an event that will wake up the watcher thread on filesystem changes.
+        overlapped.hEvent = CreateEvent(NULL, FALSE, FALSE, NULL);
+        ASSERT_MSG(overlapped.hEvent != INVALID_HANDLE_VALUE, "Unable to create watch event");
+
+        // Create the watcher thread.
+        watch_thread = std::thread([this] { WatcherThread(); });
+    }
+
+    ~Impl() {
+        // Signal watcher thread to terminate.
+        SetEvent(termination_event);
+
+        // Wait for said termination.
+        if (watch_thread.joinable()) {
+            watch_thread.join();
+        }
+
+        // Close used handles.
+        CancelIo(dir_handle);
+        GetOverlappedResult(dir_handle, &overlapped, &num_bytes_read, TRUE);
+        CloseHandle(termination_event);
+        CloseHandle(overlapped.hEvent);
+    }
+
+    void WatcherThread() {
+        const std::array wait_handles{overlapped.hEvent, termination_event};
+        while (is_running) {
+            bool result =
+                ReadDirectoryChangesW(dir_handle, buffer.data(), buffer.size(), TRUE,
+                                      FILE_NOTIFY_CHANGE_FILE_NAME |
+                                      FILE_NOTIFY_CHANGE_DIR_NAME |
+                                      FILE_NOTIFY_CHANGE_LAST_WRITE, NULL, &overlapped, NULL);
+            ASSERT_MSG(result, "Unable to read directory changes: {}", GetLastErrorMsg());
+
+            // Sleep until we receive a file changed notification or a termination event.
+            switch (
+                WaitForMultipleObjects(wait_handles.size(), wait_handles.data(), FALSE, INFINITE)) {
+            case WAIT_OBJECT_0: {
+                // Retrieve asynchronously the data from ReadDirectoryChangesW.
+                result = GetOverlappedResult(dir_handle, &overlapped, &num_bytes_read, TRUE);
+                ASSERT_MSG(result, "Unable to retrieve overlapped result: {}", GetLastErrorMsg());
+
+                // Notify about file changes.
+                NotifyFileChanges();
+                break;
+            }
+            case WAIT_OBJECT_0 + 1:
+                is_running = false;
+                break;
+            case WAIT_FAILED:
+                UNREACHABLE_MSG("Failed waiting for file watcher events: {}", GetLastErrorMsg());
+                break;
+            }
+        }
+    }
+
+    void NotifyFileChanges() {
+        // If no data was read we have nothing to do.
+        if (num_bytes_read == 0) [[unlikely]] {
+            return;
+        }
+
+        u32 next_entry_offset{};
+        while (true) {
+            // Retrieve file notify information.
+            auto fni = reinterpret_cast<FILE_NOTIFY_INFORMATION*>(buffer.data() + next_entry_offset);
+
+            // Call the callback function informing about the change.
+            if (fni->Action != 0) {
+                std::string file_name(fni->FileNameLength / sizeof(WCHAR), ' ');
+                WideCharToMultiByte(CP_UTF8, 0, fni->FileName,
+                                    fni->FileNameLength / sizeof(WCHAR), file_name.data(), file_name.size(), NULL, NULL);
+                const FileAction action = Win32ActionToFileAction(fni->Action);
+                callback(file_name, action);
+            }
+
+            // If this was the last action, break.
+            if (fni->NextEntryOffset == 0) {
+                break;
+            }
+
+            // Move to next fni structure.
+            next_entry_offset += fni->NextEntryOffset;
+        }
+    }
+
+private:
+    static constexpr size_t DirectoryWatcherBufferSize = 4096;
+    FileWatcher::Callback callback;
+    HANDLE dir_handle{};
+    HANDLE termination_event{};
+    OVERLAPPED overlapped{};
+    std::array<u8, DirectoryWatcherBufferSize> buffer{};
+    std::atomic_bool is_running{true};
+    DWORD num_bytes_read{};
+    std::thread watch_thread;
+};
+
+FileWatcher::FileWatcher(const std::string& log_dir, Callback&& callback)
+    : impl{std::make_unique<Impl>(log_dir, std::move(callback))} {}
+
+FileWatcher::~FileWatcher() = default;
+
+} // namespace Common

src/common/file_watcher.h

@@ -0,0 +1,33 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <functional>
+#include <string>
+
+namespace Common {
+
+enum class FileAction : u8 {
+    Added,
+    Removed,
+    Modified,
+    RenamedOldName,
+    RenamedNewName,
+    Invalid = std::numeric_limits<u8>::max(),
+};
+
+class FileWatcher {
+    using Callback = std::function<void(const std::string&, FileAction)>;
+
+public:
+    explicit FileWatcher(const std::string& log_dir, Callback&& callback);
+    ~FileWatcher();
+
+private:
+    struct Impl;
+    std::unique_ptr<Impl> impl;
+};
+
+} // namespace Common

src/core/file_sys/romfs_reader.cpp

@@ -25,7 +25,7 @@ std::size_t DirectRomFSReader::ReadFile(std::size_t offset, std::size_t length,
             d.Seek(crypto_offset + offset);
             d.ProcessData(buffer, buffer, length);
         }
-        // LOG_INFO(Service_FS, "Cache SKIP: offset={}, length={}", offset, length);
+        LOG_TRACE(Service_FS, "RomFS Cache SKIP: offset={}, length={}", offset, length);
         return length;
     }
 
@@ -44,11 +44,11 @@ std::size_t DirectRomFSReader::ReadFile(std::size_t offset, std::size_t length,
                 d.Seek(crypto_offset + page);
                 d.ProcessData(cache_entry.second.data(), cache_entry.second.data(), read_size);
             }
-            // LOG_INFO(Service_FS, "Cache MISS: page={}, length={}, into={}", page, seg.second,
+            LOG_TRACE(Service_FS, "RomFS Cache MISS: page={}, length={}, into={}", page, seg.second,
-            //          (seg.first - page));
+                      (seg.first - page));
         } else {
-            // LOG_INFO(Service_FS, "Cache HIT: page={}, length={}, into={}", page, seg.second,
+            LOG_TRACE(Service_FS, "RomFS Cache HIT: page={}, length={}, into={}", page, seg.second,
-            //          (seg.first - page));
+                      (seg.first - page));
         }
         size_t copy_amount =
             (read_size > (seg.first - page))

src/core/frontend/emu_window.h

@@ -77,6 +77,11 @@ class GraphicsContext {
 public:
     virtual ~GraphicsContext();
 
+    /// Checks whether this context uses OpenGL ES.
+    virtual bool IsGLES() {
+        return false;
+    }
+
     /// Inform the driver to swap the front/back buffers and present the current image
     virtual void SwapBuffers(){};
 

src/core/hle/kernel/address_arbiter.cpp

@@ -6,7 +6,6 @@
 #include "common/archives.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
-#include "core/global.h"
 #include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/kernel.h"
@@ -21,7 +20,7 @@ void AddressArbiter::WaitThread(std::shared_ptr<Thread> thread, VAddr wait_addre
     waiting_threads.emplace_back(std::move(thread));
 }
 
-void AddressArbiter::ResumeAllThreads(VAddr address) {
+u64 AddressArbiter::ResumeAllThreads(VAddr address) {
     // Determine which threads are waiting on this address, those should be woken up.
     auto itr = std::stable_partition(waiting_threads.begin(), waiting_threads.end(),
                                      [address](const auto& thread) {
@@ -31,13 +30,15 @@ void AddressArbiter::ResumeAllThreads(VAddr address) {
                                      });
 
     // Wake up all the found threads
+    const u64 num_threads = std::distance(itr, waiting_threads.end());
     std::for_each(itr, waiting_threads.end(), [](auto& thread) { thread->ResumeFromWait(); });
 
     // Remove the woken up threads from the wait list.
     waiting_threads.erase(itr, waiting_threads.end());
+    return num_threads;
 }
 
-std::shared_ptr<Thread> AddressArbiter::ResumeHighestPriorityThread(VAddr address) {
+bool AddressArbiter::ResumeHighestPriorityThread(VAddr address) {
     // Determine which threads are waiting on this address, those should be considered for wakeup.
     auto matches_start = std::stable_partition(
         waiting_threads.begin(), waiting_threads.end(), [address](const auto& thread) {
@@ -54,14 +55,15 @@ std::shared_ptr<Thread> AddressArbiter::ResumeHighestPriorityThread(VAddr addres
                                     return lhs->current_priority < rhs->current_priority;
                                 });
 
-    if (itr == waiting_threads.end())
+    if (itr == waiting_threads.end()) {
-        return nullptr;
+        return false;
+    }
 
     auto thread = *itr;
     thread->ResumeFromWait();
-
     waiting_threads.erase(itr);
-    return thread;
+
+    return true;
 }
 
 AddressArbiter::AddressArbiter(KernelSystem& kernel)
@@ -107,17 +109,28 @@ ResultCode AddressArbiter::ArbitrateAddress(std::shared_ptr<Thread> thread, Arbi
     switch (type) {
 
     // Signal thread(s) waiting for arbitrate address...
-    case ArbitrationType::Signal:
+    case ArbitrationType::Signal: {
+        u64 num_threads{};
+
         // Negative value means resume all threads
         if (value < 0) {
-            ResumeAllThreads(address);
+            num_threads = ResumeAllThreads(address);
         } else {
             // Resume first N threads
-            for (int i = 0; i < value; i++)
+            for (s32 i = 0; i < value; i++) {
-                ResumeHighestPriorityThread(address);
+                num_threads += ResumeHighestPriorityThread(address);
+            }
+        }
+
+        // Prevents lag from low priority threads that spam svcArbitrateAddress and wake no threads
+        // The tick count is taken directly from official HOS kernel. The priority value is one less
+        // than official kernel as the affected FMV threads dont meet the priority threshold of 50.
+        // TODO: Revisit this when scheduler is rewritten and adjust if there isn't a problem there.
+        if (num_threads == 0 && thread->current_priority >= 49) {
+            kernel.current_cpu->GetTimer().AddTicks(1614u);
         }
         break;
-
+    }
     // Wait current thread (acquire the arbiter)...
     case ArbitrationType::WaitIfLessThan:
         if ((s32)kernel.memory.Read32(address) < value) {

src/core/hle/kernel/address_arbiter.h

@@ -65,11 +65,11 @@ private:
     void WaitThread(std::shared_ptr<Thread> thread, VAddr wait_address);
 
     /// Resume all threads found to be waiting on the address under this address arbiter
-    void ResumeAllThreads(VAddr address);
+    u64 ResumeAllThreads(VAddr address);
 
     /// Resume one thread found to be waiting on the address under this address arbiter and return
     /// the resumed thread.
-    std::shared_ptr<Thread> ResumeHighestPriorityThread(VAddr address);
+    bool ResumeHighestPriorityThread(VAddr address);
 
     /// Threads waiting for the address arbiter to be signaled.
     std::vector<std::shared_ptr<Thread>> waiting_threads;

src/tests/CMakeLists.txt

@@ -12,11 +12,10 @@ 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
-    video_core/shader/shader_jit_x64_compiler.cpp
+    video_core/shader/shader_jit_compiler.cpp
 )
 
 create_target_directory_groups(tests)

src/tests/audio_core/hle/adts_reader.cpp

@@ -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);
-    }
-}

src/tests/video_core/shader/shader_jit_compiler.cpp

@@ -1,9 +1,9 @@
-// Copyright 2017 Citra Emulator Project
+// Copyright 2023 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
 #include "common/arch.h"
-#if CITRA_ARCH(x86_64)
+#if CITRA_ARCH(x86_64) || CITRA_ARCH(arm64)
 
 #include <algorithm>
 #include <cmath>
@@ -14,7 +14,11 @@
 #include <fmt/format.h>
 #include <nihstro/inline_assembly.h>
 #include "video_core/shader/shader_interpreter.h"
+#if CITRA_ARCH(x86_64)
 #include "video_core/shader/shader_jit_x64_compiler.h"
+#elif CITRA_ARCH(arm64)
+#include "video_core/shader/shader_jit_a64_compiler.h"
+#endif
 
 using JitShader = Pica::Shader::JitShader;
 using ShaderInterpreter = Pica::Shader::InterpreterEngine;
@@ -31,6 +35,18 @@ static constexpr Common::Vec4f vec4_zero = Common::Vec4f::AssignToAll(0.0f);
 
 namespace Catch {
 template <>
+struct StringMaker<Common::Vec2f> {
+    static std::string convert(Common::Vec2f value) {
+        return fmt::format("({}, {})", value.x, value.y);
+    }
+};
+template <>
+struct StringMaker<Common::Vec3f> {
+    static std::string convert(Common::Vec3f value) {
+        return fmt::format("({}, {}, {})", value.r(), value.g(), value.b());
+    }
+};
+template <>
 struct StringMaker<Common::Vec4f> {
     static std::string convert(Common::Vec4f value) {
         return fmt::format("({}, {}, {}, {})", value.r(), value.g(), value.b(), value.a());
@@ -59,6 +75,11 @@ public:
         shader_jit.Compile(&shader_setup->program_code, &shader_setup->swizzle_data);
     }
 
+    explicit ShaderTest(std::unique_ptr<Pica::Shader::ShaderSetup> input_shader_setup)
+        : shader_setup(std::move(input_shader_setup)) {
+        shader_jit.Compile(&shader_setup->program_code, &shader_setup->swizzle_data);
+    }
+
     Common::Vec4f Run(std::span<const Common::Vec4f> inputs) {
         Pica::Shader::UnitState shader_unit;
         RunJit(shader_unit, inputs);
@@ -144,6 +165,41 @@ TEST_CASE("ADD", "[video_core][shader][shader_jit]") {
     REQUIRE(std::isinf(shader.Run({INFINITY, -1.0f}).x));
 }
 
+TEST_CASE("CALL", "[video_core][shader][shader_jit]") {
+    const auto sh_input = SourceRegister::MakeInput(0);
+    const auto sh_output = DestRegister::MakeOutput(0);
+
+    auto shader_setup = CompileShaderSetup({
+        {OpCode::Id::NOP}, // call foo
+        {OpCode::Id::END},
+        // .proc foo
+        {OpCode::Id::NOP}, // call ex2
+        {OpCode::Id::END},
+        // .proc ex2
+        {OpCode::Id::EX2, sh_output, sh_input},
+        {OpCode::Id::END},
+    });
+
+    // nihstro does not support the CALL* instructions, so the instruction-binary must be manually
+    // inserted here:
+    nihstro::Instruction CALL = {};
+    CALL.opcode = nihstro::OpCode(nihstro::OpCode::Id::CALL);
+
+    // call foo
+    CALL.flow_control.dest_offset = 2;
+    CALL.flow_control.num_instructions = 1;
+    shader_setup->program_code[0] = CALL.hex;
+
+    // call ex2
+    CALL.flow_control.dest_offset = 4;
+    CALL.flow_control.num_instructions = 1;
+    shader_setup->program_code[2] = CALL.hex;
+
+    auto shader = ShaderTest(std::move(shader_setup));
+
+    REQUIRE(shader.Run(0.f).x == Catch::Approx(1.f));
+}
+
 TEST_CASE("DP3", "[video_core][shader][shader_jit]") {
     const auto sh_input1 = SourceRegister::MakeInput(0);
     const auto sh_input2 = SourceRegister::MakeInput(1);
@@ -395,6 +451,39 @@ TEST_CASE("RSQ", "[video_core][shader][shader_jit]") {
     REQUIRE(shader.Run({0.0625f}).x == Catch::Approx(4.0f).margin(0.004f));
 }
 
+TEST_CASE("Uniform Read", "[video_core][shader][shader_jit]") {
+    const auto sh_input = SourceRegister::MakeInput(0);
+    const auto sh_c0 = SourceRegister::MakeFloat(0);
+    const auto sh_output = DestRegister::MakeOutput(0);
+
+    auto shader = ShaderTest({
+        // mova a0.x, sh_input.x
+        {OpCode::Id::MOVA, DestRegister{}, "x", sh_input, "x", SourceRegister{}, "",
+         nihstro::InlineAsm::RelativeAddress::A1},
+        // mov sh_output.xyzw, c0[a0.x].xyzw
+        {OpCode::Id::MOV, sh_output, "xyzw", sh_c0, "xyzw", SourceRegister{}, "",
+         nihstro::InlineAsm::RelativeAddress::A1},
+        {OpCode::Id::END},
+    });
+
+    // Prepare shader uniforms
+    std::array<Common::Vec4f, 96> f_uniforms = {};
+    for (u32 i = 0; i < 96; ++i) {
+        const float color = (i * 2.0f) / 255.0f;
+        const auto color_f24 = Pica::f24::FromFloat32(color);
+        shader.shader_setup->uniforms.f[i] = {color_f24, color_f24, color_f24, Pica::f24::One()};
+        f_uniforms[i] = {color, color, color, 1.0f};
+    }
+
+    for (u32 i = 0; i < 96; ++i) {
+        const float index = static_cast<float>(i);
+        // Add some fractional values to test proper float->integer truncation
+        const float fractional = (i % 17) / 17.0f;
+
+        REQUIRE(shader.Run(index + fractional) == f_uniforms[i]);
+    }
+}
+
 TEST_CASE("Address Register Offset", "[video_core][shader][shader_jit]") {
     const auto sh_input = SourceRegister::MakeInput(0);
     const auto sh_c40 = SourceRegister::MakeFloat(40);
@@ -445,23 +534,83 @@ TEST_CASE("Address Register Offset", "[video_core][shader][shader_jit]") {
     REQUIRE(shader.Run(-129.f) == f_uniforms[40]);
 }
 
-// TODO: Requires fix from https://github.com/neobrain/nihstro/issues/68
+TEST_CASE("Dest Mask", "[video_core][shader][shader_jit]") {
-// TEST_CASE("MAD", "[video_core][shader][shader_jit]") {
+    const auto sh_input = SourceRegister::MakeInput(0);
-//     const auto sh_input1 = SourceRegister::MakeInput(0);
+    const auto sh_output = DestRegister::MakeOutput(0);
-//     const auto sh_input2 = SourceRegister::MakeInput(1);
-//     const auto sh_input3 = SourceRegister::MakeInput(2);
-//     const auto sh_output = DestRegister::MakeOutput(0);
 
-//     auto shader = ShaderTest({
+    const auto shader = [&sh_input, &sh_output](const char* dest_mask) {
-//         {OpCode::Id::MAD, sh_output, sh_input1, sh_input2, sh_input3},
+        return std::unique_ptr<ShaderTest>(new ShaderTest{
-//         {OpCode::Id::END},
+            {OpCode::Id::MOV, sh_output, dest_mask, sh_input, "xyzw", SourceRegister{}, ""},
-//     });
+            {OpCode::Id::END},
+        });
+    };
 
