libPad: ABIs

This reverts commit eccd6f6cac.

.github/workflows/build.yml

@@ -208,34 +208,66 @@ jobs:
         run: |
           archive="$RUNNER_TEMP/MoltenVK-macos.tar"
           package="$RUNNER_TEMP/MoltenVK"
+          install_dir="_Build/macos/install"
+          app_dir="$install_dir/KytyPS5.app"
+          moltenvk="$install_dir/libMoltenVK.dylib"
           curl --fail --location --retry 3 \
             --output "$archive" \
             "https://github.com/KhronosGroup/MoltenVK/releases/download/$MOLTENVK_VERSION/MoltenVK-macos.tar"
           echo "$MOLTENVK_SHA256  $archive" | shasum -a 256 --check
           tar -xf "$archive" -C "$RUNNER_TEMP"
+          test -d "$app_dir"
           install -m 755 \
             "$package/MoltenVK/dynamic/dylib/macOS/libMoltenVK.dylib" \
-            _Build/macos/install/libMoltenVK.dylib
-          install -m 644 "$package/LICENSE" _Build/macos/install/LICENSE.MoltenVK
-          codesign --force --sign - --timestamp=none _Build/macos/install/libMoltenVK.dylib
+            "$moltenvk"
+          install -m 644 "$package/LICENSE" "$install_dir/LICENSE.MoltenVK"
+          codesign --force --sign - --timestamp=none "$moltenvk"
+          install -d -m 755 "$app_dir/Contents/Frameworks" "$app_dir/Contents/Resources"
+          install -m 755 "$moltenvk" "$app_dir/Contents/Frameworks/libMoltenVK.dylib"
+          install -m 644 "$package/LICENSE" "$app_dir/Contents/Resources/LICENSE.MoltenVK"
+          codesign --force --sign - --timestamp=none "$app_dir"
 
       - name: Verify artifacts
         shell: bash
         run: |
-          file _Build/macos/install/launcher
-          file _Build/macos/install/kyty_emulator
-          file _Build/macos/install/libMoltenVK.dylib
-          lipo _Build/macos/install/launcher -verify_arch x86_64
-          lipo _Build/macos/install/kyty_emulator -verify_arch x86_64
-          lipo _Build/macos/install/libMoltenVK.dylib -verify_arch x86_64
-          codesign --verify --strict _Build/macos/install/kyty_emulator
-          codesign --verify --strict _Build/macos/install/libMoltenVK.dylib
+          install_dir="_Build/macos/install"
+          app_dir="$install_dir/KytyPS5.app"
+          test -d "$app_dir"
+          for binary in \
+            "$install_dir/kyty_emulator" \
+            "$install_dir/libMoltenVK.dylib" \
+            "$app_dir/Contents/MacOS/KytyPS5" \
+            "$app_dir/Contents/MacOS/kyty_emulator" \
+            "$app_dir/Contents/Frameworks/libMoltenVK.dylib"; do
+            file "$binary"
+            lipo "$binary" -verify_arch x86_64
+          done
+          for binary in \
+            "$install_dir/kyty_emulator" \
+            "$install_dir/libMoltenVK.dylib" \
+            "$app_dir/Contents/MacOS/KytyPS5" \
+            "$app_dir/Contents/MacOS/kyty_emulator" \
+            "$app_dir/Contents/Frameworks/libMoltenVK.dylib"; do
+            codesign --verify --strict "$binary"
+          done
+          codesign --verify --deep --strict --verbose=2 "$app_dir"
+
+      - name: Package macOS artifacts
+        shell: bash
+        run: |
+          install_dir="_Build/macos/install"
+          archive="$RUNNER_TEMP/KytyPS5-macOS-x86_64.zip"
+          (
+            cd "$install_dir"
+            zip -qry "$archive" .
+          )
+          unzip -t "$archive"
 
       - name: Upload macOS artifacts
         uses: actions/upload-artifact@v4
         with:
           name: KytyPS5-macOS-x86_64
-          path: _Build/macos/install/**
+          path: ${{ runner.temp }}/KytyPS5-macOS-x86_64.zip
           if-no-files-found: error
 
   linux:
@@ -424,16 +456,9 @@ jobs:
         if: needs.macos.result == 'success'
         shell: bash
         run: |
-          macos_dir="artifacts/KytyPS5-macOS-x86_64"
-          test -d "$macos_dir"
-          chmod a+x \
-            "$macos_dir/launcher" \
-            "$macos_dir/kyty_emulator" \
-            "$macos_dir/libMoltenVK.dylib"
-          (
-            cd "$macos_dir"
-            zip -r "$GITHUB_WORKSPACE/$RELEASE_NAME-macOS-x86_64.zip" .
-          )
+          macos_archive="artifacts/KytyPS5-macOS-x86_64/KytyPS5-macOS-x86_64.zip"
+          test -f "$macos_archive"
+          cp "$macos_archive" "$RELEASE_NAME-macOS-x86_64.zip"
 
       - name: Package Linux build
         if: needs.linux.result == 'success'

.gitignore

@@ -5,3 +5,4 @@ build/
 _Build/vscode-clang/
 _Build/
 cmake-build-debug/
+_PipelineCache/

.vscode/launch.json

@@ -10,7 +10,14 @@
             "stopAtEntry": false,
             "cwd": "${workspaceFolder}\\_Build\\vscode-clang\\launcher",
             "environment": [],
-            "externalConsole": false
+            "externalConsole": false,
+            "linux": {
+                "type": "cppdbg",
+                "program": "${workspaceFolder}/_Build/vscode-clang/launcher/launcher",
+                "cwd": "${workspaceFolder}/_Build/vscode-clang/launcher",
+                "MIMode": "gdb",
+                "miDebuggerPath": "/usr/bin/gdb"
+            }
         },
         {
             "name": "Debug kyty_emulator",
@@ -25,13 +32,40 @@
             "stopAtEntry": false,
             "cwd": "${workspaceFolder}\\_Build\\vscode-clang",
             "environment": [],
-            "externalConsole": true
+            "externalConsole": true,
+            "linux": {
+                "type": "cppdbg",
+                "program": "${workspaceFolder}/_Build/vscode-clang/kyty_emulator",
+                "cwd": "${workspaceFolder}/_Build/vscode-clang",
+                "externalConsole": false,
+                "MIMode": "gdb",
+                "miDebuggerPath": "/usr/bin/gdb",
+                "setupCommands": [
+                    {
+                        "description": "Pass Kyty's internal signals to its handlers",
+                        "text": "handle SIG37 SIGSEGV SIGBUS SIGILL nostop noprint pass",
+                        "ignoreFailures": false
+                    }
+                ]
+            }
         },
         {
             "name": "Attach to running kyty_emulator",
             "type": "cppvsdbg",
             "request": "attach",
-            "processId": "${command:pickProcess}"
+            "processId": "${command:pickProcess}",
+            "linux": {
+                "type": "cppdbg",
+                "MIMode": "gdb",
+                "miDebuggerPath": "/usr/bin/gdb",
+                "setupCommands": [
+                    {
+                        "description": "Pass Kyty's internal signals to its handlers",
+                        "text": "handle SIG37 SIGSEGV SIGBUS SIGILL nostop noprint pass",
+                        "ignoreFailures": false
+                    }
+                ]
+            }
         }
     ]
 }

CMakeLists.txt

@@ -431,16 +431,15 @@ add_executable(shader_vertex_metadata_tests EXCLUDE_FROM_ALL
 )
 target_include_directories(shader_vertex_metadata_tests PRIVATE ${inc_headers})
 
-add_executable(shader_stage_runtime_tests EXCLUDE_FROM_ALL
-	"${KYTY_TESTS_DIR}/ShaderStageRuntimeTests.cpp"
+add_executable(resource_materialization_tests EXCLUDE_FROM_ALL
+	"${KYTY_TESTS_DIR}/ResourceMaterializationTests.cpp"
 	"${KYTY_SOURCE_DIR}/graphics/guest_gpu/gpu_format.cpp"
 	"${KYTY_SOURCE_DIR}/graphics/host_gpu/hostMemory.cpp"
-	"${KYTY_SOURCE_DIR}/graphics/shader/shaderStageRuntime.cpp"
 	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/passes/ResourceMaterialization.cpp"
 	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/passes/SrtWalker.cpp"
 )
-target_link_libraries(shader_stage_runtime_tests fmt::fmt)
-target_include_directories(shader_stage_runtime_tests PRIVATE ${inc_headers})
+target_link_libraries(resource_materialization_tests fmt::fmt)
+target_include_directories(resource_materialization_tests PRIVATE ${inc_headers})
 
 add_executable(resource_tracking_tests EXCLUDE_FROM_ALL
 	"${KYTY_TESTS_DIR}/ResourceTrackingTests.cpp"
@@ -524,6 +523,16 @@ if(NOT KYTY_CLANG_CL)
 endif()
 
 if(BUILD_TESTING)
+	add_test(NAME emulator_user_name_cli
+	         COMMAND $<TARGET_FILE:kyty_emulator> --help --user-name 1234567890123456)
+	add_test(NAME emulator_user_name_too_long
+	         COMMAND $<TARGET_FILE:kyty_emulator> --help --user-name 12345678901234567)
+	set_tests_properties(emulator_user_name_too_long PROPERTIES WILL_FAIL TRUE)
+	add_test(NAME emulator_present_mode_cli
+	         COMMAND $<TARGET_FILE:kyty_emulator> --help --present-mode Immediate)
+	add_test(NAME emulator_present_mode_invalid
+	         COMMAND $<TARGET_FILE:kyty_emulator> --help --present-mode Invalid)
+	set_tests_properties(emulator_present_mode_invalid PROPERTIES WILL_FAIL TRUE)
 	add_test(NAME ime_dialog COMMAND $<TARGET_FILE:ime_dialog_tests>)
 	add_test(NAME shader_cfg COMMAND $<TARGET_FILE:shader_cfg_tests>)
 	add_test(NAME scalar_provenance COMMAND $<TARGET_FILE:scalar_provenance_tests>)
@@ -532,7 +541,7 @@ if(BUILD_TESTING)
 	add_test(NAME page_manager COMMAND $<TARGET_FILE:page_manager_tests>)
 	add_test(NAME bit_array COMMAND $<TARGET_FILE:bit_array_tests>)
 	add_test(NAME shader_vertex_metadata COMMAND $<TARGET_FILE:shader_vertex_metadata_tests>)
-	add_test(NAME shader_stage_runtime COMMAND $<TARGET_FILE:shader_stage_runtime_tests>)
+	add_test(NAME resource_materialization COMMAND $<TARGET_FILE:resource_materialization_tests>)
 	add_test(NAME resource_tracking COMMAND $<TARGET_FILE:resource_tracking_tests>)
 	add_test(NAME event_queue_lifetime COMMAND $<TARGET_FILE:event_queue_lifetime_tests>)
 	add_test(NAME kernel_file_system COMMAND $<TARGET_FILE:kernel_file_system_tests>)
@@ -564,6 +573,8 @@ if(BUILD_TESTING)
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --storage-sampled-only)
 	add_test(NAME texture_cache_depth_readback
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --depth-readback-only)
+	add_test(NAME compute_meta_clear_classification
+		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --compute-meta-clear-only)
 	add_test(NAME buffer_cache_dirty_gc
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --buffer-cache-gc-only)
 
@@ -586,7 +597,7 @@ if(BUILD_TESTING)
 		page_manager_tests
 		bit_array_tests
 		shader_vertex_metadata_tests
-		shader_stage_runtime_tests
+		resource_materialization_tests
 		resource_tracking_tests
 		event_queue_lifetime_tests
 		kernel_file_system_tests

README.md

@@ -220,17 +220,23 @@ cmake --install _Build/macos --prefix _Build/macos/install
 ```
 
 The build re-signs `kyty_emulator` with the JIT entitlements it needs to execute translated
-guest code; no manual signing step is required.
+guest code; no manual signing step is required. When the launcher is built, the install
+also produces `_Build/macos/install/KytyPS5.app` — double-click to launch the GUI.
+A flat `kyty_emulator` is kept for CLI usage.
 
 Vulkan comes from MoltenVK. Download `MoltenVK-macos.tar` from the
 [MoltenVK releases](https://github.com/KhronosGroup/MoltenVK/releases), then copy
-`MoltenVK/dynamic/dylib/macOS/libMoltenVK.dylib` next to `kyty_emulator` and ad-hoc sign it:
+`MoltenVK/dynamic/dylib/macOS/libMoltenVK.dylib` next to the flat `kyty_emulator`
+(and, for the bundle, into `KytyPS5.app/Contents/Frameworks/`) and ad-hoc sign it:
 
 ```bash
 codesign --force --sign - _Build/macos/install/libMoltenVK.dylib
+# For the bundle (if present):
+codesign --force --sign - _Build/macos/install/KytyPS5.app/Contents/Frameworks/libMoltenVK.dylib
+codesign --force --sign - _Build/macos/install/KytyPS5.app
 ```
 
-Release archives already include a signed `libMoltenVK.dylib`.
+Release archives already include a signed `libMoltenVK.dylib` (both flat and inside the bundle).
 
 ### Regression tests
 
@@ -274,6 +280,10 @@ To use the graphical launcher:
 ./_Build/linux/install/launcher
 ```
 
+```bash
+open _Build/macos/install/KytyPS5.app  # or double-click in Finder
+```
+
 On first launch, add one or more game folders in the global settings. The launcher searches those
 folders recursively for game directories containing `eboot.bin`. Select a detected game and run it
 from the game list.
@@ -288,12 +298,17 @@ The emulator can also be started directly with a legally obtained game directory
 ./_Build/linux/install/kyty_emulator --game "/games/ExampleGame"
 ```
 
-On macOS, point SDL at the MoltenVK library explicitly; the hardened runtime prevents it from
-being picked up from the executable's directory:
+On macOS, the adjacent flat or app-bundled `libMoltenVK.dylib` is found automatically; no
+environment variable is required:
 
