Plumb through a parameter for creating framebuffers of different color formats

.github/workflows/build.yml

@@ -139,6 +139,12 @@ jobs:
           cxx: clang++
           args: cd android && ./ab.sh -j2 APP_ABI=armeabi-v7a UNITTEST=1 HEADLESS=1
           id: android-arm32
+        - os: ubuntu-latest
+          extra: android
+          cc: clang
+          cxx: clang++
+          args: cd android && ./ab.sh -j2 APP_ABI=x86 UNITTEST=1 HEADLESS=1
+          id: android-x86_32
         - os: ubuntu-latest
           extra: android
           cc: clang
@@ -149,8 +155,14 @@ jobs:
           extra: android
           cc: clang
           cxx: clang++
-          args: cd android && ./ab.sh -j2 APP_ABI=arm64-v8a OPENXR=1
+          args: cd android && ./ab.sh -j2 APP_ABI=arm64-v8a OPENXR=1 OPENXR_PLATFORM_QUEST=1
-          id: android-vr
+          id: android-vr-quest
+        - os: ubuntu-latest
+          extra: android
+          cc: clang
+          cxx: clang++
+          args: cd android && ./ab.sh -j2 APP_ABI=arm64-v8a OPENXR=1 OPENXR_PLATFORM_PICO=1
+          id: android-vr-pico
         - os: ubuntu-latest
           extra: android
           cc: clang
@@ -219,7 +231,7 @@ jobs:
       run: |
         sudo add-apt-repository -y "deb http://archive.ubuntu.com/ubuntu `lsb_release -sc` main universe restricted multiverse"
         sudo apt-get update -y -qq
-        sudo apt-get install libsdl2-dev libgl1-mesa-dev libglu1-mesa-dev libsdl2-ttf-dev libfontconfig1-dev
+        sudo apt-get install libsdl2-dev libgl1-mesa-dev libglu1-mesa-dev
 
     - name: Create macOS git-version.cpp for tagged release
       if: startsWith(github.ref, 'refs/tags/') && runner.os == 'macOS' && matrix.extra == 'test'
@@ -229,8 +241,6 @@ jobs:
 
     - name: Setup ccache
       uses: hendrikmuhs/ccache-action@v1.2
-      # Disable ccache on macos for now, it's become buggy for some reason.
-      if: matrix.id != 'macos'
       with:
         key: ${{ matrix.id }}
 
@@ -333,7 +343,7 @@ jobs:
       run: |
         sudo add-apt-repository -y "deb http://archive.ubuntu.com/ubuntu `lsb_release -sc` main universe restricted multiverse"
         sudo apt-get update -y -qq
-        sudo apt-get install libsdl2-dev libgl1-mesa-dev libglu1-mesa-dev libsdl2-ttf-dev libfontconfig1-dev
+        sudo apt-get install libsdl2-dev libgl1-mesa-dev libglu1-mesa-dev
 
     - name: Install macOS dependencies
       if: runner.os == 'macOS'

.gitignore

@@ -105,7 +105,6 @@ Windows/*.ipch
 # For vim
 *.swp
 tags
-*.ctags
 
 # Other VCS
 .bzr/
@@ -118,8 +117,6 @@ debian/ppsspp/
 
 # Libretro build
 *.o
-*.tmp
-*.a
 
 # YouCompleteMe file
 .ycm_extra_conf.pyc
@@ -132,7 +129,3 @@ CMakeFiles
 
 # Clangd 
 .cache/
-build
-libretro/obj/local
-
-ppsspp_retroachievements.dat

.gitlab-ci.yml

@@ -49,14 +49,6 @@ include:
   - project: 'libretro-infrastructure/ci-templates'
     file: '/android-cmake.yml'
 
-  # iOS
-  - project: 'libretro-infrastructure/ci-templates'
-    file: '/ios-cmake.yml'
-
-  # tvOS
-  - project: 'libretro-infrastructure/ci-templates'
-    file: '/tvos-cmake.yml'
-
   ################################## CONSOLES ################################
   
   #################################### MISC ##################################
@@ -141,21 +133,3 @@ libretro-build-android-x86:
   extends:
     - .libretro-android-cmake-x86
     - .core-defs
-
-# iOS arm64
-libretro-build-ios-arm64:
-  extends:
-    - .libretro-ios-cmake-arm64
-    - .core-defs
-    - .cmake-defs
-  variables:
-    CORE_ARGS: -DCMAKE_TOOLCHAIN_FILE=cmake/Toolchains/ios.cmake -DLIBRETRO=ON
-
-# tvOS arm64
-libretro-build-tvos-arm64:
-  extends:
-    - .libretro-tvos-cmake-arm64
-    - .core-defs
-    - .cmake-defs
-  variables:
-    CORE_ARGS: -DIOS_PLATFORM=TVOS -DCMAKE_TOOLCHAIN_FILE=cmake/Toolchains/ios.cmake -DLIBRETRO=ON

.gitmodules

@@ -1,6 +1,3 @@
-[submodule "libretro/libretro-common"]
-	path = libretro/libretro-common
-	url = https://github.com/libretro/libretro-common.git
 [submodule "pspautotests"]
 	path = pspautotests
 	url = https://github.com/hrydgard/pspautotests.git
@@ -44,12 +41,3 @@
 [submodule "cpu_features"]
 	path = ext/cpu_features
 	url = https://github.com/google/cpu_features.git
-[submodule "ext/rcheevos"]
-	path = ext/rcheevos
-	url = https://github.com/RetroAchievements/rcheevos.git
-[submodule "ext/naett"]
-	path = ext/naett
-	url = https://github.com/erkkah/naett.git
-[submodule "ext/libchdr"]
-	path = ext/libchdr
-	url = https://github.com/rtissera/libchdr.git

CMakeLists.txt

@@ -67,8 +67,6 @@ if(CMAKE_SYSTEM_PROCESSOR)
 		set(MIPS_DEVICE ON)
 	elseif(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^riscv64")
 		set(RISCV64_DEVICE ON)
-	elseif(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^loongarch64")
-		set(LOONGARCH64_DEVICE ON)
 	else()
 		message("Unknown CMAKE_SYSTEM_PROCESSOR: ${CMAKE_SYSTEM_PROCESSOR}")
 	endif()
@@ -104,18 +102,15 @@ if(ANDROID OR WIN32 OR (UNIX AND NOT ARM_NO_VULKAN))
 	set(VULKAN ON)
 endif()
 
-# Default to bundled SDL2 on macOS, system SDL2 elsewhere.
-if(APPLE AND NOT IOS)
-	set(DEFAULT_USE_SYSTEM_LIBSDL2 OFF)
-else()
-	set(DEFAULT_USE_SYSTEM_LIBSDL2 ON)
-endif()
-
 list(APPEND CMAKE_MODULE_PATH ${CMAKE_SOURCE_DIR}/cmake/Modules)
 if(NOT IOS)
 	list(APPEND CMAKE_MODULE_PATH ${CMAKE_SOURCE_DIR}/cmake/sdl)
 endif()
 
+if(MACOSX AND NOT USE_SYSTEM_LIBSDL2)
+	set(SDL2_LIBRARY ${CMAKE_SOURCE_DIR}/SDL/macOS/SDL2.framework)
+endif()
+
 include(ccache)
 include(GNUInstallDirs)
 
@@ -124,8 +119,16 @@ add_definitions(-DASSETS_DIR="${CMAKE_INSTALL_FULL_DATADIR}/ppsspp/assets/")
 if(OPENXR)
 	add_definitions(-DOPENXR)
 	add_library(openxr SHARED IMPORTED)
-	set_property(TARGET openxr PROPERTY IMPORTED_LOCATION "${CMAKE_SOURCE_DIR}/ext/openxr/stub/arm64-v8a/libopenxr_loader.so")
+	if(OPENXR_PLATFORM_PICO)
-	message("OpenXR enabled")
+	  add_definitions(-DOPENXR_PLATFORM_PICO)
+	  set_property(TARGET openxr PROPERTY IMPORTED_LOCATION "${CMAKE_SOURCE_DIR}/ext/openxr/libs/pico/arm64-v8a/libopenxr_loader.so")
+	  message("OpenXR for Pico enabled")
+	endif()
+	if(OPENXR_PLATFORM_QUEST)
+	  add_definitions(-DOPENXR_PLATFORM_QUEST)
+	  set_property(TARGET openxr PROPERTY IMPORTED_LOCATION "${CMAKE_SOURCE_DIR}/ext/openxr/libs/quest/arm64-v8a/libopenxr_loader.so")
+	  message("OpenXR for Quest enabled")
+	endif()
 endif()
 
 if(GOLD)
@@ -141,7 +144,6 @@ option(ARMV7 "Set to ON if targeting an ARMv7 processor" ${ARMV7_DEVICE})
 option(ARM "Set to ON if targeting an ARM processor" ${ARM_DEVICE})
 option(MIPS "Set to ON if targeting a MIPS processor" ${MIPS_DEVICE})
 option(RISCV64 "Set to ON if targeting a RISCV64 processor" ${RISCV64_DEVICE})
-option(LOONGARCH64 "Set to ON if targeting a LOONGARCH64 processor" ${LOONGARCH64_DEVICE})
 option(X86 "Set to ON if targeting an X86 processor" ${X86_DEVICE})
 option(X86_64 "Set to ON if targeting an X86_64 processor" ${X86_64_DEVICE})
 # :: Environments
@@ -149,7 +151,7 @@ option(USING_EGL "Set to ON if target environment uses EGL" ${USING_EGL})
 option(USING_FBDEV "Set to ON if target environment uses fbdev (eg. Pandora)" ${USING_FBDEV})
 option(USING_GLES2 "Set to ON if target device uses OpenGL ES 2.0" ${USING_GLES2})
 option(USING_X11_VULKAN "Set to OFF if target environment doesn't use X11 for Vulkan" ON)
-option(USE_WAYLAND_WSI "Enable or disable Wayland WSI support for Vulkan" ON)
+option(USE_WAYLAND_WSI "Enable or disable Wayland WSI support for Vulkan" ${USE_WAYLAND_WSI})
 option(USE_VULKAN_DISPLAY_KHR "Enable or disable full screen display of Vulkan" ${USE_VULKAN_DISPLAY_KHR})
 # :: Frontends
 option(USING_QT_UI "Set to ON if you wish to use the Qt frontend wrapper" ${USING_QT_UI})
@@ -167,7 +169,7 @@ option(USE_ARMIPS "Build with armips support in API/debugger" ON)
 option(USE_SYSTEM_SNAPPY "Dynamically link against system snappy" ${USE_SYSTEM_SNAPPY})
 option(USE_SYSTEM_FFMPEG "Dynamically link against system FFMPEG" ${USE_SYSTEM_FFMPEG})
 option(USE_SYSTEM_LIBZIP "Dynamically link against system libzip" ${USE_SYSTEM_LIBZIP})
-option(USE_SYSTEM_LIBSDL2 "Dynamically link against system SDL2" ${DEFAULT_USE_SYSTEM_LIBSDL2})
+option(USE_SYSTEM_LIBSDL2 "Dynamically link against system SDL2" ON)
 option(USE_SYSTEM_LIBPNG "Dynamically link against system libpng" ON)
 option(USE_SYSTEM_ZSTD "Dynamically link against system zstd" ${USE_SYSTEM_ZSTD})
 option(USE_SYSTEM_MINIUPNPC "Dynamically link against system miniUPnPc" ${USE_SYSTEM_MINIUPNPC})
@@ -177,20 +179,20 @@ option(USE_UBSAN "Use undefined behaviour sanitizer" OFF)
 if(UNIX AND NOT (APPLE OR ANDROID) AND VULKAN)
 	if(USING_X11_VULKAN)
 		message("Using X11 for Vulkan")
-		find_package(X11)
-		include_directories(${X11_Xlib_INCLUDE_PATH})
 		add_definitions(-DVK_USE_PLATFORM_XLIB_KHR)
 	else()
 		message("NOT using X11 for Vulkan")
 	endif()
 
 	# add_definitions(-DVK_USE_PLATFORM_XCB_KHR)
-	find_package(Wayland)
+	if(USE_WAYLAND_WSI)
-	if(NOT WAYLAND_FOUND)
+		find_package(Wayland)
-		message(STATUS "Could not find Wayland libraries, disabling Wayland WSI support for Vulkan.")
+		if(NOT WAYLAND_FOUND)
-	elseif(USE_WAYLAND_WSI)
+			message(STATUS "Could not find Wayland libraries, disabling Wayland WSI support for Vulkan.")
-		include_directories(${WAYLAND_INCLUDE_DIR})
+		else()
-		add_definitions(-DVK_USE_PLATFORM_WAYLAND_KHR)
+			include_directories(${WAYLAND_INCLUDE_DIR})
+			add_definitions(-DVK_USE_PLATFORM_WAYLAND_KHR)
+		endif()
 	endif()
 
 	if(USE_VULKAN_DISPLAY_KHR)
@@ -220,16 +222,6 @@ else()
 	set(CoreLinkType STATIC)
 endif()
 
-if(NOT ANDROID AND NOT WIN32 AND (NOT APPLE OR IOS))
-	set(HTTPS_NOT_AVAILABLE ON)
-endif()
-
-# Made this flag negative because it's hopefully quite temporary and didn't
-# want to have to update all build systems.
-if(HTTPS_NOT_AVAILABLE)
-	add_definitions(-DHTTPS_NOT_AVAILABLE)
-endif()
-
 # Work around for some misfeature of the current glslang build system
 include_directories(ext/glslang)
 
@@ -257,31 +249,9 @@ if(USING_EGL)
 	set(OPENGL_LIBRARIES ${OPENGL_LIBRARIES} ${EGL_LIBRARIES})
 endif()
 
-if(NOT LIBRETRO AND NOT IOS AND NOT MACOSX)
+if(NOT LIBRETRO AND NOT IOS)
 	find_package(SDL2)
-	find_package(SDL2_ttf)
-	find_package(Fontconfig)
-
-	# TODO: this can be removed once CI supports newer SDL2_ttf
-	if (NOT SDL2_ttf_FOUND)
-		find_package(PkgConfig)
-		if(PkgConfig_FOUND)
-			pkg_check_modules(SDL2_ttf_PKGCONFIG IMPORTED_TARGET SDL2_ttf)
-		endif()
-	endif()
 endif()
-
-if(MACOSX AND NOT IOS)
-	if(USE_SYSTEM_LIBSDL2)
-		find_package(SDL2)
-		find_package(SDL2_ttf)
-	else()
-		find_library(SDL2Fwk SDL2 REQUIRED PATHS SDL/macOS)
-		message(STATUS "found SDL2Fwk=${SDL2Fwk}")
-		add_definitions(-DHAVE_SYSCTLBYNAME)
-	endif()
-endif()
-
 include(FindThreads)
 
 if(APPLE)
@@ -328,9 +298,6 @@ endif()
 if(RISCV64)
 	message("Generating for RISCV64, ${CMAKE_BUILD_TYPE}")
 endif()
-if(LOONGARCH64)
-	message("Generating for LOONGARCH64, ${CMAKE_BUILD_TYPE}")
-endif()
 if(X86)
 	message("Generating for x86, ${CMAKE_BUILD_TYPE}")
 endif()
@@ -406,15 +373,9 @@ if(NOT MSVC)
 		set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -parallel -fopenmp")
 	endif()
 
-	add_definitions(-fno-math-errno)
-
 	if(X86 OR X86_64)
 		# enable sse2 code generation
 		add_definitions(-msse2)
-		if(NOT X86_64 AND NOT CLANG)
-			add_definitions(-mfpmath=sse)
-			# add_definitions(-mstackrealign)
-		 endif()
 	endif()
 
 	if(IOS)
@@ -433,13 +394,8 @@ if(NOT MSVC)
 			add_definitions(-Wno-psabi)
 		endif()
 		add_definitions(-D_XOPEN_SOURCE=700)
-		add_definitions(-D_XOPEN_SOURCE_EXTENDED -D__BSD_VISIBLE=1 -D_BSD_SOURCE -D_DEFAULT_SOURCE)
+		add_definitions(-D_XOPEN_SOURCE_EXTENDED -D__BSD_VISIBLE=1)
-		if(CMAKE_SYSTEM_NAME MATCHES "Linux|SunOS" OR MINGW)
+		add_definitions(-D_LARGEFILE64_SOURCE=1 -D_FILE_OFFSET_BITS=64)
-			add_definitions(-D_LARGEFILE64_SOURCE=1 -D_FILE_OFFSET_BITS=64)
-		endif()
-		if(${CMAKE_SYSTEM_NAME} STREQUAL "NetBSD")
-			add_definitions(-D_NETBSD_SOURCE)
-		endif()
 	elseif(ANDROID)
 		add_definitions(-fsigned-char)
 	endif()
@@ -451,7 +407,6 @@ else()
 	endif()
 	set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -D_DEBUG")
 	set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -D_NDEBUG")
-	set(CMAKE_EXE_LINKER_FLAGS /NODEFAULTLIB:"libcmt.lib")
 endif()
 
 if(WIN32)
@@ -544,14 +499,6 @@ set(CommonRISCV64
 )
 source_group(RISCV64 FILES ${CommonRISCV64})
 
-set(CommonLOONGARCH64
-	${CommonJIT}
-	Common/LoongArchCPUDetect.cpp
-	Core/MIPS/fake/FakeJit.cpp
-	Core/MIPS/fake/FakeJit.h
-)
-source_group(LOONGARCH64 FILES ${CommonLOONGARCH64})
-
 if(WIN32)
 	set(CommonD3D
 		Common/GPU/D3D9/D3D9ShaderCompiler.cpp
@@ -590,7 +537,6 @@ add_library(Common STATIC
 	${CommonARM64}
 	${CommonMIPS}
 	${CommonRISCV64}
-	${CommonLOONGARCH64}
 	${CommonD3D}
 	${CommonVR}
 	Common/Serialize/Serializer.cpp
@@ -610,7 +556,6 @@ add_library(Common STATIC
 	Common/Data/Collections/FixedSizeQueue.h
 	Common/Data/Collections/Hashmaps.h
 	Common/Data/Collections/TinySet.h
-	Common/Data/Collections/FastVec.h
 	Common/Data/Collections/ThreadSafeList.h
 	Common/Data/Color/RGBAUtil.cpp
 	Common/Data/Color/RGBAUtil.h
@@ -634,8 +579,6 @@ add_library(Common STATIC
 	Common/Data/Format/JSONReader.cpp
 	Common/Data/Format/JSONWriter.h
 	Common/Data/Format/JSONWriter.cpp
-	Common/Data/Format/DDSLoad.cpp
-	Common/Data/Format/DDSLoad.h
 	Common/Data/Format/PNGLoad.cpp
 	Common/Data/Format/PNGLoad.h
 	Common/Data/Format/ZIMLoad.cpp
@@ -653,14 +596,10 @@ add_library(Common STATIC
 	Common/Data/Random/Rng.h
 	Common/File/VFS/VFS.h
 	Common/File/VFS/VFS.cpp
-	Common/File/VFS/ZipFileReader.cpp
+	Common/File/VFS/AssetReader.cpp
-	Common/File/VFS/ZipFileReader.h
+	Common/File/VFS/AssetReader.h
-	Common/File/VFS/DirectoryReader.cpp
-	Common/File/VFS/DirectoryReader.h
 	Common/File/AndroidStorage.h
 	Common/File/AndroidStorage.cpp
-	Common/File/AndroidContentURI.h
-	Common/File/AndroidContentURI.cpp
 	Common/File/DiskFree.h
 	Common/File/DiskFree.cpp
 	Common/File/Path.h
@@ -675,8 +614,6 @@ add_library(Common STATIC
 	Common/File/FileDescriptor.h
 	Common/GPU/DataFormat.h
 	Common/GPU/MiscTypes.h
-	Common/GPU/GPUBackendCommon.cpp
-	Common/GPU/GPUBackendCommon.h
 	Common/GPU/thin3d.cpp
 	Common/GPU/thin3d.h
 	Common/GPU/thin3d_create.h
@@ -695,11 +632,7 @@ add_library(Common STATIC
 	Common/GPU/OpenGL/gl3stub.h
 	Common/GPU/OpenGL/GLFeatures.cpp
 	Common/GPU/OpenGL/GLFeatures.h
-	Common/GPU/OpenGL/GLFrameData.cpp
-	Common/GPU/OpenGL/GLFrameData.h
 	Common/GPU/OpenGL/thin3d_gl.cpp
-	Common/GPU/OpenGL/GLMemory.cpp
-	Common/GPU/OpenGL/GLMemory.h
 	Common/GPU/OpenGL/GLRenderManager.cpp
 	Common/GPU/OpenGL/GLRenderManager.h
 	Common/GPU/OpenGL/GLQueueRunner.cpp
@@ -712,8 +645,6 @@ add_library(Common STATIC
 	Common/GPU/Vulkan/VulkanDebug.h
 	Common/GPU/Vulkan/VulkanContext.cpp
 	Common/GPU/Vulkan/VulkanContext.h
-	Common/GPU/Vulkan/VulkanDescSet.cpp
-	Common/GPU/Vulkan/VulkanDescSet.h
 	Common/GPU/Vulkan/VulkanFramebuffer.cpp
 	Common/GPU/Vulkan/VulkanFramebuffer.h
 	Common/GPU/Vulkan/VulkanImage.cpp
@@ -753,10 +684,6 @@ add_library(Common STATIC
 	Common/Net/HTTPClient.h
 	Common/Net/HTTPHeaders.cpp
 	Common/Net/HTTPHeaders.h
-	Common/Net/HTTPNaettRequest.cpp
-	Common/Net/HTTPNaettRequest.h
-	Common/Net/HTTPRequest.cpp
-	Common/Net/HTTPRequest.h
 	Common/Net/HTTPServer.cpp
 	Common/Net/HTTPServer.h
 	Common/Net/NetBuffer.cpp
@@ -781,20 +708,12 @@ add_library(Common STATIC
 	Common/Render/Text/draw_text.h
 	Common/Render/Text/draw_text_android.cpp
 	Common/Render/Text/draw_text_android.h
-	Common/Render/Text/draw_text_sdl.cpp
-	Common/Render/Text/draw_text_sdl.h
 	Common/Render/Text/draw_text_win.cpp
 	Common/Render/Text/draw_text_win.h
 	Common/Render/Text/draw_text_uwp.cpp
 	Common/Render/Text/draw_text_uwp.h
 	Common/System/Display.cpp
 	Common/System/Display.h
-	Common/System/System.h
-	Common/System/NativeApp.h
-	Common/System/Request.cpp
-	Common/System/Request.h
-	Common/System/OSD.cpp
-	Common/System/OSD.h
 	Common/Thread/Channel.h
 	Common/Thread/ParallelLoop.cpp
 	Common/Thread/ParallelLoop.h
@@ -813,8 +732,6 @@ add_library(Common STATIC
 	Common/UI/UI.h
 	Common/UI/Context.cpp
 	Common/UI/Context.h
-	Common/UI/IconCache.cpp
-	Common/UI/IconCache.h
 	Common/UI/UIScreen.cpp
 	Common/UI/UIScreen.h
 	Common/UI/Tween.cpp
@@ -852,10 +769,8 @@ add_library(Common STATIC
 	Common/MemArenaDarwin.cpp
 	Common/MemArenaPosix.cpp
 	Common/MemArenaWin32.cpp
-	Common/MemArenaHorizon.cpp
 	Common/MemArena.h
 	Common/MemoryUtil.cpp
-	Common/MemoryUtilHorizon.cpp
 	Common/MemoryUtil.h
 	Common/OSVersion.cpp
 	Common/OSVersion.h
@@ -901,11 +816,7 @@ if(USE_FFMPEG)
 					set(PLATFORM_ARCH "android/x86")
 				endif()
 			elseif(IOS)
-				if(IOS_PLATFORM STREQUAL "TVOS")
+				set(PLATFORM_ARCH "ios/universal")
-					set(PLATFORM_ARCH "tvos/arm64")
-				else()
-					set(PLATFORM_ARCH "ios/universal")
-				endif()
 			elseif(MACOSX)
 				set(PLATFORM_ARCH "macosx/universal")
 			elseif(LINUX)
@@ -919,8 +830,6 @@ if(USE_FFMPEG)
 					set(PLATFORM_ARCH "linux/mips32")
 				elseif(RISCV64)
 					set(PLATFORM_ARCH "linux/riscv64")
-				elseif(LOONGARCH64)
-					set(PLATFORM_ARCH "linux/loongarch64")
 				elseif(CMAKE_SIZEOF_VOID_P EQUAL 8)
 					set(PLATFORM_ARCH "linux/x86_64")
 				elseif(X86)
@@ -1001,7 +910,6 @@ endif()
 
 find_package(LIBZIP)
 if(LIBZIP_FOUND AND USE_SYSTEM_LIBZIP)
-	include_directories(${LIBZIP_INCLUDE_DIRS})
 	add_definitions(-DSHARED_LIBZIP)
 else()
 	add_library(libzip STATIC
@@ -1193,32 +1101,19 @@ else()
 	include_directories(ext/libpng17)
 endif()
 
-add_library(basis_universal STATIC
-	ext/basis_universal/basisu.h
-	ext/basis_universal/basisu_containers.h
-	ext/basis_universal/basisu_containers_impl.h
-	ext/basis_universal/basisu_file_headers.h
-	ext/basis_universal/basisu_transcoder.cpp
-	ext/basis_universal/basisu_transcoder.h
-	ext/basis_universal/basisu_transcoder_internal.h
-	ext/basis_universal/basisu_transcoder_tables_astc.inc
-	ext/basis_universal/basisu_transcoder_tables_astc_0_255.inc
-	ext/basis_universal/basisu_transcoder_tables_atc_55.inc
-	ext/basis_universal/basisu_transcoder_tables_atc_56.inc
-	ext/basis_universal/basisu_transcoder_tables_bc7_m5_alpha.inc
-	ext/basis_universal/basisu_transcoder_tables_bc7_m5_color.inc
-	ext/basis_universal/basisu_transcoder_tables_dxt1_5.inc
-	ext/basis_universal/basisu_transcoder_tables_dxt1_6.inc
-	ext/basis_universal/basisu_transcoder_tables_pvrtc2_45.inc
-	ext/basis_universal/basisu_transcoder_tables_pvrtc2_alpha_33.inc
-	ext/basis_universal/basisu_transcoder_uastc.h
-)
-set(BASISU_LIBRARIES basis_universal)
-
 set(nativeExtra)
 set(nativeExtraLibs)
 
 if(ANDROID)
+	set(nativeExtra ${nativeExtra}
+		Common/GL/GLInterface/EGLAndroid.cpp
+		Common/GL/GLInterface/EGLAndroid.h
+		Common/GL/GLInterface/EGL.cpp
+		Common/GL/GLInterface/EGL.h
+		Common/GL/GLInterface/GLInterface.cpp
+		Common/GL/GLInterfaceBase.h
+	)
+
 	set(NativeAppSource ${NativeAppSource}
 		android/jni/app-android.cpp
 		android/jni/AndroidJavaGLContext.cpp
@@ -1235,7 +1130,7 @@ if(ANDROID)
 		set(nativeExtraLibs ${nativeExtraLibs} openxr)
 	endif()
 	# No target
-elseif(IOS AND NOT LIBRETRO)
+elseif(IOS)
 	set(nativeExtra ${nativeExtra}
 		ios/main.mm
 		ios/AppDelegate.mm
@@ -1261,7 +1156,7 @@ elseif(IOS AND NOT LIBRETRO)
 		UI/DarwinFileSystemServices.h
 		Common/Battery/AppleBatteryClient.m
 	)
-
+    
 	set(nativeExtraLibs ${nativeExtraLibs} "-framework Foundation -framework MediaPlayer -framework AudioToolbox -framework CoreGraphics -framework QuartzCore -framework UIKit -framework GLKit -framework OpenAL -framework AVFoundation -framework CoreLocation -framework CoreVideo -framework CoreMedia -framework CoreServices" )
 	if(EXISTS "${CMAKE_IOS_SDK_ROOT}/System/Library/Frameworks/GameController.framework")
 		set(nativeExtraLibs ${nativeExtraLibs} "-weak_framework GameController")
@@ -1282,14 +1177,14 @@ elseif(IOS AND NOT LIBRETRO)
 	set_source_files_properties(UI/DarwinFileSystemServices.mm PROPERTIES COMPILE_FLAGS -fobjc-arc)
 	set_source_files_properties(Common/Battery/AppleBatteryClient.m PROPERTIES COMPILE_FLAGS -fobjc-arc)
 	set(TargetBin PPSSPP)
-elseif(IOS AND LIBRETRO)
-    set(nativeExtraLibs ${nativeExtraLibs} "-framework GLKit")
 elseif(USING_QT_UI)
 	set(CMAKE_AUTOMOC ON)
 	find_package(Qt5 COMPONENTS OpenGL Gui Core Multimedia)
 	list(APPEND NativeAppSource
 		Qt/QtMain.cpp
 		Qt/QtMain.h
+		Qt/QtHost.cpp
+		Qt/QtHost.h
 		Qt/mainwindow.cpp
 		Qt/mainwindow.h
 	)
@@ -1297,15 +1192,7 @@ elseif(USING_QT_UI)
 	if(USING_GLES2)
 		add_definitions(-DQT_OPENGL_ES -DQT_OPENGL_ES_2)
 	endif()
-	if(APPLE)
+
-		list(APPEND NativeAppSource
-			UI/DarwinFileSystemServices.mm
-			UI/DarwinFileSystemServices.h
-			Common/Battery/AppleBatteryClient.m)
-		set_source_files_properties(Common/Battery/AppleBatteryClient.m PROPERTIES COMPILE_FLAGS -fobjc-arc)
-		set_source_files_properties(UI/DarwinFileSystemServices.mm PROPERTIES COMPILE_FLAGS -fobjc-arc)
-		set(nativeExtraLibs ${nativeExtraLibs} ${COCOA_LIBRARY} ${QUARTZ_CORE_LIBRARY} ${IOKIT_LIBRARY})
-	endif()
 	include_directories(Qt)
 	include_directories(${CMAKE_CURRENT_BINARY_DIR})
 	set(nativeExtraLibs ${nativeExtraLibs} Qt5::OpenGL Qt5::Gui Qt5::Core Qt5::Multimedia)
@@ -1329,13 +1216,8 @@ elseif(WIN32)
 	else()
 		link_directories(dx9sdk/Lib/x86)
 	endif()
-elseif(LIBRETRO)
+elseif(TARGET SDL2::SDL2)
-else()
+	set(TargetBin PPSSPPSDL)
-	if(GOLD)
-		set(TargetBin PPSSPPGold)
-	else()
-		set(TargetBin PPSSPPSDL)
-	endif()
 	# Require SDL
 	add_definitions(-DSDL)
 	set(nativeExtra ${nativeExtra}
@@ -1349,55 +1231,17 @@ else()
 			SDL/SDLVulkanGraphicsContext.cpp
 		)
 	endif()
-	if(SDL2_ttf_FOUND OR
+	set(nativeExtraLibs ${nativeExtraLibs} SDL2::SDL2)
-		(SDL2_ttf_PKGCONFIG_FOUND AND
-		 SDL2_ttf_PKGCONFIG_VERSION VERSION_GREATER_EQUAL "2.0.18"))
-		add_definitions(-DUSE_SDL2_TTF)
-		if(FONTCONFIG_FOUND)
-			add_definitions(-DUSE_SDL2_TTF_FONTCONFIG)
-			set(nativeExtraLibs ${nativeExtraLibs} Fontconfig::Fontconfig)
-		endif()
-	elseif(SDL2_ttf_PKGCONFIG_FOUND)
-		message(WARNING "Found SDL2_ttf <2.0.18 - this is too old, falling back to atlas")
-	endif()
-	if(SDL2_ttf_FOUND)
-		set(nativeExtraLibs ${nativeExtraLibs} SDL2_ttf::SDL2_ttf)
-	elseif(SDL2_ttf_PKGCONFIG_FOUND)
-		set(nativeExtraLibs ${nativeExtraLibs} PkgConfig::SDL2_ttf_PKGCONFIG)
-	endif()
 	if(APPLE)
-		set(nativeExtra ${nativeExtra} 
+		set(nativeExtra ${nativeExtra} SDL/SDLMain.h SDL/SDLMain.mm SDL/SDLCocoaMetalLayer.h SDL/SDLCocoaMetalLayer.mm UI/DarwinFileSystemServices.mm UI/DarwinFileSystemServices.h Common/Battery/AppleBatteryClient.m)
-		SDL/SDLMain.h 
-		SDL/SDLMain.mm 
-		SDL/SDLCocoaMetalLayer.h 
-		SDL/SDLCocoaMetalLayer.mm 
-		SDL/CocoaBarItems.mm
-		SDL/CocoaBarItems.h
-		SDL/PPSSPPAboutViewController.m
-		SDL/PPSSPPAboutViewController.h
-		UI/DarwinFileSystemServices.mm 
-		UI/DarwinFileSystemServices.h 
-		Common/Battery/AppleBatteryClient.m 
-		UI/PSPNSApplicationDelegate.mm 
-		UI/PSPNSApplicationDelegate.h)
-
         set_source_files_properties(UI/DarwinFileSystemServices.mm PROPERTIES COMPILE_FLAGS -fobjc-arc)
-		set_source_files_properties(UI/PSPNSApplicationDelegate.mm PROPERTIES COMPILE_FLAGS -fobjc-arc)
-		set_source_files_properties(SDL/CocoaBarItems.mm PROPERTIES COMPILE_FLAGS -fobjc-arc)
-		set_source_files_properties(SDL/PPSSPPAboutViewController.m PROPERTIES COMPILE_FLAGS -fobjc-arc)
 		set_source_files_properties(Common/Battery/AppleBatteryClient.m PROPERTIES COMPILE_FLAGS -fobjc-arc)
 		set(nativeExtraLibs ${nativeExtraLibs} ${COCOA_LIBRARY} ${QUARTZ_CORE_LIBRARY} ${IOKIT_LIBRARY})
-
-		if(USE_SYSTEM_LIBSDL2)
-			set(nativeExtraLibs ${nativeExtraLibs} SDL2::SDL2)
-		else()
-			set(nativeExtraLibs ${nativeExtraLibs} ${SDL2Fwk})
-		endif()
 	elseif(USING_EGL)
-		set(nativeExtraLibs ${nativeExtraLibs} pthread SDL2::SDL2)
+		set(nativeExtraLibs ${nativeExtraLibs} pthread)
-	else()
-		set(nativeExtraLibs ${nativeExtraLibs} SDL2::SDL2)
 	endif()
+elseif(NOT LIBRETRO)
+	message(FATAL_ERROR "Could not find SDL2. Failing.")
 endif()
 
 if(WIN32)
@@ -1411,13 +1255,6 @@ if(WIN32)
 	target_link_libraries(Common winmm d3d9 dsound)
 endif()
 
-if(NOT LIBRETRO)
-	list(APPEND NativeAppSource
-		UI/AudioCommon.h
-		UI/AudioCommon.cpp
-	)
-endif()
-
 list(APPEND NativeAppSource
 	android/jni/TestRunner.cpp
 	UI/DiscordIntegration.cpp
@@ -1426,8 +1263,6 @@ list(APPEND NativeAppSource
 	UI/BackgroundAudio.cpp
 	UI/ChatScreen.h
 	UI/ChatScreen.cpp
-	UI/DebugOverlay.cpp
-	UI/DebugOverlay.h
 	UI/DevScreens.cpp
 	UI/DevScreens.h
 	UI/DisplayLayoutScreen.cpp
@@ -1442,8 +1277,6 @@ list(APPEND NativeAppSource
 	UI/MiscScreens.cpp
 	UI/PauseScreen.h
 	UI/PauseScreen.cpp
-	UI/TabbedDialogScreen.h
-	UI/TabbedDialogScreen.cpp
 	UI/GameScreen.h
 	UI/GameScreen.cpp
 	UI/GameSettingsScreen.h
@@ -1480,12 +1313,10 @@ list(APPEND NativeAppSource
 	UI/MemStickScreen.cpp
 	UI/ProfilerDraw.h
 	UI/ProfilerDraw.cpp
-	UI/CustomButtonMappingScreen.h
+	UI/ComboKeyMappingScreen.h
-	UI/CustomButtonMappingScreen.cpp
+	UI/ComboKeyMappingScreen.cpp
 	UI/Theme.h
 	UI/Theme.cpp
-	UI/RetroAchievementScreens.cpp
-	UI/RetroAchievementScreens.h
 )
 
 if(ANDROID)
@@ -1519,11 +1350,11 @@ if(LINUX AND NOT ANDROID)
 endif()
 
 set(ATOMIC_LIB)
-if(ANDROID OR (LINUX AND ARM_DEVICE) OR (LINUX AND RISCV64) OR (LINUX AND LOONGARCH64))
+if(ANDROID OR (LINUX AND ARM_DEVICE))
 	set(ATOMIC_LIB atomic)
 endif()
 
-target_link_libraries(native ${LIBZIP_LIBRARY} ${PNG_LIBRARIES} ${BASISU_LIBRARIES} ${ZLIB_LIBRARY} vma gason udis86 ${RT_LIB} ${nativeExtraLibs} ${ATOMIC_LIB} Common)
+target_link_libraries(native ${LIBZIP_LIBRARY} ${PNG_LIBRARIES} ${ZLIB_LIBRARY} vma gason udis86 ${RT_LIB} ${nativeExtraLibs} ${ATOMIC_LIB} Common)
 if(TARGET Ext::GLEW)
 	target_link_libraries(native Ext::GLEW)
 endif()
@@ -1576,8 +1407,6 @@ set(CoreExtra)
 set(CoreExtraLibs)
 
 set(CoreExtra ${CoreExtra}
-	Core/MIPS/IR/IRAnalysis.cpp
-	Core/MIPS/IR/IRAnalysis.h
 	Core/MIPS/IR/IRCompALU.cpp
 	Core/MIPS/IR/IRCompBranch.cpp
 	Core/MIPS/IR/IRCompFPU.cpp
@@ -1591,8 +1420,6 @@ set(CoreExtra ${CoreExtra}
 	Core/MIPS/IR/IRInterpreter.h
 	Core/MIPS/IR/IRJit.cpp
 	Core/MIPS/IR/IRJit.h
-	Core/MIPS/IR/IRNativeCommon.cpp
-	Core/MIPS/IR/IRNativeCommon.h
 	Core/MIPS/IR/IRPassSimplify.cpp
 	Core/MIPS/IR/IRPassSimplify.h
 	Core/MIPS/IR/IRRegCache.cpp
@@ -1632,17 +1459,6 @@ list(APPEND CoreExtra
 	Core/MIPS/ARM64/Arm64RegCache.h
 	Core/MIPS/ARM64/Arm64RegCacheFPU.cpp
 	Core/MIPS/ARM64/Arm64RegCacheFPU.h
-	Core/MIPS/ARM64/Arm64IRAsm.cpp
-	Core/MIPS/ARM64/Arm64IRCompALU.cpp
-	Core/MIPS/ARM64/Arm64IRCompBranch.cpp
-	Core/MIPS/ARM64/Arm64IRCompFPU.cpp
-	Core/MIPS/ARM64/Arm64IRCompLoadStore.cpp
-	Core/MIPS/ARM64/Arm64IRCompSystem.cpp
-	Core/MIPS/ARM64/Arm64IRCompVec.cpp
-	Core/MIPS/ARM64/Arm64IRJit.cpp
-	Core/MIPS/ARM64/Arm64IRJit.h
-	Core/MIPS/ARM64/Arm64IRRegCache.cpp
-	Core/MIPS/ARM64/Arm64IRRegCache.h
 	GPU/Common/VertexDecoderArm64.cpp
 	Core/Util/DisArm64.cpp
 )
@@ -1663,17 +1479,6 @@ list(APPEND CoreExtra
 	Core/MIPS/x86/RegCache.h
 	Core/MIPS/x86/RegCacheFPU.cpp
 	Core/MIPS/x86/RegCacheFPU.h
-	Core/MIPS/x86/X64IRAsm.cpp
-	Core/MIPS/x86/X64IRCompALU.cpp
-	Core/MIPS/x86/X64IRCompBranch.cpp
-	Core/MIPS/x86/X64IRCompFPU.cpp
-	Core/MIPS/x86/X64IRCompLoadStore.cpp
-	Core/MIPS/x86/X64IRCompSystem.cpp
-	Core/MIPS/x86/X64IRCompVec.cpp
-	Core/MIPS/x86/X64IRJit.cpp
-	Core/MIPS/x86/X64IRJit.h
-	Core/MIPS/x86/X64IRRegCache.cpp
-	Core/MIPS/x86/X64IRRegCache.h
 	GPU/Common/VertexDecoderX86.cpp
 	GPU/Software/DrawPixelX86.cpp
 	GPU/Software/SamplerX86.cpp
@@ -1684,21 +1489,6 @@ list(APPEND CoreExtra
 	Core/MIPS/MIPS/MipsJit.h
 )
 
-list(APPEND CoreExtra
-	Core/MIPS/RiscV/RiscVAsm.cpp
-	Core/MIPS/RiscV/RiscVCompALU.cpp
-	Core/MIPS/RiscV/RiscVCompBranch.cpp
-	Core/MIPS/RiscV/RiscVCompFPU.cpp
-	Core/MIPS/RiscV/RiscVCompLoadStore.cpp
-	Core/MIPS/RiscV/RiscVCompSystem.cpp
-	Core/MIPS/RiscV/RiscVCompVec.cpp
-	Core/MIPS/RiscV/RiscVJit.cpp
-	Core/MIPS/RiscV/RiscVJit.h
-	Core/MIPS/RiscV/RiscVRegCache.cpp
-	Core/MIPS/RiscV/RiscVRegCache.h
-	GPU/Common/VertexDecoderRiscV.cpp
-)
-
 if(NOT MOBILE_DEVICE)
 	set(CoreExtra ${CoreExtra}
 		Core/AVIDump.cpp
@@ -1709,7 +1499,7 @@ if(NOT MOBILE_DEVICE)
 endif()
 
 set(GPU_GLES
-	GPU/GLES/StencilBufferGLES.cpp
+	GPU/GLES/DepthBufferGLES.cpp
 	GPU/GLES/GPU_GLES.cpp
 	GPU/GLES/GPU_GLES.h
 	GPU/GLES/FragmentTestCacheGLES.cpp
@@ -1790,7 +1580,6 @@ set(GPU_SOURCES
 	${GPU_NEON}
 	GPU/Common/Draw2D.cpp
 	GPU/Common/Draw2D.h
-	GPU/Common/DepthBufferCommon.cpp
 	GPU/Common/TextureShaderCommon.cpp
 	GPU/Common/TextureShaderCommon.h
 	GPU/Common/DepalettizeShaderCommon.cpp
@@ -1839,10 +1628,6 @@ set(GPU_SOURCES
 	GPU/Common/TextureScalerCommon.h
 	GPU/Common/PostShader.cpp
 	GPU/Common/PostShader.h
-	GPU/Common/TextureReplacer.cpp
-	GPU/Common/TextureReplacer.h
-	GPU/Common/ReplacedTexture.cpp
-	GPU/Common/ReplacedTexture.h
 	GPU/Debugger/Breakpoints.cpp
 	GPU/Debugger/Breakpoints.h
 	GPU/Debugger/Debugger.cpp
@@ -1865,8 +1650,6 @@ set(GPU_SOURCES
 	GPU/GPU.h
 	GPU/GPUCommon.cpp
 	GPU/GPUCommon.h
-	GPU/GPUCommonHW.cpp
-	GPU/GPUCommonHW.h
 	GPU/GPUState.cpp
 	GPU/GPUState.h
 	GPU/Math3D.cpp
@@ -1902,8 +1685,6 @@ add_library(${CoreLibName} ${CoreLinkType}
 	${CommonJIT}
 	Core/Config.cpp
 	Core/Config.h
-	Core/ConfigSettings.cpp
-	Core/ConfigSettings.h
 	Core/ConfigValues.h
 	Core/ControlMapper.cpp
 	Core/ControlMapper.h
@@ -1916,8 +1697,6 @@ add_library(${CoreLibName} ${CoreLinkType}
 	Core/CoreTiming.h
 	Core/CwCheat.cpp
 	Core/CwCheat.h
-	Core/FrameTiming.cpp
-	Core/FrameTiming.h
 	Core/HDRemaster.cpp
 	Core/HDRemaster.h
 	Core/Instance.cpp
@@ -1926,8 +1705,6 @@ add_library(${CoreLibName} ${CoreLinkType}
 	Core/KeyMap.h
 	Core/KeyMapDefaults.cpp
 	Core/KeyMapDefaults.h
-	Core/RetroAchievements.h
-	Core/RetroAchievements.cpp
 	Core/ThreadEventQueue.h
 	Core/TiltEventProcessor.h
 	Core/TiltEventProcessor.cpp
@@ -1954,8 +1731,6 @@ add_library(${CoreLibName} ${CoreLinkType}
 	Core/Debugger/WebSocket/GameBroadcaster.h
 	Core/Debugger/WebSocket/GameSubscriber.cpp
 	Core/Debugger/WebSocket/GameSubscriber.h
-	Core/Debugger/WebSocket/ClientConfigSubscriber.cpp
-	Core/Debugger/WebSocket/ClientConfigSubscriber.h
 	Core/Debugger/WebSocket/GPUBufferSubscriber.cpp
 	Core/Debugger/WebSocket/GPUBufferSubscriber.h
 	Core/Debugger/WebSocket/GPURecordSubscriber.cpp
@@ -2201,6 +1976,8 @@ add_library(${CoreLibName} ${CoreLinkType}
 	Core/HW/SasReverb.h
 	Core/HW/StereoResampler.cpp
 	Core/HW/StereoResampler.h
+	Core/Host.cpp
+	Core/Host.h
 	Core/Loaders.cpp
 	Core/Loaders.h
 	Core/FileLoaders/CachingFileLoader.cpp
@@ -2237,8 +2014,6 @@ add_library(${CoreLibName} ${CoreLinkType}
 	Core/MIPS/MIPSTables.h
 	Core/MIPS/MIPSVFPUUtils.cpp
 	Core/MIPS/MIPSVFPUUtils.h
-	Core/MIPS/MIPSVFPUFallbacks.cpp
-	Core/MIPS/MIPSVFPUFallbacks.h
 	Core/MIPS/MIPSAsm.cpp
 	Core/MIPS/MIPSAsm.h
 	Core/MemFault.cpp
@@ -2259,14 +2034,14 @@ add_library(${CoreLibName} ${CoreLinkType}
 	Core/Screenshot.h
 	Core/System.cpp
 	Core/System.h
+	Core/TextureReplacer.cpp
+	Core/TextureReplacer.h
 	Core/ThreadPools.cpp
 	Core/ThreadPools.h
 	Core/Util/AudioFormat.cpp
 	Core/Util/AudioFormat.h
 	Core/Util/GameManager.cpp
 	Core/Util/GameManager.h
-	Core/Util/GameDB.cpp
-	Core/Util/GameDB.h
 	Core/Util/PortManager.cpp
 	Core/Util/PortManager.h
 	Core/Util/BlockAllocator.cpp
@@ -2325,18 +2100,9 @@ else()
 	include_directories(ext/zstd/lib)
 endif()
 
-include_directories(ext/libchdr/include)
+target_link_libraries(${CoreLibName} Common native kirk cityhash sfmt19937 xbrz xxhash ${GlslangLibs}
-
-target_link_libraries(${CoreLibName} Common native chdr kirk cityhash sfmt19937 xbrz xxhash rcheevos ${GlslangLibs}
 	${CoreExtraLibs} ${OPENGL_LIBRARIES} ${X11_LIBRARIES} ${CMAKE_DL_LIBS})
 
-if(NOT HTTPS_NOT_AVAILABLE)
-	target_link_libraries(${CoreLibName} naett)
-	if(WIN32)
-		target_link_libraries(${CoreLibName} winhttp)
-	endif()
-endif()
-
 target_compile_features(${CoreLibName} PUBLIC cxx_std_17)
 
 if(FFmpeg_FOUND)
@@ -2483,12 +2249,10 @@ set(WindowsFiles
 	Windows/Debugger/Debugger_MemoryDlg.h
 	Windows/Debugger/Debugger_Lists.cpp
 	Windows/Debugger/Debugger_Lists.h
+	Windows/Debugger/Debugger_SymbolMap.h
 	Windows/Debugger/Debugger_VFPUDlg.cpp
 	Windows/Debugger/Debugger_VFPUDlg.h
-	Windows/Debugger/WatchItemWindow.cpp
+	Windows/Debugger/SimpleELF.h
-	Windows/Debugger/WatchItemWindow.h
-	Windows/Debugger/EditSymbolsWindow.cpp
-	Windows/Debugger/EditSymbolsWindow.h
 	Windows/GEDebugger/CtrlDisplayListView.cpp
 	Windows/GEDebugger/SimpleGLWindow.cpp
 	Windows/GEDebugger/TabState.cpp
@@ -2532,13 +2296,6 @@ set(WindowsFiles
 	Windows/W32Util/ShellUtil.h
 	Windows/W32Util/TabControl.cpp
 	Windows/W32Util/TabControl.h
-	Windows/W32Util/IatHook.h
-	Windows/W32Util/ContextMenu.h
-	Windows/W32Util/ContextMenu.cpp
-	Windows/W32Util/DarkMode.h
-	Windows/W32Util/DarkMode.cpp
-	Windows/W32Util/UAHMenuBar.h
-	Windows/W32Util/UAHMenuBar.cpp
 	Windows/WindowsHost.cpp
 	Windows/WindowsHost.h
 	Windows/MainWindow.cpp
@@ -2562,7 +2319,7 @@ set(WindowsFiles
 list(APPEND LinkCommon ${CoreLibName} ${CMAKE_THREAD_LIBS_INIT})
 
 if(WIN32)
-	list(APPEND LinkCommon kernel32 user32 gdi32 shell32 comctl32 dsound xinput d3d9 winmm dinput8 ole32 winspool ksuser mf uxtheme mfplat mfreadwrite mfuuid shlwapi)
+	list(APPEND LinkCommon kernel32 user32 gdi32 shell32 comctl32 dsound xinput d3d9 winmm dinput8 ole32 winspool ksuser mf mfplat mfreadwrite mfuuid shlwapi)
 	#setup_target_project(${TargetBin} Windows)
 	list(APPEND NativeAppSource ${WindowsFiles})
 endif()
@@ -2581,7 +2338,6 @@ set(NativeAssets
 	assets/lang
 	assets/shaders
 	assets/themes
-	assets/vfpu
 	assets/Roboto-Condensed.ttf
 	assets/7z.png
 	assets/compat.ini
@@ -2605,8 +2361,8 @@ set(NativeAssets
 if(HEADLESS)
 	set(HeadlessSource
 		headless/Headless.cpp
-		headless/HeadlessHost.cpp
+		headless/StubHost.cpp
-		headless/HeadlessHost.h
+		headless/StubHost.h
 		headless/Compare.cpp
 		headless/Compare.h
 		headless/SDLHeadlessHost.cpp
@@ -2644,7 +2400,6 @@ if(UNITTEST)
 		unittest/TestIRPassSimplify.cpp
 		unittest/TestX64Emitter.cpp
 		unittest/TestVertexJit.cpp
-		unittest/TestVFS.cpp
 		unittest/TestRiscVEmitter.cpp
 		unittest/TestSoftwareGPUJit.cpp
 		unittest/TestThreadManager.cpp
@@ -2678,18 +2433,11 @@ endif()
 if(TargetBin)
 	if(APPLE)
 		if(NOT IOS)
-			if(GOLD)
+			set(ICON_PATH_ABS ${CMAKE_CURRENT_SOURCE_DIR}/icons/ppsspp.icns)
-				set(ICON_PATH_ABS ${CMAKE_CURRENT_SOURCE_DIR}/icons/ppsspp_gold.icns)
+			set(MACOSX_BUNDLE_ICON_FILE ppsspp.icns)
-				set(MACOSX_BUNDLE_ICON_FILE ppsspp_gold.icns)
-				set(MACOSX_BUNDLE_BUNDLE_NAME "PPSSPP Gold")
-				set(MACOSX_BUNDLE_GUI_IDENTIFIER org.ppsspp.ppssppgold)
-			else()
-				set(ICON_PATH_ABS ${CMAKE_CURRENT_SOURCE_DIR}/icons/ppsspp.icns)
-				set(MACOSX_BUNDLE_ICON_FILE ppsspp.icns)
-				set(MACOSX_BUNDLE_BUNDLE_NAME "PPSSPP")
-				set(MACOSX_BUNDLE_GUI_IDENTIFIER org.ppsspp.ppsspp)
-			endif()
 			set_source_files_properties(${ICON_PATH_ABS} PROPERTIES MACOSX_PACKAGE_LOCATION "Resources")
+			set(MACOSX_BUNDLE_BUNDLE_NAME PPSSPP)
+			set(MACOSX_BUNDLE_GUI_IDENTIFIER org.ppsspp.ppsspp)
 		endif()
 
 		# TODO: there must a native way to copy these.
@@ -2700,7 +2448,6 @@ if(TargetBin)
 		file(GLOB_RECURSE SHADER_FILES assets/shaders/*)
 		file(GLOB_RECURSE THEME_FILE assets/themes/*)
 		file(GLOB_RECURSE DEBUGGER_FILES assets/debugger/*)
-		file(GLOB_RECURSE VFPU_FILES assets/vfpu/*)
 
 		if(NOT IOS)
 			set_source_files_properties(${BigFontAssets} PROPERTIES MACOSX_PACKAGE_LOCATION "Resources/assets")
@@ -2710,7 +2457,6 @@ if(TargetBin)
 			set_source_files_properties(${SHADER_FILES} PROPERTIES MACOSX_PACKAGE_LOCATION "Resources/assets/shaders")
 			set_source_files_properties(${THEME_FILE} PROPERTIES MACOSX_PACKAGE_LOCATION "Resources/assets/themes")
 			set_source_files_properties(${DEBUGGER_FILES} PROPERTIES MACOSX_PACKAGE_LOCATION "Resources/assets/debugger")
-			set_source_files_properties(${VFPU_FILES} PROPERTIES MACOSX_PACKAGE_LOCATION "Resources/assets/vfpu")
 		endif()
 
 		if(IOS)
@@ -2718,11 +2464,11 @@ if(TargetBin)
 			file(INSTALL "${CMAKE_SOURCE_DIR}/ext/vulkan/iOS/Frameworks/libMoltenVK.dylib" DESTINATION "${CMAKE_BINARY_DIR}/PPSSPP.app/Frameworks/")
 		else()
 			add_executable(${TargetBin} MACOSX_BUNDLE ${ICON_PATH_ABS} ${NativeAssets} ${BigFontAssets} ${SHADER_FILES} ${THEME_FILE} ${DEBUGGER_FILES} ${FLASH0_FILES} ${LANG_FILES} ${NativeAppSource})
-			file(INSTALL "${CMAKE_SOURCE_DIR}/ext/vulkan/macOS/Frameworks/libMoltenVK.dylib" DESTINATION "${CMAKE_BINARY_DIR}/${TargetBin}.app/Contents/Frameworks/")
+			file(INSTALL "${CMAKE_SOURCE_DIR}/ext/vulkan/macOS/Frameworks/libMoltenVK.dylib" DESTINATION "${CMAKE_BINARY_DIR}/PPSSPPSDL.app/Contents/Frameworks/")
-			if(USING_QT_UI)
+			if(TARGET SDL2::SDL2 AND NOT USING_QT_UI)
+				add_custom_command(TARGET ${TargetBin} POST_BUILD COMMAND /bin/bash "${CMAKE_SOURCE_DIR}/SDL/macbundle.sh" "${CMAKE_BINARY_DIR}/PPSSPPSDL.app")
+			elseif(USING_QT_UI)
 				add_custom_command(TARGET ${TargetBin} POST_BUILD COMMAND /bin/bash "${CMAKE_SOURCE_DIR}/Qt/macbundle.sh" "${CMAKE_BINARY_DIR}/PPSSPPQt.app")
-			elseif(NOT USE_SYSTEM_LIBSDL2)
-				add_custom_command(TARGET ${TargetBin} POST_BUILD COMMAND /bin/bash "${CMAKE_SOURCE_DIR}/SDL/macbundle.sh" "${CMAKE_BINARY_DIR}/${TargetBin}.app" "${TargetBin}")
 			endif()
 		endif()
 	elseif(WIN32)
@@ -2746,7 +2492,7 @@ if(NOT ANDROID)
 	file(INSTALL assets/flash0 DESTINATION assets)
 endif()
 # packaging and code signing
-if(IOS AND NOT LIBRETRO)
+if(IOS)
 	set(DEPLOYMENT_TARGET 11.0)
 	file(GLOB IOSAssets ios/assets/*.png)
 	list(REMOVE_ITEM IOSAssets ${CMAKE_CURRENT_SOURCE_DIR}/ios/assets/Default-568h@2x.png)
@@ -2787,18 +2533,6 @@ if(IOS AND NOT LIBRETRO)
 	)
 endif()
 
-if(MACOSX AND NOT IOS)
-	if(GOLD)
-		set_target_properties(${TargetBin} PROPERTIES
-			MACOSX_BUNDLE_INFO_PLIST "${CMAKE_CURRENT_SOURCE_DIR}/macOS/InfoGold.plist"
-		)
-	else()
-		set_target_properties(${TargetBin} PROPERTIES
-			MACOSX_BUNDLE_INFO_PLIST "${CMAKE_CURRENT_SOURCE_DIR}/macOS/Info.plist"
-		)
-	endif()
-endif()
-
 if(UNIX AND NOT ANDROID AND NOT APPLE)
 	configure_file(
 		"${CMAKE_SOURCE_DIR}/ppsspp.desktop.in"

Common/Arm64Emitter.cpp

@@ -315,14 +315,6 @@ const u8* ARM64XEmitter::AlignCodePage()
 	return m_code;
 }
 
-const u8 *ARM64XEmitter::NopAlignCode16() {
-	int bytes = ((-(intptr_t)m_code) & 15);
-	for (int i = 0; i < bytes / 4; i++) {
-		Write32(0xD503201F); // official nop instruction
-	}
-	return m_code;
-}
-
 void ARM64XEmitter::FlushIcache()
 {
 	FlushIcacheSection(m_lastCacheFlushEnd, m_code);
@@ -514,7 +506,7 @@ void ARM64XEmitter::EncodeTestBranchInst(u32 op, ARM64Reg Rt, u8 bits, const voi
 
 	distance >>= 2;
 
-	_assert_msg_(distance >= -0x2000 && distance <= 0x1FFF, "%s: Received too large distance: %llx", __FUNCTION__, distance);
+	_assert_msg_(distance >= -0x1FFF && distance < 0x1FFF, "%s: Received too large distance: %llx", __FUNCTION__, distance);
 
 	Rt = DecodeReg(Rt);
 	Write32((b64Bit << 31) | (0x36 << 24) | (op << 24) | \
@@ -2128,13 +2120,6 @@ void ARM64FloatEmitter::EmitCopy(bool Q, u32 op, u32 imm5, u32 imm4, ARM64Reg Rd
 	        (1 << 10) | (Rn << 5) | Rd);
 }
 
-void ARM64FloatEmitter::EmitScalarPairwise(bool U, u32 size, u32 opcode, ARM64Reg Rd, ARM64Reg Rn) {
-	Rd = DecodeReg(Rd);
-	Rn = DecodeReg(Rn);
-
-	Write32((1 << 30) | (U << 29) | (0b111100011 << 20) | (size << 22) | (opcode << 12) | (1 << 11) | (Rn << 5) | Rd);
-}
-
 void ARM64FloatEmitter::Emit2RegMisc(bool Q, bool U, u32 size, u32 opcode, ARM64Reg Rd, ARM64Reg Rn)
 {
 	_assert_msg_(!IsSingle(Rd), "%s doesn't support singles!", __FUNCTION__);
@@ -2234,45 +2219,6 @@ void ARM64FloatEmitter::FCVTU(ARM64Reg Rd, ARM64Reg Rn, RoundingMode round) {
 	EmitConvertScalarToInt(Rd, Rn, round, true);
 }
 
-void ARM64FloatEmitter::FCVTZS(ARM64Reg Rd, ARM64Reg Rn, int scale) {
-	if (IsScalar(Rd)) {
-		int imm = (IsDouble(Rn) ? 64 : 32) * 2 - scale;
-		Rd = DecodeReg(Rd);
-		Rn = DecodeReg(Rn);
-
-		Write32((1 << 30) | (0 << 29) | (0x1F << 24) | (imm << 16) | (0x1F << 11) | (1 << 10) | (Rn << 5) | Rd);
-	} else {
-		bool sf = Is64Bit(Rd);
-		u32 type = 0;
-		if (IsDouble(Rd))
-			type = 1;
-		int rmode = 3;
-		int opcode = 0;
-
-		Write32((sf << 31) | (0 << 29) | (0x1E << 24) | (type << 22) | (rmode << 19) | (opcode << 16) | (scale << 10) | (Rn << 5) | Rd);
-
-	}
-}
-
-void ARM64FloatEmitter::FCVTZU(ARM64Reg Rd, ARM64Reg Rn, int scale) {
-	if (IsScalar(Rd)) {
-		int imm = (IsDouble(Rn) ? 64 : 32) * 2 - scale;
-		Rd = DecodeReg(Rd);
-		Rn = DecodeReg(Rn);
-
-		Write32((1 << 30) | (1 << 29) | (0x1F << 24) | (imm << 16) | (0x1F << 11) | (1 << 10) | (Rn << 5) | Rd);
-	} else {
-		bool sf = Is64Bit(Rd);
-		u32 type = 0;
-		if (IsDouble(Rd))
-			type = 1;
-		int rmode = 3;
-		int opcode = 1;
-
-		Write32((sf << 31) | (0 << 29) | (0x1E << 24) | (type << 22) | (rmode << 19) | (opcode << 16) | (scale << 10) | (Rn << 5) | Rd);
-	}
-}
-
 void ARM64FloatEmitter::EmitConversion2(bool sf, bool S, bool direction, u32 type, u32 rmode, u32 opcode, int scale, ARM64Reg Rd, ARM64Reg Rn)
 {
 	Rd = DecodeReg(Rd);
@@ -2307,17 +2253,6 @@ void ARM64FloatEmitter::EmitCondSelect(bool M, bool S, CCFlags cond, ARM64Reg Rd
 	        (cond << 12) | (3 << 10) | (Rn << 5) | Rd);
 }
 
-void ARM64FloatEmitter::EmitCondCompare(bool M, bool S, CCFlags cond, int op, u8 nzcv, ARM64Reg Rn, ARM64Reg Rm) {
-	_assert_msg_(!IsQuad(Rn), "%s doesn't support vector!", __FUNCTION__);
-	bool is_double = IsDouble(Rn);
-
-	Rn = DecodeReg(Rn);
-	Rm = DecodeReg(Rm);
-
-	Write32((M << 31) | (S << 29) | (0xF1 << 21) | (is_double << 22) | (Rm << 16) | \
-		(cond << 12) | (1 << 10) | (Rn << 5) | (op << 4) | nzcv);
-}
-
 void ARM64FloatEmitter::EmitPermute(u32 size, u32 op, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
 {
 	_assert_msg_(!IsSingle(Rd), "%s doesn't support singles!", __FUNCTION__);
@@ -2963,22 +2898,6 @@ void ARM64FloatEmitter::FSQRT(ARM64Reg Rd, ARM64Reg Rn)
 	EmitScalar1Source(0, 0, IsDouble(Rd), 3, Rd, Rn);
 }
 
-// Scalar - pairwise
-void ARM64FloatEmitter::FADDP(ARM64Reg Rd, ARM64Reg Rn) {
-	EmitScalarPairwise(1, IsDouble(Rd), 0b01101, Rd, Rn);
-}
-void ARM64FloatEmitter::FMAXP(ARM64Reg Rd, ARM64Reg Rn) {
-	EmitScalarPairwise(1, IsDouble(Rd), 0b01111, Rd, Rn);
-}
-void ARM64FloatEmitter::FMINP(ARM64Reg Rd, ARM64Reg Rn) {
-	EmitScalarPairwise(1, IsDouble(Rd) ? 3 : 2, 0b01111, Rd, Rn);
-}
-void ARM64FloatEmitter::FMAXNMP(ARM64Reg Rd, ARM64Reg Rn) {
-	EmitScalarPairwise(1, IsDouble(Rd), 0b01100, Rd, Rn);
-}
-void ARM64FloatEmitter::FMINNMP(ARM64Reg Rd, ARM64Reg Rn) {
-	EmitScalarPairwise(1, IsDouble(Rd) ? 3 : 2, 0b01100, Rd, Rn);
-}
 
 // Scalar - 2 Source
 void ARM64FloatEmitter::FADD(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
@@ -3061,12 +2980,6 @@ void ARM64FloatEmitter::BSL(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
 {
 	EmitThreeSame(1, 1, 3, Rd, Rn, Rm);
 }
-void ARM64FloatEmitter::BIT(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-	EmitThreeSame(1, 2, 3, Rd, Rn, Rm);
-}
-void ARM64FloatEmitter::BIF(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-	EmitThreeSame(1, 3, 3, Rd, Rn, Rm);
-}
 void ARM64FloatEmitter::DUP(u8 size, ARM64Reg Rd, ARM64Reg Rn, u8 index)
 {
 	u32 imm5 = 0;
@@ -3102,9 +3015,6 @@ void ARM64FloatEmitter::FADD(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
 {
 	EmitThreeSame(0, size >> 6, 0x1A, Rd, Rn, Rm);
 }
-void ARM64FloatEmitter::FADDP(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-	EmitThreeSame(1, size >> 6, 0x1A, Rd, Rn, Rm);
-}
 void ARM64FloatEmitter::FMAX(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
 {
 	EmitThreeSame(0, size >> 6, 0x1E, Rd, Rn, Rm);
@@ -3137,14 +3047,6 @@ void ARM64FloatEmitter::FCVTZU(u8 size, ARM64Reg Rd, ARM64Reg Rn)
 {
 	Emit2RegMisc(IsQuad(Rd), 1, 2 | (size >> 6), 0x1B, Rd, Rn);
 }
-void ARM64FloatEmitter::FCVTZS(u8 size, ARM64Reg Rd, ARM64Reg Rn, int scale) {
-	int imm = size * 2 - scale;
-	EmitShiftImm(IsQuad(Rd), false, imm >> 3, imm & 7, 0x1F, Rd, Rn);
-}
-void ARM64FloatEmitter::FCVTZU(u8 size, ARM64Reg Rd, ARM64Reg Rn, int scale) {
-	int imm = size * 2 - scale;
-	EmitShiftImm(IsQuad(Rd), true, imm >> 3, imm & 7, 0x1F, Rd, Rn);
-}
 void ARM64FloatEmitter::FDIV(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
 {
 	EmitThreeSame(1, size >> 6, 0x1F, Rd, Rn, Rm);
@@ -3248,61 +3150,6 @@ void ARM64FloatEmitter::XTN2(u8 dest_size, ARM64Reg Rd, ARM64Reg Rn)
 	Emit2RegMisc(true, 0, dest_size >> 4, 0x12, Rd, Rn);
 }
 
-void ARM64FloatEmitter::CMEQ(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	EmitThreeSame(true, size >> 4, 0b10001, Rd, Rn, Rm);
-}
-
-void ARM64FloatEmitter::CMGE(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	EmitThreeSame(false, size >> 4, 0b00111, Rd, Rn, Rm);
-}
-
-void ARM64FloatEmitter::CMGT(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	EmitThreeSame(false, size >> 4, 0b00110, Rd, Rn, Rm);
-}
-
-void ARM64FloatEmitter::CMHI(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	EmitThreeSame(true, size >> 4, 0b00110, Rd, Rn, Rm);
-}
-
-void ARM64FloatEmitter::CMHS(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	EmitThreeSame(true, size >> 4, 0b00111, Rd, Rn, Rm);
-}
-
-void ARM64FloatEmitter::CMTST(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	EmitThreeSame(false, size >> 4, 0b10001, Rd, Rn, Rm);
-}
-
-void ARM64FloatEmitter::CMEQ(u8 size, ARM64Reg Rd, ARM64Reg Rn) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	Emit2RegMisc(IsQuad(Rd), false, size >> 4, 0b01001, Rd, Rn);
-}
-
-void ARM64FloatEmitter::CMGE(u8 size, ARM64Reg Rd, ARM64Reg Rn) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	Emit2RegMisc(IsQuad(Rd), true, size >> 4, 0b01000, Rd, Rn);
-}
-
-void ARM64FloatEmitter::CMGT(u8 size, ARM64Reg Rd, ARM64Reg Rn) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	Emit2RegMisc(IsQuad(Rd), false, size >> 4, 0b01000, Rd, Rn);
-}
-
-void ARM64FloatEmitter::CMLE(u8 size, ARM64Reg Rd, ARM64Reg Rn) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	Emit2RegMisc(IsQuad(Rd), true, size >> 4, 0b01001, Rd, Rn);
-}
-
-void ARM64FloatEmitter::CMLT(u8 size, ARM64Reg Rd, ARM64Reg Rn) {
-	_assert_msg_(!IsQuad(Rd) || size != 64, "%s cannot be used for scalar double", __FUNCTION__);
-	Emit2RegMisc(IsQuad(Rd), false, size >> 4, 0b01010, Rd, Rn);
-}
-
 // Move
 void ARM64FloatEmitter::DUP(u8 size, ARM64Reg Rd, ARM64Reg Rn)
 {
@@ -3435,95 +3282,6 @@ void ARM64FloatEmitter::SMOV(u8 size, ARM64Reg Rd, ARM64Reg Rn, u8 index)
 	EmitCopy(b64Bit, 0, imm5, 5, Rd, Rn);
 }
 
-void ARM64FloatEmitter::EncodeModImm(bool Q, u8 op, u8 cmode, u8 o2, ARM64Reg Rd, u8 abcdefgh) {
-	Rd = DecodeReg(Rd);
-	u8 abc = abcdefgh >> 5;
-	u8 defgh = abcdefgh & 0x1F;
-	Write32((Q << 30) | (op << 29) | (0xF << 24) | (abc << 16) | (cmode << 12) | (o2 << 11) | (1 << 10) | (defgh << 5) | Rd);
-}
-
-void ARM64FloatEmitter::FMOV(u8 size, ARM64Reg Rd, u8 imm8) {
-	_assert_msg_(!IsSingle(Rd), "%s doesn't support singles", __FUNCTION__);
-	_assert_msg_(size == 32 || size == 64, "%s: unsupported size", __FUNCTION__);
-	_assert_msg_(IsQuad(Rd) || size == 32, "Use non-SIMD FMOV to load one double imm8");
-	EncodeModImm(IsQuad(Rd), size >> 6, 0b1111, 0, Rd, imm8);
-}
-
-void ARM64FloatEmitter::MOVI(u8 size, ARM64Reg Rd, u8 imm8, u8 shift, bool MSL) {
-	_assert_msg_(!IsSingle(Rd), "%s doesn't support singles", __FUNCTION__);
-	_assert_msg_(size == 8 || size == 16 || size == 32 || size == 64, "%s: unsupported size %d", __FUNCTION__, size);
-	_assert_msg_((shift & 7) == 0 && shift < size, "%s: unsupported shift %d", __FUNCTION__, shift);
-	_assert_msg_(!MSL || (size == 32 && shift > 0 && shift <= 16), "MOVI MSL shift requires size 32, shift must be 8 or 16");
-	_assert_msg_(size != 64 || shift == 0, "MOVI 64-bit imm cannot be shifted");
-
-	u8 cmode = 0;
-	if (size == 8)
-		cmode = 0b1110;
-	else if (size == 16)
-		cmode = 0b1000 | (shift >> 2);
-	else if (MSL)
-		cmode = 0b1100 | (shift >> 3);
-	else if (size == 32)
-		cmode = (shift >> 2);
-	else if (size == 64)
-		cmode = 0b1110;
-	else
-		_assert_msg_(false, "%s: unhandled case", __FUNCTION__);
-
-	EncodeModImm(IsQuad(Rd), size >> 6, cmode, 0, Rd, imm8);
-}
-
-void ARM64FloatEmitter::MVNI(u8 size, ARM64Reg Rd, u8 imm8, u8 shift, bool MSL) {
-	_assert_msg_(!IsSingle(Rd), "%s doesn't support singles", __FUNCTION__);
-	_assert_msg_(size == 16 || size == 32, "%s: unsupported size %d", __FUNCTION__, size);
-	_assert_msg_((shift & 7) == 0 && shift < size, "%s: unsupported shift %d", __FUNCTION__, shift);
-	_assert_msg_(!MSL || (size == 32 && shift > 0 && shift <= 16), "MVNI MSL shift requires size 32, shift must be 8 or 16");
-
-	u8 cmode = 0;
-	if (size == 16)
-		cmode = 0b1000 | (shift >> 2);
-	else if (MSL)
-		cmode = 0b1100 | (shift >> 3);
-	else if (size == 32)
-		cmode = (shift >> 2);
-	else
-		_assert_msg_(false, "%s: unhandled case", __FUNCTION__);
-
-	EncodeModImm(IsQuad(Rd), 1, cmode, 0, Rd, imm8);
-}
-
-void ARM64FloatEmitter::ORR(u8 size, ARM64Reg Rd, u8 imm8, u8 shift) {
-	_assert_msg_(!IsSingle(Rd), "%s doesn't support singles", __FUNCTION__);
-	_assert_msg_(size == 16 || size == 32, "%s: unsupported size %d", __FUNCTION__, size);
-	_assert_msg_((shift & 7) == 0 && shift < size, "%s: unsupported shift %d", __FUNCTION__, shift);
-
-	u8 cmode = 0;
-	if (size == 16)
-		cmode = 0b1001 | (shift >> 2);
-	else if (size == 32)
-		cmode = 0b0001 | (shift >> 2);
-	else
-		_assert_msg_(false, "%s: unhandled case", __FUNCTION__);
-
-	EncodeModImm(IsQuad(Rd), 0, cmode, 0, Rd, imm8);
-}
-
-void ARM64FloatEmitter::BIC(u8 size, ARM64Reg Rd, u8 imm8, u8 shift) {
-	_assert_msg_(!IsSingle(Rd), "%s doesn't support singles", __FUNCTION__);
-	_assert_msg_(size == 16 || size == 32, "%s: unsupported size %d", __FUNCTION__, size);
-	_assert_msg_((shift & 7) == 0 && shift < size, "%s: unsupported shift %d", __FUNCTION__, shift);
-
-	u8 cmode = 0;
-	if (size == 16)
-		cmode = 0b1001 | (shift >> 2);
-	else if (size == 32)
-		cmode = 0b0001 | (shift >> 2);
-	else
-		_assert_msg_(false, "%s: unhandled case", __FUNCTION__);
-
-	EncodeModImm(IsQuad(Rd), 1, cmode, 0, Rd, imm8);
-}
-
 // One source
 void ARM64FloatEmitter::FCVT(u8 size_to, u8 size_from, ARM64Reg Rd, ARM64Reg Rn)
 {
@@ -3586,38 +3344,22 @@ void ARM64FloatEmitter::UCVTF(ARM64Reg Rd, ARM64Reg Rn)
 
 void ARM64FloatEmitter::SCVTF(ARM64Reg Rd, ARM64Reg Rn, int scale)
 {
-	if (IsScalar(Rn)) {
+	bool sf = Is64Bit(Rn);
-		int imm = (IsDouble(Rn) ? 64 : 32) * 2 - scale;
+	u32 type = 0;
-		Rd = DecodeReg(Rd);
+	if (IsDouble(Rd))
-		Rn = DecodeReg(Rn);
+		type = 1;
 
-		Write32((1 << 30) | (0 << 29) | (0x1F << 24) | (imm << 16) | (0x1C << 11) | (1 << 10) | (Rn << 5) | Rd);
+	EmitConversion2(sf, 0, false, type, 0, 2, 64 - scale, Rd, Rn);
-	} else {
-		bool sf = Is64Bit(Rn);
-		u32 type = 0;
-		if (IsDouble(Rd))
-			type = 1;
-
-		EmitConversion2(sf, 0, false, type, 0, 2, 64 - scale, Rd, Rn);
-	}
 }
 
 void ARM64FloatEmitter::UCVTF(ARM64Reg Rd, ARM64Reg Rn, int scale)
 {
-	if (IsScalar(Rn)) {
+	bool sf = Is64Bit(Rn);
-		int imm = (IsDouble(Rn) ? 64 : 32) * 2 - scale;
+	u32 type = 0;
-		Rd = DecodeReg(Rd);
+	if (IsDouble(Rd))
-		Rn = DecodeReg(Rn);
+		type = 1;
 