-//     REQUIRE(shader.Run({vec4_inf, vec4_zero, vec4_zero}).x == 0.0f);
+    const Common::Vec4f iota_vec = {1.0f, 2.0f, 3.0f, 4.0f};
-//     REQUIRE(std::isnan(shader.Run({vec4_nan, vec4_zero, vec4_zero}).x));
 
-//     REQUIRE(shader.Run({vec4_one, vec4_one, vec4_one}).x == 2.0f);
+    REQUIRE(shader("x")->Run({iota_vec}).x == iota_vec.x);
-// }
+    REQUIRE(shader("y")->Run({iota_vec}).y == iota_vec.y);
+    REQUIRE(shader("z")->Run({iota_vec}).z == iota_vec.z);
+    REQUIRE(shader("w")->Run({iota_vec}).w == iota_vec.w);
+    REQUIRE(shader("xy")->Run({iota_vec}).xy() == iota_vec.xy());
+    REQUIRE(shader("xz")->Run({iota_vec}).xz() == iota_vec.xz());
+    REQUIRE(shader("xw")->Run({iota_vec}).xw() == iota_vec.xw());
+    REQUIRE(shader("yz")->Run({iota_vec}).yz() == iota_vec.yz());
+    REQUIRE(shader("yw")->Run({iota_vec}).yw() == iota_vec.yw());
+    REQUIRE(shader("zw")->Run({iota_vec}).zw() == iota_vec.zw());
+    REQUIRE(shader("xyz")->Run({iota_vec}).xyz() == iota_vec.xyz());
+    REQUIRE(shader("xyw")->Run({iota_vec}).xyw() == iota_vec.xyw());
+    REQUIRE(shader("xzw")->Run({iota_vec}).xzw() == iota_vec.xzw());
+    REQUIRE(shader("yzw")->Run({iota_vec}).yzw() == iota_vec.yzw());
+    REQUIRE(shader("xyzw")->Run({iota_vec}) == iota_vec);
+}
+
+TEST_CASE("MAD", "[video_core][shader][shader_jit]") {
+    const auto sh_input1 = SourceRegister::MakeInput(0);
+    const auto sh_input2 = SourceRegister::MakeInput(1);
+    const auto sh_input3 = SourceRegister::MakeInput(2);
+    const auto sh_output = DestRegister::MakeOutput(0);
+
+    auto shader_setup = CompileShaderSetup({
+        // TODO: Requires fix from https://github.com/neobrain/nihstro/issues/68
+        // {OpCode::Id::MAD, sh_output, sh_input1, sh_input2, sh_input3},
+        {OpCode::Id::NOP},
+        {OpCode::Id::END},
+    });
+
+    // nihstro does not support the MAD* instructions, so the instruction-binary must be manually
+    // inserted here:
+    nihstro::Instruction MAD = {};
+    MAD.opcode = nihstro::OpCode::Id::MAD;
+    MAD.mad.operand_desc_id = 0;
+    MAD.mad.src1 = sh_input1;
+    MAD.mad.src2 = sh_input2;
+    MAD.mad.src3 = sh_input3;
+    MAD.mad.dest = sh_output;
+    shader_setup->program_code[0] = MAD.hex;
+
+    nihstro::SwizzlePattern swizzle = {};
+    swizzle.dest_mask = 0b1111;
+    swizzle.SetSelectorSrc1(0, SwizzlePattern::Selector::x);
+    swizzle.SetSelectorSrc1(1, SwizzlePattern::Selector::y);
+    swizzle.SetSelectorSrc1(2, SwizzlePattern::Selector::z);
+    swizzle.SetSelectorSrc1(3, SwizzlePattern::Selector::w);
+    swizzle.SetSelectorSrc2(0, SwizzlePattern::Selector::x);
+    swizzle.SetSelectorSrc2(1, SwizzlePattern::Selector::y);
+    swizzle.SetSelectorSrc2(2, SwizzlePattern::Selector::z);
+    swizzle.SetSelectorSrc2(3, SwizzlePattern::Selector::w);
+    swizzle.SetSelectorSrc3(0, SwizzlePattern::Selector::x);
+    swizzle.SetSelectorSrc3(1, SwizzlePattern::Selector::y);
+    swizzle.SetSelectorSrc3(2, SwizzlePattern::Selector::z);
+    swizzle.SetSelectorSrc3(3, SwizzlePattern::Selector::w);
+    shader_setup->swizzle_data[0] = swizzle.hex;
+
+    auto shader = ShaderTest(std::move(shader_setup));
+
+    REQUIRE(shader.Run({vec4_zero, vec4_zero, vec4_zero}) == vec4_zero);
+    REQUIRE(shader.Run({vec4_one, vec4_one, vec4_one}) == (vec4_one * 2.0f));
+    REQUIRE(shader.Run({vec4_inf, vec4_zero, vec4_zero}) == vec4_zero);
+    REQUIRE(shader.Run({vec4_nan, vec4_zero, vec4_zero}) == vec4_nan);
+}
 
 TEST_CASE("Nested Loop", "[video_core][shader][shader_jit]") {
     const auto sh_input = SourceRegister::MakeInput(0);
@@ -518,4 +667,42 @@ TEST_CASE("Nested Loop", "[video_core][shader][shader_jit]") {
     }
 }
 
-#endif // CITRA_ARCH(x86_64)
+TEST_CASE("Source Swizzle", "[video_core][shader][shader_jit]") {
+    const auto sh_input = SourceRegister::MakeInput(0);
+    const auto sh_output = DestRegister::MakeOutput(0);
+
+    const auto shader = [&sh_input, &sh_output](const char* swizzle) {
+        return std::unique_ptr<ShaderTest>(new ShaderTest{
+            {OpCode::Id::MOV, sh_output, "xyzw", sh_input, swizzle, SourceRegister{}, ""},
+            {OpCode::Id::END},
+        });
+    };
+
+    const Common::Vec4f iota_vec = {1.0f, 2.0f, 3.0f, 4.0f};
+
+    REQUIRE(shader("x")->Run({iota_vec}).x == iota_vec.x);
+    REQUIRE(shader("y")->Run({iota_vec}).x == iota_vec.y);
+    REQUIRE(shader("z")->Run({iota_vec}).x == iota_vec.z);
+    REQUIRE(shader("w")->Run({iota_vec}).x == iota_vec.w);
+    REQUIRE(shader("xy")->Run({iota_vec}).xy() == iota_vec.xy());
+    REQUIRE(shader("xz")->Run({iota_vec}).xy() == iota_vec.xz());
+    REQUIRE(shader("xw")->Run({iota_vec}).xy() == iota_vec.xw());
+    REQUIRE(shader("yz")->Run({iota_vec}).xy() == iota_vec.yz());
+    REQUIRE(shader("yw")->Run({iota_vec}).xy() == iota_vec.yw());
+    REQUIRE(shader("zw")->Run({iota_vec}).xy() == iota_vec.zw());
+    REQUIRE(shader("yy")->Run({iota_vec}).xy() == iota_vec.yy());
+    REQUIRE(shader("wx")->Run({iota_vec}).xy() == iota_vec.wx());
+    REQUIRE(shader("xyz")->Run({iota_vec}).xyz() == iota_vec.xyz());
+    REQUIRE(shader("xyw")->Run({iota_vec}).xyz() == iota_vec.xyw());
+    REQUIRE(shader("xzw")->Run({iota_vec}).xyz() == iota_vec.xzw());
+    REQUIRE(shader("yzw")->Run({iota_vec}).xyz() == iota_vec.yzw());
+    REQUIRE(shader("yyy")->Run({iota_vec}).xyz() == iota_vec.yyy());
+    REQUIRE(shader("yxw")->Run({iota_vec}).xyz() == iota_vec.yxw());
+    REQUIRE(shader("xyzw")->Run({iota_vec}) == iota_vec);
+    REQUIRE(shader("wzxy")->Run({iota_vec}) ==
+            Common::Vec4f(iota_vec.w, iota_vec.z, iota_vec.x, iota_vec.y));
+    REQUIRE(shader("yyyy")->Run({iota_vec}) ==
+            Common::Vec4f(iota_vec.y, iota_vec.y, iota_vec.y, iota_vec.y));
+}
+
+#endif // CITRA_ARCH(x86_64) || CITRA_ARCH(arm64)

src/video_core/CMakeLists.txt

@@ -135,20 +135,29 @@ add_library(video_core STATIC
     renderer_vulkan/vk_texture_runtime.cpp
     renderer_vulkan/vk_texture_runtime.h
     shader/debug_data.h
+    shader/generator/glsl_fs_shader_gen.cpp
+    shader/generator/glsl_fs_shader_gen.h
     shader/generator/glsl_shader_decompiler.cpp
     shader/generator/glsl_shader_decompiler.h
     shader/generator/glsl_shader_gen.cpp
     shader/generator/glsl_shader_gen.h
+    shader/generator/pica_fs_config.cpp
+    shader/generator/pica_fs_config.h
+    shader/generator/profile.h
     shader/generator/shader_gen.cpp
     shader/generator/shader_gen.h
     shader/generator/shader_uniforms.cpp
     shader/generator/shader_uniforms.h
-    shader/generator/spv_shader_gen.cpp
+    shader/generator/spv_fs_shader_gen.cpp
-    shader/generator/spv_shader_gen.h
+    shader/generator/spv_fs_shader_gen.h
     shader/shader.cpp
     shader/shader.h
     shader/shader_interpreter.cpp
     shader/shader_interpreter.h
+    shader/shader_jit_a64.cpp
+    shader/shader_jit_a64_compiler.cpp
+    shader/shader_jit_a64.h
+    shader/shader_jit_a64_compiler.h
     shader/shader_jit_x64.cpp
     shader/shader_jit_x64_compiler.cpp
     shader/shader_jit_x64.h
@@ -177,6 +186,10 @@ if ("x86_64" IN_LIST ARCHITECTURE)
     target_link_libraries(video_core PUBLIC xbyak)
 endif()
 
+if ("arm64" IN_LIST ARCHITECTURE)
+    target_link_libraries(video_core PUBLIC oaknut)
+endif()
+
 if (CITRA_USE_PRECOMPILED_HEADERS)
     target_precompile_headers(video_core PRIVATE precompiled_headers.h)
 endif()

src/video_core/custom_textures/custom_tex_manager.cpp

@@ -1,9 +1,10 @@
 // Copyright 2023 Citra Emulator Project
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
-
+#pragma clang optimize off
 #include <json.hpp>
 #include "common/file_util.h"
+#include "common/file_watcher.h"
 #include "common/memory_detect.h"
 #include "common/microprofile.h"
 #include "common/settings.h"
@@ -16,6 +17,7 @@
 #include "video_core/custom_textures/custom_tex_manager.h"
 #include "video_core/rasterizer_cache/surface_params.h"
 #include "video_core/rasterizer_cache/utils.h"
+#include "video_core/renderer_base.h"
 
 namespace VideoCore {
 
@@ -63,17 +65,17 @@ void CustomTexManager::TickFrame() {
         return;
     }
     std::size_t num_uploads = 0;
-    for (auto it = async_uploads.begin(); it != async_uploads.end();) {
+    for (auto it = async_actions.begin(); it != async_actions.end();) {
         if (num_uploads >= MAX_UPLOADS_PER_TICK) {
             return;
         }
         switch (it->material->state) {
         case DecodeState::Decoded:
-            it->func();
+            it->func(it->material);
             num_uploads++;
             [[fallthrough]];
         case DecodeState::Failed:
-            it = async_uploads.erase(it);
+            it = async_actions.erase(it);
             continue;
         default:
             it++;
@@ -102,7 +104,7 @@ void CustomTexManager::FindCustomTextures() {
         if (file.isDirectory) {
             continue;
         }
-        custom_textures.push_back(std::make_unique<CustomTexture>(image_interface));
+        custom_textures.emplace_back(std::make_unique<CustomTexture>(image_interface));
         CustomTexture* const texture{custom_textures.back().get()};
         if (!ParseFilename(file, texture)) {
             continue;
@@ -292,16 +294,17 @@ Material* CustomTexManager::GetMaterial(u64 data_hash) {
     return it->second.get();
 }
 
-bool CustomTexManager::Decode(Material* material, std::function<bool()>&& upload) {
+bool CustomTexManager::Decode(Material* material, AsyncFunc&& upload) {
     if (!async_custom_loading) {
         material->LoadFromDisk(flip_png_files);
-        return upload();
+        return upload(material);
     }
     if (material->IsUnloaded()) {
         material->state = DecodeState::Pending;
         workers->QueueWork([material, this] { material->LoadFromDisk(flip_png_files); });
     }
-    async_uploads.push_back({
+    std::scoped_lock lock{async_actions_mutex};
+    async_actions.push_back({
         .material = material,
         .func = std::move(upload),
     });
@@ -374,6 +377,14 @@ std::vector<FileUtil::FSTEntry> CustomTexManager::GetTextures(u64 title_id) {
         FileUtil::CreateFullPath(load_path);
     }
 
+    const auto callback = [this](const std::string& file, Common::FileAction action) {
+        OnFileAction(file, action);
+    };
+
+    // Create a file watcher to monitor any changes to the textures directory for hot-reloading.
+    file_watcher = std::make_unique<Common::FileWatcher>(load_path, callback);
+
+    // Retrieve all texture files.
     FileUtil::FSTEntry texture_dir;
     std::vector<FileUtil::FSTEntry> textures;
     FileUtil::ScanDirectoryTree(load_path, texture_dir, 64);
@@ -386,4 +397,24 @@ void CustomTexManager::CreateWorkers() {
     workers = std::make_unique<Common::ThreadWorker>(num_workers, "Custom textures");
 }
 
+void CustomTexManager::OnFileAction(const std::string& file, Common::FileAction action) {
+    const auto invalidate = [this](const Material* material) -> bool {
+        for (const SurfaceParams* params : material->loaded_to) {
+            system.Renderer().Rasterizer()->InvalidateRegion(params->addr, params->size);
+        }
+        return true;
+    };
+
+    const std::string filename{FileUtil::GetFilename(file)};
+    const auto& hashes = path_to_hash_map[filename];
+    std::scoped_lock lock{async_actions_mutex};
+
+    for (const Hash hash : hashes) {
+        async_actions.push_back({
+            .material = material_map[hash].get(),
+            .func = std::move(invalidate),
+        });
+    }
+}
+
 } // namespace VideoCore

src/video_core/custom_textures/custom_tex_manager.h

@@ -12,6 +12,11 @@
 #include "video_core/custom_textures/material.h"
 #include "video_core/rasterizer_interface.h"
 
+namespace Common {
+class FileWatcher;
+enum class FileAction : u8;
+} // namespace Common
+
 namespace Core {
 class System;
 }
@@ -24,12 +29,10 @@ namespace VideoCore {
 
 class SurfaceParams;
 
-struct AsyncUpload {
-    const Material* material;
-    std::function<bool()> func;
-};
-
 class CustomTexManager {
+    using Hash = u64;
+    using AsyncFunc = std::function<bool(const Material*)>;
+
 public:
     explicit CustomTexManager(Core::System& system);
     ~CustomTexManager();
@@ -57,7 +60,7 @@ public:
     Material* GetMaterial(u64 data_hash);
 
     /// Decodes the textures in material to a consumable format and uploads it.
-    bool Decode(Material* material, std::function<bool()>&& upload);
+    bool Decode(Material* material, AsyncFunc&& func);
 
     /// True when mipmap uploads should be skipped (legacy packs only)
     bool SkipMipmaps() const noexcept {
@@ -79,15 +82,25 @@ private:
     /// Creates the thread workers.
     void CreateWorkers();
 
+    /// Callback for when a custom texture file is modified.
+    void OnFileAction(const std::string& file, Common::FileAction action);
+
 private:
+    struct AsyncAction {
+        const Material* material;
+        AsyncFunc func;
+    };
+
     Core::System& system;
     Frontend::ImageInterface& image_interface;
-    std::unordered_set<u64> dumped_textures;
+    std::unordered_set<Hash> dumped_textures;
-    std::unordered_map<u64, std::unique_ptr<Material>> material_map;
+    std::unordered_map<Hash, std::unique_ptr<Material>> material_map;
-    std::unordered_map<std::string, std::vector<u64>> path_to_hash_map;
+    std::unordered_map<std::string, std::vector<Hash>> path_to_hash_map;
     std::vector<std::unique_ptr<CustomTexture>> custom_textures;
-    std::list<AsyncUpload> async_uploads;
+    std::mutex async_actions_mutex;
+    std::list<AsyncAction> async_actions;
     std::unique_ptr<Common::ThreadWorker> workers;
+    std::unique_ptr<Common::FileWatcher> file_watcher;
     bool textures_loaded{false};
     bool async_custom_loading{true};
     bool skip_mipmap{false};

src/video_core/custom_textures/material.h

@@ -18,6 +18,8 @@ class ImageInterface;
 
 namespace VideoCore {
 
+class SurfaceParams;
+
 enum class MapType : u32 {
     Color = 0,
     Normal = 1,
@@ -72,6 +74,7 @@ struct Material {
     u64 hash;
     CustomPixelFormat format;
     std::array<CustomTexture*, MAX_MAPS> textures;
+    mutable std::vector<SurfaceParams*> loaded_to;
     std::atomic<DecodeState> state{};
 
     void LoadFromDisk(bool flip_png) noexcept;

src/video_core/rasterizer_accelerated.h

@@ -7,6 +7,7 @@
 #include "common/vector_math.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/regs_texturing.h"
+#include "video_core/shader/generator/pica_fs_config.h"
 #include "video_core/shader/generator/shader_uniforms.h"
 
 namespace Memory {
@@ -153,6 +154,7 @@ protected:
     Pica::Regs& regs;
 
     std::vector<HardwareVertex> vertex_batch;
+    Pica::Shader::UserConfig user_config{};
     bool shader_dirty = true;
 
     VSUniformBlockData vs_uniform_block_data{};
@@ -166,4 +168,5 @@ protected:
     std::array<Common::Vec4f, 256> proctex_lut_data{};
     std::array<Common::Vec4f, 256> proctex_diff_lut_data{};
 };
+
 } // namespace VideoCore

src/video_core/rasterizer_cache/rasterizer_cache.h

@@ -653,7 +653,6 @@ FramebufferHelper<T> RasterizerCache<T>::GetFramebufferSurfaces(bool using_color
         static_cast<u32>(std::clamp(viewport_rect.bottom, 0, framebuffer_height)),
     };
 
-    // get color and depth surfaces
     SurfaceParams color_params;
     color_params.is_tiled = true;
     color_params.res_scale = resolution_scale_factor;
@@ -672,14 +671,6 @@ FramebufferHelper<T> RasterizerCache<T>::GetFramebufferSurfaces(bool using_color
     auto color_vp_interval = color_params.GetSubRectInterval(viewport_clamped);
     auto depth_vp_interval = depth_params.GetSubRectInterval(viewport_clamped);
 