 ```bash
-cd _Build/macos/install
-SDL_VULKAN_LIBRARY="$PWD/libMoltenVK.dylib" ./kyty_emulator --game "/games/ExampleGame"
+./_Build/macos/install/kyty_emulator --game "/games/ExampleGame"
+```
+
+To override the Vulkan loader, set `SDL_VULKAN_LIBRARY`:
+
+```bash
+SDL_VULKAN_LIBRARY=/path/to/libMoltenVK.dylib ./kyty_emulator --game "/games/ExampleGame"
 ```
 
 Run `kyty_emulator --help` to see the available graphics, logging, validation, profiling, and

src/common/assert.cpp

@@ -59,6 +59,7 @@ int DbgExitHandler(const char* file, int line, fmt::text_style style, std::strin
 }
 
 void DbgExit(int status) {
+	Subsystems::EmergencyShutdownActive();
 	std::fflush(nullptr);
 	std::_Exit(status);
 }

src/common/assert.h

@@ -29,11 +29,7 @@ void DbgExit(int status);
 
 } // namespace Common
 
-#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS || KYTY_PLATFORM == KYTY_PLATFORM_LINUX
 #define EXIT_HALT() (Common::DbgExit(321), 1)
-#else
-#define EXIT_HALT() (std::_Exit(321), 1)
-#endif
 
 #ifndef KYTY_FINAL
 #define EXIT_IF(x)                                                                                 \

src/common/emulatorConfig.cpp

@@ -21,6 +21,8 @@ void Shutdown() {
 
 void Load(const ConfigOptions& cfg) {
 	EXIT_IF(g_config == nullptr);
+	EXIT_IF(cfg.user_name.empty() || cfg.user_name.size() > MAX_USER_NAME_LENGTH);
+	EXIT_IF(!IsConfiguredUserIdValid(cfg.user_id));
 
 	*g_config = cfg;
 }
@@ -33,6 +35,22 @@ uint32_t GetScreenHeight() {
 	return g_config->screen_height;
 }
 
+const std::string& GetUserName() {
+	return g_config->user_name;
+}
+
+int32_t GetUserId() {
+	return g_config->user_id;
+}
+
+PresentMode GetPresentMode() {
+	return g_config->present_mode;
+}
+
+int32_t GetGpuIndex() {
+	return g_config->gpu_index;
+}
+
 bool FullscreenEnabled() {
 	return g_config->fullscreen_enabled;
 }

src/common/emulatorConfig.h

@@ -3,6 +3,7 @@
 
 #include "common/common.h"
 
+#include <cstddef>
 #include <filesystem>
 #include <string>
 #include <vector>
@@ -26,14 +27,28 @@ enum class ProfilerDirection { None, Network };
 
 enum class OutputDirection { Silent, Console, File };
 
+enum class PresentMode { Fifo, Mailbox, Immediate };
+
 using Keymap = std::vector<std::string>;
 
 constexpr uint32_t DEFAULT_CONSOLE_LANGUAGE = 1;
 constexpr uint32_t MAX_CONSOLE_LANGUAGE     = 29;
+constexpr std::size_t MAX_USER_NAME_LENGTH = 16;
+constexpr int32_t DEFAULT_USER_ID           = 1000;
+
+constexpr bool IsConfiguredUserIdValid(int32_t user_id) {
+	constexpr int32_t USER_ID_EVERYONE = 0xfe;
+	constexpr int32_t USER_ID_SYSTEM   = 0xff;
+	return user_id >= 0 && user_id != USER_ID_EVERYONE && user_id != USER_ID_SYSTEM;
+}
 
 struct ConfigOptions {
 	uint32_t               screen_width                = 1280;
 	uint32_t               screen_height               = 720;
+	std::string            user_name                   = "Kyty";
+	int32_t                user_id                     = DEFAULT_USER_ID;
+	PresentMode            present_mode                = PresentMode::Fifo;
+	int32_t                gpu_index                   = -1;
 	bool                   fullscreen_enabled          = false;
 	uint32_t               vblank_frequency            = 60;
 	uint32_t               console_language            = DEFAULT_CONSOLE_LANGUAGE;
@@ -45,7 +60,7 @@ struct ConfigOptions {
 	bool                   command_buffer_dump_enabled = false;
 	std::filesystem::path  command_buffer_dump_folder  = "_Buffers";
 	bool                   graphics_debug_dump_enabled = false;
-	OutputDirection        printf_direction            = OutputDirection::Console;
+	OutputDirection        printf_direction            = OutputDirection::Silent;
 	std::filesystem::path  printf_output_file          = "_kyty.txt";
 	ProfilerDirection      profiler_direction          = ProfilerDirection::None;
 	bool                   spirv_debug_printf_enabled  = false;
@@ -63,6 +78,10 @@ void Load(const ConfigOptions& cfg);
 
 uint32_t GetScreenWidth();
 uint32_t GetScreenHeight();
+const std::string& GetUserName();
+int32_t  GetUserId();
+PresentMode GetPresentMode();
+int32_t GetGpuIndex();
 bool     FullscreenEnabled();
 uint32_t GetVblankFrequency();
 uint32_t GetConsoleLanguage();

src/common/file.cpp

@@ -25,9 +25,6 @@
 #ifdef CopyFile
 #undef CopyFile
 #endif
-#ifdef MoveFile
-#undef MoveFile
-#endif
 #endif
 
 namespace Common {
@@ -72,7 +69,7 @@ File::File(): m_p(std::make_unique<FilePrivate>()) {
 }
 
 File::~File() {
-	EXIT_IF(m_p->f != nullptr);
+	Close();
 }
 
 File::File(const std::filesystem::path& name): m_p(std::make_unique<FilePrivate>()) {
@@ -544,14 +541,14 @@ bool File::CopyFile(const std::filesystem::path& src,
 	return SysFileCopyFile(src, dst);
 }
 
-bool File::MoveFile(const std::filesystem::path& src,
-                    const std::filesystem::path& dst) // @suppress("Member declaration not found")
+bool File::RenameFile(const std::filesystem::path& src,
+                      const std::filesystem::path& dst) // @suppress("Member declaration not found")
 {
 	if (IsFileExisting(dst)) {
 		DeleteFile(dst); // @suppress("Invalid arguments")
 	}
 
-	return SysFileMoveFile(src, dst);
+	return SysFileRenameFile(src, dst);
 }
 
 void File::RemoveReadonly(const std::filesystem::path& name) {

src/common/file.h

@@ -20,9 +20,6 @@
 #ifdef CopyFile
 #undef CopyFile
 #endif
-#ifdef MoveFile
-#undef MoveFile
-#endif
 #endif
 
 namespace Common {
@@ -107,7 +104,7 @@ public:
 	static std::vector<FindInfo> FindFiles(const std::filesystem::path& path);
 	static std::vector<DirEntry> GetDirEntries(const std::filesystem::path& path);
 	static bool CopyFile(const std::filesystem::path& src, const std::filesystem::path& dst);
-	static bool MoveFile(const std::filesystem::path& src, const std::filesystem::path& dst);
+	static bool RenameFile(const std::filesystem::path& src, const std::filesystem::path& dst);
 	static void RemoveReadonly(const std::filesystem::path& name);
 
 	KYTY_CLASS_NO_COPY(File);

src/common/platform/sysFileIO.h

@@ -80,7 +80,7 @@ void SysFileFindFiles(const std::filesystem::path& path, std::vector<sys_file_fi
 // NOLINTNEXTLINE(google-runtime-references)
 void SysFileGetDents(const std::filesystem::path& path, std::vector<sys_dir_entry_t>& out);
 bool SysFileCopyFile(const std::filesystem::path& src, const std::filesystem::path& dst);
-bool SysFileMoveFile(const std::filesystem::path& src, const std::filesystem::path& dst);
+bool SysFileRenameFile(const std::filesystem::path& src, const std::filesystem::path& dst);
 void SysFileRemoveReadonly(const std::filesystem::path& name);
 
 #endif /* KYTY_COMMON_PLATFORM_SYSFILEIO_H_ */

src/common/platform/sysLinuxFileIO.cpp

@@ -693,7 +693,7 @@ bool SysFileCopyFile(const std::filesystem::path& src, const std::filesystem::pa
 	       !error;
 }
 
-bool SysFileMoveFile(const std::filesystem::path& src, const std::filesystem::path& dst) {
+bool SysFileRenameFile(const std::filesystem::path& src, const std::filesystem::path& dst) {
 	auto real_src = get_internal_name(src);
 	auto real_dst = get_internal_name(dst);
 

src/common/platform/sysWindowsFileIO.cpp

@@ -565,7 +565,7 @@ bool SysFileCopyFile(const std::filesystem::path& src, const std::filesystem::pa
 	return CopyFileW(src_wide.c_str(), dst_wide.c_str(), FALSE) != 0;
 }
 
-bool SysFileMoveFile(const std::filesystem::path& src, const std::filesystem::path& dst) {
+bool SysFileRenameFile(const std::filesystem::path& src, const std::filesystem::path& dst) {
 	auto src_wide = src.wstring();
 	auto dst_wide = dst.wstring();
 	return MoveFileW(src_wide.c_str(), dst_wide.c_str()) != 0;

src/emulator.cpp

@@ -200,6 +200,10 @@ void Run(const RunOptions& options) {
 	int ok = atexit(KytyClose);
 	EXIT_NOT_IMPLEMENTED(ok != 0);
 
+	// Guest threads are still running, so skip KytyClose() and only flush emergency state.
+	ok = at_quick_exit(Common::Subsystems::EmergencyShutdownActive);
+	EXIT_NOT_IMPLEMENTED(ok != 0);
+
 	Libs::LibKernel::FileSystem::Mount(options.app0_dir, "/app0");
 	Libs::LibKernel::FileSystem::Mount(options.app0_dir, "/hostapp");
 

src/generate_version.cmake

@@ -1,5 +1,6 @@
 set(KYTY_GIT_VERSION "unknown")
 set(KYTY_GIT_HASH "unknown")
+set(KYTY_GIT_REVISION "unknown")
 if(GIT_EXECUTABLE)
 	execute_process(
 		COMMAND "${GIT_EXECUTABLE}" describe --tags --always --dirty
@@ -14,16 +15,18 @@ if(GIT_EXECUTABLE)
 	endif()
 
 	execute_process(
-		COMMAND "${GIT_EXECUTABLE}" rev-parse --short=7 HEAD
+		COMMAND "${GIT_EXECUTABLE}" rev-parse HEAD
 		WORKING_DIRECTORY "${GIT_WORKING_DIRECTORY}"
-		OUTPUT_VARIABLE KYTY_GIT_HASH
+		OUTPUT_VARIABLE KYTY_GIT_REVISION
 		OUTPUT_STRIP_TRAILING_WHITESPACE
 		RESULT_VARIABLE GIT_HASH_RESULT
 		ERROR_QUIET
 	)
 	if(NOT GIT_HASH_RESULT EQUAL 0)
 		set(KYTY_GIT_HASH "unknown")
+		set(KYTY_GIT_REVISION "unknown")
 	else()
+		string(SUBSTRING "${KYTY_GIT_REVISION}" 0 7 KYTY_GIT_HASH)
 		execute_process(
 			COMMAND "${GIT_EXECUTABLE}" diff-index --quiet HEAD --
 			WORKING_DIRECTORY "${GIT_WORKING_DIRECTORY}"

src/graphics/guest_gpu/command_processor/commandProcessor.h

@@ -73,13 +73,9 @@ public:
 	void SetDrawIndirectArgsBaseAddress(uint64_t draw_indirect_args_base_addr);
 	void SetDispatchIndirectArgsBaseAddress(uint64_t dispatch_indirect_args_base_addr);
 	void SetNumInstances(uint32_t num_instances);
-	void DrawIndex(uint32_t index_count, const void* index_addr, uint32_t flags, uint32_t type,
-	               uint32_t instance_count = 0, const void* object_ids = nullptr,
-	               uint32_t render_target_slice_offset = 0, int32_t vertex_offset_add = 0,
-	               uint32_t first_instance = 0);
-	void DrawIndexOffset(uint32_t index_offset, uint32_t index_count, uint32_t flags);
-	void DrawIndexAuto(uint32_t index_count, uint32_t flags,
-	                   uint32_t render_target_slice_offset = 0);
+	void DrawIndex(DrawIndexArgs args);
+	void DrawIndexOffset(uint32_t index_offset, uint32_t index_count);
+	void DrawIndexAuto(DrawAutoArgs args);
 	void DrawIndirect(uint32_t data_offset, uint32_t draw_initiator, bool indexed);
 	void DrawIndirectMulti(uint32_t data_offset, uint32_t max_count_or_count,
 	                       const volatile uint32_t* count_addr, uint32_t stride_in_bytes,
@@ -149,10 +145,6 @@ private:
 	                      uint32_t interrupt_context_id);
 	void ProcessPm4(Pm4Execution& execution, size_t stop_depth);
 	void SuspendPm4();
-	void SubmitNonIndexedDraw(uint32_t vertex_count, uint32_t flags,
-	                          uint32_t render_target_slice_offset, uint32_t first_vertex,
-	                          uint32_t first_instance);
-
 	CommandScheduler&   GetScheduler() const { return m_renderer.GetCommandScheduler(); }
 	CommandBuffer&      CurrentBuffer() { return GetScheduler().Current(); }
 	void                CheckBuffer() const { GetScheduler().CheckActive(); }

src/graphics/guest_gpu/command_processor/pm4Dispatch.cpp

@@ -23,9 +23,7 @@ constexpr auto MakeContextDispatchTable() {
 	g_hw_ctx_func[Pm4::DB_STENCIL_CLEAR]             = HwCtxSetStencilClear;
 	g_hw_ctx_func[Pm4::DB_DEPTH_CLEAR]               = HwCtxSetDepthClear;
 	g_hw_ctx_func[Pm4::PA_SC_SCREEN_SCISSOR_TL]      = HwCtxSetScreenScissor;
-	g_hw_ctx_func[Pm4::DB_Z_INFO]                    = HwCtxSetDepthRenderTarget;
 	g_hw_ctx_func[Pm4::DB_STENCIL_INFO]              = HwCtxSetStencilInfo;
-	g_hw_ctx_func[Pm4::DB_HTILE_SURFACE]             = HwCtxSetDepthHtileSurface;
 	g_hw_ctx_func[Pm4::PA_SU_HARDWARE_SCREEN_OFFSET] = HwCtxSetHardwareScreenOffset;
 	g_hw_ctx_func[Pm4::PA_SC_WINDOW_OFFSET]          = HwCtxSetWindowOffset;
 	g_hw_ctx_func[Pm4::PA_SC_WINDOW_SCISSOR_TL]      = HwCtxSetWindowScissor;
@@ -75,7 +73,6 @@ constexpr auto MakeContextDispatchTable() {
 	g_hw_ctx_func[Pm4::PA_SC_BINNER_CNTL_0]                   = HwCtxSetPaScExtendedControl;
 	g_hw_ctx_func[Pm4::PA_SC_BINNER_CNTL_1]                   = HwCtxSetPaScExtendedControl;
 	g_hw_ctx_func[Pm4::PA_SC_CONSERVATIVE_RASTERIZATION_CNTL] = HwCtxSetPaScExtendedControl;
-	g_hw_ctx_func[Pm4::PA_SC_NGG_MODE_CNTL]                   = HwCtxSetPaScExtendedControl;
 	g_hw_ctx_func[Pm4::DB_ALPHA_TO_MASK]                      = HwCtxSetAlphaToMask;
 	g_hw_ctx_func[Pm4::VGT_DRAW_PAYLOAD_CNTL]                 = HwCtxSetDrawPayloadControl;
 	g_hw_ctx_func[Pm4::VGT_PRIMITIVEID_RESET]                 = HwCtxSetPrimitiveIdReset;
@@ -83,7 +80,6 @@ constexpr auto MakeContextDispatchTable() {
 	g_hw_ctx_func[Pm4::VGT_SHADER_STAGES_EN]                  = HwCtxSetShaderStages;
 
 	for (uint32_t slot = 0; slot < 8; slot++) {
-		g_hw_ctx_func[Pm4::CB_COLOR0_BASE + slot * 15] = HwCtxSetRenderTarget;
 		g_hw_ctx_func[Pm4::CB_COLOR0_INFO + slot * 15] = HwCtxSetColorInfo;
 