-		Write32((1 << 30) | (1 << 29) | (0x1F << 24) | (imm << 16) | (0x1C << 11) | (1 << 10) | (Rn << 5) | Rd);
+	EmitConversion2(sf, 0, false, type, 0, 3, 64 - scale, Rd, Rn);
-	} else {
-		bool sf = Is64Bit(Rn);
-		u32 type = 0;
-		if (IsDouble(Rd))
-			type = 1;
-
-		EmitConversion2(sf, 0, false, type, 0, 3, 64 - scale, Rd, Rn);
-	}
 }
 
 void ARM64FloatEmitter::FCMP(ARM64Reg Rn, ARM64Reg Rm)
@@ -3674,14 +3416,6 @@ void ARM64FloatEmitter::FCSEL(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm, CCFlags con
 	EmitCondSelect(0, 0, cond, Rd, Rn, Rm);
 }
 
-void ARM64FloatEmitter::FCCMP(ARM64Reg Rn, ARM64Reg Rm, u8 nzcv, CCFlags cond) {
-	EmitCondCompare(0, 0, cond, 0, nzcv, Rn, Rm);
-}
-
-void ARM64FloatEmitter::FCCMPE(ARM64Reg Rn, ARM64Reg Rm, u8 nzcv, CCFlags cond) {
-	EmitCondCompare(0, 0, cond, 1, nzcv, Rn, Rm);
-}
-
 // Permute
 void ARM64FloatEmitter::UZP1(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
 {
@@ -3708,20 +3442,6 @@ void ARM64FloatEmitter::ZIP2(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
 	EmitPermute(size, 7, Rd, Rn, Rm);
 }
 
-void ARM64FloatEmitter::EXT(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm, int index) {
-	_assert_msg_(!IsSingle(Rd), "%s doesn't support singles!", __FUNCTION__);
-
-	bool quad = IsQuad(Rd);
-	_assert_msg_(index >= 0 && index < 16 && (quad || index < 8), "%s start index out of bounds", __FUNCTION__);
-	_assert_msg_(IsQuad(Rd) == IsQuad(Rn) && IsQuad(Rd) == IsQuad(Rm), "%s operands not same size", __FUNCTION__);
-
-	Rd = DecodeReg(Rd);
-	Rn = DecodeReg(Rn);
-	Rm = DecodeReg(Rm);
-
-	Write32((quad << 30) | (0x17 << 25) | (Rm << 16) | (index << 11) | (Rn << 5) | Rd);
-}
-
 // Shift by immediate
 void ARM64FloatEmitter::SSHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift)
 {
@@ -3747,12 +3467,6 @@ void ARM64FloatEmitter::USHLL2(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift)
 {
 	USHLL(src_size, Rd, Rn, shift, true);
 }
-void ARM64FloatEmitter::SHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn) {
-	SHLL(src_size, Rd, Rn, false);
-}
-void ARM64FloatEmitter::SHLL2(u8 src_size, ARM64Reg Rd, ARM64Reg Rn) {
-	SHLL(src_size, Rd, Rn, true);
-}
 void ARM64FloatEmitter::SXTL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn)
 {
 	SXTL(src_size, Rd, Rn, false);
@@ -3792,11 +3506,6 @@ void ARM64FloatEmitter::USHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift,
 	EmitShiftImm(upper, 1, imm >> 3, imm & 7, 0x14, Rd, Rn);
 }
 
-void ARM64FloatEmitter::SHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, bool upper) {
-	_assert_msg_(src_size <= 32, "%s shift amount cannot be 64", __FUNCTION__);
-	Emit2RegMisc(upper, 1, src_size >> 4, 0b10011, Rd, Rn);
-}
-
 void ARM64FloatEmitter::SHRN(u8 dest_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift, bool upper)
 {
 	_assert_msg_(shift > 0, "%s shift amount must be greater than zero!", __FUNCTION__);
@@ -4187,131 +3896,20 @@ void ARM64FloatEmitter::MOVI2F(ARM64Reg Rd, float value, ARM64Reg scratch, bool
 }
 
 // TODO: Quite a few values could be generated easily using the MOVI instruction and friends.
-void ARM64FloatEmitter::MOVI2FDUP(ARM64Reg Rd, float value, ARM64Reg scratch, bool negate) {
+void ARM64FloatEmitter::MOVI2FDUP(ARM64Reg Rd, float value, ARM64Reg scratch) {
-	_assert_msg_(!IsSingle(Rd), "%s doesn't support singles", __FUNCTION__);
+	// TODO: Make it work with more element sizes
+	// TODO: Optimize - there are shorter solution for many values
+	ARM64Reg s = (ARM64Reg)(S0 + DecodeReg(Rd));
 	int ival;
 	memcpy(&ival, &value, 4);
-	uint8_t imm8;
 	if (ival == 0) {  // Make sure to not catch negative zero here
-		// Prefer MOVI 0, which may have no latency on some CPUs.
+		EOR(Rd, Rd, Rd);
-		MOVI(32, Rd, 0);
-		if (negate)
-			FNEG(32, Rd, Rd);
-	} else if (negate && FPImm8FromFloat(-value, &imm8)) {
-		FMOV(32, Rd, imm8);
-	} else if (FPImm8FromFloat(value, &imm8)) {
-		FMOV(32, Rd, imm8);
-		if (negate) {
-			FNEG(32, Rd, Rd);
-		}
-	} else if (TryAnyMOVI(32, Rd, ival)) {
-		if (negate) {
-			FNEG(32, Rd, Rd);
-		}
-	} else if (TryAnyMOVI(32, Rd, ival ^ 0x80000000)) {
-		if (!negate) {
-			FNEG(32, Rd, Rd);
-		}
 	} else {
-		_assert_msg_(scratch != INVALID_REG, "Failed to find a way to generate FP immediate %f without scratch", value);
+		MOVI2F(s, value, scratch);
-		if (negate) {
+		DUP(32, Rd, Rd, 0);
-			ival ^= 0x80000000;
-		}
-		m_emit->MOVI2R(scratch, ival);
-		DUP(32, Rd, scratch);
 	}
 }
 
-bool ARM64FloatEmitter::TryMOVI(u8 size, ARM64Reg Rd, uint64_t elementValue) {
-	if (size == 8) {
-		// Can always do 8.
-		MOVI(size, Rd, elementValue & 0xFF);
-		return true;
-	} else if (size == 16) {
-		if ((elementValue & 0xFF00) == 0) {
-			MOVI(size, Rd, elementValue & 0xFF, 0);
-			return true;
-		} else if ((elementValue & 0x00FF) == 0) {
-			MOVI(size, Rd, (elementValue >> 8) & 0xFF, 8);
-			return true;
-		} else if ((elementValue & 0xFF00) == 0xFF00) {
-			MVNI(size, Rd, ~elementValue & 0xFF, 0);
-			return true;
-		} else if ((elementValue & 0x00FF) == 0x00FF) {
-			MVNI(size, Rd, (~elementValue >> 8) & 0xFF, 8);
-			return true;
-		}
-
-		return false;
-	} else if (size == 32) {
-		for (int shift = 0; shift < 32; shift += 8) {
-			uint32_t mask = 0xFFFFFFFF &~ (0xFF << shift);
-			if ((elementValue & mask) == 0) {
-				MOVI(size, Rd, (elementValue >> shift) & 0xFF, shift);
-				return true;
-			} else if ((elementValue & mask) == mask) {
-				MVNI(size, Rd, (~elementValue >> shift) & 0xFF, shift);
-				return true;
-			}
-		}
-
-		// Maybe an MSL shift will work?
-		for (int shift = 8; shift <= 16; shift += 8) {
-			uint32_t mask = 0xFFFFFFFF & ~(0xFF << shift);
-			uint32_t ones = (1 << shift) - 1;
-			uint32_t notOnes = 0xFFFFFF00 << shift;
-			if ((elementValue & mask) == ones) {
-				MOVI(size, Rd, (elementValue >> shift) & 0xFF, shift, true);
-				return true;
-			} else if ((elementValue & mask) == notOnes) {
-				MVNI(size, Rd, (elementValue >> shift) & 0xFF, shift, true);
-				return true;
-			}
-		}
-
-		return false;
-	} else if (size == 64) {
-		uint8_t imm8 = 0;
-		for (int i = 0; i < 8; ++i) {
-			uint8_t byte = (elementValue >> (i * 8)) & 0xFF;
-			if (byte != 0 && byte != 0xFF)
-				return false;
-
-			if (byte == 0xFF)
-				imm8 |= 1 << i;
-		}
-
-		// Didn't run into any partial bytes, so size 64 is doable.
-		MOVI(size, Rd, imm8);
-		return true;
-	}
-	return false;
-}
-
-bool ARM64FloatEmitter::TryAnyMOVI(u8 size, ARM64Reg Rd, uint64_t elementValue) {
-	// Try the original size first in case that's more optimal.
-	if (TryMOVI(size, Rd, elementValue))
-		return true;
-
-	uint64_t value = elementValue;
-	if (size != 64) {
-		uint64_t masked = elementValue & ((1 << size) - 1);
-		for (int i = size; i < 64; ++i) {
-			value |= masked << i;
-		}
-	}
-
-	for (int attempt = 8; attempt <= 64; attempt += attempt) {
-		// Original size was already attempted above.
-		if (attempt != size) {
-			if (TryMOVI(attempt, Rd, value))
-				return true;
-		}
-	}
-
-	return false;
-}
-
 void ARM64XEmitter::SUBSI2R(ARM64Reg Rd, ARM64Reg Rn, u64 imm, ARM64Reg scratch) {
 	u32 val;
 	bool shift;

Common/Arm64Emitter.h

@@ -94,7 +94,7 @@ enum ARM64Reg
 
 // R19-R28. R29 (FP), R30 (LR) are always saved and FP updated appropriately.
 const u32 ALL_CALLEE_SAVED = 0x1FF80000;
-const u32 ALL_CALLEE_SAVED_FP = 0x0000FF00;  // q8-q15
+const u32 ALL_CALLEE_SAVED_FP = 0x0000FF00;  // d8-d15
 
 inline bool Is64Bit(ARM64Reg reg) { return (reg & 0x20) != 0; }
 inline bool IsSingle(ARM64Reg reg) { return (reg & 0xC0) == 0x40; }
@@ -290,23 +290,6 @@ public:
 		}
 		m_shifttype = ST_LSL;
 	}
-	ArithOption(ARM64Reg Rd, bool index, bool signExtend) {
-		if (index)
-			m_shift = 4;
-		else
-			m_shift = 0;
-
-		m_destReg = Rd;
-		m_type = TYPE_EXTENDEDREG;
-		if (Is64Bit(Rd)) {
-			m_width = WIDTH_64BIT;
-			m_extend = EXTEND_UXTX;
-		} else {
-			m_width = WIDTH_32BIT;
-			m_extend = signExtend ? EXTEND_SXTW : EXTEND_UXTW;
-		}
-		m_shifttype = ST_LSL;
-	}
 	ArithOption(ARM64Reg Rd, ShiftType shift_type, u32 shift)
 	{
 		m_destReg = Rd;
@@ -418,7 +401,6 @@ public:
 	void ReserveCodeSpace(u32 bytes);
 	const u8* AlignCode16();
 	const u8* AlignCodePage();
-	const u8 *NopAlignCode16();
 	void FlushIcache();
 	void FlushIcacheSection(const u8* start, const u8* end);
 	u8* GetWritableCodePtr();
@@ -820,13 +802,6 @@ public:
 	void FSQRT(ARM64Reg Rd, ARM64Reg Rn);
 	void FMOV(ARM64Reg Rd, ARM64Reg Rn, bool top = false);  // Also generalized move between GPR/FP
 
-	// Scalar - pairwise
-	void FADDP(ARM64Reg Rd, ARM64Reg Rn);
-	void FMAXP(ARM64Reg Rd, ARM64Reg Rn);
-	void FMINP(ARM64Reg Rd, ARM64Reg Rn);
-	void FMAXNMP(ARM64Reg Rd, ARM64Reg Rn);
-	void FMINNMP(ARM64Reg Rd, ARM64Reg Rn);
-
 	// Scalar - 2 Source
 	void FADD(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void FMUL(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
@@ -851,12 +826,9 @@ public:
 	void AND(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void EOR(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void BSL(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void BIT(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void BIF(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void DUP(u8 size, ARM64Reg Rd, ARM64Reg Rn, u8 index);
 	void FABS(u8 size, ARM64Reg Rd, ARM64Reg Rn);
 	void FADD(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void FADDP(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void FMAX(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void FMLA(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void FMLS(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
@@ -866,8 +838,6 @@ public:
 	void FCVTN(u8 dest_size, ARM64Reg Rd, ARM64Reg Rn);
 	void FCVTZS(u8 size, ARM64Reg Rd, ARM64Reg Rn);
 	void FCVTZU(u8 size, ARM64Reg Rd, ARM64Reg Rn);
-	void FCVTZS(u8 size, ARM64Reg Rd, ARM64Reg Rn, int scale);
-	void FCVTZU(u8 size, ARM64Reg Rd, ARM64Reg Rn, int scale);
 	void FDIV(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void FMUL(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void FNEG(u8 size, ARM64Reg Rd, ARM64Reg Rn);
@@ -898,18 +868,6 @@ public:
 	void XTN(u8 dest_size, ARM64Reg Rd, ARM64Reg Rn);
 	void XTN2(u8 dest_size, ARM64Reg Rd, ARM64Reg Rn);
 
-	void CMEQ(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void CMGE(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void CMGT(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void CMHI(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void CMHS(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void CMTST(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void CMEQ(u8 size, ARM64Reg Rd, ARM64Reg Rn);
-	void CMGE(u8 size, ARM64Reg Rd, ARM64Reg Rn);
-	void CMGT(u8 size, ARM64Reg Rd, ARM64Reg Rn);
-	void CMLE(u8 size, ARM64Reg Rd, ARM64Reg Rn);
-	void CMLT(u8 size, ARM64Reg Rd, ARM64Reg Rn);
-
 	// Move
 	void DUP(u8 size, ARM64Reg Rd, ARM64Reg Rn);
 	void INS(u8 size, ARM64Reg Rd, u8 index, ARM64Reg Rn);
@@ -917,18 +875,6 @@ public:
 	void UMOV(u8 size, ARM64Reg Rd, ARM64Reg Rn, u8 index);
 	void SMOV(u8 size, ARM64Reg Rd, ARM64Reg Rn, u8 index);
 
-	// Vector immediates
-	void FMOV(u8 size, ARM64Reg Rd, u8 imm8);
-	// MSL means bits shifted in are 1s.  For size=64, each bit of imm8 is expanded to 8 actual bits.
-	void MOVI(u8 size, ARM64Reg Rd, u8 imm8, u8 shift = 0, bool MSL = false);
-	void MVNI(u8 size, ARM64Reg Rd, u8 imm8, u8 shift = 0, bool MSL = false);
-	void ORR(u8 size, ARM64Reg Rd, u8 imm8, u8 shift = 0);
-	void BIC(u8 size, ARM64Reg Rd, u8 imm8, u8 shift = 0);
-
-	bool TryMOVI(u8 size, ARM64Reg Rd, uint64_t value);
-	// Allow using a different size.  Unclear if there's a penalty.
-	bool TryAnyMOVI(u8 size, ARM64Reg Rd, uint64_t value);
-
 	// One source
 	void FCVT(u8 size_to, u8 size_from, ARM64Reg Rd, ARM64Reg Rn);
 
@@ -937,8 +883,6 @@ public:
 	// and one that outputs to a scalar fp register.
 	void FCVTS(ARM64Reg Rd, ARM64Reg Rn, RoundingMode round);
 	void FCVTU(ARM64Reg Rd, ARM64Reg Rn, RoundingMode round);
-	void FCVTZS(ARM64Reg Rd, ARM64Reg Rn, int scale);
-	void FCVTZU(ARM64Reg Rd, ARM64Reg Rn, int scale);
 
 	// Scalar convert int to float. No rounding mode specifier necessary.
 	void SCVTF(ARM64Reg Rd, ARM64Reg Rn);
@@ -965,10 +909,6 @@ public:
 	// Conditional select
 	void FCSEL(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm, CCFlags cond);
 
-	// Conditional compare
-	void FCCMP(ARM64Reg Rn, ARM64Reg Rm, u8 nzcv, CCFlags cond);
-	void FCCMPE(ARM64Reg Rn, ARM64Reg Rm, u8 nzcv, CCFlags cond);
-
 	// Permute
 	void UZP1(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void TRN1(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
@@ -976,17 +916,12 @@ public:
 	void UZP2(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void TRN2(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void ZIP2(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	// Related to permute, extract vector from pair (always by byte arrangement.)
-	void EXT(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm, int index);
 
 	// Shift by immediate
 	void SSHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift);
 	void SSHLL2(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift);
 	void USHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift);
 	void USHLL2(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift);
-	// Shift == src_size for these.
-	void SHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn);
-	void SHLL2(u8 src_size, ARM64Reg Rd, ARM64Reg Rn);
 	void SHRN(u8 dest_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift);
 	void SHRN2(u8 dest_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift);
 	void SXTL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn);
@@ -1003,7 +938,7 @@ public:
 	void FMLA(u8 esize, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm, u8 index);
 
 	void MOVI2F(ARM64Reg Rd, float value, ARM64Reg scratch = INVALID_REG, bool negate = false);
-	void MOVI2FDUP(ARM64Reg Rd, float value, ARM64Reg scratch = INVALID_REG, bool negate = false);
+	void MOVI2FDUP(ARM64Reg Rd, float value, ARM64Reg scratch = INVALID_REG);
 
 	// ABI related
 	void ABI_PushRegisters(uint32_t gpr_registers, uint32_t fp_registers);
@@ -1018,7 +953,6 @@ private:
 	void EmitScalar2Source(bool M, bool S, u32 type, u32 opcode, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void EmitThreeSame(bool U, u32 size, u32 opcode, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void EmitCopy(bool Q, u32 op, u32 imm5, u32 imm4, ARM64Reg Rd, ARM64Reg Rn);
-	void EmitScalarPairwise(bool U, u32 size, u32 opcode, ARM64Reg Rd, ARM64Reg Rn);
 	void Emit2RegMisc(bool Q, bool U, u32 size, u32 opcode, ARM64Reg Rd, ARM64Reg Rn);
 	void EmitLoadStoreSingleStructure(bool L, bool R, u32 opcode, bool S, u32 size, ARM64Reg Rt, ARM64Reg Rn);
 	void EmitLoadStoreSingleStructure(bool L, bool R, u32 opcode, bool S, u32 size, ARM64Reg Rt, ARM64Reg Rn, ARM64Reg Rm);
@@ -1027,7 +961,6 @@ private:
 	void EmitConversion2(bool sf, bool S, bool direction, u32 type, u32 rmode, u32 opcode, int scale, ARM64Reg Rd, ARM64Reg Rn);
 	void EmitCompare(bool M, bool S, u32 op, u32 opcode2, ARM64Reg Rn, ARM64Reg Rm);
 	void EmitCondSelect(bool M, bool S, CCFlags cond, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-	void EmitCondCompare(bool M, bool S, CCFlags cond, int op, u8 nzcv, ARM64Reg Rn, ARM64Reg Rm);
 	void EmitPermute(u32 size, u32 op, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
 	void EmitScalarImm(bool M, bool S, u32 type, u32 imm5, ARM64Reg Rd, u32 imm8);
 	void EmitShiftImm(bool Q, bool U, u32 immh, u32 immb, u32 opcode, ARM64Reg Rd, ARM64Reg Rn);
@@ -1041,11 +974,9 @@ private:
 	void EmitScalar3Source(bool isDouble, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm, ARM64Reg Ra, int opcode);
 	void EncodeLoadStorePair(u32 size, bool load, IndexType type, ARM64Reg Rt, ARM64Reg Rt2, ARM64Reg Rn, s32 imm);
 	void EncodeLoadStoreRegisterOffset(u32 size, bool load, ARM64Reg Rt, ARM64Reg Rn, ArithOption Rm);
-	void EncodeModImm(bool Q, u8 op, u8 cmode, u8 o2, ARM64Reg Rd, u8 abcdefgh);
 
 	void SSHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift, bool upper);
 	void USHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift, bool upper);
-	void SHLL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, bool upper);
 	void SHRN(u8 dest_size, ARM64Reg Rd, ARM64Reg Rn, u32 shift, bool upper);
 	void SXTL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, bool upper);
 	void UXTL(u8 src_size, ARM64Reg Rd, ARM64Reg Rn, bool upper);

Common/ArmCPUDetect.cpp

@@ -29,18 +29,17 @@
 #if PPSSPP_ARCH(ARM)
 #include "ext/cpu_features/include/cpuinfo_arm.h"
 
-#if defined(CPU_FEATURES_OS_LINUX)
+#if defined(CPU_FEATURES_OS_LINUX) || defined(CPU_FEATURES_OS_ANDROID)
 #define USE_CPU_FEATURES 1
 #endif
-#elif PPSSPP_ARCH(ARM64)
+#elif PPSSPP_ARCH(ARM64) && defined(__aarch64__)
 #include "ext/cpu_features/include/cpuinfo_aarch64.h"
 
-#if defined(CPU_FEATURES_OS_LINUX) || defined(CPU_FEATURES_OS_ANDROID) || defined(CPU_FEATURES_OS_WINDOWS)
+#if defined(CPU_FEATURES_OS_LINUX) || defined(CPU_FEATURES_OS_ANDROID)
 #define USE_CPU_FEATURES 1
 #endif
 #endif
 
-#include <cstring>
 #include <ctype.h>
 
 #include "Common/CommonTypes.h"
@@ -55,7 +54,7 @@
 std::string GetCPUBrandString();
 #else
 // No CPUID on ARM, so we'll have to read the registry
-#include "Common/CommonWindows.h"
+#include <windows.h>
 std::string GetCPUBrandString() {
 	std::string cpu_string;
 	

Common/ArmEmitter.cpp

@@ -613,14 +613,6 @@ const u8 *ARMXEmitter::AlignCode16()
 	return code;
 }
 
-const u8 *ARMXEmitter::NopAlignCode16() {
-	int bytes = ((-(intptr_t)code) & 15);
-	for (int i = 0; i < bytes / 4; i++) {
-		Write32(0xE320F000); // one of many possible nops
-	}
-	return code;
-}
-
 const u8 *ARMXEmitter::AlignCodePage()
 {
 	ReserveCodeSpace((-(intptr_t)code) & 4095);

Common/ArmEmitter.h

@@ -446,8 +446,6 @@ public:
 	void ReserveCodeSpace(u32 bytes);
 	const u8 *AlignCode16();
 	const u8 *AlignCodePage();
-	const u8 *NopAlignCode16();
-
 	void FlushIcache();
 	void FlushIcacheSection(u8 *start, u8 *end);
 	u8 *GetWritableCodePtr();

Common/Buffer.cpp

@@ -42,7 +42,7 @@ void Buffer::Append(const Buffer &other) {
 void Buffer::AppendValue(int value) {
 	char buf[16];
 	// This is slow.
-	snprintf(buf, sizeof(buf), "%i", value);
+	sprintf(buf, "%i", value);
 	Append(buf);
 }
 

Common/CPUDetect.cpp

@@ -17,7 +17,7 @@
 
 // Reference : https://stackoverflow.com/questions/6121792/how-to-check-if-a-cpu-supports-the-sse3-instruction-set
 #include "ppsspp_config.h"
-#if (PPSSPP_ARCH(X86) || PPSSPP_ARCH(AMD64)) && !defined(__EMSCRIPTEN__)
+#if PPSSPP_ARCH(X86) || PPSSPP_ARCH(AMD64)
 
 #include "ext/cpu_features/include/cpuinfo_x86.h"
 

Common/CPUDetect.h

@@ -109,27 +109,8 @@ struct CPUInfo {
 	bool RiscV_D;
 	bool RiscV_C;
 	bool RiscV_V;
+	bool RiscV_B;
 	bool RiscV_Zicsr;
-	bool RiscV_Zba;
-	bool RiscV_Zbb;
-	bool RiscV_Zbc;
-	bool RiscV_Zbs;
-
-	// LoongArch specific extension flags.
-	bool LOONGARCH_CPUCFG;
-	bool LOONGARCH_LAM;
-	bool LOONGARCH_UAL;
-	bool LOONGARCH_FPU;
-	bool LOONGARCH_LSX;
-	bool LOONGARCH_LASX;
-	bool LOONGARCH_CRC32;
-	bool LOONGARCH_COMPLEX;
-	bool LOONGARCH_CRYPTO;
-	bool LOONGARCH_LVZ;
-	bool LOONGARCH_LBT_X86;
-	bool LOONGARCH_LBT_ARM;
-	bool LOONGARCH_LBT_MIPS;
-	bool LOONGARCH_PTW;
 
 	// Quirks
 	struct {

Common/CodeBlock.h

@@ -10,11 +10,6 @@
 #include "Common/Log.h"
 #include "Common/MemoryUtil.h"
 
-#if PPSSPP_PLATFORM(SWITCH)
-#include <cstdio>
-#include <switch.h>
-#endif // PPSSPP_PLATFORM(SWITCH)
-
 // Everything that needs to generate code should inherit from this.
 // You get memory management for free, plus, you can use all emitter functions without
 // having to prefix them with gen-> or something similar.
@@ -32,7 +27,7 @@ public:
 
 	virtual const u8 *GetCodePtr() const = 0;
 
-	u8 *GetBasePtr() const {
+	u8 *GetBasePtr() {
 		return region;
 	}
 
@@ -70,20 +65,9 @@ public:
 	// Call this before you generate any code.
 	void AllocCodeSpace(int size) {
 		region_size = size;
-#if PPSSPP_PLATFORM(SWITCH)
-		Result rc = jitCreate(&jitController, size);
-		if(R_FAILED(rc)) {
-			printf("Failed to create Jitbuffer of size 0x%x err: 0x%x\n", size, rc);
-		}
-		printf("[NXJIT]: Initialized RX: %p RW: %p\n", jitController.rx_addr, jitController.rw_addr);
-
-		region = (u8 *)jitController.rx_addr;
-		writableRegion = (u8 *)jitController.rw_addr;
-#else // PPSSPP_PLATFORM(SWITCH)
 		// The protection will be set to RW if PlatformIsWXExclusive.
 		region = (u8 *)AllocateExecutableMemory(region_size);
 		writableRegion = region;
-#endif // !PPSSPP_PLATFORM(SWITCH)
 		T::SetCodePointer(region, writableRegion);
 	}
 
@@ -151,13 +135,8 @@ public:
 
 	// Call this when shutting down. Don't rely on the destructor, even though it'll do the job.
 	void FreeCodeSpace() {
-#if !PPSSPP_PLATFORM(SWITCH)
 		ProtectMemoryPages(region, region_size, MEM_PROT_READ | MEM_PROT_WRITE);
-		FreeExecutableMemory(region, region_size);
+		FreeMemoryPages(region, region_size);
-#else // !PPSSPP_PLATFORM(SWITCH)
-		jitClose(&jitController);
-		printf("[NXJIT]: Jit closed\n");
-#endif // PPSSPP_PLATFORM(SWITCH)
 		region = nullptr;
 		writableRegion = nullptr;
 		region_size = 0;
@@ -197,7 +176,5 @@ private:
 	const uint8_t *writeStart_ = nullptr;
 	uint8_t *writableRegion = nullptr;
 	size_t writeEstimated_ = 0;
-#if PPSSPP_PLATFORM(SWITCH)
-	Jit jitController;
-#endif // PPSSPP_PLATFORM(SWITCH)
 };
+

Common/Common.vcxproj

@@ -143,7 +143,6 @@
       <RuntimeTypeInfo>false</RuntimeTypeInfo>
       <MinimalRebuild>false</MinimalRebuild>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
       <SubSystem>Windows</SubSystem>
@@ -171,7 +170,6 @@
       <RuntimeTypeInfo>false</RuntimeTypeInfo>
       <MinimalRebuild>false</MinimalRebuild>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
       <SubSystem>Windows</SubSystem>
@@ -199,7 +197,6 @@
       <RuntimeTypeInfo>false</RuntimeTypeInfo>
       <MinimalRebuild>false</MinimalRebuild>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
       <SubSystem>Windows</SubSystem>
@@ -228,7 +225,6 @@
       <RuntimeTypeInfo>false</RuntimeTypeInfo>
       <MinimalRebuild>false</MinimalRebuild>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
       <SubSystem>Windows</SubSystem>
@@ -260,7 +256,6 @@
       <RuntimeTypeInfo>false</RuntimeTypeInfo>
       <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
       <SubSystem>Windows</SubSystem>
@@ -296,7 +291,6 @@
       <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
       <StringPooling>true</StringPooling>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
       <SubSystem>Windows</SubSystem>
@@ -332,7 +326,6 @@
       <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
       <StringPooling>true</StringPooling>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
       <SubSystem>Windows</SubSystem>
@@ -368,7 +361,6 @@
       <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
       <StringPooling>true</StringPooling>
       <DebugInformationFormat>ProgramDatabase</DebugInformationFormat>
-      <LanguageStandard>stdcpp17</LanguageStandard>
     </ClCompile>
     <Link>
       <SubSystem>Windows</SubSystem>
@@ -385,13 +377,6 @@
     </Lib>
   </ItemDefinitionGroup>
   <ItemGroup>
-    <ClInclude Include="..\ext\basis_universal\basisu.h" />
-    <ClInclude Include="..\ext\basis_universal\basisu_containers.h" />
-    <ClInclude Include="..\ext\basis_universal\basisu_containers_impl.h" />
-    <ClInclude Include="..\ext\basis_universal\basisu_file_headers.h" />
-    <ClInclude Include="..\ext\basis_universal\basisu_transcoder.h" />
-    <ClInclude Include="..\ext\basis_universal\basisu_transcoder_internal.h" />
-    <ClInclude Include="..\ext\basis_universal\basisu_transcoder_uastc.h" />
     <ClInclude Include="..\ext\libpng17\png.h" />
     <ClInclude Include="..\ext\libpng17\pngconf.h" />
     <ClInclude Include="..\ext\libpng17\pngdebug.h" />
@@ -399,7 +384,6 @@
     <ClInclude Include="..\ext\libpng17\pnglibconf.h" />
     <ClInclude Include="..\ext\libpng17\pngpriv.h" />
     <ClInclude Include="..\ext\libpng17\pngstruct.h" />
-    <ClInclude Include="..\ext\naett\naett.h" />
     <ClInclude Include="..\ext\vma\vk_mem_alloc.h" />
     <ClInclude Include="ABI.h" />
     <ClInclude Include="Arm64Emitter.h" />
@@ -421,7 +405,6 @@
     <ClInclude Include="Data\Encoding\Shiftjis.h" />
     <ClInclude Include="Data\Encoding\Utf16.h" />
     <ClInclude Include="Data\Encoding\Utf8.h" />
-    <ClInclude Include="Data\Format\DDSLoad.h" />
     <ClInclude Include="Data\Format\IniFile.h" />
     <ClInclude Include="Data\Format\JSONReader.h" />
     <ClInclude Include="Data\Format\JSONWriter.h" />
@@ -435,7 +418,6 @@
     <ClInclude Include="Data\Text\Parsers.h" />
     <ClInclude Include="Data\Text\WrapText.h" />
     <ClInclude Include="FakeEmitter.h" />
-    <ClInclude Include="File\AndroidContentURI.h" />
     <ClInclude Include="File\AndroidStorage.h" />
     <ClInclude Include="File\DirListing.h" />
     <ClInclude Include="File\DiskFree.h" />
@@ -443,21 +425,17 @@
     <ClInclude Include="File\FileUtil.h" />
     <ClInclude Include="File\Path.h" />
     <ClInclude Include="File\PathBrowser.h" />
-    <ClInclude Include="File\VFS\DirectoryReader.h" />
     <ClInclude Include="File\VFS\VFS.h" />
-    <ClInclude Include="File\VFS\ZipFileReader.h" />
+    <ClInclude Include="File\VFS\AssetReader.h" />
     <ClInclude Include="GPU\D3D11\D3D11Loader.h" />
     <ClInclude Include="GPU\D3D9\D3DCompilerLoader.h" />
     <ClInclude Include="GPU\D3D9\D3D9ShaderCompiler.h" />
     <ClInclude Include="GPU\D3D9\D3D9StateCache.h" />
     <ClInclude Include="GPU\DataFormat.h" />
-    <ClInclude Include="GPU\GPUBackendCommon.h" />
     <ClInclude Include="GPU\MiscTypes.h" />
     <ClInclude Include="GPU\OpenGL\DataFormatGL.h" />
     <ClInclude Include="GPU\OpenGL\gl3stub.h" />
     <ClInclude Include="GPU\OpenGL\GLFeatures.h" />
-    <ClInclude Include="GPU\OpenGL\GLFrameData.h" />
-    <ClInclude Include="GPU\OpenGL\GLMemory.h" />
     <ClInclude Include="GPU\OpenGL\GLQueueRunner.h" />
     <ClInclude Include="GPU\OpenGL\GLRenderManager.h" />
     <ClInclude Include="GPU\OpenGL\GLSLProgram.h" />
@@ -472,7 +450,6 @@
     <ClInclude Include="GPU\Vulkan\VulkanBarrier.h" />
     <ClInclude Include="GPU\Vulkan\VulkanContext.h" />
     <ClInclude Include="GPU\Vulkan\VulkanDebug.h" />
-    <ClInclude Include="GPU\Vulkan\VulkanDescSet.h" />
     <ClInclude Include="GPU\Vulkan\VulkanFramebuffer.h" />
     <ClInclude Include="GPU\Vulkan\VulkanFrameData.h" />
     <ClInclude Include="GPU\Vulkan\VulkanImage.h" />
@@ -492,12 +469,10 @@
     <ClInclude Include="Math\lin\vec3.h" />
     <ClInclude Include="Math\math_util.h" />
     <ClInclude Include="Math\Statistics.h" />
-    <ClInclude Include="Net\HTTPNaettRequest.h" />
     <ClInclude Include="Net\NetBuffer.h" />
     <ClInclude Include="Net\HTTPClient.h" />
     <ClInclude Include="Net\HTTPHeaders.h" />
     <ClInclude Include="Net\HTTPServer.h" />
-    <ClInclude Include="Net\HTTPRequest.h" />
     <ClInclude Include="Net\Resolve.h" />
     <ClInclude Include="Net\Sinks.h" />
     <ClInclude Include="Net\URL.h" />
@@ -509,7 +484,6 @@
     <ClInclude Include="Render\Text\draw_text.h" />
     <ClInclude Include="Render\Text\draw_text_android.h" />
     <ClInclude Include="Render\Text\draw_text_qt.h" />
-    <ClInclude Include="Render\Text\draw_text_sdl.h" />
     <ClInclude Include="Render\Text\draw_text_uwp.h" />
     <ClInclude Include="Render\Text\draw_text_win.h" />
     <ClInclude Include="LogReporting.h" />
@@ -532,6 +506,27 @@
     <ClInclude Include="Crypto\sha256.h" />
     <ClInclude Include="DbgNew.h" />
     <ClInclude Include="ExceptionHandlerSetup.h" />
+    <ClInclude Include="GL\GLInterfaceBase.h" />
+    <ClInclude Include="GL\GLInterface\EGL.h">
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
+    </ClInclude>
+    <ClInclude Include="GL\GLInterface\EGLAndroid.h">
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
+    </ClInclude>
     <ClInclude Include="GraphicsContext.h" />
     <ClInclude Include="Log.h" />
     <ClInclude Include="LogManager.h" />
@@ -545,8 +540,6 @@
     <ClInclude Include="Swap.h" />
     <ClInclude Include="SysError.h" />
     <ClInclude Include="System\Display.h" />
-    <ClInclude Include="System\OSD.h" />
-    <ClInclude Include="System\Request.h" />
     <ClInclude Include="System\NativeApp.h" />
     <ClInclude Include="System\System.h" />
     <ClInclude Include="Thread\Barrier.h" />
@@ -561,7 +554,6 @@
     <ClInclude Include="TimeUtil.h" />
     <ClInclude Include="UI\AsyncImageFileView.h" />
     <ClInclude Include="UI\Context.h" />
-    <ClInclude Include="UI\IconCache.h" />
     <ClInclude Include="UI\PopupScreens.h" />
     <ClInclude Include="UI\Root.h" />
     <ClInclude Include="UI\Screen.h" />
@@ -582,7 +574,6 @@
     <ClInclude Include="x64Emitter.h" />
   </ItemGroup>
   <ItemGroup>
-    <ClCompile Include="..\ext\basis_universal\basisu_transcoder.cpp" />
     <ClCompile Include="..\ext\libpng17\png.c">
       <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">NotUsing</PrecompiledHeader>
       <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">NotUsing</PrecompiledHeader>
@@ -831,7 +822,6 @@
       <ForcedIncludeFiles Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
       </ForcedIncludeFiles>
     </ClCompile>
-    <ClCompile Include="..\ext\naett\naett.c" />
     <ClCompile Include="..\ext\vma\vk_mem_alloc.cpp" />
     <ClCompile Include="ABI.cpp" />
     <ClCompile Include="Arm64Emitter.cpp" />
@@ -847,13 +837,11 @@
     </ClCompile>
     <ClCompile Include="ArmEmitter.cpp" />
     <ClCompile Include="Buffer.cpp" />
-    <ClCompile Include="Data\Collections\FastVec.h" />
     <ClCompile Include="Data\Color\RGBAUtil.cpp" />
     <ClCompile Include="Data\Convert\SmallDataConvert.cpp" />
     <ClCompile Include="Data\Encoding\Base64.cpp" />
     <ClCompile Include="Data\Encoding\Compression.cpp" />
     <ClCompile Include="Data\Encoding\Utf8.cpp" />
-    <ClCompile Include="Data\Format\DDSLoad.cpp" />
     <ClCompile Include="Data\Format\IniFile.cpp" />
     <ClCompile Include="Data\Format\JSONReader.cpp" />
     <ClCompile Include="Data\Format\JSONWriter.cpp" />
@@ -865,7 +853,6 @@
     <ClCompile Include="Data\Text\I18n.cpp" />
     <ClCompile Include="Data\Text\Parsers.cpp" />
     <ClCompile Include="Data\Text\WrapText.cpp" />
-    <ClCompile Include="File\AndroidContentURI.cpp" />
     <ClCompile Include="File\AndroidStorage.cpp" />
     <ClCompile Include="File\DirListing.cpp" />
     <ClCompile Include="File\DiskFree.cpp" />
@@ -873,21 +860,17 @@
     <ClCompile Include="File\FileUtil.cpp" />
     <ClCompile Include="File\Path.cpp" />
     <ClCompile Include="File\PathBrowser.cpp" />
-    <ClCompile Include="File\VFS\DirectoryReader.cpp" />
     <ClCompile Include="File\VFS\VFS.cpp" />
-    <ClCompile Include="File\VFS\ZipFileReader.cpp" />
+    <ClCompile Include="File\VFS\AssetReader.cpp" />
     <ClCompile Include="GPU\D3D11\D3D11Loader.cpp" />
     <ClCompile Include="GPU\D3D11\thin3d_d3d11.cpp" />
     <ClCompile Include="GPU\D3D9\D3DCompilerLoader.cpp" />
     <ClCompile Include="GPU\D3D9\D3D9ShaderCompiler.cpp" />
     <ClCompile Include="GPU\D3D9\D3D9StateCache.cpp" />
     <ClCompile Include="GPU\D3D9\thin3d_d3d9.cpp" />
-    <ClCompile Include="GPU\GPUBackendCommon.cpp" />
     <ClCompile Include="GPU\OpenGL\DataFormatGL.cpp" />
     <ClCompile Include="GPU\OpenGL\gl3stub.c" />
     <ClCompile Include="GPU\OpenGL\GLFeatures.cpp" />
-    <ClCompile Include="GPU\OpenGL\GLFrameData.cpp" />
-    <ClCompile Include="GPU\OpenGL\GLMemory.cpp" />
     <ClCompile Include="GPU\OpenGL\GLQueueRunner.cpp" />
     <ClCompile Include="GPU\OpenGL\GLRenderManager.cpp" />
     <ClCompile Include="GPU\OpenGL\GLSLProgram.cpp" />
@@ -901,7 +884,6 @@
     <ClCompile Include="GPU\Vulkan\VulkanBarrier.cpp" />
     <ClCompile Include="GPU\Vulkan\VulkanContext.cpp" />
     <ClCompile Include="GPU\Vulkan\VulkanDebug.cpp" />
-    <ClCompile Include="GPU\Vulkan\VulkanDescSet.cpp" />
     <ClCompile Include="GPU\Vulkan\VulkanFramebuffer.cpp" />
     <ClCompile Include="GPU\Vulkan\VulkanFrameData.cpp" />
     <ClCompile Include="GPU\Vulkan\VulkanImage.cpp" />
@@ -920,12 +902,10 @@
     <ClCompile Include="Math\lin\vec3.cpp" />
     <ClCompile Include="Math\math_util.cpp" />
     <ClCompile Include="Math\Statistics.cpp" />
-    <ClCompile Include="Net\HTTPNaettRequest.cpp" />
     <ClCompile Include="Net\NetBuffer.cpp" />
     <ClCompile Include="Net\HTTPClient.cpp" />
     <ClCompile Include="Net\HTTPHeaders.cpp" />
     <ClCompile Include="Net\HTTPServer.cpp" />
-    <ClCompile Include="Net\HTTPRequest.cpp" />
     <ClCompile Include="Net\Resolve.cpp" />
     <ClCompile Include="Net\Sinks.cpp" />
     <ClCompile Include="Net\URL.cpp" />
@@ -937,13 +917,11 @@
     <ClCompile Include="Render\Text\draw_text.cpp" />
     <ClCompile Include="Render\Text\draw_text_android.cpp" />
     <ClCompile Include="Render\Text\draw_text_qt.cpp" />
-    <ClCompile Include="Render\Text\draw_text_sdl.cpp" />
     <ClCompile Include="Render\Text\draw_text_uwp.cpp" />
     <ClCompile Include="Render\Text\draw_text_win.cpp" />
     <ClCompile Include="LogReporting.cpp" />
     <ClCompile Include="RiscVCPUDetect.cpp" />
     <ClCompile Include="RiscVEmitter.cpp" />
-    <ClCompile Include="LoongArchCPUDetect.cpp" />
     <ClCompile Include="Serialize\Serializer.cpp" />
     <ClCompile Include="Data\Convert\ColorConv.cpp" />
     <ClCompile Include="ConsoleListener.cpp" />
@@ -972,6 +950,36 @@
     <ClCompile Include="Crypto\sha1.cpp" />
     <ClCompile Include="Crypto\sha256.cpp" />
     <ClCompile Include="ExceptionHandlerSetup.cpp" />
+    <ClCompile Include="GL\GLInterface\EGL.cpp">
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
+    </ClCompile>
+    <ClCompile Include="GL\GLInterface\EGLAndroid.cpp">
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
+    </ClCompile>
+    <ClCompile Include="GL\GLInterface\GLInterface.cpp">
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Debug|ARM'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|x64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM64'">true</ExcludedFromBuild>
+      <ExcludedFromBuild Condition="'$(Configuration)|$(Platform)'=='Release|ARM'">true</ExcludedFromBuild>
+    </ClCompile>
     <ClCompile Include="LogManager.cpp" />
     <ClCompile Include="MemArenaAndroid.cpp" />
     <ClCompile Include="MemArenaPosix.cpp" />
@@ -983,8 +991,6 @@
     <ClCompile Include="OSVersion.cpp" />
     <ClCompile Include="StringUtils.cpp" />
     <ClCompile Include="System\Display.cpp" />
-    <ClCompile Include="System\OSD.cpp" />
-    <ClCompile Include="System\Request.cpp" />
     <ClCompile Include="Thread\ParallelLoop.cpp" />
     <ClCompile Include="Thread\ThreadManager.cpp" />
     <ClCompile Include="Thread\ThreadUtil.cpp" />
@@ -992,7 +998,6 @@
     <ClCompile Include="TimeUtil.cpp" />
     <ClCompile Include="UI\AsyncImageFileView.cpp" />
     <ClCompile Include="UI\Context.cpp" />
-    <ClCompile Include="UI\IconCache.cpp" />
     <ClCompile Include="UI\PopupScreens.cpp" />
     <ClCompile Include="UI\Root.cpp" />
     <ClCompile Include="UI\Screen.cpp" />
@@ -1023,18 +1028,6 @@
       <Project>{f761046e-6c38-4428-a5f1-38391a37bb34}</Project>
     </ProjectReference>
   </ItemGroup>
-  <ItemGroup>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_astc.inc" />
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_astc_0_255.inc" />
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_atc_55.inc" />
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_atc_56.inc" />
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_bc7_m5_alpha.inc" />
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_bc7_m5_color.inc" />
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_dxt1_5.inc" />
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_dxt1_6.inc" />
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_pvrtc2_45.inc" />
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_pvrtc2_alpha_33.inc" />
-  </ItemGroup>
   <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
   <ImportGroup Label="ExtensionTargets">
   </ImportGroup>

Common/Common.vcxproj.filters

@@ -32,6 +32,15 @@
     <ClInclude Include="ArmCommon.h" />
     <ClInclude Include="BitSet.h" />
     <ClInclude Include="CodeBlock.h" />
+    <ClInclude Include="GL\GLInterface\EGL.h">
+      <Filter>GL\GLInterface</Filter>
+    </ClInclude>
+    <ClInclude Include="GL\GLInterface\EGLAndroid.h">
+      <Filter>GL\GLInterface</Filter>
+    </ClInclude>
+    <ClInclude Include="GL\GLInterfaceBase.h">
+      <Filter>GL</Filter>
+    </ClInclude>
     <ClInclude Include="GraphicsContext.h" />
     <ClInclude Include="DbgNew.h" />
     <ClInclude Include="OSVersion.h" />
@@ -167,6 +176,9 @@
     <ClInclude Include="File\VFS\VFS.h">
       <Filter>File\VFS</Filter>
     </ClInclude>
+    <ClInclude Include="File\VFS\AssetReader.h">
+      <Filter>File\VFS</Filter>
+    </ClInclude>
     <ClInclude Include="Data\Format\IniFile.h">
       <Filter>Data\Format</Filter>
     </ClInclude>
@@ -440,6 +452,9 @@
     <ClInclude Include="Render\ManagedTexture.h">
       <Filter>Render</Filter>
     </ClInclude>
+    <ClInclude Include="GPU\MiscTypes.h">
+      <Filter>GPU</Filter>
+    </ClInclude>
     <ClInclude Include="GPU\Vulkan\VulkanFramebuffer.h">
       <Filter>GPU\Vulkan</Filter>
     </ClInclude>
@@ -449,75 +464,6 @@
     <ClInclude Include="UI\PopupScreens.h">
       <Filter>UI</Filter>
     </ClInclude>
-    <ClInclude Include="GPU\OpenGL\GLFrameData.h">
-      <Filter>GPU\OpenGL</Filter>
-    </ClInclude>
-    <ClInclude Include="File\VFS\DirectoryReader.h">
-      <Filter>File\VFS</Filter>
-    </ClInclude>
-    <ClInclude Include="File\VFS\ZipFileReader.h">
-      <Filter>File\VFS</Filter>
-    </ClInclude>
-    <ClInclude Include="Data\Format\DDSLoad.h">
-      <Filter>Data\Format</Filter>
-    </ClInclude>
-    <ClInclude Include="..\ext\basis_universal\basisu.h">
-      <Filter>ext\basis_universal</Filter>
-    </ClInclude>
-    <ClInclude Include="..\ext\basis_universal\basisu_containers.h">
-      <Filter>ext\basis_universal</Filter>
-    </ClInclude>
-    <ClInclude Include="..\ext\basis_universal\basisu_containers_impl.h">
-      <Filter>ext\basis_universal</Filter>
-    </ClInclude>
-    <ClInclude Include="..\ext\basis_universal\basisu_file_headers.h">
-      <Filter>ext\basis_universal</Filter>
-    </ClInclude>
-    <ClInclude Include="..\ext\basis_universal\basisu_transcoder.h">
-      <Filter>ext\basis_universal</Filter>
-    </ClInclude>
-    <ClInclude Include="..\ext\basis_universal\basisu_transcoder_internal.h">
-      <Filter>ext\basis_universal</Filter>
-    </ClInclude>
-    <ClInclude Include="..\ext\basis_universal\basisu_transcoder_uastc.h">
-      <Filter>ext\basis_universal</Filter>
-    </ClInclude>
-    <ClInclude Include="System\Request.h">
-      <Filter>System</Filter>
-    </ClInclude>
-    <ClInclude Include="File\AndroidContentURI.h">
-      <Filter>File</Filter>
-    </ClInclude>
-    <ClInclude Include="GPU\OpenGL\GLMemory.h">
-      <Filter>GPU\OpenGL</Filter>
-    </ClInclude>
-    <ClInclude Include="GPU\GPUBackendCommon.h">
-      <Filter>GPU</Filter>
-    </ClInclude>
-    <ClInclude Include="GPU\MiscTypes.h">
-      <Filter>GPU</Filter>
-    </ClInclude>
-    <ClInclude Include="UI\IconCache.h">
-      <Filter>UI</Filter>
-    </ClInclude>
-    <ClInclude Include="System\OSD.h">
-      <Filter>System</Filter>
-    </ClInclude>
-    <ClInclude Include="Net\HTTPRequest.h">
-      <Filter>Net</Filter>
-    </ClInclude>
-    <ClInclude Include="..\ext\naett\naett.h">
-      <Filter>ext\naett</Filter>
-    </ClInclude>
-    <ClInclude Include="Net\HTTPNaettRequest.h">
-      <Filter>Net</Filter>
-    </ClInclude>
-    <ClInclude Include="Render\Text\draw_text_sdl.h">
-      <Filter>Render\Text</Filter>
-    </ClInclude>
-    <ClInclude Include="GPU\Vulkan\VulkanDescSet.h">
-      <Filter>GPU\Vulkan</Filter>
-    </ClInclude>
   </ItemGroup>
   <ItemGroup>
     <ClCompile Include="ABI.cpp" />
@@ -544,6 +490,15 @@
     </ClCompile>
     <ClCompile Include="MipsEmitter.cpp" />
     <ClCompile Include="Arm64Emitter.cpp" />
+    <ClCompile Include="GL\GLInterface\EGL.cpp">
+      <Filter>GL\GLInterface</Filter>
+    </ClCompile>
+    <ClCompile Include="GL\GLInterface\EGLAndroid.cpp">
+      <Filter>GL\GLInterface</Filter>
+    </ClCompile>
+    <ClCompile Include="GL\GLInterface\GLInterface.cpp">
+      <Filter>GL\GLInterface</Filter>
+    </ClCompile>
     <ClCompile Include="MemArenaPosix.cpp" />
     <ClCompile Include="MemArenaWin32.cpp" />
     <ClCompile Include="MemArenaAndroid.cpp" />
@@ -670,6 +625,9 @@
     <ClCompile Include="File\VFS\VFS.cpp">
       <Filter>File\VFS</Filter>
     </ClCompile>
+    <ClCompile Include="File\VFS\AssetReader.cpp">
+      <Filter>File\VFS</Filter>
+    </ClCompile>
     <ClCompile Include="Data\Format\IniFile.cpp">
       <Filter>Data\Format</Filter>
     </ClCompile>
@@ -878,7 +836,6 @@
       <Filter>GPU\Vulkan</Filter>
     </ClCompile>
     <ClCompile Include="RiscVEmitter.cpp" />
-    <ClCompile Include="LoongArchCPUDetect.cpp" />
     <ClCompile Include="GPU\Vulkan\VulkanFrameData.cpp">
       <Filter>GPU\Vulkan</Filter>
     </ClCompile>
@@ -921,62 +878,17 @@
     <ClCompile Include="UI\PopupScreens.cpp">
       <Filter>UI</Filter>
     </ClCompile>
-    <ClCompile Include="GPU\OpenGL\GLFrameData.cpp">
-      <Filter>GPU\OpenGL</Filter>
-    </ClCompile>
-    <ClCompile Include="File\VFS\DirectoryReader.cpp">
-      <Filter>File\VFS</Filter>
-    </ClCompile>
-    <ClCompile Include="File\VFS\ZipFileReader.cpp">
-      <Filter>File\VFS</Filter>
-    </ClCompile>
-    <ClCompile Include="Data\Format\DDSLoad.cpp">
-      <Filter>Data\Format</Filter>
-    </ClCompile>
-    <ClCompile Include="..\ext\basis_universal\basisu_transcoder.cpp">
-      <Filter>ext\basis_universal</Filter>
-    </ClCompile>
-    <ClCompile Include="System\Request.cpp">
-      <Filter>System</Filter>
-    </ClCompile>
-    <ClCompile Include="File\AndroidContentURI.cpp">
-      <Filter>File</Filter>
-    </ClCompile>
-    <ClCompile Include="GPU\OpenGL\GLMemory.cpp">
-      <Filter>GPU\OpenGL</Filter>
-    </ClCompile>
-    <ClCompile Include="Data\Collections\FastVec.h">
-      <Filter>Data\Collections</Filter>
-    </ClCompile>
-    <ClCompile Include="GPU\GPUBackendCommon.cpp">
-      <Filter>GPU</Filter>
-    </ClCompile>
-    <ClCompile Include="UI\IconCache.cpp">
-      <Filter>UI</Filter>
-    </ClCompile>
-    <ClCompile Include="System\OSD.cpp">
-      <Filter>System</Filter>
-    </ClCompile>
-    <ClCompile Include="Net\HTTPRequest.cpp">
-      <Filter>Net</Filter>
-    </ClCompile>
-    <ClCompile Include="..\ext\naett\naett.c">
-      <Filter>ext\naett</Filter>
-    </ClCompile>
-    <ClCompile Include="Net\HTTPNaettRequest.cpp">
-      <Filter>Net</Filter>
-    </ClCompile>
-    <ClCompile Include="Render\Text\draw_text_sdl.cpp">
-      <Filter>Render\Text</Filter>
-    </ClCompile>
-    <ClCompile Include="GPU\Vulkan\VulkanDescSet.cpp">
-      <Filter>GPU\Vulkan</Filter>
-    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <Filter Include="Crypto">
       <UniqueIdentifier>{1b593f03-7b28-4707-9228-4981796f5589}</UniqueIdentifier>
     </Filter>
+    <Filter Include="GL">
+      <UniqueIdentifier>{2f2ca112-9e26-499e-9cb9-38a78b4ac09d}</UniqueIdentifier>
+    </Filter>
+    <Filter Include="GL\GLInterface">
+      <UniqueIdentifier>{2c723cf4-75b6-406a-90c0-ebb7a13ba476}</UniqueIdentifier>
+    </Filter>
     <Filter Include="Serialize">
       <UniqueIdentifier>{7be79ad5-3520-46a1-a370-dce2a943978c}</UniqueIdentifier>
     </Filter>
@@ -1073,51 +985,10 @@
     <Filter Include="VR">
       <UniqueIdentifier>{9d1c29fd-8ac7-4475-8ea6-c8c759b695fe}</UniqueIdentifier>
     </Filter>
-    <Filter Include="ext\basis_universal">
-      <UniqueIdentifier>{d6d5f6e0-1c72-496b-af11-6d52d5123033}</UniqueIdentifier>
-    </Filter>
-    <Filter Include="ext\naett">
-      <UniqueIdentifier>{34f45db9-5c08-49cb-b349-b9e760ce3213}</UniqueIdentifier>
-    </Filter>
-    <Filter Include="ext\libchdr">
-      <UniqueIdentifier>{b681797d-7747-487f-b448-5ef5b2d2805b}</UniqueIdentifier>
-    </Filter>
   </ItemGroup>
   <ItemGroup>
     <Text Include="..\ext\libpng17\CMakeLists.txt">
       <Filter>ext\libpng17</Filter>
     </Text>
   </ItemGroup>
-  <ItemGroup>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_astc.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_astc_0_255.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_atc_55.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_atc_56.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_bc7_m5_alpha.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_bc7_m5_color.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_dxt1_5.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_dxt1_6.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_pvrtc2_45.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-    <None Include="..\ext\basis_universal\basisu_transcoder_tables_pvrtc2_alpha_33.inc">
-      <Filter>ext\basis_universal</Filter>
-    </None>
-  </ItemGroup>
 </Project>

Common/CommonFuncs.h

@@ -30,17 +30,12 @@
 #include <unistd.h>
 #include <errno.h>
 
-#if (PPSSPP_ARCH(X86) || PPSSPP_ARCH(AMD64)) && !defined(__EMSCRIPTEN__)
+#if PPSSPP_ARCH(X86) || PPSSPP_ARCH(AMD64)
 #define Crash() {asm ("int $3");}
-#elif PPSSPP_PLATFORM(SWITCH)
-// TODO: Implement Crash() for Switch, lets not use breakpoint for the time being
-#define Crash() {*((volatile u32 *)0x0) = 0xDEADC0DE;}
 #elif PPSSPP_ARCH(ARM)
 #define Crash() {asm ("bkpt #0");}
 #elif PPSSPP_ARCH(ARM64)
 #define Crash() {asm ("brk #0");}
-#elif PPSSPP_ARCH(RISCV64)
-#define Crash() {asm ("ebreak");}
 #else
 #include <signal.h>
 #define Crash() {kill(getpid(), SIGINT);}

Common/CommonTypes.h

@@ -36,39 +36,6 @@ typedef signed __int64 s64;
 
 #else
 
-#ifdef __SWITCH__
-// Some HID conflicts
-#define KEY_UP PKEY_UP
-#define KEY_DOWN PKEY_DOWN
-// Other conflicts
-#define Event _Event
-#define Framebuffer _Framebuffer
-#define Waitable _Waitable
-#define ThreadContext _ThreadContext
-#include <switch.h>
-// Cleanup
-#undef KEY_UP
-#undef KEY_DOWN
-#undef Event
-#undef Framebuffer
-#undef Waitable
-#undef ThreadContext
-
-// Conflicting types with libnx
-#ifndef _u64
-#define u64 _u64
-#endif // _u64
-
-#ifndef s64
-#define s64 _s64
-#endif // _s64
-
-typedef unsigned char   u_char;
-typedef unsigned short  u_short;
-typedef unsigned int    u_int;
-typedef unsigned long   u_long;
-#endif // __SWITCH__
-
 typedef unsigned char u8;
 typedef unsigned short u16;
 typedef unsigned int u32;

Common/ConsoleListener.cpp

@@ -119,10 +119,7 @@ void ConsoleListener::Init(bool AutoOpen, int Width, int Height, const char *Tit
 	title_ = ConvertUTF8ToWString(Title);
 
 	if (AutoOpen)
-	{
 		Open();
-		bHidden = false;
-	}
 #endif
 }
 
@@ -358,7 +355,7 @@ void ConsoleListener::LogWriterThread()
 
 		for (char *Text = logLocal, *End = logLocal + logLocalSize; Text < End; )
 		{
-			LogLevel Level = LogLevel::LINFO;
+			LogTypes::LOG_LEVELS Level = LogTypes::LINFO;
 
 			char *next = (char *) memchr(Text + 1, '\033', End - Text);
 			size_t Len = next - Text;
@@ -367,7 +364,7 @@ void ConsoleListener::LogWriterThread()
 
 			if (Text[0] == '\033' && Text + 1 < End)
 			{
-				Level = (LogLevel)(Text[1] - '0');
+				Level = (LogTypes::LOG_LEVELS) (Text[1] - '0');
 				Len -= 2;
 				Text += 2;
 			}
@@ -384,7 +381,7 @@ void ConsoleListener::LogWriterThread()
 	delete [] logLocal;
 }
 
-void ConsoleListener::SendToThread(LogLevel Level, const char *Text)
+void ConsoleListener::SendToThread(LogTypes::LOG_LEVELS Level, const char *Text)
 {
 	// Oops, we're already quitting.  Just do nothing.
 	if (logPendingWritePos == (u32) -1)
@@ -462,7 +459,7 @@ void ConsoleListener::SendToThread(LogLevel Level, const char *Text)
 	SetEvent(hTriggerEvent);
 }
 
-void ConsoleListener::WriteToConsole(LogLevel Level, const char *Text, size_t Len)
+void ConsoleListener::WriteToConsole(LogTypes::LOG_LEVELS Level, const char *Text, size_t Len)
 {
 	_dbg_assert_msg_(IsOpen(), "Don't call this before opening the console.");
 
@@ -479,20 +476,21 @@ void ConsoleListener::WriteToConsole(LogLevel Level, const char *Text, size_t Le
 	WORD Color;
 	static wchar_t tempBuf[2048];
 
-	switch (Level) {
+	switch (Level)
-	case LogLevel::LNOTICE: // light green
+	{
+	case NOTICE_LEVEL: // light green
 		Color = FOREGROUND_GREEN | FOREGROUND_INTENSITY;
 		break;
-	case LogLevel::LERROR: // light red
+	case ERROR_LEVEL: // light red
 		Color = FOREGROUND_RED | FOREGROUND_INTENSITY;
 		break;
-	case LogLevel::LWARNING: // light yellow
+	case WARNING_LEVEL: // light yellow
 		Color = FOREGROUND_RED | FOREGROUND_GREEN | FOREGROUND_INTENSITY;
 		break;
-	case LogLevel::LINFO: // cyan
+	case INFO_LEVEL: // cyan
 		Color = FOREGROUND_GREEN | FOREGROUND_BLUE | FOREGROUND_INTENSITY;
 		break;
-	case LogLevel::LDEBUG: // gray
+	case DEBUG_LEVEL: // gray
 		Color = FOREGROUND_INTENSITY;
 		break;
 	default: // off-white
@@ -592,7 +590,7 @@ void ConsoleListener::PixelSpace(int Left, int Top, int Width, int Height, bool
 	COORD Coo = GetCoordinates(OldCursor, LBufWidth);
 	SetConsoleCursorPosition(hConsole, Coo);
 
-	// if (SLog.length() > 0) Log(LogLevel::LNOTICE, SLog.c_str());
+	// if (SLog.length() > 0) Log(LogTypes::LNOTICE, SLog.c_str());
 
 	// Resize the window too
 	if (Resize) MoveWindow(GetConsoleWindow(), Left,Top, (Width + 100),Height, true);
@@ -614,16 +612,18 @@ void ConsoleListener::Log(const LogMessage &msg) {
 	char ColorAttr[16] = "";
 	char ResetAttr[16] = "";
 
-	if (bUseColor) {
+	if (bUseColor)
+	{
 		strcpy(ResetAttr, "\033[0m");
-		switch (msg.level) {
+		switch (msg.level)
-		case LogLevel::LNOTICE: // light green
+		{
+		case NOTICE_LEVEL: // light green
 			strcpy(ColorAttr, "\033[92m");
 			break;
-		case LogLevel::LERROR: // light red
+		case ERROR_LEVEL: // light red
 			strcpy(ColorAttr, "\033[91m");
 			break;
-		case LogLevel::LWARNING: // light yellow
+		case WARNING_LEVEL: // light yellow
 			strcpy(ColorAttr, "\033[93m");
 			break;
 		default:
@@ -653,3 +653,5 @@ void ConsoleListener::ClearScreen(bool Cursor)
 	if (Cursor) SetConsoleCursorPosition(hConsole, coordScreen); 
 #endif
 }
+
+

Common/ConsoleListener.h

@@ -54,8 +54,8 @@ private:
 
 	static unsigned int WINAPI RunThread(void *lpParam);
 	void LogWriterThread();
-	void SendToThread(LogLevel Level, const char *Text);
+	void SendToThread(LogTypes::LOG_LEVELS Level, const char *Text);
-	void WriteToConsole(LogLevel Level, const char *Text, size_t Len);
+	void WriteToConsole(LogTypes::LOG_LEVELS Level, const char *Text, size_t Len);
 
 	static int refCount;
 	static HANDLE hThread;

Common/Crypto/md5.cpp

@@ -62,7 +62,7 @@
 /*
  * MD5 context setup
  */
-void ppsspp_md5_starts( md5_context *ctx )
+void md5_starts( md5_context *ctx )
 {
     ctx->total[0] = 0;
     ctx->total[1] = 0;
@@ -73,7 +73,7 @@ void ppsspp_md5_starts( md5_context *ctx )
     ctx->state[3] = 0x10325476;
 }
 
-static void ppsspp_md5_process( md5_context *ctx, unsigned char data[64] )
+static void md5_process( md5_context *ctx, unsigned char data[64] )
 {
     unsigned long X[16], A, B, C, D;
 
@@ -199,7 +199,7 @@ static void ppsspp_md5_process( md5_context *ctx, unsigned char data[64] )
 /*
  * MD5 process buffer
  */
-void ppsspp_md5_update( md5_context *ctx, unsigned char *input, int ilen )
+void md5_update( md5_context *ctx, unsigned char *input, int ilen )
 {
     int fill;
     unsigned long left;
@@ -220,7 +220,7 @@ void ppsspp_md5_update( md5_context *ctx, unsigned char *input, int ilen )
     {
         memcpy( (void *) (ctx->buffer + left),
                 (void *) input, fill );
-        ppsspp_md5_process( ctx, ctx->buffer );
+        md5_process( ctx, ctx->buffer );
         input += fill;
         ilen  -= fill;
         left = 0;
@@ -228,7 +228,7 @@ void ppsspp_md5_update( md5_context *ctx, unsigned char *input, int ilen )
 
     while( ilen >= 64 )
     {
-        ppsspp_md5_process( ctx, input );
+        md5_process( ctx, input );
         input += 64;
         ilen  -= 64;
     }
@@ -251,7 +251,7 @@ static const unsigned char md5_padding[64] =
 /*
  * MD5 final digest
  */
-void ppsspp_md5_finish( md5_context *ctx, unsigned char output[16] )
+void md5_finish( md5_context *ctx, unsigned char output[16] )
 {
     unsigned long last, padn;
     unsigned long high, low;
@@ -267,8 +267,8 @@ void ppsspp_md5_finish( md5_context *ctx, unsigned char output[16] )
     last = ctx->total[0] & 0x3F;
     padn = ( last < 56 ) ? ( 56 - last ) : ( 120 - last );
 
-    ppsspp_md5_update( ctx, (unsigned char *) md5_padding, padn );
+    md5_update( ctx, (unsigned char *) md5_padding, padn );
-    ppsspp_md5_update( ctx, msglen, 8 );
+    md5_update( ctx, msglen, 8 );
 
     PUT_ULONG_LE( ctx->state[0], output,  0 );
     PUT_ULONG_LE( ctx->state[1], output,  4 );
@@ -279,13 +279,13 @@ void ppsspp_md5_finish( md5_context *ctx, unsigned char output[16] )
 /*
  * output = MD5( input buffer )
  */
-void ppsspp_md5( unsigned char *input, int ilen, unsigned char output[16] )
+void md5( unsigned char *input, int ilen, unsigned char output[16] )
 {
     md5_context ctx;
 
-    ppsspp_md5_starts( &ctx );
+    md5_starts( &ctx );
-    ppsspp_md5_update( &ctx, input, ilen );
+    md5_update( &ctx, input, ilen );
-    ppsspp_md5_finish( &ctx, output );
+    md5_finish( &ctx, output );
 
     memset( &ctx, 0, sizeof( md5_context ) );
 }
@@ -293,14 +293,14 @@ void ppsspp_md5( unsigned char *input, int ilen, unsigned char output[16] )
 /*
  * MD5 HMAC context setup
  */
-void ppsspp_md5_hmac_starts( md5_context *ctx, unsigned char *key, int keylen )
+void md5_hmac_starts( md5_context *ctx, unsigned char *key, int keylen )
 {
     int i;
     unsigned char sum[16];
 
     if( keylen > 64 )
     {
-        ppsspp_md5( key, keylen, sum );
+        md5( key, keylen, sum );
         keylen = 16;
         key = sum;
     }
@@ -314,8 +314,8 @@ void ppsspp_md5_hmac_starts( md5_context *ctx, unsigned char *key, int keylen )
         ctx->opad[i] = (unsigned char)( ctx->opad[i] ^ key[i] );
     }
 
-    ppsspp_md5_starts( ctx );
+    md5_starts( ctx );
-    ppsspp_md5_update( ctx, ctx->ipad, 64 );
+    md5_update( ctx, ctx->ipad, 64 );
 
     memset( sum, 0, sizeof( sum ) );
 }
@@ -323,23 +323,23 @@ void ppsspp_md5_hmac_starts( md5_context *ctx, unsigned char *key, int keylen )
 /*
  * MD5 HMAC process buffer
  */
-void ppsspp_md5_hmac_update( md5_context *ctx, unsigned char *input, int ilen )
+void md5_hmac_update( md5_context *ctx, unsigned char *input, int ilen )
 {
-    ppsspp_md5_update( ctx, input, ilen );
+    md5_update( ctx, input, ilen );
 }
 
 /*
  * MD5 HMAC final digest
  */
-void ppsspp_md5_hmac_finish( md5_context *ctx, unsigned char output[16] )
+void md5_hmac_finish( md5_context *ctx, unsigned char output[16] )
 {
     unsigned char tmpbuf[16];
 
-    ppsspp_md5_finish( ctx, tmpbuf );
+    md5_finish( ctx, tmpbuf );
-    ppsspp_md5_starts( ctx );
+    md5_starts( ctx );
-    ppsspp_md5_update( ctx, ctx->opad, 64 );
+    md5_update( ctx, ctx->opad, 64 );
-    ppsspp_md5_update( ctx, tmpbuf, 16 );
+    md5_update( ctx, tmpbuf, 16 );
-    ppsspp_md5_finish( ctx, output );
+    md5_finish( ctx, output );
 
     memset( tmpbuf, 0, sizeof( tmpbuf ) );
 }
@@ -347,14 +347,14 @@ void ppsspp_md5_hmac_finish( md5_context *ctx, unsigned char output[16] )
 /*
  * output = HMAC-MD5( hmac key, input buffer )
  */
-void ppsspp_md5_hmac( unsigned char *key, int keylen, unsigned char *input, int ilen,
+void md5_hmac( unsigned char *key, int keylen, unsigned char *input, int ilen,
                unsigned char output[16] )
 {
     md5_context ctx;
 
-    ppsspp_md5_hmac_starts( &ctx, key, keylen );
+    md5_hmac_starts( &ctx, key, keylen );
-    ppsspp_md5_hmac_update( &ctx, input, ilen );
+    md5_hmac_update( &ctx, input, ilen );
-    ppsspp_md5_hmac_finish( &ctx, output );
+    md5_hmac_finish( &ctx, output );
 
     memset( &ctx, 0, sizeof( md5_context ) );
 }
@@ -464,7 +464,7 @@ static const unsigned char md5_hmac_test_sum[7][16] =
 /*
  * Checkup routine
  */
-int ppsspp_md5_self_test( int verbose )
+int md5_self_test( int verbose )
 {
     int i, buflen;
     unsigned char buf[1024];

Common/Crypto/md5.h

@@ -46,7 +46,7 @@ extern "C" {
  *
  * \param ctx      context to be initialized
  */
-void ppsspp_md5_starts( md5_context *ctx );
+void md5_starts( md5_context *ctx );
 
 /**
  * \brief          MD5 process buffer
@@ -55,7 +55,7 @@ void ppsspp_md5_starts( md5_context *ctx );
  * \param input    buffer holding the  data
  * \param ilen     length of the input data
  */
-void ppsspp_md5_update( md5_context *ctx, unsigned char *input, int ilen );
+void md5_update( md5_context *ctx, unsigned char *input, int ilen );
 
 /**
  * \brief          MD5 final digest
@@ -63,7 +63,7 @@ void ppsspp_md5_update( md5_context *ctx, unsigned char *input, int ilen );
  * \param ctx      MD5 context
  * \param output   MD5 checksum result
  */
-void ppsspp_md5_finish( md5_context *ctx, unsigned char output[16] );
+void md5_finish( md5_context *ctx, unsigned char output[16] );
 
 /**
  * \brief          Output = MD5( input buffer )
@@ -72,7 +72,7 @@ void ppsspp_md5_finish( md5_context *ctx, unsigned char output[16] );
  * \param ilen     length of the input data
  * \param output   MD5 checksum result
  */
-void ppsspp_md5( unsigned char *input, int ilen, unsigned char output[16] );
+void md5( unsigned char *input, int ilen, unsigned char output[16] );
 
 /**
  * \brief          Output = MD5( file contents )
@@ -83,7 +83,7 @@ void ppsspp_md5( unsigned char *input, int ilen, unsigned char output[16] );
  * \return         0 if successful, 1 if fopen failed,
  *                 or 2 if fread failed
  */
-int ppsspp_md5_file( char *path, unsigned char output[16] );
+int md5_file( char *path, unsigned char output[16] );
 
 /**
  * \brief          MD5 HMAC context setup
@@ -92,7 +92,7 @@ int ppsspp_md5_file( char *path, unsigned char output[16] );
  * \param key      HMAC secret key
  * \param keylen   length of the HMAC key
  */
-void ppsspp_md5_hmac_starts( md5_context *ctx, unsigned char *key, int keylen );
+void md5_hmac_starts( md5_context *ctx, unsigned char *key, int keylen );
 
 /**
  * \brief          MD5 HMAC process buffer
@@ -101,7 +101,7 @@ void ppsspp_md5_hmac_starts( md5_context *ctx, unsigned char *key, int keylen );
  * \param input    buffer holding the  data
  * \param ilen     length of the input data
  */
-void ppsspp_md5_hmac_update( md5_context *ctx, unsigned char *input, int ilen );
+void md5_hmac_update( md5_context *ctx, unsigned char *input, int ilen );
 
 /**
  * \brief          MD5 HMAC final digest
@@ -109,7 +109,7 @@ void ppsspp_md5_hmac_update( md5_context *ctx, unsigned char *input, int ilen );
  * \param ctx      HMAC context
  * \param output   MD5 HMAC checksum result
  */
-void ppsspp_md5_hmac_finish( md5_context *ctx, unsigned char output[16] );
+void md5_hmac_finish( md5_context *ctx, unsigned char output[16] );
 
 /**
  * \brief          Output = HMAC-MD5( hmac key, input buffer )
@@ -120,7 +120,7 @@ void ppsspp_md5_hmac_finish( md5_context *ctx, unsigned char output[16] );
  * \param ilen     length of the input data
  * \param output   HMAC-MD5 result
  */
-void ppsspp_md5_hmac( unsigned char *key, int keylen,
+void md5_hmac( unsigned char *key, int keylen,
                unsigned char *input, int ilen,
                unsigned char output[16] );
 
@@ -129,7 +129,7 @@ void ppsspp_md5_hmac( unsigned char *key, int keylen,
  *
  * \return         0 if successful, or 1 if the test failed
  */
-int ppsspp_md5_self_test( int verbose );
+int md5_self_test( int verbose );
 
 #ifdef __cplusplus
 }

Common/Data/Collections/FastVec.h

@@ -1,225 +0,0 @@
-#pragma once
-
-// Yet another replacement for std::vector, this time for use in graphics queues.
-// Its major difference is that you can append uninitialized structures and initialize them after.
-// This is not allows by std::vector but is very useful for our sometimes oversized unions.
-// Also, copies during resize are done by memcpy, not by any move constructor or similar.
-
-#include <cstdlib>
-#include <cstring>
-
-#ifdef _DEBUG
-#include "Common/Log.h"
-#endif
-
-template<class T>
-class FastVec {
-public:
-	FastVec() {}
-	FastVec(size_t initialCapacity) {
-		capacity_ = initialCapacity;
-		data_ = (T *)malloc(initialCapacity * sizeof(T));
-	}
-	~FastVec() { if (data_) free(data_); }
-
-	T &push_uninitialized() {
-		if (size_ < capacity_) {
-			size_++;
-			return data_[size_ - 1];
-		} else {
-			ExtendByOne();
-			return data_[size_ - 1];
-		}
-	}
-
-	void push_back(const T &t) {
-		T &dest = push_uninitialized();
-		dest = t;
-	}
-
-	// Move constructor
-	FastVec(FastVec &&other) {
-		data_ = other.data_;
-		size_ = other.size_;
-		capacity_ = other.capacity_;
-		other.data_ = nullptr;
-		other.size_ = 0;
-		other.capacity_ = 0;
-	}
-
-	FastVec &operator=(FastVec &&other)	{
-		if (this != &other) {
-			delete[] data_;
-			data_ = other.data_;
-			size_ = other.size_;
-			capacity_ = other.capacity_;
-			other.data_ = nullptr;
-			other.size_ = 0;
-			other.capacity_ = 0;
-		}
-		return *this;
-	}
-
-	// No copy constructor.
-	FastVec(const FastVec &other) = delete;
-	FastVec &operator=(const FastVec &other) = delete;
-
-	size_t size() const { return size_; }
-	size_t capacity() const { return capacity_; }
-	void clear() { size_ = 0; }
-	bool empty() const { return size_ == 0; }
-
-	const T *data() { return data_; }
-	T *begin() { return data_; }
-	T *end() { return data_ + size_; }
-	const T *begin() const { return data_; }
-	const T *end() const { return data_ + size_; }
-
-	// Out of bounds (past size() - 1) is undefined behavior.
-	T &operator[] (const size_t index) { return data_[index]; }
-	const T &operator[] (const size_t index) const { return data_[index]; }
-	T &at(const size_t index) { return data_[index]; }
-	const T &at(const size_t index) const { return data_[index]; }
-
-	// These two are invalid if empty().
-	const T &back() const { return (*this)[size() - 1]; }
-	const T &front() const { return (*this)[0]; }
-
-	// Limited functionality for inserts and similar, add as needed.
-	T &insert(T *iter) {
-		int pos = iter - data_;
-		ExtendByOne();
-		if (pos + 1 < (int)size_) {
-			memmove(data_ + pos + 1, data_ + pos, (size_ - pos) * sizeof(T));
-		}
-		return data_[pos];
-	}
-
-	void insert(T *destIter, const T *beginIter, const T *endIter) {
-		int pos = destIter - data_;
-		if (beginIter == endIter)
-			return;
-		size_t newItems = endIter - beginIter;
-		IncreaseCapacityTo(size_ + newItems);
-		memmove(data_ + pos + newItems, data_ + pos, (size_ - pos) * sizeof(T));
-		memcpy(data_ + pos, beginIter, newItems * sizeof(T));
-		size_ += newItems;
-	}
-
-	void resize(size_t size) {
-		if (size < size_) {
-			size_ = size;
-		} else {
-			// TODO
-		}
-	}
-
-	void reserve(size_t newCapacity) {
-		IncreaseCapacityTo(newCapacity);
-	}
-
-	void extend(const T *newData, size_t count) {
-		IncreaseCapacityTo(size_ + count);
-		memcpy(data_ + size_, newData, count * sizeof(T));
-		size_ += count;
-	}
-
-	T *extend_uninitialized(size_t count) {
-		size_t sz = size_;
-		if (size_ + count <= capacity_) {
-			size_ += count;
-			return &data_[sz];
-		} else {
-			size_t newCapacity = size_ + count * 2;  // Leave some extra room when growing in all cases
-			if (newCapacity < capacity_ * 2) {
-				// Standard amortized O(1).
-				newCapacity = capacity_ * 2;
-			}
-			IncreaseCapacityTo(newCapacity);
-			size_ += count;
-			return &data_[sz];
-		}
-	}
-
-	void LockCapacity() {
-#ifdef _DEBUG
-		capacityLocked_ = true;
-#endif
-	}
-
-private:
-	void IncreaseCapacityTo(size_t newCapacity) {
-#ifdef _DEBUG
-		_dbg_assert_(!capacityLocked_);
-#endif
-		if (newCapacity <= capacity_)
-			return;
-		T *oldData = data_;
-		data_ = (T *)malloc(sizeof(T) * newCapacity);
-		if (capacity_ != 0) {
-			memcpy(data_, oldData, sizeof(T) * size_);
-			free(oldData);
-		}
-		capacity_ = newCapacity;
-	}
-
-	void ExtendByOne() {
-		size_t newCapacity = capacity_ * 2;
-		if (newCapacity < 16) {
-			newCapacity = 16;
-		}
-		IncreaseCapacityTo(newCapacity);
-		size_++;
-	}
-
-	size_t size_ = 0;
-	size_t capacity_ = 0;
-	T *data_ = nullptr;
-#ifdef _DEBUG
-	bool capacityLocked_ = false;
-#endif
-};
-
-// Simple cyclical vector.
-template <class T, size_t size>
-class HistoryBuffer {
-public:
-	T &Add(size_t index) {
-#ifdef _DEBUG
-		_dbg_assert_((int64_t)index >= 0);
-#endif
-		if (index > maxIndex_)
-			maxIndex_ = index;
-		T &entry = data_[index % size];
-		entry = T{};
-		return entry;
-	}
-
-	const T &Back(size_t index) const {
-#ifdef _DEBUG
-		_dbg_assert_(index < maxIndex_ && index < size);
-#endif
-		return data_[(maxIndex_ - index) % size];
-	}
-
-	// Out of bounds (past size() - 1) is undefined behavior.
-	T &operator[] (const size_t index) {
-#ifdef _DEBUG
-		_dbg_assert_(index <= maxIndex_);
-#endif
-		return data_[index % size];
-	}
-	const T &operator[] (const size_t index) const {
-#ifdef _DEBUG
-		_dbg_assert_(index <= maxIndex_);
-#endif
-		return data_[index % size];
-	}
-	size_t MaxIndex() const {
-		return maxIndex_;
-	}
-
-private:
-	T data_[size]{};
-	size_t maxIndex_ = 0;
-};

Common/Data/Collections/FixedSizeQueue.h

@@ -30,11 +30,15 @@ template <class T, int N>
 class FixedSizeQueue {
 public:
 	FixedSizeQueue() {
+		// Allocate aligned memory, just because.
+		//int sizeInBytes = N * sizeof(T);
+		//storage_ = (T *)AllocateMemoryPages(sizeInBytes);
 		storage_ = new T[N];
 		clear();
 	}
 
 	~FixedSizeQueue() {
+		// FreeMemoryPages((void *)storage_, N * sizeof(T));
 		delete [] storage_;
 	}
 
@@ -222,3 +226,4 @@ private:
 	volatile int curReadBlock;
 	volatile int curWriteBlock;
 };
+

Common/Data/Collections/Hashmaps.h

@@ -29,7 +29,7 @@ enum class BucketState : uint8_t {
 // we always use very small values, so it's probably better to have them in the same
 // cache-line as the corresponding key.
 // Enforces that value are pointers to make sure that combined storage makes sense.
-template <class Key, class Value>
+template <class Key, class Value, Value NullValue>
 class DenseHashMap {
 public:
 	DenseHashMap(int initialCapacity) : capacity_(initialCapacity) {
@@ -37,47 +37,26 @@ public:
 		state.resize(initialCapacity);
 	}
 
-	// Returns true if the entry was found, and writes the entry to *value.
+	// Returns nullptr if no entry was found.
-	// Returns false and does not write to value if no entry was found.
+	Value Get(const Key &key) {
-	// Note that nulls can be stored.
-	bool Get(const Key &key, Value *value) const {
 		uint32_t mask = capacity_ - 1;
 		uint32_t pos = HashKey(key) & mask;
 		// No? Let's go into search mode. Linear probing.
 		uint32_t p = pos;
 		while (true) {
-			if (state[p] == BucketState::TAKEN && KeyEquals(key, map[p].key)) {
+			if (state[p] == BucketState::TAKEN && KeyEquals(key, map[p].key))
-				*value = map[p].value;
+				return map[p].value;
-				return true;
+			else if (state[p] == BucketState::FREE)
-			} else if (state[p] == BucketState::FREE) {
+				return NullValue;
-				return false;
-			}
 			p = (p + 1) & mask;  // If the state is REMOVED, we just keep on walking. 
 			if (p == pos) {
-				// We looped around the whole map.
 				_assert_msg_(false, "DenseHashMap: Hit full on Get()");
 			}
 		}
-		return false;
+		return NullValue;
 	}
 
-	// Only works if Value can be nullptr
+	// Returns false if we already had the key! Which is a bit different.
-	Value GetOrNull(const Key &key) const {
-		Value value;
-		if (Get(key, &value)) {
-			return value;
-		} else {
-			return (Value)nullptr;
-		}
-	}
-
-	bool ContainsKey(const Key &key) const {
-		// Slightly wasteful, though compiler might optimize it.
-		Value value;
-		return Get(key, &value);
-	}
-
-	// Asserts if we already had the key!
 	bool Insert(const Key &key, Value value) {
 		// Check load factor, resize if necessary. We never shrink.
 		if (count_ > capacity_ / 2) {
@@ -135,7 +114,6 @@ public:
 		return false;
 	}
 
-	// This will never crash if you call it without locking - but, the value might not be right.
 	size_t size() const {
 		return count_;
 	}
@@ -212,7 +190,7 @@ private:
 
 // Like the above, uses linear probing for cache-friendliness.
 // Does not perform hashing at all so expects well-distributed keys.
-template <class Value>
+template <class Value, Value NullValue>
 class PrehashMap {
 public:
 	PrehashMap(int initialCapacity) : capacity_(initialCapacity) {
@@ -221,24 +199,22 @@ public:
 	}
 
 	// Returns nullptr if no entry was found.
-	bool Get(uint32_t hash, Value *value) {
+	Value Get(uint32_t hash) {
 		uint32_t mask = capacity_ - 1;
 		uint32_t pos = hash & mask;
 		// No? Let's go into search mode. Linear probing.
 		uint32_t p = pos;
 		while (true) {
-			if (state[p] == BucketState::TAKEN && hash == map[p].hash) {
+			if (state[p] == BucketState::TAKEN && hash == map[p].hash)
-				*value = map[p].value;
+				return map[p].value;
-				return true;
+			else if (state[p] == BucketState::FREE)
-			} else if (state[p] == BucketState::FREE) {
+				return NullValue;
-				return false;
-			}
 			p = (p + 1) & mask;  // If the state is REMOVED, we just keep on walking. 
 			if (p == pos) {
 				_assert_msg_(false, "PrehashMap: Hit full on Get()");
 			}
 		}
-		return false;
+		return NullValue;
 	}
 
 	// Returns false if we already had the key! Which is a bit different.