-    // Make sure that framebuffers don't overlap if both color and depth are being used
-    if (using_color_fb && using_depth_fb &&
-        boost::icl::length(color_vp_interval & depth_vp_interval)) {
-        LOG_CRITICAL(HW_GPU, "Color and depth framebuffer memory regions overlap; "
-                             "overlapping framebuffers not supported!");
-        using_depth_fb = false;
-    }
-
     Common::Rectangle<u32> color_rect{};
     SurfaceId color_id{};
     u32 color_level{};
@@ -713,11 +704,13 @@ FramebufferHelper<T> RasterizerCache<T>::GetFramebufferSurfaces(bool using_color
 
     if (color_id) {
         color_level = color_surface->LevelOf(color_params.addr);
+        color_surface->flags |= SurfaceFlagBits::RenderTarget;
         ValidateSurface(color_id, boost::icl::first(color_vp_interval),
                         boost::icl::length(color_vp_interval));
     }
     if (depth_id) {
         depth_level = depth_surface->LevelOf(depth_params.addr);
+        depth_surface->flags |= SurfaceFlagBits::RenderTarget;
         ValidateSurface(depth_id, boost::icl::first(depth_vp_interval),
                         boost::icl::length(depth_vp_interval));
     }
@@ -991,7 +984,9 @@ void RasterizerCache<T>::UploadSurface(Surface& surface, SurfaceInterval interva
     DecodeTexture(load_info, load_info.addr, load_info.end, upload_data, staging.mapped,
                   runtime.NeedsConversion(surface.pixel_format));
 
-    if (dump_textures && False(surface.flags & SurfaceFlagBits::Custom)) {
+    const bool should_dump = False(surface.flags & SurfaceFlagBits::Custom) &&
+                             False(surface.flags & SurfaceFlagBits::RenderTarget);
+    if (dump_textures && should_dump) {
         const u64 hash = ComputeHash(load_info, upload_data);
         const u32 level = surface.LevelOf(load_info.addr);
         custom_tex_manager.DumpTexture(load_info, level, upload_data, hash);
@@ -1048,7 +1043,7 @@ bool RasterizerCache<T>::UploadCustomSurface(SurfaceId surface_id, SurfaceInterv
 
     surface.flags |= SurfaceFlagBits::Custom;
 
-    const auto upload = [this, level, surface_id, material]() -> bool {
+    const auto upload = [this, level, surface_id](const Material* material) -> bool {
         ASSERT_MSG(True(slot_surfaces[surface_id].flags & SurfaceFlagBits::Custom),
                    "Surface is not suitable for custom upload, aborting!");
         if (!slot_surfaces[surface_id].IsCustom()) {

src/video_core/rasterizer_cache/surface_base.cpp

@@ -11,7 +11,18 @@ namespace VideoCore {
 
 SurfaceBase::SurfaceBase(const SurfaceParams& params) : SurfaceParams{params} {}
 
-SurfaceBase::~SurfaceBase() = default;
+SurfaceBase::SurfaceBase(const SurfaceParams& params, const Material* mat)
+    : SurfaceParams{params}, material{mat} {
+    custom_format = material->format;
+    material->loaded_to.push_back(this);
+}
+
+SurfaceBase::~SurfaceBase() {
+    if (!material) {
+        return;
+    }
+    std::erase_if(material->loaded_to, [this](SurfaceParams* params) { return params == this; });
+}
 
 bool SurfaceBase::CanFill(const SurfaceParams& dest_surface, SurfaceInterval fill_interval) const {
     if (type == SurfaceType::Fill && IsRegionValid(fill_interval) &&

src/video_core/rasterizer_cache/surface_base.h

@@ -15,17 +15,19 @@ using SurfaceRegions = boost::icl::interval_set<PAddr, std::less, SurfaceInterva
 struct Material;
 
 enum class SurfaceFlagBits : u32 {
-    Registered = 1 << 0, ///< Surface is registed in the rasterizer cache.
+    Registered = 1 << 0,   ///< Surface is registed in the rasterizer cache.
-    Picked = 1 << 1,     ///< Surface has been picked when searching for a match.
+    Picked = 1 << 1,       ///< Surface has been picked when searching for a match.
-    Tracked = 1 << 2,    ///< Surface is part of a texture cube and should be tracked.
+    Tracked = 1 << 2,      ///< Surface is part of a texture cube and should be tracked.
-    Custom = 1 << 3,     ///< Surface texture has been replaced with a custom texture.
+    Custom = 1 << 3,       ///< Surface texture has been replaced with a custom texture.
-    ShadowMap = 1 << 4,  ///< Surface is used during shadow rendering.
+    ShadowMap = 1 << 4,    ///< Surface is used during shadow rendering.
+    RenderTarget = 1 << 5, ///< Surface was a render target.
 };
 DECLARE_ENUM_FLAG_OPERATORS(SurfaceFlagBits);
 
 class SurfaceBase : public SurfaceParams {
 public:
-    SurfaceBase(const SurfaceParams& params);
+    explicit SurfaceBase(const SurfaceParams& params);
+    explicit SurfaceBase(const SurfaceParams& params, const Material* mat);
     ~SurfaceBase();
 
     /// Returns true when this surface can be used to fill the fill_interval of dest_surface

src/video_core/renderer_opengl/gl_driver.cpp

@@ -179,8 +179,8 @@ void Driver::CheckExtensionSupport() {
     arb_texture_compression_bptc = GLAD_GL_ARB_texture_compression_bptc;
     clip_cull_distance = !is_gles || GLAD_GL_EXT_clip_cull_distance;
     ext_texture_compression_s3tc = GLAD_GL_EXT_texture_compression_s3tc;
-    shader_framebuffer_fetch =
+    ext_shader_framebuffer_fetch = GLAD_GL_EXT_shader_framebuffer_fetch;
-        GLAD_GL_EXT_shader_framebuffer_fetch || GLAD_GL_ARM_shader_framebuffer_fetch;
+    arm_shader_framebuffer_fetch = GLAD_GL_ARM_shader_framebuffer_fetch;
     blend_minmax_factor = GLAD_GL_AMD_blend_minmax_factor || GLAD_GL_NV_blend_minmax_factor;
     is_suitable = GLAD_GL_VERSION_4_3 || GLAD_GL_ES_VERSION_3_1;
 }

src/video_core/renderer_opengl/gl_driver.h

@@ -107,7 +107,15 @@ public:
 
     /// Returns true if the implementation supports (EXT/ARM)_shader_framebuffer_fetch
     bool HasShaderFramebufferFetch() const {
-        return shader_framebuffer_fetch;
+        return ext_shader_framebuffer_fetch || arm_shader_framebuffer_fetch;
+    }
+
+    bool HasExtFramebufferFetch() const {
+        return ext_shader_framebuffer_fetch;
+    }
+
+    bool HasArmShaderFramebufferFetch() const {
+        return arm_shader_framebuffer_fetch;
     }
 
     /// Returns true if the implementation supports (NV/AMD)_blend_minmax_factor
@@ -136,7 +144,8 @@ private:
     bool clip_cull_distance{};
     bool ext_texture_compression_s3tc{};
     bool arb_texture_compression_bptc{};
-    bool shader_framebuffer_fetch{};
+    bool arm_shader_framebuffer_fetch{};
+    bool ext_shader_framebuffer_fetch{};
     bool blend_minmax_factor{};
 
     std::string_view gl_version{};

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -426,7 +426,7 @@ bool RasterizerOpenGL::Draw(bool accelerate, bool is_indexed) {
 
     // Sync and bind the shader
     if (shader_dirty) {
-        shader_manager.UseFragmentShader(regs, use_custom_normal);
+        shader_manager.UseFragmentShader(regs, user_config);
         shader_dirty = false;
     }
 
@@ -479,7 +479,7 @@ void RasterizerOpenGL::SyncTextureUnits(const Framebuffer* framebuffer) {
 
     // Reset transient draw state
     state.color_buffer.texture_2d = 0;
-    use_custom_normal = false;
+    user_config = {};
 
     const auto pica_textures = regs.texturing.GetTextures();
     for (u32 texture_index = 0; texture_index < pica_textures.size(); ++texture_index) {
@@ -577,20 +577,15 @@ void RasterizerOpenGL::BindMaterial(u32 texture_index, Surface& surface) {
         return;
     }
 
-    const auto bind_texture = [&](const TextureUnits::TextureUnit& unit, GLuint texture,
-                                  GLuint sampler) {
-        glActiveTexture(unit.Enum());
-        glBindTexture(GL_TEXTURE_2D, texture);
-        glBindSampler(unit.id, sampler);
-    };
-
     const GLuint sampler = state.texture_units[texture_index].sampler;
     if (surface.HasNormalMap()) {
         if (regs.lighting.disable) {
             LOG_WARNING(Render_OpenGL, "Custom normal map used but scene has no light enabled");
         }
-        bind_texture(TextureUnits::TextureNormalMap, surface.Handle(2), sampler);
+        glActiveTexture(TextureUnits::TextureNormalMap.Enum());
-        use_custom_normal = true;
+        glBindTexture(GL_TEXTURE_2D, surface.Handle(2));
+        glBindSampler(TextureUnits::TextureNormalMap.id, sampler);
+        user_config.use_custom_normal.Assign(1);
     }
 }
 

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -155,7 +155,6 @@ private:
     OGLTexture texture_buffer_lut_lf;
     OGLTexture texture_buffer_lut_rg;
     OGLTexture texture_buffer_lut_rgba;
-    bool use_custom_normal{};
     bool emulate_minmax_blend{};
 };
 

src/video_core/renderer_opengl/gl_shader_manager.cpp

@@ -14,10 +14,12 @@
 #include "video_core/renderer_opengl/gl_shader_disk_cache.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/gl_state.h"
-#include "video_core/shader/generator/shader_uniforms.h"
+#include "video_core/shader/generator/glsl_fs_shader_gen.h"
+#include "video_core/shader/generator/profile.h"
 #include "video_core/video_core.h"
 
 using namespace Pica::Shader::Generator;
+using Pica::Shader::FSConfig;
 
 namespace OpenGL {
 
@@ -146,17 +148,20 @@ private:
     OGLShaderStage program;
 };
 
-template <typename KeyConfigType, std::string (*CodeGenerator)(const KeyConfigType&, bool),
+template <typename KeyConfigType, auto CodeGenerator, GLenum ShaderType>
-          GLenum ShaderType>
 class ShaderCache {
 public:
-    explicit ShaderCache(bool separable) : separable(separable) {}
+    explicit ShaderCache(bool separable_) : separable{separable_} {}
-    std::tuple<GLuint, std::optional<std::string>> Get(const KeyConfigType& config) {
+    ~ShaderCache() = default;
+
+    template <typename... Args>
+    std::tuple<GLuint, std::optional<std::string>> Get(const KeyConfigType& config,
+                                                       Args&&... args) {
         auto [iter, new_shader] = shaders.emplace(config, OGLShaderStage{separable});
         OGLShaderStage& cached_shader = iter->second;
         std::optional<std::string> result{};
         if (new_shader) {
-            result = CodeGenerator(config, separable);
+            result = CodeGenerator(config, args...);
             cached_shader.Create(result->c_str(), ShaderType);
         }
         return {cached_shader.GetHandle(), std::move(result)};
@@ -243,8 +248,7 @@ using ProgrammableVertexShaders =
 using FixedGeometryShaders =
     ShaderCache<PicaFixedGSConfig, &GLSL::GenerateFixedGeometryShader, GL_GEOMETRY_SHADER>;
 
-using FragmentShaders =
+using FragmentShaders = ShaderCache<FSConfig, &GLSL::GenerateFragmentShader, GL_FRAGMENT_SHADER>;
-    ShaderCache<PicaFSConfig, &GLSL::GenerateFragmentShader, GL_FRAGMENT_SHADER>;
 
 class ShaderProgramManager::Impl {
 public:
@@ -252,8 +256,24 @@ public:
         : separable(separable), programmable_vertex_shaders(separable),
           trivial_vertex_shader(driver, separable), fixed_geometry_shaders(separable),
           fragment_shaders(separable), disk_cache(separable) {
-        if (separable)
+        if (separable) {
             pipeline.Create();
+        }
+        profile = Pica::Shader::Profile{
+            .has_separable_shaders = separable,
+            .has_clip_planes = driver.HasClipCullDistance(),
+            .has_geometry_shader = true,
+            .has_custom_border_color = true,
+            .has_fragment_shader_interlock = false,
+            .has_blend_minmax_factor = driver.HasBlendMinMaxFactor(),
+            .has_minus_one_to_one_range = true,
+            .has_logic_op = !driver.IsOpenGLES(),
+            .has_gl_ext_framebuffer_fetch = driver.HasExtFramebufferFetch(),
+            .has_gl_arm_framebuffer_fetch = driver.HasArmShaderFramebufferFetch(),
+            .has_gl_nv_fragment_shader_interlock = driver.GetVendor() == Vendor::Nvidia,
+            .has_gl_intel_fragment_shader_interlock = driver.GetVendor() == Vendor::Intel,
+            .is_vulkan = false,
+        };
     }
 
     struct ShaderTuple {
@@ -283,7 +303,7 @@ public:
                   "ShaderTuple layout changed!");
 
     bool separable;
-
+    Pica::Shader::Profile profile{};
     ShaderTuple current;
 
     ProgrammableVertexShaders programmable_vertex_shaders;
@@ -336,7 +356,7 @@ void ShaderProgramManager::UseTrivialVertexShader() {
 
 void ShaderProgramManager::UseFixedGeometryShader(const Pica::Regs& regs) {
     PicaFixedGSConfig gs_config(regs, driver.HasClipCullDistance());
-    auto [handle, _] = impl->fixed_geometry_shaders.Get(gs_config);
+    auto [handle, _] = impl->fixed_geometry_shaders.Get(gs_config, impl->separable);
     impl->current.gs = handle;
     impl->current.gs_hash = gs_config.Hash();
 }
@@ -346,12 +366,12 @@ void ShaderProgramManager::UseTrivialGeometryShader() {
     impl->current.gs_hash = 0;
 }
 
-void ShaderProgramManager::UseFragmentShader(const Pica::Regs& regs, bool use_normal) {
+void ShaderProgramManager::UseFragmentShader(const Pica::Regs& regs,
-    PicaFSConfig config(regs, false, driver.IsOpenGLES(), false, driver.HasBlendMinMaxFactor(),
+                                             const Pica::Shader::UserConfig& user) {
-                        use_normal);
+    const FSConfig fs_config{regs, user, impl->profile};
-    auto [handle, result] = impl->fragment_shaders.Get(config);
+    auto [handle, result] = impl->fragment_shaders.Get(fs_config, impl->profile);
     impl->current.fs = handle;
-    impl->current.fs_hash = config.Hash();
+    impl->current.fs_hash = fs_config.Hash();
     // Save FS to the disk cache if its a new shader
     if (result) {
         auto& disk_cache = impl->disk_cache;
@@ -470,8 +490,8 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
                     impl->programmable_vertex_shaders.Inject(conf, decomp->second.code,
                                                              std::move(shader));
                 } else if (raw.GetProgramType() == ProgramType::FS) {
-                    PicaFSConfig conf(raw.GetRawShaderConfig(), false, driver.IsOpenGLES(), false,
+                    // TODO: Support UserConfig in disk shader cache
-                                      driver.HasBlendMinMaxFactor());
+                    const FSConfig conf(raw.GetRawShaderConfig(), {}, impl->profile);
                     std::scoped_lock lock(mutex);
                     impl->fragment_shaders.Inject(conf, std::move(shader));
                 } else {
@@ -581,14 +601,14 @@ void ShaderProgramManager::LoadDiskCache(const std::atomic_bool& stop_loading,
                 std::scoped_lock lock(mutex);
                 impl->programmable_vertex_shaders.Inject(conf, code, std::move(stage));
             } else if (raw.GetProgramType() == ProgramType::FS) {
-                PicaFSConfig conf(raw.GetRawShaderConfig(), false, driver.IsOpenGLES(), false,
+                // TODO: Support UserConfig in disk shader cache
-                                  driver.HasBlendMinMaxFactor());
+                const FSConfig fs_config{raw.GetRawShaderConfig(), {}, impl->profile};
-                code = GLSL::GenerateFragmentShader(conf, impl->separable);
+                code = GLSL::GenerateFragmentShader(fs_config, impl->profile);
                 OGLShaderStage stage{impl->separable};
                 stage.Create(code.c_str(), GL_FRAGMENT_SHADER);
                 handle = stage.GetHandle();
                 std::scoped_lock lock(mutex);
-                impl->fragment_shaders.Inject(conf, std::move(stage));
+                impl->fragment_shaders.Inject(fs_config, std::move(stage));
             } else {
                 // Unsupported shader type got stored somehow so nuke the cache
                 LOG_ERROR(Frontend, "failed to load raw ProgramType {}", raw.GetProgramType());

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -17,7 +17,8 @@ struct Regs;
 
 namespace Pica::Shader {
 struct ShaderSetup;
-}
+union UserConfig;
+} // namespace Pica::Shader
 
 namespace OpenGL {
 
@@ -47,7 +48,7 @@ public:
 
     void UseTrivialGeometryShader();
 
-    void UseFragmentShader(const Pica::Regs& config, bool use_normal);
+    void UseFragmentShader(const Pica::Regs& config, const Pica::Shader::UserConfig& user);
 
     void ApplyTo(OpenGLState& state);
 

src/video_core/renderer_opengl/gl_texture_runtime.cpp

@@ -333,7 +333,7 @@ Surface::Surface(TextureRuntime& runtime_, const VideoCore::SurfaceParams& param
 
 Surface::Surface(TextureRuntime& runtime, const VideoCore::SurfaceBase& surface,
                  const VideoCore::Material* mat)
-    : SurfaceBase{surface}, tuple{runtime.GetFormatTuple(mat->format)} {
+    : SurfaceBase{surface, mat}, tuple{runtime.GetFormatTuple(mat->format)} {
     if (mat && !driver->IsCustomFormatSupported(mat->format)) {
         return;
     }
@@ -342,9 +342,6 @@ Surface::Surface(TextureRuntime& runtime, const VideoCore::SurfaceBase& surface,
     const GLenum target =
         texture_type == VideoCore::TextureType::CubeMap ? GL_TEXTURE_CUBE_MAP : GL_TEXTURE_2D;
 
-    custom_format = mat->format;
-    material = mat;
-
     textures[0] = MakeHandle(target, mat->width, mat->height, levels, tuple, DebugName(false));
     if (res_scale != 1) {
         textures[1] = MakeHandle(target, mat->width, mat->height, levels, DEFAULT_TUPLE,

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -3,10 +3,10 @@
 // Refer to the license.txt file included.
 