 		g_hw_ctx_func[Pm4::CB_BLEND0_CONTROL + slot * 1] = HwCtxSetBlendControl;
@@ -116,25 +112,21 @@ constexpr auto MakeShaderDispatchTable() {
 	}
 
 	for (uint32_t slot = 0; slot < 4; slot++) {
-		g_hw_sh_func[Pm4::SPI_SHADER_USER_ACCUM_PS_0 + slot] = HwShSetPsUserAccumSgpr;
+		g_hw_sh_func[Pm4::SPI_SHADER_USER_ACCUM_PS_0 + slot] = HwShIgnoreUserAccumulator;
 	}
 
 	for (uint32_t slot = 0; slot < 4; slot++) {
-		g_hw_sh_func[Pm4::SPI_SHADER_USER_ACCUM_ESGS_0 + slot] = HwShSetGsUserAccumSgpr;
+		g_hw_sh_func[Pm4::SPI_SHADER_USER_ACCUM_ESGS_0 + slot] = HwShIgnoreUserAccumulator;
 	}
 	for (uint32_t slot = 0; slot < 4; slot++) {
-		g_hw_sh_func[Pm4::SPI_SHADER_USER_ACCUM_LSHS_0 + slot] = HwShSetHsUserAccumSgpr;
+		g_hw_sh_func[Pm4::SPI_SHADER_USER_ACCUM_LSHS_0 + slot] = HwShIgnoreUserAccumulator;
 	}
 
-	for (uint32_t slot = 0; Pm4::COMPUTE_USER_ACCUM_0 + slot < Pm4::COMPUTE_USER_DATA_0; slot++) {
-		g_hw_sh_func[Pm4::COMPUTE_USER_ACCUM_0 + slot] = HwShSetCsUserAccumSgpr;
+	for (uint32_t slot = 0; slot < 4; slot++) {
+		g_hw_sh_func[Pm4::COMPUTE_USER_ACCUM_0 + slot] = HwShIgnoreUserAccumulator;
 	}
 	g_hw_sh_func[Pm4::COMPUTE_PGM_LO]                  = HwShSetCsRegisters;
 	g_hw_sh_func[Pm4::COMPUTE_PGM_HI]                  = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_TBA_LO]                  = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_TBA_HI]                  = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_TMA_LO]                  = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_TMA_HI]                  = HwShSetCsRegisters;
 	g_hw_sh_func[Pm4::COMPUTE_PGM_RSRC1]               = HwShSetCsRegisters;
 	g_hw_sh_func[Pm4::COMPUTE_PGM_RSRC2]               = HwShSetCsRegisters;
 	g_hw_sh_func[Pm4::COMPUTE_START_X]                 = HwShSetCsRegisters;
@@ -144,30 +136,16 @@ constexpr auto MakeShaderDispatchTable() {
 	g_hw_sh_func[Pm4::COMPUTE_NUM_THREAD_Y]            = HwShSetCsRegisters;
 	g_hw_sh_func[Pm4::COMPUTE_NUM_THREAD_Z]            = HwShSetCsRegisters;
 	g_hw_sh_func[Pm4::COMPUTE_RESOURCE_LIMITS]         = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_DESTINATION_EN_SE0]      = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_DESTINATION_EN_SE1]      = HwShSetCsRegisters;
 	g_hw_sh_func[Pm4::COMPUTE_TMPRING_SIZE]            = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_DESTINATION_EN_SE2]      = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_DESTINATION_EN_SE3]      = HwShSetCsRegisters;
 	g_hw_sh_func[Pm4::COMPUTE_PGM_RSRC3]               = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_SHADER_CHKSUM]           = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::COMPUTE_DISPATCH_TUNNEL]         = HwShSetCsRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_TBA_LO_PS]            = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_PGM_RSRC3_PS]         = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_PGM_RSRC4_PS]         = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_TMA_LO_PS]            = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_TMA_HI_PS]            = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_REQ_CTRL_PS]          = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_PGM_RSRC3_GS]         = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_PGM_RSRC4_GS]         = HwShIgnoreRegisters;
+	g_hw_sh_func[Pm4::COMPUTE_PACE_ID]                 = HwShSetCsRegisters;
+	g_hw_sh_func[Pm4::SPI_GRAPHICS_SHADER_CONTROL_PS]  = HwShIgnoreRegisters;
+	g_hw_sh_func[Pm4::SPI_GRAPHICS_SHADER_CONTROL_GS]  = HwShIgnoreRegisters;
 	g_hw_sh_func[Pm4::SPI_SHADER_USER_DATA_ADDR_LO_GS] = HwShIgnoreRegisters;
 	g_hw_sh_func[Pm4::SPI_SHADER_USER_DATA_ADDR_HI_GS] = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_REQ_CTRL_ESGS]        = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_PGM_RSRC3_HS]         = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_PGM_RSRC4_HS]         = HwShIgnoreRegisters;
+	g_hw_sh_func[Pm4::SPI_GRAPHICS_SHADER_CONTROL_HS]  = HwShIgnoreRegisters;
 	g_hw_sh_func[Pm4::SPI_SHADER_USER_DATA_ADDR_LO_HS] = HwShIgnoreRegisters;
 	g_hw_sh_func[Pm4::SPI_SHADER_USER_DATA_ADDR_HI_HS] = HwShIgnoreRegisters;
-	g_hw_sh_func[Pm4::SPI_SHADER_REQ_CTRL_LSHS]        = HwShIgnoreRegisters;
 
 	return g_hw_sh_func;
 }
@@ -179,6 +157,7 @@ constexpr auto MakeUconfigDispatchTable() {
 	g_hw_uc_func[Pm4::VGT_INDEX_TYPE]            = HwUcSetIndexType;
 	g_hw_uc_func[Pm4::VGT_OBJECT_ID]             = HwUcSetObjectId;
 	g_hw_uc_func[Pm4::TEXTURE_GRADIENT_FACTORS]  = HwUcSetTextureGradientFactors;
+	g_hw_uc_func[Pm4::TEXTURE_GRADIENT_CONTROL]  = HwUcSetTextureGradientFactors;
 	g_hw_uc_func[Pm4::IA_MULTI_VGT_PARAM]        = HwUcSetIaMultiVgtParam;
 	g_hw_uc_func[Pm4::GE_MULTI_PRIM_IB_RESET_EN] = HwUcSetMultiPrimIbReset;
 	g_hw_uc_func[Pm4::TA_CS_BC_BASE_ADDR]        = HwUcSetBorderColorTableAddr;

src/graphics/guest_gpu/command_processor/pm4Dispatch.h

@@ -44,11 +44,7 @@ uint32_t HwCtxSetBorderColorTableAddr(CommandProcessor&, uint32_t, uint32_t, con
 uint32_t HwCtxSetStencilClear(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetDepthClear(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetScreenScissor(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwCtxSetDepthRenderTarget(CommandProcessor&, uint32_t, uint32_t, const uint32_t*,
-                                   uint32_t);
 uint32_t HwCtxSetStencilInfo(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwCtxSetDepthHtileSurface(CommandProcessor&, uint32_t, uint32_t, const uint32_t*,
-                                   uint32_t);
 uint32_t HwCtxSetHardwareScreenOffset(CommandProcessor&, uint32_t, uint32_t, const uint32_t*,
                                       uint32_t);
 uint32_t HwCtxSetWindowOffset(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
@@ -89,7 +85,6 @@ uint32_t HwCtxSetDrawPayloadControl(CommandProcessor&, uint32_t, uint32_t, const
 uint32_t HwCtxSetPrimitiveIdReset(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetObjprimIdControl(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetShaderStages(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwCtxSetRenderTarget(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetColorInfo(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetBlendControl(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetViewportScissor(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
@@ -100,10 +95,8 @@ uint32_t HwShSetPsUserSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t
 uint32_t HwShSetGsUserSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwShSetHsUserSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwShSetCsUserSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwShSetPsUserAccumSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwShSetGsUserAccumSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwShSetHsUserAccumSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwShSetCsUserAccumSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
+uint32_t HwShIgnoreUserAccumulator(CommandProcessor&, uint32_t, uint32_t, const uint32_t*,
+                                   uint32_t);
 uint32_t HwShSetCsRegisters(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwShIgnoreRegisters(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwUcSetPrimitiveType(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);

src/graphics/guest_gpu/gpu_defs.h

@@ -407,7 +407,19 @@ enum class BufferFormat : uint32_t {
 	k5_5_5_1UNorm       = 134,
 	k1_5_5_5UNorm       = 135,
 	k4_4_4_4UNorm       = 136,
+	kFmask8_S2_F1       = 156,
+	kFmask8_S4_F1       = 157,
+	kFmask8_S8_F1       = 158,
+	kFmask8_S2_F2       = 159,
+	kFmask8_S4_F2       = 160,
 	kFmask8_S4_F4       = 161,
+	kFmask16_S16_F1     = 162,
+	kFmask16_S8_F2      = 163,
+	kFmask32_S16_F2     = 164,
+	kFmask32_S8_F4      = 165,
+	kFmask32_S8_F8      = 166,
+	kFmask64_S16_F4     = 167,
+	kFmask64_S16_F8     = 168,
 	kBc1UNorm           = 169,
 	kBc1Srgb            = 170,
 	kBc2UNorm           = 171,

src/graphics/guest_gpu/gpu_format.cpp

@@ -15,6 +15,7 @@ struct FormatInfo {
 	uint32_t           render_target_bytes_per_element;
 	bool               sampled_texture;
 	bool               uint_texture;
+	bool               sint_texture = false;
 };
 
 constexpr FormatInfo kFormatInfo[] = {
@@ -24,19 +25,19 @@ constexpr FormatInfo kFormatInfo[] = {
 	{BufferFormat::k16UNorm, 2, 0, 2, true, false},
 	{BufferFormat::k16SNorm, 2, 0, 2, true, false},
 	{BufferFormat::k16UInt, 2, 0, 2, true, true},
-	{BufferFormat::k16SInt, 2, 0, 2, false, false},
+	{BufferFormat::k16SInt, 2, 0, 2, true, false, true},
 	{BufferFormat::k16Float, 2, 0, 2, true, false},
 	{BufferFormat::k8_8UNorm, 2, 0, 2, true, false},
 	{BufferFormat::k8_8SNorm, 2, 0, 2, true, false},
 	{BufferFormat::k8_8UInt, 2, 0, 2, true, true},
-	{BufferFormat::k8_8SInt, 2, 0, 2, false, false},
+	{BufferFormat::k8_8SInt, 2, 0, 2, true, false, true},
 	{BufferFormat::k32UInt, 4, 0, 4, true, true},
-	{BufferFormat::k32SInt, 4, 0, 4, false, false},
+	{BufferFormat::k32SInt, 4, 0, 4, true, false, true},
 	{BufferFormat::k32Float, 4, 0, 4, true, false},
 	{BufferFormat::k16_16UNorm, 4, 0, 4, true, false},
 	{BufferFormat::k16_16SNorm, 4, 0, 4, true, false},
 	{BufferFormat::k16_16UInt, 4, 0, 4, true, true},
-	{BufferFormat::k16_16SInt, 4, 0, 4, false, false},
+	{BufferFormat::k16_16SInt, 4, 0, 4, true, false, true},
 	{BufferFormat::k16_16Float, 4, 0, 4, true, false},
 	{BufferFormat::k11_11_10UInt, 4, 0, 4, true, true},
 	{BufferFormat::k11_11_10Float, 4, 0, 4, true, false},
@@ -45,20 +46,20 @@ constexpr FormatInfo kFormatInfo[] = {
 	{BufferFormat::k8_8_8_8UNorm, 4, 0, 4, true, false},
 	{BufferFormat::k8_8_8_8SNorm, 4, 0, 4, true, false},
 	{BufferFormat::k8_8_8_8UInt, 4, 0, 4, true, true},
-	{BufferFormat::k8_8_8_8SInt, 4, 0, 4, false, false},
+	{BufferFormat::k8_8_8_8SInt, 4, 0, 4, true, false, true},
 	{BufferFormat::k32_32UInt, 8, 0, 8, true, true},
-	{BufferFormat::k32_32SInt, 8, 0, 8, false, false},
+	{BufferFormat::k32_32SInt, 8, 0, 8, true, false, true},
 	{BufferFormat::k32_32Float, 8, 0, 8, true, false},
 	{BufferFormat::k16_16_16_16UNorm, 8, 0, 8, true, false},
 	{BufferFormat::k16_16_16_16SNorm, 8, 0, 8, true, false},
 	{BufferFormat::k16_16_16_16UInt, 8, 0, 8, true, true},
-	{BufferFormat::k16_16_16_16SInt, 8, 0, 8, false, false},
+	{BufferFormat::k16_16_16_16SInt, 8, 0, 8, true, false, true},
 	{BufferFormat::k16_16_16_16Float, 8, 0, 8, true, false},
 	{BufferFormat::k32_32_32UInt, 12, 0, 12, true, true},
-	{BufferFormat::k32_32_32SInt, 12, 0, 12, false, false},
+	{BufferFormat::k32_32_32SInt, 12, 0, 12, true, false, true},
 	{BufferFormat::k32_32_32Float, 12, 0, 12, true, false},
 	{BufferFormat::k32_32_32_32UInt, 16, 0, 16, true, true},
-	{BufferFormat::k32_32_32_32SInt, 16, 0, 16, false, false},
+	{BufferFormat::k32_32_32_32SInt, 16, 0, 16, true, false, true},
 	{BufferFormat::k32_32_32_32Float, 16, 0, 16, true, false},
 	{BufferFormat::k8Srgb, 1, 0, 0, true, false},
 	{BufferFormat::k8_8Srgb, 2, 0, 0, true, false},
@@ -68,7 +69,19 @@ constexpr FormatInfo kFormatInfo[] = {
 	{BufferFormat::k5_5_5_1UNorm, 2, 0, 2, true, false},
 	{BufferFormat::k1_5_5_5UNorm, 0, 0, 2, false, false},
 	{BufferFormat::k4_4_4_4UNorm, 2, 0, 2, true, false},
+	{BufferFormat::kFmask8_S2_F1, 1, 0, 1, true, false},
+	{BufferFormat::kFmask8_S4_F1, 1, 0, 1, true, false},
+	{BufferFormat::kFmask8_S8_F1, 1, 0, 1, true, false},
+	{BufferFormat::kFmask8_S2_F2, 1, 0, 1, true, false},
+	{BufferFormat::kFmask8_S4_F2, 1, 0, 1, true, false},
 	{BufferFormat::kFmask8_S4_F4, 1, 0, 1, true, false},
+	{BufferFormat::kFmask16_S16_F1, 2, 0, 2, true, false},
+	{BufferFormat::kFmask16_S8_F2, 2, 0, 2, true, false},
+	{BufferFormat::kFmask32_S16_F2, 4, 0, 4, true, false},
+	{BufferFormat::kFmask32_S8_F4, 4, 0, 4, true, false},
+	{BufferFormat::kFmask32_S8_F8, 4, 0, 4, true, false},
+	{BufferFormat::kFmask64_S16_F4, 8, 0, 8, true, false},
+	{BufferFormat::kFmask64_S16_F8, 8, 0, 8, true, false},
 	{BufferFormat::kBc1UNorm, 0, 8, 0, true, false},
 	{BufferFormat::kBc1Srgb, 0, 8, 0, true, false},
 	{BufferFormat::kBc2UNorm, 0, 16, 0, true, false},
@@ -203,17 +216,18 @@ uint32_t RenderTargetBytesPerElement(BufferFormat format) {
 }
 
 bool IsFmaskTextureFormat(BufferFormat format) {
-	return format == BufferFormat::kFmask8_S4_F4;
+	return format >= BufferFormat::kFmask8_S2_F1 && format <= BufferFormat::kFmask64_S16_F8;
 }
 
-bool IsSampledTextureFormat(BufferFormat format) {
+TextureNumericClass SampledTextureNumericClass(BufferFormat format) {
 	const auto* info = FindFormatInfo(format);
-	return info != nullptr && info->sampled_texture;