Common/Data/Collections/TinySet.h

@@ -134,68 +134,3 @@ private:
 	T fastLookup_[MaxFastSize];
 	std::vector<T> *slowLookup_ = nullptr;
 };
-
-template <class T, int MaxSize>
-struct FixedTinyVec {
-	~FixedTinyVec() {}
-	// WARNING: Can fail if you exceed MaxSize!
-	inline bool push_back(const T &t) {
-		if (count_ < MaxSize) {
-			data_[count_++] = t;
-			return true;
-		} else {
-			return false;
-		}
-	}
-	// WARNING: Can fail if you exceed MaxSize!
-	inline T *add_back() {
-		if (count_ < MaxSize) {
-			return &data_[count_++];
-		}
-		return nullptr;
-	}
-	// Invalid if empty().
-	void pop_back() { count_--;	}
-
-	// Unlike TinySet, we can trivially support begin/end as pointers.
-	T *begin() { return data_; }
-	T *end() { return data_ + count_; }
-	const T *begin() const { return data_; }
-	const T *end() const { return data_ + count_; }
-
-	size_t capacity() const { return MaxSize; }
-	void clear() { count_ = 0; }
-	bool empty() const { return count_ == 0; }
-	size_t size() const { return count_; }
-
-	bool contains(T t) const {
-		for (int i = 0; i < count_; i++) {
-			if (data_[i] == t)
-				return true;
-		}
-		return false;
-	}
-
-	// Out of bounds (past size() - 1) is undefined behavior.
-	T &operator[] (const size_t index) { return data_[index]; }
-	const T &operator[] (const size_t index) const { return data_[index]; }
-
-	// These two are invalid if empty().
-	const T &back() const { return (*this)[size() - 1]; }
-	const T &front() const { return (*this)[0]; }
-
-	bool operator == (const FixedTinyVec<T, MaxSize> &other) const {
-		if (count_ != other.count_)
-			return false;
-		for (int i = 0; i < count_; i++) {
-			if (!(data_[i] == other.data_[i])) {
-				return false;
-			}
-		}
-		return true;
-	}
-
-private:
-	int count_ = 0;  // first in the struct just so it's more visible in the VS debugger.
-	T data_[MaxSize];
-};

Common/Data/Convert/ColorConv.cpp

@@ -617,7 +617,6 @@ void ConvertRGB565ToBGR565(u16 *dst, const u16 *src, u32 numPixels) {
 	u32 i = 0;
 #endif
 
-	// TODO: Add a 64-bit loop too.
 	const u32 *src32 = (const u32 *)src;
 	u32 *dst32 = (u32 *)dst;
 	for (; i < numPixels / 2; i++) {

Common/Data/Encoding/Utf8.cpp

@@ -219,15 +219,13 @@ int u8_strlen(const char *s)
 }
 
 /* reads the next utf-8 sequence out of a string, updating an index */
-uint32_t u8_nextchar(const char *s, int *index) {
+uint32_t u8_nextchar(const char *s, int *i) {
 	uint32_t ch = 0;
 	int sz = 0;
-	int i = *index;
 	do {
-		ch = (ch << 6) + (unsigned char)s[i++];
+		ch = (ch << 6) + (unsigned char)s[(*i)++];
 		sz++;
-	} while (s[i] && ((s[i]) & 0xC0) == 0x80);
+	} while (s[*i] && ((s[*i]) & 0xC0) == 0x80);
-	*index = i;
 	return ch - offsetsFromUTF8[sz - 1];
 }
 
@@ -428,17 +426,6 @@ int u8_is_locale_utf8(const char *locale)
   return 0;
 }
 
-bool AnyEmojiInString(const char *s, size_t byteCount) {
-	int i = 0;
-	while (i < byteCount) {
-		uint32_t c = u8_nextchar(s, &i);
-		if (CodepointIsProbablyEmoji(c)) {
-			return true;
-		}
-	}
-	return false;
-}
-
 int UTF8StringNonASCIICount(const char *utf8string) {
 	UTF8 utf(utf8string);
 	int count = 0;
@@ -571,12 +558,6 @@ std::u16string ConvertUTF8ToUCS2(const std::string &source) {
 	return dst;
 }
 
-std::string CodepointToUTF8(uint32_t codePoint) {
-	char temp[16]{};
-	UTF8::encode(temp, codePoint);
-	return std::string(temp);
-}
-
 #ifndef _WIN32
 
 // Replacements for the Win32 wstring functions. Not to be used from emulation code!

Common/Data/Encoding/Utf8.h

@@ -26,15 +26,6 @@ int u8_strlen(const char *s);
 void u8_inc(const char *s, int *i);
 void u8_dec(const char *s, int *i);
 
-inline bool CodepointIsProbablyEmoji(uint32_t c) {
-	// Original check was some ranges grabbed from https://stackoverflow.com/a/62898106.
-	// But let's just go with checking if outside the BMP, it's not a big deal if we accidentally
-	// switch to color when not needed if someone uses a weird glyph.
-	return c > 0xFFFF;
-}
-
-bool AnyEmojiInString(const char *s, size_t byteCount);
-
 class UTF8 {
 public:
 	static const uint32_t INVALID = (uint32_t)-1;
@@ -98,8 +89,6 @@ bool UTF8StringHasNonASCII(const char *utf8string);
 // Removes overlong encodings and similar.
 std::string SanitizeUTF8(const std::string &utf8string);
 
-std::string CodepointToUTF8(uint32_t codePoint);
-
 
 // UTF8 to Win32 UTF-16
 // Should be used when calling Win32 api calls

Common/Data/Format/DDSLoad.cpp

@@ -1,5 +0,0 @@
-#include "Common/Data/Format/DDSLoad.h"
-
-bool DetectDDSParams(const DDSHeader *header, DDSLoadInfo *info) {
-	return false;
-}

Common/Data/Format/DDSLoad.h

@@ -1,83 +0,0 @@
-#pragma once
-
-#include <cstdint>
-
-// DDSPixelFormat.dwFlags bits
-enum {
-	DDPF_ALPHAPIXELS = 1,   // Texture contains alpha data; dwRGBAlphaBitMask contains valid data.
-	DDPF_ALPHA       = 2,   // Used in some older DDS files for alpha channel only uncompressed data(dwRGBBitCount contains the alpha channel bitcount; dwABitMask contains valid data)
-	DDPF_FOURCC      = 4,   // Texture contains compressed RGB data; dwFourCC contains valid data.	0x4
-	DDPF_RGB         = 8,   // Texture contains uncompressed RGB data; dwRGBBitCount and the RGB masks(dwRBitMask, dwGBitMask, dwBBitMask) contain valid data.	0x40
-	DDPF_YUV         = 16,  //Used in some older DDS files for YUV uncompressed data(dwRGBBitCount contains the YUV bit count; dwRBitMask contains the Y mask, dwGBitMask contains the U mask, dwBBitMask contains the V mask)
-	DDPF_LUMINANCE   = 32,  // Used in some older DDS files for single channel color uncompressed data (dwRGBBitCount contains the luminance channel bit count; dwRBitMask contains the channel mask). Can be combined with DDPF_ALPHAPIXELS for a two channel DDS file.
-};
-
-// dwCaps members
-enum {
-	DDSCAPS_COMPLEX = 8,
-	DDSCAPS_MIPMAP = 0x400000,
-	DDSCAPS_TEXTURE = 0x1000,  // Required
-};
-
-// Not using any D3D headers here, this is cross platform and minimal.
-// Boiled down from the public documentation.
-struct DDSPixelFormat {
-	uint32_t dwSize;  // must be 32
-	uint32_t dwFlags;
-	uint32_t dwFourCC;
-	uint32_t dwRGBBitCount;
-	uint32_t dwRBitMask;
-	uint32_t dwGBitMask;
-	uint32_t dwBBitMask;
-	uint32_t dwABitMask;
-};
-
-struct DDSHeader {
-	uint32_t dwMagic;  // Magic is not technically part of the header struct but convenient to have here when reading the files.
-	uint32_t dwSize;  // must be 124
-	uint32_t dwFlags;
-	uint32_t dwHeight;
-	uint32_t dwWidth;
-	uint32_t dwPitchOrLinearSize;  // The pitch or number of bytes per scan line in an uncompressed texture; the total number of bytes in the top level texture for a compressed texture
-	uint32_t dwDepth; // we'll always use 1 here
-	uint32_t dwMipMapCount;
-	uint32_t dwReserved1[11];
-	DDSPixelFormat ddspf;
-	uint32_t dwCaps;
-	uint32_t dwCaps2; // nothing we care about
-	uint32_t dwCaps3; // unused
-	uint32_t dwCaps4; // unused
-	uint32_t dwReserved2;
-};
-
-// DDS header extension to handle resource arrays, DXGI pixel formats that don't map to the legacy Microsoft DirectDraw pixel format structures, and additional metadata.
-struct DDSHeaderDXT10 {
-	uint32_t dxgiFormat;
-	uint32_t resourceDimension;  // 1d = 2, 2d = 3, 3d = 4. very intuitive
-	uint32_t miscFlag;
-	uint32_t arraySize;  // we only support 1 here
-	uint32_t miscFlags2;  // sets alpha interpretation, let's not bother
-};
-
-// Simple DDS parser, suitable for texture replacement packs.
-// Doesn't actually load, only does some logic to fill out DDSLoadInfo so the caller can then
-// do the actual load with a simple series of memcpys or whatever is appropriate.
-struct DDSLoadInfo {
-	uint32_t bytesToCopy;
-};
-
-bool DetectDDSParams(const DDSHeader *header, DDSLoadInfo *info);
-
-// We use the Basis library for the actual reading, but before we do, we pre-scan using this, similarly to the png trick.
-struct KTXHeader {
-	uint8_t identifier[12];
-	uint32_t vkFormat;
-	uint32_t typeSize;
-	uint32_t pixelWidth;
-	uint32_t pixelHeight;
-	uint32_t pixelDepth;
-	uint32_t layerCount;
-	uint32_t faceCount;
-	uint32_t levelCount;
-	uint32_t supercompressionScheme;
-};

Common/Data/Format/IniFile.cpp

@@ -5,9 +5,6 @@
 #include <cstdlib>
 #include <cstdio>
 
-#include <inttypes.h>
-
-// Hm, what's this for?
 #ifndef _MSC_VER
 #include <strings.h>
 #endif
@@ -20,17 +17,17 @@
 #include <vector>
 
 #include "Common/Data/Format/IniFile.h"
-#include "Common/Data/Text/Parsers.h"
 #include "Common/File/VFS/VFS.h"
 #include "Common/File/FileUtil.h"
-#include "Common/Log.h"
+#include "Common/Data/Text/Parsers.h"
-#include "Common/Math/math_util.h"
+
+#ifdef _WIN32
+#include "Common/Data/Encoding/Utf8.h"
+#endif
 
 #include "Common/StringUtils.h"
 
-// This unescapes # signs.
+static bool ParseLineKey(const std::string &line, size_t &pos, std::string *keyOut) {
-// NOTE: These parse functions can make better use of the string_view - the pos argument should not be needed, for example.
-static bool ParseLineKey(std::string_view line, size_t &pos, std::string *keyOut) {
 	std::string key = "";
 
 	while (pos < line.size()) {
@@ -45,8 +42,7 @@ static bool ParseLineKey(std::string_view line, size_t &pos, std::string *keyOut
 			}
 
 			// Escaped.
-			key += line.substr(pos, next - pos - 1);
+			key += line.substr(pos, next - pos - 1) + "#";
-			key.push_back('#');
 			pos = next + 1;
 		} else if (line[next] == '=') {
 			// Hurray, done.
@@ -62,11 +58,11 @@ static bool ParseLineKey(std::string_view line, size_t &pos, std::string *keyOut
 	return true;
 }
 
-static bool ParseLineValue(std::string_view line, size_t &pos, std::string *valueOut) {
+static bool ParseLineValue(const std::string &line, size_t &pos, std::string *valueOut) {
 	std::string value = "";
 
-	std::string_view strippedLine = StripSpaces(line.substr(pos));
+	std::string strippedLine = StripSpaces(line.substr(pos));
-	if (strippedLine.size() >= 2 && strippedLine[0] == '"' && strippedLine[strippedLine.size() - 1] == '"') {
+	if (strippedLine[0] == '"' && strippedLine[strippedLine.size()-1] == '"') {
 		// Don't remove comment if is surrounded by " "
 		value += line.substr(pos);
 		pos = line.npos; // Won't enter the while below
@@ -86,8 +82,7 @@ static bool ParseLineValue(std::string_view line, size_t &pos, std::string *valu
 			break;
 		} else {
 			// Escaped.
-			value += line.substr(pos, next - pos - 1);
+			value += line.substr(pos, next - pos - 1) + "#";
-			value.push_back('#');
 			pos = next + 1;
 		}
 	}
@@ -99,7 +94,7 @@ static bool ParseLineValue(std::string_view line, size_t &pos, std::string *valu
 	return true;
 }
 
-static bool ParseLineComment(std::string_view line, size_t &pos, std::string *commentOut) {
+static bool ParseLineComment(const std::string& line, size_t &pos, std::string *commentOut) {
 	// Don't bother with anything if we don't need the comment data.
 	if (commentOut) {
 		// Include any whitespace/formatting in the comment.
@@ -120,7 +115,8 @@ static bool ParseLineComment(std::string_view line, size_t &pos, std::string *co
 	return true;
 }
 
-static bool ParseLine(std::string_view line, std::string* keyOut, std::string* valueOut, std::string* commentOut)
+// Ugh, this is ugly.
+static bool ParseLine(const std::string& line, std::string* keyOut, std::string* valueOut, std::string* commentOut)
 {
 	// Rules:
 	// 1. A line starting with ; is commented out.
@@ -144,7 +140,7 @@ static bool ParseLine(std::string_view line, std::string* keyOut, std::string* v
 	return true;
 }
 
-static std::string EscapeHash(std::string_view value) {
+static std::string EscapeComments(const std::string &value) {
 	std::string result = "";
 
 	for (size_t pos = 0; pos < value.size(); ) {
@@ -153,8 +149,7 @@ static std::string EscapeHash(std::string_view value) {
 			result += value.substr(pos);
 			pos = value.npos;
 		} else {
-			result += value.substr(pos, next - pos);
+			result += value.substr(pos, next - pos) + "\\#";
-			result += "\\#";
 			pos = next + 1;
 		}
 	}
@@ -162,56 +157,34 @@ static std::string EscapeHash(std::string_view value) {
 	return result;
 }
 
-void ParsedIniLine::ParseFrom(std::string_view line) {
-	line = StripSpaces(line);
-	if (line.empty()) {
-		key.clear();
-		value.clear();
-		comment.clear();
-	} else if (line[0] == '#') {
-		key.clear();
-		value.clear();
-		comment = line;
-	} else {
-		ParseLine(line, &key, &value, &comment);
-	}
-}
-
-void ParsedIniLine::Reconstruct(std::string *output) const {
-	if (!key.empty()) {
-		*output = EscapeHash(key) + " = " + EscapeHash(value) + comment;
-	} else {
-		*output = comment;
-	}
-}
-
 void Section::Clear() {
-	lines_.clear();
+	lines.clear();
 }
 
-bool Section::GetKeys(std::vector<std::string> &keys) const {
+std::string* Section::GetLine(const char* key, std::string* valueOut, std::string* commentOut)
-	keys.clear();
+{
-	for (auto liter = lines_.begin(); liter != lines_.end(); ++liter) {
+	for (std::vector<std::string>::iterator iter = lines.begin(); iter != lines.end(); ++iter)
-		if (!liter->Key().empty())
+	{
-			keys.push_back(std::string(liter->Key()));
+		std::string& line = *iter;
-	}
+		std::string lineKey;
-	return true;
+		ParseLine(line, &lineKey, valueOut, commentOut);
-}
+		if (!strcasecmp(lineKey.c_str(), key))
-
-ParsedIniLine *Section::GetLine(const char *key) {
-	for (auto &line : lines_) {
-		if (equalsNoCase(line.Key(), key))
 			return &line;
 	}
-	return nullptr;
+	return 0;
 }
 
-const ParsedIniLine *Section::GetLine(const char* key) const {
+const std::string* Section::GetLine(const char* key, std::string* valueOut, std::string* commentOut) const
-	for (auto &line : lines_) {
+{
-		if (equalsNoCase(line.Key(), key))
+	for (std::vector<std::string>::const_iterator iter = lines.begin(); iter != lines.end(); ++iter)
+	{
+		const std::string& line = *iter;
+		std::string lineKey;
+		ParseLine(line, &lineKey, valueOut, commentOut);
+		if (!strcasecmp(lineKey.c_str(), key))
 			return &line;
 	}
-	return nullptr;
+	return 0;
 }
 
 void Section::Set(const char* key, uint32_t newValue) {
@@ -219,11 +192,10 @@ void Section::Set(const char* key, uint32_t newValue) {
 }
 
 void Section::Set(const char* key, uint64_t newValue) {
-	Set(key, StringFromFormat("0x%016" PRIx64, newValue).c_str());
+	Set(key, StringFromFormat("0x%016lx", newValue).c_str());
 }
 
 void Section::Set(const char* key, float newValue) {
-	_dbg_assert_(!my_isnanorinf(newValue));
 	Set(key, StringFromFormat("%f", newValue).c_str());
 }
 
@@ -235,13 +207,19 @@ void Section::Set(const char* key, int newValue) {
 	Set(key, StringFromInt(newValue).c_str());
 }
 
-void Section::Set(const char* key, const char* newValue) {
+void Section::Set(const char* key, const char* newValue)
-	ParsedIniLine *line = GetLine(key);
+{
-	if (line) {
+	std::string value, commented;
-		line->SetValue(newValue);
+	std::string* line = GetLine(key, &value, &commented);
-	} else {
+	if (line)
+	{
+		// Change the value - keep the key and comment
+		*line = StripSpaces(key) + " = " + EscapeComments(newValue) + commented;
+	}
+	else
+	{
 		// The key did not already exist in this section - let's add it.
-		lines_.emplace_back(ParsedIniLine(key, newValue));
+		lines.emplace_back(std::string(key) + " = " + EscapeComments(newValue));
 	}
 }
 
@@ -253,15 +231,16 @@ void Section::Set(const char* key, const std::string& newValue, const std::strin
 		Delete(key);
 }
 
-bool Section::Get(const char* key, std::string* value, const char* defaultValue) const {
+bool Section::Get(const char* key, std::string* value, const char* defaultValue) const
-	const ParsedIniLine *line = GetLine(key);
+{
-	if (!line) {
+	const std::string* line = GetLine(key, value, 0);
-		if (defaultValue) {
+	if (!line)
+	{
+		if (defaultValue)
+		{
 			*value = defaultValue;
 		}
 		return false;
-	} else {
-		*value = line->Value();
 	}
 	return true;
 }
@@ -306,7 +285,7 @@ void Section::Set(const char* key, const std::vector<std::string>& newValues)
 }
 
 void Section::AddComment(const std::string &comment) {
-	lines_.emplace_back(ParsedIniLine::CommentOnly("# " + comment));
+	lines.emplace_back("# " + comment);
 }
 
 bool Section::Get(const char* key, std::vector<std::string>& values) const
@@ -341,7 +320,7 @@ bool Section::Get(const char* key, int* value, int defaultValue) const
 {
 	std::string temp;
 	bool retval = Get(key, &temp, 0);
-	if (retval && TryParse(temp, value))
+	if (retval && TryParse(temp.c_str(), value))
 		return true;
 	*value = defaultValue;
 	return false;
@@ -371,7 +350,7 @@ bool Section::Get(const char* key, bool* value, bool defaultValue) const
 {
 	std::string temp;
 	bool retval = Get(key, &temp, 0);
-	if (retval && TryParse(temp, value))
+	if (retval && TryParse(temp.c_str(), value))
 		return true;
 	*value = defaultValue;
 	return false;
@@ -381,7 +360,7 @@ bool Section::Get(const char* key, float* value, float defaultValue) const
 {
 	std::string temp;
 	bool retval = Get(key, &temp, 0);
-	if (retval && TryParse(temp, value))
+	if (retval && TryParse(temp.c_str(), value))
 		return true;
 	*value = defaultValue;
 	return false;
@@ -391,35 +370,46 @@ bool Section::Get(const char* key, double* value, double defaultValue) const
 {
 	std::string temp;
 	bool retval = Get(key, &temp, 0);
-	if (retval && TryParse(temp, value))
+	if (retval && TryParse(temp.c_str(), value))
 		return true;
 	*value = defaultValue;
 	return false;
 }
 
-bool Section::Exists(const char *key) const {
+bool Section::Exists(const char *key) const
-	for (auto &line : lines_) {
+{
-		if (equalsNoCase(key, line.Key()))
+	for (std::vector<std::string>::const_iterator iter = lines.begin(); iter != lines.end(); ++iter)
+	{
+		std::string lineKey;
+		ParseLine(*iter, &lineKey, NULL, NULL);
+		if (!strcasecmp(lineKey.c_str(), key))
 			return true;
 	}
 	return false;
 }
 
-std::map<std::string, std::string> Section::ToMap() const {
+std::map<std::string, std::string> Section::ToMap() const
+{
 	std::map<std::string, std::string> outMap;
-	for (auto &line : lines_) {
+	for (std::vector<std::string>::const_iterator iter = lines.begin(); iter != lines.end(); ++iter)
-		if (!line.Key().empty()) {
+	{
-			outMap[std::string(line.Key())] = line.Value();
+		std::string lineKey, lineValue;
+		if (ParseLine(*iter, &lineKey, &lineValue, NULL)) {
+			outMap[lineKey] = lineValue;
 		}
 	}
 	return outMap;
 }
 
-bool Section::Delete(const char *key) {
+
-	ParsedIniLine *line = GetLine(key);
+bool Section::Delete(const char *key)
-	for (auto liter = lines_.begin(); liter != lines_.end(); ++liter) {
+{
-		if (line == &*liter) {
+	std::string* line = GetLine(key, 0, 0);
-			lines_.erase(liter);
+	for (std::vector<std::string>::iterator liter = lines.begin(); liter != lines.end(); ++liter)
+	{
+		if (line == &*liter)
+		{
+			lines.erase(liter);
 			return true;
 		}
 	}
@@ -428,36 +418,42 @@ bool Section::Delete(const char *key) {
 
 // IniFile
 
-const Section* IniFile::GetSection(const char* sectionName) const {
+const Section* IniFile::GetSection(const char* sectionName) const
-	for (const auto &iter : sections)
+{
+	for (std::vector<Section>::const_iterator iter = sections.begin(); iter != sections.end(); ++iter)
 		if (!strcasecmp(iter->name().c_str(), sectionName))
-			return iter.get();
+			return (&(*iter));
-	return nullptr;
+	return 0;
 }
 
-Section* IniFile::GetSection(const char* sectionName) {
+Section* IniFile::GetSection(const char* sectionName)
-	for (const auto &iter : sections)
+{
+	for (std::vector<Section>::iterator iter = sections.begin(); iter != sections.end(); ++iter)
 		if (!strcasecmp(iter->name().c_str(), sectionName))
-			return iter.get();
+			return (&(*iter));
-	return nullptr;
+	return 0;
 }
 
-Section* IniFile::GetOrCreateSection(const char* sectionName) {
+Section* IniFile::GetOrCreateSection(const char* sectionName)
+{
 	Section* section = GetSection(sectionName);
-	if (!section) {
+	if (!section)
-		sections.push_back(std::unique_ptr<Section>(new Section(sectionName)));
+	{
-		section = sections.back().get();
+		sections.push_back(Section(sectionName));
+		section = &sections[sections.size() - 1];
 	}
 	return section;
 }
 
-bool IniFile::DeleteSection(const char* sectionName) {
+bool IniFile::DeleteSection(const char* sectionName)
+{
 	Section* s = GetSection(sectionName);
 	if (!s)
 		return false;
-
+	for (std::vector<Section>::iterator iter = sections.begin(); iter != sections.end(); ++iter)
-	for (auto iter = sections.begin(); iter != sections.end(); ++iter) {
+	{
-		if (iter->get() == s) {
+		if (&(*iter) == s)
+		{
 			sections.erase(iter);
 			return true;
 		}
@@ -465,21 +461,35 @@ bool IniFile::DeleteSection(const char* sectionName) {
 	return false;
 }
 
-bool IniFile::Exists(const char* sectionName, const char* key) const {
+bool IniFile::Exists(const char* sectionName, const char* key) const
+{
 	const Section* section = GetSection(sectionName);
 	if (!section)
 		return false;
 	return section->Exists(key);
 }
 
-bool IniFile::DeleteKey(const char* sectionName, const char* key) {
+void IniFile::SetLines(const char* sectionName, const std::vector<std::string> &lines)
+{
+	Section* section = GetOrCreateSection(sectionName);
+	section->lines.clear();
+	for (std::vector<std::string>::const_iterator iter = lines.begin(); iter != lines.end(); ++iter)
+	{
+		section->lines.push_back(*iter);
+	}
+}
+
+bool IniFile::DeleteKey(const char* sectionName, const char* key)
+{
 	Section* section = GetSection(sectionName);
 	if (!section)
 		return false;
-	ParsedIniLine *line = section->GetLine(key);
+	std::string* line = section->GetLine(key, 0, 0);
-	for (auto liter = section->lines_.begin(); liter != section->lines_.end(); ++liter) {
+	for (std::vector<std::string>::iterator liter = section->lines.begin(); liter != section->lines.end(); ++liter)
-		if (line == &(*liter)) {
+	{
-			section->lines_.erase(liter);
+		if (line == &(*liter))
+		{
+			section->lines.erase(liter);
 			return true;
 		}
 	}
@@ -487,13 +497,55 @@ bool IniFile::DeleteKey(const char* sectionName, const char* key) {
 }
 
 // Return a list of all keys in a section
-bool IniFile::GetKeys(const char* sectionName, std::vector<std::string>& keys) const {
+bool IniFile::GetKeys(const char* sectionName, std::vector<std::string>& keys) const
-	const Section *section = GetSection(sectionName);
+{
+	const Section* section = GetSection(sectionName);
 	if (!section)
 		return false;
-	return section->GetKeys(keys);
+	keys.clear();
+	for (std::vector<std::string>::const_iterator liter = section->lines.begin(); liter != section->lines.end(); ++liter)
+	{
+		std::string key;
+		ParseLine(*liter, &key, 0, 0);
+		if (!key.empty())
+			keys.push_back(key);
+	}
+	return true;
 }
 
+// Return a list of all lines in a section
+bool IniFile::GetLines(const char* sectionName, std::vector<std::string>& lines, const bool remove_comments) const
+{
+	const Section* section = GetSection(sectionName);
+	if (!section)
+		return false;
+
+	lines.clear();
+	for (std::vector<std::string>::const_iterator iter = section->lines.begin(); iter != section->lines.end(); ++iter)
+	{
+		std::string line = StripSpaces(*iter);
+
+		if (remove_comments)
+		{
+			int commentPos = (int)line.find('#');
+			if (commentPos == 0)
+			{
+				continue;
+			}
+
+			if (commentPos != (int)std::string::npos)
+			{
+				line = StripSpaces(line.substr(0, commentPos));
+			}
+		}
+
+		lines.push_back(line);
+	}
+
+	return true;
+}
+
+
 void IniFile::SortSections()
 {
 	std::sort(sections.begin(), sections.end());
@@ -502,7 +554,7 @@ void IniFile::SortSections()
 bool IniFile::Load(const Path &path)
 {
 	sections.clear();
-	sections.push_back(std::unique_ptr<Section>(new Section("")));
+	sections.push_back(Section(""));
 	// first section consists of the comments before the first real section
 
 	// Open file
@@ -515,9 +567,9 @@ bool IniFile::Load(const Path &path)
 	return success;
 }
 
-bool IniFile::LoadFromVFS(VFSInterface &vfs, const std::string &filename) {
+bool IniFile::LoadFromVFS(const std::string &filename) {
 	size_t size;
-	uint8_t *data = vfs.ReadFile(filename.c_str(), &size);
+	uint8_t *data = VFSReadFile(filename.c_str(), &size);
 	if (!data)
 		return false;
 	std::string str((const char*)data, size);
@@ -530,10 +582,10 @@ bool IniFile::LoadFromVFS(VFSInterface &vfs, const std::string &filename) {
 bool IniFile::Load(std::istream &in) {
 	// Maximum number of letters in a line
 	static const int MAX_BYTES = 1024*32;
-	char *templine = new char[MAX_BYTES];  // avoid using up massive stack space
 
 	while (!(in.eof() || in.fail()))
 	{
+		char templine[MAX_BYTES];
 		in.getline(templine, MAX_BYTES);
 		std::string line = templine;
 
@@ -558,23 +610,20 @@ bool IniFile::Load(std::istream &in) {
 			if (sectionNameEnd != std::string::npos) {
 				// New section!
 				std::string sub = line.substr(1, sectionNameEnd - 1);
-				sections.push_back(std::unique_ptr<Section>(new Section(sub)));
+				sections.push_back(Section(sub));
 
 				if (sectionNameEnd + 1 < line.size()) {
-					sections.back()->comment = line.substr(sectionNameEnd + 1);
+					sections[sections.size() - 1].comment = line.substr(sectionNameEnd + 1);
 				}
 			} else {
 				if (sections.empty()) {
-					sections.push_back(std::unique_ptr<Section>(new Section("")));
+					sections.push_back(Section(""));
 				}
-				ParsedIniLine parsedLine;
+				sections[sections.size() - 1].lines.push_back(line);
-				parsedLine.ParseFrom(line);
-				sections.back()->lines_.push_back(parsedLine);
 			}
 		}
 	}
 
-	delete[] templine;
 	return true;
 }
 
@@ -589,14 +638,13 @@ bool IniFile::Save(const Path &filename)
 	// TODO: Do we still need this? It's annoying.
 	fprintf(file, "\xEF\xBB\xBF");
 
-	for (const auto &section : sections) {
+	for (const Section &section : sections) {
-		if (!section->name().empty() && (!section->lines_.empty() || !section->comment.empty())) {
+		if (!section.name().empty() && (!section.lines.empty() || !section.comment.empty())) {
-			fprintf(file, "[%s]%s\n", section->name().c_str(), section->comment.c_str());
+			fprintf(file, "[%s]%s\n", section.name().c_str(), section.comment.c_str());
 		}
-		for (const auto &line : section->lines_) {
+
-			std::string buffer;
+		for (const std::string &s : section.lines) {
-			line.Reconstruct(&buffer);
+			fprintf(file, "%s\n", s.c_str());
-			fprintf(file, "%s\n", buffer.c_str());
 		}
 	}
 

Common/Data/Format/IniFile.h

@@ -5,50 +5,12 @@
 #pragma once
 
 #include <istream>
-#include <memory>
 #include <map>
 #include <string>
-#include <string_view>
 #include <vector>
-#include <cstdint>
 
 #include "Common/File/Path.h"
 
-class VFSInterface;
-
-class ParsedIniLine {
-public:
-	ParsedIniLine() {}
-	ParsedIniLine(std::string_view key, std::string_view value) {
-		this->key = key;
-		this->value = value;
-	}
-	ParsedIniLine(std::string_view key, std::string_view value, std::string_view comment) {
-		this->key = key;
-		this->value = value;
-		this->comment = comment;
-	}
-	static ParsedIniLine CommentOnly(std::string_view comment) {
-		return ParsedIniLine(std::string_view(), std::string_view(), comment);
-	}
-
-	// Comments only come from "ParseFrom".
-	void ParseFrom(std::string_view line);
-	void Reconstruct(std::string *output) const;
-
-	// Having these as views allows a more efficient internal representation, like one joint string.
-	std::string_view Key() const { return key; }
-	std::string_view Value() const { return value; }
-	std::string_view Comment() const { return comment; }
-
-	void SetValue(std::string_view newValue) { value = newValue; }
-
-private:
-	std::string key;
-	std::string value;
-	std::string comment;
-};
-
 class Section {
 	friend class IniFile;
 
@@ -63,8 +25,8 @@ public:
 
 	std::map<std::string, std::string> ToMap() const;
 
-	ParsedIniLine *GetLine(const char *key);
+	std::string *GetLine(const char* key, std::string* valueOut, std::string* commentOut);
-	const ParsedIniLine *GetLine(const char *key) const;
+	const std::string *GetLine(const char* key, std::string* valueOut, std::string* commentOut) const;
 
 	void Set(const char* key, const char* newValue);
 	void Set(const char* key, const std::string& newValue, const std::string& defaultValue);
@@ -105,9 +67,6 @@ public:
 	bool Get(const char* key, double* value, double defaultValue = false) const;
 	bool Get(const char* key, std::vector<std::string>& values) const;
 
-	// Return a list of all keys in this section
-	bool GetKeys(std::vector<std::string> &keys) const;
-
 	bool operator < (const Section& other) const {
 		return name_ < other.name_;
 	}
@@ -117,7 +76,7 @@ public:
 	}
 
 protected:
-	std::vector<ParsedIniLine> lines_;
+	std::vector<std::string> lines;
 	std::string name_;
 	std::string comment;
 };
@@ -125,10 +84,12 @@ protected:
 class IniFile {
 public:
 	bool Load(const Path &path);
+	bool Load(const std::string &filename) { return Load(Path(filename)); }
 	bool Load(std::istream &istream);
-	bool LoadFromVFS(VFSInterface &vfs, const std::string &filename);
+	bool LoadFromVFS(const std::string &filename);
 
 	bool Save(const Path &path);
+	bool Save(const std::string &filename) { return Save(Path(filename)); }
 
 	// Returns true if key exists in section
 	bool Exists(const char* sectionName, const char* key) const;
@@ -173,19 +134,21 @@ public:
 
 	bool GetKeys(const char* sectionName, std::vector<std::string>& keys) const;
 
+	void SetLines(const char* sectionName, const std::vector<std::string> &lines);
+	bool GetLines(const char* sectionName, std::vector<std::string>& lines, const bool remove_comments = true) const;
+
 	bool DeleteKey(const char* sectionName, const char* key);
 	bool DeleteSection(const char* sectionName);
 
 	void SortSections();
-
+	std::vector<Section> &Sections() { return sections; }
-	std::vector<std::unique_ptr<Section>> &Sections() { return sections; }
 
 	bool HasSection(const char *section) { return GetSection(section) != 0; }
 
 	Section* GetOrCreateSection(const char* section);
 
 private:
-	std::vector<std::unique_ptr<Section>> sections;
+	std::vector<Section> sections;
 
 	const Section* GetSection(const char* section) const;
 	Section* GetSection(const char* section);

Common/Data/Format/JSONReader.cpp

@@ -8,7 +8,7 @@ namespace json {
 
 JsonReader::JsonReader(const std::string &filename) {
 	size_t buf_size;
-	buffer_ = (char *)g_VFS.ReadFile(filename.c_str(), &buf_size);
+	buffer_ = (char *)VFSReadFile(filename.c_str(), &buf_size);
 	if (buffer_) {
 		parse();
 	} else {

Common/Data/Format/JSONReader.h

@@ -1,5 +1,3 @@
-#pragma once
-
 #include <cstring>
 #include <string>
 #include <vector>

Common/Data/Format/JSONWriter.h

@@ -9,8 +9,6 @@
 //
 // Zero dependencies apart from stdlib (if you remove the vhjson usage.)
 
-#pragma once
-
 #include <string>
 #include <vector>
 #include <sstream>

Common/Data/Format/PNGLoad.cpp

@@ -59,14 +59,3 @@ int pngLoadPtr(const unsigned char *input_ptr, size_t input_len, int *pwidth, in
 	png_image_finish_read(&png, NULL, *image_data_ptr, stride, NULL);
 	return 1;
 }
-
-bool PNGHeaderPeek::IsValidPNGHeader() const {
-	if (magic != 0x474e5089 || ihdrTag != 0x52444849) {
-		return false;
-	}
-	// Reject crazy sized images, too.
-	if (Width() > 32768 && Height() > 32768) {
-		return false;
-	}
-	return true;
-}

Common/Data/Format/PNGLoad.h

@@ -1,8 +1,5 @@
-#pragma once
+#ifndef _PNG_LOAD_H
-
+#define _PNG_LOAD_H
-#include <cstdint>
-
-#include "Common/BitSet.h"
 
 // *image_data_ptr should be deleted with free()
 // return value of 1 == success.
@@ -12,22 +9,4 @@ int pngLoad(const char *file, int *pwidth,
 int pngLoadPtr(const unsigned  char *input_ptr, size_t input_len, int *pwidth,
             int *pheight, unsigned char **image_data_ptr);
 
-// PNG peeker - just read the start of a PNG straight into this struct, in order to
+#endif  // _PNG_LOAD_H
-// look at basic parameters like width and height. Note that while PNG is a chunk-based
-// format, the IHDR chunk is REQUIRED to be the first one, so this will work.
-// Does not handle Apple's weirdo extension CgBI. http://iphonedevwiki.net/index.php/CgBI_file_format
-// That should not be an issue.
-struct PNGHeaderPeek {
-	uint32_t magic;
-	uint32_t ignore0;
-	uint32_t ignore1;
-	uint32_t ihdrTag;
-	uint32_t be_width;  // big endian
-	uint32_t be_height;
-	uint8_t bitDepth;   // bits per channel, can be 1, 2, 4, 8, 16
-	uint8_t colorType;  // really, pixel format. 0 = grayscale, 2 = rgb, 3 = palette index, 4 = gray+alpha, 6 = rgba
-
-	bool IsValidPNGHeader() const;
-	int Width() const { return swap32(be_width); }
-	int Height() const { return swap32(be_height); }
-};

Common/Data/Format/ZIMLoad.cpp

@@ -126,7 +126,7 @@ int LoadZIMPtr(const uint8_t *zim, size_t datasize, int *width, int *height, int
 
 int LoadZIM(const char *filename, int *width, int *height, int *format, uint8_t **image) {
 	size_t size;
-	uint8_t *buffer = g_VFS.ReadFile(filename, &size);
+	uint8_t *buffer = VFSReadFile(filename, &size);
 	if (!buffer) {
 		ERROR_LOG(IO, "Couldn't read data for '%s'", filename);
 		return 0;

Common/Data/Text/I18n.cpp

@@ -1,88 +1,38 @@
-#include <cstring>
-
 #include "Common/Data/Text/I18n.h"
 #include "Common/Data/Format/IniFile.h"
 #include "Common/File/VFS/VFS.h"
-#include "Common/Log.h"
 
 #include "Common/StringUtils.h"
 
-static const char * const g_categoryNames[(size_t)I18NCat::CATEGORY_COUNT] = {
+I18NRepo i18nrepo;
-	"Audio",
-	"Controls",
-	"CwCheats",
-	"DesktopUI",
-	"Developer",
-	"Dialog",
-	"Error",
-	"Game",
-	"Graphics",
-	"InstallZip",
-	"KeyMapping",
-	"MainMenu",
-	"MainSettings",
-	"MappableControls",
-	"Networking",
-	"Pause",
-	"PostShaders",
-	"PSPCredits",
-	"MemStick",
-	"RemoteISO",
-	"Reporting",
-	"Savedata",
-	"Screen",
-	"Search",
-	"Store",
-	"SysInfo",
-	"System",
-	"TextureShaders",
-	"Themes",
-	"UI Elements",
-	"Upgrade",
-	"VR",
-	"Achievements",
-	"PSPSettings",
-};
 
-I18NRepo g_i18nrepo;
+I18NRepo::~I18NRepo() {
+	Clear();
+}
 
 std::string I18NRepo::LanguageID() {
 	return languageID_;
 }
 
-I18NRepo::I18NRepo() {
-	Clear();
-}
-
 void I18NRepo::Clear() {
 	std::lock_guard<std::mutex> guard(catsLock_);
-	for (auto &iter : cats_) {
+	for (auto iter = cats_.begin(); iter != cats_.end(); ++iter) {
-		// Initialize with empty categories, so that early lookups don't crash.
+		iter->second.reset();
-		iter = std::shared_ptr<I18NCategory>(new I18NCategory());
 	}
-}
+	cats_.clear();
-
-I18NCategory::I18NCategory(const Section &section) {
-	std::map<std::string, std::string> sectionMap = section.ToMap();
-	SetMap(sectionMap);
-}
-
-void I18NCategory::Clear() {
-	map_.clear();
-	missedKeyLog_.clear();
 }
 
 const char *I18NCategory::T(const char *key, const char *def) {
 	if (!key) {
 		return "ERROR";
 	}
-
 	// Replace the \n's with \\n's so that key values with newlines will be found correctly.
 	std::string modifiedKey = key;
 	modifiedKey = ReplaceAll(modifiedKey, "\n", "\\n");
 
 	auto iter = map_.find(modifiedKey);
 	if (iter != map_.end()) {
+//		INFO_LOG(SYSTEM, "translation key found in %s: %s", name_.c_str(), key);
 		return iter->second.text.c_str();
 	} else {
 		std::lock_guard<std::mutex> guard(missedKeyLock_);
@@ -90,6 +40,7 @@ const char *I18NCategory::T(const char *key, const char *def) {
 			missedKeyLog_[key] = def;
 		else
 			missedKeyLog_[key] = modifiedKey;
+//		INFO_LOG(SYSTEM, "Missed translation key in %s: %s", name_.c_str(), key);
 		return def ? def : key;
 	}
 }
@@ -99,16 +50,21 @@ void I18NCategory::SetMap(const std::map<std::string, std::string> &m) {
 		if (map_.find(iter->first) == map_.end()) {
 			std::string text = ReplaceAll(iter->second, "\\n", "\n");
 			map_[iter->first] = I18NEntry(text);
+//			INFO_LOG(SYSTEM, "Language entry: %s -> %s", iter->first.c_str(), text.c_str());
 		}
 	}
 }
 
-std::shared_ptr<I18NCategory> I18NRepo::GetCategory(I18NCat category) {
+std::shared_ptr<I18NCategory> I18NRepo::GetCategory(const char *category) {
 	std::lock_guard<std::mutex> guard(catsLock_);
-	if (category != I18NCat::NONE)
+	auto iter = cats_.find(category);
-		return cats_[(size_t)category];
+	if (iter != cats_.end()) {
-	else
+		return iter->second;
-		return nullptr;
+	} else {
+		I18NCategory *c = new I18NCategory(this, category);
+		cats_[category].reset(c);
+		return cats_[category];
+	}
 }
 
 Path I18NRepo::GetIniPath(const std::string &languageID) const {
@@ -117,7 +73,7 @@ Path I18NRepo::GetIniPath(const std::string &languageID) const {
 
 bool I18NRepo::IniExists(const std::string &languageID) const {
 	File::FileInfo info;
-	if (!g_VFS.GetFileInfo(GetIniPath(languageID).ToString().c_str(), &info))
+	if (!VFSGetFileInfo(GetIniPath(languageID).ToString().c_str(), &info))
 		return false;
 	if (!info.exists)
 		return false;
@@ -135,19 +91,17 @@ bool I18NRepo::LoadIni(const std::string &languageID, const Path &overridePath)
 		iniPath = GetIniPath(languageID);
 	}
 
-	if (!ini.LoadFromVFS(g_VFS, iniPath.ToString()))
+	if (!ini.LoadFromVFS(iniPath.ToString()))
 		return false;
 
 	Clear();
 
-	const std::vector<std::unique_ptr<Section>> &sections = ini.Sections();
+	const std::vector<Section> &sections = ini.Sections();
 
 	std::lock_guard<std::mutex> guard(catsLock_);
-	for (auto &section : sections) {
+	for (auto iter = sections.begin(); iter != sections.end(); ++iter) {
-		for (size_t i = 0; i < (size_t)I18NCat::CATEGORY_COUNT; i++) {
+		if (iter->name() != "") {
-			if (!strcmp(section->name().c_str(), g_categoryNames[i])) {
+			cats_[iter->name()].reset(LoadSection(&(*iter), iter->name().c_str()));
-				cats_[i].reset(new I18NCategory(*section.get()));
-			}
 		}
 	}
 
@@ -155,21 +109,53 @@ bool I18NRepo::LoadIni(const std::string &languageID, const Path &overridePath)
 	return true;
 }
 
-void I18NRepo::LogMissingKeys() const {
+std::map<std::string, std::vector<std::string>> I18NRepo::GetMissingKeys() const {
+	std::map<std::string, std::vector<std::string>> ret;
 	std::lock_guard<std::mutex> guard(catsLock_);
-	for (size_t i = 0; i < (size_t)I18NCat::CATEGORY_COUNT; i++) {
+	for (auto &cat : cats_) {
-		auto &cat = cats_[i];
+		for (auto &key : cat.second->Missed()) {
-		for (auto &key : cat->Missed()) {
+			ret[cat.first].push_back(key.first);
-			INFO_LOG(SYSTEM, "Missing translation [%s]: %s (%s)", g_categoryNames[i], key.first.c_str(), key.second.c_str());
 		}
 	}
+	return ret;
 }
 
-std::shared_ptr<I18NCategory> GetI18NCategory(I18NCat category) {
+I18NCategory *I18NRepo::LoadSection(const Section *section, const char *name) {
-	if (category == I18NCat::NONE) {
+	I18NCategory *cat = new I18NCategory(this, name);
-		return std::shared_ptr<I18NCategory>();
+	std::map<std::string, std::string> sectionMap = section->ToMap();
-	}
+	cat->SetMap(sectionMap);
-	std::shared_ptr<I18NCategory> cat = g_i18nrepo.GetCategory(category);
-	_dbg_assert_(cat);
 	return cat;
 }
+
+// This is a very light touched save variant - it won't overwrite 
+// anything, only create new entries.
+void I18NRepo::SaveIni(const std::string &languageID) {
+	IniFile ini;
+	ini.Load(GetIniPath(languageID));
+	std::lock_guard<std::mutex> guard(catsLock_);
+	for (auto iter = cats_.begin(); iter != cats_.end(); ++iter) {
+		std::string categoryName = iter->first;
+		Section *section = ini.GetOrCreateSection(categoryName.c_str());
+		SaveSection(ini, section, iter->second);
+	}
+	ini.Save(GetIniPath(languageID));
+}
+
+void I18NRepo::SaveSection(IniFile &ini, Section *section, std::shared_ptr<I18NCategory> cat) {
+	const std::map<std::string, std::string> &missed = cat->Missed();
+
+	for (auto iter = missed.begin(); iter != missed.end(); ++iter) {
+		if (!section->Exists(iter->first.c_str())) {
+			std::string text = ReplaceAll(iter->second, "\n", "\\n");
+			section->Set(iter->first, text);
+		}
+	}
+
+	const std::map<std::string, I18NEntry> &entries = cat->GetMap();
+	for (auto iter = entries.begin(); iter != entries.end(); ++iter) {
+		std::string text = ReplaceAll(iter->second.text, "\n", "\\n");
+		section->Set(iter->first, text);
+	}
+
+	cat->ClearMissed();
+}

Common/Data/Text/I18n.h

@@ -9,10 +9,10 @@
 // As usual, everything is UTF-8. Nothing else allowed.
 
 #include <map>
+#include <memory>
 #include <mutex>
 #include <string>
 #include <vector>
-#include <memory>
 
 #include "Common/Common.h"
 #include "Common/File/Path.h"
@@ -23,45 +23,6 @@ class I18NRepo;
 class IniFile;
 class Section;
 
-enum class I18NCat : uint8_t {
-	AUDIO = 0,
-	CONTROLS,
-	CWCHEATS,
-	DESKTOPUI,
-	DEVELOPER,
-	DIALOG,
-	ERRORS,  // Can't name it ERROR, clashes with many defines.
-	GAME,
-	GRAPHICS,
-	INSTALLZIP,
-	KEYMAPPING,
-	MAINMENU,
-	MAINSETTINGS,
-	MAPPABLECONTROLS,
-	NETWORKING,
-	PAUSE,
-	POSTSHADERS,
-	PSPCREDITS,
-	MEMSTICK,
-	REMOTEISO,
-	REPORTING,
-	SAVEDATA,
-	SCREEN,
-	SEARCH,
-	STORE,
-	SYSINFO,
-	SYSTEM,
-	TEXTURESHADERS,
-	THEMES,
-	UI_ELEMENTS,
-	UPGRADE,
-	VR,
-	ACHIEVEMENTS,
-	PSPSETTINGS,
-	CATEGORY_COUNT,
-	NONE = CATEGORY_COUNT,
-};
-
 struct I18NEntry {
 	I18NEntry(const std::string &t) : text(t), readFlag(false) {}
 	I18NEntry() : readFlag(false) {}
@@ -78,10 +39,9 @@ struct I18NCandidate {
 
 class I18NCategory {
 public:
-	I18NCategory() {}
+	// NOTE: Name must be a global constant string - it is not copied.
-	explicit I18NCategory(const Section &section);
+	I18NCategory(const char *name) : name_(name) {}
-
+	const char *T(const char *key, const char *def = 0);
-	const char *T(const char *key, const char *def = nullptr);
 	const char *T(const std::string &key) {
 		return T(key.c_str(), nullptr);
 	}
@@ -93,55 +53,76 @@ public:
 
 	const std::map<std::string, I18NEntry> &GetMap() { return map_; }
 	void ClearMissed() { missedKeyLog_.clear(); }
-	void Clear();
+	const char *GetName() const { return name_.c_str(); }
 
 private:
-	I18NCategory(I18NRepo *repo, const char *name) {}
+	I18NCategory(I18NRepo *repo, const char *name) : name_(name) {}
 	void SetMap(const std::map<std::string, std::string> &m);
 
+	std::string name_;
+
 	std::map<std::string, I18NEntry> map_;
 	mutable std::mutex missedKeyLock_;
 	std::map<std::string, std::string> missedKeyLog_;
 
 	// Noone else can create these.
 	friend class I18NRepo;
+
+	DISALLOW_COPY_AND_ASSIGN(I18NCategory);
 };
 
 class I18NRepo {
 public:
-	I18NRepo();
+	I18NRepo() {}
+	~I18NRepo();
+
 	bool IniExists(const std::string &languageID) const;
 	bool LoadIni(const std::string &languageID, const Path &overridePath = Path()); // NOT the filename!
+	void SaveIni(const std::string &languageID);
 
 	std::string LanguageID();
 
-	std::shared_ptr<I18NCategory> GetCategory(I18NCat category);
+	std::shared_ptr<I18NCategory> GetCategory(const char *categoryName);
-
+	bool HasCategory(const char *categoryName) const {
-	// Translate the string, by looking up "key" in the file, and falling back to either def or key, in that order, if the lookup fails.
+		std::lock_guard<std::mutex> guard(catsLock_);
-	// def can (and usually is) set to nullptr.
+		return cats_.find(categoryName) != cats_.end();
-	const char *T(I18NCat category, const char *key, const char *def = nullptr) {
-		if (category == I18NCat::NONE)
-			return def ? def : key;
-		return cats_[(size_t)category]->T(key, def);
 	}
+	const char *T(const char *category, const char *key, const char *def = 0);
 
-	void LogMissingKeys() const;
+	std::map<std::string, std::vector<std::string>> GetMissingKeys() const;
 
 private:
 	Path GetIniPath(const std::string &languageID) const;
 	void Clear();
+	I18NCategory *LoadSection(const Section *section, const char *name);
+	void SaveSection(IniFile &ini, Section *section, std::shared_ptr<I18NCategory> cat);
 
 	mutable std::mutex catsLock_;
-	std::shared_ptr<I18NCategory> cats_[(size_t)I18NCat::CATEGORY_COUNT];
+	std::map<std::string, std::shared_ptr<I18NCategory>> cats_;
 	std::string languageID_;
+
+	DISALLOW_COPY_AND_ASSIGN(I18NRepo);
 };
 
-extern I18NRepo g_i18nrepo;
+extern I18NRepo i18nrepo;
 
 // These are simply talking to the one global instance of I18NRepo.
 
-std::shared_ptr<I18NCategory> GetI18NCategory(I18NCat cat);
+inline std::shared_ptr<I18NCategory> GetI18NCategory(const char *categoryName) {
-
+	if (!categoryName)
-inline const char *T(I18NCat category, const char *key, const char *def = nullptr) {
+		return nullptr;
-	return g_i18nrepo.T(category, key, def);
+	return i18nrepo.GetCategory(categoryName);
 }
+
+inline bool I18NCategoryLoaded(const char *categoryName) {
+	return i18nrepo.HasCategory(categoryName);
+}
+
+inline const char *T(const char *category, const char *key, const char *def = 0) {
+	return i18nrepo.T(category, key, def);
+}
+
+inline std::map<std::string, std::vector<std::string>> GetI18NMissingKeys() {
+	return i18nrepo.GetMissingKeys();
+}
+

Common/Data/Text/Parsers.cpp

@@ -19,7 +19,7 @@ void NiceSizeFormat(uint64_t size, char *out, size_t bufSize) {
 	if (s == 0)
 		snprintf(out, bufSize, "%d B", (int)size);
 	else
-		snprintf(out, bufSize, "%3.2f %s", f, sizes[s]);
+		snprintf(out, bufSize, "%3.1f %s", f, sizes[s]);
 }
 
 std::string NiceSizeFormat(uint64_t size) {
@@ -43,7 +43,7 @@ bool Version::ParseVersionString(std::string str) {
 
 std::string Version::ToString() const {
 	char temp[128];
-	snprintf(temp, sizeof(temp), "%i.%i.%i", major, minor, sub);
+	sprintf(temp, "%i.%i.%i", major, minor, sub);
 	return std::string(temp);
 }
 

Common/Data/Text/WrapText.cpp

@@ -177,18 +177,20 @@ void WordWrapper::AppendWord(int endIndex, int lastChar, bool addNewline) {
 }
 
 void WordWrapper::Wrap() {
+	out_.clear();
+
 	// First, let's check if it fits as-is.
 	size_t len = strlen(str_);
+
+	// We know it'll be approximately this size. It's fine if the guess is a little off.
+	out_.reserve(len + len / 16);
+
 	if (MeasureWidth(str_, len) <= maxW_) {
 		// If it fits, we don't need to go through each character.
 		out_ = str_;
 		return;
 	}
 
-	out_.clear();
-	// We know it'll be approximately this size. It's fine if the guess is a little off.
-	out_.reserve(len + len / 16);
-
 	if (flags_ & FLAG_ELLIPSIZE_TEXT) {
 		ellipsisWidth_ = MeasureWidth("...", 3);
 	}

Common/FakeCPUDetect.cpp

@@ -18,14 +18,12 @@
 #include "ppsspp_config.h"
 #if PPSSPP_ARCH(ARM) || PPSSPP_ARCH(ARM64)
 #define REAL_CPUDETECT_AVAIL 1
-#elif (PPSSPP_ARCH(X86) || PPSSPP_ARCH(AMD64)) && !defined(__EMSCRIPTEN__)
+#elif PPSSPP_ARCH(X86) || PPSSPP_ARCH(AMD64)
 #define REAL_CPUDETECT_AVAIL 1
 #elif PPSSPP_ARCH(MIPS) || PPSSPP_ARCH(MIPS64)
 #define REAL_CPUDETECT_AVAIL 1
 #elif PPSSPP_ARCH(RISCV64)
 #define REAL_CPUDETECT_AVAIL 1
-#elif PPSSPP_ARCH(LOONGARCH64)
-#define REAL_CPUDETECT_AVAIL 1
 #endif
 
 #ifndef REAL_CPUDETECT_AVAIL

Common/File/AndroidContentURI.cpp

@@ -1,207 +0,0 @@
-#include "Common/File/AndroidContentURI.h"
-
-bool AndroidContentURI::Parse(std::string_view path) {
-	const char *prefix = "content://";
-	if (!startsWith(path, prefix)) {
-		return false;
-	}
-
-	std::string_view components = path.substr(strlen(prefix));
-
-	std::vector<std::string_view> parts;
-	SplitString(components, '/', parts);
-	if (parts.size() == 3) {
-		// Single file URI.
-		provider = parts[0];
-		if (parts[1] == "tree") {
-			// Single directory URI.
-			// Not sure when we encounter these?
-			// file empty signals this type.
-			root = UriDecode(parts[2]);
-			return true;
-		} else if (parts[1] == "document") {
-			// root empty signals this type.
-			file = UriDecode(parts[2]);
-			return true;
-		} else {
-			// What's this?
-			return false;
-		}
-	} else if (parts.size() == 5) {
-		// Tree URI
-		provider = parts[0];
-		if (parts[1] != "tree") {
-			return false;
-		}
-		root = UriDecode(parts[2]);
-		if (parts[3] != "document") {
-			return false;
-		}
-		file = UriDecode(parts[4]);
-		// Sanity check file path.
-		return startsWith(file, root);
-	} else {
-		// Invalid Content URI
-		return false;
-	}
-}
-
-AndroidContentURI AndroidContentURI::WithRootFilePath(const std::string &filePath) {
-	if (root.empty()) {
-		ERROR_LOG(SYSTEM, "WithRootFilePath cannot be used with single file URIs.");
-		return *this;
-	}
-
-	AndroidContentURI uri = *this;
-	uri.file = uri.root;
-	if (!filePath.empty()) {
-		uri.file += "/" + filePath;
-	}
-	return uri;
-}
-
-AndroidContentURI AndroidContentURI::WithComponent(std::string_view filePath) {
-	AndroidContentURI uri = *this;
-	if (uri.file.empty()) {
-		// Not sure what to do.
-		return uri;
-	}
-	if (uri.file.back() == ':') {
-		// Special case handling for Document URIs: Treat the ':' as a directory separator too (but preserved in the filename).
-		uri.file.append(filePath);
-	} else {
-		uri.file.push_back('/');
-		uri.file.append(filePath);
-	}
-	return uri;
-}
-
-AndroidContentURI AndroidContentURI::WithExtraExtension(std::string_view extension) {
-	AndroidContentURI uri = *this;
-	uri.file.append(extension);
-	return uri;
-}
-
-AndroidContentURI AndroidContentURI::WithReplacedExtension(const std::string &oldExtension, const std::string &newExtension) const {
-	_dbg_assert_(!oldExtension.empty() && oldExtension[0] == '.');
-	_dbg_assert_(!newExtension.empty() && newExtension[0] == '.');
-	AndroidContentURI uri = *this;
-	if (endsWithNoCase(file, oldExtension)) {
-		uri.file = file.substr(0, file.size() - oldExtension.size()) + newExtension;
-	}
-	return uri;
-}
-
-AndroidContentURI AndroidContentURI::WithReplacedExtension(const std::string &newExtension) const {
-	_dbg_assert_(!newExtension.empty() && newExtension[0] == '.');
-	AndroidContentURI uri = *this;
-	if (file.empty()) {
-		return uri;
-	}
-	std::string extension = GetFileExtension();
-	uri.file = file.substr(0, file.size() - extension.size()) + newExtension;
-	return uri;
-}
-
-bool AndroidContentURI::CanNavigateUp() const {
-	if (IsTreeURI()) {
-		return file.size() > root.size();
-	} else {
-		return file.find(':') != std::string::npos && file.back() != ':';
-	}
-}
-
-// Only goes downwards in hierarchies. No ".." will ever be generated.
-bool AndroidContentURI::ComputePathTo(const AndroidContentURI &other, std::string &path) const {
-	size_t offset = FilePath().size() + 1;
-	std::string otherFilePath = other.FilePath();
-	if (offset >= otherFilePath.size()) {
-		ERROR_LOG(SYSTEM, "Bad call to PathTo. '%s' -> '%s'", FilePath().c_str(), other.FilePath().c_str());
-		return false;
-	}
-
-	path = other.FilePath().substr(FilePath().size() + 1);
-	return true;
-}
-
-std::string AndroidContentURI::GetFileExtension() const {
-	size_t pos = file.rfind(".");
-	if (pos == std::string::npos) {
-		return "";
-	}
-	size_t slash_pos = file.rfind("/");
-	if (slash_pos != std::string::npos && slash_pos > pos) {
-		// Don't want to detect "df/file" from "/as.df/file"
-		return "";
-	}
-	std::string ext = file.substr(pos);
-	for (size_t i = 0; i < ext.size(); i++) {
-		ext[i] = tolower(ext[i]);
-	}
-	return ext;
-}
-
-std::string AndroidContentURI::GetLastPart() const {
-	if (file.empty()) {
-		// Can't do anything anyway.
-		return std::string();
-	}
-
-	if (!CanNavigateUp()) {
-		size_t colon = file.rfind(':');
-		if (colon == std::string::npos) {
-			return std::string();
-		}
-		if (file.back() == ':') {
-			return file;
-		}
-		return file.substr(colon + 1);
-	}
-
-	size_t slash = file.rfind('/');
-	if (slash == std::string::npos) {
-		// ok, look for the final colon. If it's the last char, we would have been caught above in !CanNavigateUp.
-		size_t colon = file.rfind(':');
-		if (colon == std::string::npos) {
-			return std::string();
-		}
-		return file.substr(colon + 1);
-	}
-
-	std::string part = file.substr(slash + 1);
-	return part;
-}
-
-bool AndroidContentURI::NavigateUp() {
-	if (!CanNavigateUp()) {
-		return false;
-	}
-
-	size_t slash = file.rfind('/');
-	if (slash == std::string::npos) {
-		// ok, look for the final colon.
-		size_t colon = file.rfind(':');
-		if (colon == std::string::npos) {
-			return false;
-		}
-		file = file.substr(0, colon + 1);  // Note: we include the colon in these paths.
-		return true;
-	}
-
-	file = file.substr(0, slash);
-	return true;
-}
-
-
-std::string AndroidContentURI::ToString() const {
-	if (file.empty()) {
-		// Tree URI
-		return StringFromFormat("content://%s/tree/%s", provider.c_str(), UriEncode(root).c_str());
-	} else if (root.empty()) {
-		// Single file URI
-		return StringFromFormat("content://%s/document/%s", provider.c_str(), UriEncode(file).c_str());
-	} else {
-		// File URI from Tree
-		return StringFromFormat("content://%s/tree/%s/document/%s", provider.c_str(), UriEncode(root).c_str(), UriEncode(file).c_str());
-	}
-}

Common/File/AndroidContentURI.h

@@ -1,73 +0,0 @@
-#pragma once
-
-#include <string>
-
-#include "Common/StringUtils.h"
-#include "Common/Net/URL.h"
-#include "Common/Log.h"
-
-// Utility to deal with Android storage URIs of the forms:
-// content://com.android.externalstorage.documents/tree/primary%3APSP%20ISO
-// content://com.android.externalstorage.documents/tree/primary%3APSP%20ISO/document/primary%3APSP%20ISO
-
-// This file compiles on all platforms, to reduce the need for ifdefs.
-
-// I am not 100% sure it's OK to rely on the internal format of file content URIs.
-// On the other hand, I'm sure tons of apps would break if these changed, so I think we can
-// consider them pretty stable. Additionally, the official Document library just manipulates the URIs
-// in similar ways...
-class AndroidContentURI {
-private:
-	std::string provider;
-	std::string root;
-	std::string file;
-public:
-	AndroidContentURI() {}
-	explicit AndroidContentURI(std::string_view path) {
-		Parse(path);
-	}
-
-	bool Parse(std::string_view path);
-
-	AndroidContentURI WithRootFilePath(const std::string &filePath);
-	AndroidContentURI WithComponent(std::string_view filePath);
-	AndroidContentURI WithExtraExtension(std::string_view extension);  // The ext string contains the dot.
-	AndroidContentURI WithReplacedExtension(const std::string &oldExtension, const std::string &newExtension) const;
-	AndroidContentURI WithReplacedExtension(const std::string &newExtension) const;
-
-	bool CanNavigateUp() const;
-
-	// Only goes downwards in hierarchies. No ".." will ever be generated.
-	bool ComputePathTo(const AndroidContentURI &other, std::string &path) const;
-
-	std::string GetFileExtension() const;
-	std::string GetLastPart() const;
-
-	bool NavigateUp();
-
-	bool TreeContains(const AndroidContentURI &fileURI) {
-		if (root.empty()) {
-			return false;
-		}
-		return startsWith(fileURI.file, root);
-	}
-
-	std::string ToString() const;
-
-	// Never store the output of this, only show it to the user.
-	std::string ToVisualString() const {
-		return file;
-	}
-
-	const std::string &FilePath() const {
-		return file;
-	}
-
-	const std::string &RootPath() const {
-		return root.empty() ? file : root;
-	}
-
-	bool IsTreeURI() const {
-		return !root.empty();
-	}
-};

Common/File/AndroidStorage.cpp

@@ -1,5 +1,3 @@
-#include <inttypes.h>
-
 #include "Common/File/AndroidStorage.h"
 #include "Common/StringUtils.h"
 #include "Common/Log.h"
@@ -61,16 +59,16 @@ void Android_RegisterStorageCallbacks(JNIEnv * env, jobject obj) {
 	_dbg_assert_(computeRecursiveDirectorySize);
 }
 
-bool Android_IsContentUri(std::string_view filename) {
+bool Android_IsContentUri(const std::string &filename) {
 	return startsWith(filename, "content://");
 }
 
-int Android_OpenContentUriFd(std::string_view filename, Android_OpenContentUriMode mode) {
+int Android_OpenContentUriFd(const std::string &filename, Android_OpenContentUriMode mode) {
 	if (!g_nativeActivity) {
 		return -1;
 	}
 
-	std::string fname(filename);
+	std::string fname = filename;
 	// PPSSPP adds an ending slash to directories before looking them up.
 	// TODO: Fix that in the caller (or don't call this for directories).
 	if (fname.back() == '/')

Common/File/AndroidStorage.h

@@ -2,7 +2,6 @@
 
 #include <vector>
 #include <string>
-#include <string_view>
 
 #include "Common/File/DirListing.h"
 
@@ -40,8 +39,8 @@ extern std::string g_externalDir;
 
 void Android_StorageSetNativeActivity(jobject nativeActivity);
 
-bool Android_IsContentUri(std::string_view uri);
+bool Android_IsContentUri(const std::string &uri);
-int Android_OpenContentUriFd(std::string_view uri, const Android_OpenContentUriMode mode);
+int Android_OpenContentUriFd(const std::string &uri, const Android_OpenContentUriMode mode);
 StorageError Android_CreateDirectory(const std::string &parentTreeUri, const std::string &dirName);
 StorageError Android_CreateFile(const std::string &parentTreeUri, const std::string &fileName);
 StorageError Android_MoveFile(const std::string &fileUri, const std::string &srcParentUri, const std::string &destParentUri);
@@ -64,8 +63,8 @@ void Android_RegisterStorageCallbacks(JNIEnv * env, jobject obj);
 
 // Stub out the Android Storage wrappers, so that we can avoid ifdefs everywhere.
 