 #include "common/assert.h"
+#include "common/file_watcher.h"
 #include "common/logging/log.h"
 #include "common/microprofile.h"
 #include "common/settings.h"
-#include "common/texture.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
 #include "core/hw/gpu.h"
@@ -19,7 +19,6 @@
 #include "video_core/host_shaders/vulkan_present_frag_spv.h"
 #include "video_core/host_shaders/vulkan_present_interlaced_frag_spv.h"
 #include "video_core/host_shaders/vulkan_present_vert_spv.h"
-#include "vulkan/vulkan_format_traits.hpp"
 
 #include <vk_mem_alloc.h>
 

src/video_core/renderer_vulkan/vk_graphics_pipeline.h

@@ -4,9 +4,9 @@
 
 #include "common/thread_worker.h"
 #include "video_core/rasterizer_cache/pixel_format.h"
+#include "video_core/regs_pipeline.h"
+#include "video_core/regs_rasterizer.h"
 #include "video_core/renderer_vulkan/vk_common.h"
-#include "video_core/shader/generator/glsl_shader_gen.h"
-#include "video_core/shader/generator/spv_shader_gen.h"
 
 namespace Common {
 
@@ -89,6 +89,9 @@ struct DynamicState {
     u8 stencil_compare_mask;
     u8 stencil_write_mask;
 
+    Common::Rectangle<u32> scissor;
+    Common::Rectangle<s32> viewport;
+
     bool operator==(const DynamicState& other) const noexcept {
         return std::memcmp(this, &other, sizeof(DynamicState)) == 0;
     }

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -15,8 +15,12 @@
 #include "video_core/renderer_vulkan/vk_renderpass_cache.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_shader_util.h"
+#include "video_core/shader/generator/glsl_fs_shader_gen.h"
+#include "video_core/shader/generator/glsl_shader_gen.h"
+#include "video_core/shader/generator/spv_fs_shader_gen.h"
 
 using namespace Pica::Shader::Generator;
+using Pica::Shader::FSConfig;
 
 MICROPROFILE_DEFINE(Vulkan_Bind, "Vulkan", "Pipeline Bind", MP_RGB(192, 32, 32));
 
@@ -86,6 +90,17 @@ PipelineCache::PipelineCache(const Instance& instance_, Scheduler& scheduler_,
       trivial_vertex_shader{
           instance, vk::ShaderStageFlagBits::eVertex,
           GLSL::GenerateTrivialVertexShader(instance.IsShaderClipDistanceSupported(), true)} {
+    profile = Pica::Shader::Profile{
+        .has_separable_shaders = true,
+        .has_clip_planes = instance.IsShaderClipDistanceSupported(),
+        .has_geometry_shader = instance.UseGeometryShaders(),
+        .has_custom_border_color = instance.IsCustomBorderColorSupported(),
+        .has_fragment_shader_interlock = instance.IsFragmentShaderInterlockSupported(),
+        .has_blend_minmax_factor = false,
+        .has_minus_one_to_one_range = false,
+        .has_logic_op = !instance.NeedsLogicOpEmulation(),
+        .is_vulkan = true,
+    };
     BuildLayout();
 }
 
@@ -206,6 +221,32 @@ bool PipelineCache::BindPipeline(const PipelineInfo& info, bool wait_built) {
                       current_depth_stencil = current_info.depth_stencil,
                       rasterization = info.rasterization,
                       depth_stencil = info.depth_stencil](vk::CommandBuffer cmdbuf) {
+        if (dynamic.viewport != current_dynamic.viewport || is_dirty) {
+            const vk::Viewport vk_viewport = {
+                .x = static_cast<f32>(dynamic.viewport.left),
+                .y = static_cast<f32>(dynamic.viewport.top),
+                .width = static_cast<f32>(dynamic.viewport.GetWidth()),
+                .height = static_cast<f32>(dynamic.viewport.GetHeight()),
+                .minDepth = 0.f,
+                .maxDepth = 1.f,
+            };
+            cmdbuf.setViewport(0, vk_viewport);
+        }
+
+        if (dynamic.scissor != current_dynamic.scissor || is_dirty) {
+            const vk::Rect2D scissor = {
+                .offset{
+                    .x = static_cast<s32>(dynamic.scissor.left),
+                    .y = static_cast<s32>(dynamic.scissor.bottom),
+                },
+                .extent{
+                    .width = dynamic.scissor.GetWidth(),
+                    .height = dynamic.scissor.GetHeight(),
+                },
+            };
+            cmdbuf.setScissor(0, scissor);
+        }
+
         if (dynamic.stencil_compare_mask != current_dynamic.stencil_compare_mask || is_dirty) {
             cmdbuf.setStencilCompareMask(vk::StencilFaceFlagBits::eFrontAndBack,
                                          dynamic.stencil_compare_mask);
@@ -377,35 +418,30 @@ void PipelineCache::UseTrivialGeometryShader() {
     shader_hashes[ProgramType::GS] = 0;
 }
 
-void PipelineCache::UseFragmentShader(const Pica::Regs& regs) {
+void PipelineCache::UseFragmentShader(const Pica::Regs& regs,
-    const PicaFSConfig config{regs, instance.IsFragmentShaderInterlockSupported(),
+                                      const Pica::Shader::UserConfig& user) {
-                              instance.NeedsLogicOpEmulation(),
+    const FSConfig fs_config{regs, user, profile};
-                              !instance.IsCustomBorderColorSupported(), false};
+    const auto [it, new_shader] = fragment_shaders.try_emplace(fs_config, instance);
-
-    const auto [it, new_shader] = fragment_shaders.try_emplace(config, instance);
     auto& shader = it->second;
 
     if (new_shader) {
         const bool use_spirv = Settings::values.spirv_shader_gen.GetValue();
-        const auto texture0_type = config.state.texture0_type.Value();
+        if (use_spirv && !fs_config.UsesShadowPipeline()) {
-        const bool is_shadow = texture0_type == Pica::TexturingRegs::TextureConfig::Shadow2D ||
+            const std::vector code = SPIRV::GenerateFragmentShader(fs_config);
-                               texture0_type == Pica::TexturingRegs::TextureConfig::ShadowCube ||
-                               config.state.shadow_rendering.Value();
-        if (use_spirv && !is_shadow) {
-            const std::vector code = SPIRV::GenerateFragmentShader(config);
             shader.module = CompileSPV(code, instance.GetDevice());
             shader.MarkDone();
         } else {
-            workers.QueueWork([config, device = instance.GetDevice(), &shader]() {
+            workers.QueueWork([fs_config, this, &shader]() {
-                const std::string code = GLSL::GenerateFragmentShader(config, true);
+                const std::string code = GLSL::GenerateFragmentShader(fs_config, profile);
-                shader.module = Compile(code, vk::ShaderStageFlagBits::eFragment, device);
+                shader.module =
+                    Compile(code, vk::ShaderStageFlagBits::eFragment, instance.GetDevice());
                 shader.MarkDone();
             });
         }
     }
 
     current_shaders[ProgramType::FS] = &shader;
-    shader_hashes[ProgramType::FS] = config.Hash();
+    shader_hashes[ProgramType::FS] = fs_config.Hash();
 }
 
 void PipelineCache::BindTexture(u32 binding, vk::ImageView image_view, vk::Sampler sampler) {

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -9,13 +9,18 @@
 
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
 #include "video_core/renderer_vulkan/vk_graphics_pipeline.h"
-#include "video_core/shader/generator/glsl_shader_gen.h"
+#include "video_core/shader/generator/pica_fs_config.h"
-#include "video_core/shader/generator/spv_shader_gen.h"
+#include "video_core/shader/generator/profile.h"
+#include "video_core/shader/generator/shader_gen.h"
 
 namespace Pica {
 struct Regs;
 }
 
+namespace Pica::Shader {
+struct ShaderSetup;
+}
+
 namespace Vulkan {
 
 class Instance;
@@ -62,7 +67,7 @@ public:
     void UseTrivialGeometryShader();
 
     /// Binds a fragment shader generated from PICA state
-    void UseFragmentShader(const Pica::Regs& regs);
+    void UseFragmentShader(const Pica::Regs& regs, const Pica::Shader::UserConfig& user);
 
     /// Binds a texture to the specified binding
     void BindTexture(u32 binding, vk::ImageView image_view, vk::Sampler sampler);
@@ -98,6 +103,7 @@ private:
     RenderpassCache& renderpass_cache;
     DescriptorPool& pool;
 
+    Pica::Shader::Profile profile{};
     vk::UniquePipelineCache pipeline_cache;
     vk::UniquePipelineLayout pipeline_layout;
     std::size_t num_worker_threads;
@@ -118,7 +124,7 @@ private:
     std::unordered_map<Pica::Shader::Generator::PicaVSConfig, Shader*> programmable_vertex_map;
     std::unordered_map<std::string, Shader> programmable_vertex_cache;
     std::unordered_map<Pica::Shader::Generator::PicaFixedGSConfig, Shader> fixed_geometry_shaders;
-    std::unordered_map<Pica::Shader::Generator::PicaFSConfig, Shader> fragment_shaders;
+    std::unordered_map<Pica::Shader::FSConfig, Shader> fragment_shaders;
     Shader trivial_vertex_shader;
 };
 

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -497,7 +497,7 @@ bool RasterizerVulkan::Draw(bool accelerate, bool is_indexed) {
 
     // Sync and bind the shader
     if (shader_dirty) {
-        pipeline_cache.UseFragmentShader(regs);
+        pipeline_cache.UseFragmentShader(regs, user_config);
         shader_dirty = false;
     }
 
@@ -512,30 +512,13 @@ bool RasterizerVulkan::Draw(bool accelerate, bool is_indexed) {
 
     // Configure viewport and scissor
     const auto viewport = fb_helper.Viewport();
-    scheduler.Record([viewport, draw_rect](vk::CommandBuffer cmdbuf) {
+    pipeline_info.dynamic.viewport = Common::Rectangle<s32>{
-        const vk::Viewport vk_viewport = {
+        viewport.x,
-            .x = static_cast<f32>(viewport.x),
+        viewport.y,
-            .y = static_cast<f32>(viewport.y),
+        viewport.x + viewport.width,
-            .width = static_cast<f32>(viewport.width),
+        viewport.y + viewport.height,
-            .height = static_cast<f32>(viewport.height),
+    };
-            .minDepth = 0.f,
+    pipeline_info.dynamic.scissor = draw_rect;
-            .maxDepth = 1.f,
-        };
-
-        const vk::Rect2D scissor = {
-            .offset{
-                .x = static_cast<s32>(draw_rect.left),
-                .y = static_cast<s32>(draw_rect.bottom),
-            },
-            .extent{
-                .width = draw_rect.GetWidth(),
-                .height = draw_rect.GetHeight(),
-            },
-        };
-
-        cmdbuf.setViewport(0, vk_viewport);
-        cmdbuf.setScissor(0, scissor);
-    });
 
     // Draw the vertex batch
     bool succeeded = true;

src/video_core/renderer_vulkan/vk_stream_buffer.cpp

@@ -44,11 +44,12 @@ vk::MemoryPropertyFlags MakePropertyFlags(BufferType type) {
 }
 
 /// Find a memory type with the passed requirements
-std::optional<u32> FindMemoryType(const vk::PhysicalDeviceMemoryProperties& properties,
+std::optional<u32> FindMemoryType(
-                                  vk::MemoryPropertyFlags wanted) {
+    const vk::PhysicalDeviceMemoryProperties& properties, vk::MemoryPropertyFlags wanted,
+    vk::MemoryPropertyFlags excluded = vk::MemoryPropertyFlagBits::eProtected) {
     for (u32 i = 0; i < properties.memoryTypeCount; ++i) {
         const auto flags = properties.memoryTypes[i].propertyFlags;
-        if ((flags & wanted) == wanted) {
+        if (((flags & wanted) == wanted) && (!(flags & excluded))) {
             return i;
         }
     }
@@ -166,25 +167,46 @@ void StreamBuffer::CreateBuffers(u64 prefered_size) {
         static_cast<bool>(mem_type.propertyFlags & vk::MemoryPropertyFlagBits::eHostCoherent);
 
     // Substract from the preferred heap size some bytes to avoid getting out of memory.
-    const VkDeviceSize heap_size = memory_properties.memoryHeaps[preferred_heap].size;
+    const vk::DeviceSize heap_size = memory_properties.memoryHeaps[preferred_heap].size;
     // As per DXVK's example, using `heap_size / 2`
-    const VkDeviceSize allocable_size = heap_size / 2;
+    const vk::DeviceSize allocable_size = heap_size / 2;
     buffer = device.createBuffer({
         .size = std::min(prefered_size, allocable_size),
         .usage = usage,
     });
 
-    const auto requirements = device.getBufferMemoryRequirements(buffer);
+    const auto requirements_chain =
-    stream_buffer_size = static_cast<u64>(requirements.size);
+        device
+            .getBufferMemoryRequirements2<vk::MemoryRequirements2, vk::MemoryDedicatedRequirements>(
+                {.buffer = buffer});
+
+    const auto& requirements = requirements_chain.get<vk::MemoryRequirements2>();
+    const auto& dedicated_requirements = requirements_chain.get<vk::MemoryDedicatedRequirements>();
+
+    stream_buffer_size = static_cast<u64>(requirements.memoryRequirements.size);
 
     LOG_INFO(Render_Vulkan, "Creating {} buffer with size {} KB with flags {}",
              BufferTypeName(type), stream_buffer_size / 1024,
              vk::to_string(mem_type.propertyFlags));
 
-    memory = device.allocateMemory({
+    if (dedicated_requirements.prefersDedicatedAllocation) {
-        .allocationSize = requirements.size,
+        vk::StructureChain<vk::MemoryAllocateInfo, vk::MemoryDedicatedAllocateInfo> alloc_chain =
-        .memoryTypeIndex = preferred_type,
+            {};
-    });
+
+        auto& alloc_info = alloc_chain.get<vk::MemoryAllocateInfo>();
+        alloc_info.allocationSize = requirements.memoryRequirements.size;
+        alloc_info.memoryTypeIndex = preferred_type;
+
+        auto& dedicated_alloc_info = alloc_chain.get<vk::MemoryDedicatedAllocateInfo>();
+        dedicated_alloc_info.buffer = buffer;
+
+        memory = device.allocateMemory(alloc_chain.get());
+    } else {
+        memory = device.allocateMemory({
+            .allocationSize = requirements.memoryRequirements.size,
+            .memoryTypeIndex = preferred_type,
+        });
+    }
 
     device.bindBufferMemory(buffer, memory, 0);
     mapped = reinterpret_cast<u8*>(device.mapMemory(memory, 0, VK_WHOLE_SIZE));

src/video_core/renderer_vulkan/vk_texture_runtime.cpp

@@ -750,7 +750,7 @@ Surface::Surface(TextureRuntime& runtime_, const VideoCore::SurfaceParams& param
 
 Surface::Surface(TextureRuntime& runtime_, const VideoCore::SurfaceBase& surface,
                  const VideoCore::Material* mat)
-    : SurfaceBase{surface}, runtime{&runtime_}, instance{&runtime_.GetInstance()},
+    : SurfaceBase{surface, mat}, runtime{&runtime_}, instance{&runtime_.GetInstance()},
       scheduler{&runtime_.GetScheduler()}, traits{instance->GetTraits(mat->format)} {
     if (!traits.transfer_support) {
         return;
@@ -791,9 +791,6 @@ Surface::Surface(TextureRuntime& runtime_, const VideoCore::SurfaceBase& surface
                                vk::PipelineStageFlagBits::eTopOfPipe,
                                vk::DependencyFlagBits::eByRegion, {}, {}, barriers);
     });
-
-    custom_format = mat->format;
-    material = mat;
 }
 
 Surface::~Surface() {

src/video_core/shader/generator/glsl_fs_shader_gen.h

@@ -0,0 +1,100 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "video_core/shader/generator/pica_fs_config.h"
+
+namespace Pica::Shader::Generator::GLSL {
+
+class FragmentModule {
+public:
+    explicit FragmentModule(const FSConfig& config, const Profile& profile);
+    ~FragmentModule();
+
+    /// Emits GLSL source corresponding to the provided pica fragment configuration
+    std::string Generate();
+
+private:
+    /// Undos the host perspective transformation and applies the PICA one
+    void WriteDepth();
+
+    /// Emits code to emulate the scissor rectangle
+    void WriteScissor();
+
+    /// Writes the code to emulate fragment lighting
+    void WriteLighting();
+
+    /// Writes the code to emulate fog
+    void WriteFog();
+
+    /// Writes the code to emulate gas rendering
+    void WriteGas();
+
+    /// Writes the code to emulate shadow-map rendering
+    void WriteShadow();
+
+    /// Writes the code to emulate logic ops in the fragment shader
+    void WriteLogicOp();
+
+    /// Writes the code to emulate PICA min/max blending factors
+    void WriteBlending();
+
+    /// Writes the specified TEV stage source component(s)
+    void AppendSource(Pica::TexturingRegs::TevStageConfig::Source source, u32 tev_index);
+
+    /// Writes the color components to use for the specified TEV stage color modifier
+    void AppendColorModifier(Pica::TexturingRegs::TevStageConfig::ColorModifier modifier,
+                             Pica::TexturingRegs::TevStageConfig::Source source, u32 tev_index);
+
+    /// Writes the alpha component to use for the specified TEV stage alpha modifier
+    void AppendAlphaModifier(Pica::TexturingRegs::TevStageConfig::AlphaModifier modifier,
+                             Pica::TexturingRegs::TevStageConfig::Source source, u32 tev_index);
+
+    /// Writes the combiner function for the color components for the specified TEV stage operation
+    void AppendColorCombiner(Pica::TexturingRegs::TevStageConfig::Operation operation);
+
+    /// Writes the combiner function for the alpha component for the specified TEV stage operation
+    void AppendAlphaCombiner(Pica::TexturingRegs::TevStageConfig::Operation operation);
+
+    /// Writes the if-statement condition used to evaluate alpha testing
+    void WriteAlphaTestCondition(Pica::FramebufferRegs::CompareFunc func);
+
+    /// Writes the code to emulate the specified TEV stage
+    void WriteTevStage(u32 index);
+
+    void AppendProcTexShiftOffset(std::string_view v, Pica::TexturingRegs::ProcTexShift mode,
+                                  Pica::TexturingRegs::ProcTexClamp clamp_mode);
+
+    void AppendProcTexClamp(std::string_view var, Pica::TexturingRegs::ProcTexClamp mode);
+
+    void AppendProcTexCombineAndMap(Pica::TexturingRegs::ProcTexCombiner combiner,
+                                    std::string_view offset);
+
+    void DefineExtensions();
+    void DefineInterface();
+    void DefineBindings();
+    void DefineHelpers();
+    void DefineLightingHelpers();
+    void DefineShadowHelpers();
+    void DefineProcTexSampler();
+    void DefineTexUnitSampler(u32 i);
+
+private:
+    const FSConfig& config;
+    const Profile& profile;
+    std::string out;
+    bool use_blend_fallback{};
+    bool use_fragment_shader_interlock{};
+};
+
+/**
+ * Generates the GLSL fragment shader program source code for the current Pica state
+ * @param config ShaderCacheKey object generated for the current Pica state, used for the shader
+ *               configuration (NOTE: Use state in this struct only, not the Pica registers!)
+ * @returns String of the shader source code
+ */
+std::string GenerateFragmentShader(const FSConfig& config, const Profile& profile);
+
+} // namespace Pica::Shader::Generator::GLSL

src/video_core/shader/generator/glsl_shader_gen.h

@@ -46,12 +46,4 @@ std::string GenerateVertexShader(const Pica::Shader::ShaderSetup& setup, const P
  */
 std::string GenerateFixedGeometryShader(const PicaFixedGSConfig& config, bool separable_shader);
 