-}
-
-bool IsUintTextureFormat(BufferFormat format) {
-	const auto* info = FindFormatInfo(format);
-	return info != nullptr && info->uint_texture;
+	if (info == nullptr || !info->sampled_texture) {
+		return TextureNumericClass::Unsupported;
+	}
+	if (info->sint_texture) {
+		return TextureNumericClass::Sint;
+	}
+	return info->uint_texture ? TextureNumericClass::Uint : TextureNumericClass::Float;
 }
 
 BufferFormat RemapTextureFormat(BufferFormat format) {

src/graphics/guest_gpu/gpu_format.h

@@ -37,13 +37,14 @@ struct RenderTargetFormatEncoding {
 	}
 };
 
+enum class TextureNumericClass { Unsupported, Float, Uint, Sint };
+
 RenderTargetFormatEncoding ResolveRenderTargetFormat(ChannelLayout layout, ChannelType type);
 uint32_t                   NumBytesPerElement(BufferFormat format);
 uint32_t                   BlockCompressedBytesPerBlock(BufferFormat format);
 uint32_t                   RenderTargetBytesPerElement(BufferFormat format);
 bool                       IsFmaskTextureFormat(BufferFormat format);
-bool                       IsSampledTextureFormat(BufferFormat format);
-bool                       IsUintTextureFormat(BufferFormat format);
+TextureNumericClass        SampledTextureNumericClass(BufferFormat format);
 BufferFormat               RemapTextureFormat(BufferFormat format);
 
 } // namespace Libs::Graphics::Prospero

src/graphics/guest_gpu/graphicsRun.cpp

@@ -40,6 +40,21 @@ static thread_local bool              g_gpu_mutex_owned   = false;
 static thread_local bool              g_gpu_thread        = false;
 static thread_local GuestGpu*         g_gpu_state         = nullptr;
 
+struct DrawIndirectArgs {
+	uint32_t vertex_count_per_instance;
+	uint32_t instance_count;
+	uint32_t start_vertex_location;
+	uint32_t start_instance_location;
+};
+
+struct DrawIndexedIndirectArgs {
+	uint32_t index_count_per_instance;
+	uint32_t instance_count;
+	uint32_t start_index_location;
+	uint32_t base_vertex_location;
+	uint32_t start_instance_location;
+};
+
 class GpuMutexLock final {
 public:
 	explicit GpuMutexLock(Common::Mutex& mutex): m_mutex(mutex) {
@@ -847,36 +862,21 @@ void CommandProcessor::SetPredication(uint32_t condition, uint32_t op, uint32_t
 	}
 }
 
-void CommandProcessor::DrawIndex(uint32_t index_count, const void* index_addr, uint32_t flags,
-                                 uint32_t type, uint32_t instance_count, const void* object_ids,
-                                 uint32_t render_target_slice_offset, int32_t vertex_offset_add,
-                                 uint32_t first_instance) {
+void CommandProcessor::DrawIndex(DrawIndexArgs args) {
 	CheckBuffer();
 
-	if (instance_count == 0) {
-		instance_count = m_num_instances;
+	args.index_type_and_size = m_index_type_and_size;
+	if (args.instance_count == 0) {
+		args.instance_count = m_num_instances;
 	}
-	if (object_ids != nullptr) {
-		LOGF("\t draw indexed multi-instanced objectIds = 0x%016" PRIx64 "\n",
-		     reinterpret_cast<uint64_t>(object_ids));
+	if (args.base_vertex != 0 || args.first_instance != 0) {
+		LOGF("\t draw indexed offsets: base_vertex = %" PRId32 ", first_instance = %" PRIu32 "\n",
+		     args.base_vertex, args.first_instance);
 	}
-	if (render_target_slice_offset != 0) {
-		LOGF("\t draw render target slice offset = %" PRIu32 "\n", render_target_slice_offset);
-	}
-	if (vertex_offset_add != 0 || first_instance != 0) {
-		LOGF("\t draw indexed offsets: vertex_offset_add = %" PRId32 ", first_instance = %" PRIu32
-		     "\n",
-		     vertex_offset_add, first_instance);
-	}
-	m_renderer.GetRenderExecutor().DrawIndex(
-	    m_submit_id, CurrentBuffer(), m_index_type_and_size, index_count, index_addr, flags, type,
-	    instance_count, render_target_slice_offset, vertex_offset_add, first_instance);
+	m_renderer.GetRenderExecutor().DrawIndex(m_submit_id, CurrentBuffer(), args);
 }
 
-void CommandProcessor::DrawIndexOffset(uint32_t index_offset, uint32_t index_count,
-                                       uint32_t flags) {
-	CheckBuffer();
-
+void CommandProcessor::DrawIndexOffset(uint32_t index_offset, uint32_t index_count) {
 	uint64_t index_size = 0;
 	switch (m_index_type_and_size) {
 		case 0: index_size = 2; break;
@@ -888,25 +888,10 @@ void CommandProcessor::DrawIndexOffset(uint32_t index_offset, uint32_t index_cou
 	auto* index_addr = reinterpret_cast<const void*>(
 	    m_index_base_addr + static_cast<uint64_t>(index_offset) * index_size);
 
-	m_renderer.GetRenderExecutor().DrawIndex(m_submit_id, CurrentBuffer(), m_index_type_and_size,
-	                                         index_count, index_addr, flags, 1, m_num_instances);
+	DrawIndex({.index_count = index_count, .index_addr = index_addr});
 }
 
 void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiator, bool indexed) {
-	struct DrawIndirectArgs {
-		uint32_t vertex_count_per_instance;
-		uint32_t instance_count;
-		uint32_t start_vertex_location;
-		uint32_t start_instance_location;
-	};
-	struct DrawIndexedIndirectArgs {
-		uint32_t index_count_per_instance;
-		uint32_t instance_count;
-		uint32_t start_index_location;
-		uint32_t base_vertex_location;
-		uint32_t start_instance_location;
-	};
-
 	EXIT_NOT_IMPLEMENTED((draw_initiator & ~0x20u) != 2u);
 	EXIT_NOT_IMPLEMENTED(m_draw_indirect_args_base_addr == 0);
 
@@ -928,8 +913,11 @@ void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiato
 			}
 		}
 		m_num_instances = args.instance_count;
-		SubmitNonIndexedDraw(args.vertex_count_per_instance, 0, 0, args.start_vertex_location,
-		                     args.start_instance_location);
+		DrawIndexAuto({.vertex_count   = args.vertex_count_per_instance,
+		               .instance_count = args.instance_count,
+		               .first_vertex   = args.start_vertex_location,
+		               .first_instance = args.start_instance_location,
+		               .offset_source  = DrawOffsetSource::IndirectArgs});
 		return;
 	}
 
@@ -970,28 +958,18 @@ void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiato
 	}
 
 	m_num_instances = args.instance_count;
-	DrawIndex(index_count, index_addr, 0, 1, args.instance_count, nullptr, 0,
-	          static_cast<int32_t>(args.base_vertex_location), args.start_instance_location);
+	DrawIndex({.index_count    = index_count,
+	           .index_addr     = index_addr,
+	           .instance_count = args.instance_count,
+	           .base_vertex    = static_cast<int32_t>(args.base_vertex_location),
+	           .first_instance = args.start_instance_location,
+	           .offset_source  = DrawOffsetSource::IndirectArgs});
 }
 
 void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_count_or_count,
                                          const volatile uint32_t* count_addr,
                                          uint32_t stride_in_bytes, uint32_t draw_initiator,
                                          bool indexed) {
-	struct DrawIndirectArgs {
-		uint32_t vertex_count_per_instance;
-		uint32_t instance_count;
-		uint32_t start_vertex_location;
-		uint32_t start_instance_location;
-	};
-	struct DrawIndexedIndirectArgs {
-		uint32_t index_count_per_instance;
-		uint32_t instance_count;
-		uint32_t start_index_location;
-		uint32_t base_vertex_location;
-		uint32_t start_instance_location;
-	};
-
 	EXIT_NOT_IMPLEMENTED((draw_initiator & ~0x20u) != 2u);
 	EXIT_NOT_IMPLEMENTED(m_draw_indirect_args_base_addr == 0);
 
@@ -1028,8 +1006,11 @@ void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_coun
 				}
 			}
 			m_num_instances = args->instance_count;
-			SubmitNonIndexedDraw(args->vertex_count_per_instance, 0, 0, args->start_vertex_location,
-			                     args->start_instance_location);
+			DrawIndexAuto({.vertex_count   = args->vertex_count_per_instance,
+			               .instance_count = args->instance_count,
+			               .first_vertex   = args->start_vertex_location,
+			               .first_instance = args->start_instance_location,
+			               .offset_source  = DrawOffsetSource::IndirectArgs});
 			continue;
 		}
 
@@ -1071,8 +1052,12 @@ void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_coun
 		}
 
 		m_num_instances = args->instance_count;
-		DrawIndex(index_count, index_addr, 0, 1, args->instance_count, nullptr, 0,
-		          static_cast<int32_t>(args->base_vertex_location), args->start_instance_location);
+		DrawIndex({.index_count    = index_count,
+		           .index_addr     = index_addr,
+		           .instance_count = args->instance_count,
+		           .base_vertex    = static_cast<int32_t>(args->base_vertex_location),
+		           .first_instance = args->start_instance_location,
+		           .offset_source  = DrawOffsetSource::IndirectArgs});
 	}
 }
 
@@ -1163,19 +1148,13 @@ void CommandProcessor::DispatchIndirect(uint32_t data_offset, uint32_t mode) {
 	DispatchDirect(args->thread_group_x, args->thread_group_y, args->thread_group_z, mode);
 }
 
-void CommandProcessor::DrawIndexAuto(uint32_t index_count, uint32_t flags,
-                                     uint32_t render_target_slice_offset) {
-	SubmitNonIndexedDraw(index_count, flags, render_target_slice_offset, 0, 0);
-}
-
-void CommandProcessor::SubmitNonIndexedDraw(uint32_t vertex_count, uint32_t flags,
-                                            uint32_t render_target_slice_offset,
-                                            uint32_t first_vertex, uint32_t first_instance) {
+void CommandProcessor::DrawIndexAuto(DrawAutoArgs args) {
 	CheckBuffer();
 
-	m_renderer.GetRenderExecutor().DrawAuto(m_submit_id, CurrentBuffer(), vertex_count, flags,
-	                                        render_target_slice_offset, m_num_instances,
-	                                        first_vertex, first_instance);
+	if (args.instance_count == 0) {
+		args.instance_count = m_num_instances;
+	}
+	m_renderer.GetRenderExecutor().DrawAuto(m_submit_id, CurrentBuffer(), args);
 }
 
 void CommandProcessor::WaitFlipDone(uint32_t video_out_handle, uint32_t display_buffer_index) {

src/graphics/guest_gpu/hardwareContext.h

@@ -11,15 +11,6 @@ struct ColorBase {
 	uint64_t addr = 0;
 };
 
-struct ColorPitch {
-	uint32_t pitch_div8_minus1       = 0;
-	uint32_t fmask_pitch_div8_minus1 = 0;
-};
-
-struct ColorSlice {
-	uint32_t slice_div64_minus1 = 0;
-};
-
 struct ColorView {
 	uint32_t base_array_slice_index = 0;
 	uint32_t last_array_slice_index = 0;
@@ -36,20 +27,15 @@ struct ColorInfo {
 	bool                    blend_clamp                    = false;
 	bool                    blend_bypass                   = false;
 	bool                    round_mode                     = false;
-	uint32_t                cmask_is_linear                = 0;
-	uint32_t                cmask_addr_type                = 0;
-	bool                    alt_tile_mode                  = false;
 	Prospero::ChannelLayout format                         = Prospero::ChannelLayout::kInvalid;
 	Prospero::ChannelType   channel_type                   = Prospero::ChannelType::kUNorm;
 	Prospero::ChannelOrder  channel_order                  = Prospero::ChannelOrder::kStandard;
 };
 
 struct ColorAttrib {
-	bool               force_dest_alpha_to_one = false;
-	Prospero::TileMode tile_mode               = Prospero::TileMode::kLinear;
-	Prospero::TileMode fmask_tile_mode         = Prospero::TileMode::kLinear;
-	uint32_t           num_samples             = 0;
-	uint32_t           num_fragments           = 0;
+	bool     force_dest_alpha_to_one = false;
+	uint32_t num_samples             = 0;
+	uint32_t num_fragments           = 0;
 };
 
 struct ColorAttrib2 {
@@ -59,41 +45,33 @@ struct ColorAttrib2 {
 };
 
 struct ColorAttrib3 {
-	uint32_t           depth              = 0;
-	Prospero::TileMode tile_mode          = Prospero::TileMode::kLinear;
-	uint32_t           dimension          = 0;
-	bool               cmask_pipe_aligned = false;
-	bool               dcc_pipe_aligned   = false;
+	uint32_t           depth                        = 0;
+	Prospero::TileMode tile_mode                    = Prospero::TileMode::kRenderTarget;
+	uint32_t           dimension                    = 0;
+	bool               metadata_pipe_aligned        = false;
+	bool               write_vrs_rate_hint_to_cmask = false;
 };
 
 struct ColorDccControl {
-	uint32_t max_uncompressed_block_size    = 0;
-	uint32_t max_compressed_block_size      = 0;
-	uint32_t min_compressed_block_size      = 0;
-	uint32_t color_transform                = 0;
-	bool     dcc_clear_key_enable           = false;
-	bool     overwrite_combiner_disable     = false;
-	bool     independent_64b_blocks         = false;
-	bool     independent_128b_blocks        = false;
-	bool     data_write_on_dcc_clear_to_reg = false;