-inline bool Android_IsContentUri(std::string_view uri) { return false; }
+inline bool Android_IsContentUri(const std::string &uri) { return false; }
-inline int Android_OpenContentUriFd(std::string_view uri, const Android_OpenContentUriMode mode) { return -1; }
+inline int Android_OpenContentUriFd(const std::string &uri, const Android_OpenContentUriMode mode) { return -1; }
 inline StorageError Android_CreateDirectory(const std::string &parentTreeUri, const std::string &dirName) { return StorageError::UNKNOWN; }
 inline StorageError Android_CreateFile(const std::string &parentTreeUri, const std::string &fileName) { return StorageError::UNKNOWN; }
 inline StorageError Android_MoveFile(const std::string &fileUri, const std::string &srcParentUri, const std::string &destParentUri) { return StorageError::UNKNOWN; }

Common/File/DirListing.cpp

@@ -6,7 +6,6 @@
 #include <direct.h>
 #if PPSSPP_PLATFORM(UWP)
 #include <fileapifromapp.h>
-#include <UWP/UWPHelpers/StorageManager.h>
 #endif
 #else
 #include <strings.h>
@@ -184,7 +183,7 @@ bool GetFilesInDir(const Path &directory, std::vector<FileInfo> *files, const ch
 		std::string tmp;
 		while (*filter) {
 			if (*filter == ':') {
-				filters.insert(tmp);
+				filters.insert(std::move(tmp));
 				tmp.clear();
 			} else {
 				tmp.push_back(*filter);
@@ -192,7 +191,7 @@ bool GetFilesInDir(const Path &directory, std::vector<FileInfo> *files, const ch
 			filter++;
 		}
 		if (!tmp.empty())
-			filters.insert(tmp);
+			filters.insert(std::move(tmp));
 	}
 
 #if PPSSPP_PLATFORM(WINDOWS)
@@ -221,13 +220,6 @@ bool GetFilesInDir(const Path &directory, std::vector<FileInfo> *files, const ch
 	HANDLE hFind = FindFirstFileEx((directory.ToWString() + L"\\*").c_str(), FindExInfoStandard, &ffd, FindExSearchNameMatch, NULL, 0);
 #endif
 	if (hFind == INVALID_HANDLE_VALUE) {
-#if PPSSPP_PLATFORM(UWP)
-		// This step just to avoid empty results by adding fake folders
-		// it will help also to navigate back between selected folder
-		// we must ignore this function for any request other than UI navigation
-		if (GetFakeFolders(directory, files, filter, filters))
-			return true;
-#endif
 		return false;
 	}
 	do {

Common/File/DirListing.h

@@ -2,8 +2,10 @@
 
 #include <string>
 #include <vector>
+
 #include <cstdio>
-#include <cstdint>
+
+#include <inttypes.h>
 
 #include "Common/File/Path.h"
 

Common/File/DiskFree.cpp

@@ -23,19 +23,9 @@
 #include "Common/File/AndroidStorage.h"
 #include "Common/Data/Encoding/Utf8.h"
 
-#if PPSSPP_PLATFORM(UWP)
-#include "UWP/UWPHelpers/StorageManager.h"
-#endif
-
 bool free_disk_space(const Path &path, int64_t &space) {
 #ifdef _WIN32
 	ULARGE_INTEGER free;
-#if PPSSPP_PLATFORM(UWP)
-	if (GetDriveFreeSpace(path, space)) {
-		return true;
-	}
-	else
-#endif
 	if (GetDiskFreeSpaceExW(path.ToWString().c_str(), &free, nullptr, nullptr)) {
 		space = free.QuadPart;
 		return true;

Common/File/FileUtil.cpp

@@ -25,6 +25,7 @@
 #include "ppsspp_config.h"
 
 #include "android/jni/app-android.h"
+#include "android/jni/AndroidContentURI.h"
 
 #ifdef __MINGW32__
 #include <unistd.h>
@@ -32,14 +33,12 @@
 #define _POSIX_THREAD_SAFE_FUNCTIONS 200112L
 #endif
 #endif
-
 #include <cstring>
 #include <ctime>
 #include <memory>
 
 #include "Common/Log.h"
 #include "Common/LogReporting.h"
-#include "Common/File/AndroidContentURI.h"
 #include "Common/File/FileUtil.h"
 #include "Common/StringUtils.h"
 #include "Common/SysError.h"
@@ -54,7 +53,6 @@
 #include <direct.h>		// getcwd
 #if PPSSPP_PLATFORM(UWP)
 #include <fileapifromapp.h>
-#include "UWP/UWPHelpers/StorageManager.h"
 #endif
 #else
 #include <sys/param.h>
@@ -137,20 +135,14 @@ FILE *OpenCFile(const Path &path, const char *mode) {
 			}
 
 			// TODO: Support append modes and stuff... For now let's go with the most common one.
-			Android_OpenContentUriMode openMode = Android_OpenContentUriMode::READ_WRITE_TRUNCATE;
+			int descriptor = Android_OpenContentUriFd(path.ToString(), Android_OpenContentUriMode::READ_WRITE_TRUNCATE);
-			const char *fmode = "wb";
-			if (!strcmp(mode, "at") || !strcmp(mode, "a")) {
-				openMode = Android_OpenContentUriMode::READ_WRITE;
-				fmode = "ab";
-			}
-			int descriptor = Android_OpenContentUriFd(path.ToString(), openMode);
 			if (descriptor < 0) {
 				INFO_LOG(COMMON, "Opening '%s' for write failed", path.ToString().c_str());
 				return nullptr;
 			}
-			FILE *f = fdopen(descriptor, fmode);
+			FILE *f = fdopen(descriptor, "wb");
 			if (f && (!strcmp(mode, "at") || !strcmp(mode, "a"))) {
-				// Append mode - not sure we got a "true" append mode, so seek to the end.
+				// Append mode.
 				fseek(f, 0, SEEK_END);
 			}
 			return f;
@@ -165,18 +157,7 @@ FILE *OpenCFile(const Path &path, const char *mode) {
 	}
 
 #if defined(_WIN32) && defined(UNICODE)
-#if PPSSPP_PLATFORM(UWP) && !defined(__LIBRETRO__)
-	// We shouldn't use _wfopen here, 
-	// this function is not allowed to read outside Local and Installation folders
-	// FileSystem (broadFileSystemAccess) doesn't apply on _wfopen
-	// if we have custom memory stick location _wfopen will return null
-	// 'GetFileStreamFromApp' will convert 'mode' to [access, share, creationDisposition]
-	// then it will call 'CreateFile2FromAppW' -> convert HANDLE to FILE*
-	FILE* file = GetFileStreamFromApp(path.ToString(), mode);
-	return file;
-#else
 	return _wfopen(path.ToWString().c_str(), ConvertUTF8ToWString(mode).c_str());
-#endif
 #else
 	return fopen(path.c_str(), mode);
 #endif
@@ -592,7 +573,7 @@ bool CreateFullPath(const Path &path) {
 		return false;
 	}
 
-	std::vector<std::string_view> parts;
+	std::vector<std::string> parts;
 	SplitString(diff, '/', parts);
 
 	// Probably not necessary sanity check, ported from the old code.
@@ -602,7 +583,7 @@ bool CreateFullPath(const Path &path) {
 	}
 
 	Path curPath = root;
-	for (auto part : parts) {
+	for (auto &part : parts) {
 		curPath /= part;
 		if (!File::Exists(curPath)) {
 			File::CreateDir(curPath);
@@ -677,15 +658,10 @@ bool Rename(const Path &srcFilename, const Path &destFilename) {
 	INFO_LOG(COMMON, "Rename: %s --> %s", srcFilename.c_str(), destFilename.c_str());
 
 #if defined(_WIN32) && defined(UNICODE)
-#if PPSSPP_PLATFORM(UWP)
-	if (MoveFileFromAppW(srcFilename.ToWString().c_str(), destFilename.ToWString().c_str()))
-		return true;
-#else
 	std::wstring srcw = srcFilename.ToWString();
 	std::wstring destw = destFilename.ToWString();
 	if (_wrename(srcw.c_str(), destw.c_str()) == 0)
 		return true;
-#endif
 #else
 	if (rename(srcFilename.c_str(), destFilename.c_str()) == 0)
 		return true;
@@ -969,7 +945,7 @@ bool OpenFileInEditor(const Path &fileName) {
 
 #if PPSSPP_PLATFORM(WINDOWS)
 #if PPSSPP_PLATFORM(UWP)
-	OpenFile(fileName.ToString());
+	// Do nothing.
 #else
 	ShellExecuteW(nullptr, L"open", fileName.ToWString().c_str(), nullptr, nullptr, SW_SHOW);
 #endif
@@ -1181,7 +1157,6 @@ uint8_t *ReadLocalFile(const Path &filename, size_t *size) {
 		return nullptr;
 	}
 	fseek(file, 0, SEEK_SET);
-	// NOTE: If you find ~10 memory leaks from here, with very varying sizes, it might be the VFPU LUTs.
 	uint8_t *contents = new uint8_t[f_size + 1];
 	if (fread(contents, 1, f_size, file) != f_size) {
 		delete[] contents;

Common/File/Path.cpp

@@ -4,17 +4,13 @@
 #include <cstring>
 
 #include "Common/File/Path.h"
-#include "Common/File/AndroidContentURI.h"
 #include "Common/File/FileUtil.h"
 #include "Common/StringUtils.h"
 #include "Common/Log.h"
 #include "Common/Data/Encoding/Utf8.h"
 
 #include "android/jni/app-android.h"
-
+#include "android/jni/AndroidContentURI.h"
-#if PPSSPP_PLATFORM(UWP) && !defined(__LIBRETRO__)
-#include "UWP/UWPHelpers/StorageManager.h"
-#endif
 
 #if HOST_IS_CASE_SENSITIVE
 #include <dirent.h>
@@ -22,7 +18,7 @@
 #include <sys/stat.h>
 #endif
 
-Path::Path(std::string_view str) {
+Path::Path(const std::string &str) {
 	Init(str);
 }
 
@@ -33,7 +29,7 @@ Path::Path(const std::wstring &str) {
 }
 #endif
 
-void Path::Init(std::string_view str) {
+void Path::Init(const std::string &str) {
 	if (str.empty()) {
 		type_ = PathType::UNDEFINED;
 		path_.clear();
@@ -81,7 +77,7 @@ void Path::Init(std::string_view str) {
 
 // We always use forward slashes internally, we convert to backslash only when
 // converted to a wstring.
-Path Path::operator /(std::string_view subdir) const {
+Path Path::operator /(const std::string &subdir) const {
 	if (type_ == PathType::CONTENT_URI) {
 		AndroidContentURI uri(path_);
 		return Path(uri.WithComponent(subdir).ToString());
@@ -104,18 +100,18 @@ Path Path::operator /(std::string_view subdir) const {
 	return Path(fullPath);
 }
 
-void Path::operator /=(std::string_view subdir) {
+void Path::operator /=(const std::string &subdir) {
 	*this = *this / subdir;
 }
 
-Path Path::WithExtraExtension(std::string_view ext) const {
+Path Path::WithExtraExtension(const std::string &ext) const {
 	if (type_ == PathType::CONTENT_URI) {
 		AndroidContentURI uri(path_);
 		return Path(uri.WithExtraExtension(ext).ToString());
 	}
 
 	_dbg_assert_(!ext.empty() && ext[0] == '.');
-	return Path(path_ + std::string(ext));
+	return Path(path_ + ext);
 }
 
 Path Path::WithReplacedExtension(const std::string &oldExtension, const std::string &newExtension) const {
@@ -135,11 +131,6 @@ Path Path::WithReplacedExtension(const std::string &oldExtension, const std::str
 }
 
 Path Path::WithReplacedExtension(const std::string &newExtension) const {
-	if (type_ == PathType::CONTENT_URI) {
-		AndroidContentURI uri(path_);
-		return Path(uri.WithReplacedExtension(newExtension).ToString());
-	}
-
 	_dbg_assert_(!newExtension.empty() && newExtension[0] == '.');
 	if (path_.empty()) {
 		return Path(*this);
@@ -161,7 +152,7 @@ std::string Path::GetFilename() const {
 	return path_;
 }
 
-std::string GetExtFromString(std::string_view str) {
+static std::string GetExtFromString(const std::string &str) {
 	size_t pos = str.rfind(".");
 	if (pos == std::string::npos) {
 		return "";
@@ -171,7 +162,7 @@ std::string GetExtFromString(std::string_view str) {
 		// Don't want to detect "df/file" from "/as.df/file"
 		return "";
 	}
-	std::string ext(str.substr(pos));
+	std::string ext = str.substr(pos);
 	for (size_t i = 0; i < ext.size(); i++) {
 		ext[i] = tolower(ext[i]);
 	}
@@ -181,7 +172,7 @@ std::string GetExtFromString(std::string_view str) {
 std::string Path::GetFileExtension() const {
 	if (type_ == PathType::CONTENT_URI) {
 		AndroidContentURI uri(path_);
-		return uri.GetFileExtension();
+		return GetExtFromString(uri.FilePath());
 	}
 	return GetExtFromString(path_);
 }
@@ -262,49 +253,14 @@ std::wstring Path::ToWString() const {
 	}
 	return w;
 }
-std::string Path::ToCString() const {
-	std::string w = path_;
-	for (size_t i = 0; i < w.size(); i++) {
-		if (w[i] == '/') {
-			w[i] = '\\';
-		}
-	}
-	return w;
-}
 #endif
 
-std::string Path::ToVisualString(const char *relativeRoot) const {
+std::string Path::ToVisualString() const {
 	if (type_ == PathType::CONTENT_URI) {
 		return AndroidContentURI(path_).ToVisualString();
 #if PPSSPP_PLATFORM(WINDOWS)
 	} else if (type_ == PathType::NATIVE) {
-#if PPSSPP_PLATFORM(UWP) && !defined(__LIBRETRO__)
+		return ReplaceAll(path_, "/", "\\");
-		return GetPreviewPath(path_);
-#else
-		// It can be useful to show the path as relative to the memstick
-		if (relativeRoot) {
-			std::string root = ReplaceAll(relativeRoot, "/", "\\");
-			std::string path = ReplaceAll(path_, "/", "\\");
-			if (startsWithNoCase(path, root)) {
-				return path.substr(root.size());
-			} else {
-				return path;
-			}
-		} else {
-			return ReplaceAll(path_, "/", "\\");
-		}
-#endif
-#else
-		if (relativeRoot) {
-			std::string root = relativeRoot;
-			if (startsWithNoCase(path_, root)) {
-				return path_.substr(root.size());
-			} else {
-				return path_;
-			}
-		} else {
-			return path_;
-		}
 #endif
 	} else {
 		return path_;

Common/File/Path.h

@@ -3,7 +3,6 @@
 #include "ppsspp_config.h"
 
 #include <string>
-#include <string_view>
 
 #if defined(__APPLE__)
 
@@ -37,11 +36,11 @@ enum class PathType {
 
 class Path {
 private:
-	void Init(std::string_view str);
+	void Init(const std::string &str);
 
 public:
 	Path() : type_(PathType::UNDEFINED) {}
-	explicit Path(std::string_view str);
+	explicit Path(const std::string &str);
 
 #if PPSSPP_PLATFORM(WINDOWS)
 	explicit Path(const std::wstring &str);
@@ -72,34 +71,28 @@ public:
 	bool IsAbsolute() const;
 
 	// Returns a path extended with a subdirectory.
-	Path operator /(std::string_view subdir) const;
+	Path operator /(const std::string &subdir) const;
 
 	// Navigates down into a subdir.
-	void operator /=(std::string_view subdir);
+	void operator /=(const std::string &subdir);
 
 	// File extension manipulation.
-	Path WithExtraExtension(std::string_view ext) const;
+	Path WithExtraExtension(const std::string &ext) const;
 	Path WithReplacedExtension(const std::string &oldExtension, const std::string &newExtension) const;
 	Path WithReplacedExtension(const std::string &newExtension) const;
 
+	// Removes the last component.
 	std::string GetFilename() const;  // Really, GetLastComponent. Could be a file or directory. Includes the extension.
 	std::string GetFileExtension() const;  // Always lowercase return. Includes the dot.
-	// Removes the last component.
 	std::string GetDirectory() const;
 
 	const std::string &ToString() const;
 
 #if PPSSPP_PLATFORM(WINDOWS)
 	std::wstring ToWString() const;
-	std::string ToCString() const;  // Flips the slashes back to Windows standard, but string still UTF-8.
-#else
-	std::string ToCString() const {
-		return ToString();
-	}
 #endif
 
-	// Pass in a relative root to turn the path into a relative path - if it is one!
+	std::string ToVisualString() const;
-	std::string ToVisualString(const char *relativeRoot = nullptr) const;
 
 	bool CanNavigateUp() const;
 	Path NavigateUp() const;
@@ -139,8 +132,6 @@ private:
 	PathType type_;
 };
 
-// Utility function for parsing out file extensions.
-std::string GetExtFromString(std::string_view str);
 
 // Utility function for fixing the case of paths. Only present on Unix-like systems.
 

Common/File/PathBrowser.cpp

@@ -17,6 +17,7 @@
 
 #if PPSSPP_PLATFORM(ANDROID)
 #include "android/jni/app-android.h"
+#include "android/jni/AndroidContentURI.h"
 #endif
 
 bool LoadRemoteFileList(const Path &url, const std::string &userAgent, bool *cancel, std::vector<File::FileInfo> &files) {
@@ -38,7 +39,7 @@ bool LoadRemoteFileList(const Path &url, const std::string &userAgent, bool *can
 	http::RequestParams req(baseURL.Resource(), "text/plain, text/html; q=0.9, */*; q=0.8");
 	if (http.Resolve(baseURL.Host().c_str(), baseURL.Port())) {
 		if (http.Connect(2, 20.0, cancel)) {
-			net::RequestProgress progress(cancel);
+			http::RequestProgress progress(cancel);
 			code = http.GET(req, &result, responseHeaders, &progress);
 			http.Disconnect();
 		}
@@ -78,7 +79,7 @@ bool LoadRemoteFileList(const Path &url, const std::string &userAgent, bool *can
 		return false;
 	}
 
-	for (auto &item : items) {
+	for (std::string item : items) {
 		// Apply some workarounds.
 		if (item.empty())
 			continue;
@@ -210,7 +211,7 @@ std::string PathBrowser::GetFriendlyPath() const {
 bool PathBrowser::GetListing(std::vector<File::FileInfo> &fileInfo, const char *filter, bool *cancel) {
 	std::unique_lock<std::mutex> guard(pendingLock_);
 	while (!IsListingReady() && (!cancel || !*cancel)) {
-		// In case cancel changes, just sleep. TODO: Replace with condition variable.
+		// In case cancel changes, just sleep.
 		guard.unlock();
 		sleep_ms(50);
 		guard.lock();
@@ -221,6 +222,14 @@ bool PathBrowser::GetListing(std::vector<File::FileInfo> &fileInfo, const char *
 }
 
 bool PathBrowser::CanNavigateUp() {
+/* Leaving this commented out, not sure if there's a use in UWP for navigating up from the user data folder.
+#if PPSSPP_PLATFORM(UWP)
+	// Can't navigate up from memstick folder :(
+	if (path_ == GetSysDirectory(DIRECTORY_MEMSTICK_ROOT)) {
+		return false;
+	}
+#endif
+*/
 	return path_.CanNavigateUp();
 }
 

Common/File/VFS/AssetReader.cpp

@@ -0,0 +1,204 @@
+#include <algorithm>
+#include <ctype.h>
+#include <set>
+#include <stdio.h>
+
+#ifdef __ANDROID__
+#include <zip.h>
+#endif
+
+#include "Common/Common.h"
+#include "Common/Log.h"
+#include "Common/File/VFS/AssetReader.h"
+
+#ifdef __ANDROID__
+uint8_t *ReadFromZip(zip *archive, const char* filename, size_t *size) {
+	// Figure out the file size first.
+	struct zip_stat zstat;
+	zip_file *file = zip_fopen(archive, filename, ZIP_FL_NOCASE|ZIP_FL_UNCHANGED);
+	if (!file) {
+		ERROR_LOG(IO, "Error opening %s from ZIP", filename);
+		return 0;
+	}
+	zip_stat(archive, filename, ZIP_FL_NOCASE|ZIP_FL_UNCHANGED, &zstat);
+
+	uint8_t *contents = new uint8_t[zstat.size + 1];
+	zip_fread(file, contents, zstat.size);
+	zip_fclose(file);
+	contents[zstat.size] = 0;
+
+	*size = zstat.size;
+	return contents;
+}
+
+#endif
+
+#ifdef __ANDROID__
+
+ZipAssetReader::ZipAssetReader(const char *zip_file, const char *in_zip_path) {
+	zip_file_ = zip_open(zip_file, 0, NULL);
+	strcpy(in_zip_path_, in_zip_path);
+	if (!zip_file_) {
+		ERROR_LOG(IO, "Failed to open %s as a zip file", zip_file);
+	}
+
+	std::vector<File::FileInfo> info;
+	GetFileListing("assets", &info, 0);
+	for (size_t i = 0; i < info.size(); i++) {
+		if (info[i].isDirectory) {
+			DEBUG_LOG(IO, "Directory: %s", info[i].name.c_str());
+		} else {
+			DEBUG_LOG(IO, "File: %s", info[i].name.c_str());
+		}
+	}
+}
+
+ZipAssetReader::~ZipAssetReader() {
+	std::lock_guard<std::mutex> guard(lock_);
+	zip_close(zip_file_);
+}
+
+uint8_t *ZipAssetReader::ReadAsset(const char *path, size_t *size) {
+	char temp_path[1024];
+	strcpy(temp_path, in_zip_path_);
+	strcat(temp_path, path);
+
+	std::lock_guard<std::mutex> guard(lock_);
+	return ReadFromZip(zip_file_, temp_path, size);
+}
+
+bool ZipAssetReader::GetFileListing(const char *orig_path, std::vector<File::FileInfo> *listing, const char *filter = 0) {
+	char path[1024];
+	strcpy(path, in_zip_path_);
+	strcat(path, orig_path);
+
+	std::set<std::string> filters;
+	std::string tmp;
+	if (filter) {
+		while (*filter) {
+			if (*filter == ':') {
+				filters.insert("." + tmp);
+				tmp.clear();
+			} else {
+				tmp.push_back(*filter);
+			}
+			filter++;
+		}
+	}
+	if (tmp.size())
+		filters.insert("." + tmp);
+
+	// We just loop through the whole ZIP file and deduce what files are in this directory, and what subdirectories there are.
+	std::set<std::string> files;
+	std::set<std::string> directories;
+	GetZipListings(path, files, directories);
+
+	for (auto diter = directories.begin(); diter != directories.end(); ++diter) {
+		File::FileInfo info;
+		info.name = *diter;
+
+		// Remove the "inzip" part of the fullname.
+		info.fullName = Path(std::string(path).substr(strlen(in_zip_path_))) / *diter;
+		info.exists = true;
+		info.isWritable = false;
+		info.isDirectory = true;
+		listing->push_back(info);
+	}
+
+	for (auto fiter = files.begin(); fiter != files.end(); ++fiter) {
+		std::string fpath = path;
+		File::FileInfo info;
+		info.name = *fiter;
+		info.fullName = Path(std::string(path).substr(strlen(in_zip_path_))) / *fiter;
+		info.exists = true;
+		info.isWritable = false;
+		info.isDirectory = false;
+		std::string ext = info.fullName.GetFileExtension();
+		if (filter) {
+			if (filters.find(ext) == filters.end()) {
+				continue;
+			}
+		}
+		listing->push_back(info);
+	}
+
+	std::sort(listing->begin(), listing->end());
+	return true;
+}
+
+void ZipAssetReader::GetZipListings(const char *path, std::set<std::string> &files, std::set<std::string> &directories) {
+	size_t pathlen = strlen(path);
+	if (path[pathlen - 1] == '/')
+		pathlen--;
+
+	std::lock_guard<std::mutex> guard(lock_);
+	int numFiles = zip_get_num_files(zip_file_);
+	for (int i = 0; i < numFiles; i++) {
+		const char* name = zip_get_name(zip_file_, i, 0);
+		if (!name)
+			continue;
+		if (!memcmp(name, path, pathlen)) {
+			// The prefix is right. Let's see if this is a file or path.
+			const char *slashPos = strchr(name + pathlen + 1, '/');
+			if (slashPos != 0) {
+				// A directory.
+				std::string dirName = std::string(name + pathlen + 1, slashPos - (name + pathlen + 1));
+				directories.insert(dirName);
+			} else if (name[pathlen] == '/') {
+				const char *fn = name + pathlen + 1;
+				files.insert(std::string(fn));
+			}  // else, it was a file with the same prefix as the path. like langregion.ini next to lang/.
+		}
+	}
+}
+
+bool ZipAssetReader::GetFileInfo(const char *path, File::FileInfo *info) {
+	struct zip_stat zstat;
+	char temp_path[1024];
+	strcpy(temp_path, in_zip_path_);
+	strcat(temp_path, path);
+	if (0 != zip_stat(zip_file_, temp_path, ZIP_FL_NOCASE|ZIP_FL_UNCHANGED, &zstat)) {
+		// ZIP files do not have real directories, so we'll end up here if we
+		// try to stat one. For now that's fine.
+		info->exists = false;
+		info->size = 0;
+		return false;
+	}
+
+	info->fullName = Path(path);
+	info->exists = true; // TODO
+	info->isWritable = false;
+	info->isDirectory = false;    // TODO
+	info->size = zstat.size;
+	return true;
+}
+
+#endif
+
+DirectoryAssetReader::DirectoryAssetReader(const Path &path) {
+	path_ = path;
+}
+
+uint8_t *DirectoryAssetReader::ReadAsset(const char *path, size_t *size) {
+	Path new_path = Path(path).StartsWith(path_) ? Path(path) : path_ / path;
+	return File::ReadLocalFile(new_path, size);
+}
+
+bool DirectoryAssetReader::GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter = nullptr) {
+	Path new_path = Path(path).StartsWith(path_) ? Path(path) : path_ / path;
+
+	File::FileInfo info;
+	if (!File::GetFileInfo(new_path, &info))
+		return false;
+
+	if (info.isDirectory) {
+		File::GetFilesInDir(new_path, listing, filter);
+		return true;
+	}
+	return false;
+}
+
+bool DirectoryAssetReader::GetFileInfo(const char *path, File::FileInfo *info) {
+	Path new_path = Path(path).StartsWith(path_) ? Path(path) : path_ / path;
+	return File::GetFileInfo(new_path, info);
+}

Common/File/VFS/AssetReader.h

@@ -0,0 +1,65 @@
+// TODO: Move much of this code to vfs.cpp
+#pragma once
+
+#ifdef __ANDROID__
+#include <zip.h>
+#endif
+
+#include <mutex>
+#include <set>
+#include <string.h>
+#include <string>
+
+#include "Common/File/VFS/VFS.h"
+#include "Common/File/FileUtil.h"
+#include "Common/File/Path.h"
+
+class AssetReader {
+public:
+	virtual ~AssetReader() {}
+	// use delete[]
+	virtual uint8_t *ReadAsset(const char *path, size_t *size) = 0;
+	// Filter support is optional but nice to have
+	virtual bool GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter = 0) = 0;
+	virtual bool GetFileInfo(const char *path, File::FileInfo *info) = 0;
+	virtual std::string toString() const = 0;
+};
+
+#ifdef __ANDROID__
+uint8_t *ReadFromZip(zip *archive, const char* filename, size_t *size);
+class ZipAssetReader : public AssetReader {
+public:
+	ZipAssetReader(const char *zip_file, const char *in_zip_path);
+	~ZipAssetReader();
+	// use delete[]
+	uint8_t *ReadAsset(const char *path, size_t *size) override;
+	bool GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter) override;
+	bool GetFileInfo(const char *path, File::FileInfo *info) override;
+	std::string toString() const override {
+		return in_zip_path_;
+	}
+
+private:
+	void GetZipListings(const char *path, std::set<std::string> &files, std::set<std::string> &directories);
+
+	zip *zip_file_;
+	std::mutex lock_;
+	char in_zip_path_[256];
+};
+#endif
+
+class DirectoryAssetReader : public AssetReader {
+public:
+	explicit DirectoryAssetReader(const Path &path);
+	// use delete[]
+	uint8_t *ReadAsset(const char *path, size_t *size) override;
+	bool GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter) override;
+	bool GetFileInfo(const char *path, File::FileInfo *info) override;
+	std::string toString() const override {
+		return path_.ToString();
+	}
+
+private:
+	Path path_;
+};
+

Common/File/VFS/DirectoryReader.cpp

@@ -1,99 +0,0 @@
-#include <cstdio>
-
-#include "Common/Common.h"
-#include "Common/Log.h"
-#include "Common/File/VFS/DirectoryReader.h"
-
-DirectoryReader::DirectoryReader(const Path &path) {
-	path_ = path;
-}
-
-uint8_t *DirectoryReader::ReadFile(const char *path, size_t *size) {
-	Path new_path = Path(path).StartsWith(path_) ? Path(path) : path_ / path;
-	return File::ReadLocalFile(new_path, size);
-}
-
-bool DirectoryReader::GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter = nullptr) {
-	Path new_path = Path(path).StartsWith(path_) ? Path(path) : path_ / path;
-
-	File::FileInfo info;
-	if (!File::GetFileInfo(new_path, &info))
-		return false;
-
-	if (info.isDirectory) {
-		File::GetFilesInDir(new_path, listing, filter);
-		return true;
-	}
-	return false;
-}
-
-bool DirectoryReader::GetFileInfo(const char *path, File::FileInfo *info) {
-	Path new_path = Path(path).StartsWith(path_) ? Path(path) : path_ / path;
-	return File::GetFileInfo(new_path, info);
-}
-
-class DirectoryReaderFileReference : public VFSFileReference {
-public:
-	Path path;
-};
-
-class DirectoryReaderOpenFile : public VFSOpenFile {
-public:
-	~DirectoryReaderOpenFile() {
-		_dbg_assert_(file == nullptr);
-	}
-	FILE *file = nullptr;
-};
-
-VFSFileReference *DirectoryReader::GetFile(const char *path) {
-	Path filePath = path_ / path;
-	if (!File::Exists(filePath)) {
-		return nullptr;
-	}
-
-	DirectoryReaderFileReference *reference = new DirectoryReaderFileReference();
-	reference->path = filePath;
-	return reference;
-}
-
-bool DirectoryReader::GetFileInfo(VFSFileReference *vfsReference, File::FileInfo *fileInfo) {
-	DirectoryReaderFileReference *reference = (DirectoryReaderFileReference *)vfsReference;
-	return File::GetFileInfo(reference->path, fileInfo);
-}
-
-void DirectoryReader::ReleaseFile(VFSFileReference *vfsReference) {
-	DirectoryReaderFileReference *reference = (DirectoryReaderFileReference *)vfsReference;
-	delete reference;
-}
-
-VFSOpenFile *DirectoryReader::OpenFileForRead(VFSFileReference *vfsReference, size_t *size) {
-	DirectoryReaderFileReference *reference = (DirectoryReaderFileReference *)vfsReference;
-	FILE *file = File::OpenCFile(reference->path, "rb");
-	if (!file) {
-		return nullptr;
-	}
-	fseek(file, 0, SEEK_END);
-	*size = ftell(file);
-	fseek(file, 0, SEEK_SET);
-	DirectoryReaderOpenFile *openFile = new DirectoryReaderOpenFile();
-	openFile->file = file;
-	return openFile;
-}
-
-void DirectoryReader::Rewind(VFSOpenFile *vfsOpenFile) {
-	DirectoryReaderOpenFile *openFile = (DirectoryReaderOpenFile *)vfsOpenFile;
-	fseek(openFile->file, 0, SEEK_SET);
-}
-
-size_t DirectoryReader::Read(VFSOpenFile *vfsOpenFile, void *buffer, size_t length) {
-	DirectoryReaderOpenFile *openFile = (DirectoryReaderOpenFile *)vfsOpenFile;
-	return fread(buffer, 1, length, openFile->file);
-}
-
-void DirectoryReader::CloseFile(VFSOpenFile *vfsOpenFile) {
-	DirectoryReaderOpenFile *openFile = (DirectoryReaderOpenFile *)vfsOpenFile;
-	_dbg_assert_(openFile->file != nullptr);
-	fclose(openFile->file);
-	openFile->file = nullptr;
-	delete openFile;
-}

Common/File/VFS/DirectoryReader.h

@@ -1,30 +0,0 @@
-#pragma once
-
-#include "Common/File/VFS/VFS.h"
-#include "Common/File/FileUtil.h"
-#include "Common/File/Path.h"
-
-class DirectoryReader : public VFSBackend {
-public:
-	explicit DirectoryReader(const Path &path);
-	// use delete[] on the returned value.
-	uint8_t *ReadFile(const char *path, size_t *size) override;
-
-	VFSFileReference *GetFile(const char *path) override;
-	bool GetFileInfo(VFSFileReference *vfsReference, File::FileInfo *fileInfo) override;
-	void ReleaseFile(VFSFileReference *vfsReference) override;
-
-	VFSOpenFile *OpenFileForRead(VFSFileReference *vfsReference, size_t *size) override;
-	void Rewind(VFSOpenFile *vfsOpenFile) override;
-	size_t Read(VFSOpenFile *vfsOpenFile, void *buffer, size_t length) override;
-	void CloseFile(VFSOpenFile *vfsOpenFile) override;
-
-	bool GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter) override;
-	bool GetFileInfo(const char *path, File::FileInfo *info) override;
-	std::string toString() const override {
-		return path_.ToString();
-	}
-
-private:
-	Path path_;
-};

Common/File/VFS/VFS.cpp

@@ -1,34 +1,35 @@
-#include <cstring>
-
 #include "Common/Log.h"
 #include "Common/File/VFS/VFS.h"
-#include "Common/File/FileUtil.h"
+#include "Common/File/VFS/AssetReader.h"
 #include "Common/File/AndroidStorage.h"
-#include "Common/StringUtils.h"
 
-VFS g_VFS;
+struct VFSEntry {
+	const char *prefix;
+	AssetReader *reader;
+};
 
-void VFS::Register(const char *prefix, VFSBackend *reader) {
+static VFSEntry entries[16];
-	if (reader) {
+static int num_entries = 0;
-		entries_.push_back(VFSEntry{ prefix, reader });
+
-		DEBUG_LOG(IO, "Registered VFS for prefix %s: %s", prefix, reader->toString().c_str());
+void VFSRegister(const char *prefix, AssetReader *reader) {
-	} else {
+	entries[num_entries].prefix = prefix;
-		ERROR_LOG(IO, "Trying to register null VFS backend for prefix %s", prefix);
+	entries[num_entries].reader = reader;
-	}
+	DEBUG_LOG(IO, "Registered VFS for prefix %s: %s", prefix, reader->toString().c_str());
+	num_entries++;
 }
 
-void VFS::Clear() {
+void VFSShutdown() {
-	for (auto &entry : entries_) {
+	for (int i = 0; i < num_entries; i++) {
-		delete entry.reader;
+		delete entries[i].reader;
 	}
-	entries_.clear();
+	num_entries = 0;
 }
 
 // TODO: Use Path more.
 static bool IsLocalAbsolutePath(const char *path) {
 	bool isUnixLocal = path[0] == '/';
 #ifdef _WIN32
-	bool isWindowsLocal = (isalpha(path[0]) && path[1] == ':') || startsWith(path, "\\\\") || startsWith(path, "//");
+	bool isWindowsLocal = isalpha(path[0]) && path[1] == ':';
 #else
 	bool isWindowsLocal = false;
 #endif
@@ -37,7 +38,7 @@ static bool IsLocalAbsolutePath(const char *path) {
 }
 
 // The returned data should be free'd with delete[].
-uint8_t *VFS::ReadFile(const char *filename, size_t *size) {
+uint8_t *VFSReadFile(const char *filename, size_t *size) {
 	if (IsLocalAbsolutePath(filename)) {
 		// Local path, not VFS.
 		// INFO_LOG(IO, "Not a VFS path: %s . Reading local file.", filename);
@@ -46,13 +47,13 @@ uint8_t *VFS::ReadFile(const char *filename, size_t *size) {
 
 	int fn_len = (int)strlen(filename);
 	bool fileSystemFound = false;
-	for (const auto &entry : entries_) {
+	for (int i = 0; i < num_entries; i++) {
-		int prefix_len = (int)strlen(entry.prefix);
+		int prefix_len = (int)strlen(entries[i].prefix);
 		if (prefix_len >= fn_len) continue;
-		if (0 == memcmp(filename, entry.prefix, prefix_len)) {
+		if (0 == memcmp(filename, entries[i].prefix, prefix_len)) {
 			fileSystemFound = true;
 			// INFO_LOG(IO, "Prefix match: %s (%s) -> %s", entries[i].prefix, filename, filename + prefix_len);
-			uint8_t *data = entry.reader->ReadFile(filename + prefix_len, size);
+			uint8_t *data = entries[i].reader->ReadAsset(filename + prefix_len, size);
 			if (data)
 				return data;
 			else
@@ -66,7 +67,7 @@ uint8_t *VFS::ReadFile(const char *filename, size_t *size) {
 	return 0;
 }
 
-bool VFS::GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter) {
+bool VFSGetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter) {
 	if (IsLocalAbsolutePath(path)) {
 		// Local path, not VFS.
 		// INFO_LOG(IO, "Not a VFS path: %s . Reading local directory.", path);
@@ -76,12 +77,12 @@ bool VFS::GetFileListing(const char *path, std::vector<File::FileInfo> *listing,
 
 	int fn_len = (int)strlen(path);
 	bool fileSystemFound = false;
-	for (const auto &entry : entries_) {
+	for (int i = 0; i < num_entries; i++) {
-		int prefix_len = (int)strlen(entry.prefix);
+		int prefix_len = (int)strlen(entries[i].prefix);
 		if (prefix_len >= fn_len) continue;
-		if (0 == memcmp(path, entry.prefix, prefix_len)) {
+		if (0 == memcmp(path, entries[i].prefix, prefix_len)) {
 			fileSystemFound = true;
-			if (entry.reader->GetFileListing(path + prefix_len, listing, filter)) {
+			if (entries[i].reader->GetFileListing(path + prefix_len, listing, filter)) {
 				return true;
 			}
 		}
@@ -93,7 +94,7 @@ bool VFS::GetFileListing(const char *path, std::vector<File::FileInfo> *listing,
 	return false;
 }
 
-bool VFS::GetFileInfo(const char *path, File::FileInfo *info) {
+bool VFSGetFileInfo(const char *path, File::FileInfo *info) {
 	if (IsLocalAbsolutePath(path)) {
 		// Local path, not VFS.
 		// INFO_LOG(IO, "Not a VFS path: %s . Getting local file info.", path);
@@ -102,19 +103,19 @@ bool VFS::GetFileInfo(const char *path, File::FileInfo *info) {
 
 	bool fileSystemFound = false;
 	int fn_len = (int)strlen(path);
-	for (const auto &entry : entries_) {
+	for (int i = 0; i < num_entries; i++) {
-		int prefix_len = (int)strlen(entry.prefix);
+		int prefix_len = (int)strlen(entries[i].prefix);
 		if (prefix_len >= fn_len) continue;
-		if (0 == memcmp(path, entry.prefix, prefix_len)) {
+		if (0 == memcmp(path, entries[i].prefix, prefix_len)) {
 			fileSystemFound = true;
-			if (entry.reader->GetFileInfo(path + prefix_len, info))
+			if (entries[i].reader->GetFileInfo(path + prefix_len, info))
 				return true;
 			else
 				continue;
 		}
 	}
 	if (!fileSystemFound) {
-		ERROR_LOG(IO, "Missing filesystem for '%s'", path);
+		ERROR_LOG(IO, "Missing filesystem for %s", path);
 	}  // Otherwise, the file was just missing. No need to log.
 	return false;
 }

Common/File/VFS/VFS.h

@@ -1,83 +1,21 @@
 #pragma once
 
 #include <vector>
-#include <cstdint>
 
 #include "Common/File/DirListing.h"
 
-// Basic read-only virtual file system. Used to manage assets on Android, where we have to
+// Basic virtual file system. Used to manage assets on Android, where we have to
 // read them manually out of the APK zipfile, while being able to run on other
 // platforms as well with the appropriate directory set-up.
 
-// Note that this is kinda similar in concept to Core/MetaFileSystem.h, but that one
+class AssetReader;
-// is specifically for operations done by the emulated PSP, while this is for operations
-// on the system level, like loading assets, and maybe texture packs. Also, as mentioned,
-// this one is read-only, so a bit smaller and simpler.
 
-// VFSBackend instances can be used on their own, without the VFS, to serve as an abstraction of
+void VFSRegister(const char *prefix, AssetReader *reader);
-// a single directory or ZIP file.
+void VFSShutdown();
 
-// The VFSFileReference level of abstraction is there to hold things like zip file indices,
+// Use delete [] to release the returned memory.
-// for fast re-open etc.
+// Always allocates an extra zero byte at the end, so that it
-
+// can be used for text like shader sources.
-class VFSFileReference {
+uint8_t *VFSReadFile(const char *filename, size_t *size);
-public:
+bool VFSGetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter = 0);
-	virtual ~VFSFileReference() {}
+bool VFSGetFileInfo(const char *filename, File::FileInfo *fileInfo);
-};
-
-class VFSOpenFile {
-public:
-	virtual ~VFSOpenFile() {}
-};
-
-// Common interface parts between VFSBackend and VFS.
-// Sometimes you don't need the VFS multiplexing and only have a VFSBackend *, sometimes you do need it,
-// and it would be cool to be able to use the same interface, like when loading INI files.
-class VFSInterface {
-public:
-	virtual ~VFSInterface() {}
-	virtual uint8_t *ReadFile(const char *path, size_t *size) = 0;
-	// If listing already contains files, it'll be cleared.
-	virtual bool GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter = nullptr) = 0;
-};
-
-class VFSBackend : public VFSInterface {
-public:
-	virtual VFSFileReference *GetFile(const char *path) = 0;
-	virtual bool GetFileInfo(VFSFileReference *vfsReference, File::FileInfo *fileInfo) = 0;
-	virtual void ReleaseFile(VFSFileReference *vfsReference) = 0;
-
-	// Must write the size of the file to *size. Both backends can do this efficiently here,
-	// avoiding a call to GetFileInfo.
-	virtual VFSOpenFile *OpenFileForRead(VFSFileReference *vfsReference, size_t *size) = 0;
-	virtual void Rewind(VFSOpenFile *vfsOpenFile) = 0;
-	virtual size_t Read(VFSOpenFile *vfsOpenFile, void *buffer, size_t length) = 0;
-	virtual void CloseFile(VFSOpenFile *vfsOpenFile) = 0;
-
-	// Filter support is optional but nice to have
-	virtual bool GetFileInfo(const char *path, File::FileInfo *info) = 0;
-	virtual std::string toString() const = 0;
-};
-
-class VFS : public VFSInterface {
-public:
-	~VFS() { Clear(); }
-	void Register(const char *prefix, VFSBackend *reader);
-	void Clear();
-
-	// Use delete [] to release the returned memory.
-	// Always allocates an extra zero byte at the end, so that it
-	// can be used for text like shader sources.
-	uint8_t *ReadFile(const char *filename, size_t *size) override;
-	bool GetFileInfo(const char *filename, File::FileInfo *fileInfo);
-	bool GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter = nullptr) override;
-
-private:
-	struct VFSEntry {
-		const char *prefix;
-		VFSBackend *reader;
-	};
-	std::vector<VFSEntry> entries_;
-};
-
-extern VFS g_VFS;

Common/File/VFS/ZipFileReader.cpp

@@ -1,310 +0,0 @@
-#include <algorithm>
-#include <ctype.h>
-#include <set>
-#include <cstdio>
-#include <cstring>
-
-#ifdef SHARED_LIBZIP
-#include <zip.h>
-#else
-#include "ext/libzip/zip.h"
-#endif
-
-#include "Common/Common.h"
-#include "Common/Log.h"
-#include "Common/File/VFS/ZipFileReader.h"
-#include "Common/StringUtils.h"
-
-ZipFileReader *ZipFileReader::Create(const Path &zipFile, const char *inZipPath, bool logErrors) {
-	int error = 0;
-	zip *zip_file;
-	if (zipFile.Type() == PathType::CONTENT_URI) {
-		int fd = File::OpenFD(zipFile, File::OPEN_READ);
-		if (!fd) {
-			if (logErrors) {
-				ERROR_LOG(IO, "Failed to open FD for '%s' as zip file", zipFile.c_str());
-			}
-			return nullptr;
-		}
-		zip_file = zip_fdopen(fd, 0, &error);
-	} else {
-		zip_file = zip_open(zipFile.c_str(), 0, &error);
-	}
-
-	if (!zip_file) {
-		if (logErrors) {
-			ERROR_LOG(IO, "Failed to open %s as a zip file", zipFile.c_str());
-		}
-		return nullptr;
-	}
-
-	// The inZipPath is supposed to be a folder, and internally in this class, we suffix
-	// folder paths with '/', matching how the zip library works.
-	std::string path = inZipPath;
-	if (!path.empty() && path.back() != '/') {
-		path.push_back('/');
-	}
-	return new ZipFileReader(zip_file, path);
-}
-
-ZipFileReader::~ZipFileReader() {
-	std::lock_guard<std::mutex> guard(lock_);
-	zip_close(zip_file_);
-}
-
-uint8_t *ZipFileReader::ReadFile(const char *path, size_t *size) {
-	std::string temp_path = inZipPath_ + path;
-
-	std::lock_guard<std::mutex> guard(lock_);
-	// Figure out the file size first.
-	struct zip_stat zstat;
-	zip_stat(zip_file_, temp_path.c_str(), ZIP_FL_NOCASE | ZIP_FL_UNCHANGED, &zstat);
-	zip_file *file = zip_fopen(zip_file_, temp_path.c_str(), ZIP_FL_NOCASE | ZIP_FL_UNCHANGED);
-	if (!file) {
-		ERROR_LOG(IO, "Error opening %s from ZIP", temp_path.c_str());
-		return 0;
-	}
-	uint8_t *contents = new uint8_t[zstat.size + 1];
-	zip_fread(file, contents, zstat.size);
-	zip_fclose(file);
-	contents[zstat.size] = 0;
-
-	*size = zstat.size;
-	return contents;
-}
-
-bool ZipFileReader::GetFileListing(const char *orig_path, std::vector<File::FileInfo> *listing, const char *filter = 0) {
-	std::string path = std::string(inZipPath_) + orig_path;
-	if (!path.empty() && path.back() != '/') {
-		path.push_back('/');
-	}
-
-	std::set<std::string> filters;
-	std::string tmp;
-	if (filter) {
-		while (*filter) {
-			if (*filter == ':') {
-				filters.insert("." + tmp);
-				tmp.clear();
-			} else {
-				tmp.push_back(*filter);
-			}
-			filter++;
-		}
-	}
-
-	if (tmp.size())
-		filters.insert("." + tmp);
-
-	// We just loop through the whole ZIP file and deduce what files are in this directory, and what subdirectories there are.
-	std::set<std::string> files;
-	std::set<std::string> directories;
-	bool success = GetZipListings(path, files, directories);
-	if (!success) {
-		// This means that no file prefix matched the path.
-		return false;
-	}
-
-	listing->clear();
-
-	INFO_LOG(SYSTEM, "Listing %s", orig_path);
-
-	for (auto diter = directories.begin(); diter != directories.end(); ++diter) {
-		File::FileInfo info;
-		info.name = *diter;
-
-		// Remove the "inzip" part of the fullname.
-		std::string relativePath = std::string(path).substr(inZipPath_.size());
-		info.fullName = Path(relativePath + *diter);
-		info.exists = true;
-		info.isWritable = false;
-		info.isDirectory = true;
-		// INFO_LOG(SYSTEM, "Found file: %s (%s)", info.name.c_str(), info.fullName.c_str());
-		listing->push_back(info);
-	}
-
-	for (auto fiter = files.begin(); fiter != files.end(); ++fiter) {
-		std::string fpath = path;
-		File::FileInfo info;
-		info.name = *fiter;
-		std::string relativePath = std::string(path).substr(inZipPath_.size());
-		info.fullName = Path(relativePath + *fiter);
-		info.exists = true;
-		info.isWritable = false;
-		info.isDirectory = false;
-		std::string ext = info.fullName.GetFileExtension();
-		if (filter) {
-			if (filters.find(ext) == filters.end()) {
-				continue;
-			}
-		}
-		// INFO_LOG(SYSTEM, "Found dir: %s (%s)", info.name.c_str(), info.fullName.c_str());
-		listing->push_back(info);
-	}
-
-	std::sort(listing->begin(), listing->end());
-	return true;
-}
-
-// path here is from the root, so inZipPath needs to already be added.
-bool ZipFileReader::GetZipListings(const std::string &path, std::set<std::string> &files, std::set<std::string> &directories) {
-	_dbg_assert_(path.empty() || path.back() == '/');
-
-	std::lock_guard<std::mutex> guard(lock_);
-	int numFiles = zip_get_num_files(zip_file_);
-	bool anyPrefixMatched = false;
-	for (int i = 0; i < numFiles; i++) {
-		const char* name = zip_get_name(zip_file_, i, 0);
-		if (!name)
-			continue;  // shouldn't happen, I think
-		if (startsWith(name, path)) {
-			if (strlen(name) == path.size()) {
-				// Don't want to return the same folder.
-				continue;
-			}
-			const char *slashPos = strchr(name + path.size(), '/');
-			if (slashPos != 0) {
-				anyPrefixMatched = true;
-				// A directory. Let's pick off the only part we care about.
-				size_t offset = path.size();
-				std::string dirName = std::string(name + offset, slashPos - (name + offset));
-				// We might get a lot of these if the tree is deep. The std::set deduplicates.
-				directories.insert(dirName);
-			} else {
-				anyPrefixMatched = true;
-				// It's a file.
-				const char *fn = name + path.size();
-				files.insert(std::string(fn));
-			}
-		}
-	}
-	return anyPrefixMatched;
-}
-
-bool ZipFileReader::GetFileInfo(const char *path, File::FileInfo *info) {
-	struct zip_stat zstat;
-	std::string temp_path = inZipPath_ + path;
-
-	// Clear some things to start.
-	info->isDirectory = false;
-	info->isWritable = false;
-	info->size = 0;
-
-	{
-		std::lock_guard<std::mutex> guard(lock_);
-		if (0 != zip_stat(zip_file_, temp_path.c_str(), ZIP_FL_NOCASE | ZIP_FL_UNCHANGED, &zstat)) {
-			// ZIP files do not have real directories, so we'll end up here if we
-			// try to stat one. For now that's fine.
-			info->exists = false;
-			return false;
-		}
-	}
-
-	// Zips usually don't contain directory entries, but they may.
-	if ((zstat.valid & ZIP_STAT_NAME) != 0 && zstat.name) {
-		info->isDirectory = zstat.name[strlen(zstat.name) - 1] == '/';
-	}
-	if ((zstat.valid & ZIP_STAT_SIZE) != 0) {
-		info->size = zstat.size;
-	}
-
-	info->fullName = Path(path);
-	info->exists = true;
-	return true;
-}
-
-class ZipFileReaderFileReference : public VFSFileReference {
-public:
-	int zi;
-};
-
-class ZipFileReaderOpenFile : public VFSOpenFile {
-public:
-	~ZipFileReaderOpenFile() {
-		// Needs to be closed properly and unlocked.
-		_dbg_assert_(zf == nullptr);
-	}
-	ZipFileReaderFileReference *reference;
-	zip_file_t *zf = nullptr;
-};
-
-VFSFileReference *ZipFileReader::GetFile(const char *path) {
-	std::lock_guard<std::mutex> guard(lock_);
-	int zi = zip_name_locate(zip_file_, path, ZIP_FL_NOCASE);
-	if (zi < 0) {
-		// Not found.
-		return nullptr;
-	}
-	ZipFileReaderFileReference *ref = new ZipFileReaderFileReference();
-	ref->zi = zi;
-	return ref;
-}
-
-bool ZipFileReader::GetFileInfo(VFSFileReference *vfsReference, File::FileInfo *fileInfo) {
-	ZipFileReaderFileReference *reference = (ZipFileReaderFileReference *)vfsReference;
-	// If you crash here, you called this while having the lock held by having the file open.
-	// Don't do that, check the info before you open the file.
-	std::lock_guard<std::mutex> guard(lock_);
-	zip_stat_t zstat;
-	if (zip_stat_index(zip_file_, reference->zi, 0, &zstat) != 0)
-		return false;
-	*fileInfo = File::FileInfo{};
-	fileInfo->size = 0;
-	if (zstat.valid & ZIP_STAT_SIZE)
-		fileInfo->size = zstat.size;
-	return zstat.size;
-}
-
-void ZipFileReader::ReleaseFile(VFSFileReference *vfsReference) {
-	ZipFileReaderFileReference *reference = (ZipFileReaderFileReference *)vfsReference;
-	// Don't do anything other than deleting it.
-	delete reference;
-}
-
-VFSOpenFile *ZipFileReader::OpenFileForRead(VFSFileReference *vfsReference, size_t *size) {
-	ZipFileReaderFileReference *reference = (ZipFileReaderFileReference *)vfsReference;
-	ZipFileReaderOpenFile *openFile = new ZipFileReaderOpenFile();
-	openFile->reference = reference;
-	*size = 0;
-	// We only allow one file to be open for read concurrently. It's possible that this can be improved,
-	// especially if we only access by index like this.
-	lock_.lock();
-	zip_stat_t zstat;
-	if (zip_stat_index(zip_file_, reference->zi, 0, &zstat) != 0) {
-		lock_.unlock();
-		delete openFile;
-		return nullptr;
-	}
-
-	openFile->zf = zip_fopen_index(zip_file_, reference->zi, 0);
-	if (!openFile->zf) {
-		WARN_LOG(G3D, "File with index %d not found in zip", reference->zi);
-		lock_.unlock();
-		return nullptr;
-	}
-
-	*size = zstat.size;
-	// Intentionally leaving the mutex locked, will be closed in CloseFile.
-	return openFile;
-}
-
-void ZipFileReader::Rewind(VFSOpenFile *vfsOpenFile) {
-	ZipFileReaderOpenFile *openFile = (ZipFileReaderOpenFile *)vfsOpenFile;
-	// Close and re-open.
-	zip_fclose(openFile->zf);
-	openFile->zf = zip_fopen_index(zip_file_, openFile->reference->zi, 0);
-}
-
-size_t ZipFileReader::Read(VFSOpenFile *vfsOpenFile, void *buffer, size_t length) {
-	ZipFileReaderOpenFile *file = (ZipFileReaderOpenFile *)vfsOpenFile;
-	return zip_fread(file->zf, buffer, length);
-}
-
-void ZipFileReader::CloseFile(VFSOpenFile *vfsOpenFile) {
-	ZipFileReaderOpenFile *file = (ZipFileReaderOpenFile *)vfsOpenFile;
-	_dbg_assert_(file->zf != nullptr);
-	zip_fclose(file->zf);
-	file->zf = nullptr;
-	lock_.unlock();
-	delete file;
-}

Common/File/VFS/ZipFileReader.h

@@ -1,50 +0,0 @@
-#pragma once
-
-#ifdef SHARED_LIBZIP
-#include <zip.h>
-#else
-#include "ext/libzip/zip.h"
-#endif
-
-#include <mutex>
-#include <set>
-#include <string>
-
-#include "Common/File/VFS/VFS.h"
-#include "Common/File/FileUtil.h"
-#include "Common/File/Path.h"
-
-class ZipFileReader : public VFSBackend {
-public:
-	static ZipFileReader *Create(const Path &zipFile, const char *inZipPath, bool logErrors = true);
-	~ZipFileReader();
-
-	bool IsValid() const { return zip_file_ != nullptr; }
-
-	// use delete[] on the returned value.
-	uint8_t *ReadFile(const char *path, size_t *size) override;
-
-	VFSFileReference *GetFile(const char *path) override;
-	bool GetFileInfo(VFSFileReference *vfsReference, File::FileInfo *fileInfo) override;
-	void ReleaseFile(VFSFileReference *vfsReference) override;
-
-	VFSOpenFile *OpenFileForRead(VFSFileReference *vfsReference, size_t *size) override;
-	void Rewind(VFSOpenFile *vfsOpenFile) override;
-	size_t Read(VFSOpenFile *vfsOpenFile, void *buffer, size_t length) override;
-	void CloseFile(VFSOpenFile *vfsOpenFile) override;
-
-	bool GetFileListing(const char *path, std::vector<File::FileInfo> *listing, const char *filter) override;
-	bool GetFileInfo(const char *path, File::FileInfo *info) override;
-	std::string toString() const override {
-		return inZipPath_;
-	}
-
-private:
-	ZipFileReader(zip *zip_file, const std::string &inZipPath) : zip_file_(zip_file), inZipPath_(inZipPath) {}
-	// Path has to be either an empty string, or a string ending with a /.
-	bool GetZipListings(const std::string &path, std::set<std::string> &files, std::set<std::string> &directories);
-
-	zip *zip_file_ = nullptr;
-	std::mutex lock_;
-	std::string inZipPath_;
-};

Common/GL/GLInterface/EGL.cpp

@@ -0,0 +1,402 @@
+// Copyright 2012 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#include <array>
+#include <cstdlib>
+
+#include "Common/Log.h"
+#include "Common/GL/GLInterface/EGL.h"
+
+// Show the current FPS
+void cInterfaceEGL::Swap() {
+	eglSwapBuffers(egl_dpy, egl_surf);
+}
+
+void cInterfaceEGL::SwapInterval(int Interval) {
+	eglSwapInterval(egl_dpy, Interval);
+}
+
+void* cInterfaceEGL::GetFuncAddress(const std::string& name) {
+	return (void*)eglGetProcAddress(name.c_str());
+}
+
+void cInterfaceEGL::DetectMode() {
+	EGLint num_configs;
+	bool supportsGL = false, supportsGLES2 = false, supportsGLES3 = false;
+	static const int renderable_types[3] = {
+		EGL_OPENGL_BIT,
+		(1 << 6), /* EGL_OPENGL_ES3_BIT_KHR */
+		EGL_OPENGL_ES2_BIT,
+	};
+
+	static const char *renderable_names[3] = {
+		"OpenGL", "OpenGL ES 3", "OpenGL ES 2"
+	};
+
+	for (int i = 0; i < 3; i++) {
+		int renderable_type = renderable_types[i];
+		const char *renderable_name = renderable_names[i];
+		// attributes for a visual in RGBA format with at least
+		// 8 bits per color
+		int attribs[] = {
+			EGL_RENDERABLE_TYPE, renderable_type,
+			EGL_RED_SIZE, 8,
+			EGL_GREEN_SIZE, 8,
+			EGL_BLUE_SIZE, 8,
+			EGL_ALPHA_SIZE, 8,
+			EGL_DEPTH_SIZE, 16,
+			EGL_STENCIL_SIZE, 8,
+			EGL_SURFACE_TYPE, EGL_WINDOW_BIT,
+			EGL_TRANSPARENT_TYPE, EGL_NONE,
+			EGL_SAMPLES, 0,
+			EGL_NONE
+		};
+
+		// Get how many configs there are
+		if (!eglChooseConfig( egl_dpy, attribs, nullptr, 0, &num_configs)) {
+			EGL_ILOG("DetectMode: couldn't get an EGL visual config with renderable_type=%s", renderable_name);
+			continue;
+		}
+		EGL_ILOG("DetectMode: got an EGL visual config with renderable_type=%s", renderable_name);
+
+		EGLConfig* config = new EGLConfig[num_configs];
+
+		// Get all the configurations
+		if (!eglChooseConfig(egl_dpy, attribs, config, num_configs, &num_configs)) {
+			EGL_ILOG("DetectMode: couldn't choose an EGL visual config\n");
+			delete[] config;
+			continue;
+		}
+
+		for (int i = 0; i < num_configs; ++i) {
+			EGLint attribVal;
+			bool ret;
+			ret = eglGetConfigAttrib(egl_dpy, config[i], EGL_RENDERABLE_TYPE, &attribVal);
+			if (ret) {
+				if ((attribVal & EGL_OPENGL_BIT) && s_opengl_mode != GLInterfaceMode::MODE_DETECT_ES)
+					supportsGL = true;
+				if (attribVal & (1 << 6)) /* EGL_OPENGL_ES3_BIT_KHR */
+					supportsGLES3 = true;  // Apparently, this cannot be completely trusted so we implement a fallback to ES 2.0 below.
+				if (attribVal & EGL_OPENGL_ES2_BIT)
+					supportsGLES2 = true;
+			}
+		}
+		delete[] config;
+	}
+
+	if (supportsGL)
+		s_opengl_mode = GLInterfaceMode::MODE_OPENGL;
+	else if (supportsGLES3)
+		s_opengl_mode = GLInterfaceMode::MODE_OPENGLES3;
+	else if (supportsGLES2)
+		s_opengl_mode = GLInterfaceMode::MODE_OPENGLES2;
+
+	if (s_opengl_mode == GLInterfaceMode::MODE_DETECT) // Errored before we found a mode
+		s_opengl_mode = GLInterfaceMode::MODE_OPENGL; // Fall back to OpenGL
+}
+
+static void LogEGLConfig(EGLDisplay egl_dpy, EGLConfig config) {
+	EGLint red = 0, green = 0, blue = 0, alpha = 0, depth = 0, stencil = 0, format = -1, type;
+
+	struct {
+		EGLint value;
+		const char *name;
+	} vals[] = {
+		{ EGL_RED_SIZE, "EGL_RED_SIZE" },
+		{ EGL_GREEN_SIZE, "EGL_GREEN_SIZE" },
+		{ EGL_BLUE_SIZE, "EGL_BLUE_SIZE" },
+		{ EGL_ALPHA_SIZE, "EGL_ALPHA_SIZE" },
+		{ EGL_DEPTH_SIZE, "EGL_DEPTH_SIZE" },
+		{ EGL_STENCIL_SIZE, "EGL_STENCIL_SIZE" },
+		{ EGL_NATIVE_VISUAL_ID, "EGL_NATIVE_VISUAL_ID" },
+		{ EGL_NATIVE_VISUAL_TYPE, "EGL_NATIVE_VISUAL_TYPE" },
+		{ EGL_MAX_SWAP_INTERVAL, "EGL_MAX_SWAP_INTERVAL" },
+		{ EGL_MIN_SWAP_INTERVAL, "EGL_MIN_SWAP_INTERVAL" },
+		{ EGL_MIN_SWAP_INTERVAL, "EGL_MIN_SWAP_INTERVAL" },
+		{ EGL_NATIVE_RENDERABLE, "EGL_NATIVE_RENDERABLE" },
+		{ EGL_COLOR_BUFFER_TYPE, "EGL_COLOR_BUFFER_TYPE" },
+		{ EGL_BUFFER_SIZE, "EGL_BUFFER_SIZE" },
+		{ EGL_CONFIG_ID, "EGL_CONFIG_ID" },
+		{ EGL_SAMPLES, "EGL_SAMPLES" },
+	};
+	
+	for (int i = 0; i < (int)(sizeof(vals)/sizeof(vals[0])); i++) {
+		EGLint value;
+		eglGetConfigAttrib(egl_dpy, config, vals[i].value, &value);
+		EGL_ILOG("  %s = %d", vals[i].name, value);
+	}
+}
+
+const char *cInterfaceEGL::EGLGetErrorString(EGLint error) {
+	switch (error) {
+	case EGL_SUCCESS: return "EGL_SUCCESS";
+	case EGL_NOT_INITIALIZED: return "EGL_NOT_INITIALIZED";
+	case EGL_BAD_ACCESS: return "EGL_BAD_ACCESS";
+	case EGL_BAD_ALLOC: return "EGL_BAD_ALLOC";
+	case EGL_BAD_ATTRIBUTE: return "EGL_BAD_ATTRIBUTE";
+	case EGL_BAD_CONTEXT: return "EGL_BAD_CONTEXT";
+	case EGL_BAD_CONFIG: return "EGL_BAD_CONFIG";
+	case EGL_BAD_CURRENT_SURFACE: return "EGL_BAD_CURRENT_SURFACE";
+	case EGL_BAD_DISPLAY: return "EGL_BAD_DISPLAY";
+	case EGL_BAD_SURFACE: return "EGL_BAD_SURFACE";
+	case EGL_BAD_MATCH: return "EGL_BAD_MATCH";
+	case EGL_BAD_PARAMETER: return "EGL_BAD_PARAMETER";
+	case EGL_BAD_NATIVE_PIXMAP: return "EGL_BAD_NATIVE_PIXMAP";
+	case EGL_BAD_NATIVE_WINDOW: return "EGL_BAD_NATIVE_WINDOW";
+	case EGL_CONTEXT_LOST: return "EGL_CONTEXT_LOST";
+	default:
+		return "(UNKNOWN)";
+	}
+}
+
+bool cInterfaceEGL::ChooseAndCreate(void *window_handle, bool core, bool use565) {
+	int attribs32[] = {
+		EGL_RENDERABLE_TYPE, EGL_OPENGL_ES2_BIT,  // Keep this first!
+		EGL_RED_SIZE, 8,
+		EGL_GREEN_SIZE, 8,
+		EGL_BLUE_SIZE, 8,
+		EGL_ALPHA_SIZE, 8,
+		EGL_DEPTH_SIZE, 16,
+		EGL_STENCIL_SIZE, 8,
+		EGL_TRANSPARENT_TYPE, EGL_NONE,
+		EGL_NONE, 0
+	};
+	int attribs16[] = {
+		EGL_RENDERABLE_TYPE, EGL_OPENGL_ES2_BIT,  // Keep this first!
+		EGL_RED_SIZE, 5,
+		EGL_GREEN_SIZE, 6,
+		EGL_BLUE_SIZE, 5,
+		EGL_ALPHA_SIZE, 0,
+		EGL_DEPTH_SIZE, 16,
+		EGL_STENCIL_SIZE, 8,
+		EGL_TRANSPARENT_TYPE, EGL_NONE,
+		EGL_NONE, 0
+	};
+	int attribsFallback32[] = {
+		EGL_RENDERABLE_TYPE, EGL_OPENGL_ES2_BIT,  // Keep this first!
+		EGL_RED_SIZE, 8,
+		EGL_GREEN_SIZE, 8,
+		EGL_BLUE_SIZE, 8,
+		EGL_ALPHA_SIZE, 8,
+		EGL_DEPTH_SIZE, 16,
+		EGL_NONE, 0
+	};
+	int attribsFallback16[] = {
+		EGL_RENDERABLE_TYPE, EGL_OPENGL_ES2_BIT,  // Keep this first!
+		EGL_RED_SIZE, 5,
+		EGL_GREEN_SIZE, 6,
+		EGL_BLUE_SIZE, 5,
+		EGL_ALPHA_SIZE, 0,
+		EGL_DEPTH_SIZE, 16,
+		EGL_NONE, 0
+	};
+
+	int *attribs = attribs32;
+	int *attribsFallback = attribsFallback32;
+	if (use565) {
+		attribs = attribs16;
+		attribsFallback = attribsFallback16;
+	}
+
+	EGLint ctx_attribs[] = {
+		EGL_CONTEXT_CLIENT_VERSION, 2,
+		EGL_NONE, 0,
+		EGL_NONE, 0,
+		EGL_NONE, 0,
+		EGL_NONE, 0,
+	};
+
+	switch (s_opengl_mode) {
+	case MODE_OPENGL:
+		EGL_ILOG("Setting RENDERABLE_TYPE to EGL_OPENGL_BIT");
+		attribs[1] = EGL_OPENGL_BIT;
+		// 1 will be major version, and 3 the minor version.
+		ctx_attribs[2] = 0x30FB; /* EGL_CONTEXT_MINOR_VERSION_KHR */
+		// Let's always use a core profile here.
+		ctx_attribs[4] = 0x30FD; /* EGL_CONTEXT_OPENGL_PROFILE_MASK_KHR */
+		ctx_attribs[5] = 1; /* EGL_CONTEXT_OPENGL_CORE_PROFILE_BIT_KHR */
+		break;
+	case MODE_OPENGLES2:
+		EGL_ILOG("Setting RENDERABLE_TYPE to EGL_OPENGL_ES2_BIT");
+		attribs[1] = EGL_OPENGL_ES2_BIT;
+		ctx_attribs[1] = 2;
+		break;
+	case MODE_OPENGLES3:
+		EGL_ILOG("Setting RENDERABLE_TYPE to EGL_OPENGL_ES3_BIT_KHR");
+		attribs[1] = (1 << 6); /* EGL_OPENGL_ES3_BIT_KHR */
+		ctx_attribs[1] = 3;
+		break;
+	default:
+		EGL_ELOG("Unknown OpenGL mode set\n");
+		return false;
+		break;
+	}
+
+	EGL_ILOG("Calling eglChooseConfig to get number of configs (use16bit=%d)...", (int)use565);
+
+	EGLConfig *configs;
+	EGLint num_configs = 0;
+	if (!eglChooseConfig(egl_dpy, attribs, NULL, 0, &num_configs) || num_configs == 0) {
+		EGL_ILOG("Error: couldn't get a number of configs. Trying with fallback config (no stencil, not specifying transparent:none)\n");
+		attribsFallback[1] = attribs[1];
+		attribs = attribsFallback;
+		if (!eglChooseConfig(egl_dpy, attribs, NULL, 0, &num_configs) || num_configs == 0) {
+			eglTerminate(egl_dpy);
+			return false;
+		}
+	}
+
+	EGL_ILOG("Got %d configs. Now choosing...", num_configs);
+	configs = new EGLConfig[num_configs];
+
+	if (!eglChooseConfig(egl_dpy, attribs, configs, num_configs, &num_configs)) {
+		EGL_ELOG("Error: couldn't get an EGL visual config (num_configs=%d)! Terminating EGL.\n", num_configs);
+		eglTerminate(egl_dpy);
+		return false;
+	}
+
+	int chosenConfig = -1;
+	// Find our ideal config in the list. If it's there, use it, otherwise pick whatever the device wanted (#0)
+	int wantedAlpha = 8;
+	// Requiring alpha seems to be a problem on older devices. Let's see if this helps...
+	if (attribs[1] == EGL_OPENGL_ES2_BIT)
+		wantedAlpha = 0;
+	for (int i = 0; i < num_configs; i++) {
+		EGL_ILOG("Config %d:", i);
+		LogEGLConfig(egl_dpy, configs[i]);
+		int red, green, blue, alpha, depth, stencil;
+		eglGetConfigAttrib(egl_dpy, configs[i], EGL_RED_SIZE, &red);
+		eglGetConfigAttrib(egl_dpy, configs[i], EGL_GREEN_SIZE, &green);
+		eglGetConfigAttrib(egl_dpy, configs[i], EGL_BLUE_SIZE, &blue);
+		eglGetConfigAttrib(egl_dpy, configs[i], EGL_ALPHA_SIZE, &alpha);
+		eglGetConfigAttrib(egl_dpy, configs[i], EGL_DEPTH_SIZE, &depth);
+		eglGetConfigAttrib(egl_dpy, configs[i], EGL_STENCIL_SIZE, &stencil);
+		if (chosenConfig == -1 && red == 8 && green == 8 && blue == 8 && alpha == wantedAlpha && depth == 24 && stencil == 8) {
+			chosenConfig = i;
+		}
+	}
+	if (chosenConfig == -1)
+		chosenConfig = 0;
+
+	EGL_ILOG("eglChooseConfig successful: num_configs=%d, choosing config %d", num_configs, chosenConfig);
+
+	if (s_opengl_mode == MODE_OPENGL) {
+		EGL_ILOG("eglBindAPI(OPENGL)");
+		eglBindAPI(EGL_OPENGL_API);
+	} else {
+		EGL_ILOG("eglBindAPI(OPENGL_ES)");
+		eglBindAPI(EGL_OPENGL_ES_API);
+	}
+
+	EGLNativeWindowType host_window = (EGLNativeWindowType)window_handle;
+	EGLNativeWindowType native_window = InitializePlatform(host_window, configs[chosenConfig]);
+
+	const char *s = eglQueryString(egl_dpy, EGL_VERSION);
+	EGL_ILOG("EGL_VERSION = %s\n", s);
+
+	s = eglQueryString(egl_dpy, EGL_VENDOR);
+	EGL_ILOG("EGL_VENDOR = %s\n", s);
+
+	s = eglQueryString(egl_dpy, EGL_EXTENSIONS);
+	EGL_ILOG("EGL_EXTENSIONS = %s\n", s);
+
+	s = eglQueryString(egl_dpy, EGL_CLIENT_APIS);
+	EGL_ILOG("EGL_CLIENT_APIS = %s\n", s);
+
+
+	if (s_opengl_mode == MODE_OPENGL) {
+		EGL_ILOG("Finding a good GL version");
+		egl_ctx = nullptr;
+		for (int minor = 6; minor >= 0 && !egl_ctx; --minor) {
+			ctx_attribs[1] = 4;
+			ctx_attribs[3] = minor;
+			egl_ctx = eglCreateContext(egl_dpy, configs[chosenConfig], EGL_NO_CONTEXT, ctx_attribs);
+		}
+		if (!egl_ctx) {
+			ctx_attribs[1] = 3;
+			ctx_attribs[3] = 3;
+			egl_ctx = eglCreateContext(egl_dpy, configs[chosenConfig], EGL_NO_CONTEXT, ctx_attribs);
+		}
+	} else {
+		egl_ctx = eglCreateContext(egl_dpy, configs[chosenConfig], EGL_NO_CONTEXT, ctx_attribs);
+	}
+	if (!egl_ctx) {
+		EGL_ILOG("Error: eglCreateContext failed: %s\n", EGLGetErrorString(eglGetError()));
+		delete[] configs;
+		return false;
+	}
+	EGL_ILOG("Successfully created EGL context.\n");
+
+	egl_surf = eglCreateWindowSurface(egl_dpy, configs[chosenConfig], native_window, nullptr);
+	if (!egl_surf) {
+		EGL_ILOG("Error: eglCreateWindowSurface failed: native_window=%p error=%s ctx_attribs[1]==%d\n", native_window, EGLGetErrorString(eglGetError()), ctx_attribs[1]);
+		eglDestroyContext(egl_dpy, egl_ctx);
+		delete[] configs;
+		return false;
+	}
+	EGL_ILOG("Successfully created EGL window surface (window=%p).\n", native_window);
+
+	delete[] configs;
+	return true;
+}
+
+// Create rendering window.
+bool cInterfaceEGL::Create(void *window_handle, bool core, bool use565) {
+	EGLint egl_major, egl_minor;
+
+	egl_dpy = OpenDisplay();
+
+	if (!egl_dpy) {
+		EGL_ILOG("Error: eglGetDisplay() failed\n");
+		return false;
+	}
+
+	if (!eglInitialize(egl_dpy, &egl_major, &egl_minor)) {
+		EGL_ILOG("Error: eglInitialize() failed\n");
+		return false;
+	}
+	EGL_ILOG("eglInitialize() succeeded (use565=%d)\n", (int)use565);
+
+	if (s_opengl_mode == MODE_DETECT || s_opengl_mode == MODE_DETECT_ES)
+		DetectMode();
+
+	if (!ChooseAndCreate(window_handle, core, use565) && (s_opengl_mode == MODE_OPENGLES3 || s_opengl_mode == MODE_OPENGL)) {
+		// Fallback to ES 2.0 and try again.
+		s_opengl_mode = MODE_OPENGLES2;
+		if (!ChooseAndCreate(window_handle, core, use565)) {
+			eglTerminate(egl_dpy);
+			egl_dpy = nullptr;
+			return false;
+		}
+	}
+
+	return true;
+}
+
+bool cInterfaceEGL::MakeCurrent() {
+	return eglMakeCurrent(egl_dpy, egl_surf, egl_surf, egl_ctx);
+}
+
+bool cInterfaceEGL::ClearCurrent() {
+	return eglMakeCurrent(egl_dpy, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
+}
+
+void cInterfaceEGL::Shutdown() {
+	ShutdownPlatform();
+	if (egl_ctx && !eglMakeCurrent(egl_dpy, egl_surf, egl_surf, egl_ctx)) {
+		NOTICE_LOG(G3D, "Could not release drawing context.");
+	}
+	if (egl_ctx) {
+		eglMakeCurrent(egl_dpy, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
+		if (!eglDestroySurface(egl_dpy, egl_surf))
+			NOTICE_LOG(G3D, "Could not destroy window surface.");
+		if (!eglDestroyContext(egl_dpy, egl_ctx))
+			NOTICE_LOG(G3D, "Could not destroy drawing context.");
+		if (!eglTerminate(egl_dpy))
+			NOTICE_LOG(G3D, "Could not destroy display connection.");
+		egl_ctx = nullptr;
+		egl_dpy = nullptr;
+		egl_surf = nullptr;
+	}
+}

Common/GL/GLInterface/EGL.h

@@ -0,0 +1,42 @@
+// Copyright 2008 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <string>
+#include <EGL/egl.h>
+
+#include "Common/Log.h"
+#include "Common/GL/GLInterfaceBase.h"
+
+#define EGL_ILOG(...) INFO_LOG(G3D, __VA_ARGS__)
+#define EGL_ELOG(...) INFO_LOG(G3D, __VA_ARGS__)
+
+
+class cInterfaceEGL : public cInterfaceBase {
+public:
+	void SwapInterval(int Interval) override;
+	void Swap() override;
+	void SetMode(u32 mode) override { s_opengl_mode = mode; }
+	void* GetFuncAddress(const std::string& name) override;
+	bool Create(void *window_handle, bool core, bool use565) override;
+	bool MakeCurrent() override;
+	bool ClearCurrent() override;
+	void Shutdown() override;
+
+protected:
+	EGLSurface egl_surf;
+	EGLContext egl_ctx;
+	EGLDisplay egl_dpy;
+
+	virtual EGLDisplay OpenDisplay() = 0;
+	virtual EGLNativeWindowType InitializePlatform(EGLNativeWindowType host_window, EGLConfig config) = 0;
+	virtual void ShutdownPlatform() = 0;
+	virtual void SetInternalResolution(int internalWidth, int internalHeight) {}
+	const char *EGLGetErrorString(EGLint error);
+
+private:
+	bool ChooseAndCreate(void *window_handle, bool core, bool use565);
+	void DetectMode();
+};

Common/GL/GLInterface/EGLAndroid.cpp

@@ -0,0 +1,31 @@
+// Copyright 2014 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#include <android/native_window.h>
+#include "Common/Log.h"
+#include "Common/GL/GLInterface/EGLAndroid.h"
+
+EGLDisplay cInterfaceEGLAndroid::OpenDisplay() {
+	return eglGetDisplay(EGL_DEFAULT_DISPLAY);
+}
+
+EGLNativeWindowType cInterfaceEGLAndroid::InitializePlatform(EGLNativeWindowType host_window, EGLConfig config) {
+	EGLint format;
+	if (EGL_FALSE == eglGetConfigAttrib(egl_dpy, config, EGL_NATIVE_VISUAL_ID, &format)) {
+		EGL_ELOG("Failed getting EGL_NATIVE_VISUAL_ID: error %s", EGLGetErrorString(eglGetError()));
+		return NULL;
+	}
+
+	int32_t result = ANativeWindow_setBuffersGeometry(host_window, internalWidth_, internalHeight_, format);
+	EGL_ILOG("ANativeWindow_setBuffersGeometry returned %d", result);
+
+	const int width = ANativeWindow_getWidth(host_window);
+	const int height = ANativeWindow_getHeight(host_window);
+	SetBackBufferDimensions(width, height);
+
+	return host_window;
+}
+
+void cInterfaceEGLAndroid::ShutdownPlatform() {
+}