-/**
- * Generates the GLSL fragment shader program source code for the current Pica state
- * @param config ShaderCacheKey object generated for the current Pica state, used for the shader
- *               configuration (NOTE: Use state in this struct only, not the Pica registers!)
- * @returns String of the shader source code
- */
-std::string GenerateFragmentShader(const PicaFSConfig& config, bool separable_shader);
-
 } // namespace Pica::Shader::Generator::GLSL

src/video_core/shader/generator/pica_fs_config.cpp

@@ -0,0 +1,193 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "video_core/shader/generator/pica_fs_config.h"
+
+namespace Pica::Shader {
+
+FramebufferConfig::FramebufferConfig(const Pica::Regs& regs, const Profile& profile) {
+    const auto& output_merger = regs.framebuffer.output_merger;
+    scissor_test_mode.Assign(regs.rasterizer.scissor_test.mode);
+    depthmap_enable.Assign(regs.rasterizer.depthmap_enable);
+    shadow_rendering.Assign(regs.framebuffer.IsShadowRendering());
+    alpha_test_func.Assign(output_merger.alpha_test.enable
+                               ? output_merger.alpha_test.func.Value()
+                               : Pica::FramebufferRegs::CompareFunc::Always);
+
+    // Emulate logic op in the shader if needed and not supported.
+    logic_op.Assign(Pica::FramebufferRegs::LogicOp::Copy);
+    if (!profile.has_logic_op && !regs.framebuffer.output_merger.alphablend_enable) {
+        logic_op.Assign(regs.framebuffer.output_merger.logic_op);
+    }
+
+    const auto alpha_eq = output_merger.alpha_blending.blend_equation_a.Value();
+    const auto rgb_eq = output_merger.alpha_blending.blend_equation_rgb.Value();
+    if (!profile.has_blend_minmax_factor && output_merger.alphablend_enable) {
+        if (rgb_eq == Pica::FramebufferRegs::BlendEquation::Max ||
+            rgb_eq == Pica::FramebufferRegs::BlendEquation::Min) {
+            rgb_blend.eq = rgb_eq;
+            rgb_blend.src_factor = output_merger.alpha_blending.factor_source_rgb;
+            rgb_blend.dst_factor = output_merger.alpha_blending.factor_dest_rgb;
+        }
+        if (alpha_eq == Pica::FramebufferRegs::BlendEquation::Max ||
+            alpha_eq == Pica::FramebufferRegs::BlendEquation::Min) {
+            alpha_blend.eq = alpha_eq;
+            alpha_blend.src_factor = output_merger.alpha_blending.factor_source_a;
+            alpha_blend.dst_factor = output_merger.alpha_blending.factor_dest_a;
+        }
+    }
+}
+
+TextureConfig::TextureConfig(const Pica::TexturingRegs& regs, const Profile& profile) {
+    texture0_type.Assign(regs.texture0.type);
+    texture2_use_coord1.Assign(regs.main_config.texture2_use_coord1 != 0);
+    combiner_buffer_input.Assign(regs.tev_combiner_buffer_input.update_mask_rgb.Value() |
+                                 regs.tev_combiner_buffer_input.update_mask_a.Value() << 4);
+    fog_mode.Assign(regs.fog_mode);
+    fog_flip.Assign(regs.fog_flip != 0);
+    shadow_texture_orthographic.Assign(regs.shadow.orthographic != 0);
+
+    // Emulate custom border color if needed and not supported.
+    const auto pica_textures = regs.GetTextures();
+    for (u32 tex_index = 0; tex_index < 3; tex_index++) {
+        const auto& config = pica_textures[tex_index].config;
+        texture_border_color[tex_index].enable_s.Assign(
+            !profile.has_custom_border_color &&
+            config.wrap_s == Pica::TexturingRegs::TextureConfig::WrapMode::ClampToBorder);
+        texture_border_color[tex_index].enable_t.Assign(
+            !profile.has_custom_border_color &&
+            config.wrap_t == Pica::TexturingRegs::TextureConfig::WrapMode::ClampToBorder);
+    }
+
+    const auto& stages = regs.GetTevStages();
+    for (std::size_t i = 0; i < tev_stages.size(); i++) {
+        const auto& tev_stage = stages[i];
+        tev_stages[i].sources_raw = tev_stage.sources_raw;
+        tev_stages[i].modifiers_raw = tev_stage.modifiers_raw;
+        tev_stages[i].ops_raw = tev_stage.ops_raw;
+        tev_stages[i].scales_raw = tev_stage.scales_raw;
+        if (tev_stage.color_op == Pica::TexturingRegs::TevStageConfig::Operation::Dot3_RGBA) {
+            tev_stages[i].sources_raw &= 0xFFF;
+            tev_stages[i].modifiers_raw &= 0xFFF;
+            tev_stages[i].ops_raw &= 0xF;
+        }
+    }
+}
+
+LightConfig::LightConfig(const Pica::LightingRegs& regs) {
+    if (regs.disable) {
+        return;
+    }
+
+    enable.Assign(1);
+    src_num.Assign(regs.max_light_index + 1);
+    config.Assign(regs.config0.config);
+    enable_primary_alpha.Assign(regs.config0.enable_primary_alpha);
+    enable_secondary_alpha.Assign(regs.config0.enable_secondary_alpha);
+    bump_mode.Assign(regs.config0.bump_mode);
+    bump_selector.Assign(regs.config0.bump_selector);
+    bump_renorm.Assign(regs.config0.disable_bump_renorm == 0);
+    clamp_highlights.Assign(regs.config0.clamp_highlights != 0);
+
+    enable_shadow.Assign(regs.config0.enable_shadow != 0);
+    if (enable_shadow) {
+        shadow_primary.Assign(regs.config0.shadow_primary != 0);
+        shadow_secondary.Assign(regs.config0.shadow_secondary != 0);
+        shadow_invert.Assign(regs.config0.shadow_invert != 0);
+        shadow_alpha.Assign(regs.config0.shadow_alpha != 0);
+        shadow_selector.Assign(regs.config0.shadow_selector);
+    }
+
+    for (u32 light_index = 0; light_index <= regs.max_light_index; ++light_index) {
+        const u32 num = regs.light_enable.GetNum(light_index);
+        const auto& light = regs.light[num];
+        lights[light_index].num.Assign(num);
+        lights[light_index].directional.Assign(light.config.directional != 0);
+        lights[light_index].two_sided_diffuse.Assign(light.config.two_sided_diffuse != 0);
+        lights[light_index].geometric_factor_0.Assign(light.config.geometric_factor_0 != 0);
+        lights[light_index].geometric_factor_1.Assign(light.config.geometric_factor_1 != 0);
+        lights[light_index].dist_atten_enable.Assign(!regs.IsDistAttenDisabled(num));
+        lights[light_index].spot_atten_enable.Assign(!regs.IsSpotAttenDisabled(num));
+        lights[light_index].shadow_enable.Assign(!regs.IsShadowDisabled(num));
+    }
+
+    lut_d0.enable.Assign(regs.config1.disable_lut_d0 == 0);
+    if (lut_d0.enable) {
+        lut_d0.abs_input.Assign(regs.abs_lut_input.disable_d0 == 0);
+        lut_d0.type.Assign(regs.lut_input.d0.Value());
+        lut_d0.scale = regs.lut_scale.GetScale(regs.lut_scale.d0);
+    }
+
+    lut_d1.enable.Assign(regs.config1.disable_lut_d1 == 0);
+    if (lut_d1.enable) {
+        lut_d1.abs_input.Assign(regs.abs_lut_input.disable_d1 == 0);
+        lut_d1.type.Assign(regs.lut_input.d1.Value());
+        lut_d1.scale = regs.lut_scale.GetScale(regs.lut_scale.d1);
+    }
+
+    // This is a dummy field due to lack of the corresponding register
+    lut_sp.enable.Assign(1);
+    lut_sp.abs_input.Assign(regs.abs_lut_input.disable_sp == 0);
+    lut_sp.type.Assign(regs.lut_input.sp.Value());
+    lut_sp.scale = regs.lut_scale.GetScale(regs.lut_scale.sp);
+
+    lut_fr.enable.Assign(regs.config1.disable_lut_fr == 0);
+    if (lut_fr.enable) {
+        lut_fr.abs_input.Assign(regs.abs_lut_input.disable_fr == 0);
+        lut_fr.type.Assign(regs.lut_input.fr.Value());
+        lut_fr.scale = regs.lut_scale.GetScale(regs.lut_scale.fr);
+    }
+
+    lut_rr.enable.Assign(regs.config1.disable_lut_rr == 0);
+    if (lut_rr.enable) {
+        lut_rr.abs_input.Assign(regs.abs_lut_input.disable_rr == 0);
+        lut_rr.type.Assign(regs.lut_input.rr.Value());
+        lut_rr.scale = regs.lut_scale.GetScale(regs.lut_scale.rr);
+    }
+
+    lut_rg.enable.Assign(regs.config1.disable_lut_rg == 0);
+    if (lut_rg.enable) {
+        lut_rg.abs_input.Assign(regs.abs_lut_input.disable_rg == 0);
+        lut_rg.type.Assign(regs.lut_input.rg.Value());
+        lut_rg.scale = regs.lut_scale.GetScale(regs.lut_scale.rg);
+    }
+
+    lut_rb.enable.Assign(regs.config1.disable_lut_rb == 0);
+    if (lut_rb.enable) {
+        lut_rb.abs_input.Assign(regs.abs_lut_input.disable_rb == 0);
+        lut_rb.type.Assign(regs.lut_input.rb.Value());
+        lut_rb.scale = regs.lut_scale.GetScale(regs.lut_scale.rb);
+    }
+}
+
+ProcTexConfig::ProcTexConfig(const Pica::TexturingRegs& regs) {
+    if (!regs.main_config.texture3_enable) {
+        return;
+    }
+
+    enable.Assign(1);
+    coord.Assign(regs.main_config.texture3_coordinates);
+    u_clamp.Assign(regs.proctex.u_clamp);
+    v_clamp.Assign(regs.proctex.v_clamp);
+    color_combiner.Assign(regs.proctex.color_combiner);
+    alpha_combiner.Assign(regs.proctex.alpha_combiner);
+    separate_alpha.Assign(regs.proctex.separate_alpha);
+    noise_enable.Assign(regs.proctex.noise_enable);
+    u_shift.Assign(regs.proctex.u_shift);
+    v_shift.Assign(regs.proctex.v_shift);
+    lut_width = regs.proctex_lut.width;
+    lut_offset0 = regs.proctex_lut_offset.level0;
+    lut_offset1 = regs.proctex_lut_offset.level1;
+    lut_offset2 = regs.proctex_lut_offset.level2;
+    lut_offset3 = regs.proctex_lut_offset.level3;
+    lod_min = regs.proctex_lut.lod_min;
+    lod_max = regs.proctex_lut.lod_max;
+    lut_filter.Assign(regs.proctex_lut.filter);
+}
+
+FSConfig::FSConfig(const Pica::Regs& regs, const UserConfig& user_, const Profile& profile)
+    : framebuffer{regs, profile}, texture{regs.texturing, profile}, lighting{regs.lighting},
+      proctex{regs.texturing}, user{user_} {}
+
+} // namespace Pica::Shader

src/video_core/shader/generator/pica_fs_config.h

@@ -0,0 +1,207 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/hash.h"
+#include "video_core/regs.h"
+#include "video_core/shader/generator/profile.h"
+
+namespace Pica::Shader {
+
+struct BlendConfig {
+    Pica::FramebufferRegs::BlendEquation eq;
+    Pica::FramebufferRegs::BlendFactor src_factor;
+    Pica::FramebufferRegs::BlendFactor dst_factor;
+};
+
+struct FramebufferConfig {
+    explicit FramebufferConfig(const Pica::Regs& regs, const Profile& profile);
+
+    union {
+        u32 raw{};
+        BitField<0, 3, Pica::FramebufferRegs::CompareFunc> alpha_test_func;
+        BitField<3, 2, Pica::RasterizerRegs::ScissorMode> scissor_test_mode;
+        BitField<5, 1, Pica::RasterizerRegs::DepthBuffering> depthmap_enable;
+        BitField<6, 4, Pica::FramebufferRegs::LogicOp> logic_op;
+        BitField<10, 1, u32> shadow_rendering;
+    };
+    BlendConfig rgb_blend{};
+    BlendConfig alpha_blend{};
+};
+static_assert(std::has_unique_object_representations_v<FramebufferConfig>);
+
+struct TevStageConfigRaw {
+    u32 sources_raw;
+    u32 modifiers_raw;
+    u32 ops_raw;
+    u32 scales_raw;
+    operator Pica::TexturingRegs::TevStageConfig() const noexcept {
+        return {
+            .sources_raw = sources_raw,
+            .modifiers_raw = modifiers_raw,
+            .ops_raw = ops_raw,
+            .const_color = 0,
+            .scales_raw = scales_raw,
+        };
+    }
+};
+
+union TextureBorder {
+    BitField<0, 1, u32> enable_s;
+    BitField<1, 1, u32> enable_t;
+};
+
+struct TextureConfig {
+    explicit TextureConfig(const Pica::TexturingRegs& regs, const Profile& profile);
+
+    union {
+        u32 raw{};
+        BitField<0, 3, Pica::TexturingRegs::TextureConfig::TextureType> texture0_type;
+        BitField<3, 1, u32> texture2_use_coord1;
+        BitField<4, 8, u32> combiner_buffer_input;
+        BitField<12, 3, Pica::TexturingRegs::FogMode> fog_mode;
+        BitField<15, 1, u32> fog_flip;
+        BitField<16, 1, u32> shadow_texture_orthographic;
+    };
+    std::array<TextureBorder, 3> texture_border_color{};
+    std::array<TevStageConfigRaw, 6> tev_stages{};
+};
+static_assert(std::has_unique_object_representations_v<TextureConfig>);
+
+union Light {
+    u16 raw;
+    BitField<0, 3, u16> num;
+    BitField<3, 1, u16> directional;
+    BitField<4, 1, u16> two_sided_diffuse;
+    BitField<5, 1, u16> dist_atten_enable;
+    BitField<6, 1, u16> spot_atten_enable;
+    BitField<7, 1, u16> geometric_factor_0;
+    BitField<8, 1, u16> geometric_factor_1;
+    BitField<9, 1, u16> shadow_enable;
+};
+static_assert(std::has_unique_object_representations_v<Light>);
+
+struct LutConfig {
+    union {
+        u32 raw;
+        BitField<0, 1, u32> enable;
+        BitField<1, 1, u32> abs_input;
+        BitField<2, 3, Pica::LightingRegs::LightingLutInput> type;
+    };
+    f32 scale;
+};
+
+struct LightConfig {
+    explicit LightConfig(const Pica::LightingRegs& regs);
+
+    union {
+        u32 raw{};
+        BitField<0, 1, u32> enable;
+        BitField<1, 4, u32> src_num;
+        BitField<5, 2, Pica::LightingRegs::LightingBumpMode> bump_mode;
+        BitField<7, 2, u32> bump_selector;
+        BitField<9, 1, u32> bump_renorm;
+        BitField<10, 1, u32> clamp_highlights;
+        BitField<11, 4, Pica::LightingRegs::LightingConfig> config;
+        BitField<15, 1, u32> enable_primary_alpha;
+        BitField<16, 1, u32> enable_secondary_alpha;
+        BitField<17, 1, u32> enable_shadow;
+        BitField<18, 1, u32> shadow_primary;
+        BitField<19, 1, u32> shadow_secondary;
+        BitField<20, 1, u32> shadow_invert;
+        BitField<21, 1, u32> shadow_alpha;
+        BitField<22, 2, u32> shadow_selector;
+    };
+    LutConfig lut_d0{};
+    LutConfig lut_d1{};
+    LutConfig lut_sp{};
+    LutConfig lut_fr{};
+    LutConfig lut_rr{};
+    LutConfig lut_rg{};
+    LutConfig lut_rb{};
+    std::array<Light, 8> lights{};
+};
+
+struct ProcTexConfig {
+    explicit ProcTexConfig(const Pica::TexturingRegs& regs);
+
+    union {
+        u32 raw{};
+        BitField<0, 1, u32> enable;
+        BitField<1, 2, u32> coord;
+        BitField<3, 3, Pica::TexturingRegs::ProcTexClamp> u_clamp;
+        BitField<6, 3, Pica::TexturingRegs::ProcTexClamp> v_clamp;
+        BitField<9, 4, Pica::TexturingRegs::ProcTexCombiner> color_combiner;
+        BitField<13, 4, Pica::TexturingRegs::ProcTexCombiner> alpha_combiner;
+        BitField<17, 3, Pica::TexturingRegs::ProcTexFilter> lut_filter;
+        BitField<20, 1, u32> separate_alpha;
+        BitField<21, 1, u32> noise_enable;
+        BitField<22, 2, Pica::TexturingRegs::ProcTexShift> u_shift;
+        BitField<24, 2, Pica::TexturingRegs::ProcTexShift> v_shift;
+    };
+    s32 lut_width{};
+    s32 lut_offset0{};
+    s32 lut_offset1{};
+    s32 lut_offset2{};
+    s32 lut_offset3{};
+    u16 lod_min{};
+    u16 lod_max{};
+};
+static_assert(std::has_unique_object_representations_v<ProcTexConfig>);
+
+union UserConfig {
+    u32 raw{};
+    BitField<0, 1, u32> use_custom_normal;
+};
+static_assert(std::has_unique_object_representations_v<UserConfig>);
+
+struct FSConfig {
+    explicit FSConfig(const Pica::Regs& regs, const UserConfig& user, const Profile& profile);
+
+    [[nodiscard]] bool TevStageUpdatesCombinerBufferColor(u32 stage_index) const {
+        return (stage_index < 4) && (texture.combiner_buffer_input & (1 << stage_index));
+    }
+
+    [[nodiscard]] bool TevStageUpdatesCombinerBufferAlpha(u32 stage_index) const {
+        return (stage_index < 4) && ((texture.combiner_buffer_input >> 4) & (1 << stage_index));
+    }
+
+    [[nodiscard]] bool EmulateBlend() const {
+        return framebuffer.rgb_blend.eq != Pica::FramebufferRegs::BlendEquation::Add ||
+               framebuffer.alpha_blend.eq != Pica::FramebufferRegs::BlendEquation::Add;
+    }
+
+    [[nodiscard]] bool UsesShadowPipeline() const {
+        const auto texture0_type = texture.texture0_type.Value();
+        return texture0_type == Pica::TexturingRegs::TextureConfig::Shadow2D ||
+               texture0_type == Pica::TexturingRegs::TextureConfig::ShadowCube ||
+               framebuffer.shadow_rendering.Value();
+    }
+
+    bool operator==(const FSConfig& other) const noexcept {
+        return std::memcmp(this, &other, sizeof(FSConfig)) == 0;
+    }
+
+    std::size_t Hash() const noexcept {
+        return Common::ComputeHash64(this, sizeof(FSConfig));
+    }
+
+    FramebufferConfig framebuffer;
+    TextureConfig texture;
+    LightConfig lighting;
+    ProcTexConfig proctex;
+    UserConfig user;
+};
+
+} // namespace Pica::Shader
+
+namespace std {
+template <>
+struct hash<Pica::Shader::FSConfig> {
+    std::size_t operator()(const Pica::Shader::FSConfig& k) const noexcept {
+        return k.Hash();
+    }
+};
+} // namespace std