+	enum class IndependentBlockSize : uint8_t { Disabled, Bytes64, Bytes128 };
+
+	uint32_t             max_uncompressed_block_size    = 2;
+	uint32_t             max_compressed_block_size      = 2;
+	uint32_t             color_transform                = 0;
+	bool                 dcc_clear_key_enable           = false;
+	bool                 overwrite_combiner_disable     = false;
+	IndependentBlockSize independent_block_size         = IndependentBlockSize::Disabled;
+	bool                 data_write_on_dcc_clear_to_reg = false;
 };
 
 struct ColorCmask {
 	uint64_t addr = 0;
 };
 
-struct ColorCmaskSlice {
-	uint32_t slice_minus1 = 0;
-};
-
 struct ColorFmask {
 	uint64_t addr = 0;
 };
 
-struct ColorFmaskSlice {
-	uint32_t slice_minus1 = 0;
-};
-
 struct ColorClearWord0 {
 	uint32_t word0 = 0;
 };
@@ -106,15 +84,8 @@ struct ColorDccAddr {
 	uint64_t addr = 0;
 };
 
-struct ColorSize {
-	uint32_t width  = 0;
-	uint32_t height = 0;
-};
-
 struct RenderTarget {
 	ColorBase       base;
-	ColorPitch      pitch;
-	ColorSlice      slice;
 	ColorView       view;
 	ColorInfo       info;
 	ColorAttrib     attrib;
@@ -122,13 +93,10 @@ struct RenderTarget {
 	ColorAttrib3    attrib3;
 	ColorDccControl dcc;
 	ColorCmask      cmask;
-	ColorCmaskSlice cmask_slice;
 	ColorFmask      fmask;
-	ColorFmaskSlice fmask_slice;
 	ColorClearWord0 clear_word0;
 	ColorClearWord1 clear_word1;
 	ColorDccAddr    dcc_addr;
-	ColorSize       size;
 };
 
 struct DepthZInfo {
@@ -136,7 +104,7 @@ struct DepthZInfo {
 	uint32_t                                    num_samples = 0;
 	Prospero::TextureCompatiblePlaneCompression texture_compatibility =
 	    Prospero::TextureCompatiblePlaneCompression::kDisable;
-	Prospero::ZCompareBase z_compare_base     = Prospero::ZCompareBase::kZMin;
+	Prospero::ZCompareBase z_compare_base     = Prospero::ZCompareBase::kZMax;
 	bool                   htile_acceleration = false;
 	bool                   expclear_enabled   = false;
 	bool                   partially_resident = false;
@@ -171,7 +139,7 @@ struct DepthStencilInfo {
 	Prospero::TextureCompatibleStencil texture_compatibility =
 	    Prospero::TextureCompatibleStencil::kDisable;
 	bool expclear_enabled       = false;
-	bool htile_stencil_disabled = false;
+	bool htile_stencil_disabled = true;
 	bool partially_resident     = false;
 
 	[[nodiscard]] static DepthStencilInfo Decode(uint32_t value) {
@@ -194,16 +162,6 @@ struct DepthStencilInfo {
 	}
 };
 
-struct DepthRenderTargetDepthInfo {
-	uint32_t addr5_swizzle_mask = 0;
-	uint32_t array_mode         = 0;
-	uint32_t pipe_config        = 0;
-	uint32_t bank_width         = 0;
-	uint32_t bank_height        = 0;
-	uint32_t macro_tile_aspect  = 0;
-	uint32_t num_banks          = 0;
-};
-
 struct DepthDepthView {
 	uint32_t slice_start           = 0;
 	uint32_t slice_max             = 0;
@@ -218,36 +176,18 @@ struct DepthDepthSizeXY {
 	bool     valid = false;
 };
 
-struct DepthRenderTargetHTileSurface {
-	uint32_t linear                  = 0;
-	uint32_t full_cache              = 0;
-	uint32_t htile_uses_preload_win  = 0;
-	uint32_t preload                 = 0;
-	uint32_t prefetch_width          = 0;
-	uint32_t prefetch_height         = 0;
-	uint32_t dst_outside_zero_to_one = 0;
-};
-
 struct DepthRenderTarget {
-	DepthZInfo                    z_info;
-	DepthStencilInfo              stencil_info;
-	DepthDepthView                depth_view;
-	DepthDepthSizeXY              size;
-	DepthRenderTargetDepthInfo    depth_info;
-	DepthRenderTargetHTileSurface htile_surface;
+	DepthZInfo       z_info;
+	DepthStencilInfo stencil_info;
+	DepthDepthView   depth_view;
+	DepthDepthSizeXY size;
 
 	uint64_t z_read_base_addr        = 0;
 	uint64_t stencil_read_base_addr  = 0;
 	uint64_t z_write_base_addr       = 0;
 	uint64_t stencil_write_base_addr = 0;
-	uint32_t pitch_div8_minus1       = 0;
-	uint32_t height_div8_minus1      = 0;
-	bool     pitch_height_valid      = false;
-	uint32_t slice_div64_minus1      = 0;
 	uint64_t htile_data_base_addr    = 0;
-	uint32_t width                   = 0;
-	uint32_t height                  = 0;
-	bool     width_height_valid      = false;
+	uint8_t  shading_rate_encoding   = 0;
 };
 
 struct RenderControl {
@@ -256,6 +196,8 @@ struct RenderControl {
 	bool    resummarize_enable       = false;
 	bool    stencil_compress_disable = false;
 	bool    depth_compress_disable   = false;
+	bool    copy_depth_to_color      = false;
+	bool    copy_stencil_to_color    = false;
 	bool    copy_centroid            = false;
 	uint8_t copy_sample              = 0;
 };
@@ -299,16 +241,14 @@ struct ClipControl {
 };
 
 struct DepthControl {
-	bool    stencil_enable                     = false;
-	bool    z_enable                           = false;
-	bool    z_write_enable                     = false;
-	bool    depth_bounds_enable                = false;
-	uint8_t zfunc                              = 0;
-	bool    backface_enable                    = false;
-	uint8_t stencilfunc                        = 0;
-	uint8_t stencilfunc_bf                     = 0;
-	bool    color_writes_on_depth_fail_enable  = false;
-	bool    color_writes_on_depth_pass_disable = false;
+	bool    stencil_enable      = false;
+	bool    z_enable            = false;
+	bool    z_write_enable      = false;
+	bool    depth_bounds_enable = false;
+	uint8_t zfunc               = 0;
+	bool    backface_enable     = false;
+	uint8_t stencilfunc         = 0;
+	uint8_t stencilfunc_bf      = 0;
 };
 
 struct StencilControl {
@@ -468,27 +408,6 @@ struct ScreenViewport {
 	bool     clip_rect_window_offset_enable[4]    = {};
 };
 
-struct VsShaderResource1 {
-	uint8_t vgprs                    = 0;
-	uint8_t sgprs                    = 0;
-	uint8_t priority                 = 0;
-	uint8_t float_mode               = 0;
-	bool    dx10_clamp               = false;
-	bool    ieee_mode                = false;
-	uint8_t vgpr_component_count     = 0;
-	bool    cu_group_enable          = false;
-	bool    require_forward_progress = false;
-	bool    fp16_overflow            = false;
-};
-
-struct VsShaderResource2 {
-	bool    scratch_en        = false;
-	uint8_t user_sgpr         = 0;
-	bool    offchip_lds       = false;
-	bool    streamout_enabled = false;
-	uint8_t shared_vgprs      = 0;
-};
-
 struct HsShaderResource1 {
 	uint8_t vgprs                     = 0;
 	uint8_t priority                  = 0;
@@ -511,7 +430,6 @@ struct HsShaderResource2 {
 
 struct PsShaderResource1 {
 	uint8_t vgprs                    = 0;
-	uint8_t sgprs                    = 0;
 	uint8_t priority                 = 0;
 	uint8_t float_mode               = 0;
 	bool    dx10_clamp               = false;
@@ -535,7 +453,6 @@ struct PsStageRegisters {
 	uint64_t          data_addr = 0;
 	PsShaderResource1 rsrc1;
 	PsShaderResource2 rsrc2;
-	uint64_t          chksum = 0;
 };
 
 struct CsStageRegisters {
@@ -545,12 +462,13 @@ struct CsStageRegisters {
 	uint32_t num_thread_y              = 0;
 	uint32_t num_thread_z              = 0;
 	uint8_t  vgprs                     = 0;
-	uint8_t  sgprs                     = 0;
+	uint8_t  priority                  = 0;
 	uint8_t  float_mode                = 0;
 	bool     dx10_clamp                = false;
+	bool     debug_mode                = false;
 	bool     ieee_mode                 = false;
+	bool     require_forward_progress  = false;
 	bool     fp16_overflow             = false;
-	bool     bulky                     = false;
 	bool     threadgroup_configuration = false;
 	uint8_t  wave_size                 = 64;
 	bool     scratch_en                = false;
@@ -561,6 +479,7 @@ struct CsStageRegisters {
 	bool     tg_size_en                = false;
 	uint8_t  tidig_comp_cnt            = 0;
 	uint16_t lds_size                  = 0;
+	uint8_t  shared_vgprs              = 0;
 };
 
 struct EsStageRegisters {
@@ -568,21 +487,17 @@ struct EsStageRegisters {
 };
 
 struct LsStageRegisters {
-	uint64_t          data_addr = 0;
-	VsShaderResource1 rsrc1;
-	VsShaderResource2 rsrc2;
+	uint64_t data_addr = 0;
 };
 
 struct HsStageRegisters {
 	uint64_t          data_addr = 0;
 	HsShaderResource1 rsrc1;
 	HsShaderResource2 rsrc2;
-	uint64_t          chksum = 0;
 };
 
 struct GsShaderResource1 {
 	uint8_t vgprs                     = 0;
-	uint8_t sgprs                     = 0;
 	uint8_t priority                  = 0;
 	uint8_t float_mode                = 0;
 	bool    dx10_clamp                = false;
@@ -608,7 +523,6 @@ struct GsStageRegisters {
 	uint64_t          data_addr = 0;
 	GsShaderResource1 rsrc1;
 	GsShaderResource2 rsrc2;
-	uint64_t          chksum = 0;
 };
 
 struct ShaderRegisters {
@@ -664,8 +578,7 @@ struct ShaderRegisters {
 enum class UserSgprType { Unknown, Region, Vsharp };
 
 struct UserSgprInfo {
-	// Changed from 16 to 64
-	static constexpr int SGPRS_MAX = 64;
+	static constexpr int SGPRS_MAX = 32;
 
 	uint32_t     value[SGPRS_MAX] = {0};
 	UserSgprType type[SGPRS_MAX]  = {};
@@ -712,12 +625,6 @@ public:
 	void Reset() { *this = Context(); }
 
 	void SetColorBase(uint32_t slot, const ColorBase& base) { m_render_targets[slot].base = base; }
-	void SetColorPitch(uint32_t slot, const ColorPitch& pitch) {
-		m_render_targets[slot].pitch = pitch;
-	}
-	void SetColorSlice(uint32_t slot, const ColorSlice& slice) {
-		m_render_targets[slot].slice = slice;
-	}
 	void SetColorView(uint32_t slot, const ColorView& view) { m_render_targets[slot].view = view; }
 	void SetColorInfo(uint32_t slot, const ColorInfo& info) { m_render_targets[slot].info = info; }
 	void SetColorAttrib(uint32_t slot, const ColorAttrib& attrib) {
@@ -735,15 +642,9 @@ public:
 	void SetColorCmask(uint32_t slot, const ColorCmask& cmask) {
 		m_render_targets[slot].cmask = cmask;
 	}
-	void SetColorCmaskSlice(uint32_t slot, const ColorCmaskSlice& cmask_slice) {
-		m_render_targets[slot].cmask_slice = cmask_slice;
-	}
 	void SetColorFmask(uint32_t slot, const ColorFmask& fmask) {
 		m_render_targets[slot].fmask = fmask;
 	}
-	void SetColorFmaskSlice(uint32_t slot, const ColorFmaskSlice& fmask_slice) {
-		m_render_targets[slot].fmask_slice = fmask_slice;
-	}
 	void SetColorClearWord0(uint32_t slot, const ColorClearWord0& clear_word0) {
 		m_render_targets[slot].clear_word0 = clear_word0;
 	}
@@ -753,7 +654,6 @@ public:
 	void SetColorDccAddr(uint32_t slot, const ColorDccAddr& dcc_addr) {
 		m_render_targets[slot].dcc_addr = dcc_addr;
 	}
-	void SetColorSize(uint32_t slot, const ColorSize& size) { m_render_targets[slot].size = size; }
 	[[nodiscard]] const RenderTarget& GetRenderTarget(uint32_t slot) const {
 		return m_render_targets[slot];
 	}
@@ -1072,21 +972,16 @@ public:
 
 	void SetEsShaderBase(uint64_t addr) { m_vs.es_regs.data_addr = addr; }
 	void SetLsShaderBase(uint64_t addr) { m_vs.ls_regs.data_addr = addr; }
-	void SetLsShaderResource1(const VsShaderResource1& rsrc1) { m_vs.ls_regs.rsrc1 = rsrc1; }
-	void SetLsShaderResource2(const VsShaderResource2& rsrc2) { m_vs.ls_regs.rsrc2 = rsrc2; }
 	void SetHsShaderBase(uint64_t addr) { m_vs.hs_regs.data_addr = addr; }
 	void SetHsShaderResource1(const HsShaderResource1& rsrc1) { m_vs.hs_regs.rsrc1 = rsrc1; }
 	void SetHsShaderResource2(const HsShaderResource2& rsrc2) { m_vs.hs_regs.rsrc2 = rsrc2; }
-	void SetHsShaderChksum(uint32_t value) { m_vs.hs_regs.chksum = value; }
 	void SetGsShaderBase(uint64_t addr) { m_vs.gs_regs.data_addr = addr; }
 	void SetGsShaderResource1(const GsShaderResource1& rsrc1) { m_vs.gs_regs.rsrc1 = rsrc1; }
 	void SetGsShaderResource2(const GsShaderResource2& rsrc2) { m_vs.gs_regs.rsrc2 = rsrc2; }
-	void SetGsShaderChksum(uint32_t value) { m_vs.gs_regs.chksum = value; }
 
 	void SetPsShaderBase(uint64_t addr) { m_ps.ps_regs.data_addr = addr; }
 	void SetPsShaderResource1(const PsShaderResource1& rsrc1) { m_ps.ps_regs.rsrc1 = rsrc1; }
 	void SetPsShaderResource2(const PsShaderResource2& rsrc2) { m_ps.ps_regs.rsrc2 = rsrc2; }
-	void SetPsShaderChksum(uint32_t value) { m_ps.ps_regs.chksum = value; }
 
 	void SetCsShader(const CsStageRegisters& cs_regs) { m_cs.cs_regs = cs_regs; }
 	void SetCsWaveSize(uint8_t wave_size) { m_cs.cs_regs.wave_size = wave_size; }

src/graphics/guest_gpu/pm4.h

@@ -98,6 +98,10 @@ constexpr uint32_t DB_RENDER_CONTROL_DEPTH_CLEAR_ENABLE_SHIFT       = 0;
 constexpr uint32_t DB_RENDER_CONTROL_DEPTH_CLEAR_ENABLE_MASK        = 0x1;
 constexpr uint32_t DB_RENDER_CONTROL_STENCIL_CLEAR_ENABLE_SHIFT     = 1;
 constexpr uint32_t DB_RENDER_CONTROL_STENCIL_CLEAR_ENABLE_MASK      = 0x1;
+constexpr uint32_t DB_RENDER_CONTROL_COPY_DEPTH_TO_COLOR_SHIFT      = 2;
+constexpr uint32_t DB_RENDER_CONTROL_COPY_DEPTH_TO_COLOR_MASK       = 0x1;
+constexpr uint32_t DB_RENDER_CONTROL_COPY_STENCIL_TO_COLOR_SHIFT    = 3;