Common/GL/GLInterface/EGLAndroid.h

@@ -0,0 +1,24 @@
+// Copyright 2014 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "Common/GL/GLInterface/EGL.h"
+
+class cInterfaceEGLAndroid : public cInterfaceEGL {
+public:
+	cInterfaceEGLAndroid() : internalWidth_(0), internalHeight_(0) {}
+protected:
+	EGLDisplay OpenDisplay() override;
+	EGLNativeWindowType InitializePlatform(EGLNativeWindowType host_window, EGLConfig config) override;
+	void ShutdownPlatform() override;
+	void OverrideBackbufferDimensions(int internalWidth, int internalHeight) override {
+		internalWidth_ = internalWidth;
+		internalHeight_ = internalHeight;
+	}
+
+private:
+	int internalWidth_;
+	int internalHeight_;
+};

Common/GL/GLInterface/EGLSwitch.cpp

@@ -0,0 +1,21 @@
+// Copyright 2014 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#include "ppsspp_config.h"
+#if PPSSPP_PLATFORM(SWITCH)
+#include <switch.h>
+#include "Common/Log.h"
+#include "Common/GL/GLInterface/EGLSwitch.h"
+
+EGLDisplay cInterfaceEGLSwitch::OpenDisplay() {
+	return eglGetDisplay(EGL_DEFAULT_DISPLAY);
+}
+
+EGLNativeWindowType cInterfaceEGLSwitch::InitializePlatform(EGLNativeWindowType host_window, EGLConfig config) {
+	return nwindowGetDefault();
+}
+
+void cInterfaceEGLSwitch::ShutdownPlatform() {
+}
+#endif

Common/GL/GLInterface/EGLSwitch.h

@@ -0,0 +1,24 @@
+// Copyright 2014 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "Common/GL/GLInterface/EGL.h"
+
+class cInterfaceEGLSwitch : public cInterfaceEGL {
+public:
+	cInterfaceEGLSwitch() {}
+protected:
+	EGLDisplay OpenDisplay() override;
+	EGLNativeWindowType InitializePlatform(EGLNativeWindowType host_window, EGLConfig config) override;
+	void ShutdownPlatform() override;
+	void OverrideBackbufferDimensions(int internalWidth, int internalHeight) override {
+		internalWidth_ = internalWidth;
+		internalHeight_ = internalHeight;
+	}
+
+private:
+	int internalWidth_ = 0;
+	int internalHeight_ = 0;
+};

Common/GL/GLInterface/GLInterface.cpp

@@ -0,0 +1,44 @@
+// Copyright 2014 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#include "ppsspp_config.h"
+#include "Common/GL/GLInterfaceBase.h"
+
+#ifdef __ANDROID__
+#include "Common/GL/GLInterface/EGLAndroid.h"
+#elif PPSSPP_PLATFORM(SWITCH)
+#include "Common/GL/GLInterface/EGLSwitch.h"
+#elif defined(__APPLE__)
+#include "Common/GL/GLInterface/AGL.h"
+#elif defined(_WIN32)
+#include "Common/GL/GLInterface/WGL.h"
+#elif HAVE_X11
+#if defined(USE_EGL) && USE_EGL
+#include "Common/GL/GLInterface/EGLX11.h"
+#else
+#include "Common/GL/GLInterface/GLX.h"
+#endif
+#else
+#error Platform doesnt have a GLInterface
+#endif
+
+cInterfaceBase* HostGL_CreateGLInterface(){
+	#ifdef __ANDROID__
+		return new cInterfaceEGLAndroid;
+	#elif PPSSPP_PLATFORM(SWITCH)
+		return new cInterfaceEGLSwitch;
+	#elif defined(__APPLE__)
+		return new cInterfaceAGL;
+	#elif defined(_WIN32)
+		return new cInterfaceWGL;
+	#elif defined(HAVE_X11) && HAVE_X11
+	#if defined(USE_EGL) && USE_EGL
+		return new cInterfaceEGLX11;
+	#else
+		return new cInterfaceGLX;
+	#endif
+	#else
+		return nullptr;
+	#endif
+}

Common/GL/GLInterfaceBase.h

@@ -0,0 +1,50 @@
+// Copyright 2008 Dolphin Emulator Project
+// Licensed under GPLv2+
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <string>
+
+#include "Common/CommonTypes.h"
+
+enum GLInterfaceMode {
+	MODE_DETECT = 0,
+	MODE_DETECT_ES,
+	MODE_OPENGL,
+	MODE_OPENGLES2,
+	MODE_OPENGLES3,
+};
+
+class cInterfaceBase {
+protected:
+	// Window dimensions.
+	u32 s_backbuffer_width;
+	u32 s_backbuffer_height;
+
+	u32 s_opengl_mode;
+public:
+	cInterfaceBase() : s_backbuffer_width(0), s_backbuffer_height(0), s_opengl_mode(MODE_DETECT) {}
+	virtual ~cInterfaceBase() {}
+	virtual void Swap() {}
+	virtual void SetMode(u32 mode) { s_opengl_mode = GLInterfaceMode::MODE_OPENGL; }
+	virtual u32 GetMode() { return s_opengl_mode; }
+	virtual void* GetFuncAddress(const std::string& name) { return nullptr; }
+	virtual bool Create(void *window_handle, bool core = true, bool use16bit = false) = 0;
+	virtual bool MakeCurrent() { return true; }
+	virtual bool ClearCurrent() { return true; }
+	virtual void Shutdown() {}
+
+	virtual void SwapInterval(int Interval) { }
+	virtual u32 GetBackBufferWidth() { return s_backbuffer_width; }
+	virtual u32 GetBackBufferHeight() { return s_backbuffer_height; }
+
+	virtual void OverrideBackbufferDimensions(int w, int h) = 0;
+
+	virtual void SetBackBufferDimensions(u32 W, u32 H) {s_backbuffer_width = W; s_backbuffer_height = H; }
+	virtual void Update() { }
+	virtual bool PeekMessages() { return false; }
+};
+
+
+cInterfaceBase* HostGL_CreateGLInterface();

Common/GPU/D3D11/D3D11Loader.cpp

@@ -11,6 +11,7 @@ static HMODULE g_D3DCompileModule;
 
 LPCREATEDXGIFACTORY ptr_CreateDXGIFactory;
 LPD3D11CREATEDEVICE ptr_D3D11CreateDevice;
+LPD3D11CREATEDEVICEANDSWAPCHAIN ptr_D3D11CreateDeviceAndSwapChain;
 pD3DCompile ptr_D3DCompile;
 
 LoadD3D11Error LoadD3D11() {
@@ -21,6 +22,7 @@ LoadD3D11Error LoadD3D11() {
 	g_D3D11Module = LoadLibrary(L"d3d11.dll");
 	if (g_D3D11Module) {
 		ptr_D3D11CreateDevice = (LPD3D11CREATEDEVICE)GetProcAddress(g_D3D11Module, "D3D11CreateDevice");
+		ptr_D3D11CreateDeviceAndSwapChain = (LPD3D11CREATEDEVICEANDSWAPCHAIN)GetProcAddress(g_D3D11Module, "D3D11CreateDeviceAndSwapChain");
 	} else {
 		return LoadD3D11Error::FAIL_NO_D3D11;
 	}

Common/GPU/D3D11/D3D11Loader.h

@@ -14,10 +14,12 @@
 #endif
 
 typedef HRESULT (WINAPI *LPCREATEDXGIFACTORY)(REFIID, void **);
+typedef HRESULT (WINAPI *LPD3D11CREATEDEVICEANDSWAPCHAIN)(__in_opt IDXGIAdapter *pAdapter, D3D_DRIVER_TYPE DriverType, HMODULE Software, UINT Flags, __in_ecount_opt(FeatureLevels) CONST D3D_FEATURE_LEVEL *pFeatureLevels, UINT FeatureLevels, UINT SDKVersion, __in_opt CONST DXGI_SWAP_CHAIN_DESC *pSwapChainDesc, __out_opt IDXGISwapChain **ppSwapChain, __out_opt ID3D11Device **ppDevice, __out_opt D3D_FEATURE_LEVEL *pFeatureLevel, __out_opt ID3D11DeviceContext **ppImmediateContext);
 typedef HRESULT (WINAPI *LPD3D11CREATEDEVICE)(IDXGIAdapter *, D3D_DRIVER_TYPE, HMODULE, UINT32, D3D_FEATURE_LEVEL *, UINT, UINT32, ID3D11Device **, D3D_FEATURE_LEVEL *, ID3D11DeviceContext **);
 
 extern LPCREATEDXGIFACTORY ptr_CreateDXGIFactory;
 extern LPD3D11CREATEDEVICE ptr_D3D11CreateDevice;
+extern LPD3D11CREATEDEVICEANDSWAPCHAIN ptr_D3D11CreateDeviceAndSwapChain;
 extern pD3DCompile ptr_D3DCompile;
 
 enum class LoadD3D11Error {

Common/GPU/D3D11/thin3d_d3d11.cpp

@@ -11,7 +11,6 @@
 #include "Common/Data/Convert/ColorConv.h"
 #include "Common/Data/Convert/SmallDataConvert.h"
 #include "Common/Data/Encoding/Utf8.h"
-#include "Common/TimeUtil.h"
 #include "Common/Log.h"
 
 #include <map>
@@ -63,7 +62,7 @@ public:
 
 class D3D11DrawContext : public DrawContext {
 public:
-	D3D11DrawContext(ID3D11Device *device, ID3D11DeviceContext *deviceContext, ID3D11Device1 *device1, ID3D11DeviceContext1 *deviceContext1, IDXGISwapChain *swapChain, D3D_FEATURE_LEVEL featureLevel, HWND hWnd, std::vector<std::string> deviceList, int maxInflightFrames);
+	D3D11DrawContext(ID3D11Device *device, ID3D11DeviceContext *deviceContext, ID3D11Device1 *device1, ID3D11DeviceContext1 *deviceContext1, D3D_FEATURE_LEVEL featureLevel, HWND hWnd, std::vector<std::string> deviceList);
 	~D3D11DrawContext();
 
 	const DeviceCaps &GetDeviceCaps() const override {
@@ -89,11 +88,10 @@ public:
 	Framebuffer *CreateFramebuffer(const FramebufferDesc &desc) override;
 
 	void UpdateBuffer(Buffer *buffer, const uint8_t *data, size_t offset, size_t size, UpdateBufferFlags flags) override;
-	void UpdateTextureLevels(Texture *texture, const uint8_t **data, TextureCallback initDataCallback, int numLevels) override;
 
 	void CopyFramebufferImage(Framebuffer *src, int level, int x, int y, int z, Framebuffer *dst, int dstLevel, int dstX, int dstY, int dstZ, int width, int height, int depth, int channelBits, const char *tag) override;
 	bool BlitFramebuffer(Framebuffer *src, int srcX1, int srcY1, int srcX2, int srcY2, Framebuffer *dst, int dstX1, int dstY1, int dstX2, int dstY2, int channelBits, FBBlitFilter filter, const char *tag) override;
-	bool CopyFramebufferToMemory(Framebuffer *src, int channelBits, int x, int y, int w, int h, Draw::DataFormat format, void *pixels, int pixelStride, ReadbackMode mode, const char *tag) override;
+	bool CopyFramebufferToMemorySync(Framebuffer *src, int channelBits, int x, int y, int w, int h, Draw::DataFormat format, void *pixels, int pixelStride, const char *tag) override;
 
 	// These functions should be self explanatory.
 	void BindFramebufferAsRenderTarget(Framebuffer *fbo, const RenderPassInfo &rp, const char *tag) override;
@@ -106,7 +104,7 @@ public:
 	void BindTextures(int start, int count, Texture **textures, TextureBindFlags flags) override;
 	void BindNativeTexture(int index, void *nativeTexture) override;
 	void BindSamplerStates(int start, int count, SamplerState **states) override;
-	void BindVertexBuffer(Buffer *buffers, int offset) override;
+	void BindVertexBuffers(int start, int count, Buffer **buffers, const int *offsets) override;
 	void BindIndexBuffer(Buffer *indexBuffer, int offset) override;
 	void BindPipeline(Pipeline *pipeline) override;
 
@@ -114,7 +112,7 @@ public:
 
 	// Raster state
 	void SetScissorRect(int left, int top, int width, int height) override;
-	void SetViewport(const Viewport &viewport) override;
+	void SetViewports(int count, Viewport *viewports) override;
 	void SetBlendFactor(float color[4]) override {
 		if (memcmp(blendFactor_, color, sizeof(float) * 4)) {
 			memcpy(blendFactor_, color, sizeof(float) * 4);
@@ -128,17 +126,14 @@ public:
 		stencilDirty_ = true;
 	}
 
+	void EndFrame() override;
 
 	void Draw(int vertexCount, int offset) override;
 	void DrawIndexed(int vertexCount, int offset) override;
 	void DrawUP(const void *vdata, int vertexCount) override;
 	void Clear(int mask, uint32_t colorval, float depthVal, int stencilVal) override;
 
-	void BeginFrame(DebugFlags debugFlags) override;
+	void BeginFrame() override;
-	void EndFrame() override;
-	void Present(PresentMode presentMode, int vblanks) override;
-
-	int GetFrameCount() override { return frameCount_; }
 
 	std::string GetInfoString(InfoField info) const override {
 		switch (info) {
@@ -179,10 +174,9 @@ private:
 
 	HWND hWnd_;
 	ID3D11Device *device_;
-	ID3D11Device1 *device1_;
 	ID3D11DeviceContext *context_;
+	ID3D11Device1 *device1_;
 	ID3D11DeviceContext1 *context1_;
-	IDXGISwapChain *swapChain_;
 
 	ID3D11Texture2D *bbRenderTargetTex_ = nullptr; // NOT OWNED
 	ID3D11RenderTargetView *bbRenderTargetView_ = nullptr;
@@ -214,15 +208,14 @@ private:
 	ID3D11GeometryShader *curGS_ = nullptr;
 	D3D11_PRIMITIVE_TOPOLOGY curTopology_ = D3D11_PRIMITIVE_TOPOLOGY_UNDEFINED;
 
-	ID3D11Buffer *nextVertexBuffer_ = nullptr;
+	ID3D11Buffer *nextVertexBuffers_[4]{};
-	UINT nextVertexBufferOffset_ = 0;
+	int nextVertexBufferOffsets_[4]{};
 
 	bool dirtyIndexBuffer_ = false;
 	ID3D11Buffer *nextIndexBuffer_ = nullptr;
-	UINT nextIndexBufferOffset_ = 0;
+	int nextIndexBufferOffset_ = 0;
 
 	InvalidationCallback invalidationCallback_;
-	int frameCount_ = FRAME_TIME_HISTORY_LENGTH;
 
 	// Dynamic state
 	float blendFactor_[4]{};
@@ -243,14 +236,13 @@ private:
 	std::vector<std::string> deviceList_;
 };
 
-D3D11DrawContext::D3D11DrawContext(ID3D11Device *device, ID3D11DeviceContext *deviceContext, ID3D11Device1 *device1, ID3D11DeviceContext1 *deviceContext1, IDXGISwapChain *swapChain, D3D_FEATURE_LEVEL featureLevel, HWND hWnd, std::vector<std::string> deviceList, int maxInflightFrames)
+D3D11DrawContext::D3D11DrawContext(ID3D11Device *device, ID3D11DeviceContext *deviceContext, ID3D11Device1 *device1, ID3D11DeviceContext1 *deviceContext1, D3D_FEATURE_LEVEL featureLevel, HWND hWnd, std::vector<std::string> deviceList)
 	: hWnd_(hWnd),
 		device_(device),
 		context_(deviceContext1),
 		device1_(device1),
 		context1_(deviceContext1),
 		featureLevel_(featureLevel),
-		swapChain_(swapChain),
 		deviceList_(deviceList) {
 
 	// We no longer support Windows Phone.
@@ -281,10 +273,6 @@ D3D11DrawContext::D3D11DrawContext(ID3D11Device *device, ID3D11DeviceContext *de
 	caps_.blendMinMaxSupported = true;
 	caps_.multiSampleLevelsMask = 1;   // More could be supported with some work.
 
-	caps_.presentInstantModeChange = true;
-	caps_.presentMaxInterval = 4;
-	caps_.presentModesSupported = PresentMode::FIFO | PresentMode::IMMEDIATE;
-
 	D3D11_FEATURE_DATA_D3D11_OPTIONS options{};
 	HRESULT result = device_->CheckFeatureSupport(D3D11_FEATURE_D3D11_OPTIONS, &options, sizeof(options));
 	if (SUCCEEDED(result)) {
@@ -350,13 +338,6 @@ D3D11DrawContext::D3D11DrawContext(ID3D11Device *device, ID3D11DeviceContext *de
 	const size_t UP_MAX_BYTES = 65536 * 24;
 
 	upBuffer_ = CreateBuffer(UP_MAX_BYTES, BufferUsageFlag::DYNAMIC | BufferUsageFlag::VERTEXDATA);
-
-	IDXGIDevice1 *dxgiDevice1 = nullptr;
-	hr = device_->QueryInterface(__uuidof(IDXGIDevice), reinterpret_cast<void **>(&dxgiDevice1));
-	if (SUCCEEDED(hr)) {
-		caps_.setMaxFrameLatencySupported = true;
-		dxgiDevice1->SetMaximumFrameLatency(maxInflightFrames);
-	}
 }
 
 D3D11DrawContext::~D3D11DrawContext() {
@@ -427,43 +408,32 @@ void D3D11DrawContext::HandleEvent(Event ev, int width, int height, void *param1
 		curRTHeight_ = height;
 		break;
 	}
+	case Event::PRESENTED:
+		// Make sure that we don't eliminate the next time the render target is set.
+		curRenderTargetView_ = nullptr;
+		curDepthStencilView_ = nullptr;
+		break;
 	}
 }
 
 void D3D11DrawContext::EndFrame() {
-	// Fake a submit time.
-	frameTimeHistory_[frameCount_].firstSubmit = time_now_d();
 	curPipeline_ = nullptr;
 }
 
-void D3D11DrawContext::Present(PresentMode presentMode, int vblanks) {
+void D3D11DrawContext::SetViewports(int count, Viewport *viewports) {
-	frameTimeHistory_[frameCount_].queuePresent = time_now_d();
+	D3D11_VIEWPORT vp[4];
-
+	for (int i = 0; i < count; i++) {
-	int interval = vblanks;
+		DisplayRect<float> rc{ viewports[i].TopLeftX , viewports[i].TopLeftY, viewports[i].Width, viewports[i].Height };
-	if (presentMode != PresentMode::FIFO) {
+		if (curRenderTargetView_ == bbRenderTargetView_)  // Only the backbuffer is actually rotated wrong!
-		interval = 0;
+			RotateRectToDisplay(rc, curRTWidth_, curRTHeight_);
+		vp[i].TopLeftX = rc.x;
+		vp[i].TopLeftY = rc.y;
+		vp[i].Width = rc.w;
+		vp[i].Height = rc.h;
+		vp[i].MinDepth = viewports[i].MinDepth;
+		vp[i].MaxDepth = viewports[i].MaxDepth;
 	}
-	// Safety for libretro
+	context_->RSSetViewports(count, vp);
-	if (swapChain_) {
-		swapChain_->Present(interval, 0);
-	}
-	curRenderTargetView_ = nullptr;
-	curDepthStencilView_ = nullptr;
-	frameCount_++;
-}
-
-void D3D11DrawContext::SetViewport(const Viewport &viewport) {
-	DisplayRect<float> rc{ viewport.TopLeftX , viewport.TopLeftY, viewport.Width, viewport.Height };
-	if (curRenderTargetView_ == bbRenderTargetView_)  // Only the backbuffer is actually rotated wrong!
-		RotateRectToDisplay(rc, curRTWidth_, curRTHeight_);
-	D3D11_VIEWPORT vp;
-	vp.TopLeftX = rc.x;
-	vp.TopLeftY = rc.y;
-	vp.Width = rc.w;
-	vp.Height = rc.h;
-	vp.MinDepth = viewport.MinDepth;
-	vp.MaxDepth = viewport.MaxDepth;
-	context_->RSSetViewports(1, &vp);
 }
 
 void D3D11DrawContext::SetScissorRect(int left, int top, int width, int height) {
@@ -523,12 +493,7 @@ static DXGI_FORMAT dataFormatToD3D11(DataFormat format) {
 	case DataFormat::D16: return DXGI_FORMAT_D16_UNORM;
 	case DataFormat::D32F: return DXGI_FORMAT_D32_FLOAT;
 	case DataFormat::D32F_S8: return DXGI_FORMAT_D32_FLOAT_S8X24_UINT;
-	case DataFormat::BC1_RGBA_UNORM_BLOCK: return DXGI_FORMAT_BC1_UNORM;
+	case DataFormat::ETC1:
-	case DataFormat::BC2_UNORM_BLOCK: return DXGI_FORMAT_BC2_UNORM;
-	case DataFormat::BC3_UNORM_BLOCK: return DXGI_FORMAT_BC3_UNORM;
-	case DataFormat::BC4_UNORM_BLOCK: return DXGI_FORMAT_BC4_UNORM;
-	case DataFormat::BC5_UNORM_BLOCK: return DXGI_FORMAT_BC5_UNORM;
-	case DataFormat::BC7_UNORM_BLOCK: return DXGI_FORMAT_BC7_UNORM;
 	default:
 		return DXGI_FORMAT_UNKNOWN;
 	}
@@ -725,7 +690,7 @@ public:
 	D3D11InputLayout() {}
 	InputLayoutDesc desc;
 	std::vector<D3D11_INPUT_ELEMENT_DESC> elements;
-	UINT stride;  // type to match function parameter
+	std::vector<int> strides;
 };
 
 const char *semanticToD3D11(int semantic, UINT *index) {
@@ -752,13 +717,15 @@ InputLayout *D3D11DrawContext::CreateInputLayout(const InputLayoutDesc &desc) {
 		D3D11_INPUT_ELEMENT_DESC el;
 		el.AlignedByteOffset = desc.attributes[i].offset;
 		el.Format = dataFormatToD3D11(desc.attributes[i].format);
-		el.InstanceDataStepRate = 0;
+		el.InstanceDataStepRate = desc.bindings[desc.attributes[i].binding].instanceRate ? 1 : 0;
-		el.InputSlot = 0;
+		el.InputSlot = desc.attributes[i].binding;
 		el.SemanticName = semanticToD3D11(desc.attributes[i].location, &el.SemanticIndex);
-		el.InputSlotClass = D3D11_INPUT_PER_VERTEX_DATA;
+		el.InputSlotClass = desc.bindings[desc.attributes[i].binding].instanceRate ? D3D11_INPUT_PER_INSTANCE_DATA : D3D11_INPUT_PER_VERTEX_DATA;
 		inputLayout->elements.push_back(el);
 	}
-	inputLayout->stride = desc.stride;
+	for (size_t i = 0; i < desc.bindings.size(); i++) {
+		inputLayout->strides.push_back(desc.bindings[i].stride);
+	}
 	return inputLayout;
 }
 
@@ -821,83 +788,35 @@ public:
 		width_ = desc.width;
 		height_ = desc.height;
 		depth_ = desc.depth;
-		format_ = desc.format;
-		mipLevels_ = desc.mipLevels;
 	}
 	~D3D11Texture() {
-		if (tex_)
+		if (tex)
-			tex_->Release();
+			tex->Release();
-		if (stagingTex_)
+		if (stagingTex)
-			stagingTex_->Release();
+			stagingTex->Release();
-		if (view_)
+		if (view)
-			view_->Release();
+			view->Release();
 	}
 
-	bool Create(ID3D11DeviceContext *context, ID3D11Device *device, const TextureDesc &desc, bool generateMips);
+	ID3D11Texture2D *tex = nullptr;
-
+	ID3D11Texture2D *stagingTex = nullptr;
-	bool CreateStagingTexture(ID3D11Device *device);
+	ID3D11ShaderResourceView *view = nullptr;
-	void UpdateTextureLevels(ID3D11DeviceContext *context, ID3D11Device *device, Texture *texture, const uint8_t *const *data, TextureCallback initDataCallback, int numLevels);
-
-	ID3D11ShaderResourceView *View() { return view_; }
-
-private:
-	bool FillLevel(ID3D11DeviceContext *context, int level, int w, int h, int d, const uint8_t *const *data, TextureCallback initDataCallback);
-
-	ID3D11Texture2D *tex_ = nullptr;
-	ID3D11Texture2D *stagingTex_ = nullptr;
-	ID3D11ShaderResourceView *view_ = nullptr;
-	int mipLevels_ = 0;
 };
 
-bool D3D11Texture::FillLevel(ID3D11DeviceContext *context, int level, int w, int h, int d, const uint8_t *const *data, TextureCallback initDataCallback) {
+Texture *D3D11DrawContext::CreateTexture(const TextureDesc &desc) {
-	D3D11_MAPPED_SUBRESOURCE mapped;
+	if (!(GetDataFormatSupport(desc.format) & FMT_TEXTURE)) {
-	HRESULT hr = context->Map(stagingTex_, level, D3D11_MAP_WRITE, 0, &mapped);
+		// D3D11 does not support this format as a texture format.
-	if (!SUCCEEDED(hr)) {
+		return nullptr;
-		tex_->Release();
-		tex_ = nullptr;
-		return false;
 	}
 
-	if (!initDataCallback((uint8_t *)mapped.pData, data[level], w, h, d, mapped.RowPitch, mapped.DepthPitch)) {
+	D3D11Texture *tex = new D3D11Texture(desc);
-		for (int s = 0; s < d; ++s) {
+
-			for (int y = 0; y < h; ++y) {
+	bool generateMips = desc.generateMips;
-				void *dest = (uint8_t *)mapped.pData + mapped.DepthPitch * s + mapped.RowPitch * y;
+	if (desc.generateMips && !(GetDataFormatSupport(desc.format) & FMT_AUTOGEN_MIPS)) {
-				uint32_t byteStride = w * (uint32_t)DataFormatSizeInBytes(format_);
+		// D3D11 does not support autogenerating mipmaps for this format.
-				const void *src = data[level] + byteStride * (y + h * s);
+		generateMips = false;
-				memcpy(dest, src, byteStride);
-			}
-		}
 	}
-	context->Unmap(stagingTex_, level);
-	return true;
-}
 
-bool D3D11Texture::CreateStagingTexture(ID3D11Device *device) {
-	if (stagingTex_)
-		return true;
-	D3D11_TEXTURE2D_DESC descColor{};
-	descColor.Width = width_;
-	descColor.Height = height_;
-	descColor.MipLevels = mipLevels_;
-	descColor.ArraySize = 1;
-	descColor.Format = dataFormatToD3D11(format_);
-	descColor.SampleDesc.Count = 1;
-	descColor.SampleDesc.Quality = 0;
-	descColor.Usage = D3D11_USAGE_STAGING;
-	descColor.BindFlags = 0;
-	descColor.MiscFlags = 0;
-	descColor.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;
-
-	HRESULT hr = device->CreateTexture2D(&descColor, nullptr, &stagingTex_);
-	if (!SUCCEEDED(hr)) {
-		stagingTex_->Release();
-		stagingTex_ = nullptr;
-		return false;
-	}
-	return true;
-}
-
-bool D3D11Texture::Create(ID3D11DeviceContext *context, ID3D11Device *device, const TextureDesc &desc, bool generateMips) {
 	D3D11_TEXTURE2D_DESC descColor{};
 	descColor.Width = desc.width;
 	descColor.Height = desc.height;
@@ -906,16 +825,25 @@ bool D3D11Texture::Create(ID3D11DeviceContext *context, ID3D11Device *device, co
 	descColor.Format = dataFormatToD3D11(desc.format);
 	descColor.SampleDesc.Count = 1;
 	descColor.SampleDesc.Quality = 0;
+
+	if (desc.initDataCallback) {
+		descColor.Usage = D3D11_USAGE_STAGING;
+		descColor.BindFlags = 0;
+		descColor.MiscFlags = 0;
+		descColor.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;
+
+		HRESULT hr = device_->CreateTexture2D(&descColor, nullptr, &tex->stagingTex);
+		if (!SUCCEEDED(hr)) {
+			delete tex;
+			return nullptr;
+		}
+	}
+
 	descColor.Usage = D3D11_USAGE_DEFAULT;
 	descColor.BindFlags = generateMips ? (D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_RENDER_TARGET) : D3D11_BIND_SHADER_RESOURCE;
 	descColor.MiscFlags = generateMips ? D3D11_RESOURCE_MISC_GENERATE_MIPS : 0;
 	descColor.CPUAccessFlags = 0;
 
-	// Make sure we have a staging texture if we'll need it.
-	if (desc.initDataCallback && !CreateStagingTexture(device)) {
-		return false;
-	}
-
 	D3D11_SUBRESOURCE_DATA *initDataParam = nullptr;
 	D3D11_SUBRESOURCE_DATA initData[12]{};
 	std::vector<uint8_t> initDataBuffer[12];
@@ -935,39 +863,62 @@ bool D3D11Texture::Create(ID3D11DeviceContext *context, ID3D11Device *device, co
 		initDataParam = initData;
 	}
 
-	HRESULT hr = device->CreateTexture2D(&descColor, initDataParam, &tex_);
+	HRESULT hr = device_->CreateTexture2D(&descColor, initDataParam, &tex->tex);
 	if (!SUCCEEDED(hr)) {
-		tex_ = nullptr;
+		delete tex;
-		return false;
+		return nullptr;
 	}
-	hr = device->CreateShaderResourceView(tex_, nullptr, &view_);
+	hr = device_->CreateShaderResourceView(tex->tex, nullptr, &tex->view);
 	if (!SUCCEEDED(hr)) {
-		return false;
+		delete tex;
+		return nullptr;
 	}
 
+	auto populateLevelCallback = [&](int level, int w, int h, int d) {
+		D3D11_MAPPED_SUBRESOURCE mapped;
+		hr = context_->Map(tex->stagingTex, level, D3D11_MAP_WRITE, 0, &mapped);
+		if (!SUCCEEDED(hr)) {
+			return false;
+		}
+
+		if (!desc.initDataCallback((uint8_t *)mapped.pData, desc.initData[level], w, h, d, mapped.RowPitch, mapped.DepthPitch)) {
+			for (int s = 0; s < d; ++s) {
+				for (int y = 0; y < h; ++y) {
+					void *dest = (uint8_t *)mapped.pData + mapped.DepthPitch * s + mapped.RowPitch * y;
+					uint32_t byteStride = w * (uint32_t)DataFormatSizeInBytes(desc.format);
+					const void *src = desc.initData[level] + byteStride * (y + h * d);
+					memcpy(dest, src, byteStride);
+				}
+			}
+		}
+		context_->Unmap(tex->stagingTex, level);
+		return true;
+	};
+
 	if (generateMips && desc.initData.size() >= 1) {
 		if (desc.initDataCallback) {
-			if (!FillLevel(context, 0, desc.width, desc.height, desc.depth, desc.initData.data(), desc.initDataCallback)) {
+			if (!populateLevelCallback(0, desc.width, desc.height, desc.depth)) {
-				tex_->Release();
+				delete tex;
-				return false;
+				return nullptr;
 			}
 
-			context->CopyResource(tex_, stagingTex_);
+			context_->CopyResource(tex->stagingTex, tex->stagingTex);
-			stagingTex_->Release();
+			tex->stagingTex->Release();
-			stagingTex_ = nullptr;
+			tex->stagingTex = nullptr;
 		} else {
 			uint32_t byteStride = desc.width * (uint32_t)DataFormatSizeInBytes(desc.format);
-			context->UpdateSubresource(tex_, 0, nullptr, desc.initData[0], byteStride, 0);
+			context_->UpdateSubresource(tex->tex, 0, nullptr, desc.initData[0], byteStride, 0);
 		}
-		context->GenerateMips(view_);
+		context_->GenerateMips(tex->view);
 	} else if (desc.initDataCallback) {
 		int w = desc.width;
 		int h = desc.height;
 		int d = desc.depth;
 		for (int i = 0; i < (int)desc.initData.size(); i++) {
-			if (!FillLevel(context, i, w, h, d, desc.initData.data(), desc.initDataCallback)) {
+			if (!populateLevelCallback(i, desc.width, desc.height, desc.depth)) {
 				if (i == 0) {
-					return false;
+					delete tex;
+					return nullptr;
 				} else {
 					break;
 				}
@@ -978,62 +929,13 @@ bool D3D11Texture::Create(ID3D11DeviceContext *context, ID3D11Device *device, co
 			d = (d + 1) / 2;
 		}
 
-		context->CopyResource(tex_, stagingTex_);
+		context_->CopyResource(tex->tex, tex->stagingTex);
-		stagingTex_->Release();
+		tex->stagingTex->Release();
-		stagingTex_ = nullptr;
+		tex->stagingTex = nullptr;
 	}
-	return true;
-}
-
-void D3D11Texture::UpdateTextureLevels(ID3D11DeviceContext *context, ID3D11Device *device, Texture *texture, const uint8_t * const*data, TextureCallback initDataCallback, int numLevels) {
-	if (!CreateStagingTexture(device)) {
-		return;
-	}
-
-	int w = width_;
-	int h = height_;
-	int d = depth_;
-	for (int i = 0; i < (int)numLevels; i++) {
-		if (!FillLevel(context, i, w, h, d, data, initDataCallback)) {
-			break;
-		}
-
-		w = (w + 1) / 2;
-		h = (h + 1) / 2;
-		d = (d + 1) / 2;
-	}
-
-	context->CopyResource(tex_, stagingTex_);
-	stagingTex_->Release();
-	stagingTex_ = nullptr;
-}
-
-Texture *D3D11DrawContext::CreateTexture(const TextureDesc &desc) {
-	if (!(GetDataFormatSupport(desc.format) & FMT_TEXTURE)) {
-		// D3D11 does not support this format as a texture format.
-		return nullptr;
-	}
-
-	D3D11Texture *tex = new D3D11Texture(desc);
-	bool generateMips = desc.generateMips;
-	if (desc.generateMips && !(GetDataFormatSupport(desc.format) & FMT_AUTOGEN_MIPS)) {
-		// D3D11 does not support autogenerating mipmaps for this format.
-		generateMips = false;
-	}
-	if (!tex->Create(context_, device_, desc, generateMips)) {
-		tex->Release();
-		return nullptr;
-	}
-
 	return tex;
 }
 
-
-void D3D11DrawContext::UpdateTextureLevels(Texture *texture, const uint8_t **data, TextureCallback initDataCallback, int numLevels) {
-	D3D11Texture *tex = (D3D11Texture *)texture;
-	tex->UpdateTextureLevels(context_, device_, texture, data, initDataCallback, numLevels);
-}
-
 ShaderModule *D3D11DrawContext::CreateShaderModule(ShaderStage stage, ShaderLanguage language, const uint8_t *data, size_t dataSize, const char *tag) {
 	if (language != ShaderLanguage::HLSL_D3D11) {
 		ERROR_LOG(G3D, "Unsupported shader language");
@@ -1251,7 +1153,8 @@ void D3D11DrawContext::ApplyCurrentState() {
 	}
 
 	if (curPipeline_->input != nullptr) {
-		context_->IASetVertexBuffers(0, 1, &nextVertexBuffer_, &curPipeline_->input->stride, &nextVertexBufferOffset_);
+		int numVBs = (int)curPipeline_->input->strides.size();
+		context_->IASetVertexBuffers(0, numVBs, nextVertexBuffers_, (UINT *)curPipeline_->input->strides.data(), (UINT *)nextVertexBufferOffsets_);
 	}
 	if (dirtyIndexBuffer_) {
 		context_->IASetIndexBuffer(nextIndexBuffer_, DXGI_FORMAT_R16_UINT, nextIndexBufferOffset_);
@@ -1320,11 +1223,14 @@ void D3D11DrawContext::UpdateBuffer(Buffer *buffer, const uint8_t *data, size_t
 	context_->UpdateSubresource(buf->buf, 0, &box, data, 0, 0);
 }
 
-void D3D11DrawContext::BindVertexBuffer(Buffer *buffer, int offset) {
+void D3D11DrawContext::BindVertexBuffers(int start, int count, Buffer **buffers, const int *offsets) {
+	_assert_(start + count <= ARRAY_SIZE(nextVertexBuffers_));
 	// Lazy application
-	D3D11Buffer *buf = (D3D11Buffer *)buffer;
+	for (int i = 0; i < count; i++) {
-	nextVertexBuffer_ = buf->buf;
+		D3D11Buffer *buf = (D3D11Buffer *)buffers[i];
-	nextVertexBufferOffset_ = offset;
+		nextVertexBuffers_[start + i] = buf->buf;
+		nextVertexBufferOffsets_[start + i] = offsets ? offsets[i] : 0;
+	}
 }
 
 void D3D11DrawContext::BindIndexBuffer(Buffer *indexBuffer, int offset) {
@@ -1348,10 +1254,10 @@ void D3D11DrawContext::DrawIndexed(int indexCount, int offset) {
 void D3D11DrawContext::DrawUP(const void *vdata, int vertexCount) {
 	ApplyCurrentState();
 
-	int byteSize = vertexCount * curPipeline_->input->stride;
+	int byteSize = vertexCount * curPipeline_->input->strides[0];
 
 	UpdateBuffer(upBuffer_, (const uint8_t *)vdata, 0, byteSize, Draw::UPDATE_DISCARD);
-	BindVertexBuffer(upBuffer_, 0);
+	BindVertexBuffers(0, 1, &upBuffer_, nullptr);
 	int offset = 0;
 	Draw(vertexCount, offset);
 }
@@ -1419,7 +1325,17 @@ Framebuffer *D3D11DrawContext::CreateFramebuffer(const FramebufferDesc &desc) {
 	// Texture arrays are supported but we don't have any other use cases yet.
 	_dbg_assert_(desc.numLayers == 1);
 
-	fb->colorFormat = DXGI_FORMAT_R8G8B8A8_UNORM;
+	DXGI_FORMAT colorFormat;
+	switch (desc.colorFormat) {
+	case DataFormat::R8G8B8A8_UNORM:
+		colorFormat = DXGI_FORMAT_R8G8B8A8_UNORM;
+		break;
+	default:
+		_assert_msg_(false, "Framebuffer format not supported");
+		return nullptr;
+	}
+
+	fb->colorFormat = colorFormat;
 	D3D11_TEXTURE2D_DESC descColor{};
 	descColor.Width = desc.width;
 	descColor.Height = desc.height;
@@ -1498,7 +1414,7 @@ void D3D11DrawContext::BindTextures(int start, int count, Texture **textures, Te
 	_assert_(start + count <= ARRAY_SIZE(views));
 	for (int i = 0; i < count; i++) {
 		D3D11Texture *tex = (D3D11Texture *)textures[i];
-		views[i] = tex ? tex->View() : nullptr;
+		views[i] = tex ? tex->view : nullptr;
 	}
 	context_->PSSetShaderResources(start, count, views);
 }
@@ -1535,11 +1451,7 @@ void D3D11DrawContext::Clear(int mask, uint32_t colorval, float depthVal, int st
 	}
 }
 
-void D3D11DrawContext::BeginFrame(DebugFlags debugFlags) {
+void D3D11DrawContext::BeginFrame() {
-	FrameTimeData &frameTimeData = frameTimeHistory_.Add(frameCount_);
-	frameTimeData.afterFenceWait = time_now_d();
-	frameTimeData.frameBegin = frameTimeData.afterFenceWait;
-
 	context_->OMSetRenderTargets(1, &curRenderTargetView_, curDepthStencilView_);
 
 	if (curBlend_ != nullptr) {
@@ -1559,7 +1471,7 @@ void D3D11DrawContext::BeginFrame(DebugFlags debugFlags) {
 		context_->IASetPrimitiveTopology(curTopology_);
 	}
 	if (curPipeline_ != nullptr) {
-		context_->IASetVertexBuffers(0, 1, &nextVertexBuffer_, &curPipeline_->input->stride, &nextVertexBufferOffset_);
+		context_->IASetVertexBuffers(0, 1, nextVertexBuffers_, (UINT *)curPipeline_->input->strides.data(), (UINT *)nextVertexBufferOffsets_);
 		context_->IASetIndexBuffer(nextIndexBuffer_, DXGI_FORMAT_R16_UINT, nextIndexBufferOffset_);
 		if (curPipeline_->dynamicUniforms) {
 			context_->VSSetConstantBuffers(0, 1, &curPipeline_->dynamicUniforms);
@@ -1623,7 +1535,7 @@ bool D3D11DrawContext::BlitFramebuffer(Framebuffer *srcfb, int srcX1, int srcY1,
 	return false;
 }
 
-bool D3D11DrawContext::CopyFramebufferToMemory(Framebuffer *src, int channelBits, int bx, int by, int bw, int bh, Draw::DataFormat destFormat, void *pixels, int pixelStride, ReadbackMode mode, const char *tag) {
+bool D3D11DrawContext::CopyFramebufferToMemorySync(Framebuffer *src, int channelBits, int bx, int by, int bw, int bh, Draw::DataFormat destFormat, void *pixels, int pixelStride, const char *tag) {
 	D3D11Framebuffer *fb = (D3D11Framebuffer *)src;
 
 	if (fb) {
@@ -1862,7 +1774,7 @@ uint64_t D3D11DrawContext::GetNativeObject(NativeObject obj, void *srcObject) {
 	case NativeObject::FEATURE_LEVEL:
 		return (uint64_t)(uintptr_t)featureLevel_;
 	case NativeObject::TEXTURE_VIEW:
-		return (uint64_t)(((D3D11Texture *)srcObject)->View());
+		return (uint64_t)(((D3D11Texture *)srcObject)->view);
 	default:
 		return 0;
 	}
@@ -1879,8 +1791,8 @@ void D3D11DrawContext::GetFramebufferDimensions(Framebuffer *fbo, int *w, int *h
 	}
 }
 
-DrawContext *T3DCreateD3D11Context(ID3D11Device *device, ID3D11DeviceContext *context, ID3D11Device1 *device1, ID3D11DeviceContext1 *context1, IDXGISwapChain *swapChain, D3D_FEATURE_LEVEL featureLevel, HWND hWnd, std::vector<std::string> adapterNames, int maxInflightFrames) {
+DrawContext *T3DCreateD3D11Context(ID3D11Device *device, ID3D11DeviceContext *context, ID3D11Device1 *device1, ID3D11DeviceContext1 *context1, D3D_FEATURE_LEVEL featureLevel, HWND hWnd, std::vector<std::string> adapterNames) {
-	return new D3D11DrawContext(device, context, device1, context1, swapChain, featureLevel, hWnd, adapterNames, maxInflightFrames);
+	return new D3D11DrawContext(device, context, device1, context1, featureLevel, hWnd, adapterNames);
 }
 
 }  // namespace Draw

Common/GPU/D3D9/thin3d_d3d9.cpp

@@ -25,7 +25,6 @@
 #include "Common/GPU/D3D9/D3D9StateCache.h"
 #include "Common/OSVersion.h"
 #include "Common/StringUtils.h"
-#include "Common/TimeUtil.h"
 
 #include "Common/Log.h"
 
@@ -114,7 +113,7 @@ static const D3DSTENCILOP stencilOpToD3D9[] = {
 	D3DSTENCILOP_DECR,
 };
 
-static D3DFORMAT FormatToD3DFMT(DataFormat fmt) {
+D3DFORMAT FormatToD3DFMT(DataFormat fmt) {
 	switch (fmt) {
 	case DataFormat::R16_UNORM: return D3DFMT_L16;  // closest match, should be a fine substitution if we ignore channels except R.
 	case DataFormat::R8G8B8A8_UNORM: return D3DFMT_A8R8G8B8;
@@ -126,9 +125,6 @@ static D3DFORMAT FormatToD3DFMT(DataFormat fmt) {
 	case DataFormat::A1R5G5B5_UNORM_PACK16: return D3DFMT_A1R5G5B5;
 	case DataFormat::D24_S8: return D3DFMT_D24S8;
 	case DataFormat::D16: return D3DFMT_D16;
-	case DataFormat::BC1_RGBA_UNORM_BLOCK: return D3DFMT_DXT1;
-	case DataFormat::BC2_UNORM_BLOCK: return D3DFMT_DXT3;  // DXT3 is indeed BC2.
-	case DataFormat::BC3_UNORM_BLOCK: return D3DFMT_DXT5;  // DXT5 is indeed BC3
 	default: return D3DFMT_UNKNOWN;
 	}
 }
@@ -231,14 +227,14 @@ public:
 			decl_->Release();
 		}
 	}
-	int GetStride() const { return stride_; }
+	int GetStride(int binding) const { return stride_[binding]; }
 	void Apply(LPDIRECT3DDEVICE9 device) {
 		device->SetVertexDeclaration(decl_);
 	}
 
 private:
 	LPDIRECT3DVERTEXDECLARATION9 decl_;
-	int stride_;
+	int stride_[4];
 };
 
 class D3D9ShaderModule : public ShaderModule {
@@ -309,14 +305,14 @@ public:
 			return nullptr;
 		}
 	}
-	void UpdateTextureLevels(const uint8_t * const *data, int numLevels, TextureCallback initDataCallback);
 
 private:
-	void SetImageData(int x, int y, int z, int width, int height, int depth, int level, int stride, const uint8_t *data, TextureCallback initDataCallback);
+	void SetImageData(int x, int y, int z, int width, int height, int depth, int level, int stride, const uint8_t *data, TextureCallback callback);
 	bool Create(const TextureDesc &desc);
 	LPDIRECT3DDEVICE9 device_;
 	LPDIRECT3DDEVICE9EX deviceEx_;
 	TextureType type_;
+	DataFormat format_;
 	D3DFORMAT d3dfmt_;
 	LPDIRECT3DTEXTURE9 tex_ = nullptr;
 	LPDIRECT3DVOLUMETEXTURE9 volTex_ = nullptr;
@@ -375,31 +371,27 @@ bool D3D9Texture::Create(const TextureDesc &desc) {
 		break;
 	}
 	if (FAILED(hr)) {
-		ERROR_LOG(G3D, "D3D9 Texture creation failed");
+		ERROR_LOG(G3D,  "Texture creation failed");
 		return false;
 	}
 
 	if (desc.initData.size()) {
 		// In D3D9, after setting D3DUSAGE_AUTOGENMIPS, we can only access the top layer. The rest will be
 		// automatically generated.
-		int numLevels = desc.generateMips ? 1 : (int)desc.initData.size();
+		int maxLevel = desc.generateMips ? 1 : (int)desc.initData.size();
-		UpdateTextureLevels(desc.initData.data(), numLevels, desc.initDataCallback);
+		int w = desc.width;
+		int h = desc.height;
+		int d = desc.depth;
+		for (int i = 0; i < maxLevel; i++) {
+			SetImageData(0, 0, 0, w, h, d, i, 0, desc.initData[i], desc.initDataCallback);
+			w = (w + 1) / 2;
+			h = (h + 1) / 2;
+			d = (d + 1) / 2;
+		}
 	}
 	return true;
 }
 
-void D3D9Texture::UpdateTextureLevels(const uint8_t * const *data, int numLevels, TextureCallback initDataCallback) {
-	int w = width_;
-	int h = height_;
-	int d = depth_;
-	for (int i = 0; i < numLevels; i++) {
-		SetImageData(0, 0, 0, w, h, d, i, 0, data[i], initDataCallback);
-		w = (w + 1) / 2;
-		h = (h + 1) / 2;
-		d = (d + 1) / 2;
-	}
-}
-
 // Just switches R and G.
 inline uint32_t Shuffle8888(uint32_t x) {
 	return (x & 0xFF00FF00) | ((x >> 16) & 0xFF) | ((x << 16) & 0xFF0000);
@@ -533,13 +525,12 @@ public:
 	Framebuffer *CreateFramebuffer(const FramebufferDesc &desc) override;
 
 	void UpdateBuffer(Buffer *buffer, const uint8_t *data, size_t offset, size_t size, UpdateBufferFlags flags) override;
-	void UpdateTextureLevels(Texture *texture, const uint8_t **data, TextureCallback initDataCallback, int numLevels) override;
 
 	void CopyFramebufferImage(Framebuffer *src, int level, int x, int y, int z, Framebuffer *dst, int dstLevel, int dstX, int dstY, int dstZ, int width, int height, int depth, int channelBits, const char *tag) override {
 		// Not implemented
 	}
 	bool BlitFramebuffer(Framebuffer *src, int srcX1, int srcY1, int srcX2, int srcY2, Framebuffer *dst, int dstX1, int dstY1, int dstX2, int dstY2, int channelBits, FBBlitFilter filter, const char *tag) override;
-	bool CopyFramebufferToMemory(Framebuffer *src, int channelBits, int x, int y, int w, int h, Draw::DataFormat format, void *pixels, int pixelStride, ReadbackMode mode, const char *tag) override;
+	bool CopyFramebufferToMemorySync(Framebuffer *src, int channelBits, int x, int y, int w, int h, Draw::DataFormat format, void *pixels, int pixelStride, const char *tag) override;
 
 	// These functions should be self explanatory.
 	void BindFramebufferAsRenderTarget(Framebuffer *fbo, const RenderPassInfo &rp, const char *tag) override;
@@ -560,9 +551,12 @@ public:
 				s->Apply(device_, start + i);
 		}
 	}
-	void BindVertexBuffer(Buffer *vertexBuffer, int offset) override {
+	void BindVertexBuffers(int start, int count, Buffer **buffers, const int *offsets) override {
-		curVBuffer_ = (D3D9Buffer *)vertexBuffer;
+		_assert_(start + count <= ARRAY_SIZE(curVBuffers_));
-		curVBufferOffset_ = offset;
+		for (int i = 0; i < count; i++) {
+			curVBuffers_[i + start] = (D3D9Buffer *)buffers[i];
+			curVBufferOffsets_[i + start] = offsets ? offsets[i] : 0;
+		}
 	}
 	void BindIndexBuffer(Buffer *indexBuffer, int offset) override {
 		curIBuffer_ = (D3D9Buffer *)indexBuffer;
@@ -573,17 +567,13 @@ public:
 		curPipeline_ = (D3D9Pipeline *)pipeline;
 	}
 
-	void BeginFrame(Draw::DebugFlags debugFlags) override;
 	void EndFrame() override;
-	void Present(PresentMode presentMode, int vblanks) override;
-
-	int GetFrameCount() override { return frameCount_; }
 
 	void UpdateDynamicUniformBuffer(const void *ub, size_t size) override;
 
 	// Raster state
 	void SetScissorRect(int left, int top, int width, int height) override;
-	void SetViewport(const Viewport &viewport) override;
+	void SetViewports(int count, Viewport *viewports) override;
 	void SetBlendFactor(float color[4]) override;
 	void SetStencilParams(uint8_t refValue, uint8_t writeMask, uint8_t compareMask) override;
 
@@ -638,12 +628,11 @@ private:
 	D3DCAPS9 d3dCaps_;
 	char shadeLangVersion_[64]{};
 	DeviceCaps caps_{};
-	int frameCount_ = FRAME_TIME_HISTORY_LENGTH;
 
 	// Bound state
 	AutoRef<D3D9Pipeline> curPipeline_;
-	AutoRef<D3D9Buffer> curVBuffer_;
+	AutoRef<D3D9Buffer> curVBuffers_[4];
-	int curVBufferOffset_ = 0;
+	int curVBufferOffsets_[4]{};
 	AutoRef<D3D9Buffer> curIBuffer_;
 	int curIBufferOffset_ = 0;
 	AutoRef<Framebuffer> curRenderTarget_;
@@ -755,10 +744,10 @@ D3D9Context::D3D9Context(IDirect3D9 *d3d, IDirect3D9Ex *d3dEx, int adapterId, ID
 	}
 
 	if (SUCCEEDED(result)) {
-		snprintf(shadeLangVersion_, sizeof(shadeLangVersion_), "PS: %04x VS: %04x", d3dCaps_.PixelShaderVersion & 0xFFFF, d3dCaps_.VertexShaderVersion & 0xFFFF);
+		sprintf(shadeLangVersion_, "PS: %04x VS: %04x", d3dCaps_.PixelShaderVersion & 0xFFFF, d3dCaps_.VertexShaderVersion & 0xFFFF);
 	} else {
 		WARN_LOG(G3D, "Direct3D9: Failed to get the device caps!");
-		truncate_cpy(shadeLangVersion_, "N/A");
+		strcpy(shadeLangVersion_, "N/A");
 	}
 
 	caps_.deviceID = identifier_.DeviceId;
@@ -780,9 +769,6 @@ D3D9Context::D3D9Context(IDirect3D9 *d3d, IDirect3D9Ex *d3dEx, int adapterId, ID
 	caps_.multiSampleLevelsMask = 1;  // More could be supported with some work.
 
 	caps_.clipPlanesSupported = caps.MaxUserClipPlanes;
-	caps_.presentInstantModeChange = false;
-	caps_.presentMaxInterval = 1;
-	caps_.presentModesSupported = PresentMode::FIFO;
 
 	if ((caps.RasterCaps & D3DPRASTERCAPS_ANISOTROPY) != 0 && caps.MaxAnisotropy > 1) {
 		caps_.anisoSupported = true;
@@ -941,12 +927,6 @@ Texture *D3D9Context::CreateTexture(const TextureDesc &desc) {
 	return tex;
 }
 
-void D3D9Context::UpdateTextureLevels(Texture *texture, const uint8_t **data, TextureCallback initDataCallback, int numLevels) {
-	D3D9Texture *tex = (D3D9Texture *)texture;
-	tex->UpdateTextureLevels(data, numLevels, initDataCallback);
-}
-
-
 void D3D9Context::BindTextures(int start, int count, Texture **textures, TextureBindFlags flags) {
 	_assert_(start + count <= MAX_BOUND_TEXTURES);
 	for (int i = start; i < start + count; i++) {
@@ -964,31 +944,10 @@ void D3D9Context::BindNativeTexture(int index, void *nativeTexture) {
 	device_->SetTexture(index, texture);
 }
 
-void D3D9Context::BeginFrame(Draw::DebugFlags debugFlags) {
-	FrameTimeData frameTimeData = frameTimeHistory_.Add(frameCount_);
-	frameTimeData.frameBegin = time_now_d();
-	frameTimeData.afterFenceWait = frameTimeData.frameBegin;  // no fence wait
-}
-
 void D3D9Context::EndFrame() {
-	frameTimeHistory_[frameCount_].firstSubmit = time_now_d();
 	curPipeline_ = nullptr;
 }
 
-void D3D9Context::Present(PresentMode presentMode, int vblanks) {
-	frameTimeHistory_[frameCount_].queuePresent = time_now_d();
-	if (deviceEx_) {
-		deviceEx_->EndScene();
-		deviceEx_->PresentEx(NULL, NULL, NULL, NULL, 0);
-		deviceEx_->BeginScene();
-	} else {
-		device_->EndScene();
-		device_->Present(NULL, NULL, NULL, NULL);
-		device_->BeginScene();
-	}
-	frameCount_++;
-}
-
 static void SemanticToD3D9UsageAndIndex(int semantic, BYTE *usage, BYTE *index) {
 	*index = 0;
 	switch (semantic) {
@@ -1025,7 +984,7 @@ D3D9InputLayout::D3D9InputLayout(LPDIRECT3DDEVICE9 device, const InputLayoutDesc
 	D3DVERTEXELEMENT9 *elements = new D3DVERTEXELEMENT9[desc.attributes.size() + 1];
 	size_t i;
 	for (i = 0; i < desc.attributes.size(); i++) {
-		elements[i].Stream = 0;
+		elements[i].Stream = desc.attributes[i].binding;
 		elements[i].Offset = desc.attributes[i].offset;
 		elements[i].Method = D3DDECLMETHOD_DEFAULT;
 		SemanticToD3D9UsageAndIndex(desc.attributes[i].location, &elements[i].Usage, &elements[i].UsageIndex);
@@ -1035,7 +994,9 @@ D3D9InputLayout::D3D9InputLayout(LPDIRECT3DDEVICE9 device, const InputLayoutDesc
 	// Zero the last one.
 	memcpy(&elements[i], &end, sizeof(elements[i]));
 
-	stride_ = desc.stride;
+	for (i = 0; i < desc.bindings.size(); i++) {
+		stride_[i] = desc.bindings[i].stride;
+	}
 
 	HRESULT hr = device->CreateVertexDeclaration(elements, &decl_);
 	if (FAILED(hr)) {
@@ -1169,7 +1130,7 @@ inline int D3DPrimCount(D3DPRIMITIVETYPE prim, int size) {
 }
 
 void D3D9Context::Draw(int vertexCount, int offset) {
-	device_->SetStreamSource(0, curVBuffer_->vbuffer_, curVBufferOffset_, curPipeline_->inputLayout->GetStride());
+	device_->SetStreamSource(0, curVBuffers_[0]->vbuffer_, curVBufferOffsets_[0], curPipeline_->inputLayout->GetStride(0));
 	curPipeline_->inputLayout->Apply(device_);
 	curPipeline_->Apply(device_, stencilRef_, stencilWriteMask_, stencilCompareMask_);
 	ApplyDynamicState();
@@ -1180,7 +1141,7 @@ void D3D9Context::DrawIndexed(int vertexCount, int offset) {
 	curPipeline_->inputLayout->Apply(device_);
 	curPipeline_->Apply(device_, stencilRef_, stencilWriteMask_, stencilCompareMask_);
 	ApplyDynamicState();
-	device_->SetStreamSource(0, curVBuffer_->vbuffer_, curVBufferOffset_, curPipeline_->inputLayout->GetStride());
+	device_->SetStreamSource(0, curVBuffers_[0]->vbuffer_, curVBufferOffsets_[0], curPipeline_->inputLayout->GetStride(0));
 	device_->SetIndices(curIBuffer_->ibuffer_);
 	device_->DrawIndexedPrimitive(curPipeline_->prim, 0, 0, vertexCount, offset, D3DPrimCount(curPipeline_->prim, vertexCount));
 }
@@ -1190,7 +1151,7 @@ void D3D9Context::DrawUP(const void *vdata, int vertexCount) {
 	curPipeline_->Apply(device_, stencilRef_, stencilWriteMask_, stencilCompareMask_);
 	ApplyDynamicState();
 
-	device_->DrawPrimitiveUP(curPipeline_->prim, D3DPrimCount(curPipeline_->prim, vertexCount), vdata, curPipeline_->inputLayout->GetStride());
+	device_->DrawPrimitiveUP(curPipeline_->prim, D3DPrimCount(curPipeline_->prim, vertexCount), vdata, curPipeline_->inputLayout->GetStride(0));
 }
 
 static uint32_t SwapRB(uint32_t c) {
@@ -1212,12 +1173,12 @@ void D3D9Context::SetScissorRect(int left, int top, int width, int height) {
 	dxstate.scissorTest.set(true);
 }
 
-void D3D9Context::SetViewport(const Viewport &viewport) {
+void D3D9Context::SetViewports(int count, Viewport *viewports) {
-	int x = (int)viewport.TopLeftX;
+	int x = (int)viewports[0].TopLeftX;
-	int y = (int)viewport.TopLeftY;
+	int y = (int)viewports[0].TopLeftY;
-	int w = (int)viewport.Width;
+	int w = (int)viewports[0].Width;
-	int h = (int)viewport.Height;
+	int h = (int)viewports[0].Height;
-	dxstate.viewport.set(x, y, w, h, viewport.MinDepth, viewport.MaxDepth);
+	dxstate.viewport.set(x, y, w, h, viewports[0].MinDepth, viewports[0].MaxDepth);
 }
 
 void D3D9Context::SetBlendFactor(float color[4]) {
@@ -1303,7 +1264,18 @@ Framebuffer *D3D9Context::CreateFramebuffer(const FramebufferDesc &desc) {
 	D3D9Framebuffer *fbo = new D3D9Framebuffer(desc.width, desc.height);
 	fbo->depthstenciltex = nullptr;
 
-	HRESULT rtResult = device_->CreateTexture(desc.width, desc.height, 1, D3DUSAGE_RENDERTARGET, D3DFMT_A8R8G8B8, D3DPOOL_DEFAULT, &fbo->tex, nullptr);
+	D3DFORMAT colorFormat;
+	switch (desc.colorFormat) {
+	case DataFormat::R8G8B8A8_UNORM:
+		// We pretend to support this, although in reality we use the reverse format.
+		colorFormat = D3DFMT_A8R8G8B8;
+		break;
+	default:
+		_assert_msg_(false, "Framebuffer format not supported");
+		return nullptr;
+	}
+
+	HRESULT rtResult = device_->CreateTexture(desc.width, desc.height, 1, D3DUSAGE_RENDERTARGET, colorFormat, D3DPOOL_DEFAULT, &fbo->tex, nullptr);
 	if (FAILED(rtResult)) {
 		ERROR_LOG(G3D,  "Failed to create render target");
 		fbo->Release();
@@ -1465,7 +1437,7 @@ bool D3D9Context::BlitFramebuffer(Framebuffer *srcfb, int srcX1, int srcY1, int
 	return SUCCEEDED(device_->StretchRect(srcSurf, &srcRect, dstSurf, &dstRect, (filter == FB_BLIT_LINEAR && channelBits == FB_COLOR_BIT) ? D3DTEXF_LINEAR : D3DTEXF_POINT));
 }
 
-bool D3D9Context::CopyFramebufferToMemory(Framebuffer *src, int channelBits, int bx, int by, int bw, int bh, Draw::DataFormat destFormat, void *pixels, int pixelStride, ReadbackMode mode, const char *tag) {
+bool D3D9Context::CopyFramebufferToMemorySync(Framebuffer *src, int channelBits, int bx, int by, int bw, int bh, Draw::DataFormat destFormat, void *pixels, int pixelStride, const char *tag) {
 	D3D9Framebuffer *fb = (D3D9Framebuffer *)src;
 
 	if (fb) {
@@ -1628,7 +1600,6 @@ uint32_t D3D9Context::GetDataFormatSupport(DataFormat fmt) const {
 	case DataFormat::BC1_RGBA_UNORM_BLOCK:
 	case DataFormat::BC2_UNORM_BLOCK:
 	case DataFormat::BC3_UNORM_BLOCK:
-		// DXT1, DXT3, DXT5.
 		return FMT_TEXTURE;
 	default:
 		return 0;

Common/GPU/DataFormat.h

@@ -46,20 +46,22 @@ enum class DataFormat : uint8_t {
 	// Block compression formats.
 	// These are modern names for DXT and friends, now patent free.
 	// https://msdn.microsoft.com/en-us/library/bb694531.aspx
-	BC1_RGBA_UNORM_BLOCK,  // 64 bits per 4x4 block. Used by Basis, along with ETC2_R8G8B8_UNORM_BLOCK.
+	BC1_RGBA_UNORM_BLOCK,
-	BC2_UNORM_BLOCK,  // 4-bit straight alpha + DXT1 color. 128 bits per block. Usually not worth using
+	BC1_RGBA_SRGB_BLOCK,
-	BC3_UNORM_BLOCK,  // 3-bit alpha with 2 ref values (+ magic) + DXT1 color. 128 bits per block.
+	BC2_UNORM_BLOCK,  // 4-bit straight alpha + DXT1 color. Usually not worth using
-	BC4_UNORM_BLOCK,  // 1-channel, same storage as BC3 alpha. 64 bits per block.
+	BC2_SRGB_BLOCK,
-	BC5_UNORM_BLOCK,  // 2-channel RG, each has same storage as BC3 alpha. 128 bits per block.
+	BC3_UNORM_BLOCK,  // 3-bit alpha with 2 ref values (+ magic) + DXT1 color
-	BC7_UNORM_BLOCK,  // Highly advanced RGBA, very expensive to compress, very good quality. 128 bits per block.
+	BC3_SRGB_BLOCK,
+	BC4_UNORM_BLOCK,  // 1-channel, same storage as BC3 alpha
+	BC4_SNORM_BLOCK,
+	BC5_UNORM_BLOCK,  // 2-channel RG, each has same storage as BC3 alpha
+	BC5_SNORM_BLOCK,
+	BC6H_UFLOAT_BLOCK,  // TODO
+	BC6H_SFLOAT_BLOCK,
+	BC7_UNORM_BLOCK,    // Highly advanced, very expensive to compress, very good quality.
+	BC7_SRGB_BLOCK,
 
-	// Ericsson texture compression.
+	ETC1,
-	ETC2_R8G8B8_UNORM_BLOCK,  // Color-only, 64 bits per 4x4 block.
-	ETC2_R8G8B8A1_UNORM_BLOCK,  // Color + alpha, 128 bits per 4x4 block.
-	ETC2_R8G8B8A8_UNORM_BLOCK,  // Color + alpha, 128 bits per 4x4 block.
-
-	// This is the one ASTC format used by UASTC / basis Universal.
-	ASTC_4x4_UNORM_BLOCK,
 
 	S8,
 	D16,
@@ -74,7 +76,6 @@ bool DataFormatIsDepthStencil(DataFormat fmt);
 inline bool DataFormatIsColor(DataFormat fmt) {
 	return !DataFormatIsDepthStencil(fmt);
 }
-bool DataFormatIsBlockCompressed(DataFormat fmt, int *blockSize);
 
 // Limited format support for now.
 const char *DataFormatToString(DataFormat fmt);

Common/GPU/GPUBackendCommon.cpp

@@ -1,28 +0,0 @@
-#include <mutex>
-#include <set>
-
-#include "Common/GPU/GPUBackendCommon.h"
-
-// Global push buffer tracker for GPU memory profiling.
-// Don't want to manually dig up all the active push buffers.
-static std::mutex g_pushBufferListMutex;
-static std::set<GPUMemoryManager *> g_pushBuffers;
-
-std::vector<GPUMemoryManager *> GetActiveGPUMemoryManagers() {
-	std::vector<GPUMemoryManager *> buffers;
-	std::lock_guard<std::mutex> guard(g_pushBufferListMutex);
-	for (auto iter : g_pushBuffers) {
-		buffers.push_back(iter);
-	}
-	return buffers;
-}
-
-void RegisterGPUMemoryManager(GPUMemoryManager *manager) {
-	std::lock_guard<std::mutex> guard(g_pushBufferListMutex);
-	g_pushBuffers.insert(manager);
-}
-
-void UnregisterGPUMemoryManager(GPUMemoryManager *manager) {
-	std::lock_guard<std::mutex> guard(g_pushBufferListMutex);
-	g_pushBuffers.erase(manager);
-}

Common/GPU/GPUBackendCommon.h

@@ -1,17 +0,0 @@
-#pragma once
-
-#include <vector>
-
-// Just an abstract thing to get debug information.
-class GPUMemoryManager {
-public:
-	virtual ~GPUMemoryManager() {}
-
-	virtual void GetDebugString(char *buffer, size_t bufSize) const = 0;
-	virtual const char *Name() const = 0;  // for sorting
-};
-
-std::vector<GPUMemoryManager *> GetActiveGPUMemoryManagers();
-
-void RegisterGPUMemoryManager(GPUMemoryManager *manager);
-void UnregisterGPUMemoryManager(GPUMemoryManager *manager);

Common/GPU/MiscTypes.h

@@ -2,8 +2,6 @@
 
 #include "Common/Common.h"
 
-// Flags and structs shared between backends that haven't found a good home.
-
 enum class InvalidationFlags {
 	CACHED_RENDER_STATE = 1,
 };
@@ -16,22 +14,3 @@ enum class InvalidationCallbackFlags {
 ENUM_CLASS_BITOPS(InvalidationCallbackFlags);
 
 typedef std::function<void(InvalidationCallbackFlags)> InvalidationCallback;
-
-// These are separate from FrameData because we store some history of these.
-// Also, this might be joined with more non-GPU timing information later.
-struct FrameTimeData {
-	uint64_t frameId;
-
-	int waitCount;
-
-	double frameBegin;
-	double afterFenceWait;
-	double firstSubmit;
-	double queuePresent;
-
-	double actualPresent;
-	double desiredPresentTime;
-	double earliestPresentTime;
-	double presentMargin;
-};
-constexpr size_t FRAME_TIME_HISTORY_LENGTH = 32;

Common/GPU/OpenGL/DataFormatGL.cpp

@@ -55,7 +55,7 @@ bool Thin3DFormatToGLFormatAndType(DataFormat fmt, GLuint &internalFormat, GLuin
 		internalFormat = GL_RGB;
 		format = GL_RGB;
 		type = GL_UNSIGNED_BYTE;
-		alignment = 3;
+		alignment = 1;
 		break;
 
 	case DataFormat::R4G4B4A4_UNORM_PACK16:
@@ -86,75 +86,6 @@ bool Thin3DFormatToGLFormatAndType(DataFormat fmt, GLuint &internalFormat, GLuin
 		alignment = 16;
 		break;
 
-#ifndef USING_GLES2
-	case DataFormat::BC1_RGBA_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RGB_S3TC_DXT1_EXT;
-		format = GL_RGB;
-		type = GL_FLOAT;
-		alignment = 8;
-		break;
-	case DataFormat::BC2_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RGBA_S3TC_DXT3_EXT;
-		format = GL_RGBA;
-		type = GL_FLOAT;
-		alignment = 16;
-		break;
-	case DataFormat::BC3_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RGBA_S3TC_DXT5_EXT;
-		format = GL_RGBA;
-		type = GL_FLOAT;
-		alignment = 16;
-		break;
-	case DataFormat::BC4_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RED_RGTC1;
-		format = GL_R;
-		type = GL_FLOAT;
-		alignment = 16;
-		break;
-	case DataFormat::BC5_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RG_RGTC2;
-		format = GL_RG;
-		type = GL_FLOAT;
-		alignment = 16;
-		break;
-	case DataFormat::BC7_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RGBA_BPTC_UNORM;
-		format = GL_RGBA;
-		type = GL_FLOAT;
-		alignment = 16;
-		break;
-#endif
-
-	case DataFormat::ETC2_R8G8B8_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RGB8_ETC2;
-		format = GL_RGB;
-		type = GL_FLOAT;
-		alignment = 8;
-		break;
-
-	case DataFormat::ETC2_R8G8B8A1_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RGB8_PUNCHTHROUGH_ALPHA1_ETC2;
-		format = GL_RGBA;
-		type = GL_FLOAT;
-		alignment = 16;
-		break;
-
-	case DataFormat::ETC2_R8G8B8A8_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RGBA8_ETC2_EAC;
-		format = GL_RGBA;
-		type = GL_FLOAT;
-		alignment = 16;
-		break;
-
-#ifdef GL_COMPRESSED_RGBA_ASTC_4x4_KHR
-	case DataFormat::ASTC_4x4_UNORM_BLOCK:
-		internalFormat = GL_COMPRESSED_RGBA_ASTC_4x4_KHR;
-		format = GL_RGBA;
-		type = GL_FLOAT;
-		alignment = 16;
-		break;
-#endif
-
 	default:
 		return false;
 	}

Common/GPU/OpenGL/GLFeatures.cpp

@@ -182,9 +182,7 @@ bool CheckGLExtensions() {
 			gl_extensions.gpuVendor = GPU_VENDOR_IMGTEC;
 		} else if (vendor == "Qualcomm") {
 			gl_extensions.gpuVendor = GPU_VENDOR_QUALCOMM;
-			if (1 != sscanf(renderer, "Adreno (TM) %d", &gl_extensions.modelNumber)) {
+			sscanf(renderer, "Adreno (TM) %d", &gl_extensions.modelNumber);
-				gl_extensions.modelNumber = 300;  // or what should we default to?
-			}
 		} else if (vendor == "Broadcom") {
 			gl_extensions.gpuVendor = GPU_VENDOR_BROADCOM;
 			// Just for reference: Galaxy Y has renderer == "VideoCore IV HW"
@@ -380,12 +378,6 @@ bool CheckGLExtensions() {
 	gl_extensions.ARB_explicit_attrib_location = g_set_gl_extensions.count("GL_ARB_explicit_attrib_location") != 0;
 	gl_extensions.ARB_texture_non_power_of_two = g_set_gl_extensions.count("GL_ARB_texture_non_power_of_two") != 0;
 	gl_extensions.ARB_shader_stencil_export = g_set_gl_extensions.count("GL_ARB_shader_stencil_export") != 0;
-	gl_extensions.ARB_texture_compression_bptc = g_set_gl_extensions.count("GL_ARB_texture_compression_bptc") != 0;
-	gl_extensions.ARB_texture_compression_rgtc = g_set_gl_extensions.count("GL_ARB_texture_compression_rgtc") != 0;
-	gl_extensions.KHR_texture_compression_astc_ldr = g_set_gl_extensions.count("GL_KHR_texture_compression_astc_ldr") != 0;
-	gl_extensions.EXT_texture_compression_s3tc = g_set_gl_extensions.count("GL_EXT_texture_compression_s3tc") != 0;
-	gl_extensions.OES_texture_compression_astc = g_set_gl_extensions.count("GL_OES_texture_compression_astc") != 0;
-
 	if (gl_extensions.IsGLES) {
 		gl_extensions.EXT_blend_func_extended = g_set_gl_extensions.count("GL_EXT_blend_func_extended") != 0;
 		gl_extensions.OES_texture_npot = g_set_gl_extensions.count("GL_OES_texture_npot") != 0;
@@ -583,40 +575,6 @@ bool CheckGLExtensions() {
 		gl_extensions.EXT_clip_cull_distance = false;
 	}
 
-	// Check the old query API. It doesn't seem to be very reliable (can miss stuff).
-	GLint numCompressedFormats = 0;
-	glGetIntegerv(GL_NUM_COMPRESSED_TEXTURE_FORMATS, &numCompressedFormats);
-	GLint *compressedFormats = new GLint[numCompressedFormats];
-	if (numCompressedFormats > 0) {
-		glGetIntegerv(GL_COMPRESSED_TEXTURE_FORMATS, compressedFormats);
-		for (int i = 0; i < numCompressedFormats; i++) {
-			switch (compressedFormats[i]) {
-			case GL_COMPRESSED_RGB8_ETC2: gl_extensions.supportsETC2 = true; break;
-#ifdef GL_COMPRESSED_RGBA_ASTC_4x4_KHR
-			case GL_COMPRESSED_RGBA_ASTC_4x4_KHR: gl_extensions.supportsASTC = true; break;
-#endif
-#ifndef USING_GLES2
-			case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: gl_extensions.supportsBC123 = true; break;
-			case GL_COMPRESSED_RGBA_BPTC_UNORM: gl_extensions.supportsBC7 = true; break;
-#endif
-			}
-		}
-	}
-
-	// Enable additional formats based on extensions.
-	if (gl_extensions.EXT_texture_compression_s3tc) gl_extensions.supportsBC123 = true;
-	if (gl_extensions.ARB_texture_compression_bptc) gl_extensions.supportsBC7 = true;
-	if (gl_extensions.ARB_texture_compression_rgtc) gl_extensions.supportsBC45 = true;
-	if (gl_extensions.KHR_texture_compression_astc_ldr) gl_extensions.supportsASTC = true;
-	if (gl_extensions.OES_texture_compression_astc) gl_extensions.supportsASTC = true;
-
-	// Now, disable known-emulated texture formats.
-	if (gl_extensions.gpuVendor == GPU_VENDOR_NVIDIA && !gl_extensions.IsGLES) {
-		gl_extensions.supportsETC2 = false;
-		gl_extensions.supportsASTC = false;
-	}
-	delete[] compressedFormats;
-
 	ProcessGPUFeatures();
 
 	int error = glGetError();
@@ -628,13 +586,11 @@ bool CheckGLExtensions() {
 }
 
 void SetGLCoreContext(bool flag) {
-	if (!extensionsDone) {
+	_assert_msg_(!extensionsDone, "SetGLCoreContext() after CheckGLExtensions()");
-		useCoreContext = flag;
+
-		// For convenience, it'll get reset later.
+	useCoreContext = flag;
-		gl_extensions.IsCoreContext = useCoreContext;
+	// For convenience, it'll get reset later.
-	} else {
+	gl_extensions.IsCoreContext = useCoreContext;
-		_assert_(flag == useCoreContext);
-	}
 }
 
 void ResetGLExtensions() {

Common/GPU/OpenGL/GLFeatures.h

@@ -5,7 +5,6 @@
 #pragma once
 
 #include <string>
-#include <cstdint>
 
 // TODO: Replace with thin3d's vendor enum.
 enum {
@@ -18,7 +17,6 @@ enum {
 	GPU_VENDOR_BROADCOM = 7,  // Raspberry PI etc
 	GPU_VENDOR_VIVANTE = 8,
 	GPU_VENDOR_APPLE = 9,
-	GPU_VENDOR_MESA = 10,
 	GPU_VENDOR_UNKNOWN = 0,
 };
 