src/video_core/shader/generator/profile.h

@@ -0,0 +1,25 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+namespace Pica::Shader {
+
+struct Profile {
+    bool has_separable_shaders{};
+    bool has_clip_planes{};
+    bool has_geometry_shader{};
+    bool has_custom_border_color{};
+    bool has_fragment_shader_interlock{};
+    bool has_blend_minmax_factor{};
+    bool has_minus_one_to_one_range{};
+    bool has_logic_op{};
+    bool has_gl_ext_framebuffer_fetch{};
+    bool has_gl_arm_framebuffer_fetch{};
+    bool has_gl_nv_fragment_shader_interlock{};
+    bool has_gl_intel_fragment_shader_interlock{};
+    bool is_vulkan{};
+};
+
+} // namespace Pica::Shader

src/video_core/shader/generator/shader_gen.cpp

@@ -9,218 +9,6 @@
 
 namespace Pica::Shader::Generator {
 
-PicaFSConfig::PicaFSConfig(const Pica::Regs& regs, bool has_fragment_shader_interlock,
-                           bool emulate_logic_op, bool emulate_custom_border_color,
-                           bool emulate_blend_minmax_factor, bool use_custom_normal_map) {
-    state.scissor_test_mode.Assign(regs.rasterizer.scissor_test.mode);
-
-    state.depthmap_enable.Assign(regs.rasterizer.depthmap_enable);
-
-    state.alpha_test_func.Assign(regs.framebuffer.output_merger.alpha_test.enable
-                                     ? regs.framebuffer.output_merger.alpha_test.func.Value()
-                                     : Pica::FramebufferRegs::CompareFunc::Always);
-
-    state.texture0_type.Assign(regs.texturing.texture0.type);
-
-    state.texture2_use_coord1.Assign(regs.texturing.main_config.texture2_use_coord1 != 0);
-
-    const auto pica_textures = regs.texturing.GetTextures();
-    for (u32 tex_index = 0; tex_index < 3; tex_index++) {
-        const auto config = pica_textures[tex_index].config;
-        state.texture_border_color[tex_index].enable_s.Assign(
-            emulate_custom_border_color &&
-            config.wrap_s == Pica::TexturingRegs::TextureConfig::WrapMode::ClampToBorder);
-        state.texture_border_color[tex_index].enable_t.Assign(
-            emulate_custom_border_color &&
-            config.wrap_t == Pica::TexturingRegs::TextureConfig::WrapMode::ClampToBorder);
-    }
-
-    // Emulate logic op in the shader if not supported. This is mostly for mobile GPUs
-    const bool needs_emulate_logic_op =
-        emulate_logic_op && !regs.framebuffer.output_merger.alphablend_enable;
-
-    state.emulate_logic_op.Assign(needs_emulate_logic_op);
-    if (needs_emulate_logic_op) {
-        state.logic_op.Assign(regs.framebuffer.output_merger.logic_op);
-    } else {
-        state.logic_op.Assign(Pica::FramebufferRegs::LogicOp::NoOp);
-    }
-
-    // Copy relevant tev stages fields.
-    // We don't sync const_color here because of the high variance, it is a
-    // shader uniform instead.
-    const auto& tev_stages = regs.texturing.GetTevStages();
-    DEBUG_ASSERT(state.tev_stages.size() == tev_stages.size());
-    for (std::size_t i = 0; i < tev_stages.size(); i++) {
-        const auto& tev_stage = tev_stages[i];
-        state.tev_stages[i].sources_raw = tev_stage.sources_raw;
-        state.tev_stages[i].modifiers_raw = tev_stage.modifiers_raw;
-        state.tev_stages[i].ops_raw = tev_stage.ops_raw;
-        state.tev_stages[i].scales_raw = tev_stage.scales_raw;
-        if (tev_stage.color_op == Pica::TexturingRegs::TevStageConfig::Operation::Dot3_RGBA) {
-            state.tev_stages[i].sources_raw &= 0xFFF;
-            state.tev_stages[i].modifiers_raw &= 0xFFF;
-            state.tev_stages[i].ops_raw &= 0xF;
-        }
-    }
-
-    state.fog_mode.Assign(regs.texturing.fog_mode);
-    state.fog_flip.Assign(regs.texturing.fog_flip != 0);
-
-    state.combiner_buffer_input.Assign(
-        regs.texturing.tev_combiner_buffer_input.update_mask_rgb.Value() |
-        regs.texturing.tev_combiner_buffer_input.update_mask_a.Value() << 4);
-
-    // Fragment lighting
-    state.lighting.enable.Assign(!regs.lighting.disable);
-    if (state.lighting.enable) {
-        state.lighting.src_num.Assign(regs.lighting.max_light_index + 1);
-
-        for (u32 light_index = 0; light_index < state.lighting.src_num; ++light_index) {
-            const u32 num = regs.lighting.light_enable.GetNum(light_index);
-            const auto& light = regs.lighting.light[num];
-            state.lighting.light[light_index].num.Assign(num);
-            state.lighting.light[light_index].directional.Assign(light.config.directional != 0);
-            state.lighting.light[light_index].two_sided_diffuse.Assign(
-                light.config.two_sided_diffuse != 0);
-            state.lighting.light[light_index].geometric_factor_0.Assign(
-                light.config.geometric_factor_0 != 0);
-            state.lighting.light[light_index].geometric_factor_1.Assign(
-                light.config.geometric_factor_1 != 0);
-            state.lighting.light[light_index].dist_atten_enable.Assign(
-                !regs.lighting.IsDistAttenDisabled(num));
-            state.lighting.light[light_index].spot_atten_enable.Assign(
-                !regs.lighting.IsSpotAttenDisabled(num));
-            state.lighting.light[light_index].shadow_enable.Assign(
-                !regs.lighting.IsShadowDisabled(num));
-        }
-
-        state.lighting.lut_d0.enable.Assign(regs.lighting.config1.disable_lut_d0 == 0);
-        if (state.lighting.lut_d0.enable) {
-            state.lighting.lut_d0.abs_input.Assign(regs.lighting.abs_lut_input.disable_d0 == 0);
-            state.lighting.lut_d0.type.Assign(regs.lighting.lut_input.d0.Value());
-            state.lighting.lut_d0.scale =
-                regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.d0);
-        }
-
-        state.lighting.lut_d1.enable.Assign(regs.lighting.config1.disable_lut_d1 == 0);
-        if (state.lighting.lut_d1.enable) {
-            state.lighting.lut_d1.abs_input.Assign(regs.lighting.abs_lut_input.disable_d1 == 0);
-            state.lighting.lut_d1.type.Assign(regs.lighting.lut_input.d1.Value());
-            state.lighting.lut_d1.scale =
-                regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.d1);
-        }
-
-        // this is a dummy field due to lack of the corresponding register
-        state.lighting.lut_sp.enable.Assign(1);
-        state.lighting.lut_sp.abs_input.Assign(regs.lighting.abs_lut_input.disable_sp == 0);
-        state.lighting.lut_sp.type.Assign(regs.lighting.lut_input.sp.Value());
-        state.lighting.lut_sp.scale = regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.sp);
-
-        state.lighting.lut_fr.enable.Assign(regs.lighting.config1.disable_lut_fr == 0);
-        if (state.lighting.lut_fr.enable) {
-            state.lighting.lut_fr.abs_input.Assign(regs.lighting.abs_lut_input.disable_fr == 0);
-            state.lighting.lut_fr.type.Assign(regs.lighting.lut_input.fr.Value());
-            state.lighting.lut_fr.scale =
-                regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.fr);
-        }
-
-        state.lighting.lut_rr.enable.Assign(regs.lighting.config1.disable_lut_rr == 0);
-        if (state.lighting.lut_rr.enable) {
-            state.lighting.lut_rr.abs_input.Assign(regs.lighting.abs_lut_input.disable_rr == 0);
-            state.lighting.lut_rr.type.Assign(regs.lighting.lut_input.rr.Value());
-            state.lighting.lut_rr.scale =
-                regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.rr);
-        }
-
-        state.lighting.lut_rg.enable.Assign(regs.lighting.config1.disable_lut_rg == 0);
-        if (state.lighting.lut_rg.enable) {
-            state.lighting.lut_rg.abs_input.Assign(regs.lighting.abs_lut_input.disable_rg == 0);
-            state.lighting.lut_rg.type.Assign(regs.lighting.lut_input.rg.Value());
-            state.lighting.lut_rg.scale =
-                regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.rg);
-        }
-
-        state.lighting.lut_rb.enable.Assign(regs.lighting.config1.disable_lut_rb == 0);
-        if (state.lighting.lut_rb.enable) {
-            state.lighting.lut_rb.abs_input.Assign(regs.lighting.abs_lut_input.disable_rb == 0);
-            state.lighting.lut_rb.type.Assign(regs.lighting.lut_input.rb.Value());
-            state.lighting.lut_rb.scale =
-                regs.lighting.lut_scale.GetScale(regs.lighting.lut_scale.rb);
-        }
-
-        state.lighting.config.Assign(regs.lighting.config0.config);
-        state.lighting.enable_primary_alpha.Assign(regs.lighting.config0.enable_primary_alpha);
-        state.lighting.enable_secondary_alpha.Assign(regs.lighting.config0.enable_secondary_alpha);
-        state.lighting.bump_mode.Assign(regs.lighting.config0.bump_mode);
-        state.lighting.bump_selector.Assign(regs.lighting.config0.bump_selector);
-        state.lighting.bump_renorm.Assign(regs.lighting.config0.disable_bump_renorm == 0);
-        state.lighting.clamp_highlights.Assign(regs.lighting.config0.clamp_highlights != 0);
-
-        state.lighting.enable_shadow.Assign(regs.lighting.config0.enable_shadow != 0);
-        if (state.lighting.enable_shadow) {
-            state.lighting.shadow_primary.Assign(regs.lighting.config0.shadow_primary != 0);
-            state.lighting.shadow_secondary.Assign(regs.lighting.config0.shadow_secondary != 0);
-            state.lighting.shadow_invert.Assign(regs.lighting.config0.shadow_invert != 0);
-            state.lighting.shadow_alpha.Assign(regs.lighting.config0.shadow_alpha != 0);
-            state.lighting.shadow_selector.Assign(regs.lighting.config0.shadow_selector);
-        }
-    }
-
-    state.proctex.enable.Assign(regs.texturing.main_config.texture3_enable);
-    if (state.proctex.enable) {
-        state.proctex.coord.Assign(regs.texturing.main_config.texture3_coordinates);
-        state.proctex.u_clamp.Assign(regs.texturing.proctex.u_clamp);
-        state.proctex.v_clamp.Assign(regs.texturing.proctex.v_clamp);
-        state.proctex.color_combiner.Assign(regs.texturing.proctex.color_combiner);
-        state.proctex.alpha_combiner.Assign(regs.texturing.proctex.alpha_combiner);
-        state.proctex.separate_alpha.Assign(regs.texturing.proctex.separate_alpha);
-        state.proctex.noise_enable.Assign(regs.texturing.proctex.noise_enable);
-        state.proctex.u_shift.Assign(regs.texturing.proctex.u_shift);
-        state.proctex.v_shift.Assign(regs.texturing.proctex.v_shift);
-        state.proctex.lut_width = regs.texturing.proctex_lut.width;
-        state.proctex.lut_offset0 = regs.texturing.proctex_lut_offset.level0;
-        state.proctex.lut_offset1 = regs.texturing.proctex_lut_offset.level1;
-        state.proctex.lut_offset2 = regs.texturing.proctex_lut_offset.level2;
-        state.proctex.lut_offset3 = regs.texturing.proctex_lut_offset.level3;
-        state.proctex.lod_min = regs.texturing.proctex_lut.lod_min;
-        state.proctex.lod_max = regs.texturing.proctex_lut.lod_max;
-        state.proctex.lut_filter.Assign(regs.texturing.proctex_lut.filter);
-    }
-
-    const auto alpha_eq = regs.framebuffer.output_merger.alpha_blending.blend_equation_a.Value();
-    const auto rgb_eq = regs.framebuffer.output_merger.alpha_blending.blend_equation_rgb.Value();
-    if (emulate_blend_minmax_factor && regs.framebuffer.output_merger.alphablend_enable) {
-        if (rgb_eq == Pica::FramebufferRegs::BlendEquation::Max ||
-            rgb_eq == Pica::FramebufferRegs::BlendEquation::Min) {
-            state.rgb_blend.emulate_blending = true;
-            state.rgb_blend.eq = rgb_eq;
-            state.rgb_blend.src_factor =
-                regs.framebuffer.output_merger.alpha_blending.factor_source_rgb;
-            state.rgb_blend.dst_factor =
-                regs.framebuffer.output_merger.alpha_blending.factor_dest_rgb;
-        }
-        if (alpha_eq == Pica::FramebufferRegs::BlendEquation::Max ||
-            alpha_eq == Pica::FramebufferRegs::BlendEquation::Min) {
-            state.alpha_blend.emulate_blending = true;
-            state.alpha_blend.eq = alpha_eq;
-            state.alpha_blend.src_factor =
-                regs.framebuffer.output_merger.alpha_blending.factor_source_a;
-            state.alpha_blend.dst_factor =
-                regs.framebuffer.output_merger.alpha_blending.factor_dest_a;
-        }
-    }
-
-    state.shadow_rendering.Assign(regs.framebuffer.output_merger.fragment_operation_mode ==
-                                  Pica::FramebufferRegs::FragmentOperationMode::Shadow);
-    state.shadow_texture_orthographic.Assign(regs.texturing.shadow.orthographic != 0);
-
-    // We only need fragment shader interlock when shadow rendering.
-    state.use_fragment_shader_interlock.Assign(state.shadow_rendering &&
-                                               has_fragment_shader_interlock);
-    state.use_custom_normal_map.Assign(use_custom_normal_map);
-}
-
 void PicaGSConfigState::Init(const Pica::Regs& regs, bool use_clip_planes_) {
     use_clip_planes = use_clip_planes_;
 

src/video_core/shader/generator/shader_gen.h

@@ -28,141 +28,6 @@ enum Attributes {
     ATTRIBUTE_VIEW,
 };
 