+constexpr uint32_t DB_RENDER_CONTROL_COPY_STENCIL_TO_COLOR_MASK     = 0x1;
 constexpr uint32_t DB_RENDER_CONTROL_RESUMMARIZE_ENABLE_SHIFT       = 4;
 constexpr uint32_t DB_RENDER_CONTROL_RESUMMARIZE_ENABLE_MASK        = 0x1;
 constexpr uint32_t DB_RENDER_CONTROL_STENCIL_COMPRESS_DISABLE_SHIFT = 5;
@@ -163,22 +167,6 @@ constexpr uint32_t PA_SC_SCREEN_SCISSOR_BR_BR_Y_MASK  = 0xFFFF;
 
 constexpr uint32_t DB_DFSM_CONTROL = 0xE;
 
-constexpr uint32_t DB_DEPTH_INFO                          = 0xF;
-constexpr uint32_t DB_DEPTH_INFO_ADDR5_SWIZZLE_MASK_SHIFT = 0;
-constexpr uint32_t DB_DEPTH_INFO_ADDR5_SWIZZLE_MASK_MASK  = 0xF;
-constexpr uint32_t DB_DEPTH_INFO_ARRAY_MODE_SHIFT         = 4;
-constexpr uint32_t DB_DEPTH_INFO_ARRAY_MODE_MASK          = 0xF;
-constexpr uint32_t DB_DEPTH_INFO_PIPE_CONFIG_SHIFT        = 8;
-constexpr uint32_t DB_DEPTH_INFO_PIPE_CONFIG_MASK         = 0x1F;
-constexpr uint32_t DB_DEPTH_INFO_BANK_WIDTH_SHIFT         = 13;
-constexpr uint32_t DB_DEPTH_INFO_BANK_WIDTH_MASK          = 0x3;
-constexpr uint32_t DB_DEPTH_INFO_BANK_HEIGHT_SHIFT        = 15;
-constexpr uint32_t DB_DEPTH_INFO_BANK_HEIGHT_MASK         = 0x3;
-constexpr uint32_t DB_DEPTH_INFO_MACRO_TILE_ASPECT_SHIFT  = 17;
-constexpr uint32_t DB_DEPTH_INFO_MACRO_TILE_ASPECT_MASK   = 0x3;
-constexpr uint32_t DB_DEPTH_INFO_NUM_BANKS_SHIFT          = 19;
-constexpr uint32_t DB_DEPTH_INFO_NUM_BANKS_MASK           = 0x3;
-
 constexpr uint32_t DB_Z_INFO                               = 0x10;
 constexpr uint32_t DB_Z_INFO_FORMAT_SHIFT                  = 0;
 constexpr uint32_t DB_Z_INFO_FORMAT_MASK                   = 0x3;
@@ -222,16 +210,6 @@ constexpr uint32_t DB_STENCIL_READ_BASE  = 0x13;
 constexpr uint32_t DB_Z_WRITE_BASE       = 0x14;
 constexpr uint32_t DB_STENCIL_WRITE_BASE = 0x15;
 
-constexpr uint32_t DB_DEPTH_SIZE                       = 0x16;
-constexpr uint32_t DB_DEPTH_SIZE_PITCH_TILE_MAX_SHIFT  = 0;
-constexpr uint32_t DB_DEPTH_SIZE_PITCH_TILE_MAX_MASK   = 0x7FF;
-constexpr uint32_t DB_DEPTH_SIZE_HEIGHT_TILE_MAX_SHIFT = 11;
-constexpr uint32_t DB_DEPTH_SIZE_HEIGHT_TILE_MAX_MASK  = 0x7FF;
-
-constexpr uint32_t DB_DEPTH_SLICE                      = 0x17;
-constexpr uint32_t DB_DEPTH_SLICE_SLICE_TILE_MAX_SHIFT = 0;
-constexpr uint32_t DB_DEPTH_SLICE_SLICE_TILE_MAX_MASK  = 0x3FFFFF;
-
 constexpr uint32_t DB_Z_READ_BASE_HI                  = 0x1A;
 constexpr uint32_t DB_STENCIL_READ_BASE_HI            = 0x1B;
 constexpr uint32_t DB_Z_WRITE_BASE_HI                 = 0x1C;
@@ -424,11 +402,6 @@ constexpr uint32_t DB_DEPTH_CONTROL_STENCILFUNC_SHIFT                        = 8
 constexpr uint32_t DB_DEPTH_CONTROL_STENCILFUNC_MASK                         = 0x7;
 constexpr uint32_t DB_DEPTH_CONTROL_STENCILFUNC_BF_SHIFT                     = 20;
 constexpr uint32_t DB_DEPTH_CONTROL_STENCILFUNC_BF_MASK                      = 0x7;
-constexpr uint32_t DB_DEPTH_CONTROL_ENABLE_COLOR_WRITES_ON_DEPTH_FAIL_SHIFT  = 30;
-constexpr uint32_t DB_DEPTH_CONTROL_ENABLE_COLOR_WRITES_ON_DEPTH_FAIL_MASK   = 0x1;
-constexpr uint32_t DB_DEPTH_CONTROL_DISABLE_COLOR_WRITES_ON_DEPTH_PASS_SHIFT = 31;
-constexpr uint32_t DB_DEPTH_CONTROL_DISABLE_COLOR_WRITES_ON_DEPTH_PASS_MASK  = 0x1;
-
 constexpr uint32_t DB_EQAA                                  = 0x201;
 constexpr uint32_t DB_EQAA_MAX_ANCHOR_SAMPLES_SHIFT         = 0;
 constexpr uint32_t DB_EQAA_MAX_ANCHOR_SAMPLES_MASK          = 0x7;
@@ -560,21 +533,9 @@ constexpr uint32_t VGT_DRAW_PAYLOAD_CNTL  = 0x2A6;
 constexpr uint32_t VGT_ESGS_RING_ITEMSIZE = 0x2AB;
 constexpr uint32_t VGT_REUSE_OFF          = 0x2AD;
 
-constexpr uint32_t DB_HTILE_SURFACE                               = 0x2AF;
-constexpr uint32_t DB_HTILE_SURFACE_LINEAR_SHIFT                  = 0;
-constexpr uint32_t DB_HTILE_SURFACE_LINEAR_MASK                   = 0x1;
-constexpr uint32_t DB_HTILE_SURFACE_FULL_CACHE_SHIFT              = 1;
-constexpr uint32_t DB_HTILE_SURFACE_FULL_CACHE_MASK               = 0x1;
-constexpr uint32_t DB_HTILE_SURFACE_HTILE_USES_PRELOAD_WIN_SHIFT  = 2;
-constexpr uint32_t DB_HTILE_SURFACE_HTILE_USES_PRELOAD_WIN_MASK   = 0x1;
-constexpr uint32_t DB_HTILE_SURFACE_PRELOAD_SHIFT                 = 3;
-constexpr uint32_t DB_HTILE_SURFACE_PRELOAD_MASK                  = 0x1;
-constexpr uint32_t DB_HTILE_SURFACE_PREFETCH_WIDTH_SHIFT          = 4;
-constexpr uint32_t DB_HTILE_SURFACE_PREFETCH_WIDTH_MASK           = 0x3F;
-constexpr uint32_t DB_HTILE_SURFACE_PREFETCH_HEIGHT_SHIFT         = 10;
-constexpr uint32_t DB_HTILE_SURFACE_PREFETCH_HEIGHT_MASK          = 0x3F;
-constexpr uint32_t DB_HTILE_SURFACE_DST_OUTSIDE_ZERO_TO_ONE_SHIFT = 16;
-constexpr uint32_t DB_HTILE_SURFACE_DST_OUTSIDE_ZERO_TO_ONE_MASK  = 0x1;
+constexpr uint32_t DB_SHADING_RATE_ENCODING                             = 0x2AF;
+constexpr uint32_t DB_SHADING_RATE_ENCODING_SHADING_RATE_ENCODING_SHIFT = 19;
+constexpr uint32_t DB_SHADING_RATE_ENCODING_SHADING_RATE_ENCODING_MASK  = 0x3;
 
 constexpr uint32_t DB_SRESULTS_COMPARE_STATE0                          = 0x2B0;
 constexpr uint32_t DB_SRESULTS_COMPARE_STATE1                          = 0x2B1;
@@ -621,7 +582,6 @@ constexpr uint32_t PA_SC_SHADER_CONTROL                  = 0x310;
 constexpr uint32_t PA_SC_BINNER_CNTL_0                   = 0x311;
 constexpr uint32_t PA_SC_BINNER_CNTL_1                   = 0x312;
 constexpr uint32_t PA_SC_CONSERVATIVE_RASTERIZATION_CNTL = 0x313;
-constexpr uint32_t PA_SC_NGG_MODE_CNTL                   = 0x314;
 constexpr uint32_t CB_COLOR0_BASE                        = 0x318;
 
 constexpr uint32_t CB_COLOR0_VIEW                   = 0x31B;
@@ -649,24 +609,13 @@ constexpr uint32_t CB_COLOR0_INFO_BLEND_BYPASS_SHIFT              = 16;
 constexpr uint32_t CB_COLOR0_INFO_BLEND_BYPASS_MASK               = 0x1;
 constexpr uint32_t CB_COLOR0_INFO_ROUND_MODE_SHIFT                = 18;
 constexpr uint32_t CB_COLOR0_INFO_ROUND_MODE_MASK                 = 0x1;
-constexpr uint32_t CB_COLOR0_INFO_CMASK_IS_LINEAR_SHIFT           = 19;
-constexpr uint32_t CB_COLOR0_INFO_CMASK_IS_LINEAR_MASK            = 0x1;
 constexpr uint32_t CB_COLOR0_INFO_FMASK_COMPRESSION_DISABLE_SHIFT = 26;
 constexpr uint32_t CB_COLOR0_INFO_FMASK_COMPRESSION_DISABLE_MASK  = 0x1;
 constexpr uint32_t CB_COLOR0_INFO_FMASK_COMPRESS_1FRAG_ONLY_SHIFT = 27;
 constexpr uint32_t CB_COLOR0_INFO_FMASK_COMPRESS_1FRAG_ONLY_MASK  = 0x1;
 constexpr uint32_t CB_COLOR0_INFO_DCC_ENABLE_SHIFT                = 28;
 constexpr uint32_t CB_COLOR0_INFO_DCC_ENABLE_MASK                 = 0x1;
-constexpr uint32_t CB_COLOR0_INFO_CMASK_ADDR_TYPE_SHIFT           = 29;
-constexpr uint32_t CB_COLOR0_INFO_CMASK_ADDR_TYPE_MASK            = 0x3;
-constexpr uint32_t CB_COLOR0_INFO_ALT_TILE_MODE_SHIFT             = 31;
-constexpr uint32_t CB_COLOR0_INFO_ALT_TILE_MODE_MASK              = 0x1;
-
 constexpr uint32_t CB_COLOR0_ATTRIB                             = 0x31D;
-constexpr uint32_t CB_COLOR0_ATTRIB_TILE_MODE_INDEX_SHIFT       = 0;
-constexpr uint32_t CB_COLOR0_ATTRIB_TILE_MODE_INDEX_MASK        = 0x1F;
-constexpr uint32_t CB_COLOR0_ATTRIB_FMASK_TILE_MODE_INDEX_SHIFT = 5;
-constexpr uint32_t CB_COLOR0_ATTRIB_FMASK_TILE_MODE_INDEX_MASK  = 0x1F;
 constexpr uint32_t CB_COLOR0_ATTRIB_NUM_SAMPLES_SHIFT           = 12;
 constexpr uint32_t CB_COLOR0_ATTRIB_NUM_SAMPLES_MASK            = 0x7;
 constexpr uint32_t CB_COLOR0_ATTRIB_NUM_FRAGMENTS_SHIFT         = 15;
@@ -681,8 +630,6 @@ constexpr uint32_t CB_COLOR0_DCC_CONTROL_KEY_CLEAR_ENABLE_SHIFT
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_KEY_CLEAR_ENABLE_MASK                  = 0x1;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_MAX_UNCOMPRESSED_BLOCK_SIZE_SHIFT      = 2;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_MAX_UNCOMPRESSED_BLOCK_SIZE_MASK       = 0x3;
-constexpr uint32_t CB_COLOR0_DCC_CONTROL_MIN_COMPRESSED_BLOCK_SIZE_SHIFT        = 4;
-constexpr uint32_t CB_COLOR0_DCC_CONTROL_MIN_COMPRESSED_BLOCK_SIZE_MASK         = 0x1;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_MAX_COMPRESSED_BLOCK_SIZE_SHIFT        = 5;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_MAX_COMPRESSED_BLOCK_SIZE_MASK         = 0x3;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_COLOR_TRANSFORM_SHIFT                  = 7;
@@ -739,31 +686,25 @@ constexpr uint32_t CB_COLOR0_ATTRIB3_CMASK_PIPE_ALIGNED_SHIFT = 26;
 constexpr uint32_t CB_COLOR0_ATTRIB3_CMASK_PIPE_ALIGNED_MASK  = 0x1;
 constexpr uint32_t CB_COLOR0_ATTRIB3_DCC_PIPE_ALIGNED_SHIFT   = 30;
 constexpr uint32_t CB_COLOR0_ATTRIB3_DCC_PIPE_ALIGNED_MASK    = 0x1;
+constexpr uint32_t CB_COLOR0_ATTRIB3_WRITE_VRS_RATE_HINT_TO_CMASK_SHIFT = 31;
+constexpr uint32_t CB_COLOR0_ATTRIB3_WRITE_VRS_RATE_HINT_TO_CMASK_MASK  = 0x1;
 
 constexpr uint32_t CB_COLOR7_ATTRIB3 = 0x3BF;
 
-/* Fake codes. Don't exist on real HW */
-
-constexpr uint32_t CX_NOP = 0x800003FF;
+constexpr uint32_t CX_NOP = 0xDB;
 
 constexpr uint32_t CX_NUM = 0x3FF + 1;
 
 /* Shader registers */
 
-constexpr uint32_t SPI_SHADER_TBA_LO_PS     = 0x0;
-constexpr uint32_t SPI_SHADER_PGM_RSRC4_PS  = 0x1;
-constexpr uint32_t SPI_SHADER_TMA_LO_PS     = 0x2;
-constexpr uint32_t SPI_SHADER_TMA_HI_PS     = 0x3;
-constexpr uint32_t SPI_SHADER_PGM_CHKSUM_PS = 0x6;
-constexpr uint32_t SPI_SHADER_PGM_RSRC3_PS  = 0x7;
-constexpr uint32_t SPI_SHADER_PGM_LO_PS     = 0x8;
-constexpr uint32_t SPI_SHADER_PGM_HI_PS     = 0x9;
+constexpr uint32_t SPI_SHADER_PACE_ID_PS            = 0x6;
+constexpr uint32_t SPI_GRAPHICS_SHADER_CONTROL_PS    = 0x7;
+constexpr uint32_t SPI_SHADER_PGM_LO_PS              = 0x8;
+constexpr uint32_t SPI_SHADER_PGM_HI_PS              = 0x9;
 
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS                        = 0xA;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_VGPRS_SHIFT            = 0;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_VGPRS_MASK             = 0x3F;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_SGPRS_SHIFT            = 6;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_SGPRS_MASK             = 0xF;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_PRIORITY_SHIFT         = 10;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_PRIORITY_MASK          = 0x3;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_FLOAT_MODE_SHIFT       = 12;
@@ -800,14 +741,11 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_SHARED_VGPR_CNT_MASK         = 0xF;
 constexpr uint32_t SPI_SHADER_USER_DATA_PS_0  = 0xC;
 constexpr uint32_t SPI_SHADER_USER_DATA_PS_15 = 0x1B;
 constexpr uint32_t SPI_SHADER_USER_DATA_PS_31 = 0x2B;
-constexpr uint32_t SPI_SHADER_REQ_CTRL_PS     = 0x30;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_PS_0 = 0x32;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_PS_3 = 0x35;
 
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_VGPRS_SHIFT           = 0;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_VGPRS_MASK            = 0x3F;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_SGPRS_SHIFT           = 6;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_SGPRS_MASK            = 0xF;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_PRIORITY_SHIFT        = 10;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_PRIORITY_MASK         = 0x3;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_FLOAT_MODE_SHIFT      = 12;
@@ -838,19 +776,20 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC2_VS_USER_SGPR_MSB_MASK    = 0x1;
 constexpr uint32_t SPI_SHADER_PGM_RSRC2_VS_SHARED_VGPR_CNT_SHIFT = 28;
 constexpr uint32_t SPI_SHADER_PGM_RSRC2_VS_SHARED_VGPR_CNT_MASK  = 0xF;
 
-constexpr uint32_t SPI_SHADER_PGM_CHKSUM_GS        = 0x80;
+constexpr uint32_t SPI_SHADER_PACE_ID_GS           = 0x80;
+// AGC compiler records use this numeric tag as opaque metadata. It is not a shader checksum
+// register and intentionally has no checksum-state consumer in HardwareContext.
+constexpr uint32_t SPI_SHADER_PGM_CHKSUM_GS        = SPI_SHADER_PACE_ID_GS;
 constexpr uint32_t SPI_SHADER_PGM_RSRC4_GS         = 0x81;
 constexpr uint32_t SPI_SHADER_USER_DATA_ADDR_LO_GS = 0x82;
 constexpr uint32_t SPI_SHADER_USER_DATA_ADDR_HI_GS = 0x83;
-constexpr uint32_t SPI_SHADER_PGM_RSRC3_GS         = 0x87;
+constexpr uint32_t SPI_GRAPHICS_SHADER_CONTROL_GS  = 0x87;
 constexpr uint32_t SPI_SHADER_PGM_LO_GS            = 0x88;
 constexpr uint32_t SPI_SHADER_PGM_HI_GS            = 0x89;
 
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS                        = 0x8A;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_VGPRS_SHIFT            = 0;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_VGPRS_MASK             = 0x3F;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_SGPRS_SHIFT            = 6;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_SGPRS_MASK             = 0xF;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_PRIORITY_SHIFT         = 10;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_PRIORITY_MASK          = 0x3;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_FLOAT_MODE_SHIFT       = 12;