@@ -54,7 +52,6 @@ struct GLExtensions {
 	bool OES_copy_image;
 	bool OES_texture_float;
 	bool OES_texture_3D;
-	bool OES_texture_compression_astc;
 
 	// ARB
 	bool ARB_framebuffer_object;
@@ -76,14 +73,8 @@ struct GLExtensions {
 	bool ARB_texture_non_power_of_two;
 	bool ARB_stencil_texturing;
 	bool ARB_shader_stencil_export;
-	bool ARB_texture_compression_bptc;
-	bool ARB_texture_compression_rgtc;
-
-	// KHR
-	bool KHR_texture_compression_astc_ldr;
 
 	// EXT
-	bool EXT_texture_compression_s3tc;
 	bool EXT_swap_control_tear;
 	bool EXT_discard_framebuffer;
 	bool EXT_unpack_subimage;  // always supported on desktop and ES3
@@ -124,12 +115,6 @@ struct GLExtensions {
 
 	int maxVertexTextureUnits;
 
-	bool supportsETC2;
-	bool supportsBC123;
-	bool supportsBC45;
-	bool supportsBC7;
-	bool supportsASTC;
-
 	// greater-or-equal than
 	bool VersionGEThan(int major, int minor, int sub = 0);
 	int GLSLVersion();

Common/GPU/OpenGL/GLFrameData.cpp

@@ -1,75 +0,0 @@
-#include "Common/GPU/OpenGL/GLCommon.h"
-#include "Common/GPU/OpenGL/GLFrameData.h"
-#include "Common/GPU/OpenGL/GLRenderManager.h"
-#include "Common/Log.h"
-
-void GLDeleter::Take(GLDeleter &other) {
-	_assert_msg_(IsEmpty(), "Deleter already has stuff");
-	shaders = std::move(other.shaders);
-	programs = std::move(other.programs);
-	buffers = std::move(other.buffers);
-	textures = std::move(other.textures);
-	inputLayouts = std::move(other.inputLayouts);
-	framebuffers = std::move(other.framebuffers);
-	pushBuffers = std::move(other.pushBuffers);
-	other.shaders.clear();
-	other.programs.clear();
-	other.buffers.clear();
-	other.textures.clear();
-	other.inputLayouts.clear();
-	other.framebuffers.clear();
-	other.pushBuffers.clear();
-}
-
-// Runs on the GPU thread.
-void GLDeleter::Perform(GLRenderManager *renderManager, bool skipGLCalls) {
-	for (auto pushBuffer : pushBuffers) {
-		renderManager->UnregisterPushBuffer(pushBuffer);
-		if (skipGLCalls) {
-			pushBuffer->Destroy(false);
-		}
-		delete pushBuffer;
-	}
-	pushBuffers.clear();
-	for (auto shader : shaders) {
-		if (skipGLCalls && shader)
-			shader->shader = 0;  // prevent the glDeleteShader
-		delete shader;
-	}
-	shaders.clear();
-	for (auto program : programs) {
-		if (skipGLCalls && program)
-			program->program = 0;  // prevent the glDeleteProgram
-		delete program;
-	}
-	programs.clear();
-	for (auto buffer : buffers) {
-		if (skipGLCalls && buffer)
-			buffer->buffer_ = 0;
-		delete buffer;
-	}
-	buffers.clear();
-	for (auto texture : textures) {
-		if (skipGLCalls && texture)
-			texture->texture = 0;
-		delete texture;
-	}
-	textures.clear();
-	for (auto inputLayout : inputLayouts) {
-		// No GL objects in an inputLayout yet
-		delete inputLayout;
-	}
-	inputLayouts.clear();
-	for (auto framebuffer : framebuffers) {
-		if (skipGLCalls) {
-			framebuffer->handle = 0;
-			framebuffer->color_texture.texture = 0;
-			framebuffer->z_stencil_buffer = 0;
-			framebuffer->z_stencil_texture.texture = 0;
-			framebuffer->z_buffer = 0;
-			framebuffer->stencil_buffer = 0;
-		}
-		delete framebuffer;
-	}
-	framebuffers.clear();
-}

Common/GPU/OpenGL/GLFrameData.h

@@ -1,68 +0,0 @@
-#pragma once
-
-#include <mutex>
-#include <condition_variable>
-#include <vector>
-#include <string>
-#include <set>
-
-#include "Common/GPU/OpenGL/GLCommon.h"
-
-class GLRShader;
-class GLRBuffer;
-class GLRTexture;
-class GLRInputLayout;
-class GLRFramebuffer;
-class GLPushBuffer;
-class GLRProgram;
-class GLRenderManager;
-
-class GLDeleter {
-public:
-	void Perform(GLRenderManager *renderManager, bool skipGLCalls);
-
-	bool IsEmpty() const {
-		return shaders.empty() && programs.empty() && buffers.empty() && textures.empty() && inputLayouts.empty() && framebuffers.empty() && pushBuffers.empty();
-	}
-
-	void Take(GLDeleter &other);
-
-	std::vector<GLRShader *> shaders;
-	std::vector<GLRProgram *> programs;
-	std::vector<GLRBuffer *> buffers;
-	std::vector<GLRTexture *> textures;
-	std::vector<GLRInputLayout *> inputLayouts;
-	std::vector<GLRFramebuffer *> framebuffers;
-	std::vector<GLPushBuffer *> pushBuffers;
-};
-
-struct GLQueueProfileContext {
-	bool enabled;
-	double cpuStartTime;
-	double cpuEndTime;
-	std::string passesString;
-	int commandCounts[25];  // Can't grab count from the enum as it would mean a circular include. Might clean this up later.
-};
-
-
-// Per-frame data, round-robin so we can overlap submission with execution of the previous frame.
-struct GLFrameData {
-	bool skipSwap = false;
-
-	// Frames need unique IDs to wait for present on, let's keep them here.
-	// Also used for indexing into the frame timing history buffer.
-	uint64_t frameId;
-
-	std::mutex fenceMutex;
-	std::condition_variable fenceCondVar;
-	bool readyForFence = true;
-
-	// Swapchain.
-	bool hasBegun = false;
-
-	GLDeleter deleter;
-	GLDeleter deleter_prev;
-	std::set<GLPushBuffer *> activePushBuffers;
-
-	GLQueueProfileContext profile;
-};

Common/GPU/OpenGL/GLMemory.cpp

@@ -1,288 +0,0 @@
-#include "Common/MemoryUtil.h"
-#include "Common/GPU/OpenGL/GLMemory.h"
-#include "Common/GPU/OpenGL/GLRenderManager.h"
-#include "Common/GPU/OpenGL/GLFeatures.h"
-#include "Common/Data/Text/Parsers.h"
-
-extern std::thread::id renderThreadId;
-#if MAX_LOGLEVEL >= DEBUG_LEVEL
-static bool OnRenderThread() {
-	return std::this_thread::get_id() == renderThreadId;
-}
-#endif
-
-void *GLRBuffer::Map(GLBufferStrategy strategy) {
-	_assert_(buffer_ != 0);
-
-	GLbitfield access = GL_MAP_WRITE_BIT;
-	if ((strategy & GLBufferStrategy::MASK_FLUSH) != 0) {
-		access |= GL_MAP_FLUSH_EXPLICIT_BIT;
-	}
-	if ((strategy & GLBufferStrategy::MASK_INVALIDATE) != 0) {
-		access |= GL_MAP_INVALIDATE_BUFFER_BIT;
-	}
-
-	void *p = nullptr;
-	bool allowNativeBuffer = strategy != GLBufferStrategy::SUBDATA;
-	if (allowNativeBuffer) {
-		glBindBuffer(target_, buffer_);
-
-		if (gl_extensions.ARB_buffer_storage || gl_extensions.EXT_buffer_storage) {
-#if !PPSSPP_PLATFORM(IOS)
-			if (!hasStorage_) {
-				GLbitfield storageFlags = access & ~(GL_MAP_INVALIDATE_BUFFER_BIT | GL_MAP_FLUSH_EXPLICIT_BIT);
-#ifdef USING_GLES2
-#ifdef GL_EXT_buffer_storage
-				glBufferStorageEXT(target_, size_, nullptr, storageFlags);
-#endif
-#else
-				glBufferStorage(target_, size_, nullptr, storageFlags);
-#endif
-				hasStorage_ = true;
-			}
-#endif
-			p = glMapBufferRange(target_, 0, size_, access);
-		} else if (gl_extensions.VersionGEThan(3, 0, 0)) {
-			// GLES3 or desktop 3.
-			p = glMapBufferRange(target_, 0, size_, access);
-		} else if (!gl_extensions.IsGLES) {
-#ifndef USING_GLES2
-			p = glMapBuffer(target_, GL_READ_WRITE);
-#endif
-		}
-	}
-
-	mapped_ = p != nullptr;
-	return p;
-}
-
-bool GLRBuffer::Unmap() {
-	glBindBuffer(target_, buffer_);
-	mapped_ = false;
-	return glUnmapBuffer(target_) == GL_TRUE;
-}
-
-GLPushBuffer::GLPushBuffer(GLRenderManager *render, GLuint target, size_t size, const char *tag) : render_(render), size_(size), target_(target), tag_(tag) {
-	bool res = AddBuffer();
-	_assert_(res);
-	RegisterGPUMemoryManager(this);
-}
-
-GLPushBuffer::~GLPushBuffer() {
-	UnregisterGPUMemoryManager(this);
-	Destroy(true);
-}
-
-void GLPushBuffer::Map() {
-	_assert_(!writePtr_);
-	auto &info = buffers_[buf_];
-	writePtr_ = info.deviceMemory ? info.deviceMemory : info.localMemory;
-	info.flushOffset = 0;
-	// Force alignment.  This is needed for PushAligned() to work as expected.
-	while ((intptr_t)writePtr_ & 15) {
-		writePtr_++;
-		offset_++;
-		info.flushOffset++;
-	}
-	_assert_(writePtr_);
-}
-
-void GLPushBuffer::Unmap() {
-	_assert_(writePtr_);
-	if (!buffers_[buf_].deviceMemory) {
-		// Here we simply upload the data to the last buffer.
-		// Might be worth trying with size_ instead of offset_, so the driver can replace
-		// the whole buffer. At least if it's close.
-		render_->BufferSubdata(buffers_[buf_].buffer, 0, offset_, buffers_[buf_].localMemory, false);
-	} else {
-		buffers_[buf_].flushOffset = offset_;
-	}
-	writePtr_ = nullptr;
-}
-
-void GLPushBuffer::Flush() {
-	// Must be called from the render thread.
-	_dbg_assert_(OnRenderThread());
-
-	if (buf_ >= buffers_.size()) {
-		_dbg_assert_msg_(false, "buf_ somehow got out of sync: %d vs %d", (int)buf_, (int)buffers_.size());
-		return;
-	}
-
-	buffers_[buf_].flushOffset = offset_;
-	if (!buffers_[buf_].deviceMemory && writePtr_) {
-		auto &info = buffers_[buf_];
-		if (info.flushOffset != 0) {
-			_assert_(info.buffer->buffer_);
-			glBindBuffer(target_, info.buffer->buffer_);
-			glBufferSubData(target_, 0, info.flushOffset, info.localMemory);
-		}
-
-		// Here we will submit all the draw calls, with the already known buffer and offsets.
-		// Might as well reset the write pointer here and start over the current buffer.
-		writePtr_ = info.localMemory;
-		offset_ = 0;
-		info.flushOffset = 0;
-	}
-
-	// For device memory, we flush all buffers here.
-	if ((strategy_ & GLBufferStrategy::MASK_FLUSH) != 0) {
-		for (auto &info : buffers_) {
-			if (info.flushOffset == 0 || !info.deviceMemory)
-				continue;
-
-			glBindBuffer(target_, info.buffer->buffer_);
-			glFlushMappedBufferRange(target_, 0, info.flushOffset);
-			info.flushOffset = 0;
-		}
-	}
-}
-
-bool GLPushBuffer::AddBuffer() {
-	// INFO_LOG(G3D, "GLPushBuffer(%s): Allocating %d bytes", tag_, size_);
-	BufInfo info;
-	info.localMemory = (uint8_t *)AllocateAlignedMemory(size_, 16);
-	if (!info.localMemory)
-		return false;
-	info.buffer = render_->CreateBuffer(target_, size_, GL_DYNAMIC_DRAW);
-	info.size = size_;
-	buf_ = buffers_.size();
-	buffers_.push_back(info);
-	return true;
-}
-
-void GLPushBuffer::Destroy(bool onRenderThread) {
-	if (buf_ == -1)
-		return;  // Already destroyed
-	for (BufInfo &info : buffers_) {
-		// This will automatically unmap device memory, if needed.
-		// NOTE: We immediately delete the buffer, don't go through the deleter, if we're on the render thread.
-		if (onRenderThread) {
-			delete info.buffer;
-		} else {
-			render_->DeleteBuffer(info.buffer);
-		}
-		FreeAlignedMemory(info.localMemory);
-	}
-	buffers_.clear();
-	buf_ = -1;
-}
-
-void GLPushBuffer::NextBuffer(size_t minSize) {
-	// First, unmap the current memory.
-	Unmap();
-
-	buf_++;
-	if (buf_ >= buffers_.size() || minSize > size_) {
-		// Before creating the buffer, adjust to the new size_ if necessary.
-		while (size_ < minSize) {
-			size_ <<= 1;
-		}
-
-		bool res = AddBuffer();
-		_assert_(res);
-		if (!res) {
-			// Let's try not to crash at least?
-			buf_ = 0;
-		}
-	}
-
-	// Now, move to the next buffer and map it.
-	offset_ = 0;
-	Map();
-}
-
-void GLPushBuffer::Defragment() {
-	_dbg_assert_msg_(!OnRenderThread(), "Defragment must not run on the render thread");
-
-	if (buffers_.size() <= 1) {
-		// Let's take this opportunity to jettison any localMemory we don't need.
-		for (auto &info : buffers_) {
-			if (info.deviceMemory) {
-				FreeAlignedMemory(info.localMemory);
-				info.localMemory = nullptr;
-			}
-		}
-
-		return;
-	}
-
-	// Okay, we have more than one.  Destroy them all and start over with a larger one.
-
-	// When calling AddBuffer, we sometimes increase size_. So if we allocated multiple buffers in a frame,
-	// they won't all have the same size. Sum things up properly.
-	size_t newSize = 0;
-	for (int i = 0; i < (int)buffers_.size(); i++) {
-		newSize += buffers_[i].size;
-	}
-
-	Destroy(false);
-
-	// Set some sane but very free limits. If there's another spike, we'll just allocate more anyway.
-	size_ = std::min(std::max(newSize, (size_t)65536), (size_t)(512 * 1024 * 1024));
-	bool res = AddBuffer();
-	_assert_msg_(res, "AddBuffer failed");
-}
-
-size_t GLPushBuffer::GetTotalSize() const {
-	size_t sum = 0;
-	// When calling AddBuffer, we sometimes increase size_. So if we allocated multiple buffers in a frame,
-	// they won't all have the same size. Sum things up properly.
-	if (buffers_.size() > 1) {
-		for (int i = 0; i < (int)buffers_.size() - 1; i++) {
-			sum += buffers_[i].size;
-		}
-	}
-	sum += offset_;
-	return sum;
-}
-
-void GLPushBuffer::MapDevice(GLBufferStrategy strategy) {
-	_dbg_assert_msg_(OnRenderThread(), "MapDevice must run on render thread");
-
-	strategy_ = strategy;
-	if (strategy_ == GLBufferStrategy::SUBDATA) {
-		return;
-	}
-
-	bool mapChanged = false;
-	for (auto &info : buffers_) {
-		if (!info.buffer->buffer_ || info.deviceMemory) {
-			// Can't map - no device buffer associated yet or already mapped.
-			continue;
-		}
-
-		info.deviceMemory = (uint8_t *)info.buffer->Map(strategy_);
-		mapChanged = mapChanged || info.deviceMemory != nullptr;
-
-		if (!info.deviceMemory && !info.localMemory) {
-			// Somehow it failed, let's dodge crashing.
-			info.localMemory = (uint8_t *)AllocateAlignedMemory(info.buffer->size_, 16);
-			mapChanged = true;
-		}
-
-		_dbg_assert_msg_(info.localMemory || info.deviceMemory, "Local or device memory must succeed");
-	}
-
-	if (writePtr_ && mapChanged) {
-		// This can happen during a sync.  Remap.
-		writePtr_ = nullptr;
-		Map();
-	}
-}
-
-void GLPushBuffer::UnmapDevice() {
-	_dbg_assert_msg_(OnRenderThread(), "UnmapDevice must run on render thread");
-
-	for (auto &info : buffers_) {
-		if (info.deviceMemory) {
-			// TODO: Technically this can return false?
-			info.buffer->Unmap();
-			info.deviceMemory = nullptr;
-		}
-	}
-}
-
-void GLPushBuffer::GetDebugString(char *buffer, size_t bufSize) const {
-	snprintf(buffer, bufSize, "%s: %s/%s (%d)", tag_, NiceSizeFormat(this->offset_).c_str(), NiceSizeFormat(this->size_).c_str(), (int)buffers_.size());
-}

Common/GPU/OpenGL/GLMemory.h

@@ -1,185 +0,0 @@
-#pragma once
-
-#include <vector>
-#include <cstdint>
-#include <cstring>
-
-#include "Common/GPU/GPUBackendCommon.h"
-#include "Common/GPU/OpenGL/GLCommon.h"
-#include "Common/Log.h"
-
-enum class GLBufferStrategy {
-	SUBDATA = 0,
-
-	MASK_FLUSH = 0x10,
-	MASK_INVALIDATE = 0x20,
-
-	// Map/unmap the buffer each frame.
-	FRAME_UNMAP = 1,
-	// Map/unmap and also invalidate the buffer on map.
-	INVALIDATE_UNMAP = MASK_INVALIDATE,
-	// Map/unmap and explicitly flushed changed ranges.
-	FLUSH_UNMAP = MASK_FLUSH,
-	// Map/unmap, invalidate on map, and explicit flush.
-	FLUSH_INVALIDATE_UNMAP = MASK_FLUSH | MASK_INVALIDATE,
-};
-
-static inline int operator &(const GLBufferStrategy &lhs, const GLBufferStrategy &rhs) {
-	return (int)lhs & (int)rhs;
-}
-
-class GLRBuffer {
-public:
-	GLRBuffer(GLuint target, size_t size) : target_(target), size_((int)size) {}
-	~GLRBuffer() {
-		if (buffer_) {
-			glDeleteBuffers(1, &buffer_);
-		}
-	}
-
-	void *Map(GLBufferStrategy strategy);
-	bool Unmap();
-
-	bool Mapped() const {
-		return mapped_;
-	}
-
-	GLuint buffer_ = 0;
-	GLuint target_;
-	int size_;
-
-private:
-	bool mapped_ = false;
-	bool hasStorage_ = false;
-};
-
-class GLRenderManager;
-
-// Similar to VulkanPushBuffer but is currently less efficient - it collects all the data in
-// RAM then does a big memcpy/buffer upload at the end of the frame. This is at least a lot
-// faster than the hundreds of buffer uploads or memory array buffers we used before.
-// On modern GL we could avoid the copy using glBufferStorage but not sure it's worth the
-// trouble.
-// We need to manage the lifetime of this together with the other resources so its destructor
-// runs on the render thread.
-class GLPushBuffer : public GPUMemoryManager {
-public:
-	friend class GLRenderManager;
-
-	struct BufInfo {
-		GLRBuffer *buffer = nullptr;
-		uint8_t *localMemory = nullptr;
-		uint8_t *deviceMemory = nullptr;
-		size_t flushOffset = 0;
-		size_t size;
-	};
-
-	GLPushBuffer(GLRenderManager *render, GLuint target, size_t size, const char *tag);
-	~GLPushBuffer();
-
-	void Reset() { offset_ = 0; }
-
-	void GetDebugString(char *buffer, size_t bufSize) const override;
-
-	const char *Name() const override { return tag_; };  // for sorting
-
-	// Utility for users of this class, not used internally.
-	enum { INVALID_OFFSET = 0xFFFFFFFF };
-
-private:
-	// Needs context in case of defragment.
-	void Begin() {
-		buf_ = 0;
-		offset_ = 0;
-		// Note: we must defrag because some buffers may be smaller than size_.
-		Defragment();
-		Map();
-		_dbg_assert_(writePtr_);
-	}
-
-	void BeginNoReset() {
-		Map();
-	}
-
-	void End() {
-		Unmap();
-	}
-
-public:
-	void Map();
-	void Unmap();
-
-	bool IsReady() const {
-		return writePtr_ != nullptr;
-	}
-
-	// Recommended - lets you write directly into the buffer through the returned pointer.
-	// If you didn't end up using all the memory you grabbed here, then before calling Allocate or Push
-	// again, call Rewind (see below).
-	uint8_t *Allocate(uint32_t numBytes, uint32_t alignment, GLRBuffer **buf, uint32_t *bindOffset) {
-		uint32_t offset = ((uint32_t)offset_ + alignment - 1) & ~(alignment - 1);
-		if (offset + numBytes <= size_) {
-			// Common path.
-			offset_ = offset + numBytes;
-			*buf = buffers_[buf_].buffer;
-			*bindOffset = offset;
-			return writePtr_ + offset;
-		}
-
-		NextBuffer(numBytes);
-		*bindOffset = 0;
-		*buf = buffers_[buf_].buffer;
-		// Need to mark the allocated range used in the new buffer. How did things work before this?
-		offset_ = numBytes;
-		return writePtr_;
-	}
-
-	// For convenience if all you'll do is to copy.
-	uint32_t Push(const void *data, uint32_t numBytes, int alignment, GLRBuffer **buf) {
-		uint32_t bindOffset;
-		uint8_t *ptr = Allocate(numBytes, alignment, buf, &bindOffset);
-		memcpy(ptr, data, numBytes);
-		return bindOffset;
-	}
-
-	uint8_t *GetPtr(uint32_t offset) {
-		return writePtr_ + offset;
-	}
-
-	// If you didn't use all of the previous allocation you just made (obviously can't be another one),
-	// you can return memory to the buffer by specifying the offset up until which you wrote data.
-	// Pass in the buffer you got last time. If that buffer has been filled already, no rewind can be safely done.
-	// (well technically would be possible but not worth the trouble).
-	void Rewind(GLRBuffer *buffer, uint32_t offset) {
-		if (buffer == buffers_[buf_].buffer) {
-			_dbg_assert_(offset != INVALID_OFFSET);
-			_dbg_assert_(offset <= offset_);
-			offset_ = offset;
-		}
-	}
-
-	size_t GetOffset() const { return offset_; }
-	size_t GetTotalSize() const;
-
-	void Destroy(bool onRenderThread);
-	void Flush();
-
-protected:
-	void MapDevice(GLBufferStrategy strategy);
-	void UnmapDevice();
-
-private:
-	bool AddBuffer();
-	void NextBuffer(size_t minSize);
-	void Defragment();
-
-	GLRenderManager *render_;
-	std::vector<BufInfo> buffers_;
-	size_t buf_ = 0;
-	size_t offset_ = 0;
-	size_t size_ = 0;
-	uint8_t *writePtr_ = nullptr;
-	GLuint target_;
-	GLBufferStrategy strategy_ = GLBufferStrategy::SUBDATA;
-	const char *tag_;
-};

Common/GPU/OpenGL/GLQueueRunner.cpp

@@ -73,19 +73,24 @@ void GLQueueRunner::CreateDeviceObjects() {
 	populate(GL_SHADING_LANGUAGE_VERSION);
 	CHECK_GL_ERROR_IF_DEBUG();
 
-#if !PPSSPP_ARCH(X86)  // Doesn't work on AMD for some reason. See issue #17787
 	useDebugGroups_ = !gl_extensions.IsGLES && gl_extensions.VersionGEThan(4, 3);
-#endif
 }
 
 void GLQueueRunner::DestroyDeviceObjects() {
 	CHECK_GL_ERROR_IF_DEBUG();
+	if (!nameCache_.empty()) {
+		glDeleteTextures((GLsizei)nameCache_.size(), &nameCache_[0]);
+		nameCache_.clear();
+	}
 	if (gl_extensions.ARB_vertex_array_object) {
 		glDeleteVertexArrays(1, &globalVAO_);
 	}
 	delete[] readbackBuffer_;
 	readbackBuffer_ = nullptr;
 	readbackBufferSize_ = 0;
+	delete[] tempBuffer_;
+	tempBuffer_ = nullptr;
+	tempBufferSize_ = 0;
 	CHECK_GL_ERROR_IF_DEBUG();
 }
 
@@ -107,20 +112,20 @@ static std::string GetInfoLog(GLuint name, Getiv getiv, GetLog getLog) {
 	return infoLog;
 }
 
-static int GetStereoBufferIndex(const char *uniformName) {
+int GLQueueRunner::GetStereoBufferIndex(const char *uniformName) {
 	if (!uniformName) return -1;
 	else if (strcmp(uniformName, "u_view") == 0) return 0;
 	else if (strcmp(uniformName, "u_proj_lens") == 0) return 1;
 	else return -1;
 }
 
-static std::string GetStereoBufferLayout(const char *uniformName) {
+std::string GLQueueRunner::GetStereoBufferLayout(const char *uniformName) {
 	if (strcmp(uniformName, "u_view") == 0) return "ViewMatrices";
 	else if (strcmp(uniformName, "u_proj_lens") == 0) return "ProjectionMatrix";
 	else return "undefined";
 }
 
-void GLQueueRunner::RunInitSteps(const FastVec<GLRInitStep> &steps, bool skipGLCalls) {
+void GLQueueRunner::RunInitSteps(const std::vector<GLRInitStep> &steps, bool skipGLCalls) {
 	if (skipGLCalls) {
 		// Some bookkeeping still needs to be done.
 		for (size_t i = 0; i < steps.size(); i++) {
@@ -325,8 +330,8 @@ void GLQueueRunner::RunInitSteps(const FastVec<GLRInitStep> &steps, bool skipGLC
 			glCompileShader(shader);
 			GLint success = 0;
 			glGetShaderiv(shader, GL_COMPILE_STATUS, &success);
+			std::string infoLog = GetInfoLog(shader, glGetShaderiv, glGetShaderInfoLog);
 			if (!success) {
-				std::string infoLog = GetInfoLog(shader, glGetShaderiv, glGetShaderInfoLog);
 				std::string errorString = StringFromFormat(
 					"Error in shader compilation for: %s\n"
 					"Info log: %s\n"
@@ -334,10 +339,10 @@ void GLQueueRunner::RunInitSteps(const FastVec<GLRInitStep> &steps, bool skipGLC
 					step.create_shader.shader->desc.c_str(),
 					infoLog.c_str(),
 					LineNumberString(code).c_str());
-				std::vector<std::string_view> lines;
+				std::vector<std::string> lines;
 				SplitString(errorString, '\n', lines);
-				for (auto line : lines) {
+				for (auto &line : lines) {
-					ERROR_LOG(G3D, "%.*s", (int)line.size(), line.data());
+					ERROR_LOG(G3D, "%s", line.c_str());
 				}
 				if (errorCallback_) {
 					std::string desc = StringFromFormat("Shader compilation failed: %s", step.create_shader.stage == GL_VERTEX_SHADER ? "vertex" : "fragment");
@@ -385,23 +390,14 @@ void GLQueueRunner::RunInitSteps(const FastVec<GLRInitStep> &steps, bool skipGLC
 
 			// For things to show in RenderDoc, need to split into glTexImage2D(..., nullptr) and glTexSubImage.
 
-			int blockSize = 0;
-			bool bc = Draw::DataFormatIsBlockCompressed(step.texture_image.format, &blockSize);
-
 			GLenum internalFormat, format, type;
 			int alignment;
 			Thin3DFormatToGLFormatAndType(step.texture_image.format, internalFormat, format, type, alignment);
 			if (step.texture_image.depth == 1) {
-				if (bc) {
+				glTexImage2D(tex->target,
-					int dataSize = ((step.texture_image.width + 3) & ~3) * ((step.texture_image.height + 3) & ~3) * blockSize / 16;
+					step.texture_image.level, internalFormat,
-					glCompressedTexImage2D(tex->target, step.texture_image.level, internalFormat,
+					step.texture_image.width, step.texture_image.height, 0,
-						step.texture_image.width, step.texture_image.height, 0, dataSize, step.texture_image.data);
+					format, type, step.texture_image.data);
-				} else {
-					glTexImage2D(tex->target,
-						step.texture_image.level, internalFormat,
-						step.texture_image.width, step.texture_image.height, 0,
-						format, type, step.texture_image.data);
-				}
 			} else {
 				glTexImage3D(tex->target,
 					step.texture_image.level, internalFormat,
@@ -512,8 +508,8 @@ void GLQueueRunner::InitCreateFramebuffer(const GLRInitStep &step) {
 
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, tex.wrapS);
 		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, tex.wrapT);
-		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, tex.magFilter);
+		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, tex.magFilter);
-		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, tex.minFilter);
+		glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, tex.minFilter);
 		if (!gl_extensions.IsGLES || gl_extensions.GLES3) {
 			glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 0);
 		}
@@ -521,7 +517,22 @@ void GLQueueRunner::InitCreateFramebuffer(const GLRInitStep &step) {
 
 	// Color texture is same everywhere
 	glGenFramebuffers(1, &fbo->handle);
-	initFBOTexture(fbo->color_texture, GL_RGBA, GL_RGBA, GL_UNSIGNED_BYTE, true);
+
+	GLint colorInternalFormat;
+	GLint colorFormat;
+	GLint colorElementType;
+	switch (fbo->colorFormat) {
+	case Draw::DataFormat::R8G8B8A8_UNORM:
+		colorInternalFormat = GL_RGBA;
+		colorFormat = GL_RGBA;
+		colorElementType = GL_UNSIGNED_BYTE;
+		break;
+	default:
+		_assert_msg_(false, "Data format not supported");
+		return;
+	}
+
+	initFBOTexture(fbo->color_texture, colorInternalFormat, colorFormat, colorElementType, true);
 
 retry_depth:
 	if (!fbo->z_stencil_) {
@@ -653,7 +664,7 @@ retry_depth:
 	currentReadHandle_ = fbo->handle;
 }
 
-void GLQueueRunner::RunSteps(const std::vector<GLRStep *> &steps, GLFrameData &frameData, bool skipGLCalls, bool keepSteps, bool useVR) {
+void GLQueueRunner::RunSteps(const std::vector<GLRStep *> &steps, bool skipGLCalls, bool keepSteps, bool useVR) {
 	if (skipGLCalls) {
 		if (keepSteps) {
 			return;
@@ -702,7 +713,7 @@ void GLQueueRunner::RunSteps(const std::vector<GLRStep *> &steps, GLFrameData &f
 	CHECK_GL_ERROR_IF_DEBUG();
 	size_t renderCount = 0;
 	for (size_t i = 0; i < steps.size(); i++) {
-		GLRStep &step = *steps[i];
+		const GLRStep &step = *steps[i];
 
 #if !defined(USING_GLES2)
 		if (useDebugGroups_)
@@ -713,10 +724,11 @@ void GLQueueRunner::RunSteps(const std::vector<GLRStep *> &steps, GLFrameData &f
 		case GLRStepType::RENDER:
 			renderCount++;
 			if (IsVREnabled()) {
-				PreprocessStepVR(&step);
+				GLRStep vrStep = step;
-				PerformRenderPass(step, renderCount == 1, renderCount == totalRenderCount, frameData.profile);
+				PreprocessStepVR(&vrStep);
+				PerformRenderPass(vrStep, renderCount == 1, renderCount == totalRenderCount);
 			} else {
-				PerformRenderPass(step, renderCount == 1, renderCount == totalRenderCount, frameData.profile);
+				PerformRenderPass(step, renderCount == 1, renderCount == totalRenderCount);
 			}
 			break;
 		case GLRStepType::COPY:
@@ -742,17 +754,19 @@ void GLQueueRunner::RunSteps(const std::vector<GLRStep *> &steps, GLFrameData &f
 		if (useDebugGroups_)
 			glPopDebugGroup();
 #endif
-		if (frameData.profile.enabled) {
+
-			frameData.profile.passesString += StepToString(step);
-		}
 		if (!keepSteps) {
 			delete steps[i];
 		}
 	}
-
 	CHECK_GL_ERROR_IF_DEBUG();
 }
 
+void GLQueueRunner::LogSteps(const std::vector<GLRStep *> &steps) {
+
+}
+
+
 void GLQueueRunner::PerformBlit(const GLRStep &step) {
 	CHECK_GL_ERROR_IF_DEBUG();
 	// Without FBO_ARB / GLES3, this will collide with bind_for_read, but there's nothing
@@ -782,20 +796,7 @@ void GLQueueRunner::PerformBlit(const GLRStep &step) {
 	}
 }
 
-static void EnableDisableVertexArrays(uint32_t prevAttr, uint32_t newAttr) {
+void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last) {
-	int enable = (~prevAttr) & newAttr;
-	int disable = prevAttr & (~newAttr);
-	for (int i = 0; i < 7; i++) {  // SEM_MAX
-		if (enable & (1 << i)) {
-			glEnableVertexAttribArray(i);
-		}
-		if (disable & (1 << i)) {
-			glDisableVertexAttribArray(i);
-		}
-	}
-}
-
-void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last, GLQueueProfileContext &profile) {
 	CHECK_GL_ERROR_IF_DEBUG();
 
 	PerformBindFramebufferAsRenderTarget(step);
@@ -812,9 +813,10 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 			glDisable(GL_COLOR_LOGIC_OP);
 		}
 #endif
-		if (gl_extensions.ARB_vertex_array_object) {
+	}
-			glBindVertexArray(globalVAO_);
+
-		}
+	if (first && gl_extensions.ARB_vertex_array_object) {
+		glBindVertexArray(globalVAO_);
 	}
 
 	GLRProgram *curProgram = nullptr;
@@ -840,43 +842,12 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 	bool clipDistanceEnabled[8]{};
 	GLuint blendEqColor = (GLuint)-1;
 	GLuint blendEqAlpha = (GLuint)-1;
-	GLenum blendSrcColor = (GLenum)-1;
-	GLenum blendDstColor = (GLenum)-1;
-	GLenum blendSrcAlpha = (GLenum)-1;
-	GLenum blendDstAlpha = (GLenum)-1;
 
-	GLuint stencilWriteMask = (GLuint)-1;
-	GLenum stencilFunc = (GLenum)-1;
-	GLuint stencilRef = (GLuint)-1;
-	GLuint stencilCompareMask = (GLuint)-1;
-	GLenum stencilSFail = (GLenum)-1;
-	GLenum stencilZFail = (GLenum)-1;
-	GLenum stencilPass = (GLenum)-1;
-	GLenum frontFace = (GLenum)-1;
-	GLenum cullFace = (GLenum)-1;
 	GLRTexture *curTex[MAX_GL_TEXTURE_SLOTS]{};
 
-	GLRViewport viewport = {
-		-1000000000.0f,
-		-1000000000.0f,
-		-1000000000.0f,
-		-1000000000.0f,
-		-1000000000.0f,
-		-1000000000.0f,
-	};
-
-	GLRect2D scissorRc = { -1, -1, -1, -1 };
-
 	CHECK_GL_ERROR_IF_DEBUG();
 	auto &commands = step.commands;
 	for (const auto &c : commands) {
-#ifdef _DEBUG
-		if (profile.enabled) {
-			if ((size_t)c.cmd < ARRAY_SIZE(profile.commandCounts)) {
-				profile.commandCounts[(size_t)c.cmd]++;
-			}
-		}
-#endif
 		switch (c.cmd) {
 		case GLRRenderCommand::DEPTH:
 			if (c.depth.enabled) {
@@ -892,39 +863,28 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 					glDepthFunc(c.depth.func);
 					depthFunc = c.depth.func;
 				}
-			} else if (/* !c.depth.enabled && */ depthEnabled) {
+			} else if (!c.depth.enabled && depthEnabled) {
 				glDisable(GL_DEPTH_TEST);
 				depthEnabled = false;
 			}
 			break;
-		case GLRRenderCommand::STENCIL:
+		case GLRRenderCommand::STENCILFUNC:
-			if (c.stencil.enabled) {
+			if (c.stencilFunc.enabled) {
 				if (!stencilEnabled) {
 					glEnable(GL_STENCIL_TEST);
 					stencilEnabled = true;
 				}
-				if (c.stencil.func != stencilFunc || c.stencil.ref != stencilRef || c.stencil.compareMask != stencilCompareMask) {
+				glStencilFunc(c.stencilFunc.func, c.stencilFunc.ref, c.stencilFunc.compareMask);
-					glStencilFunc(c.stencil.func, c.stencil.ref, c.stencil.compareMask);
+			} else if (stencilEnabled) {
-					stencilFunc = c.stencil.func;
-					stencilRef = c.stencil.ref;
-					stencilCompareMask = c.stencil.compareMask;
-				}
-				if (c.stencil.sFail != stencilSFail || c.stencil.zFail != stencilZFail || c.stencil.pass != stencilPass) {
-					glStencilOp(c.stencil.sFail, c.stencil.zFail, c.stencil.pass);
-					stencilSFail = c.stencil.sFail;
-					stencilZFail = c.stencil.zFail;
-					stencilPass = c.stencil.pass;
-				}
-				if (c.stencil.writeMask != stencilWriteMask) {
-					glStencilMask(c.stencil.writeMask);
-					stencilWriteMask = c.stencil.writeMask;
-				}
-			} else if (/* !c.stencilFunc.enabled && */stencilEnabled) {
 				glDisable(GL_STENCIL_TEST);
 				stencilEnabled = false;
 			}
 			CHECK_GL_ERROR_IF_DEBUG();
 			break;
+		case GLRRenderCommand::STENCILOP:
+			glStencilOp(c.stencilOp.sFail, c.stencilOp.zFail, c.stencilOp.pass);
+			glStencilMask(c.stencilOp.writeMask);
+			break;
 		case GLRRenderCommand::BLEND:
 			if (c.blend.enabled) {
 				if (!blendEnabled) {
@@ -936,14 +896,8 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 					blendEqColor = c.blend.funcColor;
 					blendEqAlpha = c.blend.funcAlpha;
 				}
-				if (blendSrcColor != c.blend.srcColor || blendDstColor != c.blend.dstColor || blendSrcAlpha != c.blend.srcAlpha || blendDstAlpha != c.blend.dstAlpha) {
+				glBlendFuncSeparate(c.blend.srcColor, c.blend.dstColor, c.blend.srcAlpha, c.blend.dstAlpha);
-					glBlendFuncSeparate(c.blend.srcColor, c.blend.dstColor, c.blend.srcAlpha, c.blend.dstAlpha);
+			} else if (!c.blend.enabled && blendEnabled) {
-					blendSrcColor = c.blend.srcColor;
-					blendDstColor = c.blend.dstColor;
-					blendSrcAlpha = c.blend.srcAlpha;
-					blendDstAlpha = c.blend.dstAlpha;
-				}
-			} else if (/* !c.blend.enabled && */ blendEnabled) {
 				glDisable(GL_BLEND);
 				blendEnabled = false;
 			}
@@ -963,7 +917,7 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 				if (logicOp != c.logic.logicOp) {
 					glLogicOp(c.logic.logicOp);
 				}
-			} else if (/* !c.logic.enabled && */ logicEnabled) {
+			} else if (!c.logic.enabled && logicEnabled) {
 				glDisable(GL_COLOR_LOGIC_OP);
 				logicEnabled = false;
 			}
@@ -1020,27 +974,16 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 				y = curFBHeight_ - y - c.viewport.vp.h;
 
 			// TODO: Support FP viewports through glViewportArrays
-			if (viewport.x != c.viewport.vp.x || viewport.y != y || viewport.w != c.viewport.vp.w || viewport.h != c.viewport.vp.h) {
+			glViewport((GLint)c.viewport.vp.x, (GLint)y, (GLsizei)c.viewport.vp.w, (GLsizei)c.viewport.vp.h);
-				glViewport((GLint)c.viewport.vp.x, (GLint)y, (GLsizei)c.viewport.vp.w, (GLsizei)c.viewport.vp.h);
-				viewport.x = c.viewport.vp.x;
-				viewport.y = y;
-				viewport.w = c.viewport.vp.w;
-				viewport.h = c.viewport.vp.h;
-			}
-
-			if (viewport.minZ != c.viewport.vp.minZ || viewport.maxZ != c.viewport.vp.maxZ) {
-				viewport.minZ = c.viewport.vp.minZ;
-				viewport.maxZ = c.viewport.vp.maxZ;
 #if !defined(USING_GLES2)
-				if (gl_extensions.IsGLES) {
+			if (gl_extensions.IsGLES) {
-					glDepthRangef(c.viewport.vp.minZ, c.viewport.vp.maxZ);
-				} else {
-					glDepthRange(c.viewport.vp.minZ, c.viewport.vp.maxZ);
-				}
-#else
 				glDepthRangef(c.viewport.vp.minZ, c.viewport.vp.maxZ);
-#endif
+			} else {
+				glDepthRange(c.viewport.vp.minZ, c.viewport.vp.maxZ);
 			}
+#else
+			glDepthRangef(c.viewport.vp.minZ, c.viewport.vp.maxZ);
+#endif
 			CHECK_GL_ERROR_IF_DEBUG();
 			break;
 		}
@@ -1049,30 +992,30 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 			int y = c.scissor.rc.y;
 			if (!curFB_)
 				y = curFBHeight_ - y - c.scissor.rc.h;
-			if (scissorRc.x != c.scissor.rc.x || scissorRc.y != y || scissorRc.w != c.scissor.rc.w || scissorRc.h != c.scissor.rc.h) {
+			glScissor(c.scissor.rc.x, y, c.scissor.rc.w, c.scissor.rc.h);
-				glScissor(c.scissor.rc.x, y, c.scissor.rc.w, c.scissor.rc.h);
-				scissorRc.x = c.scissor.rc.x;
-				scissorRc.y = y;
-				scissorRc.w = c.scissor.rc.w;
-				scissorRc.h = c.scissor.rc.h;
-			}
 			CHECK_GL_ERROR_IF_DEBUG();
 			break;
 		}
 		case GLRRenderCommand::UNIFORM4F:
 		{
-			_dbg_assert_(curProgram);
 			int loc = c.uniform4.loc ? *c.uniform4.loc : -1;
 			if (c.uniform4.name) {
 				loc = curProgram->GetUniformLoc(c.uniform4.name);
 			}
 			if (loc >= 0) {
-				_dbg_assert_(c.uniform4.count >=1 && c.uniform4.count <=4);
 				switch (c.uniform4.count) {
-				case 1: glUniform1f(loc, c.uniform4.v[0]); break;
+				case 1:
-				case 2: glUniform2fv(loc, 1, c.uniform4.v); break;
+					glUniform1f(loc, c.uniform4.v[0]);
-				case 3: glUniform3fv(loc, 1, c.uniform4.v); break;
+					break;
-				case 4: glUniform4fv(loc, 1, c.uniform4.v); break;
+				case 2:
+					glUniform2fv(loc, 1, c.uniform4.v);
+					break;
+				case 3:
+					glUniform3fv(loc, 1, c.uniform4.v);
+					break;
+				case 4:
+					glUniform4fv(loc, 1, c.uniform4.v);
+					break;
 				}
 			}
 			CHECK_GL_ERROR_IF_DEBUG();
@@ -1086,7 +1029,6 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 				loc = curProgram->GetUniformLoc(c.uniform4.name);
 			}
 			if (loc >= 0) {
-				_dbg_assert_(c.uniform4.count >=1 && c.uniform4.count <=4);
 				switch (c.uniform4.count) {
 				case 1: glUniform1uiv(loc, 1, (GLuint *)c.uniform4.v); break;
 				case 2: glUniform2uiv(loc, 1, (GLuint *)c.uniform4.v); break;
@@ -1105,7 +1047,6 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 				loc = curProgram->GetUniformLoc(c.uniform4.name);
 			}
 			if (loc >= 0) {
-				_dbg_assert_(c.uniform4.count >=1 && c.uniform4.count <=4);
 				switch (c.uniform4.count) {
 				case 1: glUniform1iv(loc, 1, (GLint *)c.uniform4.v); break;
 				case 2: glUniform2iv(loc, 1, (GLint *)c.uniform4.v); break;
@@ -1120,34 +1061,28 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 		{
 			_dbg_assert_(curProgram);
 			if (IsMultiviewSupported()) {
-				int layout = GetStereoBufferIndex(c.uniformStereoMatrix4.name);
+				int layout = GetStereoBufferIndex(c.uniformMatrix4.name);
 				if (layout >= 0) {
 					int size = 2 * 16 * sizeof(float);
-					glBindBufferBase(GL_UNIFORM_BUFFER, layout, *c.uniformStereoMatrix4.loc);
+					glBindBufferBase(GL_UNIFORM_BUFFER, layout, *c.uniformMatrix4.loc);
-					glBindBuffer(GL_UNIFORM_BUFFER, *c.uniformStereoMatrix4.loc);
+					glBindBuffer(GL_UNIFORM_BUFFER, *c.uniformMatrix4.loc);
 					void *matrices = glMapBufferRange(GL_UNIFORM_BUFFER, 0, size, GL_MAP_WRITE_BIT | GL_MAP_INVALIDATE_BUFFER_BIT);
-					memcpy(matrices, c.uniformStereoMatrix4.mData, size);
+					memcpy(matrices, c.uniformMatrix4.m, size);
 					glUnmapBuffer(GL_UNIFORM_BUFFER);
 					glBindBuffer(GL_UNIFORM_BUFFER, 0);
 				}
-				delete[] c.uniformStereoMatrix4.mData;  // We only playback once.
 			} else {
-				int loc = c.uniformStereoMatrix4.loc ? *c.uniformStereoMatrix4.loc : -1;
+				int loc = c.uniformMatrix4.loc ? *c.uniformMatrix4.loc : -1;
-				if (c.uniformStereoMatrix4.name) {
+				if (c.uniformMatrix4.name) {
-					loc = curProgram->GetUniformLoc(c.uniformStereoMatrix4.name);
+					loc = curProgram->GetUniformLoc(c.uniformMatrix4.name);
 				}
 				if (loc >= 0) {
 					if (GetVRFBOIndex() == 0) {
-						glUniformMatrix4fv(loc, 1, false, c.uniformStereoMatrix4.mData);
+						glUniformMatrix4fv(loc, 1, false, c.uniformMatrix4.m);
 					} else {
-						glUniformMatrix4fv(loc, 1, false, c.uniformStereoMatrix4.mData + 16);
+						glUniformMatrix4fv(loc, 1, false, &c.uniformMatrix4.m[16]);
 					}
 				}
-				if (GetVRFBOIndex() == 1 || GetVRPassesCount() == 1) {
-					// Only delete the data if we're rendering the only or the second eye.
-					// If we delete during the first eye, we get a use-after-free or double delete.
-					delete[] c.uniformStereoMatrix4.mData;
-				}
 			}
 			CHECK_GL_ERROR_IF_DEBUG();
 			break;
@@ -1236,37 +1171,52 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 			CHECK_GL_ERROR_IF_DEBUG();
 			break;
 		}
-		case GLRRenderCommand::DRAW:
+		case GLRRenderCommand::BIND_VERTEX_BUFFER:
 		{
-			GLRInputLayout *layout = c.draw.inputLayout;
+			// TODO: Add fast path for glBindVertexBuffer
-			GLuint buf = c.draw.vertexBuffer->buffer_;
+			GLRInputLayout *layout = c.bindVertexBuffer.inputLayout;
-			_dbg_assert_(!c.draw.vertexBuffer->Mapped());
+			GLuint buf = c.bindVertexBuffer.buffer ? c.bindVertexBuffer.buffer->buffer_ : 0;
+			_dbg_assert_(!c.bindVertexBuffer.buffer || !c.bindVertexBuffer.buffer->Mapped());
 			if (buf != curArrayBuffer) {
 				glBindBuffer(GL_ARRAY_BUFFER, buf);
 				curArrayBuffer = buf;
 			}
 			if (attrMask != layout->semanticsMask_) {
-				EnableDisableVertexArrays(attrMask, layout->semanticsMask_);
+				int enable = layout->semanticsMask_ & ~attrMask;
+				int disable = (~layout->semanticsMask_) & attrMask;
+
+				for (int i = 0; i < 7; i++) {  // SEM_MAX
+					if (enable & (1 << i)) {
+						glEnableVertexAttribArray(i);
+					}
+					if (disable & (1 << i)) {
+						glDisableVertexAttribArray(i);
+					}
+				}
 				attrMask = layout->semanticsMask_;
 			}
 			for (size_t i = 0; i < layout->entries.size(); i++) {
 				auto &entry = layout->entries[i];
-				glVertexAttribPointer(entry.location, entry.count, entry.type, entry.normalized, layout->stride, (const void *)(c.draw.vertexOffset + entry.offset));
+				glVertexAttribPointer(entry.location, entry.count, entry.type, entry.normalized, entry.stride, (const void *)(c.bindVertexBuffer.offset + entry.offset));
 			}
-			if (c.draw.indexBuffer) {
+			CHECK_GL_ERROR_IF_DEBUG();
-				GLuint buf = c.draw.indexBuffer->buffer_;
+			break;
-				_dbg_assert_(!c.draw.indexBuffer->Mapped());
+		}
+		case GLRRenderCommand::BIND_BUFFER:
+		{
+			if (c.bind_buffer.target == GL_ARRAY_BUFFER) {
+				Crash();
+			} else if (c.bind_buffer.target == GL_ELEMENT_ARRAY_BUFFER) {
+				GLuint buf = c.bind_buffer.buffer ? c.bind_buffer.buffer->buffer_ : 0;
+				_dbg_assert_(!c.bind_buffer.buffer->Mapped());
 				if (buf != curElemArrayBuffer) {
 					glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, buf);
 					curElemArrayBuffer = buf;
 				}
-				if (c.draw.instances == 1) {
-					glDrawElements(c.draw.mode, c.draw.count, c.draw.indexType, (void *)(intptr_t)c.draw.indexOffset);
-				} else {
-					glDrawElementsInstanced(c.draw.mode, c.draw.count, c.draw.indexType, (void *)(intptr_t)c.draw.indexOffset, c.draw.instances);
-				}
 			} else {
-				glDrawArrays(c.draw.mode, c.draw.first, c.draw.count);
+				GLuint buf = c.bind_buffer.buffer ? c.bind_buffer.buffer->buffer_ : 0;
+				_dbg_assert_(!c.bind_buffer.buffer->Mapped());
+				glBindBuffer(c.bind_buffer.target, buf);
 			}
 			CHECK_GL_ERROR_IF_DEBUG();
 			break;
@@ -1275,6 +1225,16 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 			// TODO: Should we include the texture handle in the command?
 			// Also, should this not be an init command?
 			glGenerateMipmap(GL_TEXTURE_2D);
+			break;
+		case GLRRenderCommand::DRAW:
+			glDrawArrays(c.draw.mode, c.draw.first, c.draw.count);
+			break;
+		case GLRRenderCommand::DRAW_INDEXED:
+			if (c.drawIndexed.instances == 1) {
+				glDrawElements(c.drawIndexed.mode, c.drawIndexed.count, c.drawIndexed.indexType, c.drawIndexed.indices);
+			} else {
+				glDrawElementsInstanced(c.drawIndexed.mode, c.drawIndexed.count, c.drawIndexed.indexType, c.drawIndexed.indices, c.drawIndexed.instances);
+			}
 			CHECK_GL_ERROR_IF_DEBUG();
 			break;
 		case GLRRenderCommand::TEXTURESAMPLER:
@@ -1322,7 +1282,7 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 		}
 		case GLRRenderCommand::TEXTURELOD:
 		{
-			GLint slot = c.textureLod.slot;
+			GLint slot = c.textureSampler.slot;
 			if (slot != activeSlot) {
 				glActiveTexture(GL_TEXTURE0 + slot);
 				activeSlot = slot;
@@ -1378,15 +1338,9 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 					glEnable(GL_CULL_FACE);
 					cullEnabled = true;
 				}
-				if (frontFace != c.raster.frontFace) {
+				glFrontFace(c.raster.frontFace);
-					glFrontFace(c.raster.frontFace);
+				glCullFace(c.raster.cullFace);
-					frontFace = c.raster.frontFace;
+			} else if (!c.raster.cullEnable && cullEnabled) {
-				}
-				if (cullFace != c.raster.cullFace) {
-					glCullFace(c.raster.cullFace);
-					cullFace = c.raster.cullFace;
-				}
-			} else if (/* !c.raster.cullEnable && */ cullEnabled) {
 				glDisable(GL_CULL_FACE);
 				cullEnabled = false;
 			}
@@ -1395,7 +1349,7 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 					glEnable(GL_DITHER);
 					ditherEnabled = true;
 				}
-			} else if (/* !c.raster.ditherEnable && */ ditherEnabled) {
+			} else if (!c.raster.ditherEnable && ditherEnabled) {
 				glDisable(GL_DITHER);
 				ditherEnabled = false;
 			}
@@ -1405,7 +1359,7 @@ void GLQueueRunner::PerformRenderPass(const GLRStep &step, bool first, bool last
 					glEnable(GL_DEPTH_CLAMP);
 					depthClampEnabled = true;
 				}
-			} else if (/* !c.raster.depthClampEnable && */ depthClampEnabled) {
+			} else if (!c.raster.depthClampEnable && depthClampEnabled) {
 				glDisable(GL_DEPTH_CLAMP);
 				depthClampEnabled = false;
 			}
@@ -1542,24 +1496,26 @@ void GLQueueRunner::PerformReadback(const GLRStep &pass) {
 	CHECK_GL_ERROR_IF_DEBUG();
 
 	// Always read back in 8888 format for the color aspect.
+	GLuint internalFormat = GL_RGBA;
 	GLuint format = GL_RGBA;
 	GLuint type = GL_UNSIGNED_BYTE;
 	int srcAlignment = 4;
+	int dstAlignment = (int)DataFormatSizeInBytes(pass.readback.dstFormat);
 
 #ifndef USING_GLES2
 	if (pass.readback.aspectMask & GL_DEPTH_BUFFER_BIT) {
+		internalFormat = GL_DEPTH_COMPONENT;
 		format = GL_DEPTH_COMPONENT;
 		type = GL_FLOAT;
 		srcAlignment = 4;
 	} else if (pass.readback.aspectMask & GL_STENCIL_BUFFER_BIT) {
+		internalFormat = GL_STENCIL_INDEX;
 		format = GL_STENCIL_INDEX;
 		type = GL_UNSIGNED_BYTE;
 		srcAlignment = 1;
 	}
 #endif
 
-	readbackAspectMask_ = pass.readback.aspectMask;
-
 	int pixelStride = pass.readback.srcRect.w;
 	// Apply the correct alignment.
 	glPixelStorei(GL_PACK_ALIGNMENT, srcAlignment);
@@ -1570,20 +1526,31 @@ void GLQueueRunner::PerformReadback(const GLRStep &pass) {
 
 	GLRect2D rect = pass.readback.srcRect;
 
-	int readbackSize = srcAlignment * rect.w * rect.h;
+	bool convert = internalFormat == GL_RGBA && pass.readback.dstFormat != DataFormat::R8G8B8A8_UNORM;
+
+	int tempSize = srcAlignment * rect.w * rect.h;
+	int readbackSize = dstAlignment * rect.w * rect.h;
+	if (convert && tempSize > tempBufferSize_) {
+		delete[] tempBuffer_;
+		tempBuffer_ = new uint8_t[tempSize];
+		tempBufferSize_ = tempSize;
+	}
 	if (readbackSize > readbackBufferSize_) {
 		delete[] readbackBuffer_;
 		readbackBuffer_ = new uint8_t[readbackSize];
 		readbackBufferSize_ = readbackSize;
 	}
 
-	glReadPixels(rect.x, rect.y, rect.w, rect.h, format, type, readbackBuffer_);
+	glReadPixels(rect.x, rect.y, rect.w, rect.h, format, type, convert ? tempBuffer_ : readbackBuffer_);
 	#ifdef DEBUG_READ_PIXELS
 	LogReadPixelsError(glGetError());
 	#endif
 	if (!gl_extensions.IsGLES || gl_extensions.GLES3) {
 		glPixelStorei(GL_PACK_ROW_LENGTH, 0);
 	}
+	if (convert && tempBuffer_ && readbackBuffer_) {
+		ConvertFromRGBA8888(readbackBuffer_, tempBuffer_, pixelStride, pixelStride, rect.w, rect.h, pass.readback.dstFormat);
+	}
 	CHECK_GL_ERROR_IF_DEBUG();
 }
 
@@ -1612,7 +1579,7 @@ void GLQueueRunner::PerformReadbackImage(const GLRStep &pass) {
 		glGetTexLevelParameteriv(GL_TEXTURE_2D, pass.readback_image.mipLevel, GL_TEXTURE_WIDTH, &w);
 		glGetTexLevelParameteriv(GL_TEXTURE_2D, pass.readback_image.mipLevel, GL_TEXTURE_HEIGHT, &h);
 
-		int size = 4 * std::max((int)w, rect.x + rect.w) * std::max((int)h, rect.y + rect.h);
+		int size = 4 * std::max((int)w, rect.x + rect.w) * std::max((int)h, rect.h);
 		if (size > readbackBufferSize_) {
 			delete[] readbackBuffer_;
 			readbackBuffer_ = new uint8_t[size];
@@ -1663,7 +1630,7 @@ void GLQueueRunner::PerformBindFramebufferAsRenderTarget(const GLRStep &pass) {
 	CHECK_GL_ERROR_IF_DEBUG();
 }
 
-void GLQueueRunner::CopyFromReadbackBuffer(GLRFramebuffer *framebuffer, int width, int height, Draw::DataFormat srcFormat, Draw::DataFormat destFormat, int pixelStride, uint8_t *pixels) {
+void GLQueueRunner::CopyReadbackBuffer(int width, int height, Draw::DataFormat srcFormat, Draw::DataFormat destFormat, int pixelStride, uint8_t *pixels) {
 	// TODO: Maybe move data format conversion here, and always read back 8888. Drivers
 	// don't usually provide very optimized conversion implementations, though some do.
 	// Just need to be careful about dithering, which may break Danganronpa.
@@ -1672,26 +1639,20 @@ void GLQueueRunner::CopyFromReadbackBuffer(GLRFramebuffer *framebuffer, int widt
 		// Something went wrong during the read and no readback buffer was allocated, probably.
 		return;
 	}
-
+	for (int y = 0; y < height; y++) {
-	// Always read back in 8888 format for the color aspect.
+		memcpy(pixels + y * pixelStride * bpp, readbackBuffer_ + y * width * bpp, width * bpp);
-	GLuint internalFormat = GL_RGBA;
-#ifndef USING_GLES2
-	if (readbackAspectMask_ & GL_DEPTH_BUFFER_BIT) {
-		internalFormat = GL_DEPTH_COMPONENT;
-	} else if (readbackAspectMask_ & GL_STENCIL_BUFFER_BIT) {
-		internalFormat = GL_STENCIL_INDEX;
 	}
-#endif
+}
 
-	bool convert = internalFormat == GL_RGBA && destFormat != Draw::DataFormat::R8G8B8A8_UNORM;
+GLuint GLQueueRunner::AllocTextureName() {
-	if (convert) {
+	if (nameCache_.empty()) {
-		// srcStride is width because we read back "packed" (with no gaps) from GL.
+		nameCache_.resize(TEXCACHE_NAME_CACHE_SIZE);
-		ConvertFromRGBA8888(pixels, readbackBuffer_, pixelStride, width, width, height, destFormat);
+		glGenTextures(TEXCACHE_NAME_CACHE_SIZE, &nameCache_[0]);
-	} else {
-		for (int y = 0; y < height; y++) {
-			memcpy(pixels + y * pixelStride * bpp, readbackBuffer_ + y * width * bpp, width * bpp);
-		}
 	}
+	u32 name = nameCache_.back();
+	nameCache_.pop_back();
+	CHECK_GL_ERROR_IF_DEBUG();
+	return name;
 }
 
 // On PC, we always use GL_DEPTH24_STENCIL8. 
@@ -1812,7 +1773,7 @@ void GLQueueRunner::fbo_unbind() {
 	glBindFramebuffer(GL_FRAMEBUFFER, g_defaultFBO);
 #endif
 
-#if PPSSPP_PLATFORM(IOS) && !defined(__LIBRETRO__)
+#if PPSSPP_PLATFORM(IOS)
 	bindDefaultFBO();
 #endif
 
@@ -1853,74 +1814,3 @@ GLRFramebuffer::~GLRFramebuffer() {
 		glDeleteRenderbuffers(1, &stencil_buffer);
 	CHECK_GL_ERROR_IF_DEBUG();
 }
-
-std::string GLQueueRunner::StepToString(const GLRStep &step) const {
-	char buffer[256];
-	switch (step.stepType) {
-	case GLRStepType::RENDER:
-	{
-		int w = step.render.framebuffer ? step.render.framebuffer->width : targetWidth_;
-		int h = step.render.framebuffer ? step.render.framebuffer->height : targetHeight_;
-		snprintf(buffer, sizeof(buffer), "RENDER %s %s (commands: %d, %dx%d)\n", step.tag, step.render.framebuffer ? step.render.framebuffer->Tag() : "", (int)step.commands.size(), w, h);
-		break;
-	}
-	case GLRStepType::COPY:
-		snprintf(buffer, sizeof(buffer), "COPY '%s' %s -> %s (%dx%d, %s)\n", step.tag, step.copy.src->Tag(), step.copy.dst->Tag(), step.copy.srcRect.w, step.copy.srcRect.h, GLRAspectToString((GLRAspect)step.copy.aspectMask));
-		break;
-	case GLRStepType::BLIT:
-		snprintf(buffer, sizeof(buffer), "BLIT '%s' %s -> %s (%dx%d->%dx%d, %s)\n", step.tag, step.copy.src->Tag(), step.copy.dst->Tag(), step.blit.srcRect.w, step.blit.srcRect.h, step.blit.dstRect.w, step.blit.dstRect.h, GLRAspectToString((GLRAspect)step.blit.aspectMask));
-		break;
-	case GLRStepType::READBACK:
-		snprintf(buffer, sizeof(buffer), "READBACK '%s' %s (%dx%d, %s)\n", step.tag, step.readback.src ? step.readback.src->Tag() : "(backbuffer)", step.readback.srcRect.w, step.readback.srcRect.h, GLRAspectToString((GLRAspect)step.readback.aspectMask));
-		break;
-	case GLRStepType::READBACK_IMAGE:
-		snprintf(buffer, sizeof(buffer), "READBACK_IMAGE '%s' (%dx%d)\n", step.tag, step.readback_image.srcRect.w, step.readback_image.srcRect.h);
-		break;
-	case GLRStepType::RENDER_SKIP:
-		snprintf(buffer, sizeof(buffer), "(RENDER_SKIP) %s\n", step.tag);
-		break;
-	default:
-		buffer[0] = 0;
-		break;
-	}
-	return std::string(buffer);
-}
-
-const char *GLRAspectToString(GLRAspect aspect) {
-	switch (aspect) {
-	case GLR_ASPECT_COLOR: return "COLOR";
-	case GLR_ASPECT_DEPTH: return "DEPTH";
-	case GLR_ASPECT_STENCIL: return "STENCIL";
-	default: return "N/A";
-	}
-}
-
-const char *RenderCommandToString(GLRRenderCommand cmd) {
-	switch (cmd) {
-	case GLRRenderCommand::DEPTH: return "DEPTH";
-	case GLRRenderCommand::STENCIL: return "STENCIL";
-	case GLRRenderCommand::BLEND: return "BLEND";
-	case GLRRenderCommand::BLENDCOLOR: return "BLENDCOLOR";
-	case GLRRenderCommand::LOGICOP: return "LOGICOP";
-	case GLRRenderCommand::UNIFORM4I: return "UNIFORM4I";
-	case GLRRenderCommand::UNIFORM4UI: return "UNIFORM4UI";
-	case GLRRenderCommand::UNIFORM4F: return "UNIFORM4F";
-	case GLRRenderCommand::UNIFORMMATRIX: return "UNIFORMMATRIX";
-	case GLRRenderCommand::UNIFORMSTEREOMATRIX: return "UNIFORMSTEREOMATRIX";
-	case GLRRenderCommand::TEXTURESAMPLER: return "TEXTURESAMPLER";
-	case GLRRenderCommand::TEXTURELOD: return "TEXTURELOD";
-	case GLRRenderCommand::VIEWPORT: return "VIEWPORT";
-	case GLRRenderCommand::SCISSOR: return "SCISSOR";
-	case GLRRenderCommand::RASTER: return "RASTER";
-	case GLRRenderCommand::CLEAR: return "CLEAR";
-	case GLRRenderCommand::INVALIDATE: return "INVALIDATE";
-	case GLRRenderCommand::BINDPROGRAM: return "BINDPROGRAM";
-	case GLRRenderCommand::BINDTEXTURE: return "BINDTEXTURE";
-	case GLRRenderCommand::BIND_FB_TEXTURE: return "BIND_FB_TEXTURE";
-	case GLRRenderCommand::BIND_VERTEX_BUFFER: return "BIND_VERTEX_BUFFER";
-	case GLRRenderCommand::GENMIPS: return "GENMIPS";
-	case GLRRenderCommand::DRAW: return "DRAW";
-	case GLRRenderCommand::TEXTURE_SUBIMAGE: return "TEXTURE_SUBIMAGE";
-	default: return "N/A";
-	}
-}

Common/GPU/OpenGL/GLQueueRunner.h

@@ -6,12 +6,11 @@
 #include <unordered_map>
 
 #include "Common/GPU/OpenGL/GLCommon.h"
-#include "Common/GPU/OpenGL/GLFrameData.h"
 #include "Common/GPU/DataFormat.h"
 #include "Common/GPU/Shader.h"
 #include "Common/GPU/thin3d.h"
 #include "Common/Data/Collections/TinySet.h"
-#include "Common/Data/Collections/FastVec.h"
+
 
 struct GLRViewport {
 	float x, y, w, h, minZ, maxZ;
@@ -25,7 +24,7 @@ struct GLOffset2D {
 	int x, y;
 };
 
-enum class GLRAllocType : uint8_t {
+enum class GLRAllocType {
 	NONE,
 	NEW,
 	ALIGNED,
@@ -40,7 +39,8 @@ class GLRInputLayout;
 
 enum class GLRRenderCommand : uint8_t {
 	DEPTH,
-	STENCIL,
+	STENCILFUNC,
+	STENCILOP,
 	BLEND,
 	BLENDCOLOR,
 	LOGICOP,
@@ -60,8 +60,10 @@ enum class GLRRenderCommand : uint8_t {
 	BINDTEXTURE,
 	BIND_FB_TEXTURE,
 	BIND_VERTEX_BUFFER,
+	BIND_BUFFER,
 	GENMIPS,
 	DRAW,
+	DRAW_INDEXED,
 	TEXTURE_SUBIMAGE,
 };
 
@@ -69,7 +71,6 @@ enum class GLRRenderCommand : uint8_t {
 // type field, smashed right after each other?)
 // Also, all GLenums are really only 16 bits.
 struct GLRRenderData {
-	GLRRenderData(GLRRenderCommand _cmd) : cmd(_cmd) {}
 	GLRRenderCommand cmd;
 	union {
 		struct {
@@ -99,23 +100,26 @@ struct GLRRenderData {
 			GLenum func;
 			uint8_t ref;
 			uint8_t compareMask;
+		} stencilFunc;
+		struct {
 			GLenum sFail;
 			GLenum zFail;
 			GLenum pass;
 			uint8_t writeMask;
-		} stencil;
+		} stencilOp;  // also write mask
 		struct {
-			GLRInputLayout *inputLayout;
-			GLRBuffer *vertexBuffer;
-			GLRBuffer *indexBuffer;
-			uint32_t vertexOffset;
-			uint32_t indexOffset;
 			GLenum mode;  // primitive
+			GLint buffer;
 			GLint first;
 			GLint count;
-			GLint indexType;
-			GLint instances;
 		} draw;
+		struct {
+			GLenum mode;  // primitive
+			GLint count;
+			GLint instances;
+			GLint indexType;
+			void *indices;
+		} drawIndexed;
 		struct {
 			const char *name;  // if null, use loc
 			const GLint *loc; // NOTE: This is a pointer so we can immediately use things that are "queried" during program creation.
@@ -125,13 +129,8 @@ struct GLRRenderData {
 		struct {
 			const char *name;  // if null, use loc
 			const GLint *loc;
-			float m[16];
+			float m[32];
 		} uniformMatrix4;
-		struct {
-			const char *name;  // if null, use loc
-			const GLint *loc;
-			float *mData;  // new'd, 32 entries
-		} uniformStereoMatrix4;
 		struct {
 			uint32_t clearColor;
 			float clearZ;
@@ -171,6 +170,11 @@ struct GLRRenderData {
 		struct {
 			GLRProgram *program;
 		} program;
+		struct {
+			GLRInputLayout *inputLayout;
+			GLRBuffer *buffer;
+			size_t offset;
+		} bindVertexBuffer;
 		struct {
 			int slot;
 			GLenum wrapS;
@@ -223,6 +227,7 @@ struct GLRInitStep {
 	union {
 		struct {
 			GLRTexture *texture;
+			GLenum target;
 		} create_texture;
 		struct {
 			GLRShader *shader;
@@ -290,17 +295,16 @@ enum class GLRRenderPassAction {
 
 class GLRFramebuffer;
 
-enum GLRAspect {
+enum {
 	GLR_ASPECT_COLOR = 1,
 	GLR_ASPECT_DEPTH = 2,
 	GLR_ASPECT_STENCIL = 3,
 };
-const char *GLRAspectToString(GLRAspect aspect);
 
 struct GLRStep {
 	GLRStep(GLRStepType _type) : stepType(_type) {}
 	GLRStepType stepType;
-	FastVec<GLRRenderData> commands;
+	std::vector<GLRRenderData> commands;
 	TinySet<const GLRFramebuffer *, 8> dependencies;
 	const char *tag;
 	union {
@@ -309,6 +313,8 @@ struct GLRStep {
 			GLRRenderPassAction color;
 			GLRRenderPassAction depth;
 			GLRRenderPassAction stencil;
+			// Note: not accurate.
+			int numDraws;
 		} render;
 		struct {
 			GLRFramebuffer *src;
@@ -352,14 +358,22 @@ public:
 		caps_ = caps;
 	}
 
-	void RunInitSteps(const FastVec<GLRInitStep> &steps, bool skipGLCalls);
+	int GetStereoBufferIndex(const char *uniformName);
+	std::string GetStereoBufferLayout(const char *uniformName);
 
-	void RunSteps(const std::vector<GLRStep *> &steps, GLFrameData &frameData, bool skipGLCalls, bool keepSteps, bool useVR);
+	void RunInitSteps(const std::vector<GLRInitStep> &steps, bool skipGLCalls);
+
+	void RunSteps(const std::vector<GLRStep *> &steps, bool skipGLCalls, bool keepSteps, bool useVR);
+	void LogSteps(const std::vector<GLRStep *> &steps);
 
 	void CreateDeviceObjects();
 	void DestroyDeviceObjects();
 
-	void CopyFromReadbackBuffer(GLRFramebuffer *framebuffer, int width, int height, Draw::DataFormat srcFormat, Draw::DataFormat destFormat, int pixelStride, uint8_t *pixels);
+	inline int RPIndex(GLRRenderPassAction color, GLRRenderPassAction depth) {
+		return (int)depth * 3 + (int)color;
+	}
+
+	void CopyReadbackBuffer(int width, int height, Draw::DataFormat srcFormat, Draw::DataFormat destFormat, int pixelStride, uint8_t *pixels);
 
 	void Resize(int width, int height) {
 		targetWidth_ = width;
@@ -379,7 +393,7 @@ private:
 	void InitCreateFramebuffer(const GLRInitStep &step);
 
 	void PerformBindFramebufferAsRenderTarget(const GLRStep &pass);
-	void PerformRenderPass(const GLRStep &pass, bool first, bool last, GLQueueProfileContext &profile);
+	void PerformRenderPass(const GLRStep &pass, bool first, bool last);
 	void PerformCopy(const GLRStep &pass);
 	void PerformBlit(const GLRStep &pass);
 	void PerformReadback(const GLRStep &pass);
@@ -390,8 +404,6 @@ private:
 	GLenum fbo_get_fb_target(bool read, GLuint **cached);
 	void fbo_unbind();
 
-	std::string StepToString(const GLRStep &step) const;
-
 	GLRFramebuffer *curFB_ = nullptr;
 
 	GLuint globalVAO_ = 0;
@@ -407,7 +419,9 @@ private:
 	// We size it generously.
 	uint8_t *readbackBuffer_ = nullptr;
 	int readbackBufferSize_ = 0;
-	uint32_t readbackAspectMask_ = 0;
+	// Temp buffer for color conversion
+	uint8_t *tempBuffer_ = nullptr;
+	int tempBufferSize_ = 0;
 
 	float maxAnisotropyLevel_ = 0.0f;
 
@@ -415,6 +429,10 @@ private:
 	GLuint currentDrawHandle_ = 0;
 	GLuint currentReadHandle_ = 0;
 
+	GLuint AllocTextureName();
+
+	// Texture name cache. Ripped straight from TextureCacheGLES.
+	std::vector<GLuint> nameCache_;
 	std::unordered_map<int, std::string> glStrings_;
 
 	bool sawOutOfMemory_ = false;
@@ -423,5 +441,3 @@ private:
 	ErrorCallbackFn errorCallback_ = nullptr;
 	void *errorCallbackUserData_ = nullptr;
 };
-
-const char *RenderCommandToString(GLRRenderCommand cmd);

Common/GPU/OpenGL/GLRenderManager.cpp

@@ -8,7 +8,6 @@
 #include "Common/Log.h"
 #include "Common/TimeUtil.h"
 #include "Common/MemoryUtil.h"
-#include "Common/StringUtils.h"
 #include "Common/Math/math_util.h"
 
 #if 0 // def _DEBUG
@@ -17,7 +16,12 @@
 #define VLOG(...)
 #endif
 
-std::thread::id renderThreadId;
+static std::thread::id renderThreadId;
+#if MAX_LOGLEVEL >= DEBUG_LEVEL
+static bool OnRenderThread() {
+	return std::this_thread::get_id() == renderThreadId;
+}
+#endif
 
 GLRTexture::GLRTexture(const Draw::DeviceCaps &caps, int width, int height, int depth, int numMips) {
 	if (caps.textureNPOTFullySupported) {
@@ -37,9 +41,76 @@ GLRTexture::~GLRTexture() {
 	}
 }
 
-GLRenderManager::GLRenderManager(HistoryBuffer<FrameTimeData, FRAME_TIME_HISTORY_LENGTH> &frameTimeHistory) : frameTimeHistory_(frameTimeHistory) {
+void GLDeleter::Take(GLDeleter &other) {
-	// size_t sz = sizeof(GLRRenderData);
+	_assert_msg_(IsEmpty(), "Deleter already has stuff");
-	// _dbg_assert_(sz == 88);
+	shaders = std::move(other.shaders);
+	programs = std::move(other.programs);
+	buffers = std::move(other.buffers);
+	textures = std::move(other.textures);
+	inputLayouts = std::move(other.inputLayouts);
+	framebuffers = std::move(other.framebuffers);
+	pushBuffers = std::move(other.pushBuffers);
+	other.shaders.clear();
+	other.programs.clear();
+	other.buffers.clear();
+	other.textures.clear();
+	other.inputLayouts.clear();
+	other.framebuffers.clear();
+	other.pushBuffers.clear();
+}
+
+// Runs on the GPU thread.
+void GLDeleter::Perform(GLRenderManager *renderManager, bool skipGLCalls) {
+	for (auto pushBuffer : pushBuffers) {
+		renderManager->UnregisterPushBuffer(pushBuffer);
+		if (skipGLCalls) {
+			pushBuffer->Destroy(false);
+		}
+		delete pushBuffer;
+	}
+	pushBuffers.clear();
+	for (auto shader : shaders) {
+		if (skipGLCalls)
+			shader->shader = 0;  // prevent the glDeleteShader
+		delete shader;
+	}
+	shaders.clear();
+	for (auto program : programs) {
+		if (skipGLCalls)
+			program->program = 0;  // prevent the glDeleteProgram
+		delete program;
+	}
+	programs.clear();
+	for (auto buffer : buffers) {
+		if (skipGLCalls)
+			buffer->buffer_ = 0;
+		delete buffer;
+	}
+	buffers.clear();
+	for (auto texture : textures) {
+		if (skipGLCalls)
+			texture->texture = 0;
+		delete texture;
+	}
+	textures.clear();
+	for (auto inputLayout : inputLayouts) {
+		// No GL objects in an inputLayout yet
+		delete inputLayout;
+	}
+	inputLayouts.clear();
+	for (auto framebuffer : framebuffers) {
+		if (skipGLCalls) {
+			framebuffer->handle = 0;
+			framebuffer->color_texture.texture = 0;
+			framebuffer->z_stencil_buffer = 0;
+			framebuffer->z_stencil_texture.texture = 0;
+			framebuffer->z_buffer = 0;
+			framebuffer->stencil_buffer = 0;
+			framebuffer->colorFormat = Draw::DataFormat::UNDEFINED;
+		}
+		delete framebuffer;
+	}
+	framebuffers.clear();
 }
 
 GLRenderManager::~GLRenderManager() {
@@ -129,25 +200,25 @@ bool GLRenderManager::ThreadFrame() {
 		return false;
 	}
 
-	GLRRenderThreadTask *task = nullptr;
+	GLRRenderThreadTask task;
 
 	// In case of syncs or other partial completion, we keep going until we complete a frame.
 	while (true) {
 		// Pop a task of the queue and execute it.
 		// NOTE: We need to actually wait for a task, we can't just bail!
+
 		{
 			std::unique_lock<std::mutex> lock(pushMutex_);
 			while (renderThreadQueue_.empty()) {
 				pushCondVar_.wait(lock);
 			}
-			task = std::move(renderThreadQueue_.front());
+			task = renderThreadQueue_.front();
 			renderThreadQueue_.pop();
 		}
 
 		// We got a task! We can now have pushMutex_ unlocked, allowing the host to
 		// push more work when it feels like it, and just start working.
-		if (task->runType == GLRRunType::EXIT) {
+		if (task.runType == GLRRunType::EXIT) {
-			delete task;
 			// Oh, host wanted out. Let's leave, and also let's notify the host.
 			// This is unlike Vulkan too which can just block on the thread existing.
 			std::unique_lock<std::mutex> lock(syncMutex_);
@@ -157,13 +228,11 @@ bool GLRenderManager::ThreadFrame() {
 		}
 
 		// Render the scene.
-		VLOG("  PULL: Frame %d RUN (%0.3f)", task->frame, time_now_d());
+		VLOG("  PULL: Frame %d RUN (%0.3f)", task.frame, time_now_d());
-		if (Run(*task)) {
+		if (Run(task)) {
 			// Swap requested, so we just bail the loop.
-			delete task;
 			break;
 		}
-		delete task;
 	};
 
 	return true;
@@ -176,21 +245,15 @@ void GLRenderManager::StopThread() {
 		run_ = false;
 
 		std::unique_lock<std::mutex> lock(pushMutex_);
-		renderThreadQueue_.push(new GLRRenderThreadTask(GLRRunType::EXIT));
+		GLRRenderThreadTask exitTask{};
+		exitTask.runType = GLRRunType::EXIT;
+		renderThreadQueue_.push(exitTask);
 		pushCondVar_.notify_one();
 	} else {
 		WARN_LOG(G3D, "GL submission thread was already paused.");
 	}
 }
 
-std::string GLRenderManager::GetGpuProfileString() const {
-	int curFrame = curFrame_;
-	const GLQueueProfileContext &profile = frameData_[curFrame].profile;
-
-	float cputime_ms = 1000.0f * (profile.cpuEndTime - profile.cpuStartTime);
-	return StringFromFormat("CPU time to run the list: %0.2f ms\n\n%s", cputime_ms, profilePassesString_.c_str());
-}
-
 void GLRenderManager::BindFramebufferAsRenderTarget(GLRFramebuffer *fb, GLRRenderPassAction color, GLRRenderPassAction depth, GLRRenderPassAction stencil, uint32_t clearColor, float clearDepth, uint8_t clearStencil, const char *tag) {
 	_assert_(insideFrame_);
 #ifdef _DEBUG
@@ -198,7 +261,7 @@ void GLRenderManager::BindFramebufferAsRenderTarget(GLRFramebuffer *fb, GLRRende
 #endif
 
 	// Eliminate dupes.
-	if (steps_.size() && steps_.back()->stepType == GLRStepType::RENDER && steps_.back()->render.framebuffer == fb) {
+	if (steps_.size() && steps_.back()->render.framebuffer == fb && steps_.back()->stepType == GLRStepType::RENDER) {
 		if (color != GLRRenderPassAction::CLEAR && depth != GLRRenderPassAction::CLEAR && stencil != GLRRenderPassAction::CLEAR) {
 			// We don't move to a new step, this bind was unnecessary and we can safely skip it.
 			curRenderStep_ = steps_.back();
@@ -206,7 +269,7 @@ void GLRenderManager::BindFramebufferAsRenderTarget(GLRFramebuffer *fb, GLRRende
 		}
 	}
 	if (curRenderStep_ && curRenderStep_->commands.size() == 0) {
-		VLOG("Empty render step. Usually happens after uploading pixels.");
+		VLOG("Empty render step. Usually happens after uploading pixels..");
 	}
 
 	GLRStep *step = new GLRStep{ GLRStepType::RENDER };
@@ -215,11 +278,13 @@ void GLRenderManager::BindFramebufferAsRenderTarget(GLRFramebuffer *fb, GLRRende
 	step->render.color = color;
 	step->render.depth = depth;
 	step->render.stencil = stencil;
+	step->render.numDraws = 0;
 	step->tag = tag;
 	steps_.push_back(step);
 
 	GLuint clearMask = 0;
-	GLRRenderData data(GLRRenderCommand::CLEAR);
+	GLRRenderData data;
+	data.cmd = GLRRenderCommand::CLEAR;
 	if (color == GLRRenderPassAction::CLEAR) {
 		clearMask |= GL_COLOR_BUFFER_BIT;
 		data.clear.clearColor = clearColor;
@@ -296,7 +361,7 @@ void GLRenderManager::BlitFramebuffer(GLRFramebuffer *src, GLRect2D srcRect, GLR
 	steps_.push_back(step);
 }
 
-bool GLRenderManager::CopyFramebufferToMemory(GLRFramebuffer *src, int aspectBits, int x, int y, int w, int h, Draw::DataFormat destFormat, uint8_t *pixels, int pixelStride, Draw::ReadbackMode mode, const char *tag) {
+bool GLRenderManager::CopyFramebufferToMemorySync(GLRFramebuffer *src, int aspectBits, int x, int y, int w, int h, Draw::DataFormat destFormat, uint8_t *pixels, int pixelStride, const char *tag) {
 	_assert_(pixels);
 
 	GLRStep *step = new GLRStep{ GLRStepType::READBACK };
@@ -323,7 +388,7 @@ bool GLRenderManager::CopyFramebufferToMemory(GLRFramebuffer *src, int aspectBit
 	} else {
 		return false;
 	}
-	queueRunner_.CopyFromReadbackBuffer(src, w, h, srcFormat, destFormat, pixelStride, pixels);
+	queueRunner_.CopyReadbackBuffer(w, h, srcFormat, destFormat, pixelStride, pixels);
 	return true;
 }
 
@@ -340,28 +405,20 @@ void GLRenderManager::CopyImageToMemorySync(GLRTexture *texture, int mipLevel, i
 	curRenderStep_ = nullptr;
 	FlushSync();
 
-	queueRunner_.CopyFromReadbackBuffer(nullptr, w, h, Draw::DataFormat::R8G8B8A8_UNORM, destFormat, pixelStride, pixels);
+	queueRunner_.CopyReadbackBuffer(w, h, Draw::DataFormat::R8G8B8A8_UNORM, destFormat, pixelStride, pixels);
 }
 
-void GLRenderManager::BeginFrame(bool enableProfiling) {
+void GLRenderManager::BeginFrame() {
 #ifdef _DEBUG
 	curProgram_ = nullptr;
 #endif
 
 	int curFrame = GetCurFrame();
 
-	FrameTimeData &frameTimeData = frameTimeHistory_.Add(frameIdGen_);
+	FrameData &frameData = frameData_[curFrame];
-	frameTimeData.frameBegin = time_now_d();
-	frameTimeData.afterFenceWait = frameTimeData.frameBegin;
-
-	GLFrameData &frameData = frameData_[curFrame];
-	frameData.frameId = frameIdGen_;
-	frameData.profile.enabled = enableProfiling;
-
-	frameIdGen_++;
 	{
-		std::unique_lock<std::mutex> lock(frameData.fenceMutex);
 		VLOG("PUSH: BeginFrame (curFrame = %d, readyForFence = %d, time=%0.3f)", curFrame, (int)frameData.readyForFence, time_now_d());
+		std::unique_lock<std::mutex> lock(frameData.fenceMutex);
 		while (!frameData.readyForFence) {
 			frameData.fenceCondVar.wait(lock);
 		}
@@ -378,99 +435,36 @@ void GLRenderManager::BeginFrame(bool enableProfiling) {
 void GLRenderManager::Finish() {
 	curRenderStep_ = nullptr;  // EndCurRenderStep is this simple here.
 