-// Doesn't include const_color because we don't sync it, see comment in BuildFromRegs()
-struct TevStageConfigRaw {
-    u32 sources_raw;
-    u32 modifiers_raw;
-    u32 ops_raw;
-    u32 scales_raw;
-    explicit operator Pica::TexturingRegs::TevStageConfig() const noexcept {
-        return {
-            .sources_raw = sources_raw,
-            .modifiers_raw = modifiers_raw,
-            .ops_raw = ops_raw,
-            .const_color = 0,
-            .scales_raw = scales_raw,
-        };
-    }
-};
-
-struct PicaFSConfigState {
-    union {
-        BitField<0, 3, Pica::FramebufferRegs::CompareFunc> alpha_test_func;
-        BitField<3, 2, Pica::RasterizerRegs::ScissorMode> scissor_test_mode;
-        BitField<5, 3, Pica::TexturingRegs::TextureConfig::TextureType> texture0_type;
-        BitField<8, 1, u32> texture2_use_coord1;
-        BitField<9, 8, u32> combiner_buffer_input;
-        BitField<17, 1, Pica::RasterizerRegs::DepthBuffering> depthmap_enable;
-        BitField<18, 3, Pica::TexturingRegs::FogMode> fog_mode;
-        BitField<21, 1, u32> fog_flip;
-        BitField<22, 1, u32> emulate_logic_op;
-        BitField<23, 4, Pica::FramebufferRegs::LogicOp> logic_op;
-        BitField<27, 1, u32> shadow_rendering;
-        BitField<28, 1, u32> shadow_texture_orthographic;
-        BitField<29, 1, u32> use_fragment_shader_interlock;
-        BitField<30, 1, u32> use_custom_normal_map;
-    };
-
-    union {
-        BitField<0, 1, u32> enable_s;
-        BitField<1, 1, u32> enable_t;
-    } texture_border_color[3];
-
-    std::array<TevStageConfigRaw, 6> tev_stages;
-
-    struct {
-        union {
-            BitField<0, 3, u16> num;
-            BitField<3, 1, u16> directional;
-            BitField<4, 1, u16> two_sided_diffuse;
-            BitField<5, 1, u16> dist_atten_enable;
-            BitField<6, 1, u16> spot_atten_enable;
-            BitField<7, 1, u16> geometric_factor_0;
-            BitField<8, 1, u16> geometric_factor_1;
-            BitField<9, 1, u16> shadow_enable;
-        } light[8];
-
-        union {
-            BitField<0, 1, u32> enable;
-            BitField<1, 4, u32> src_num;
-            BitField<5, 2, Pica::LightingRegs::LightingBumpMode> bump_mode;
-            BitField<7, 2, u32> bump_selector;
-            BitField<9, 1, u32> bump_renorm;
-            BitField<10, 1, u32> clamp_highlights;
-            BitField<11, 4, Pica::LightingRegs::LightingConfig> config;
-            BitField<15, 1, u32> enable_primary_alpha;
-            BitField<16, 1, u32> enable_secondary_alpha;
-            BitField<17, 1, u32> enable_shadow;
-            BitField<18, 1, u32> shadow_primary;
-            BitField<19, 1, u32> shadow_secondary;
-            BitField<20, 1, u32> shadow_invert;
-            BitField<21, 1, u32> shadow_alpha;
-            BitField<22, 2, u32> shadow_selector;
-        };
-
-        struct {
-            union {
-                BitField<0, 1, u32> enable;
-                BitField<1, 1, u32> abs_input;
-                BitField<2, 3, Pica::LightingRegs::LightingLutInput> type;
-            };
-            float scale;
-        } lut_d0, lut_d1, lut_sp, lut_fr, lut_rr, lut_rg, lut_rb;
-    } lighting;
-
-    struct {
-        union {
-            BitField<0, 1, u32> enable;
-            BitField<1, 2, u32> coord;
-            BitField<3, 3, Pica::TexturingRegs::ProcTexClamp> u_clamp;
-            BitField<6, 3, Pica::TexturingRegs::ProcTexClamp> v_clamp;
-            BitField<9, 4, Pica::TexturingRegs::ProcTexCombiner> color_combiner;
-            BitField<13, 4, Pica::TexturingRegs::ProcTexCombiner> alpha_combiner;
-            BitField<17, 3, Pica::TexturingRegs::ProcTexFilter> lut_filter;
-            BitField<20, 1, u32> separate_alpha;
-            BitField<21, 1, u32> noise_enable;
-            BitField<22, 2, Pica::TexturingRegs::ProcTexShift> u_shift;
-            BitField<24, 2, Pica::TexturingRegs::ProcTexShift> v_shift;
-        };
-        s32 lut_width;
-        s32 lut_offset0;
-        s32 lut_offset1;
-        s32 lut_offset2;
-        s32 lut_offset3;
-        u8 lod_min;
-        u8 lod_max;
-    } proctex;
-
-    struct {
-        bool emulate_blending;
-        Pica::FramebufferRegs::BlendEquation eq;
-        Pica::FramebufferRegs::BlendFactor src_factor;
-        Pica::FramebufferRegs::BlendFactor dst_factor;
-    } rgb_blend, alpha_blend;
-};
-
-/**
- * This struct contains all state used to generate the GLSL fragment shader that emulates the
- * current Pica register configuration. This struct is used as a cache key for generated GLSL shader
- * programs. The functions in glsl_shader_gen.cpp should retrieve state from this struct only, not
- * by directly accessing Pica registers. This should reduce the risk of bugs in shader generation
- * where Pica state is not being captured in the shader cache key, thereby resulting in (what should
- * be) two separate shaders sharing the same key.
- */
-struct PicaFSConfig : Common::HashableStruct<PicaFSConfigState> {
-    PicaFSConfig(const Pica::Regs& regs, bool has_fragment_shader_interlock, bool emulate_logic_op,
-                 bool emulate_custom_border_color, bool emulate_blend_minmax_factor,
-                 bool use_custom_normal_map = false);
-
-    [[nodiscard]] bool TevStageUpdatesCombinerBufferColor(unsigned stage_index) const {
-        return (stage_index < 4) && (state.combiner_buffer_input & (1 << stage_index));
-    }
-
-    [[nodiscard]] bool TevStageUpdatesCombinerBufferAlpha(unsigned stage_index) const {
-        return (stage_index < 4) && ((state.combiner_buffer_input >> 4) & (1 << stage_index));
-    }
-};
-
 enum class AttribLoadFlags {
     Float = 1 << 0,
     Sint = 1 << 1,
@@ -238,13 +103,6 @@ struct PicaFixedGSConfig : Common::HashableStruct<PicaGSConfigState> {
 } // namespace Pica::Shader::Generator
 
 namespace std {
-template <>
-struct hash<Pica::Shader::Generator::PicaFSConfig> {
-    std::size_t operator()(const Pica::Shader::Generator::PicaFSConfig& k) const noexcept {
-        return k.Hash();
-    }
-};
-
 template <>
 struct hash<Pica::Shader::Generator::PicaVSConfig> {
     std::size_t operator()(const Pica::Shader::Generator::PicaVSConfig& k) const noexcept {

src/video_core/shader/generator/spv_fs_shader_gen.cpp

@@ -2,9 +2,9 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "core/core.h"
+#include "video_core/shader/generator/spv_fs_shader_gen.h"
-#include "core/telemetry_session.h"
+
-#include "video_core/shader/generator/spv_shader_gen.h"
+namespace Pica::Shader::Generator::SPIRV {
 
 using Pica::FramebufferRegs;
 using Pica::LightingRegs;
@@ -12,12 +12,10 @@ using Pica::RasterizerRegs;
 using Pica::TexturingRegs;
 using TevStageConfig = TexturingRegs::TevStageConfig;
 
-namespace Pica::Shader::Generator::SPIRV {
-
 constexpr u32 SPIRV_VERSION_1_3 = 0x00010300;
 
-FragmentModule::FragmentModule(Core::TelemetrySession& telemetry_, const PicaFSConfig& config_)
+FragmentModule::FragmentModule(const FSConfig& config_)
-    : Sirit::Module{SPIRV_VERSION_1_3}, telemetry{telemetry_}, config{config_} {
+    : Sirit::Module{SPIRV_VERSION_1_3}, config{config_} {
     DefineArithmeticTypes();
     DefineUniformStructs();
     DefineInterface();
@@ -37,38 +35,32 @@ void FragmentModule::Generate() {
     secondary_fragment_color = ConstF32(0.f, 0.f, 0.f, 0.f);
 
     // Do not do any sort of processing if it's obvious we're not going to pass the alpha test
-    if (config.state.alpha_test_func == Pica::FramebufferRegs::CompareFunc::Never) {
+    if (config.framebuffer.alpha_test_func == Pica::FramebufferRegs::CompareFunc::Never) {
         OpKill();
         OpFunctionEnd();
         return;
     }
 
-    // Check if the fragment is outside scissor rectangle
+    // Append the scissor and depth tests
+    WriteDepth();
     WriteScissor();
 
     // Write shader bytecode to emulate all enabled PICA lights
-    if (config.state.lighting.enable) {
+    WriteLighting();
-        WriteLighting();
-    }
 
     combiner_buffer = ConstF32(0.f, 0.f, 0.f, 0.f);
     next_combiner_buffer = GetShaderDataMember(vec_ids.Get(4), ConstS32(26));
     last_tex_env_out = rounded_primary_color;
 
     // Write shader bytecode to emulate PICA TEV stages
-    for (std::size_t index = 0; index < config.state.tev_stages.size(); ++index) {
+    for (u32 index = 0; index < config.texture.tev_stages.size(); ++index) {
-        WriteTevStage(static_cast<s32>(index));
+        WriteTevStage(index);
     }
 
-    WriteAlphaTestCondition(config.state.alpha_test_func);
+    WriteAlphaTestCondition(config.framebuffer.alpha_test_func);
-
-    // After perspective divide, OpenGL transform z_over_w from [-1, 1] to [near, far]. Here we use
-    // default near = 0 and far = 1, and undo the transformation to get the original z_over_w, then
-    // do our own transformation according to PICA specification.
-    WriteDepth();
 
     // Emulate the fog
-    switch (config.state.fog_mode) {
+    switch (config.texture.fog_mode) {
     case TexturingRegs::FogMode::Fog:
         WriteFog();
         break;
@@ -80,29 +72,27 @@ void FragmentModule::Generate() {
     }
 
     Id color{Byteround(last_tex_env_out, 4)};
-    if (config.state.emulate_logic_op) {
+    switch (config.framebuffer.logic_op) {
-        switch (config.state.logic_op) {
+    case FramebufferRegs::LogicOp::Clear:
-        case FramebufferRegs::LogicOp::Clear:
+        color = ConstF32(0.f, 0.f, 0.f, 0.f);
-            color = ConstF32(0.f, 0.f, 0.f, 0.f);
+        break;
-            break;
+    case FramebufferRegs::LogicOp::Set:
-        case FramebufferRegs::LogicOp::Set:
+        color = ConstF32(1.f, 1.f, 1.f, 1.f);
-            color = ConstF32(1.f, 1.f, 1.f, 1.f);
+        break;
-            break;
+    case FramebufferRegs::LogicOp::Copy:
-        case FramebufferRegs::LogicOp::Copy:
+        // Take the color output as-is
-            // Take the color output as-is
+        break;
-            break;
+    case FramebufferRegs::LogicOp::CopyInverted:
-        case FramebufferRegs::LogicOp::CopyInverted:
+        // out += "color = ~color;\n";
-            // out += "color = ~color;\n";
+        break;
-            break;
+    case FramebufferRegs::LogicOp::NoOp:
-        case FramebufferRegs::LogicOp::NoOp:
+        // We need to discard the color, but not necessarily the depth. This is not possible
-            // We need to discard the color, but not necessarily the depth. This is not possible
+        // with fragment shader alone, so we emulate this behavior with the color mask.
-            // with fragment shader alone, so we emulate this behavior with the color mask.
+        break;
-            break;
+    default:
-        default:
+        LOG_CRITICAL(HW_GPU, "Unhandled logic_op {:x}",
-            LOG_CRITICAL(HW_GPU, "Unhandled logic_op {:x}",
+                     static_cast<u32>(config.framebuffer.logic_op.Value()));
-                         static_cast<u32>(config.state.logic_op.Value()));
+        UNIMPLEMENTED();
-            UNIMPLEMENTED();
-        }
     }
 
     // Write output color
@@ -119,7 +109,7 @@ void FragmentModule::WriteDepth() {
     const Id depth_scale{GetShaderDataMember(f32_id, ConstS32(2))};
     const Id depth_offset{GetShaderDataMember(f32_id, ConstS32(3))};
     depth = OpFma(f32_id, z_over_w, depth_scale, depth_offset);
-    if (config.state.depthmap_enable == Pica::RasterizerRegs::DepthBuffering::WBuffering) {
+    if (config.framebuffer.depthmap_enable == Pica::RasterizerRegs::DepthBuffering::WBuffering) {
         const Id gl_frag_coord_w{
             OpLoad(f32_id, OpAccessChain(input_pointer_id, gl_frag_coord_id, ConstU32(3u)))};
         depth = OpFDiv(f32_id, depth, gl_frag_coord_w);
@@ -128,7 +118,7 @@ void FragmentModule::WriteDepth() {
 }
 
 void FragmentModule::WriteScissor() {
-    if (config.state.scissor_test_mode == RasterizerRegs::ScissorMode::Disabled) {
+    if (config.framebuffer.scissor_test_mode == RasterizerRegs::ScissorMode::Disabled) {
         return;
     }
 
@@ -149,7 +139,7 @@ void FragmentModule::WriteScissor() {
     const Id cond2{OpFOrdLessThan(bvec_ids.Get(2), gl_frag_coord_xy, scissor_2)};
 
     Id result{OpAll(bool_id, OpCompositeConstruct(bvec_ids.Get(4), cond1, cond2))};
-    if (config.state.scissor_test_mode == RasterizerRegs::ScissorMode::Include) {
+    if (config.framebuffer.scissor_test_mode == RasterizerRegs::ScissorMode::Include) {
         result = OpLogicalNot(bool_id, result);
     }
 
@@ -167,7 +157,7 @@ void FragmentModule::WriteScissor() {
 void FragmentModule::WriteFog() {
     // Get index into fog LUT
     Id fog_index{};
-    if (config.state.fog_flip) {
+    if (config.texture.fog_flip) {
         fog_index = OpFMul(f32_id, OpFSub(f32_id, ConstF32(1.f), depth), ConstF32(128.f));
     } else {
         fog_index = OpFMul(f32_id, depth, ConstF32(128.f));
@@ -201,14 +191,17 @@ void FragmentModule::WriteFog() {
 
 void FragmentModule::WriteGas() {
     // TODO: Implement me
-    telemetry.AddField(Common::Telemetry::FieldType::Session, "VideoCore_Pica_UseGasMode", true);
     LOG_CRITICAL(Render, "Unimplemented gas mode");
     OpKill();
     OpFunctionEnd();
 }
 
 void FragmentModule::WriteLighting() {
-    const auto& lighting = config.state.lighting;
+    if (!config.lighting.enable) {
+        return;
+    }
+
+    const auto& lighting = config.lighting;
 
     // Define lighting globals
     Id diffuse_sum{ConstF32(0.f, 0.f, 0.f, 1.f)};
@@ -363,7 +356,7 @@ void FragmentModule::WriteLighting() {
         const Id sampler_index{ConstU32(static_cast<u32>(sampler))};
         if (abs) {
             // LUT index is in the range of (0.0, 1.0)
-            index = lighting.light[light_num].two_sided_diffuse
+            index = lighting.lights[light_num].two_sided_diffuse
                         ? OpFAbs(f32_id, index)
                         : OpFMax(f32_id, index, ConstF32(0.f));
             return lookup_lighting_lut_unsigned(sampler_index, index);
@@ -375,11 +368,12 @@ void FragmentModule::WriteLighting() {
 
     // Write the code to emulate each enabled light
     for (u32 light_index = 0; light_index < lighting.src_num; ++light_index) {
-        const auto& light_config = lighting.light[light_index];
+        const auto& light_config = lighting.lights[light_index];
 
         const auto GetLightMember = [&](s32 member) -> Id {
             const Id member_type = member < 6 ? vec_ids.Get(3) : f32_id;
-            const Id light_num{ConstS32(static_cast<s32>(lighting.light[light_index].num.Value()))};
+            const Id light_num{
+                ConstS32(static_cast<s32>(lighting.lights[light_index].num.Value()))};
             return GetShaderDataMember(member_type, ConstS32(24), light_num, ConstS32(member));
         };
 
@@ -595,7 +589,7 @@ void FragmentModule::WriteLighting() {
 
 void FragmentModule::WriteTevStage(s32 index) {
     const TexturingRegs::TevStageConfig stage =
-        static_cast<const TexturingRegs::TevStageConfig>(config.state.tev_stages[index]);
+        static_cast<const TexturingRegs::TevStageConfig>(config.texture.tev_stages[index]);
 
     // Detects if a TEV stage is configured to be skipped (to avoid generating unnecessary code)
     const auto is_passthrough_tev_stage = [](const TevStageConfig& stage) {
@@ -860,8 +854,6 @@ Id FragmentModule::AppendProcTexCombineAndMap(ProcTexCombiner combiner, Id u, Id
 }
 
 void FragmentModule::DefineTexSampler(u32 texture_unit) {
-    const PicaFSConfigState& state = config.state;
-
     const Id func_type{TypeFunction(vec_ids.Get(4))};
     sample_tex_unit_func[texture_unit] =
         OpFunction(vec_ids.Get(4), spv::FunctionControlMask::MaskNone, func_type);
@@ -869,14 +861,15 @@ void FragmentModule::DefineTexSampler(u32 texture_unit) {
 
     const Id zero_vec{ConstF32(0.f, 0.f, 0.f, 0.f)};
 
-    if (texture_unit == 0 && state.texture0_type == TexturingRegs::TextureConfig::Disabled) {
+    if (texture_unit == 0 &&
+        config.texture.texture0_type == TexturingRegs::TextureConfig::Disabled) {
         OpReturnValue(zero_vec);
         OpFunctionEnd();
         return;
     }
 
     if (texture_unit == 3) {
-        if (state.proctex.enable) {
+        if (config.proctex.enable) {
             OpReturnValue(ProcTexSampler());
         } else {
             OpReturnValue(zero_vec);
@@ -888,10 +881,10 @@ void FragmentModule::DefineTexSampler(u32 texture_unit) {
     const Id border_label{OpLabel()};
     const Id not_border_label{OpLabel()};
 
-    u32 texcoord_num = texture_unit == 2 && state.texture2_use_coord1 ? 1 : texture_unit;
+    u32 texcoord_num = texture_unit == 2 && config.texture.texture2_use_coord1 ? 1 : texture_unit;
     const Id texcoord{OpLoad(vec_ids.Get(2), texcoord_id[texcoord_num])};
 
-    auto& texture_border_color = state.texture_border_color[texture_unit];
+    const auto& texture_border_color = config.texture.texture_border_color[texture_unit];
     if (texture_border_color.enable_s || texture_border_color.enable_t) {
         const Id texcoord_s{OpCompositeExtract(f32_id, texcoord, 0)};
         const Id texcoord_t{OpCompositeExtract(f32_id, texcoord, 1)};
@@ -960,7 +953,7 @@ void FragmentModule::DefineTexSampler(u32 texture_unit) {
     switch (texture_unit) {
     case 0:
         // Only unit 0 respects the texturing type
-        switch (state.texture0_type) {
+        switch (config.texture.texture0_type) {
         case Pica::TexturingRegs::TextureConfig::Texture2D:
             ret_val = sample_lod(tex0_id);
             break;
@@ -976,7 +969,8 @@ void FragmentModule::DefineTexSampler(u32 texture_unit) {
             // return "shadowTextureCube(texcoord0, texcoord0_w)";
             break;
         default:
-            LOG_CRITICAL(Render, "Unhandled texture type {:x}", state.texture0_type.Value());
+            LOG_CRITICAL(Render, "Unhandled texture type {:x}",
+                         config.texture.texture0_type.Value());
             UNIMPLEMENTED();
             ret_val = zero_vec;
             break;
@@ -999,7 +993,7 @@ void FragmentModule::DefineTexSampler(u32 texture_unit) {
 
 Id FragmentModule::ProcTexSampler() {
     // Define noise tables at the beginning of the function
-    if (config.state.proctex.noise_enable) {
+    if (config.proctex.noise_enable) {
         noise1d_table =
             DefineVar<false>(TypeArray(i32_id, ConstU32(16u)), spv::StorageClass::Function);
         noise2d_table =
@@ -1008,8 +1002,8 @@ Id FragmentModule::ProcTexSampler() {
     lut_offsets = DefineVar<false>(TypeArray(i32_id, ConstU32(8u)), spv::StorageClass::Function);
 