@@ -891,22 +830,16 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC2_GS_SHARED_VGPR_CNT_MASK   = 0xF;
 constexpr uint32_t SPI_SHADER_USER_DATA_GS_0                      = 0x8C;
 constexpr uint32_t SPI_SHADER_USER_DATA_GS_15                     = 0x9B;
 constexpr uint32_t SPI_SHADER_USER_DATA_GS_31                     = 0xAB;
-constexpr uint32_t SPI_SHADER_REQ_CTRL_ESGS                       = 0xB0;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_ESGS_0                   = 0xB2;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_ESGS_3                   = 0xB5;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_ES_GS                     = 0xBC;
 constexpr uint32_t SPI_SHADER_PGM_LO_ES                           = 0xC8;
 constexpr uint32_t SPI_SHADER_PGM_HI_ES                           = 0xC9;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_ES                        = 0xCA;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_ES                        = 0xCB;
-constexpr uint32_t SPI_SHADER_USER_DATA_ES_0                      = 0xCC;
-constexpr uint32_t SPI_SHADER_USER_DATA_ES_15                     = 0xDB;
-constexpr uint32_t SPI_SHADER_USER_DATA_ES_31                     = 0xEB;
+// Private AGC registers used by shader binaries submitted through SET_SH_REG_INDIRECT.
 constexpr uint32_t SPI_SHADER_PGM_CHKSUM_HS                       = 0x100;
 constexpr uint32_t SPI_SHADER_PGM_RSRC4_HS                        = 0x101;
 constexpr uint32_t SPI_SHADER_USER_DATA_ADDR_LO_HS                = 0x102;
 constexpr uint32_t SPI_SHADER_USER_DATA_ADDR_HI_HS                = 0x103;
-constexpr uint32_t SPI_SHADER_PGM_RSRC3_HS                        = 0x107;
+constexpr uint32_t SPI_GRAPHICS_SHADER_CONTROL_HS                 = 0x107;
 constexpr uint32_t SPI_SHADER_PGM_LO_HS                           = 0x108;
 constexpr uint32_t SPI_SHADER_PGM_HI_HS                           = 0x109;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS                        = 0x10A;
@@ -944,13 +877,10 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_SHARED_VGPR_CNT_MASK   = 0xF;
 constexpr uint32_t SPI_SHADER_USER_DATA_HS_0                      = 0x10C;
 constexpr uint32_t SPI_SHADER_USER_DATA_HS_15                     = 0x11B;
 constexpr uint32_t SPI_SHADER_USER_DATA_HS_31                     = 0x12B;
-constexpr uint32_t SPI_SHADER_REQ_CTRL_LSHS                       = 0x130;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_LSHS_0                   = 0x132;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_LSHS_3                   = 0x135;
 constexpr uint32_t SPI_SHADER_PGM_LO_LS                           = 0x148;
 constexpr uint32_t SPI_SHADER_PGM_HI_LS                           = 0x149;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_LS                        = 0x14A;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_LS                        = 0x14B;
 constexpr uint32_t COMPUTE_START_X                                = 0x204;
 constexpr uint32_t COMPUTE_START_Y                                = 0x205;
 constexpr uint32_t COMPUTE_START_Z                                = 0x206;
@@ -959,28 +889,26 @@ constexpr uint32_t COMPUTE_NUM_THREAD_Y                           = 0x208;
 constexpr uint32_t COMPUTE_NUM_THREAD_Z                           = 0x209;
 constexpr uint32_t COMPUTE_PGM_LO                                 = 0x20C;
 constexpr uint32_t COMPUTE_PGM_HI                                 = 0x20D;
-constexpr uint32_t COMPUTE_TBA_LO                                 = 0x20E;
-constexpr uint32_t COMPUTE_TBA_HI                                 = 0x20F;
-constexpr uint32_t COMPUTE_TMA_LO                                 = 0x210;
-constexpr uint32_t COMPUTE_TMA_HI                                 = 0x211;
 
 constexpr uint32_t COMPUTE_PGM_RSRC1                  = 0x212;
 constexpr uint32_t COMPUTE_PGM_RSRC1_VGPRS_SHIFT      = 0;
 constexpr uint32_t COMPUTE_PGM_RSRC1_VGPRS_MASK       = 0x3F;
-constexpr uint32_t COMPUTE_PGM_RSRC1_SGPRS_SHIFT      = 6;
-constexpr uint32_t COMPUTE_PGM_RSRC1_SGPRS_MASK       = 0xF;
+constexpr uint32_t COMPUTE_PGM_RSRC1_PRIORITY_SHIFT   = 10;
+constexpr uint32_t COMPUTE_PGM_RSRC1_PRIORITY_MASK    = 0x3;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FLOAT_MODE_SHIFT = 12;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FLOAT_MODE_MASK  = 0xFF;
 constexpr uint32_t COMPUTE_PGM_RSRC1_DX10_CLAMP_SHIFT = 21;
 constexpr uint32_t COMPUTE_PGM_RSRC1_DX10_CLAMP_MASK  = 0x1;
+constexpr uint32_t COMPUTE_PGM_RSRC1_DEBUG_MODE_SHIFT = 22;
+constexpr uint32_t COMPUTE_PGM_RSRC1_DEBUG_MODE_MASK  = 0x1;
 constexpr uint32_t COMPUTE_PGM_RSRC1_IEEE_MODE_SHIFT  = 23;
 constexpr uint32_t COMPUTE_PGM_RSRC1_IEEE_MODE_MASK   = 0x1;
-constexpr uint32_t COMPUTE_PGM_RSRC1_BULKY_SHIFT      = 24;
-constexpr uint32_t COMPUTE_PGM_RSRC1_BULKY_MASK       = 0x1;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FP16_OVFL_SHIFT  = 26;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FP16_OVFL_MASK   = 0x1;
 constexpr uint32_t COMPUTE_PGM_RSRC1_WGP_MODE_SHIFT   = 29;
 constexpr uint32_t COMPUTE_PGM_RSRC1_WGP_MODE_MASK    = 0x1;
+constexpr uint32_t COMPUTE_PGM_RSRC1_FWD_PROGRESS_SHIFT = 31;
+constexpr uint32_t COMPUTE_PGM_RSRC1_FWD_PROGRESS_MASK  = 0x1;
 
 constexpr uint32_t COMPUTE_DISPATCH_INITIATOR_CS_W32_EN_SHIFT = 15;
 constexpr uint32_t COMPUTE_DISPATCH_INITIATOR_CS_W32_EN_MASK  = 0x1;
@@ -1011,21 +939,16 @@ constexpr uint32_t COMPUTE_PGM_RSRC2_LDS_SIZE_SHIFT       = 15;
 constexpr uint32_t COMPUTE_PGM_RSRC2_LDS_SIZE_MASK        = 0x1FF;
 
 constexpr uint32_t COMPUTE_RESOURCE_LIMITS    = 0x215;
-constexpr uint32_t COMPUTE_DESTINATION_EN_SE0 = 0x216;
-constexpr uint32_t COMPUTE_DESTINATION_EN_SE1 = 0x217;
 constexpr uint32_t COMPUTE_TMPRING_SIZE       = 0x218;
-constexpr uint32_t COMPUTE_DESTINATION_EN_SE2 = 0x219;
-constexpr uint32_t COMPUTE_DESTINATION_EN_SE3 = 0x21A;
 constexpr uint32_t COMPUTE_USER_ACCUM_0       = 0x224;
 constexpr uint32_t COMPUTE_PGM_RSRC3          = 0x228;
-constexpr uint32_t COMPUTE_SHADER_CHKSUM      = 0x22A;
+constexpr uint32_t COMPUTE_PGM_RSRC3_SHARED_VGPRS_SHIFT = 0;
+constexpr uint32_t COMPUTE_PGM_RSRC3_SHARED_VGPRS_MASK  = 0xF;
+constexpr uint32_t COMPUTE_PACE_ID             = 0x22A;
 constexpr uint32_t COMPUTE_USER_DATA_0        = 0x240;
 constexpr uint32_t COMPUTE_USER_DATA_15       = 0x24F;
-constexpr uint32_t COMPUTE_DISPATCH_TUNNEL    = 0x27D;
 
-/* Fake codes. Don't exist on real HW */
-
-constexpr uint32_t SH_NOP = 0x800002FF;
+constexpr uint32_t SH_NOP = 0x280;
 
 constexpr uint32_t SH_NUM = 0x2FF + 1;
 
@@ -1042,6 +965,7 @@ constexpr uint32_t GE_INDX_OFFSET            = 0x24A;
 constexpr uint32_t GE_MULTI_PRIM_IB_RESET_EN = 0x24B;
 constexpr uint32_t VGT_HS_OFFCHIP_PARAM      = 0x24F;
 constexpr uint32_t VGT_TF_MEMORY_BASE        = 0x250;
+constexpr uint32_t IA_MULTI_VGT_PARAM        = 0x258;
 
 constexpr uint32_t GE_CNTL                     = 0x25B;
 constexpr uint32_t GE_CNTL_PRIM_GRP_SIZE_SHIFT = 0;
@@ -1053,7 +977,7 @@ constexpr uint32_t GE_USER_VGPR1  = 0x25C;
 constexpr uint32_t GE_USER_VGPR2  = 0x25D;
 constexpr uint32_t GE_USER_VGPR3  = 0x25E;
 constexpr uint32_t GE_STEREO_CNTL = 0x25F;
-constexpr uint32_t GE_PC_ALLOC    = 0x260;
+constexpr uint32_t UC_PARAMETER_OVERSUBSCRIPTION = 0x260;
 
 constexpr uint32_t GE_USER_VGPR_EN                     = 0x262;
 constexpr uint32_t GE_USER_VGPR_EN_EN_USER_VGPR1_SHIFT = 0;
@@ -1063,28 +987,15 @@ constexpr uint32_t GE_USER_VGPR_EN_EN_USER_VGPR2_MASK  = 0x1;
 constexpr uint32_t GE_USER_VGPR_EN_EN_USER_VGPR3_SHIFT = 2;
 constexpr uint32_t GE_USER_VGPR_EN_EN_USER_VGPR3_MASK  = 0x1;
 
-constexpr uint32_t IA_MULTI_VGT_PARAM       = 0x258;
 constexpr uint32_t TA_CS_BC_BASE_ADDR       = 0x380;
 constexpr uint32_t TA_CS_BC_BASE_ADDR_HI    = 0x381;
-constexpr uint32_t TEXTURE_GRADIENT_FACTORS = 0x382;
+constexpr uint32_t TEXTURE_GRADIENT_FACTORS = 0x383;
+constexpr uint32_t TEXTURE_GRADIENT_CONTROL = 0x384;
 constexpr uint32_t GDS_OA_CNTL              = 0x41D;
 constexpr uint32_t GDS_OA_COUNTER           = 0x41E;
 constexpr uint32_t GDS_OA_ADDRESS           = 0x41F;
 
-/* Fake codes. Don't exist on real HW */
-
-constexpr uint32_t FSR_EXTEND_SUBPIXEL_ROUNDING = 0x80003FF4;
-constexpr uint32_t FSR_ALPHA_VALUE0             = 0x80003FF5;
-constexpr uint32_t FSR_ALPHA_VALUE1             = 0x80003FF6;
-constexpr uint32_t FSR_CONTROL_POINT0           = 0x80003FF7;
-constexpr uint32_t FSR_CONTROL_POINT1           = 0x80003FF8;
-constexpr uint32_t FSR_CONTROL_POINT2           = 0x80003FF9;
-constexpr uint32_t FSR_CONTROL_POINT3           = 0x80003FFA;
-constexpr uint32_t FSR_WINDOW0                  = 0x80003FFB;
-constexpr uint32_t FSR_WINDOW1                  = 0x80003FFC;
-constexpr uint32_t TEXTURE_GRADIENT_CONTROL     = 0x80003FFD;
-constexpr uint32_t MEMORY_MAPPING_MASK          = 0x80003FFE;
-constexpr uint32_t UC_NOP                       = 0x80003FFF;
+constexpr uint32_t UC_NOP = 0xA2;
 
 constexpr uint32_t UC_NUM = 0x3FFF + 1;
 

src/graphics/host_gpu/memoryTracker.cpp

@@ -18,8 +18,8 @@ MemoryTracker::~MemoryTracker() = default;
 #if KYTY_BUILD == KYTY_BUILD_DEBUG
 void MemoryTracker::ValidateGpuDirtyPages(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
                                           const char* operation) const noexcept {
-	if (vaddr == 0 || size == 0 || size > UINT64_MAX - vaddr ||
-	    (vaddr & (TRACKER_PAGE_SIZE - 1)) != 0 || (size & (TRACKER_PAGE_SIZE - 1)) != 0) {
+	if (!GuestRange {vaddr, size}.Valid() || (vaddr & (TRACKER_PAGE_SIZE - 1)) != 0 ||
+	    (size & (TRACKER_PAGE_SIZE - 1)) != 0) {
 		EXIT("MemoryTracker: invalid dirty-page validation range\n");
 	}
 	for (auto page = vaddr; page < vaddr + size; page += TRACKER_PAGE_SIZE) {
@@ -37,9 +37,6 @@ void MemoryTracker::ValidateGpuDirtyPages(const RangeSet& dirty, uint64_t vaddr,
 void MemoryTracker::ValidateGpuDirtyOwnership(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
                                               const char* operation) {
 	ValidateRange(vaddr, size);
-	if (vaddr + size > UINT64_MAX - (TRACKER_PAGE_SIZE - 1)) {
-		EXIT("MemoryTracker: dirty ownership range alignment overflow\n");
-	}
 	const auto begin = vaddr & ~(TRACKER_PAGE_SIZE - 1);
 	const auto end   = (vaddr + size + TRACKER_PAGE_SIZE - 1) & ~(TRACKER_PAGE_SIZE - 1);
 	for (auto page = begin; page < end; page += TRACKER_PAGE_SIZE) {
@@ -56,8 +53,7 @@ void MemoryTracker::ValidateGpuDirtyOwnership(const RangeSet& dirty, uint64_t va
 #endif
 
 void MemoryTracker::ValidateRange(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("invalid memory tracker range\n");
 	}
 }

src/graphics/host_gpu/pageManager.cpp

@@ -110,7 +110,7 @@ private:
 };
 
 void ValidateRange(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= ADDRESS_SIZE || size > ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		Fatal("invalid range vaddr=0x%016" PRIx64 ", size=0x%016" PRIx64, vaddr, size);
 	}
 }

src/graphics/host_gpu/regionDefinitions.h

@@ -4,6 +4,8 @@
 #include "common/bitArray.h"
 #include "common/common.h"
 
+#include <compare>
+
 namespace Libs::Graphics {
 
 constexpr uint64_t TRACKER_PAGE_SIZE    = 4ull * 1024ull;
@@ -11,6 +13,20 @@ constexpr uint64_t TRACKER_REGION_SIZE  = 4ull * 1024ull * 1024ull;
 constexpr uint64_t TRACKER_ADDRESS_SIZE = 1ull << 40u;
 constexpr size_t   TRACKER_REGION_PAGES = TRACKER_REGION_SIZE / TRACKER_PAGE_SIZE;
 
+struct GuestRange {
+	uint64_t address = 0;
+	uint64_t size    = 0;
+
+	[[nodiscard]] constexpr bool Empty() const noexcept { return address == 0 && size == 0; }
+	[[nodiscard]] constexpr bool Valid() const noexcept {
+		return address != 0 && size != 0 && address < TRACKER_ADDRESS_SIZE &&
+		       size <= TRACKER_ADDRESS_SIZE - address;
+	}
+	[[nodiscard]] constexpr bool     ValidOrEmpty() const noexcept { return Empty() || Valid(); }
+	[[nodiscard]] constexpr uint64_t End() const noexcept { return address + size; }
+	auto                             operator<=>(const GuestRange&) const = default;
+};
+
 enum class DirtySource { Cpu, Gpu };
 using RegionBits = Common::BitArray<TRACKER_REGION_PAGES>;
 static_assert(sizeof(RegionBits) == TRACKER_REGION_PAGES / 8);

src/graphics/host_gpu/renderer/cache/bufferCache.cpp

@@ -241,8 +241,7 @@ BufferCache::~BufferCache() {
 }
 
 void BufferCache::InvalidateMemory(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: invalid memory-invalidation range\n");