-	int curFrame = curFrame_;
+	int curFrame = GetCurFrame();
-	GLFrameData &frameData = frameData_[curFrame];
+	FrameData &frameData = frameData_[curFrame];
-
-	frameTimeHistory_[frameData.frameId].firstSubmit = time_now_d();
 
 	frameData_[curFrame].deleter.Take(deleter_);
 
-	if (frameData.profile.enabled) {
-		profilePassesString_ = std::move(frameData.profile.passesString);
-
-#ifdef _DEBUG
-		std::string cmdString;
-		for (int i = 0; i < ARRAY_SIZE(frameData.profile.commandCounts); i++) {
-			if (frameData.profile.commandCounts[i] > 0) {
-				cmdString += StringFromFormat("%s: %d\n", RenderCommandToString((GLRRenderCommand)i), frameData.profile.commandCounts[i]);
-			}
-		}
-		memset(frameData.profile.commandCounts, 0, sizeof(frameData.profile.commandCounts));
-		profilePassesString_ = cmdString + profilePassesString_;
-#endif
-
-		frameData.profile.passesString.clear();
-	}
-
 	VLOG("PUSH: Finish, pushing task. curFrame = %d", curFrame);
-	GLRRenderThreadTask *task = new GLRRenderThreadTask(GLRRunType::SUBMIT);
+	GLRRenderThreadTask task;
-	task->frame = curFrame;
+	task.frame = curFrame;
+	task.runType = GLRRunType::PRESENT;
+
 	{
 		std::unique_lock<std::mutex> lock(pushMutex_);
 		renderThreadQueue_.push(task);
-		renderThreadQueue_.back()->initSteps = std::move(initSteps_);
+		renderThreadQueue_.back().initSteps = std::move(initSteps_);
-		renderThreadQueue_.back()->steps = std::move(steps_);
+		renderThreadQueue_.back().steps = std::move(steps_);
 		initSteps_.clear();
 		steps_.clear();
 		pushCondVar_.notify_one();
 	}
-}
 
-void GLRenderManager::Present() {
+	curFrame_++;
-	GLRRenderThreadTask *presentTask = new GLRRenderThreadTask(GLRRunType::PRESENT);
+	if (curFrame_ >= inflightFrames_)
-	presentTask->frame = curFrame_;
+		curFrame_ = 0;
-	{
-		std::unique_lock<std::mutex> lock(pushMutex_);
-		renderThreadQueue_.push(presentTask);
-		pushCondVar_.notify_one();
-	}
-
-	int newCurFrame = curFrame_ + 1;
-	if (newCurFrame >= inflightFrames_) {
-		newCurFrame = 0;
-	}
-	curFrame_ = newCurFrame;
 
 	insideFrame_ = false;
 }
 
 // Render thread. Returns true if the caller should handle a swap.
 bool GLRenderManager::Run(GLRRenderThreadTask &task) {
-	_dbg_assert_(task.frame >= 0);
+	FrameData &frameData = frameData_[task.frame];
-
-	GLFrameData &frameData = frameData_[task.frame];
-
-	if (task.runType == GLRRunType::PRESENT) {
-		bool swapRequest = false;
-		if (!frameData.skipSwap) {
-			frameTimeHistory_[frameData.frameId].queuePresent = time_now_d();
-			if (swapIntervalChanged_) {
-				swapIntervalChanged_ = false;
-				if (swapIntervalFunction_) {
-					swapIntervalFunction_(swapInterval_);
-				}
-			}
-			// This is the swapchain framebuffer flip.
-			if (swapFunction_) {
-				VLOG("  PULL: SwapFunction()");
-				swapFunction_();
-			}
-			swapRequest = true;
-		} else {
-			frameData.skipSwap = false;
-		}
-		frameData.hasBegun = false;
-
-		VLOG("  PULL: Frame %d.readyForFence = true", task.frame);
-
-		{
-			std::lock_guard<std::mutex> lock(frameData.fenceMutex);
-			frameData.readyForFence = true;
-			frameData.fenceCondVar.notify_one();
-			// At this point, we're done with this framedata (for now).
-		}
-		return swapRequest;
-	}
 
 	if (!frameData.hasBegun) {
 		frameData.hasBegun = true;
@@ -490,23 +484,15 @@ bool GLRenderManager::Run(GLRRenderThreadTask &task) {
 		}
 	}
 
-	if (frameData.profile.enabled) {
-		frameData.profile.cpuStartTime = time_now_d();
-	}
-
 	if (IsVREnabled()) {
 		int passes = GetVRPassesCount();
 		for (int i = 0; i < passes; i++) {
 			PreVRFrameRender(i);
-			queueRunner_.RunSteps(task.steps, frameData, skipGLCalls_, i < passes - 1, true);
+			queueRunner_.RunSteps(task.steps, skipGLCalls_, i < passes - 1, true);
 			PostVRFrameRender();
 		}
 	} else {
-		queueRunner_.RunSteps(task.steps, frameData, skipGLCalls_, false, false);
+		queueRunner_.RunSteps(task.steps, skipGLCalls_, false, false);
-	}
-
-	if (frameData.profile.enabled) {
-		frameData.profile.cpuEndTime = time_now_d();
 	}
 
 	if (!skipGLCalls_) {
@@ -515,8 +501,43 @@ bool GLRenderManager::Run(GLRRenderThreadTask &task) {
 		}
 	}
 
+	bool swapRequest = false;
+
 	switch (task.runType) {
-	case GLRRunType::SUBMIT:
+	case GLRRunType::PRESENT:
+		if (!frameData.skipSwap) {
+			if (swapIntervalChanged_) {
+				swapIntervalChanged_ = false;
+				if (swapIntervalFunction_) {
+					swapIntervalFunction_(swapInterval_);
+				}
+			}
+			// This is the swapchain framebuffer flip.
+			if (swapFunction_) {
+				VLOG("  PULL: SwapFunction()");
+				swapFunction_();
+				if (!retainControl_) {
+					// get out of here.
+					swapRequest = true;
+				}
+			} else {
+				VLOG("  PULL: SwapRequested");
+				swapRequest = true;
+			}
+		} else {
+			frameData.skipSwap = false;
+		}
+		frameData.hasBegun = false;
+
+		VLOG("  PULL: Frame %d.readyForFence = true", task.frame);
+
+		{
+			std::lock_guard<std::mutex> lock(frameData.fenceMutex);
+			frameData.readyForFence = true;
+			frameData.fenceCondVar.notify_one();
+			// At this point, we're done with this framedata (for now).
+		}
+
 		break;
 
 	case GLRRunType::SYNC:
@@ -525,7 +546,7 @@ bool GLRenderManager::Run(GLRRenderThreadTask &task) {
 		// glFinish is not actually necessary here, and won't be unless we start using
 		// glBufferStorage. Then we need to use fences.
 		{
-			std::lock_guard<std::mutex> lock(syncMutex_);
+			std::unique_lock<std::mutex> lock(syncMutex_);
 			syncDone_ = true;
 			syncCondVar_.notify_one();
 		}
@@ -535,20 +556,21 @@ bool GLRenderManager::Run(GLRRenderThreadTask &task) {
 		_assert_(false);
 	}
 	VLOG("  PULL: ::Run(): Done running tasks");
-	return false;
+	return swapRequest;
 }
 
 void GLRenderManager::FlushSync() {
 	{
 		VLOG("PUSH: Frame[%d].readyForRun = true (sync)", curFrame_);
 
-		GLRRenderThreadTask *task = new GLRRenderThreadTask(GLRRunType::SYNC);
+		GLRRenderThreadTask task;
-		task->frame = curFrame_;
+		task.frame = curFrame_;
+		task.runType = GLRRunType::SYNC;
 
 		std::unique_lock<std::mutex> lock(pushMutex_);
 		renderThreadQueue_.push(task);
-		renderThreadQueue_.back()->initSteps = std::move(initSteps_);
+		renderThreadQueue_.back().initSteps = std::move(initSteps_);
-		renderThreadQueue_.back()->steps = std::move(steps_);
+		renderThreadQueue_.back().steps = std::move(steps_);
 		pushCondVar_.notify_one();
 		steps_.clear();
 	}
@@ -563,3 +585,254 @@ void GLRenderManager::FlushSync() {
 		syncDone_ = false;
 	}
 }
+
+GLPushBuffer::GLPushBuffer(GLRenderManager *render, GLuint target, size_t size) : render_(render), size_(size), target_(target) {
+	bool res = AddBuffer();
+	_assert_(res);
+}
+
+GLPushBuffer::~GLPushBuffer() {
+	Destroy(true);
+}
+
+void GLPushBuffer::Map() {
+	_assert_(!writePtr_);
+	auto &info = buffers_[buf_];
+	writePtr_ = info.deviceMemory ? info.deviceMemory : info.localMemory;
+	info.flushOffset = 0;
+	// Force alignment.  This is needed for PushAligned() to work as expected.
+	while ((intptr_t)writePtr_ & 15) {
+		writePtr_++;
+		offset_++;
+		info.flushOffset++;
+	}
+	_assert_(writePtr_);
+}
+
+void GLPushBuffer::Unmap() {
+	_assert_(writePtr_);
+	if (!buffers_[buf_].deviceMemory) {
+		// Here we simply upload the data to the last buffer.
+		// Might be worth trying with size_ instead of offset_, so the driver can replace
+		// the whole buffer. At least if it's close.
+		render_->BufferSubdata(buffers_[buf_].buffer, 0, offset_, buffers_[buf_].localMemory, false);
+	} else {
+		buffers_[buf_].flushOffset = offset_;
+	}
+	writePtr_ = nullptr;
+}
+
+void GLPushBuffer::Flush() {
+	// Must be called from the render thread.
+	_dbg_assert_(OnRenderThread());
+
+	buffers_[buf_].flushOffset = offset_;
+	if (!buffers_[buf_].deviceMemory && writePtr_) {
+		auto &info = buffers_[buf_];
+		if (info.flushOffset != 0) {
+			_assert_(info.buffer->buffer_);
+			glBindBuffer(target_, info.buffer->buffer_);
+			glBufferSubData(target_, 0, info.flushOffset, info.localMemory);
+		}
+
+		// Here we will submit all the draw calls, with the already known buffer and offsets.
+		// Might as well reset the write pointer here and start over the current buffer.
+		writePtr_ = info.localMemory;
+		offset_ = 0;
+		info.flushOffset = 0;
+	}
+
+	// For device memory, we flush all buffers here.
+	if ((strategy_ & GLBufferStrategy::MASK_FLUSH) != 0) {
+		for (auto &info : buffers_) {
+			if (info.flushOffset == 0 || !info.deviceMemory)
+				continue;
+
+			glBindBuffer(target_, info.buffer->buffer_);
+			glFlushMappedBufferRange(target_, 0, info.flushOffset);
+			info.flushOffset = 0;
+		}
+	}
+}
+
+bool GLPushBuffer::AddBuffer() {
+	BufInfo info;
+	info.localMemory = (uint8_t *)AllocateAlignedMemory(size_, 16);
+	if (!info.localMemory)
+		return false;
+	info.buffer = render_->CreateBuffer(target_, size_, GL_DYNAMIC_DRAW);
+	buf_ = buffers_.size();
+	buffers_.push_back(info);
+	return true;
+}
+
+void GLPushBuffer::Destroy(bool onRenderThread) {
+	if (buf_ == -1)
+		return;  // Already destroyed
+	for (BufInfo &info : buffers_) {
+		// This will automatically unmap device memory, if needed.
+		// NOTE: We immediately delete the buffer, don't go through the deleter, if we're on the render thread.
+		if (onRenderThread) {
+			delete info.buffer;
+		} else {
+			render_->DeleteBuffer(info.buffer);
+		}
+
+		FreeAlignedMemory(info.localMemory);
+	}
+	buffers_.clear();
+	buf_ = -1;
+}
+
+void GLPushBuffer::NextBuffer(size_t minSize) {
+	// First, unmap the current memory.
+	Unmap();
+
+	buf_++;
+	if (buf_ >= buffers_.size() || minSize > size_) {
+		// Before creating the buffer, adjust to the new size_ if necessary.
+		while (size_ < minSize) {
+			size_ <<= 1;
+		}
+
+		bool res = AddBuffer();
+		_assert_(res);
+		if (!res) {
+			// Let's try not to crash at least?
+			buf_ = 0;
+		}
+	}
+
+	// Now, move to the next buffer and map it.
+	offset_ = 0;
+	Map();
+}
+
+void GLPushBuffer::Defragment() {
+	_dbg_assert_msg_(!OnRenderThread(), "Defragment must not run on the render thread");
+
+	if (buffers_.size() <= 1) {
+		// Let's take this chance to jetison localMemory we don't need.
+		for (auto &info : buffers_) {
+			if (info.deviceMemory) {
+				FreeAlignedMemory(info.localMemory);
+				info.localMemory = nullptr;
+			}
+		}
+
+		return;
+	}
+
+	// Okay, we have more than one.  Destroy them all and start over with a larger one.
+	size_t newSize = size_ * buffers_.size();
+	Destroy(false);
+
+	size_ = newSize;
+	bool res = AddBuffer();
+	_assert_msg_(res, "AddBuffer failed");
+}
+
+size_t GLPushBuffer::GetTotalSize() const {
+	size_t sum = 0;
+	if (buffers_.size() > 1)
+		sum += size_ * (buffers_.size() - 1);
+	sum += offset_;
+	return sum;
+}
+
+void GLPushBuffer::MapDevice(GLBufferStrategy strategy) {
+	_dbg_assert_msg_(OnRenderThread(), "MapDevice must run on render thread");
+
+	strategy_ = strategy;
+	if (strategy_ == GLBufferStrategy::SUBDATA) {
+		return;
+	}
+
+	bool mapChanged = false;
+	for (auto &info : buffers_) {
+		if (!info.buffer->buffer_ || info.deviceMemory) {
+			// Can't map - no device buffer associated yet or already mapped.
+			continue;
+		}
+
+		info.deviceMemory = (uint8_t *)info.buffer->Map(strategy_);
+		mapChanged = mapChanged || info.deviceMemory != nullptr;
+
+		if (!info.deviceMemory && !info.localMemory) {
+			// Somehow it failed, let's dodge crashing.
+			info.localMemory = (uint8_t *)AllocateAlignedMemory(info.buffer->size_, 16);
+			mapChanged = true;
+		}
+
+		_dbg_assert_msg_(info.localMemory || info.deviceMemory, "Local or device memory must succeed");
+	}
+
+	if (writePtr_ && mapChanged) {
+		// This can happen during a sync.  Remap.
+		writePtr_ = nullptr;
+		Map();
+	}
+}
+
+void GLPushBuffer::UnmapDevice() {
+	_dbg_assert_msg_(OnRenderThread(), "UnmapDevice must run on render thread");
+
+	for (auto &info : buffers_) {
+		if (info.deviceMemory) {
+			// TODO: Technically this can return false?
+			info.buffer->Unmap();
+			info.deviceMemory = nullptr;
+		}
+	}
+}
+
+void *GLRBuffer::Map(GLBufferStrategy strategy) {
+	_assert_(buffer_ != 0);
+
+	GLbitfield access = GL_MAP_WRITE_BIT;
+	if ((strategy & GLBufferStrategy::MASK_FLUSH) != 0) {
+		access |= GL_MAP_FLUSH_EXPLICIT_BIT;
+	}
+	if ((strategy & GLBufferStrategy::MASK_INVALIDATE) != 0) {
+		access |= GL_MAP_INVALIDATE_BUFFER_BIT;
+	}
+
+	void *p = nullptr;
+	bool allowNativeBuffer = strategy != GLBufferStrategy::SUBDATA;
+	if (allowNativeBuffer) {
+		glBindBuffer(target_, buffer_);
+
+		if (gl_extensions.ARB_buffer_storage || gl_extensions.EXT_buffer_storage) {
+#if !PPSSPP_PLATFORM(IOS)
+			if (!hasStorage_) {
+				GLbitfield storageFlags = access & ~(GL_MAP_INVALIDATE_BUFFER_BIT | GL_MAP_FLUSH_EXPLICIT_BIT);
+#ifdef USING_GLES2
+#ifdef GL_EXT_buffer_storage
+				glBufferStorageEXT(target_, size_, nullptr, storageFlags);
+#endif
+#else
+				glBufferStorage(target_, size_, nullptr, storageFlags);
+#endif
+				hasStorage_ = true;
+			}
+#endif
+			p = glMapBufferRange(target_, 0, size_, access);
+		} else if (gl_extensions.VersionGEThan(3, 0, 0)) {
+			// GLES3 or desktop 3.
+			p = glMapBufferRange(target_, 0, size_, access);
+		} else if (!gl_extensions.IsGLES) {
+#ifndef USING_GLES2
+			p = glMapBuffer(target_, GL_READ_WRITE);
+#endif
+		}
+	}
+
+	mapped_ = p != nullptr;
+	return p;
+}
+
+bool GLRBuffer::Unmap() {
+	glBindBuffer(target_, buffer_);
+	mapped_ = false;
+	return glUnmapBuffer(target_) == GL_TRUE;
+}

Common/GPU/OpenGL/GLRenderManager.h

@@ -10,13 +10,11 @@
 #include <queue>
 #include <condition_variable>
 
+#include "Common/GPU/OpenGL/GLCommon.h"
 #include "Common/GPU/MiscTypes.h"
 #include "Common/Data/Convert/SmallDataConvert.h"
 #include "Common/Log.h"
-#include "Common/GPU/OpenGL/GLQueueRunner.h"
+#include "GLQueueRunner.h"
-#include "Common/GPU/OpenGL/GLFrameData.h"
-#include "Common/GPU/OpenGL/GLCommon.h"
-#include "Common/GPU/OpenGL/GLMemory.h"
 
 class GLRInputLayout;
 class GLPushBuffer;
@@ -53,13 +51,11 @@ public:
 
 class GLRFramebuffer {
 public:
-	GLRFramebuffer(const Draw::DeviceCaps &caps, int _width, int _height, bool z_stencil, const char *tag)
+	GLRFramebuffer(const Draw::DeviceCaps &caps, Draw::DataFormat _colorFormat, int _width, int _height, bool z_stencil)
 		: color_texture(caps, _width, _height, 1, 1), z_stencil_texture(caps, _width, _height, 1, 1),
-		width(_width), height(_height), z_stencil_(z_stencil) {
+		colorFormat(_colorFormat), width(_width), height(_height), z_stencil_(z_stencil) {}
-	}
-	~GLRFramebuffer();
 
-	const char *Tag() const { return tag_.c_str(); }
+	~GLRFramebuffer();
 
 	GLuint handle = 0;
 	GLRTexture color_texture;
@@ -68,14 +64,13 @@ public:
 	GLRTexture z_stencil_texture;
 	GLuint z_buffer = 0;
 	GLuint stencil_buffer = 0;
+	Draw::DataFormat colorFormat;
 
 	int width;
 	int height;
 	GLuint colorDepth = 0;
-	bool z_stencil_;
 
-private:
+	bool z_stencil_;
-	std::string tag_;
 };
 
 // We need to create some custom heap-allocated types so we can forward things that need to be created on the GL thread, before
@@ -183,6 +178,178 @@ private:
 	std::unordered_map<std::string, UniformInfo> uniformCache_;
 };
 
+enum class GLBufferStrategy {
+	SUBDATA = 0,
+
+	MASK_FLUSH = 0x10,
+	MASK_INVALIDATE = 0x20,
+
+	// Map/unmap the buffer each frame.
+	FRAME_UNMAP = 1,
+	// Map/unmap and also invalidate the buffer on map.
+	INVALIDATE_UNMAP = MASK_INVALIDATE,
+	// Map/unmap and explicitly flushed changed ranges.
+	FLUSH_UNMAP = MASK_FLUSH,
+	// Map/unmap, invalidate on map, and explicit flush.
+	FLUSH_INVALIDATE_UNMAP = MASK_FLUSH | MASK_INVALIDATE,
+};
+
+static inline int operator &(const GLBufferStrategy &lhs, const GLBufferStrategy &rhs) {
+	return (int)lhs & (int)rhs;
+}
+
+class GLRBuffer {
+public:
+	GLRBuffer(GLuint target, size_t size) : target_(target), size_((int)size) {}
+	~GLRBuffer() {
+		if (buffer_) {
+			glDeleteBuffers(1, &buffer_);
+		}
+	}
+
+	void *Map(GLBufferStrategy strategy);
+	bool Unmap();
+
+	bool Mapped() const {
+		return mapped_;
+	}
+
+	GLuint buffer_ = 0;
+	GLuint target_;
+	int size_;
+
+private:
+	bool mapped_ = false;
+	bool hasStorage_ = false;
+};
+
+class GLRenderManager;
+
+// Similar to VulkanPushBuffer but is currently less efficient - it collects all the data in
+// RAM then does a big memcpy/buffer upload at the end of the frame. This is at least a lot
+// faster than the hundreds of buffer uploads or memory array buffers we used before.
+// On modern GL we could avoid the copy using glBufferStorage but not sure it's worth the
+// trouble.
+// We need to manage the lifetime of this together with the other resources so its destructor
+// runs on the render thread.
+class GLPushBuffer {
+public:
+	friend class GLRenderManager;
+
+	struct BufInfo {
+		GLRBuffer *buffer = nullptr;
+		uint8_t *localMemory = nullptr;
+		uint8_t *deviceMemory = nullptr;
+		size_t flushOffset = 0;
+	};
+
+	GLPushBuffer(GLRenderManager *render, GLuint target, size_t size);
+	~GLPushBuffer();
+
+	void Reset() { offset_ = 0; }
+
+private:
+	// Needs context in case of defragment.
+	void Begin() {
+		buf_ = 0;
+		offset_ = 0;
+		// Note: we must defrag because some buffers may be smaller than size_.
+		Defragment();
+		Map();
+		_dbg_assert_(writePtr_);
+	}
+
+	void BeginNoReset() {
+		Map();
+	}
+
+	void End() {
+		Unmap();
+	}
+
+public:
+	void Map();
+	void Unmap();
+
+	bool IsReady() const {
+		return writePtr_ != nullptr;
+	}
+
+	// When using the returned memory, make sure to bind the returned vkbuf.
+	// This will later allow for handling overflow correctly.
+	size_t Allocate(size_t numBytes, GLRBuffer **vkbuf) {
+		size_t out = offset_;
+		if (offset_ + ((numBytes + 3) & ~3) >= size_) {
+			NextBuffer(numBytes);
+			out = offset_;
+			offset_ += (numBytes + 3) & ~3;
+		} else {
+			offset_ += (numBytes + 3) & ~3;  // Round up to 4 bytes.
+		}
+		*vkbuf = buffers_[buf_].buffer;
+		return out;
+	}
+
+	// Returns the offset that should be used when binding this buffer to get this data.
+	size_t Push(const void *data, size_t size, GLRBuffer **vkbuf) {
+		_dbg_assert_(writePtr_);
+		size_t off = Allocate(size, vkbuf);
+		memcpy(writePtr_ + off, data, size);
+		return off;
+	}
+
+	uint32_t PushAligned(const void *data, size_t size, int align, GLRBuffer **vkbuf) {
+		_dbg_assert_(writePtr_);
+		offset_ = (offset_ + align - 1) & ~(align - 1);
+		size_t off = Allocate(size, vkbuf);
+		memcpy(writePtr_ + off, data, size);
+		return (uint32_t)off;
+	}
+
+	size_t GetOffset() const {
+		return offset_;
+	}
+
+	// "Zero-copy" variant - you can write the data directly as you compute it.
+	// Recommended.
+	void *Push(size_t size, uint32_t *bindOffset, GLRBuffer **vkbuf) {
+		_dbg_assert_(writePtr_);
+		size_t off = Allocate(size, vkbuf);
+		*bindOffset = (uint32_t)off;
+		return writePtr_ + off;
+	}
+	void *PushAligned(size_t size, uint32_t *bindOffset, GLRBuffer **vkbuf, int align) {
+		_dbg_assert_(writePtr_);
+		offset_ = (offset_ + align - 1) & ~(align - 1);
+		size_t off = Allocate(size, vkbuf);
+		*bindOffset = (uint32_t)off;
+		return writePtr_ + off;
+	}
+
+	size_t GetTotalSize() const;
+
+	void Destroy(bool onRenderThread);
+	void Flush();
+
+protected:
+	void MapDevice(GLBufferStrategy strategy);
+	void UnmapDevice();
+
+private:
+	bool AddBuffer();
+	void NextBuffer(size_t minSize);
+	void Defragment();
+
+	GLRenderManager *render_;
+	std::vector<BufInfo> buffers_;
+	size_t buf_ = 0;
+	size_t offset_ = 0;
+	size_t size_ = 0;
+	uint8_t *writePtr_ = nullptr;
+	GLuint target_;
+	GLBufferStrategy strategy_ = GLBufferStrategy::SUBDATA;
+};
+
 class GLRInputLayout {
 public:
 	struct Entry {
@@ -190,15 +357,14 @@ public:
 		int count;
 		GLenum type;
 		GLboolean normalized;
+		int stride;
 		intptr_t offset;
 	};
 	std::vector<Entry> entries;
-	int stride;
 	int semanticsMask_ = 0;
 };
 
 enum class GLRRunType {
-	SUBMIT,
 	PRESENT,
 	SYNC,
 	EXIT,
@@ -207,19 +373,33 @@ enum class GLRRunType {
 class GLRenderManager;
 class GLPushBuffer;
 
-// These are enqueued from the main thread, and the render thread pops them off
+class GLDeleter {
+public:
+	void Perform(GLRenderManager *renderManager, bool skipGLCalls);
+
+	bool IsEmpty() const {
+		return shaders.empty() && programs.empty() && buffers.empty() && textures.empty() && inputLayouts.empty() && framebuffers.empty() && pushBuffers.empty();
+	}
+
+	void Take(GLDeleter &other);
+
+	std::vector<GLRShader *> shaders;
+	std::vector<GLRProgram *> programs;
+	std::vector<GLRBuffer *> buffers;
+	std::vector<GLRTexture *> textures;
+	std::vector<GLRInputLayout *> inputLayouts;
+	std::vector<GLRFramebuffer *> framebuffers;
+	std::vector<GLPushBuffer *> pushBuffers;
+};
+
+// These are enqueued from the main thread,
+// and the render thread pops them off
 struct GLRRenderThreadTask {
-	GLRRenderThreadTask(GLRRunType _runType) : runType(_runType) {}
-
 	std::vector<GLRStep *> steps;
-	FastVec<GLRInitStep> initSteps;
+	std::vector<GLRInitStep> initSteps;
 
-	int frame = -1;
+	int frame;
 	GLRRunType runType;
-
-	// Avoid copying these by accident.
-	GLRRenderThreadTask(GLRRenderThreadTask &) = delete;
-	GLRRenderThreadTask& operator =(GLRRenderThreadTask &) = delete;
 };
 
 // Note: The GLRenderManager is created and destroyed on the render thread, and the latter
@@ -227,11 +407,9 @@ struct GLRRenderThreadTask {
 // directly in the destructor.
 class GLRenderManager {
 public:
-	GLRenderManager(HistoryBuffer<FrameTimeData, FRAME_TIME_HISTORY_LENGTH> &frameTimeHistory);
+	GLRenderManager() {}
 	~GLRenderManager();
 
-	GLRenderManager(GLRenderManager &) = delete;
-	GLRenderManager &operator=(GLRenderManager &) = delete;
 
 	void SetInvalidationCallback(InvalidationCallback callback) {
 		invalidationCallback_ = callback;
@@ -249,52 +427,47 @@ public:
 		caps_ = caps;
 	}
 
-	std::string GetGpuProfileString() const;
-
 	// Makes sure that the GPU has caught up enough that we can start writing buffers of this frame again.
-	void BeginFrame(bool enableProfiling);
+	void BeginFrame();
 	// Can run on a different thread!
-	void Finish();
+	void Finish(); 
-	void Present();
+	bool Run(GLRRenderThreadTask &task);
 
 	// Creation commands. These were not needed in Vulkan since there we can do that on the main thread.
 	// We pass in width/height here even though it's not strictly needed until we support glTextureStorage
 	// and then we'll also need formats and stuff.
 	GLRTexture *CreateTexture(GLenum target, int width, int height, int depth, int numMips) {
-		_dbg_assert_(target != 0);
+		GLRInitStep step { GLRInitStepType::CREATE_TEXTURE };
-		GLRInitStep &step = initSteps_.push_uninitialized();
-		step.stepType = GLRInitStepType::CREATE_TEXTURE;
 		step.create_texture.texture = new GLRTexture(caps_, width, height, depth, numMips);
 		step.create_texture.texture->target = target;
+		initSteps_.push_back(step);
 		return step.create_texture.texture;
 	}
 
 	GLRBuffer *CreateBuffer(GLuint target, size_t size, GLuint usage) {
-		GLRInitStep &step = initSteps_.push_uninitialized();
+		GLRInitStep step{ GLRInitStepType::CREATE_BUFFER };
-		step.stepType = GLRInitStepType::CREATE_BUFFER;
 		step.create_buffer.buffer = new GLRBuffer(target, size);
 		step.create_buffer.size = (int)size;
 		step.create_buffer.usage = usage;
+		initSteps_.push_back(step);
 		return step.create_buffer.buffer;
 	}
 
 	GLRShader *CreateShader(GLuint stage, const std::string &code, const std::string &desc) {
-		GLRInitStep &step = initSteps_.push_uninitialized();
+		GLRInitStep step{ GLRInitStepType::CREATE_SHADER };
-		step.stepType = GLRInitStepType::CREATE_SHADER;
 		step.create_shader.shader = new GLRShader();
 		step.create_shader.shader->desc = desc;
 		step.create_shader.stage = stage;
 		step.create_shader.code = new char[code.size() + 1];
 		memcpy(step.create_shader.code, code.data(), code.size() + 1);
+		initSteps_.push_back(step);
 		return step.create_shader.shader;
 	}
 
-	GLRFramebuffer *CreateFramebuffer(int width, int height, bool z_stencil, const char *tag) {
+	GLRFramebuffer *CreateFramebuffer(Draw::DataFormat colorFormat, int width, int height, bool z_stencil) {
-		_dbg_assert_(width > 0 && height > 0 && tag != nullptr);
+		GLRInitStep step{ GLRInitStepType::CREATE_FRAMEBUFFER };
-
+		step.create_framebuffer.framebuffer = new GLRFramebuffer(caps_, colorFormat, width, height, z_stencil);
-		GLRInitStep &step = initSteps_.push_uninitialized();
+		initSteps_.push_back(step);
-		step.stepType = GLRInitStepType::CREATE_FRAMEBUFFER;
-		step.create_framebuffer.framebuffer = new GLRFramebuffer(caps_, width, height, z_stencil, tag);
 		return step.create_framebuffer.framebuffer;
 	}
 
@@ -303,8 +476,7 @@ public:
 	GLRProgram *CreateProgram(
 		std::vector<GLRShader *> shaders, std::vector<GLRProgram::Semantic> semantics, std::vector<GLRProgram::UniformLocQuery> queries,
 		std::vector<GLRProgram::Initializer> initializers, GLRProgramLocData *locData, const GLRProgramFlags &flags) {
-		GLRInitStep &step = initSteps_.push_uninitialized();
+		GLRInitStep step{ GLRInitStepType::CREATE_PROGRAM };
-		step.stepType = GLRInitStepType::CREATE_PROGRAM;
 		_assert_(shaders.size() <= ARRAY_SIZE(step.create_program.shaders));
 		step.create_program.program = new GLRProgram();
 		step.create_program.program->semantics_ = semantics;
@@ -328,53 +500,46 @@ public:
 		}
 #endif
 		step.create_program.num_shaders = (int)shaders.size();
+		initSteps_.push_back(step);
 		return step.create_program.program;
 	}
 
-	GLRInputLayout *CreateInputLayout(const std::vector<GLRInputLayout::Entry> &entries, int stride) {
+	GLRInputLayout *CreateInputLayout(const std::vector<GLRInputLayout::Entry> &entries) {
-		GLRInitStep &step = initSteps_.push_uninitialized();
+		GLRInitStep step{ GLRInitStepType::CREATE_INPUT_LAYOUT };
-		step.stepType = GLRInitStepType::CREATE_INPUT_LAYOUT;
 		step.create_input_layout.inputLayout = new GLRInputLayout();
 		step.create_input_layout.inputLayout->entries = entries;
-		step.create_input_layout.inputLayout->stride = stride;
 		for (auto &iter : step.create_input_layout.inputLayout->entries) {
 			step.create_input_layout.inputLayout->semanticsMask_ |= 1 << iter.location;
 		}
+		initSteps_.push_back(step);
 		return step.create_input_layout.inputLayout;
 	}
 
-	GLPushBuffer *CreatePushBuffer(int frame, GLuint target, size_t size, const char *tag) {
+	GLPushBuffer *CreatePushBuffer(int frame, GLuint target, size_t size) {
-		GLPushBuffer *push = new GLPushBuffer(this, target, size, tag);
+		GLPushBuffer *push = new GLPushBuffer(this, target, size);
 		RegisterPushBuffer(frame, push);
 		return push;
 	}
 
 	void DeleteShader(GLRShader *shader) {
-		_dbg_assert_(shader != nullptr);
 		deleter_.shaders.push_back(shader);
 	}
 	void DeleteProgram(GLRProgram *program) {
-		_dbg_assert_(program != nullptr);
 		deleter_.programs.push_back(program);
 	}
 	void DeleteBuffer(GLRBuffer *buffer) {
-		_dbg_assert_(buffer != nullptr);
 		deleter_.buffers.push_back(buffer);
 	}
 	void DeleteTexture(GLRTexture *texture) {
-		_dbg_assert_(texture != nullptr);
 		deleter_.textures.push_back(texture);
 	}
 	void DeleteInputLayout(GLRInputLayout *inputLayout) {
-		_dbg_assert_(inputLayout != nullptr);
 		deleter_.inputLayouts.push_back(inputLayout);
 	}
 	void DeleteFramebuffer(GLRFramebuffer *framebuffer) {
-		_dbg_assert_(framebuffer != nullptr);
 		deleter_.framebuffers.push_back(framebuffer);
 	}
 	void DeletePushBuffer(GLPushBuffer *pushbuffer) {
-		_dbg_assert_(pushbuffer != nullptr);
 		deleter_.pushBuffers.push_back(pushbuffer);
 	}
 
@@ -386,13 +551,9 @@ public:
 		pushbuffer->End();
 	}
 
-	bool IsInRenderPass() const {
-		return curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER;
-	}
-
 	// This starts a new step (like a "render pass" in Vulkan).
 	//
-	// After a "CopyFramebuffer" or the other functions that start "steps", you need to call this before
+	// After a "CopyFramebuffer" or the other functions that start "steps", you need to call this beforce
 	// making any new render state changes or draw calls.
 	//
 	// The following state needs to be reset by the caller after calling this (and will thus not safely carry over from
@@ -409,7 +570,7 @@ public:
 	// Binds a framebuffer as a texture, for the following draws.
 	void BindFramebufferAsTexture(GLRFramebuffer *fb, int binding, int aspectBit);
 
-	bool CopyFramebufferToMemory(GLRFramebuffer *src, int aspectBits, int x, int y, int w, int h, Draw::DataFormat destFormat, uint8_t *pixels, int pixelStride, Draw::ReadbackMode mode, const char *tag);
+	bool CopyFramebufferToMemorySync(GLRFramebuffer *src, int aspectBits, int x, int y, int w, int h, Draw::DataFormat destFormat, uint8_t *pixels, int pixelStride, const char *tag);
 	void CopyImageToMemorySync(GLRTexture *texture, int mipLevel, int x, int y, int w, int h, Draw::DataFormat destFormat, uint8_t *pixels, int pixelStride, const char *tag);
 
 	void CopyFramebuffer(GLRFramebuffer *src, GLRect2D srcRect, GLRFramebuffer *dst, GLOffset2D dstPos, int aspectMask, const char *tag);
@@ -419,8 +580,7 @@ public:
 	void BufferSubdata(GLRBuffer *buffer, size_t offset, size_t size, uint8_t *data, bool deleteData = true) {
 		// TODO: Maybe should be a render command instead of an init command? When possible it's better as
 		// an init command, that's for sure.
-		GLRInitStep &step = initSteps_.push_uninitialized();
+		GLRInitStep step{ GLRInitStepType::BUFFER_SUBDATA };
-		step.stepType = GLRInitStepType::BUFFER_SUBDATA;
 		_dbg_assert_(offset >= 0);
 		_dbg_assert_(offset <= buffer->size_ - size);
 		step.buffer_subdata.buffer = buffer;
@@ -428,12 +588,12 @@ public:
 		step.buffer_subdata.size = (int)size;
 		step.buffer_subdata.data = data;
 		step.buffer_subdata.deleteData = deleteData;
+		initSteps_.push_back(step);
 	}
 
 	// Takes ownership over the data pointer and delete[]-s it.
 	void TextureImage(GLRTexture *texture, int level, int width, int height, int depth, Draw::DataFormat format, uint8_t *data, GLRAllocType allocType = GLRAllocType::NEW, bool linearFilter = false) {
-		GLRInitStep &step = initSteps_.push_uninitialized();
+		GLRInitStep step{ GLRInitStepType::TEXTURE_IMAGE };
-		step.stepType = GLRInitStepType::TEXTURE_IMAGE;
 		step.texture_image.texture = texture;
 		step.texture_image.data = data;
 		step.texture_image.format = format;
@@ -443,11 +603,12 @@ public:
 		step.texture_image.depth = depth;
 		step.texture_image.allocType = allocType;
 		step.texture_image.linearFilter = linearFilter;
+		initSteps_.push_back(step);
 	}
 
 	void TextureSubImage(int slot, GLRTexture *texture, int level, int x, int y, int width, int height, Draw::DataFormat format, uint8_t *data, GLRAllocType allocType = GLRAllocType::NEW) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData _data(GLRRenderCommand::TEXTURE_SUBIMAGE);
+		GLRRenderData _data{ GLRRenderCommand::TEXTURE_SUBIMAGE };
 		_data.texture_subimage.texture = texture;
 		_data.texture_subimage.data = data;
 		_data.texture_subimage.format = format;
@@ -462,59 +623,80 @@ public:
 	}
 
 	void FinalizeTexture(GLRTexture *texture, int loadedLevels, bool genMips) {
-		GLRInitStep &step = initSteps_.push_uninitialized();
+		GLRInitStep step{ GLRInitStepType::TEXTURE_FINALIZE };
-		step.stepType = GLRInitStepType::TEXTURE_FINALIZE;
 		step.texture_finalize.texture = texture;
 		step.texture_finalize.loadedLevels = loadedLevels;
 		step.texture_finalize.genMips = genMips;
+		initSteps_.push_back(step);
 	}
 
 	void BindTexture(int slot, GLRTexture *tex) {
-		if (!curRenderStep_ && !tex) {
-			// Likely a pre-emptive bindtexture for D3D11 to avoid hazards. Not necessary.
-			// This can happen in BlitUsingRaster.
-			return;
-		}
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
 		_dbg_assert_(slot < MAX_GL_TEXTURE_SLOTS);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::BINDTEXTURE };
-		data.cmd = GLRRenderCommand::BINDTEXTURE;
 		data.texture.slot = slot;
 		data.texture.texture = tex;
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void BindProgram(GLRProgram *program) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::BINDPROGRAM };
-		data.cmd = GLRRenderCommand::BINDPROGRAM;
 		_dbg_assert_(program != nullptr);
 		data.program.program = program;
+		curRenderStep_->commands.push_back(data);
 #ifdef _DEBUG
 		curProgram_ = program;
 #endif
 	}
 
+	void BindPixelPackBuffer(GLRBuffer *buffer) {  // Want to support an offset but can't in ES 2.0. We supply an offset when binding the buffers instead.
+		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
+		GLRRenderData data{ GLRRenderCommand::BIND_BUFFER };
+		data.bind_buffer.buffer = buffer;
+		data.bind_buffer.target = GL_PIXEL_PACK_BUFFER;
+		curRenderStep_->commands.push_back(data);
+	}
+
+	void BindIndexBuffer(GLRBuffer *buffer) {  // Want to support an offset but can't in ES 2.0. We supply an offset when binding the buffers instead.
+		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
+		GLRRenderData data{ GLRRenderCommand::BIND_BUFFER};
+		data.bind_buffer.buffer = buffer;
+		data.bind_buffer.target = GL_ELEMENT_ARRAY_BUFFER;
+		curRenderStep_->commands.push_back(data);
+	}
+
+	void BindVertexBuffer(GLRInputLayout *inputLayout, GLRBuffer *buffer, size_t offset) {
+		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
+		_dbg_assert_(inputLayout);
+		GLRRenderData data{ GLRRenderCommand::BIND_VERTEX_BUFFER };
+		data.bindVertexBuffer.inputLayout = inputLayout;
+		data.bindVertexBuffer.offset = offset;
+		data.bindVertexBuffer.buffer = buffer;
+		curRenderStep_->commands.push_back(data);
+	}
+
 	void SetDepth(bool enabled, bool write, GLenum func) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::DEPTH };
-		data.cmd = GLRRenderCommand::DEPTH;
 		data.depth.enabled = enabled;
 		data.depth.write = write;
 		data.depth.func = func;
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetViewport(const GLRViewport &vp) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::VIEWPORT };
-		data.cmd = GLRRenderCommand::VIEWPORT;
 		data.viewport.vp = vp;
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetScissor(const GLRect2D &rc) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::SCISSOR };
-		data.cmd = GLRRenderCommand::SCISSOR;
 		data.scissor.rc = rc;
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetUniformI(const GLint *loc, int count, const int *udata) {
@@ -522,12 +704,11 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORM4I };
-		data.cmd = GLRRenderCommand::UNIFORM4I;
-		data.uniform4.name = nullptr;
 		data.uniform4.loc = loc;
 		data.uniform4.count = count;
 		memcpy(data.uniform4.v, udata, sizeof(int) * count);
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetUniformI1(const GLint *loc, int udata) {
@@ -535,12 +716,11 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORM4I };
-		data.cmd = GLRRenderCommand::UNIFORM4I;
-		data.uniform4.name = nullptr;
 		data.uniform4.loc = loc;
 		data.uniform4.count = 1;
 		memcpy(data.uniform4.v, &udata, sizeof(udata));
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetUniformUI(const GLint *loc, int count, const uint32_t *udata) {
@@ -548,12 +728,11 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORM4UI };
-		data.cmd = GLRRenderCommand::UNIFORM4UI;
-		data.uniform4.name = nullptr;
 		data.uniform4.loc = loc;
 		data.uniform4.count = count;
 		memcpy(data.uniform4.v, udata, sizeof(uint32_t) * count);
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetUniformUI1(const GLint *loc, uint32_t udata) {
@@ -561,12 +740,11 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORM4UI };
-		data.cmd = GLRRenderCommand::UNIFORM4UI;
-		data.uniform4.name = nullptr;
 		data.uniform4.loc = loc;
 		data.uniform4.count = 1;
 		memcpy(data.uniform4.v, &udata, sizeof(udata));
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetUniformF(const GLint *loc, int count, const float *udata) {
@@ -574,12 +752,11 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORM4F };
-		data.cmd = GLRRenderCommand::UNIFORM4F;
-		data.uniform4.name = nullptr;
 		data.uniform4.loc = loc;
 		data.uniform4.count = count;
 		memcpy(data.uniform4.v, udata, sizeof(float) * count);
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetUniformF1(const GLint *loc, const float udata) {
@@ -587,12 +764,11 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORM4F };
-		data.cmd = GLRRenderCommand::UNIFORM4F;
-		data.uniform4.name = nullptr;
 		data.uniform4.loc = loc;
 		data.uniform4.count = 1;
 		memcpy(data.uniform4.v, &udata, sizeof(float));
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetUniformF(const char *name, int count, const float *udata) {
@@ -600,12 +776,11 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORM4F };
-		data.cmd = GLRRenderCommand::UNIFORM4F;
 		data.uniform4.name = name;
-		data.uniform4.loc = nullptr;
 		data.uniform4.count = count;
 		memcpy(data.uniform4.v, udata, sizeof(float) * count);
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetUniformM4x4(const GLint *loc, const float *udata) {
@@ -613,11 +788,10 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORMMATRIX };
-		data.cmd = GLRRenderCommand::UNIFORMMATRIX;
-		data.uniformMatrix4.name = nullptr;
 		data.uniformMatrix4.loc = loc;
 		memcpy(data.uniformMatrix4.m, udata, sizeof(float) * 16);
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetUniformM4x4Stereo(const char *name, const GLint *loc, const float *left, const float *right) {
@@ -625,13 +799,12 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORMSTEREOMATRIX };
-		data.cmd = GLRRenderCommand::UNIFORMSTEREOMATRIX;
+		data.uniformMatrix4.name = name;
-		data.uniformStereoMatrix4.name = name;
+		data.uniformMatrix4.loc = loc;
-		data.uniformStereoMatrix4.loc = loc;
+		memcpy(&data.uniformMatrix4.m[0], left, sizeof(float) * 16);
-		data.uniformStereoMatrix4.mData = new float[32];
+		memcpy(&data.uniformMatrix4.m[16], right, sizeof(float) * 16);
-		memcpy(&data.uniformStereoMatrix4.mData[0], left, sizeof(float) * 16);
+		curRenderStep_->commands.push_back(data);
-		memcpy(&data.uniformStereoMatrix4.mData[16], right, sizeof(float) * 16);
 	}
 
 	void SetUniformM4x4(const char *name, const float *udata) {
@@ -639,19 +812,17 @@ public:
 #ifdef _DEBUG
 		_dbg_assert_(curProgram_);
 #endif
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::UNIFORMMATRIX };
-		data.cmd = GLRRenderCommand::UNIFORMMATRIX;
 		data.uniformMatrix4.name = name;
-		data.uniformMatrix4.loc = nullptr;
 		memcpy(data.uniformMatrix4.m, udata, sizeof(float) * 16);
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetBlendAndMask(int colorMask, bool blendEnabled, GLenum srcColor, GLenum dstColor, GLenum srcAlpha, GLenum dstAlpha, GLenum funcColor, GLenum funcAlpha) {
 		// Make this one only a non-debug _assert_, since it often comes first.
 		// Lets us collect info about this potential crash through assert extra data.
 		_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::BLEND };
-		data.cmd = GLRRenderCommand::BLEND;
 		data.blend.mask = colorMask;
 		data.blend.enabled = blendEnabled;
 		data.blend.srcColor = srcColor;
@@ -660,88 +831,96 @@ public:
 		data.blend.dstAlpha = dstAlpha;
 		data.blend.funcColor = funcColor;
 		data.blend.funcAlpha = funcAlpha;
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetNoBlendAndMask(int colorMask) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::BLEND };
-		data.cmd = GLRRenderCommand::BLEND;
 		data.blend.mask = colorMask;
 		data.blend.enabled = false;
+		curRenderStep_->commands.push_back(data);
 	}
 
 #ifndef USING_GLES2
 	void SetLogicOp(bool enabled, GLenum logicOp) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::LOGICOP };
-		data.cmd = GLRRenderCommand::LOGICOP;
 		data.logic.enabled = enabled;
 		data.logic.logicOp = logicOp;
+		curRenderStep_->commands.push_back(data);
 	}
 #endif
 
-	void SetStencil(bool enabled, GLenum func, uint8_t refValue, uint8_t compareMask, uint8_t writeMask, GLenum sFail, GLenum zFail, GLenum pass) {
+	void SetStencilFunc(bool enabled, GLenum func, uint8_t refValue, uint8_t compareMask) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::STENCILFUNC };
-		data.cmd = GLRRenderCommand::STENCIL;
+		data.stencilFunc.enabled = enabled;
-		data.stencil.enabled = enabled;
+		data.stencilFunc.func = func;
-		data.stencil.func = func;
+		data.stencilFunc.ref = refValue;
-		data.stencil.ref = refValue;
+		data.stencilFunc.compareMask = compareMask;
-		data.stencil.compareMask = compareMask;
+		curRenderStep_->commands.push_back(data);
-		data.stencil.writeMask = writeMask;
+	}
-		data.stencil.sFail = sFail;
+
-		data.stencil.zFail = zFail;
+	void SetStencilOp(uint8_t writeMask, GLenum sFail, GLenum zFail, GLenum pass) {
-		data.stencil.pass = pass;
+		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
+		GLRRenderData data{ GLRRenderCommand::STENCILOP };
+		data.stencilOp.writeMask = writeMask;
+		data.stencilOp.sFail = sFail;
+		data.stencilOp.zFail = zFail;
+		data.stencilOp.pass = pass;
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetStencilDisabled() {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data;
-		data.cmd = GLRRenderCommand::STENCIL;
+		data.cmd = GLRRenderCommand::STENCILFUNC;
-		data.stencil.enabled = false;
+		data.stencilFunc.enabled = false;
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetBlendFactor(const float color[4]) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::BLENDCOLOR };
-		data.cmd = GLRRenderCommand::BLENDCOLOR;
 		CopyFloat4(data.blendColor.color, color);
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetRaster(GLboolean cullEnable, GLenum frontFace, GLenum cullFace, GLboolean ditherEnable, GLboolean depthClamp) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::RASTER };
-		data.cmd = GLRRenderCommand::RASTER;
 		data.raster.cullEnable = cullEnable;
 		data.raster.frontFace = frontFace;
 		data.raster.cullFace = cullFace;
 		data.raster.ditherEnable = ditherEnable;
 		data.raster.depthClampEnable = depthClamp;
+		curRenderStep_->commands.push_back(data);
 	}
 	
 	// Modifies the current texture as per GL specs, not global state.
 	void SetTextureSampler(int slot, GLenum wrapS, GLenum wrapT, GLenum magFilter, GLenum minFilter, float anisotropy) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
 		_dbg_assert_(slot < MAX_GL_TEXTURE_SLOTS);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::TEXTURESAMPLER };
-		data.cmd = GLRRenderCommand::TEXTURESAMPLER;
 		data.textureSampler.slot = slot;
 		data.textureSampler.wrapS = wrapS;
 		data.textureSampler.wrapT = wrapT;
 		data.textureSampler.magFilter = magFilter;
 		data.textureSampler.minFilter = minFilter;
 		data.textureSampler.anisotropy = anisotropy;
+		curRenderStep_->commands.push_back(data);
 	}
 
 	void SetTextureLod(int slot, float minLod, float maxLod, float lodBias) {
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
 		_dbg_assert_(slot < MAX_GL_TEXTURE_SLOTS);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::TEXTURELOD};
-		data.cmd = GLRRenderCommand::TEXTURELOD;
 		data.textureLod.slot = slot;
 		data.textureLod.minLod = minLod;
 		data.textureLod.maxLod = maxLod;
 		data.textureLod.lodBias = lodBias;
+		curRenderStep_->commands.push_back(data);
 	}
 
 	// If scissorW == 0, no scissor is applied (the whole render target is cleared).
@@ -749,8 +928,7 @@ public:
 		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
 		if (!clearMask)
 			return;
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::CLEAR };
-		data.cmd = GLRRenderCommand::CLEAR;
 		data.clear.clearMask = clearMask;
 		data.clear.clearColor = clearColor;
 		data.clear.clearZ = clearZ;
@@ -760,36 +938,30 @@ public:
 		data.clear.scissorY = scissorY;
 		data.clear.scissorW = scissorW;
 		data.clear.scissorH = scissorH;
+		curRenderStep_->commands.push_back(data);
 	}
 
-	void Draw(GLRInputLayout *inputLayout, GLRBuffer *vertexBuffer, uint32_t vertexOffset, GLenum mode, int first, int count) {
+	void Draw(GLenum mode, int first, int count) {
-		_dbg_assert_(vertexBuffer && curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
+		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		GLRRenderData data{ GLRRenderCommand::DRAW };
-		data.cmd = GLRRenderCommand::DRAW;
-		data.draw.inputLayout = inputLayout;
-		data.draw.vertexOffset = vertexOffset;
-		data.draw.vertexBuffer = vertexBuffer;
-		data.draw.indexBuffer = nullptr;
 		data.draw.mode = mode;
 		data.draw.first = first;
 		data.draw.count = count;
-		data.draw.indexType = 0;
+		data.draw.buffer = 0;
+		curRenderStep_->commands.push_back(data);
+		curRenderStep_->render.numDraws++;
 	}
 
-	// Would really love to have a basevertex parameter, but impossible in unextended GLES, without glDrawElementsBaseVertex, unfortunately.
+	void DrawIndexed(GLenum mode, int count, GLenum indexType, void *indices, int instances = 1) {
-	void DrawIndexed(GLRInputLayout *inputLayout, GLRBuffer *vertexBuffer, uint32_t vertexOffset, GLRBuffer *indexBuffer, uint32_t indexOffset, GLenum mode, int count, GLenum indexType, int instances = 1) {
+		_dbg_assert_(curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
-		_dbg_assert_(vertexBuffer && indexBuffer && curRenderStep_ && curRenderStep_->stepType == GLRStepType::RENDER);
+		GLRRenderData data{ GLRRenderCommand::DRAW_INDEXED };
-		GLRRenderData &data = curRenderStep_->commands.push_uninitialized();
+		data.drawIndexed.mode = mode;
-		data.cmd = GLRRenderCommand::DRAW;
+		data.drawIndexed.count = count;
-		data.draw.inputLayout = inputLayout;
+		data.drawIndexed.indexType = indexType;
-		data.draw.vertexOffset = vertexOffset;
+		data.drawIndexed.instances = instances;
-		data.draw.vertexBuffer = vertexBuffer;
+		data.drawIndexed.indices = indices;
-		data.draw.indexBuffer = indexBuffer;
+		curRenderStep_->commands.push_back(data);
-		data.draw.indexOffset = indexOffset;
+		curRenderStep_->render.numDraws++;
-		data.draw.mode = mode;
-		data.draw.count = count;
-		data.draw.indexType = indexType;
-		data.draw.instances = instances;
 	}
 
 	enum { MAX_INFLIGHT_FRAMES = 3 };
@@ -820,8 +992,9 @@ public:
 		_dbg_assert_(foundCount == 1);
 	}
 
-	void SetSwapFunction(std::function<void()> swapFunction) {
+	void SetSwapFunction(std::function<void()> swapFunction, bool retainControl) {
 		swapFunction_ = swapFunction;
+		retainControl_ = retainControl;
 	}
 
 	void SetSwapIntervalFunction(std::function<void(int)> swapIntervalFunction) {
@@ -853,8 +1026,6 @@ public:
 	}
 
 private:
-	bool Run(GLRRenderThreadTask &task);
-
 	// Bad for performance but sometimes necessary for synchronous CPU readbacks (screenshots and whatnot).
 	void FlushSync();
 
@@ -864,14 +1035,30 @@ private:
 		frameData_[frame].activePushBuffers.insert(buffer);
 	}
 
-	GLFrameData frameData_[MAX_INFLIGHT_FRAMES];
+	// Per-frame data, round-robin so we can overlap submission with execution of the previous frame.
+	struct FrameData {
+		bool skipSwap = false;
+
+		std::mutex fenceMutex;
+		std::condition_variable fenceCondVar;
+		bool readyForFence = true;
+
+		// Swapchain.
+		bool hasBegun = false;
+
+		GLDeleter deleter;
+		GLDeleter deleter_prev;
+		std::set<GLPushBuffer *> activePushBuffers;
+	};
+
+	FrameData frameData_[MAX_INFLIGHT_FRAMES];
 
 	// Submission time state
 	bool insideFrame_ = false;
 
 	GLRStep *curRenderStep_ = nullptr;
 	std::vector<GLRStep *> steps_;
-	FastVec<GLRInitStep> initSteps_;
+	std::vector<GLRInitStep> initSteps_;
 
 	// Execution time state
 	bool run_ = true;
@@ -883,7 +1070,7 @@ private:
 	std::mutex pushMutex_;
 	std::condition_variable pushCondVar_;
 
-	std::queue<GLRRenderThreadTask *> renderThreadQueue_;
+	std::queue<GLRRenderThreadTask> renderThreadQueue_;
 
 	// For readbacks and other reasons we need to sync with the render thread.
 	std::mutex syncMutex_;
@@ -900,6 +1087,7 @@ private:
 
 	std::function<void()> swapFunction_;
 	std::function<void(int)> swapIntervalFunction_;
+	bool retainControl_ = false;
 	GLBufferStrategy bufferStrategy_ = GLBufferStrategy::SUBDATA;
 
 	int inflightFrames_ = MAX_INFLIGHT_FRAMES;
@@ -916,9 +1104,5 @@ private:
 #endif
 	Draw::DeviceCaps caps_{};
 
-	std::string profilePassesString_;
 	InvalidationCallback invalidationCallback_;
-
-	uint64_t frameIdGen_ = FRAME_TIME_HISTORY_LENGTH;
-	HistoryBuffer<FrameTimeData, FRAME_TIME_HISTORY_LENGTH> &frameTimeHistory_;
 };

Common/GPU/OpenGL/GLSLProgram.cpp

@@ -102,7 +102,7 @@ bool glsl_recompile(GLSLProgram *program, std::string *error_message) {
 
 	if (!program->vshader_source && !vsh_src) {
 		size_t sz;
-		vsh_src.reset((char *)g_VFS.ReadFile(program->vshader_filename, &sz));
+		vsh_src.reset((char *)VFSReadFile(program->vshader_filename, &sz));
 	}
 	if (!program->vshader_source && !vsh_src) {
 		ERROR_LOG(G3D, "File missing: %s", program->vshader_filename);
@@ -113,7 +113,7 @@ bool glsl_recompile(GLSLProgram *program, std::string *error_message) {
 	}
 	if (!program->fshader_source && !fsh_src) {
 		size_t sz;
-		fsh_src.reset((char *)g_VFS.ReadFile(program->fshader_filename, &sz));
+		fsh_src.reset((char *)VFSReadFile(program->fshader_filename, &sz));
 	}
 	if (!program->fshader_source && !fsh_src) {
 		ERROR_LOG(G3D, "File missing: %s", program->fshader_filename);

Common/GPU/OpenGL/thin3d_gl.cpp

@@ -180,9 +180,8 @@ public:
 
 	void Apply(GLRenderManager *render, uint8_t stencilRef, uint8_t stencilWriteMask, uint8_t stencilCompareMask) {
 		render->SetDepth(depthTestEnabled, depthWriteEnabled, depthComp);
-		render->SetStencil(
+		render->SetStencilFunc(stencilEnabled, stencilCompareOp, stencilRef, stencilCompareMask);
-			stencilEnabled, stencilCompareOp, stencilRef, stencilCompareMask,
+		render->SetStencilOp(stencilWriteMask, stencilFail, stencilZFail, stencilPass);
-			stencilWriteMask, stencilFail, stencilZFail, stencilPass);
 	}
 };
 
@@ -322,7 +321,7 @@ class OpenGLTexture;
 
 class OpenGLContext : public DrawContext {
 public:
-	OpenGLContext(bool canChangeSwapInterval);
+	OpenGLContext();
 	~OpenGLContext();
 
 	void SetTargetSize(int w, int h) override {
@@ -359,20 +358,14 @@ public:
 	Buffer *CreateBuffer(size_t size, uint32_t usageFlags) override;
 	Framebuffer *CreateFramebuffer(const FramebufferDesc &desc) override;
 
-	void BeginFrame(DebugFlags debugFlags) override;
+	void BeginFrame() override;
 	void EndFrame() override;
-	void Present(PresentMode mode, int vblanks) override;
-
-	int GetFrameCount() override {
-		return frameCount_;
-	}
 
 	void UpdateBuffer(Buffer *buffer, const uint8_t *data, size_t offset, size_t size, UpdateBufferFlags flags) override;
-	void UpdateTextureLevels(Texture *texture, const uint8_t **data, TextureCallback initDataCallback, int numLevels) override;
 
 	void CopyFramebufferImage(Framebuffer *src, int level, int x, int y, int z, Framebuffer *dst, int dstLevel, int dstX, int dstY, int dstZ, int width, int height, int depth, int channelBits, const char *tag) override;
 	bool BlitFramebuffer(Framebuffer *src, int srcX1, int srcY1, int srcX2, int srcY2, Framebuffer *dst, int dstX1, int dstY1, int dstX2, int dstY2, int channelBits, FBBlitFilter filter, const char *tag) override;
-	bool CopyFramebufferToMemory(Framebuffer *src, int channelBits, int x, int y, int w, int h, Draw::DataFormat format, void *pixels, int pixelStride, ReadbackMode mode, const char *tag) override;
+	bool CopyFramebufferToMemorySync(Framebuffer *src, int channelBits, int x, int y, int w, int h, Draw::DataFormat format, void *pixels, int pixelStride, const char *tag) override;
 
 	// These functions should be self explanatory.
 	void BindFramebufferAsRenderTarget(Framebuffer *fbo, const RenderPassInfo &rp, const char *tag) override;
@@ -392,9 +385,9 @@ public:
 		renderManager_.SetScissor({ left, top, width, height });
 	}
 
-	void SetViewport(const Viewport &viewport) override {
+	void SetViewports(int count, Viewport *viewports) override {
 		// Same structure, different name.
-		renderManager_.SetViewport((GLRViewport &)viewport);
+		renderManager_.SetViewport((GLRViewport &)*viewports);
 	}
 
 	void SetBlendFactor(float color[4]) override {
@@ -406,11 +399,12 @@ public:
 		stencilWriteMask_ = writeMask;
 		stencilCompareMask_ = compareMask;
 		// Do we need to update on the fly here?
-		renderManager_.SetStencil(
+		renderManager_.SetStencilFunc(
 			curPipeline_->depthStencil->stencilEnabled,
 			curPipeline_->depthStencil->stencilCompareOp,
 			refValue,
-			compareMask,
+			compareMask);
+		renderManager_.SetStencilOp(
 			writeMask,
 			curPipeline_->depthStencil->stencilFail,
 			curPipeline_->depthStencil->stencilZFail,
@@ -421,9 +415,12 @@ public:
 	void BindNativeTexture(int sampler, void *nativeTexture) override;
 
 	void BindPipeline(Pipeline *pipeline) override;
-	void BindVertexBuffer(Buffer *buffer, int offset) override {
+	void BindVertexBuffers(int start, int count, Buffer **buffers, const int *offsets) override {
-		curVBuffer_ = (OpenGLBuffer *)buffer;
+		_assert_(start + count <= ARRAY_SIZE(curVBuffers_));
-		curVBufferOffset_ = offset;
+		for (int i = 0; i < count; i++) {
+			curVBuffers_[i + start] = (OpenGLBuffer *)buffers[i];
+			curVBufferOffsets_[i + start] = offsets ? offsets[i] : 0;
+		}
 	}
 	void BindIndexBuffer(Buffer *indexBuffer, int offset) override {
 		curIBuffer_ = (OpenGLBuffer  *)indexBuffer;
@@ -460,7 +457,6 @@ public:
 				case GPUVendor::VENDOR_BROADCOM: return "VENDOR_BROADCOM";
 				case GPUVendor::VENDOR_VIVANTE: return "VENDOR_VIVANTE";
 				case GPUVendor::VENDOR_APPLE: return "VENDOR_APPLE";
-				case GPUVendor::VENDOR_MESA: return "VENDOR_MESA";
 				case GPUVendor::VENDOR_UNKNOWN:
 				default:
 					return "VENDOR_UNKNOWN";
@@ -483,15 +479,10 @@ public:
 		renderManager_.SetInvalidationCallback(callback);
 	}
 
-	std::string GetGpuProfileString() const override {
-		return renderManager_.GetGpuProfileString();
-	}
-
 private:
 	void ApplySamplers();
 
 	GLRenderManager renderManager_;
-	int frameCount_ = 0;
 
 	DeviceCaps caps_{};
 
@@ -502,9 +493,9 @@ private:
 	const GLRTexture *boundTextures_[MAX_TEXTURE_SLOTS]{};
 
 	AutoRef<OpenGLPipeline> curPipeline_;
-	AutoRef<OpenGLBuffer> curVBuffer_;
+	AutoRef<OpenGLBuffer> curVBuffers_[4]{};
+	int curVBufferOffsets_[4]{};
 	AutoRef<OpenGLBuffer> curIBuffer_;
-	int curVBufferOffset_ = 0;
 	int curIBufferOffset_ = 0;
 	AutoRef<Framebuffer> curRenderTarget_;
 
@@ -540,7 +531,7 @@ static bool HasIntelDualSrcBug(const int versions[4]) {
 	}
 }
 
-OpenGLContext::OpenGLContext(bool canChangeSwapInterval) : renderManager_(frameTimeHistory_) {
+OpenGLContext::OpenGLContext() {
 	if (gl_extensions.IsGLES) {
 		if (gl_extensions.OES_packed_depth_stencil || gl_extensions.OES_depth24) {
 			caps_.preferredDepthBufferFormat = DataFormat::D24_S8;
@@ -564,7 +555,6 @@ OpenGLContext::OpenGLContext(bool canChangeSwapInterval) : renderManager_(frameT
 		caps_.textureDepthSupported = true;
 	}
 
-	caps_.setMaxFrameLatencySupported = true;
 	caps_.dualSourceBlend = gl_extensions.ARB_blend_func_extended || gl_extensions.EXT_blend_func_extended;
 	caps_.anisoSupported = gl_extensions.EXT_texture_filter_anisotropic;
 	caps_.framebufferCopySupported = gl_extensions.OES_copy_image || gl_extensions.NV_copy_image || gl_extensions.EXT_copy_image || gl_extensions.ARB_copy_image;
@@ -613,7 +603,6 @@ OpenGLContext::OpenGLContext(bool canChangeSwapInterval) : renderManager_(frameT
 	case GPU_VENDOR_IMGTEC: caps_.vendor = GPUVendor::VENDOR_IMGTEC; break;
 	case GPU_VENDOR_VIVANTE: caps_.vendor = GPUVendor::VENDOR_VIVANTE; break;
 	case GPU_VENDOR_APPLE: caps_.vendor = GPUVendor::VENDOR_APPLE; break;
-	case GPU_VENDOR_MESA: caps_.vendor = GPUVendor::VENDOR_MESA; break;
 	case GPU_VENDOR_UNKNOWN:
 	default:
 		caps_.vendor = GPUVendor::VENDOR_UNKNOWN;
@@ -632,7 +621,7 @@ OpenGLContext::OpenGLContext(bool canChangeSwapInterval) : renderManager_(frameT
 	caps_.isTilingGPU = gl_extensions.IsGLES && caps_.vendor != GPUVendor::VENDOR_NVIDIA && caps_.vendor != GPUVendor::VENDOR_INTEL;
 
 	for (int i = 0; i < GLRenderManager::MAX_INFLIGHT_FRAMES; i++) {
-		frameData_[i].push = renderManager_.CreatePushBuffer(i, GL_ARRAY_BUFFER, 64 * 1024, "thin3d_vbuf");
+		frameData_[i].push = renderManager_.CreatePushBuffer(i, GL_ARRAY_BUFFER, 64 * 1024);
 	}
 
 	if (!gl_extensions.VersionGEThan(3, 0, 0)) {
@@ -751,9 +740,7 @@ OpenGLContext::OpenGLContext(bool canChangeSwapInterval) : renderManager_(frameT
 			// This too...
 			shaderLanguageDesc_.shaderLanguage = ShaderLanguage::GLSL_1xx;
 			if (gl_extensions.EXT_gpu_shader4) {
-				// Older macOS devices seem to have problems defining uint uniforms.
+				shaderLanguageDesc_.bitwiseOps = true;
-				// Let's just assume OpenGL 3.0+ is required.
-				shaderLanguageDesc_.bitwiseOps = gl_extensions.VersionGEThan(3, 0, 0);
 				shaderLanguageDesc_.texelFetch = "texelFetch2D";
 			}
 		}
@@ -772,16 +759,6 @@ OpenGLContext::OpenGLContext(bool canChangeSwapInterval) : renderManager_(frameT
 		}
 	}
 
-	if (canChangeSwapInterval) {
-		caps_.presentInstantModeChange = true;
-		caps_.presentMaxInterval = 4;
-		caps_.presentModesSupported = PresentMode::FIFO | PresentMode::IMMEDIATE;
-	} else {
-		caps_.presentInstantModeChange = false;
-		caps_.presentModesSupported = PresentMode::FIFO;
-		caps_.presentMaxInterval = 1;
-	}
-
 	renderManager_.SetDeviceCaps(caps_);
 }
 