     Id uv{};
-    if (config.state.proctex.coord < 3) {
+    if (config.proctex.coord < 3) {
-        const Id texcoord{OpLoad(vec_ids.Get(2), texcoord_id[config.state.proctex.coord.Value()])};
+        const Id texcoord{OpLoad(vec_ids.Get(2), texcoord_id[config.proctex.coord.Value()])};
         uv = OpFAbs(vec_ids.Get(2), texcoord);
     } else {
         LOG_CRITICAL(Render, "Unexpected proctex.coord >= 3");
@@ -1027,26 +1021,24 @@ Id FragmentModule::ProcTexSampler() {
     // unlike normal texture, the bias is inside the log2
     const Id proctex_bias{GetShaderDataMember(f32_id, ConstS32(16))};
     const Id bias{
-        OpFMul(f32_id, ConstF32(static_cast<f32>(config.state.proctex.lut_width)), proctex_bias)};
+        OpFMul(f32_id, ConstF32(static_cast<f32>(config.proctex.lut_width)), proctex_bias)};
     const Id duv_xy{
         OpFAdd(f32_id, OpCompositeExtract(f32_id, duv, 0), OpCompositeExtract(f32_id, duv, 1))};
 
     Id lod{OpLog2(f32_id, OpFMul(f32_id, OpFAbs(f32_id, bias), duv_xy))};
     lod = OpSelect(f32_id, OpFOrdEqual(bool_id, proctex_bias, ConstF32(0.f)), ConstF32(0.f), lod);
-    lod = OpFClamp(f32_id, lod,
+    lod =
-                   ConstF32(std::max(0.0f, static_cast<float>(config.state.proctex.lod_min))),
+        OpFClamp(f32_id, lod, ConstF32(std::max(0.0f, static_cast<float>(config.proctex.lod_min))),
-                   ConstF32(std::min(7.0f, static_cast<float>(config.state.proctex.lod_max))));
+                 ConstF32(std::min(7.0f, static_cast<float>(config.proctex.lod_max))));
 
     // Get shift offset before noise generation
     const Id u_shift{AppendProcTexShiftOffset(OpCompositeExtract(f32_id, uv, 1),
-                                              config.state.proctex.u_shift,
+                                              config.proctex.u_shift, config.proctex.u_clamp)};
-                                              config.state.proctex.u_clamp)};
     const Id v_shift{AppendProcTexShiftOffset(OpCompositeExtract(f32_id, uv, 0),
-                                              config.state.proctex.v_shift,
+                                              config.proctex.v_shift, config.proctex.v_clamp)};
-                                              config.state.proctex.v_clamp)};
 
     // Generate noise
-    if (config.state.proctex.noise_enable) {
+    if (config.proctex.noise_enable) {
         const Id proctex_noise_a{GetShaderDataMember(vec_ids.Get(2), ConstS32(21))};
         const Id noise_coef{ProcTexNoiseCoef(uv)};
         uv = OpFAdd(vec_ids.Get(2), uv,
@@ -1059,16 +1051,16 @@ Id FragmentModule::ProcTexSampler() {
     Id v{OpFAdd(f32_id, OpCompositeExtract(f32_id, uv, 1), v_shift)};
 
     // Clamp
-    u = AppendProcTexClamp(u, config.state.proctex.u_clamp);
+    u = AppendProcTexClamp(u, config.proctex.u_clamp);
-    v = AppendProcTexClamp(v, config.state.proctex.v_clamp);
+    v = AppendProcTexClamp(v, config.proctex.v_clamp);
 
     // Combine and map
     const Id proctex_color_map_offset{GetShaderDataMember(i32_id, ConstS32(12))};
-    const Id lut_coord{AppendProcTexCombineAndMap(config.state.proctex.color_combiner, u, v,
+    const Id lut_coord{
-                                                  proctex_color_map_offset)};
+        AppendProcTexCombineAndMap(config.proctex.color_combiner, u, v, proctex_color_map_offset)};
 
     Id final_color{};
-    switch (config.state.proctex.lut_filter) {
+    switch (config.proctex.lut_filter) {
     case ProcTexFilter::Linear:
     case ProcTexFilter::Nearest: {
         final_color = SampleProcTexColor(lut_coord, ConstS32(0));
@@ -1090,9 +1082,9 @@ Id FragmentModule::ProcTexSampler() {
     }
     }
 
-    if (config.state.proctex.separate_alpha) {
+    if (config.proctex.separate_alpha) {
         const Id proctex_alpha_map_offset{GetShaderDataMember(i32_id, ConstS32(13))};
-        const Id final_alpha{AppendProcTexCombineAndMap(config.state.proctex.alpha_combiner, u, v,
+        const Id final_alpha{AppendProcTexCombineAndMap(config.proctex.alpha_combiner, u, v,
                                                         proctex_alpha_map_offset)};
         final_color = OpCompositeInsert(vec_ids.Get(4), final_alpha, final_color, 3);
     }
@@ -1189,13 +1181,11 @@ Id FragmentModule::ProcTexNoiseCoef(Id x) {
 }
 
 Id FragmentModule::SampleProcTexColor(Id lut_coord, Id level) {
-    const Id lut_width{
+    const Id lut_width{OpShiftRightArithmetic(i32_id, ConstS32(config.proctex.lut_width), level)};
-        OpShiftRightArithmetic(i32_id, ConstS32(config.state.proctex.lut_width), level)};
     const Id lut_ptr{TypePointer(spv::StorageClass::Function, i32_id)};
     // Offsets for level 4-7 seem to be hardcoded
-    InitTableS32(lut_offsets, config.state.proctex.lut_offset0, config.state.proctex.lut_offset1,
+    InitTableS32(lut_offsets, config.proctex.lut_offset0, config.proctex.lut_offset1,
-                 config.state.proctex.lut_offset2, config.state.proctex.lut_offset3, 0xF0, 0xF8,
+                 config.proctex.lut_offset2, config.proctex.lut_offset3, 0xF0, 0xF8, 0xFC, 0xFE);
-                 0xFC, 0xFE);
     const Id lut_offset{OpLoad(i32_id, OpAccessChain(lut_ptr, lut_offsets, level))};
     // For the color lut, coord=0.0 is lut[offset] and coord=1.0 is lut[offset+width-1]
     lut_coord =
@@ -1209,7 +1199,7 @@ Id FragmentModule::SampleProcTexColor(Id lut_coord, Id level) {
     const Id proctex_lut_offset{GetShaderDataMember(i32_id, ConstS32(14))};
     const Id lut_rgba{OpImage(image_buffer_id, texture_buffer_lut_rgba)};
 
-    switch (config.state.proctex.lut_filter) {
+    switch (config.proctex.lut_filter) {
     case ProcTexFilter::Linear:
     case ProcTexFilter::LinearMipmapLinear:
     case ProcTexFilter::LinearMipmapNearest: {
@@ -1549,9 +1539,8 @@ void FragmentModule::DefineInterface() {
     Decorate(gl_frag_depth_id, spv::Decoration::BuiltIn, spv::BuiltIn::FragDepth);
 }
 
-std::vector<u32> GenerateFragmentShader(const PicaFSConfig& config) {
+std::vector<u32> GenerateFragmentShader(const FSConfig& config) {
-    auto& telemetry = Core::System::GetInstance().TelemetrySession();
+    FragmentModule module{config};
-    FragmentModule module{telemetry, config};
     module.Generate();
     return module.Assemble();
 }

src/video_core/shader/generator/spv_fs_shader_gen.h

@@ -7,11 +7,7 @@
 #include <array>
 #include <sirit/sirit.h>
 
-#include "video_core/shader/generator/shader_gen.h"
+#include "video_core/shader/generator/pica_fs_config.h"
-
-namespace Core {
-class TelemetrySession;
-}
 
 namespace Pica::Shader::Generator::SPIRV {
 
@@ -34,7 +30,7 @@ class FragmentModule : public Sirit::Module {
     static constexpr u32 NUM_NON_PROC_TEX_UNITS = 3;
 
 public:
-    explicit FragmentModule(Core::TelemetrySession& telemetry, const PicaFSConfig& config);
+    explicit FragmentModule(const FSConfig& config);
     ~FragmentModule();
 
     /// Emits SPIR-V bytecode corresponding to the provided pica fragment configuration
@@ -218,8 +214,7 @@ private:
     Id CompareShadow(Id pixel, Id z);
 
 private:
-    Core::TelemetrySession& telemetry;
+    const FSConfig& config;
-    PicaFSConfig config;
     Id void_id{};
     Id bool_id{};
     Id f32_id{};
@@ -289,6 +284,6 @@ private:
  * @param separable_shader generates shader that can be used for separate shader object
  * @returns String of the shader source code
  */
-std::vector<u32> GenerateFragmentShader(const PicaFSConfig& config);
+std::vector<u32> GenerateFragmentShader(const FSConfig& config);
 
 } // namespace Pica::Shader::Generator::SPIRV

src/video_core/shader/shader.cpp

@@ -15,7 +15,9 @@
 #include "video_core/shader/shader_interpreter.h"
 #if CITRA_ARCH(x86_64)
 #include "video_core/shader/shader_jit_x64.h"
-#endif // CITRA_ARCH(x86_64)
+#elif CITRA_ARCH(arm64)
+#include "video_core/shader/shader_jit_a64.h"
+#endif
 #include "video_core/video_core.h"
 
 namespace Pica::Shader {
@@ -141,27 +143,29 @@ MICROPROFILE_DEFINE(GPU_Shader, "GPU", "Shader", MP_RGB(50, 50, 240));
 
 #if CITRA_ARCH(x86_64)
 static std::unique_ptr<JitX64Engine> jit_engine;
-#endif // CITRA_ARCH(x86_64)
+#elif CITRA_ARCH(arm64)
+static std::unique_ptr<JitA64Engine> jit_engine;
+#endif
 static InterpreterEngine interpreter_engine;
 
 ShaderEngine* GetEngine() {
-#if CITRA_ARCH(x86_64)
+#if CITRA_ARCH(x86_64) || CITRA_ARCH(arm64)
     // TODO(yuriks): Re-initialize on each change rather than being persistent
     if (VideoCore::g_shader_jit_enabled) {
         if (jit_engine == nullptr) {
-            jit_engine = std::make_unique<JitX64Engine>();
+            jit_engine = std::make_unique<decltype(jit_engine)::element_type>();
         }
         return jit_engine.get();
     }
-#endif // CITRA_ARCH(x86_64)
+#endif // CITRA_ARCH(x86_64) || CITRA_ARCH(arm64)
 
     return &interpreter_engine;
 }
 
 void Shutdown() {
-#if CITRA_ARCH(x86_64)
+#if CITRA_ARCH(x86_64) || CITRA_ARCH(arm64)
     jit_engine = nullptr;
-#endif // CITRA_ARCH(x86_64)
+#endif // CITRA_ARCH(x86_64) || CITRA_ARCH(arm64)
 }
 
 } // namespace Pica::Shader

src/video_core/shader/shader_jit_a64.cpp

@@ -0,0 +1,51 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/arch.h"
+#if CITRA_ARCH(arm64)
+
+#include "common/assert.h"
+#include "common/microprofile.h"
+#include "video_core/shader/shader.h"
+#include "video_core/shader/shader_jit_a64.h"
+#include "video_core/shader/shader_jit_a64_compiler.h"
+
+namespace Pica::Shader {
+
+JitA64Engine::JitA64Engine() = default;
+JitA64Engine::~JitA64Engine() = default;
+
+void JitA64Engine::SetupBatch(ShaderSetup& setup, unsigned int entry_point) {
+    ASSERT(entry_point < MAX_PROGRAM_CODE_LENGTH);
+    setup.engine_data.entry_point = entry_point;
+
+    u64 code_hash = setup.GetProgramCodeHash();
+    u64 swizzle_hash = setup.GetSwizzleDataHash();
+
+    u64 cache_key = code_hash ^ swizzle_hash;
+    auto iter = cache.find(cache_key);
+    if (iter != cache.end()) {
+        setup.engine_data.cached_shader = iter->second.get();
+    } else {
+        auto shader = std::make_unique<JitShader>();
+        shader->Compile(&setup.program_code, &setup.swizzle_data);
+        setup.engine_data.cached_shader = shader.get();
+        cache.emplace_hint(iter, cache_key, std::move(shader));
+    }
+}
+
+MICROPROFILE_DECLARE(GPU_Shader);
+
+void JitA64Engine::Run(const ShaderSetup& setup, UnitState& state) const {
+    ASSERT(setup.engine_data.cached_shader != nullptr);
+
+    MICROPROFILE_SCOPE(GPU_Shader);
+
+    const JitShader* shader = static_cast<const JitShader*>(setup.engine_data.cached_shader);
+    shader->Run(setup, state, setup.engine_data.entry_point);
+}
+
+} // namespace Pica::Shader
+
+#endif // CITRA_ARCH(arm64)

src/video_core/shader/shader_jit_a64.h

@@ -0,0 +1,33 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/arch.h"
+#if CITRA_ARCH(arm64)
+
+#include <memory>
+#include <unordered_map>
+#include "common/common_types.h"
+#include "video_core/shader/shader.h"
+
+namespace Pica::Shader {
+
+class JitShader;
+
+class JitA64Engine final : public ShaderEngine {
+public:
+    JitA64Engine();
+    ~JitA64Engine() override;
+
+    void SetupBatch(ShaderSetup& setup, unsigned int entry_point) override;
+    void Run(const ShaderSetup& setup, UnitState& state) const override;
+
+private:
+    std::unordered_map<u64, std::unique_ptr<JitShader>> cache;
+};
+
+} // namespace Pica::Shader
+
+#endif // CITRA_ARCH(arm64)

src/video_core/shader/shader_jit_a64_compiler.h

@@ -0,0 +1,146 @@
+// Copyright 2023 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/arch.h"
+#if CITRA_ARCH(arm64)
+
+#include <array>
+#include <bitset>
+#include <cstddef>
+#include <optional>
+#include <utility>
+#include <vector>
+#include <nihstro/shader_bytecode.h>
+#include <oaknut/code_block.hpp>
+#include <oaknut/oaknut.hpp>
+#include "common/common_types.h"
+#include "video_core/shader/shader.h"
+
+using nihstro::Instruction;
+using nihstro::OpCode;
+using nihstro::SourceRegister;
+using nihstro::SwizzlePattern;
+
+namespace Pica::Shader {
+
+/// Memory allocated for each compiled shader
+constexpr std::size_t MAX_SHADER_SIZE = MAX_PROGRAM_CODE_LENGTH * 256;
+
+/**
+ * This class implements the shader JIT compiler. It recompiles a Pica shader program into x86_64
+ * code that can be executed on the host machine directly.
+ */
+class JitShader : private oaknut::CodeBlock, public oaknut::CodeGenerator {
+public:
+    JitShader();
+
+    void Run(const ShaderSetup& setup, UnitState& state, unsigned offset) const {
+        program(&setup.uniforms, &state, instruction_labels[offset].ptr<const std::byte*>());
+    }
+
+    void Compile(const std::array<u32, MAX_PROGRAM_CODE_LENGTH>* program_code,
+                 const std::array<u32, MAX_SWIZZLE_DATA_LENGTH>* swizzle_data);
+
+    void Compile_ADD(Instruction instr);
+    void Compile_DP3(Instruction instr);
+    void Compile_DP4(Instruction instr);
+    void Compile_DPH(Instruction instr);
+    void Compile_EX2(Instruction instr);
+    void Compile_LG2(Instruction instr);
+    void Compile_MUL(Instruction instr);
+    void Compile_SGE(Instruction instr);
+    void Compile_SLT(Instruction instr);
+    void Compile_FLR(Instruction instr);
+    void Compile_MAX(Instruction instr);
+    void Compile_MIN(Instruction instr);
+    void Compile_RCP(Instruction instr);
+    void Compile_RSQ(Instruction instr);
+    void Compile_MOVA(Instruction instr);
+    void Compile_MOV(Instruction instr);
+    void Compile_NOP(Instruction instr);
+    void Compile_END(Instruction instr);
+    void Compile_BREAKC(Instruction instr);
+    void Compile_CALL(Instruction instr);
+    void Compile_CALLC(Instruction instr);
+    void Compile_CALLU(Instruction instr);
+    void Compile_IF(Instruction instr);
+    void Compile_LOOP(Instruction instr);
+    void Compile_JMP(Instruction instr);
+    void Compile_CMP(Instruction instr);
+    void Compile_MAD(Instruction instr);
+    void Compile_EMIT(Instruction instr);
+    void Compile_SETE(Instruction instr);
+
+private:
+    void Compile_Block(unsigned end);
+    void Compile_NextInstr();
+
+    void Compile_SwizzleSrc(Instruction instr, unsigned src_num, SourceRegister src_reg,
+                            oaknut::QReg dest);
+    void Compile_DestEnable(Instruction instr, oaknut::QReg dest);
+
+    /**
+     * Compiles a `MUL src1, src2` operation, properly handling the PICA semantics when multiplying
+     * zero by inf. Clobbers `src2` and `scratch`.
+     */
+    void Compile_SanitizedMul(oaknut::QReg src1, oaknut::QReg src2, oaknut::QReg scratch0);
+
+    void Compile_EvaluateCondition(Instruction instr);
+    void Compile_UniformCondition(Instruction instr);
+
+    /**
+     * Emits the code to conditionally return from a subroutine envoked by the `CALL` instruction.
+     */
+    void Compile_Return();
+
+    std::bitset<64> PersistentCallerSavedRegs();
+
+    /**
+     * Assertion evaluated at compile-time, but only triggered if executed at runtime.
+     * @param condition Condition to be evaluated.
+     * @param msg       Message to be logged if the assertion fails.
+     */
+    void Compile_Assert(bool condition, const char* msg);
+
+    /**
+     * Analyzes the entire shader program for `CALL` instructions before emitting any code,
+     * identifying the locations where a return needs to be inserted.
+     */
+    void FindReturnOffsets();
+
+    /**
+     * Emits data and code for utility functions.
+     */
+    void CompilePrelude();
+    oaknut::Label CompilePrelude_Log2();
+    oaknut::Label CompilePrelude_Exp2();
+
+    const std::array<u32, MAX_PROGRAM_CODE_LENGTH>* program_code = nullptr;
+    const std::array<u32, MAX_SWIZZLE_DATA_LENGTH>* swizzle_data = nullptr;
+
+    /// Mapping of Pica VS instructions to pointers in the emitted code
+    std::array<oaknut::Label, MAX_PROGRAM_CODE_LENGTH> instruction_labels;
+
+    /// Labels pointing to the end of each nested LOOP block. Used by the BREAKC instruction to
+    /// break out of a loop.
+    std::vector<oaknut::Label> loop_break_labels;
+
+    /// Offsets in code where a return needs to be inserted
+    std::vector<unsigned> return_offsets;
+
+    unsigned program_counter = 0; ///< Offset of the next instruction to decode
+    u8 loop_depth = 0;            ///< Depth of the (nested) loops currently compiled
+
+    using CompiledShader = void(const void* setup, void* state, const std::byte* start_addr);
+    CompiledShader* program = nullptr;
+
+    oaknut::Label log2_subroutine;
+    oaknut::Label exp2_subroutine;
+};
+
+} // namespace Pica::Shader
+
+#endif