@@ -793,8 +770,8 @@ OpenGLContext::~OpenGLContext() {
 	}
 }
 
-void OpenGLContext::BeginFrame(DebugFlags debugFlags) {
+void OpenGLContext::BeginFrame() {
-	renderManager_.BeginFrame(debugFlags & DebugFlags::PROFILE_TIMESTAMPS);
+	renderManager_.BeginFrame();
 	FrameData &frameData = frameData_[renderManager_.GetCurFrame()];
 	renderManager_.BeginPushBuffer(frameData.push);
 }
@@ -803,12 +780,8 @@ void OpenGLContext::EndFrame() {
 	FrameData &frameData = frameData_[renderManager_.GetCurFrame()];
 	renderManager_.EndPushBuffer(frameData.push);  // upload the data!
 	renderManager_.Finish();
-	Invalidate(InvalidationFlags::CACHED_RENDER_STATE);
-}
 
-void OpenGLContext::Present(PresentMode presentMode, int vblanks) {
+	Invalidate(InvalidationFlags::CACHED_RENDER_STATE);
-	renderManager_.Present();
-	frameCount_++;
 }
 
 void OpenGLContext::Invalidate(InvalidationFlags flags) {
@@ -830,7 +803,7 @@ InputLayout *OpenGLContext::CreateInputLayout(const InputLayoutDesc &desc) {
 	return fmt;
 }
 
-static GLuint TypeToTarget(TextureType type) {
+GLuint TypeToTarget(TextureType type) {
 	switch (type) {
 #ifndef USING_GLES2
 	case TextureType::LINEAR1D: return GL_TEXTURE_1D;
@@ -868,33 +841,25 @@ public:
 		return tex_;
 	}
 
-	void UpdateTextureLevels(GLRenderManager *render, const uint8_t *const *data, int numLevels, TextureCallback initDataCallback);
-
 private:
-	void SetImageData(int x, int y, int z, int width, int height, int depth, int level, int stride, const uint8_t *data, TextureCallback initDataCallback);
+	void SetImageData(int x, int y, int z, int width, int height, int depth, int level, int stride, const uint8_t *data, TextureCallback callback);
 
 	GLRenderManager *render_;
 	GLRTexture *tex_;
 
+	DataFormat format_;
 	TextureType type_;
 	int mipLevels_;
-	bool generateMips_;  // Generate mips requested
+	bool generatedMips_;
-	bool generatedMips_;  // Has generated mips
 };
 
 OpenGLTexture::OpenGLTexture(GLRenderManager *render, const TextureDesc &desc) : render_(render) {
-	_dbg_assert_(desc.format != Draw::DataFormat::UNDEFINED);
-	_dbg_assert_(desc.width > 0 && desc.height > 0 && desc.depth > 0);
-	_dbg_assert_(desc.type != Draw::TextureType::UNKNOWN);
-
 	generatedMips_ = false;
-	generateMips_ = desc.generateMips;
 	width_ = desc.width;
 	height_ = desc.height;
 	depth_ = desc.depth;
 	format_ = desc.format;
 	type_ = desc.type;
-
 	GLenum target = TypeToTarget(desc.type);
 	tex_ = render->CreateTexture(target, desc.width, desc.height, 1, desc.mipLevels);
 
@@ -902,25 +867,21 @@ OpenGLTexture::OpenGLTexture(GLRenderManager *render, const TextureDesc &desc) :
 	if (desc.initData.empty())
 		return;
 
-	UpdateTextureLevels(render, desc.initData.data(), (int)desc.initData.size(), desc.initDataCallback);
-}
-
-void OpenGLTexture::UpdateTextureLevels(GLRenderManager *render, const uint8_t * const *data, int numLevels, TextureCallback initDataCallback) {
 	int level = 0;
 	int width = width_;
 	int height = height_;
 	int depth = depth_;
-	for (int i = 0; i < numLevels; i++) {
+	for (auto data : desc.initData) {
-		SetImageData(0, 0, 0, width, height, depth, level, 0, data[i], initDataCallback);
+		SetImageData(0, 0, 0, width, height, depth, level, 0, data, desc.initDataCallback);
 		width = (width + 1) / 2;
 		height = (height + 1) / 2;
 		depth = (depth + 1) / 2;
 		level++;
 	}
-	mipLevels_ = generateMips_ ? mipLevels_ : level;
+	mipLevels_ = desc.generateMips ? desc.mipLevels : level;
 
 	bool genMips = false;
-	if (numLevels < mipLevels_ && generateMips_) {
+	if ((int)desc.initData.size() < desc.mipLevels && desc.generateMips) {
 		// Assumes the texture is bound for editing
 		genMips = true;
 		generatedMips_ = true;
@@ -931,7 +892,7 @@ void OpenGLTexture::UpdateTextureLevels(GLRenderManager *render, const uint8_t *
 OpenGLTexture::~OpenGLTexture() {
 	if (tex_) {
 		render_->DeleteTexture(tex_);
-		tex_ = nullptr;
+		tex_ = 0;
 		generatedMips_ = false;
 	}
 }
@@ -950,7 +911,7 @@ public:
 	GLRFramebuffer *framebuffer_ = nullptr;
 };
 
-void OpenGLTexture::SetImageData(int x, int y, int z, int width, int height, int depth, int level, int stride, const uint8_t *data, TextureCallback initDataCallback) {
+void OpenGLTexture::SetImageData(int x, int y, int z, int width, int height, int depth, int level, int stride, const uint8_t *data, TextureCallback callback) {
 	if ((width != width_ || height != height_ || depth != depth_) && level == 0) {
 		// When switching to texStorage we need to handle this correctly.
 		width_ = width;
@@ -966,8 +927,8 @@ void OpenGLTexture::SetImageData(int x, int y, int z, int width, int height, int
 	uint8_t *texData = new uint8_t[(size_t)(width * height * depth * alignment)];
 
 	bool texDataPopulated = false;
-	if (initDataCallback) {
+	if (callback) {
-		texDataPopulated = initDataCallback(texData, data, width, height, depth, width * (int)alignment, height * width * (int)alignment);
+		texDataPopulated = callback(texData, data, width, height, depth, width * (int)alignment, height * width * (int)alignment);
 	}
 	if (texDataPopulated) {
 		if (format_ == DataFormat::A1R5G5B5_UNORM_PACK16) {
@@ -1027,7 +988,7 @@ static void LogReadPixelsError(GLenum error) {
 }
 #endif
 
-bool OpenGLContext::CopyFramebufferToMemory(Framebuffer *src, int channelBits, int x, int y, int w, int h, Draw::DataFormat dataFormat, void *pixels, int pixelStride, ReadbackMode mode, const char *tag) {
+bool OpenGLContext::CopyFramebufferToMemorySync(Framebuffer *src, int channelBits, int x, int y, int w, int h, Draw::DataFormat dataFormat, void *pixels, int pixelStride, const char *tag) {
 	if (gl_extensions.IsGLES && (channelBits & FB_COLOR_BIT) == 0) {
 		// Can't readback depth or stencil on GLES.
 		return false;
@@ -1040,7 +1001,8 @@ bool OpenGLContext::CopyFramebufferToMemory(Framebuffer *src, int channelBits, i
 		aspect |= GL_DEPTH_BUFFER_BIT;
 	if (channelBits & FB_STENCIL_BIT)
 		aspect |= GL_STENCIL_BUFFER_BIT;
-	return renderManager_.CopyFramebufferToMemory(fb ? fb->framebuffer_ : nullptr, aspect, x, y, w, h, dataFormat, (uint8_t *)pixels, pixelStride, mode, tag);
+	renderManager_.CopyFramebufferToMemorySync(fb ? fb->framebuffer_ : nullptr, aspect, x, y, w, h, dataFormat, (uint8_t *)pixels, pixelStride, tag);
+	return true;
 }
 
 
@@ -1048,11 +1010,6 @@ Texture *OpenGLContext::CreateTexture(const TextureDesc &desc) {
 	return new OpenGLTexture(&renderManager_, desc);
 }
 
-void OpenGLContext::UpdateTextureLevels(Texture *texture, const uint8_t **data, TextureCallback initDataCallback, int numLevels) {
-	OpenGLTexture *tex = (OpenGLTexture *)texture;
-	tex->UpdateTextureLevels(&renderManager_, data, numLevels, initDataCallback);
-}
-
 DepthStencilState *OpenGLContext::CreateDepthStencilState(const DepthStencilStateDesc &desc) {
 	OpenGLDepthStencilState *ds = new OpenGLDepthStencilState();
 	ds->depthTestEnabled = desc.depthTestEnabled;
@@ -1306,7 +1263,7 @@ bool OpenGLPipeline::LinkShaders(const PipelineDesc &desc) {
 		for (int i = 0; i < (int)std::min((const uint32_t)samplers_.size(), MAX_TEXTURE_SLOTS); i++) {
 			queries.push_back({ &locs_->samplerLocs_[i], samplers_[i].name, true });
 		}
-		samplersToCheck = (int)std::min((const uint32_t)samplers_.size(), MAX_TEXTURE_SLOTS);
+		samplersToCheck = (int)samplers_.size();
 	} else {
 		queries.push_back({ &locs_->samplerLocs_[0], "sampler0" });
 		queries.push_back({ &locs_->samplerLocs_[1], "sampler1" });
@@ -1367,39 +1324,40 @@ void OpenGLContext::UpdateDynamicUniformBuffer(const void *ub, size_t size) {
 }
 
 void OpenGLContext::Draw(int vertexCount, int offset) {
-	_dbg_assert_msg_(curVBuffer_ != nullptr, "Can't call Draw without a vertex buffer");
+	_dbg_assert_msg_(curVBuffers_[0] != nullptr, "Can't call Draw without a vertex buffer");
 	ApplySamplers();
-	_assert_(curPipeline_->inputLayout);
+	if (curPipeline_->inputLayout) {
-	renderManager_.Draw(curPipeline_->inputLayout->inputLayout_, curVBuffer_->buffer_, curVBufferOffset_, curPipeline_->prim, offset, vertexCount);
+		renderManager_.BindVertexBuffer(curPipeline_->inputLayout->inputLayout_, curVBuffers_[0]->buffer_, curVBufferOffsets_[0]);
+	}
+	renderManager_.Draw(curPipeline_->prim, offset, vertexCount);
 }
 
 void OpenGLContext::DrawIndexed(int vertexCount, int offset) {
-	_dbg_assert_msg_(curVBuffer_ != nullptr, "Can't call DrawIndexed without a vertex buffer");
+	_dbg_assert_msg_(curVBuffers_[0] != nullptr, "Can't call DrawIndexed without a vertex buffer");
 	_dbg_assert_msg_(curIBuffer_ != nullptr, "Can't call DrawIndexed without an index buffer");
 	ApplySamplers();
-	_assert_(curPipeline_->inputLayout);
+	if (curPipeline_->inputLayout) {
-	renderManager_.DrawIndexed(
+		renderManager_.BindVertexBuffer(curPipeline_->inputLayout->inputLayout_, curVBuffers_[0]->buffer_, curVBufferOffsets_[0]);
-		curPipeline_->inputLayout->inputLayout_,
+	}
-		curVBuffer_->buffer_, curVBufferOffset_,
+	renderManager_.BindIndexBuffer(curIBuffer_->buffer_);
-		curIBuffer_->buffer_, curIBufferOffset_ + offset * sizeof(uint32_t),
+	renderManager_.DrawIndexed(curPipeline_->prim, vertexCount, GL_UNSIGNED_SHORT, (void *)((intptr_t)curIBufferOffset_ + offset * sizeof(uint32_t)));
-		curPipeline_->prim, vertexCount, GL_UNSIGNED_SHORT);
 }
 
 void OpenGLContext::DrawUP(const void *vdata, int vertexCount) {
 	_assert_(curPipeline_->inputLayout != nullptr);
 	int stride = curPipeline_->inputLayout->stride;
-	uint32_t dataSize = stride * vertexCount;
+	size_t dataSize = stride * vertexCount;
 
 	FrameData &frameData = frameData_[renderManager_.GetCurFrame()];
 
 	GLRBuffer *buf;
-	uint32_t offset;
+	size_t offset = frameData.push->Push(vdata, dataSize, &buf);
-	uint8_t *dest = frameData.push->Allocate(dataSize, 4, &buf, &offset);
-	memcpy(dest, vdata, dataSize);
 
 	ApplySamplers();
-	_assert_(curPipeline_->inputLayout);
+	if (curPipeline_->inputLayout) {
-	renderManager_.Draw(curPipeline_->inputLayout->inputLayout_, buf, offset, curPipeline_->prim, 0, vertexCount);
+		renderManager_.BindVertexBuffer(curPipeline_->inputLayout->inputLayout_, buf, offset);
+	}
+	renderManager_.Draw(curPipeline_->prim, 0, vertexCount);
 }
 
 void OpenGLContext::Clear(int mask, uint32_t colorval, float depthVal, int stencilVal) {
@@ -1418,8 +1376,8 @@ void OpenGLContext::Clear(int mask, uint32_t colorval, float depthVal, int stenc
 	renderManager_.Clear(colorval, depthVal, stencilVal, glMask, 0xF, 0, 0, targetWidth_, targetHeight_);
 }
 
-DrawContext *T3DCreateGLContext(bool canChangeSwapInterval) {
+DrawContext *T3DCreateGLContext() {
-	return new OpenGLContext(canChangeSwapInterval);
+	return new OpenGLContext();
 }
 
 OpenGLInputLayout::~OpenGLInputLayout() {
@@ -1429,12 +1387,13 @@ OpenGLInputLayout::~OpenGLInputLayout() {
 void OpenGLInputLayout::Compile(const InputLayoutDesc &desc) {
 	// TODO: This is only accurate if there's only one stream. But whatever, for now we
 	// never use multiple streams anyway.
-	stride = desc.stride;
+	stride = desc.bindings.empty() ? 0 : (GLsizei)desc.bindings[0].stride;
 
 	std::vector<GLRInputLayout::Entry> entries;
 	for (auto &attr : desc.attributes) {
 		GLRInputLayout::Entry entry;
 		entry.location = attr.location;
+		entry.stride = (GLsizei)desc.bindings[attr.binding].stride;
 		entry.offset = attr.offset;
 		switch (attr.format) {
 		case DataFormat::R32G32_FLOAT:
@@ -1466,7 +1425,7 @@ void OpenGLInputLayout::Compile(const InputLayoutDesc &desc) {
 		entries.push_back(entry);
 	}
 	if (!entries.empty()) {
-		inputLayout_ = render_->CreateInputLayout(entries, stride);
+		inputLayout_ = render_->CreateInputLayout(entries);
 	} else {
 		inputLayout_ = nullptr;
 	}
@@ -1478,7 +1437,7 @@ Framebuffer *OpenGLContext::CreateFramebuffer(const FramebufferDesc &desc) {
 	// TODO: Support multiview later. (It's our only use case for multi layers).
 	_dbg_assert_(desc.numLayers == 1);
 
-	GLRFramebuffer *framebuffer = renderManager_.CreateFramebuffer(desc.width, desc.height, desc.z_stencil, desc.tag);
+	GLRFramebuffer *framebuffer = renderManager_.CreateFramebuffer(desc.colorFormat, desc.width, desc.height, desc.z_stencil);
 	OpenGLFramebuffer *fbo = new OpenGLFramebuffer(&renderManager_, framebuffer);
 	return fbo;
 }
@@ -1598,23 +1557,7 @@ uint32_t OpenGLContext::GetDataFormatSupport(DataFormat fmt) const {
 	case DataFormat::BC1_RGBA_UNORM_BLOCK:
 	case DataFormat::BC2_UNORM_BLOCK:
 	case DataFormat::BC3_UNORM_BLOCK:
-		return gl_extensions.supportsBC123 ? FMT_TEXTURE : 0;
+		return FMT_TEXTURE;
-
-	case DataFormat::BC4_UNORM_BLOCK:
-	case DataFormat::BC5_UNORM_BLOCK:
-		return gl_extensions.supportsBC45 ? FMT_TEXTURE : 0;
-
-	case DataFormat::BC7_UNORM_BLOCK:
-		return gl_extensions.supportsBC7 ? FMT_TEXTURE : 0;
-
-	case DataFormat::ASTC_4x4_UNORM_BLOCK:
-		return gl_extensions.supportsASTC ? FMT_TEXTURE : 0;
-
-	case DataFormat::ETC2_R8G8B8_UNORM_BLOCK:
-	case DataFormat::ETC2_R8G8B8A1_UNORM_BLOCK:
-	case DataFormat::ETC2_R8G8B8A8_UNORM_BLOCK:
-		return gl_extensions.supportsETC2 ? FMT_TEXTURE : 0;
-
 	default:
 		return 0;
 	}

Common/GPU/ShaderTranslation.cpp

@@ -63,10 +63,15 @@ static EShLanguage GetShLanguageFromStage(const ShaderStage stage) {
 }
 
 void ShaderTranslationInit() {
+	// TODO: We have TLS issues on UWP
+#if !PPSSPP_PLATFORM(UWP)
 	glslang::InitializeProcess();
+#endif
 }
 void ShaderTranslationShutdown() {
+#if !PPSSPP_PLATFORM(UWP)
 	glslang::FinalizeProcess();
+#endif
 }
 
 struct Builtin {
@@ -224,6 +229,11 @@ bool TranslateShader(std::string *dest, ShaderLanguage destLang, const ShaderLan
 		return result;
 	}
 
+#if PPSSPP_PLATFORM(UWP)
+	*errorMessage = "No shader translation available (UWP)";
+	return false;
+#endif
+
 	errorMessage->clear();
 
 	glslang::TProgram program;

Common/GPU/Vulkan/VulkanContext.cpp

@@ -57,13 +57,12 @@ std::string VulkanVendorString(uint32_t vendorId) {
 	case VULKAN_VENDOR_QUALCOMM: return "Qualcomm";
 	case VULKAN_VENDOR_IMGTEC: return "Imagination";
 	case VULKAN_VENDOR_APPLE: return "Apple";
-	case VULKAN_VENDOR_MESA: return "Mesa";
 	default:
 		return StringFromFormat("%08x", vendorId);
 	}
 }
 
-const char *VulkanPresentModeToString(VkPresentModeKHR presentMode) {
+const char *PresentModeString(VkPresentModeKHR presentMode) {
 	switch (presentMode) {
 	case VK_PRESENT_MODE_IMMEDIATE_KHR: return "IMMEDIATE";
 	case VK_PRESENT_MODE_MAILBOX_KHR: return "MAILBOX";
@@ -536,7 +535,7 @@ int VulkanContext::GetBestPhysicalDevice() {
 
 void VulkanContext::ChooseDevice(int physical_device) {
 	physical_device_ = physical_device;
-	INFO_LOG(G3D, "Chose physical device %d: %s", physical_device, physicalDeviceProperties_[physical_device].properties.deviceName);
+	INFO_LOG(G3D, "Chose physical device %d: %p", physical_device, physical_devices_[physical_device]);
 
 	GetDeviceLayerProperties();
 	if (!CheckLayers(device_layer_properties_, device_layer_names_)) {
@@ -550,13 +549,12 @@ void VulkanContext::ChooseDevice(int physical_device) {
 	vkGetPhysicalDeviceQueueFamilyProperties(physical_devices_[physical_device_], &queue_count, queueFamilyProperties_.data());
 	_dbg_assert_(queue_count >= 1);
 
-	// Detect preferred depth/stencil formats, in this order. All supported devices will support at least one of these.
+	// Detect preferred formats, in this order.
 	static const VkFormat depthStencilFormats[] = {
 		VK_FORMAT_D24_UNORM_S8_UINT,
 		VK_FORMAT_D32_SFLOAT_S8_UINT,
 		VK_FORMAT_D16_UNORM_S8_UINT,
 	};
-
 	deviceInfo_.preferredDepthStencilFormat = VK_FORMAT_UNDEFINED;
 	for (size_t i = 0; i < ARRAY_SIZE(depthStencilFormats); i++) {
 		VkFormatProperties props;
@@ -575,8 +573,7 @@ void VulkanContext::ChooseDevice(int physical_device) {
 		deviceInfo_.canBlitToPreferredDepthStencilFormat = true;
 	}
 
-	// This is as good a place as any to do this. Though, we don't use this much anymore after we added
+	// This is as good a place as any to do this.
-	// support for VMA.
 	vkGetPhysicalDeviceMemoryProperties(physical_devices_[physical_device_], &memory_properties_);
 	INFO_LOG(G3D, "Memory Types (%d):", memory_properties_.memoryTypeCount);
 	for (int i = 0; i < (int)memory_properties_.memoryTypeCount; i++) {
@@ -595,24 +592,31 @@ void VulkanContext::ChooseDevice(int physical_device) {
 		VkPhysicalDeviceFeatures2 features2{VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2_KHR};
 		// Add to chain even if not supported, GetPhysicalDeviceFeatures is supposed to ignore unknown structs.
 		VkPhysicalDeviceMultiviewFeatures multiViewFeatures{ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES };
-		VkPhysicalDevicePresentWaitFeaturesKHR presentWaitFeatures{ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PRESENT_WAIT_FEATURES_KHR };
-		VkPhysicalDevicePresentIdFeaturesKHR presentIdFeatures{ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PRESENT_ID_FEATURES_KHR };
-
 		features2.pNext = &multiViewFeatures;
-		multiViewFeatures.pNext = &presentWaitFeatures;
-		presentWaitFeatures.pNext = &presentIdFeatures;
-		presentIdFeatures.pNext = nullptr;
-
 		vkGetPhysicalDeviceFeatures2KHR(physical_devices_[physical_device_], &features2);
 		deviceFeatures_.available.standard = features2.features;
 		deviceFeatures_.available.multiview = multiViewFeatures;
-		deviceFeatures_.available.presentWait = presentWaitFeatures;
-		deviceFeatures_.available.presentId = presentIdFeatures;
 	} else {
 		vkGetPhysicalDeviceFeatures(physical_devices_[physical_device_], &deviceFeatures_.available.standard);
 		deviceFeatures_.available.multiview = {};
 	}
 
+	deviceFeatures_.enabled = {};
+	// Enable a few safe ones if they are available.
+	deviceFeatures_.enabled.standard.dualSrcBlend = deviceFeatures_.available.standard.dualSrcBlend;
+	deviceFeatures_.enabled.standard.logicOp = deviceFeatures_.available.standard.logicOp;
+	deviceFeatures_.enabled.standard.depthClamp = deviceFeatures_.available.standard.depthClamp;
+	deviceFeatures_.enabled.standard.depthBounds = deviceFeatures_.available.standard.depthBounds;
+	deviceFeatures_.enabled.standard.samplerAnisotropy = deviceFeatures_.available.standard.samplerAnisotropy;
+	deviceFeatures_.enabled.standard.shaderClipDistance = deviceFeatures_.available.standard.shaderClipDistance;
+	deviceFeatures_.enabled.standard.shaderCullDistance = deviceFeatures_.available.standard.shaderCullDistance;
+	deviceFeatures_.enabled.standard.geometryShader = deviceFeatures_.available.standard.geometryShader;
+	deviceFeatures_.enabled.standard.sampleRateShading = deviceFeatures_.available.standard.sampleRateShading;
+
+	deviceFeatures_.enabled.multiview = { VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES };
+	deviceFeatures_.enabled.multiview.multiview = deviceFeatures_.available.multiview.multiview;
+	// deviceFeatures_.enabled.multiview.multiviewGeometryShader = deviceFeatures_.available.multiview.multiviewGeometryShader;
+
 	GetDeviceLayerExtensionList(nullptr, device_extension_properties_);
 
 	device_extensions_enabled_.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
@@ -644,8 +648,8 @@ VkResult VulkanContext::CreateDevice() {
 		return VK_ERROR_INITIALIZATION_FAILED;
 	}
 
-	VkDeviceQueueCreateInfo queue_info{ VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO };
+	VkDeviceQueueCreateInfo queue_info{VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO};
-	float queue_priorities[1] = { 1.0f };
+	float queue_priorities[1] = {1.0f};
 	queue_info.queueCount = 1;
 	queue_info.pQueuePriorities = queue_priorities;
 	bool found = false;
@@ -668,6 +672,11 @@ VkResult VulkanContext::CreateDevice() {
 		extensionsLookup_.KHR_get_memory_requirements2 = true;
 		extensionsLookup_.KHR_dedicated_allocation = EnableDeviceExtension(VK_KHR_DEDICATED_ALLOCATION_EXTENSION_NAME);
 	}
+	if (EnableDeviceExtension(VK_KHR_EXTERNAL_MEMORY_CAPABILITIES_EXTENSION_NAME)) {
+		if (EnableDeviceExtension(VK_KHR_EXTERNAL_MEMORY_EXTENSION_NAME)) {
+			extensionsLookup_.EXT_external_memory_host = EnableDeviceExtension(VK_EXT_EXTERNAL_MEMORY_HOST_EXTENSION_NAME);
+		}
+	}
 	if (EnableDeviceExtension(VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME)) {
 		extensionsLookup_.KHR_create_renderpass2 = true;
 		extensionsLookup_.KHR_depth_stencil_resolve = EnableDeviceExtension(VK_KHR_DEPTH_STENCIL_RESOLVE_EXTENSION_NAME);
@@ -677,41 +686,6 @@ VkResult VulkanContext::CreateDevice() {
 	extensionsLookup_.EXT_fragment_shader_interlock = EnableDeviceExtension(VK_EXT_FRAGMENT_SHADER_INTERLOCK_EXTENSION_NAME);
 	extensionsLookup_.ARM_rasterization_order_attachment_access = EnableDeviceExtension(VK_ARM_RASTERIZATION_ORDER_ATTACHMENT_ACCESS_EXTENSION_NAME);
 
-#if !PPSSPP_PLATFORM(MAC) && !PPSSPP_PLATFORM(IOS)
-	extensionsLookup_.GOOGLE_display_timing = EnableDeviceExtension(VK_GOOGLE_DISPLAY_TIMING_EXTENSION_NAME);
-#endif
-	if (!extensionsLookup_.GOOGLE_display_timing) {
-		extensionsLookup_.KHR_present_id = EnableDeviceExtension(VK_KHR_PRESENT_ID_EXTENSION_NAME);
-		extensionsLookup_.KHR_present_wait = EnableDeviceExtension(VK_KHR_PRESENT_WAIT_EXTENSION_NAME);
-	}
-
-	deviceFeatures_.enabled = {};
-	// Enable a few safe ones if they are available.
-	deviceFeatures_.enabled.standard.dualSrcBlend = deviceFeatures_.available.standard.dualSrcBlend;
-	deviceFeatures_.enabled.standard.logicOp = deviceFeatures_.available.standard.logicOp;
-	deviceFeatures_.enabled.standard.depthClamp = deviceFeatures_.available.standard.depthClamp;
-	deviceFeatures_.enabled.standard.depthBounds = deviceFeatures_.available.standard.depthBounds;
-	deviceFeatures_.enabled.standard.samplerAnisotropy = deviceFeatures_.available.standard.samplerAnisotropy;
-	deviceFeatures_.enabled.standard.shaderClipDistance = deviceFeatures_.available.standard.shaderClipDistance;
-	deviceFeatures_.enabled.standard.shaderCullDistance = deviceFeatures_.available.standard.shaderCullDistance;
-	deviceFeatures_.enabled.standard.geometryShader = deviceFeatures_.available.standard.geometryShader;
-	deviceFeatures_.enabled.standard.sampleRateShading = deviceFeatures_.available.standard.sampleRateShading;
-
-	deviceFeatures_.enabled.multiview = { VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES };
-	if (extensionsLookup_.KHR_multiview) {
-		deviceFeatures_.enabled.multiview.multiview = deviceFeatures_.available.multiview.multiview;
-	}
-	// Strangely, on Intel, it reports these as available even though the extension isn't in the list.
-	deviceFeatures_.enabled.presentId = { VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PRESENT_ID_FEATURES_KHR };
-	if (extensionsLookup_.KHR_present_id) {
-		deviceFeatures_.enabled.presentId.presentId = deviceFeatures_.available.presentId.presentId;
-	}
-	deviceFeatures_.enabled.presentWait = { VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PRESENT_WAIT_FEATURES_KHR };
-	if (extensionsLookup_.KHR_present_wait) {
-		deviceFeatures_.enabled.presentWait.presentWait = deviceFeatures_.available.presentWait.presentWait;
-	}
-	// deviceFeatures_.enabled.multiview.multiviewGeometryShader = deviceFeatures_.available.multiview.multiviewGeometryShader;
-
 	VkPhysicalDeviceFeatures2 features2{ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2 };
 
 	VkDeviceCreateInfo device_info{ VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO };
@@ -726,9 +700,6 @@ VkResult VulkanContext::CreateDevice() {
 		device_info.pNext = &features2;
 		features2.features = deviceFeatures_.enabled.standard;
 		features2.pNext = &deviceFeatures_.enabled.multiview;
-		deviceFeatures_.enabled.multiview.pNext = &deviceFeatures_.enabled.presentWait;
-		deviceFeatures_.enabled.presentWait.pNext = &deviceFeatures_.enabled.presentId;
-		deviceFeatures_.enabled.presentId.pNext = nullptr;
 	} else {
 		device_info.pEnabledFeatures = &deviceFeatures_.enabled.standard;
 	}
@@ -740,9 +711,7 @@ VkResult VulkanContext::CreateDevice() {
 	} else {
 		VulkanLoadDeviceFunctions(device_, extensionsLookup_);
 	}
-	INFO_LOG(G3D, "Vulkan Device created: %s", physicalDeviceProperties_[physical_device_].properties.deviceName);
+	INFO_LOG(G3D, "Vulkan Device created");
-
-	// Since we successfully created a device (however we got here, might be interesting in debug), we force the choice to be visible in the menu.
 	VulkanSetAvailable(true);
 
 	VmaAllocatorCreateInfo allocatorInfo = {};
@@ -917,187 +886,6 @@ VkResult VulkanContext::ReinitSurface() {
 	case WINDOWSYSTEM_DISPLAY:
 	{
 		VkDisplaySurfaceCreateInfoKHR display{ VK_STRUCTURE_TYPE_DISPLAY_SURFACE_CREATE_INFO_KHR };
-#if !defined(__LIBRETRO__)
-		/*
-		And when not to use libretro need VkDisplaySurfaceCreateInfoKHR this extension,
-		then you need to use dlopen to read vulkan loader in VulkanLoader.cpp.
-		huangzihan China
-		*/
-
-		if(!vkGetPhysicalDeviceDisplayPropertiesKHR || 
-		   !vkGetPhysicalDeviceDisplayPlanePropertiesKHR || 
-		   !vkGetDisplayModePropertiesKHR || 
-		   !vkGetDisplayPlaneSupportedDisplaysKHR || 
-		   !vkGetDisplayPlaneCapabilitiesKHR ) {
-			_assert_msg_(false, "DISPLAY Vulkan cannot find any vulkan function symbols.");
-			return VK_ERROR_INITIALIZATION_FAILED;
-		}
-
-		//The following code is for reference:
-		// https://github.com/vanfanel/ppsspp
-		// When using the VK_KHR_display extension and not using LIBRETRO, a complete
-		// VkDisplaySurfaceCreateInfoKHR is needed.
-
-		uint32_t display_count;
-		uint32_t plane_count;
-
-		VkDisplayPropertiesKHR *display_props = NULL;
-		VkDisplayPlanePropertiesKHR *plane_props = NULL;
-		VkDisplayModePropertiesKHR* mode_props = NULL;
-
-		VkExtent2D image_size;
-		// This is the chosen physical_device, it has been chosen elsewhere.
-		VkPhysicalDevice phys_device = physical_devices_[physical_device_];
-		VkDisplayModeKHR display_mode = VK_NULL_HANDLE;
-		VkDisplayPlaneAlphaFlagBitsKHR alpha_mode = VK_DISPLAY_PLANE_ALPHA_OPAQUE_BIT_KHR;
-		uint32_t plane = UINT32_MAX;
-
-		// For now, use the first available (connected) display.
-		int display_index = 0;
-
-		VkResult result;
-		bool ret = false;
-		bool mode_found = false;
-
-		int i, j;
-
-		// 1 physical device can have N displays connected.
-		// Vulkan only counts the connected displays.
-
-		// Get a list of displays on the physical device.
-		display_count = 0;
-		vkGetPhysicalDeviceDisplayPropertiesKHR(phys_device, &display_count, NULL);
-		if (display_count == 0) {
-			_assert_msg_(false, "DISPLAY Vulkan couldn't find any displays.");
-			return VK_ERROR_INITIALIZATION_FAILED;
-		}
-		display_props = new VkDisplayPropertiesKHR[display_count];
-		vkGetPhysicalDeviceDisplayPropertiesKHR(phys_device, &display_count, display_props);
-
-		// Get a list of display planes on the physical device.
-		plane_count = 0;
-		vkGetPhysicalDeviceDisplayPlanePropertiesKHR(phys_device, &plane_count, NULL);
-		if (plane_count == 0) {
-			_assert_msg_(false, "DISPLAY Vulkan couldn't find any planes on the physical device");
-			return VK_ERROR_INITIALIZATION_FAILED;
-
-		}
-		plane_props = new VkDisplayPlanePropertiesKHR[plane_count];
-		vkGetPhysicalDeviceDisplayPlanePropertiesKHR(phys_device, &plane_count, plane_props);
-
-		// Get the Vulkan display we are going to use.	
-		VkDisplayKHR myDisplay = display_props[display_index].display;
-
-		// Get the list of display modes of the display
-		uint32_t mode_count = 0;
-		vkGetDisplayModePropertiesKHR(phys_device, myDisplay, &mode_count, NULL);
-		if (mode_count == 0) {
-			_assert_msg_(false, "DISPLAY Vulkan couldn't find any video modes on the display");
-			return VK_ERROR_INITIALIZATION_FAILED;
-		}
-		mode_props = new VkDisplayModePropertiesKHR[mode_count];
-		vkGetDisplayModePropertiesKHR(phys_device, myDisplay, &mode_count, mode_props);
-
-		// See if there's an appropiate mode available on the display 
-		display_mode = VK_NULL_HANDLE;
-		for (i = 0; i < mode_count; ++i)
-		{
-			const VkDisplayModePropertiesKHR* mode = &mode_props[i];
-
-			if (mode->parameters.visibleRegion.width == g_display.pixel_xres &&
-			    mode->parameters.visibleRegion.height == g_display.pixel_yres)
-			{
-				display_mode = mode->displayMode;
-				mode_found = true;
-				break;
-			}
-		}
-
-		// Free the mode list now.
-		delete [] mode_props;
-
-		// If there are no useable modes found on the display, error out
-		if (display_mode == VK_NULL_HANDLE)
-		{
-			_assert_msg_(false, "DISPLAY Vulkan couldn't find any video modes on the display");
-			return VK_ERROR_INITIALIZATION_FAILED;
-		}
-
-		/* Iterate on the list of planes of the physical device
-		   to find a plane that matches these criteria:
-		   -It must be compatible with the chosen display + mode.
-		   -It isn't currently bound to another display.
-		   -It supports per-pixel alpha, if possible. */
-		for (i = 0; i < plane_count; i++) {
-			uint32_t supported_displays_count = 0;
-			VkDisplayKHR* supported_displays;
-			VkDisplayPlaneCapabilitiesKHR plane_caps;
-
-			/* See if the plane is compatible with the current display. */
-			vkGetDisplayPlaneSupportedDisplaysKHR(phys_device, i, &supported_displays_count, NULL);
-			if (supported_displays_count == 0) {
-				/* This plane doesn't support any displays. Continue to the next plane. */
-				continue;
-			}
-
-			/* Get the list of displays supported by this plane. */
-			supported_displays = new VkDisplayKHR[supported_displays_count];
-			vkGetDisplayPlaneSupportedDisplaysKHR(phys_device, i,
-			    &supported_displays_count, supported_displays);
-
-			/* The plane must be bound to the chosen display, or not in use.
-			   If none of these is true, iterate to another plane. */
-			if ( !( (plane_props[i].currentDisplay == myDisplay) ||
-			        (plane_props[i].currentDisplay == VK_NULL_HANDLE))) 
-				continue;
-
-			/* Iterate the list of displays supported by this plane
-			   in order to find out if the chosen display is among them. */
-			bool plane_supports_display = false;
-			for (j = 0; j < supported_displays_count; j++) {
-				if (supported_displays[j] == myDisplay) {
-					plane_supports_display = true;
-					break;
-				}
-			}
-
-			/* Free the list of displays supported by this plane. */
-			delete [] supported_displays;
-
-			/* If the display is not supported by this plane, iterate to the next plane. */
-			if (!plane_supports_display)
-				continue;
-
-			/* Want a plane that supports the alpha mode we have chosen. */
-			vkGetDisplayPlaneCapabilitiesKHR(phys_device, display_mode, i, &plane_caps);
-			if (plane_caps.supportedAlpha & alpha_mode) {
-				/* Yep, this plane is alright. */
-				plane = i;
-				break;
-			}
-		}
-
-		/* If we couldn't find an appropiate plane, error out. */
-		if (plane == UINT32_MAX) {
-			_assert_msg_(false, "DISPLAY Vulkan couldn't find an appropiate plane");
-			return VK_ERROR_INITIALIZATION_FAILED;
-		}
-	
-		// Finally, create the vulkan surface.
-		image_size.width = g_display.pixel_xres;
-		image_size.height = g_display.pixel_yres;
-
-		display.displayMode = display_mode;
-		display.imageExtent = image_size;
-		display.transform = VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR;
-		display.alphaMode = alpha_mode;
-		display.globalAlpha = 1.0f;
-		display.planeIndex = plane;
-		display.planeStackIndex = plane_props[plane].currentStackIndex;
-		display.pNext = nullptr;
-		delete [] display_props;
-		delete [] plane_props;
-#endif
 		display.flags = 0;
 		retval = vkCreateDisplayPlaneSurfaceKHR(instance_, &display, nullptr, &surface_);
 		break;
@@ -1233,12 +1021,6 @@ static std::string surface_transforms_to_string(VkSurfaceTransformFlagsKHR trans
 }
 
 bool VulkanContext::InitSwapchain() {
-	_assert_(physical_device_ >= 0 && physical_device_ < physical_devices_.size());
-	if (!surface_) {
-		ERROR_LOG(G3D, "VK: No surface, can't create swapchain");
-		return false;
-	}
-
 	VkResult res = vkGetPhysicalDeviceSurfaceCapabilitiesKHR(physical_devices_[physical_device_], surface_, &surfCapabilities_);
 	if (res == VK_ERROR_SURFACE_LOST_KHR) {
 		// Not much to do.
@@ -1273,26 +1055,23 @@ bool VulkanContext::InitSwapchain() {
 		surfCapabilities_.maxImageExtent.width, surfCapabilities_.maxImageExtent.height,
 		swapChainExtent_.width, swapChainExtent_.height);
 
-	availablePresentModes_.clear();
 	// TODO: Find a better way to specify the prioritized present mode while being able
 	// to fall back in a sensible way.
 	VkPresentModeKHR swapchainPresentMode = VK_PRESENT_MODE_MAX_ENUM_KHR;
 	std::string modes = "";
 	for (size_t i = 0; i < presentModeCount; i++) {
-		modes += VulkanPresentModeToString(presentModes[i]);
+		modes += PresentModeString(presentModes[i]);
 		if (i != presentModeCount - 1) {
 			modes += ", ";
 		}
-		availablePresentModes_.push_back(presentModes[i]);
 	}
-
 	INFO_LOG(G3D, "Supported present modes: %s", modes.c_str());
 	for (size_t i = 0; i < presentModeCount; i++) {
 		bool match = false;
 		match = match || ((flags_ & VULKAN_FLAG_PRESENT_MAILBOX) && presentModes[i] == VK_PRESENT_MODE_MAILBOX_KHR);
-		match = match || ((flags_ & VULKAN_FLAG_PRESENT_IMMEDIATE) && presentModes[i] == VK_PRESENT_MODE_IMMEDIATE_KHR);
 		match = match || ((flags_ & VULKAN_FLAG_PRESENT_FIFO_RELAXED) && presentModes[i] == VK_PRESENT_MODE_FIFO_RELAXED_KHR);
 		match = match || ((flags_ & VULKAN_FLAG_PRESENT_FIFO) && presentModes[i] == VK_PRESENT_MODE_FIFO_KHR);
+		match = match || ((flags_ & VULKAN_FLAG_PRESENT_IMMEDIATE) && presentModes[i] == VK_PRESENT_MODE_IMMEDIATE_KHR);
 
 		// Default to the first present mode from the list.
 		if (match || swapchainPresentMode == VK_PRESENT_MODE_MAX_ENUM_KHR) {
@@ -1302,6 +1081,10 @@ bool VulkanContext::InitSwapchain() {
 			break;
 		}
 	}
+#ifdef __ANDROID__
+	// HACK
+	swapchainPresentMode = VK_PRESENT_MODE_FIFO_KHR;
+#endif
 	delete[] presentModes;
 	// Determine the number of VkImage's to use in the swap chain (we desire to
 	// own only 1 image at a time, besides the images being displayed and
@@ -1315,7 +1098,7 @@ bool VulkanContext::InitSwapchain() {
 	}
 
 	INFO_LOG(G3D, "Chosen present mode: %d (%s). numSwapChainImages: %d/%d",
-		swapchainPresentMode, VulkanPresentModeToString(swapchainPresentMode),
+		swapchainPresentMode, PresentModeString(swapchainPresentMode),
 		desiredNumberOfSwapChainImages, surfCapabilities_.maxImageCount);
 
 	// We mostly follow the practices from
@@ -1324,8 +1107,8 @@ bool VulkanContext::InitSwapchain() {
 	VkSurfaceTransformFlagBitsKHR preTransform;
 	std::string supportedTransforms = surface_transforms_to_string(surfCapabilities_.supportedTransforms);
 	std::string currentTransform = surface_transforms_to_string(surfCapabilities_.currentTransform);
-	g_display.rotation = DisplayRotation::ROTATE_0;
+	g_display_rotation = DisplayRotation::ROTATE_0;
-	g_display.rot_matrix.setIdentity();
+	g_display_rot_matrix.setIdentity();
 
 	uint32_t allowedRotations = VK_SURFACE_TRANSFORM_ROTATE_90_BIT_KHR | VK_SURFACE_TRANSFORM_ROTATE_180_BIT_KHR | VK_SURFACE_TRANSFORM_ROTATE_270_BIT_KHR;
 	// Hack: Don't allow 270 degrees pretransform (inverse landscape), it creates bizarre issues on some devices (see #15773).
@@ -1336,20 +1119,20 @@ bool VulkanContext::InitSwapchain() {
 	} else if (surfCapabilities_.currentTransform & allowedRotations) {
 		// Normal, sensible rotations. Let's handle it.
 		preTransform = surfCapabilities_.currentTransform;
-		g_display.rot_matrix.setIdentity();
+		g_display_rot_matrix.setIdentity();
 		switch (surfCapabilities_.currentTransform) {
 		case VK_SURFACE_TRANSFORM_ROTATE_90_BIT_KHR:
-			g_display.rotation = DisplayRotation::ROTATE_90;
+			g_display_rotation = DisplayRotation::ROTATE_90;
-			g_display.rot_matrix.setRotationZ90();
+			g_display_rot_matrix.setRotationZ90();
 			std::swap(swapChainExtent_.width, swapChainExtent_.height);
 			break;
 		case VK_SURFACE_TRANSFORM_ROTATE_180_BIT_KHR:
-			g_display.rotation = DisplayRotation::ROTATE_180;
+			g_display_rotation = DisplayRotation::ROTATE_180;
-			g_display.rot_matrix.setRotationZ180();
+			g_display_rot_matrix.setRotationZ180();
 			break;
 		case VK_SURFACE_TRANSFORM_ROTATE_270_BIT_KHR:
-			g_display.rotation = DisplayRotation::ROTATE_270;
+			g_display_rotation = DisplayRotation::ROTATE_270;
-			g_display.rot_matrix.setRotationZ270();
+			g_display_rot_matrix.setRotationZ270();
 			std::swap(swapChainExtent_.width, swapChainExtent_.height);
 			break;
 		default:
@@ -1395,8 +1178,6 @@ bool VulkanContext::InitSwapchain() {
 	swap_chain_info.clipped = true;
 	swap_chain_info.imageUsage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
 
-	presentMode_ = swapchainPresentMode;
-
 	// Don't ask for TRANSFER_DST for the swapchain image, we don't use that.
 	// if (surfCapabilities_.supportedUsageFlags & VK_IMAGE_USAGE_TRANSFER_DST_BIT)
 	//	swap_chain_info.imageUsage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
@@ -1666,101 +1447,80 @@ void VulkanDeleteList::Take(VulkanDeleteList &del) {
 }
 
 void VulkanDeleteList::PerformDeletes(VulkanContext *vulkan, VmaAllocator allocator) {
-	int deleteCount = 0;
-
 	for (auto &callback : callbacks_) {
 		callback.func(vulkan, callback.userdata);
-		deleteCount++;
 	}
 	callbacks_.clear();
 
 	VkDevice device = vulkan->GetDevice();
 	for (auto &cmdPool : cmdPools_) {
 		vkDestroyCommandPool(device, cmdPool, nullptr);
-		deleteCount++;
 	}
 	cmdPools_.clear();
 	for (auto &descPool : descPools_) {
 		vkDestroyDescriptorPool(device, descPool, nullptr);
-		deleteCount++;
 	}
 	descPools_.clear();
 	for (auto &module : modules_) {
 		vkDestroyShaderModule(device, module, nullptr);
-		deleteCount++;
 	}
 	modules_.clear();
 	for (auto &buf : buffers_) {
 		vkDestroyBuffer(device, buf, nullptr);
-		deleteCount++;
 	}
 	buffers_.clear();
 	for (auto &buf : buffersWithAllocs_) {
 		vmaDestroyBuffer(allocator, buf.buffer, buf.alloc);
-		deleteCount++;
 	}
 	buffersWithAllocs_.clear();
 	for (auto &bufView : bufferViews_) {
 		vkDestroyBufferView(device, bufView, nullptr);
-		deleteCount++;
 	}
 	bufferViews_.clear();
 	for (auto &imageWithAlloc : imagesWithAllocs_) {
 		vmaDestroyImage(allocator, imageWithAlloc.image, imageWithAlloc.alloc);
-		deleteCount++;
 	}
 	imagesWithAllocs_.clear();
 	for (auto &imageView : imageViews_) {
 		vkDestroyImageView(device, imageView, nullptr);
-		deleteCount++;
 	}
 	imageViews_.clear();
 	for (auto &mem : deviceMemory_) {
 		vkFreeMemory(device, mem, nullptr);
-		deleteCount++;
 	}
 	deviceMemory_.clear();
 	for (auto &sampler : samplers_) {
 		vkDestroySampler(device, sampler, nullptr);
-		deleteCount++;
 	}
 	samplers_.clear();
 	for (auto &pipeline : pipelines_) {
 		vkDestroyPipeline(device, pipeline, nullptr);
-		deleteCount++;
 	}
 	pipelines_.clear();
 	for (auto &pcache : pipelineCaches_) {
 		vkDestroyPipelineCache(device, pcache, nullptr);
-		deleteCount++;
 	}
 	pipelineCaches_.clear();
 	for (auto &renderPass : renderPasses_) {
 		vkDestroyRenderPass(device, renderPass, nullptr);
-		deleteCount++;
 	}
 	renderPasses_.clear();
 	for (auto &framebuffer : framebuffers_) {
 		vkDestroyFramebuffer(device, framebuffer, nullptr);
-		deleteCount++;
 	}
 	framebuffers_.clear();
 	for (auto &pipeLayout : pipelineLayouts_) {
 		vkDestroyPipelineLayout(device, pipeLayout, nullptr);
-		deleteCount++;
 	}
 	pipelineLayouts_.clear();
 	for (auto &descSetLayout : descSetLayouts_) {
 		vkDestroyDescriptorSetLayout(device, descSetLayout, nullptr);
-		deleteCount++;
 	}
 	descSetLayouts_.clear();
 	for (auto &queryPool : queryPools_) {
 		vkDestroyQueryPool(device, queryPool, nullptr);
-		deleteCount++;
 	}
 	queryPools_.clear();
-	deleteCount_ = deleteCount;
 }
 
 void VulkanContext::GetImageMemoryRequirements(VkImage image, VkMemoryRequirements *mem_reqs, bool *dedicatedAllocation) {

Common/GPU/Vulkan/VulkanContext.h

@@ -36,7 +36,6 @@ enum {
 	VULKAN_VENDOR_QUALCOMM = 0x00005143,
 	VULKAN_VENDOR_IMGTEC = 0x00001010,  // PowerVR
 	VULKAN_VENDOR_APPLE = 0x0000106b,  // Apple through MoltenVK
-	VULKAN_VENDOR_MESA = 0x00010005, // lavapipe
 };
 
 VK_DEFINE_HANDLE(VmaAllocator);
@@ -138,10 +137,6 @@ public:
 	void Take(VulkanDeleteList &del);
 	void PerformDeletes(VulkanContext *vulkan, VmaAllocator allocator);
 
-	int GetLastDeleteCount() const {
-		return deleteCount_;
-	}
-
 private:
 	std::vector<VkCommandPool> cmdPools_;
 	std::vector<VkDescriptorPool> descPools_;
@@ -161,7 +156,6 @@ private:
 	std::vector<VkDescriptorSetLayout> descSetLayouts_;
 	std::vector<VkQueryPool> queryPools_;
 	std::vector<Callback> callbacks_;
-	int deleteCount_ = 0;
 };
 
 // VulkanContext manages the device and swapchain, and deferred deletion of objects.
@@ -228,8 +222,7 @@ public:
 	// Simple workaround for the casting warning.
 	template <class T>
 	void SetDebugName(T handle, VkObjectType type, const char *name) {
-		if (extensionsLookup_.EXT_debug_utils && handle != VK_NULL_HANDLE) {
+		if (extensionsLookup_.EXT_debug_utils) {
-			_dbg_assert_(handle != VK_NULL_HANDLE);
 			SetDebugNameImpl((uint64_t)handle, type, name);
 		}
 	}
@@ -270,8 +263,6 @@ public:
 	struct AllPhysicalDeviceFeatures {
 		VkPhysicalDeviceFeatures standard;
 		VkPhysicalDeviceMultiviewFeatures multiview;
-		VkPhysicalDevicePresentWaitFeaturesKHR presentWait;
-		VkPhysicalDevicePresentIdFeaturesKHR presentId;
 	};
 
 	const PhysicalDeviceProps &GetPhysicalDeviceProperties(int i = -1) const {
@@ -297,13 +288,6 @@ public:
 		return device_extensions_enabled_;
 	}
 
-	const std::vector<VkExtensionProperties> &GetInstanceExtensionsAvailable() const {
-		return instance_extension_properties_;
-	}
-	const std::vector<const char *> &GetInstanceExtensionsEnabled() const {
-		return instance_extensions_enabled_;
-	}
-
 	const VkPhysicalDeviceMemoryProperties &GetMemoryProperties() const {
 		return memory_properties_;
 	}
@@ -327,6 +311,7 @@ public:
 		for (const auto &iter : instance_layer_properties_) {
 			for (const auto &ext : iter.extensions) {
 				if (!strcmp(extensionName, ext.extensionName)) {
+					INFO_LOG(G3D, "%s found in layer extensions: %s", extensionName, iter.properties.layerName);
 					return true;
 				}
 			}
@@ -344,7 +329,6 @@ public:
 	}
 
 	int GetInflightFrames() const {
-		// out of MAX_INFLIGHT_FRAMES.
 		return inflightFrames_;
 	}
 	// Don't call while a frame is in progress.
@@ -389,18 +373,6 @@ public:
 		return surfFormats_;
 	}
 
-	VkPresentModeKHR GetPresentMode() const {
-		return presentMode_;
-	}
-
-	std::vector<VkPresentModeKHR> GetAvailablePresentModes() const {
-		return availablePresentModes_;
-	}
-
-	int GetLastDeleteCount() const {
-		return frame_[curFrame_].deleteList.GetLastDeleteCount();
-	}
-
 private:
 	bool ChooseQueue();
 
@@ -488,9 +460,6 @@ private:
 	VkSurfaceCapabilitiesKHR surfCapabilities_{};
 	std::vector<VkSurfaceFormatKHR> surfFormats_{};
 
-	VkPresentModeKHR presentMode_;
-	std::vector<VkPresentModeKHR> availablePresentModes_;
-
 	std::vector<VkCommandBuffer> cmdQueue_;
 
 	VmaAllocator allocator_ = VK_NULL_HANDLE;
@@ -516,7 +485,6 @@ bool GLSLtoSPV(const VkShaderStageFlagBits shader_type, const char *sourceCode,
 
 const char *VulkanColorSpaceToString(VkColorSpaceKHR colorSpace);
 const char *VulkanFormatToString(VkFormat format);
-const char *VulkanPresentModeToString(VkPresentModeKHR presentMode);
 
 std::string FormatDriverVersion(const VkPhysicalDeviceProperties &props);
 

Common/GPU/Vulkan/VulkanDebug.cpp

@@ -21,7 +21,6 @@
 #include <mutex>
 
 #include "Common/Log.h"
-#include "Common/System/System.h"
 #include "Common/GPU/Vulkan/VulkanContext.h"
 #include "Common/GPU/Vulkan/VulkanDebug.h"
 
@@ -48,62 +47,22 @@ VKAPI_ATTR VkBool32 VKAPI_CALL VulkanDebugUtilsCallback(
 	const char *pMessage = pCallbackData->pMessage;
 
 	int messageCode = pCallbackData->messageIdNumber;
-	switch (messageCode) {
+	if (messageCode == 101294395) {
-	case 101294395:
 		// UNASSIGNED-CoreValidation-Shader-OutputNotConsumed - benign perf warning
 		return false;
-	case 1303270965:
+	}
+	if (messageCode == 1303270965) {
 		// Benign perf warning, image blit using GENERAL layout.
-		// TODO: Oops, turns out we filtered out a bit too much here!
+		// UNASSIGNED
-		// We really need that performance flag check to sort out the stuff that matters.
+		return false;
-		// Will enable it soon, but it'll take some fixing.
+	}
-		//
+	if (messageCode == 606910136 || messageCode == -392708513 || messageCode == -384083808) {
-		if (messageType & VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT)
-			return false;
-		break;
-
-	case 606910136:
-	case -392708513:
-	case -384083808:
 		// VUID-vkCmdDraw-None-02686
 		// Kinda false positive, or at least very unnecessary, now that I solved the real issue.
 		// See https://github.com/hrydgard/ppsspp/pull/16354
 		return false;
-	case -375211665:
-		// VUID-vkAllocateMemory-pAllocateInfo-01713
-		// Can happen when VMA aggressively tries to allocate aperture memory for upload. It gracefully
-		// falls back to regular video memory, so we just ignore this. I'd argue this is a VMA bug, actually.
-		return false;
-	case 181611958:
-		// Extended validation.
-		// UNASSIGNED-BestPractices-vkCreateDevice-deprecated-extension
-		// Doing what this one says doesn't seem very reliable - if I rely strictly on the Vulkan version, I don't get some function pointers? Like createrenderpass2.
-		return false;
-	case 657182421:
-		// Extended validation (ARM best practices)
-		// Non-fifo validation not recommended
-		return false;
-	case 337425955:
-		// False positive
-		// https://github.com/KhronosGroup/Vulkan-ValidationLayers/issues/3615
-		return false;
-	default:
-		break;
 	}
 
-	/*
-	// Can be used to temporarily turn errors into info for easier debugging.
-	switch (messageCode) {
-	case 1544472022:
-		if (messageSeverity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT) {
-			messageSeverity = (VkDebugUtilsMessageSeverityFlagBitsEXT)((messageSeverity & ~VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT) | VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT);
-		}
-		break;
-	default:
-		break;
-	}
-	*/
-
 	int count;
 	{
 		std::lock_guard<std::mutex> lock(g_errorCountMutex);
@@ -140,7 +99,7 @@ VKAPI_ATTR VkBool32 VKAPI_CALL VulkanDebugUtilsCallback(
 #ifdef _WIN32
 	OutputDebugStringA(msg.c_str());
 	if (messageSeverity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT) {
-		if (options->breakOnError && System_GetPropertyBool(SYSPROP_DEBUGGER_PRESENT)) {
+		if (options->breakOnError && IsDebuggerPresent()) {
 			DebugBreak();
 		}
 		if (options->msgBoxOnError) {
@@ -148,7 +107,7 @@ VKAPI_ATTR VkBool32 VKAPI_CALL VulkanDebugUtilsCallback(
 		}
 	} else if (messageSeverity & VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT) {
 		// Don't break on perf warnings for now, even with a debugger. We log them at least.
-		if (options->breakOnWarning && System_GetPropertyBool(SYSPROP_DEBUGGER_PRESENT) && 0 == (messageType & VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT)) {
+		if (options->breakOnWarning && IsDebuggerPresent() && 0 == (messageType & VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT)) {
 			DebugBreak();
 		}
 	}

Common/GPU/Vulkan/VulkanDescSet.cpp

@@ -1,129 +0,0 @@
-#include "Common/GPU/Vulkan/VulkanDescSet.h"
-
-VulkanDescSetPool::~VulkanDescSetPool() {
-	_assert_msg_(descPool_ == VK_NULL_HANDLE, "VulkanDescSetPool %s never destroyed", tag_);
-}
-
-void VulkanDescSetPool::Create(VulkanContext *vulkan, const BindingType *bindingTypes, uint32_t bindingTypesCount, uint32_t descriptorCount) {
-	_assert_msg_(descPool_ == VK_NULL_HANDLE, "VulkanDescSetPool::Create when already exists");
-
-	vulkan_ = vulkan;
-	info_ = { VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO };
-	info_.maxSets = descriptorCount;
-	_dbg_assert_(sizes_.empty());
-
-	uint32_t storageImageCount = 0;
-	uint32_t storageBufferCount = 0;
-	uint32_t combinedImageSamplerCount = 0;
-	uint32_t uniformBufferDynamicCount = 0;
-	for (uint32_t i = 0; i < bindingTypesCount; i++) {
-		switch (bindingTypes[i]) {
-		case BindingType::COMBINED_IMAGE_SAMPLER: combinedImageSamplerCount++; break;
-		case BindingType::UNIFORM_BUFFER_DYNAMIC_VERTEX:
-		case BindingType::UNIFORM_BUFFER_DYNAMIC_ALL: uniformBufferDynamicCount++; break;
-		case BindingType::STORAGE_BUFFER_VERTEX:
-		case BindingType::STORAGE_BUFFER_COMPUTE: storageBufferCount++; break;
-		case BindingType::STORAGE_IMAGE_COMPUTE: storageImageCount++; break;
-		}
-	}
-	if (combinedImageSamplerCount) {
-		sizes_.push_back(VkDescriptorPoolSize{ VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, combinedImageSamplerCount * descriptorCount });
-	}
-	if (uniformBufferDynamicCount) {
-		sizes_.push_back(VkDescriptorPoolSize{ VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC, uniformBufferDynamicCount * descriptorCount });
-	}
-	if (storageBufferCount) {
-		sizes_.push_back(VkDescriptorPoolSize{ VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, storageBufferCount * descriptorCount });
-	}
-	if (storageImageCount) {
-		sizes_.push_back(VkDescriptorPoolSize{ VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, storageImageCount * descriptorCount });
-	}
-	VkResult res = Recreate(false);
-	_assert_msg_(res == VK_SUCCESS, "Could not create VulkanDescSetPool %s", tag_);
-}
-
-bool VulkanDescSetPool::Allocate(VkDescriptorSet *descriptorSets, int count, const VkDescriptorSetLayout *layouts) {
-	if (descPool_ == VK_NULL_HANDLE || usage_ + count >= info_.maxSets) {
-		// Missing or out of space, need to recreate.
-		VkResult res = Recreate(grow_);
-		_assert_msg_(res == VK_SUCCESS, "Could not grow VulkanDescSetPool %s on usage %d", tag_, (int)usage_);
-	}
-
-	VkDescriptorSetAllocateInfo descAlloc{ VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO };
-	descAlloc.descriptorPool = descPool_;
-	descAlloc.descriptorSetCount = count;
-	descAlloc.pSetLayouts = layouts;
-	VkResult result = vkAllocateDescriptorSets(vulkan_->GetDevice(), &descAlloc, descriptorSets);
-
-	if (result == VK_ERROR_FRAGMENTED_POOL || result < 0) {
-		WARN_LOG(G3D, "Pool %s %s - recreating", tag_, result == VK_ERROR_FRAGMENTED_POOL ? "fragmented" : "full");
-		// There seems to have been a spec revision. Here we should apparently recreate the descriptor pool,
-		// so let's do that. See https://www.khronos.org/registry/vulkan/specs/1.0/man/html/vkAllocateDescriptorSets.html
-		// Fragmentation shouldn't really happen though since we wipe the pool every frame.
-		VkResult res = Recreate(false);
-		_assert_msg_(res == VK_SUCCESS, "Ran out of descriptor space (frag?) and failed to recreate a descriptor pool. sz=%d res=%d", usage_, (int)res);
-
-		// Need to update this pointer since we have allocated a new one.
-		descAlloc.descriptorPool = descPool_;
-		result = vkAllocateDescriptorSets(vulkan_->GetDevice(), &descAlloc, descriptorSets);
-		_assert_msg_(result == VK_SUCCESS, "Ran out of descriptor space (frag?) and failed to allocate after recreating a descriptor pool. res=%d", (int)result);
-	}
-
-	if (result != VK_SUCCESS) {
-		return false;
-	}
-
-	usage_ += count;
-	return true;
-}
-
-void VulkanDescSetPool::Reset() {
-	_assert_msg_(descPool_ != VK_NULL_HANDLE, "VulkanDescSetPool::Reset without valid pool");
-	vkResetDescriptorPool(vulkan_->GetDevice(), descPool_, 0);
-
-	usage_ = 0;
-}
-
-void VulkanDescSetPool::Destroy() {
-	if (descPool_ != VK_NULL_HANDLE) {
-		vulkan_->Delete().QueueDeleteDescriptorPool(descPool_);
-		usage_ = 0;
-	}
-	sizes_.clear();
-}
-
-void VulkanDescSetPool::DestroyImmediately() {
-	if (descPool_ != VK_NULL_HANDLE) {
-		vkDestroyDescriptorPool(vulkan_->GetDevice(), descPool_, nullptr);
-		descPool_ = VK_NULL_HANDLE;
-		usage_ = 0;
-	}
-	sizes_.clear();
-}
-
-VkResult VulkanDescSetPool::Recreate(bool grow) {
-	_assert_msg_(vulkan_ != nullptr, "VulkanDescSetPool::Recreate without VulkanContext");
-
-	uint32_t prevSize = info_.maxSets;
-	if (grow) {
-		info_.maxSets *= 2;
-		for (auto &size : sizes_)
-			size.descriptorCount *= 2;
-	}
-
-	// Delete the pool if it already exists.
-	if (descPool_ != VK_NULL_HANDLE) {
-		INFO_LOG(G3D, "Reallocating %s desc pool from %d to %d", tag_, prevSize, info_.maxSets);
-		vulkan_->Delete().QueueDeleteDescriptorPool(descPool_);
-		usage_ = 0;
-	}
-
-	info_.pPoolSizes = &sizes_[0];
-	info_.poolSizeCount = (uint32_t)sizes_.size();
-
-	VkResult result = vkCreateDescriptorPool(vulkan_->GetDevice(), &info_, nullptr, &descPool_);
-	if (result == VK_SUCCESS) {
-		vulkan_->SetDebugName(descPool_, VK_OBJECT_TYPE_DESCRIPTOR_POOL, tag_);
-	}
-	return result;
-}

Common/GPU/Vulkan/VulkanDescSet.h

@@ -1,48 +0,0 @@
-#pragma once
-
-#include "Common/Data/Collections/FastVec.h"
-#include "Common/GPU/Vulkan/VulkanContext.h"
-
-#include <vector>
-
-enum class BindingType {
-	COMBINED_IMAGE_SAMPLER,
-	UNIFORM_BUFFER_DYNAMIC_VERTEX,
-	UNIFORM_BUFFER_DYNAMIC_ALL,
-	STORAGE_BUFFER_VERTEX,
-	STORAGE_BUFFER_COMPUTE,
-	STORAGE_IMAGE_COMPUTE,
-};
-
-// Only appropriate for use in a per-frame pool.
-class VulkanDescSetPool {
-public:
-	VulkanDescSetPool(const char *tag, bool grow = true) : tag_(tag), grow_(grow) {}
-	~VulkanDescSetPool();
-
-	void Create(VulkanContext *vulkan, const BindingType *bindingTypes, uint32_t bindingTypesCount, uint32_t descriptorCount);
-	// Allocate a new set, which may resize and empty the current sets.
-	// Use only for the current frame.
-	bool Allocate(VkDescriptorSet *descriptorSets, int count, const VkDescriptorSetLayout *layouts);
-	void Reset();
-
-	// This queues up destruction.
-	void Destroy();
-	// This actually destroys immediately.
-	void DestroyImmediately();
-
-	bool IsDestroyed() const {
-		return !descPool_;
-	}
-
-private:
-	VkResult Recreate(bool grow);
-
-	const char *tag_;
-	VulkanContext *vulkan_ = nullptr;
-	VkDescriptorPool descPool_ = VK_NULL_HANDLE;
-	VkDescriptorPoolCreateInfo info_{};
-	std::vector<VkDescriptorPoolSize> sizes_;
-	uint32_t usage_ = 0;
-	bool grow_;
-};

Common/GPU/Vulkan/VulkanFrameData.cpp

@@ -4,19 +4,6 @@
 #include "Common/Log.h"
 #include "Common/StringUtils.h"
 
-#if 0 // def _DEBUG
-#define VLOG(...) NOTICE_LOG(G3D, __VA_ARGS__)
-#else
-#define VLOG(...)
-#endif
-
-void CachedReadback::Destroy(VulkanContext *vulkan) {
-	if (buffer) {
-		vulkan->Delete().QueueDeleteBufferAllocation(buffer, allocation);
-	}
-	bufferSize = 0;
-}
-
 void FrameData::Init(VulkanContext *vulkan, int index) {
 	this->index = index;
 	VkDevice device = vulkan->GetDevice();
@@ -49,6 +36,11 @@ void FrameData::Init(VulkanContext *vulkan, int index) {
 	vulkan->SetDebugName(fence, VK_OBJECT_TYPE_FENCE, StringFromFormat("fence%d", index).c_str());
 	readyForFence = true;
 
+	// This fence is used for synchronizing readbacks. Does not need preinitialization.
+	// TODO: Put this in frameDataShared, only one is needed.
+	readbackFence = vulkan->CreateFence(false);
+	vulkan->SetDebugName(fence, VK_OBJECT_TYPE_FENCE, "readbackFence");
+
 	VkQueryPoolCreateInfo query_ci{ VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO };
 	query_ci.queryCount = MAX_TIMESTAMP_QUERIES;
 	query_ci.queryType = VK_QUERY_TYPE_TIMESTAMP;
@@ -60,13 +52,8 @@ void FrameData::Destroy(VulkanContext *vulkan) {
 	vkDestroyCommandPool(device, cmdPoolInit, nullptr);
 	vkDestroyCommandPool(device, cmdPoolMain, nullptr);
 	vkDestroyFence(device, fence, nullptr);
+	vkDestroyFence(device, readbackFence, nullptr);
 	vkDestroyQueryPool(device, profile.queryPool, nullptr);
-
-	readbacks_.IterateMut([=](const ReadbackKey &key, CachedReadback *value) {
-		value->Destroy(vulkan);
-		delete value;
-	});
-	readbacks_.Clear();
 }
 
 void FrameData::AcquireNextImage(VulkanContext *vulkan, FrameDataShared &shared) {
@@ -90,10 +77,6 @@ void FrameData::AcquireNextImage(VulkanContext *vulkan, FrameDataShared &shared)
 		WARN_LOG(G3D, "%s returned from AcquireNextImage - processing the frame, but not presenting", VulkanResultToString(res));
 		skipSwap = true;
 		break;
-	case VK_ERROR_SURFACE_LOST_KHR:
-		ERROR_LOG(G3D, "%s returned from AcquireNextImage - ignoring, but this better be during shutdown", VulkanResultToString(res));
-		skipSwap = true;
-		break;
 	default:
 		// Weird, shouldn't get any other values. Maybe lost device?
 		_assert_msg_(false, "vkAcquireNextImageKHR failed! result=%s", VulkanResultToString(res));
@@ -114,25 +97,6 @@ VkResult FrameData::QueuePresent(VulkanContext *vulkan, FrameDataShared &shared)
 	present.pWaitSemaphores = &shared.renderingCompleteSemaphore;
 	present.waitSemaphoreCount = 1;
 
-	// Can't move these into the if.
-	VkPresentIdKHR presentID{ VK_STRUCTURE_TYPE_PRESENT_ID_KHR };
-	VkPresentTimesInfoGOOGLE presentGOOGLE{ VK_STRUCTURE_TYPE_PRESENT_TIMES_INFO_GOOGLE };
-
-	uint64_t frameId = this->frameId;
-	VkPresentTimeGOOGLE presentTimeGOOGLE{ (uint32_t)frameId, 0 };  // it's ok to truncate this. it'll wrap around and work (if we ever reach 4 billion frames..)
-
-	if (shared.measurePresentTime) {
-		if (vulkan->Extensions().KHR_present_id && vulkan->GetDeviceFeatures().enabled.presentId.presentId) {
-			presentID.pPresentIds = &frameId;
-			presentID.swapchainCount = 1;
-			present.pNext = &presentID;
-		} else if (vulkan->Extensions().GOOGLE_display_timing) {
-			presentGOOGLE.pTimes = &presentTimeGOOGLE;
-			presentGOOGLE.swapchainCount = 1;
-			present.pNext = &presentGOOGLE;
-		}
-	}
-
 	return vkQueuePresentKHR(vulkan->GetGraphicsQueue(), &present);
 }
 
@@ -150,7 +114,7 @@ VkCommandBuffer FrameData::GetInitCmd(VulkanContext *vulkan) {
 		}
 
 		// Good spot to reset the query pool.
-		if (profile.enabled) {
+		if (profilingEnabled_) {
 			vkCmdResetQueryPool(initCmd, profile.queryPool, 0, MAX_TIMESTAMP_QUERIES);
 			vkCmdWriteTimestamp(initCmd, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, profile.queryPool, 0);
 		}
@@ -167,7 +131,7 @@ void FrameData::SubmitPending(VulkanContext *vulkan, FrameSubmitType type, Frame
 	VkFence fenceToTrigger = VK_NULL_HANDLE;
 
 	if (hasInitCommands) {
-		if (profile.enabled) {
+		if (profilingEnabled_) {
 			// Pre-allocated query ID 1 - end of init cmdbuf.
 			vkCmdWriteTimestamp(initCmd, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, profile.queryPool, 1);
 		}
@@ -180,7 +144,7 @@ void FrameData::SubmitPending(VulkanContext *vulkan, FrameSubmitType type, Frame
 	}
 
 	if ((hasMainCommands || hasPresentCommands) && type == FrameSubmitType::Sync) {
-		fenceToTrigger = sharedData.readbackFence;
+		fenceToTrigger = readbackFence;
 	}
 
 	if (hasMainCommands) {
@@ -225,16 +189,12 @@ void FrameData::SubmitPending(VulkanContext *vulkan, FrameSubmitType type, Frame
 
 	VkResult res;
 	if (fenceToTrigger == fence) {
-		VLOG("Doing queue submit, fencing frame %d", this->index);
 		// The fence is waited on by the main thread, they are not allowed to access it simultaneously.
 		res = vkQueueSubmit(vulkan->GetGraphicsQueue(), 1, &submit_info, fenceToTrigger);
-		if (sharedData.useMultiThreading) {
+		std::lock_guard<std::mutex> lock(fenceMutex);
-			std::lock_guard<std::mutex> lock(fenceMutex);
+		readyForFence = true;
-			readyForFence = true;
+		fenceCondVar.notify_one();
-			fenceCondVar.notify_one();
-		}
 	} else {
-		VLOG("Doing queue submit, fencing something (%p)", fenceToTrigger);
 		res = vkQueueSubmit(vulkan->GetGraphicsQueue(), 1, &submit_info, fenceToTrigger);
 	}
 
@@ -246,31 +206,23 @@ void FrameData::SubmitPending(VulkanContext *vulkan, FrameSubmitType type, Frame
 
 	if (type == FrameSubmitType::Sync) {
 		// Hard stall of the GPU, not ideal, but necessary so the CPU has the contents of the readback.
-		vkWaitForFences(vulkan->GetDevice(), 1, &sharedData.readbackFence, true, UINT64_MAX);
+		vkWaitForFences(vulkan->GetDevice(), 1, &readbackFence, true, UINT64_MAX);
-		vkResetFences(vulkan->GetDevice(), 1, &sharedData.readbackFence);
+		vkResetFences(vulkan->GetDevice(), 1, &readbackFence);
 		syncDone = true;
 	}
 }
 
-void FrameDataShared::Init(VulkanContext *vulkan, bool useMultiThreading, bool measurePresentTime) {
+void FrameDataShared::Init(VulkanContext *vulkan) {
 	VkSemaphoreCreateInfo semaphoreCreateInfo = { VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO };
 	semaphoreCreateInfo.flags = 0;
 	VkResult res = vkCreateSemaphore(vulkan->GetDevice(), &semaphoreCreateInfo, nullptr, &acquireSemaphore);
 	_dbg_assert_(res == VK_SUCCESS);
 	res = vkCreateSemaphore(vulkan->GetDevice(), &semaphoreCreateInfo, nullptr, &renderingCompleteSemaphore);
 	_dbg_assert_(res == VK_SUCCESS);
-
-	// This fence is used for synchronizing readbacks. Does not need preinitialization.
-	readbackFence = vulkan->CreateFence(false);
-	vulkan->SetDebugName(readbackFence, VK_OBJECT_TYPE_FENCE, "readbackFence");
-
-	this->useMultiThreading = useMultiThreading;
-	this->measurePresentTime = measurePresentTime;
 }
 
 void FrameDataShared::Destroy(VulkanContext *vulkan) {
 	VkDevice device = vulkan->GetDevice();
 	vkDestroySemaphore(device, acquireSemaphore, nullptr);
 	vkDestroySemaphore(device, renderingCompleteSemaphore, nullptr);
-	vkDestroyFence(device, readbackFence, nullptr);
 }

Common/GPU/Vulkan/VulkanFrameData.h

@@ -6,46 +6,23 @@
 #include <condition_variable>
 
 #include "Common/GPU/Vulkan/VulkanContext.h"
-#include "Common/Data/Collections/Hashmaps.h"
 
 enum {
 	MAX_TIMESTAMP_QUERIES = 128,
 };
 
 enum class VKRRunType {
-	SUBMIT,
 	PRESENT,
 	SYNC,
 	EXIT,
 };
 
 struct QueueProfileContext {
-	bool enabled = false;
-	bool timestampsEnabled = false;
 	VkQueryPool queryPool;
 	std::vector<std::string> timestampDescriptions;
 	std::string profileSummary;
 	double cpuStartTime;
 	double cpuEndTime;
-	double descWriteTime;
-	int descriptorsWritten;
-};
-
-class VKRFramebuffer;
-
-struct ReadbackKey {
-	const VKRFramebuffer *framebuf;
-	int width;
-	int height;
-};
-
-struct CachedReadback {
-	VkBuffer buffer;
-	VmaAllocation allocation;
-	VkDeviceSize bufferSize;
-	bool isCoherent;
-
-	void Destroy(VulkanContext *vulkan);
 };
 
 struct FrameDataShared {
@@ -53,12 +30,7 @@ struct FrameDataShared {
 	VkSemaphore acquireSemaphore = VK_NULL_HANDLE;
 	VkSemaphore renderingCompleteSemaphore = VK_NULL_HANDLE;
 
-	// For synchronous readbacks.
+	void Init(VulkanContext *vulkan);
-	VkFence readbackFence = VK_NULL_HANDLE;
-	bool useMultiThreading;
-	bool measurePresentTime;
-
-	void Init(VulkanContext *vulkan, bool useMultiThreading, bool measurePresentTime);
 	void Destroy(VulkanContext *vulkan);
 };
 
@@ -77,6 +49,7 @@ struct FrameData {
 	bool readyForFence = true;
 
 	VkFence fence = VK_NULL_HANDLE;
+	VkFence readbackFence = VK_NULL_HANDLE;  // Strictly speaking we might only need one global of these.
 
 	// These are on different threads so need separate pools.
 	VkCommandPool cmdPoolInit = VK_NULL_HANDLE;  // Written to from main thread
@@ -98,17 +71,9 @@ struct FrameData {
 	// Swapchain.
 	uint32_t curSwapchainImage = -1;
 
-	// Frames need unique IDs to wait for present on, let's keep them here.
-	// Also used for indexing into the frame timing history buffer.
-	uint64_t frameId = 0;
-
 	// Profiling.
-	QueueProfileContext profile{};
+	QueueProfileContext profile;
-
+	bool profilingEnabled_ = false;
-	// Async readback cache.
-	DenseHashMap<ReadbackKey, CachedReadback *> readbacks_;
-
-	FrameData() : readbacks_(8) {}
 
 	void Init(VulkanContext *vulkan, int index);
 	void Destroy(VulkanContext *vulkan);
@@ -124,5 +89,5 @@ struct FrameData {
 
 private:
 	// Metadata for logging etc
-	int index = -1;
+	int index;
 };