 	}
 	m_memory_tracker.InvalidateRegion(vaddr, size,
@@ -314,7 +313,7 @@ BufferId BufferCache::FindBuffer(uint64_t vaddr, uint64_t size) {
 	if (vaddr == 0) {
 		return NULL_BUFFER_ID;
 	}
-	if (size == 0 || vaddr >= TRACKER_ADDRESS_SIZE || size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: invalid buffer discovery request\n");
 	}
 	const auto* owner = m_page_table.Find(vaddr >> PageTable::kPageBits);
@@ -443,8 +442,7 @@ std::pair<Buffer*, uint64_t> BufferCache::ObtainBuffer(uint64_t vaddr, uint64_t
                                                        bool is_written, bool is_texel_buffer,
                                                        BufferId id) {
 	auto& command = m_scheduler.Current();
-	if (command.IsInvalid() || vaddr == 0 || size == 0 ||
-	    vaddr >= TRACKER_ADDRESS_SIZE || size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (command.IsInvalid() || !GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: buffer request requires a recording command buffer\n");
 	}
 
@@ -474,8 +472,7 @@ std::pair<Buffer*, uint64_t> BufferCache::ObtainBuffer(uint64_t vaddr, uint64_t
 }
 
 std::pair<Buffer*, uint64_t> BufferCache::ObtainBufferForImage(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: invalid image source\n");
 	}
 	auto find_owner = [&]() -> Buffer* {
@@ -661,8 +658,7 @@ void BufferCache::CopyBuffer(uint64_t dst_vaddr, uint64_t src_vaddr, uint64_t si
 }
 
 bool BufferCache::IsRegionRegistered(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: invalid registered-region query\n");
 	}
 	// Cached buffers are ordered and non-overlapping. The last buffer beginning before the query

src/graphics/host_gpu/renderer/cache/gpuResourceManager.cpp

@@ -35,8 +35,7 @@ bool GpuResourceManager::InvalidateMemory(uint64_t vaddr, uint64_t size) {
 }
 
 bool GpuResourceManager::IsMapped(uint64_t vaddr, uint64_t size) const noexcept {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		return false;
 	}
 	std::shared_lock lock(m_mapped_ranges_mutex);

src/graphics/host_gpu/renderer/colorRenderTarget.cpp

@@ -28,12 +28,27 @@ static void ResolveDccClearInfo(RenderColorInfo& info, vk::Format format, bool h
 	info.metadata_clear_supported =
 	    has_dcc && DecodePackedColorClear(format, packed_clear, info.color_clear_value);
 	switch (format) {
+		case vk::Format::eR8Unorm:
+		case vk::Format::eR8G8Unorm:
 		case vk::Format::eR8G8B8A8Unorm:
 		case vk::Format::eR8G8B8A8Srgb:
 		case vk::Format::eB8G8R8A8Unorm:
 		case vk::Format::eB8G8R8A8Srgb:
 		case vk::Format::eA2B10G10R10UnormPack32:
 		case vk::Format::eA2R10G10B10UnormPack32:
+		case vk::Format::eR5G6B5UnormPack16:
+		case vk::Format::eA1R5G5B5UnormPack16:
+		case vk::Format::eR4G4B4A4UnormPack16:
+		case vk::Format::eR16Unorm:
+		case vk::Format::eR16G16Unorm:
+		case vk::Format::eR16G16B16A16Unorm:
+		case vk::Format::eR16Sfloat:
+		case vk::Format::eR16G16Sfloat:
+		case vk::Format::eR16G16B16A16Sfloat:
+		case vk::Format::eR32Sfloat:
+		case vk::Format::eR32G32Sfloat:
+		case vk::Format::eR32G32B32A32Sfloat:
+		case vk::Format::eB10G11R11UfloatPack32:
 			info.metadata_fixed_clear_supported = has_dcc;
 			break;
 		default: info.metadata_fixed_clear_supported = false; break;
@@ -204,13 +219,11 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 	if (texture_tile &&
 	    (!TileGetTextureBlockLayout(transfer_format, rt.attrib3.tile_mode, volume,
 	                                texture_tile_layout) ||
-	     rt.pitch.pitch_div8_minus1 != 0 ||
 	     (rt.base.addr & (texture_tile_layout.block.block_size - 1u)) != 0 ||
 	     rt.info.fmask_compression_enable || rt.info.fmask_data_compression_disable ||
 	     rt.info.fmask_one_frag_mode || rt.info.cmask_fast_clear_enable ||
-	     rt.info.dcc_compression_enable || rt.info.cmask_is_linear != 0 ||
-	     rt.info.cmask_addr_type != 0 || rt.info.alt_tile_mode || rt.cmask.addr != 0 ||
-	     rt.fmask.addr != 0 || rt.dcc_addr.addr != 0 || rt.dcc.data_write_on_dcc_clear_to_reg)) {
+	     rt.info.dcc_compression_enable || rt.cmask.addr != 0 || rt.fmask.addr != 0 ||
+	     rt.dcc_addr.addr != 0 || rt.dcc.data_write_on_dcc_clear_to_reg)) {
 		EXIT("unsupported texture-tiled render target: addr=0x%016" PRIx64 " tile=%u"
 		     " dimension=%u depth=%u levels=%u layer=%u/%u samples=%u fragments=%u bpe=%u"
 		     " cmask=0x%016" PRIx64 " fmask=0x%016" PRIx64 " dcc=0x%016" PRIx64 "\n",
@@ -219,15 +232,14 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 		     rt.attrib.num_fragments, bytes_per_element, rt.cmask.addr, rt.fmask.addr,
 		     rt.dcc_addr.addr);
 	}
-	if ((standard64 || depth_tile) && (rt.attrib3.dimension != 1 || rt.attrib3.depth != 0 ||
-	                                   view.base_layer != 0 || view.image_layers != 1)) {
+	if ((standard64 || depth_tile) &&
+	    (rt.attrib3.dimension != 1 || rt.attrib3.depth != 0 ||
+	     (depth_tile && (view.base_layer != 0 || view.image_layers != 1)))) {
 		EXIT("unsupported 64KB texture-tiled render-target view: dimension=%u depth=%u"
 		     " layer=%u/%u\n",
 		     rt.attrib3.dimension, rt.attrib3.depth, view.base_layer, view.image_layers);
 	}
-	if (rt.pitch.pitch_div8_minus1 != 0) {
-		pitch = (rt.pitch.pitch_div8_minus1 + 1u) << 3u;
-	} else if (tile) {
+	if (tile) {
 		if (volume) {
 			pitch = TileGetTexturePitch(transfer_format, width, rt.attrib3.tile_mode);
 		} else if (texture_tile) {
@@ -293,12 +305,6 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 		mip_sizes[0]  = {static_cast<uint32_t>(size), 0, 0, 0, 0, 0};
 		mip_padded[0] = {pitch, height};
 	}
-	if (rt.slice.slice_div64_minus1 != 0 &&
-	    (static_cast<uint64_t>(rt.slice.slice_div64_minus1) + 1u) * 64u != size) {
-		EXIT("render-target slice span mismatch: encoded=0x%016" PRIx64 " derived=0x%016" PRIx64
-		     "\n",
-		     (static_cast<uint64_t>(rt.slice.slice_div64_minus1) + 1u) * 64u, size);
-	}
 	if (size == 0 || (!volume && size > UINT64_MAX / view.image_layers)) {
 		EXIT("render-target memory footprint is invalid\n");
 	}
@@ -308,7 +314,7 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 	if (backing_size == 0) {
 		EXIT("render-target backing is empty\n");
 	}
-	if (backing_size > TRACKER_ADDRESS_SIZE - rt.base.addr) {
+	if (!GuestRange {rt.base.addr, backing_size}.Valid()) {
 		EXIT("render-target backing range is invalid\n");
 	}
 

src/graphics/host_gpu/renderer/debug.cpp

@@ -84,17 +84,8 @@ void uc_print(const char* func, const HW::UserConfig& uc) {
 }
 
 void uc_check(const HW::UserConfig& uc) {
-	const auto& ge_cntl = uc.GetGeControl();
 	const auto& user_en = uc.GetGeUserVgprEn();
 
-	if (ge_cntl.primitive_group_size > 0x0040) {
-		LOGF("\t warning: unsupported GE_CNTL primitive_group_size = 0x%04" PRIx16 "\n",
-		     ge_cntl.primitive_group_size);
-	}
-	if (ge_cntl.vertex_group_size > 0x0040) {
-		LOGF("\t warning: unsupported GE_CNTL vertex_group_size = 0x%04" PRIx16 "\n",
-		     ge_cntl.vertex_group_size);
-	}
 	EXIT_NOT_IMPLEMENTED(user_en.vgpr1 != false);
 	EXIT_NOT_IMPLEMENTED(user_en.vgpr2 != false);
 	EXIT_NOT_IMPLEMENTED(user_en.vgpr3 != false);
@@ -106,17 +97,11 @@ void sh_print(const char* func, const HW::Shader& /*uc*/) {
 
 std::vector<std::string> rt_print(const char* func, const HW::RenderTarget& rt) {
 	std::vector<std::string> dst;
-	dst.reserve(53);
+	dst.reserve(40);
 
 	dst.push_back(fmt::format("{}\n", func));
 
 	dst.push_back(fmt::format("\t base.addr                       = 0x{:016x}\n", rt.base.addr));
-	dst.push_back(
-	    fmt::format("\t pitch.pitch_div8_minus1         = 0x{:08x}\n", rt.pitch.pitch_div8_minus1));
-	dst.push_back(fmt::format("\t pitch.fmask_pitch_div8_minus1   = 0x{:08x}\n",
-	                          rt.pitch.fmask_pitch_div8_minus1));
-	dst.push_back(fmt::format("\t slice.slice_div64_minus1        = 0x{:08x}\n",
-	                          rt.slice.slice_div64_minus1));
 	dst.push_back(fmt::format("\t view.base_array_slice_index     = 0x{:08x}\n",
 	                          rt.view.base_array_slice_index));
 	dst.push_back(fmt::format("\t view.last_array_slice_index     = 0x{:08x}\n",
@@ -137,12 +122,6 @@ std::vector<std::string> rt_print(const char* func, const HW::RenderTarget& rt)
 	                          rt.info.cmask_fast_clear_enable ? "true" : "false"));
 	dst.push_back(fmt::format("\t info.dcc_compression_enable     = {}\n",
 	                          rt.info.dcc_compression_enable ? "true" : "false"));
-	dst.push_back(
-	    fmt::format("\t info.cmask_is_linear            = 0x{:08x}\n", rt.info.cmask_is_linear));
-	dst.push_back(
-	    fmt::format("\t info.cmask_addr_type            = 0x{:08x}\n", rt.info.cmask_addr_type));
-	dst.push_back(fmt::format("\t info.alt_tile_mode              = {}\n",
-	                          rt.info.alt_tile_mode ? "true" : "false"));
 	dst.push_back(fmt::format("\t info.format                     = 0x{:08x}\n",
 	                          static_cast<uint32_t>(rt.info.format)));
 	dst.push_back(fmt::format("\t info.channel_type               = 0x{:08x}\n",
@@ -157,10 +136,6 @@ std::vector<std::string> rt_print(const char* func, const HW::RenderTarget& rt)
 	                          rt.info.round_mode ? "true" : "false"));
 	dst.push_back(fmt::format("\t attrib.force_dest_alpha_to_one  = {}\n",
 	                          rt.attrib.force_dest_alpha_to_one ? "true" : "false"));
-	dst.push_back(fmt::format("\t attrib.tile_mode                = 0x{:08x}\n",
-	                          static_cast<uint32_t>(rt.attrib.tile_mode)));
-	dst.push_back(fmt::format("\t attrib.fmask_tile_mode          = 0x{:08x}\n",
-	                          static_cast<uint32_t>(rt.attrib.fmask_tile_mode)));
 	dst.push_back(
 	    fmt::format("\t attrib.num_samples              = 0x{:08x}\n", rt.attrib.num_samples));
 	dst.push_back(
@@ -175,42 +150,32 @@ std::vector<std::string> rt_print(const char* func, const HW::RenderTarget& rt)
 	                          static_cast<uint32_t>(rt.attrib3.tile_mode)));
 	dst.push_back(
 	    fmt::format("\t attrib3.dimension               = 0x{:08x}\n", rt.attrib3.dimension));
-	dst.push_back(fmt::format("\t attrib3.cmask_pipe_aligned      = {}\n",
-	                          rt.attrib3.cmask_pipe_aligned ? "true" : "false"));
-	dst.push_back(fmt::format("\t attrib3.dcc_pipe_aligned        = {}\n",
-	                          rt.attrib3.dcc_pipe_aligned ? "true" : "false"));
+	dst.push_back(fmt::format("\t attrib3.metadata_pipe_aligned   = {}\n",
+	                          rt.attrib3.metadata_pipe_aligned ? "true" : "false"));
+	dst.push_back(fmt::format("\t attrib3.write_vrs_rate_hint_to_cmask = {}\n",
+	                          rt.attrib3.write_vrs_rate_hint_to_cmask ? "true" : "false"));
 	dst.push_back(fmt::format("\t dcc.max_uncompressed_block_size = 0x{:08x}\n",
 	                          rt.dcc.max_uncompressed_block_size));
 	dst.push_back(fmt::format("\t dcc.max_compressed_block_size   = 0x{:08x}\n",
 	                          rt.dcc.max_compressed_block_size));
-	dst.push_back(fmt::format("\t dcc.min_compressed_block_size   = 0x{:08x}\n",
-	                          rt.dcc.min_compressed_block_size));
 	dst.push_back(
 	    fmt::format("\t dcc.color_transform             = 0x{:08x}\n", rt.dcc.color_transform));
 	dst.push_back(fmt::format("\t dcc.overwrite_combiner_disable  = {}\n",
 	                          rt.dcc.overwrite_combiner_disable ? "true" : "false"));
-	dst.push_back(fmt::format("\t dcc.independent_64b_blocks      = {}\n",
-	                          rt.dcc.independent_64b_blocks ? "true" : "false"));
-	dst.push_back(fmt::format("\t dcc.independent_128b_blocks     = {}\n",
-	                          rt.dcc.independent_128b_blocks ? "true" : "false"));
+	dst.push_back(fmt::format("\t dcc.independent_block_size      = 0x{:02x}\n",
+	                          static_cast<uint8_t>(rt.dcc.independent_block_size)));
 	dst.push_back(fmt::format("\t data_write_on_dcc_clear_to_reg  = {}\n",
 	                          rt.dcc.data_write_on_dcc_clear_to_reg ? "true" : "false"));
 	dst.push_back(fmt::format("\t dcc.dcc_clear_key_enable        = {}\n",
 	                          rt.dcc.dcc_clear_key_enable ? "true" : "false"));
 	dst.push_back(fmt::format("\t cmask.addr                      = 0x{:016x}\n", rt.cmask.addr));
-	dst.push_back(fmt::format("\t cmask_slice.slice_minus1        = 0x{:08x}\n",
-	                          rt.cmask_slice.slice_minus1));
 	dst.push_back(fmt::format("\t fmask.addr                      = 0x{:016x}\n", rt.fmask.addr));
-	dst.push_back(fmt::format("\t fmask_slice.slice_minus1        = 0x{:08x}\n",
-	                          rt.fmask_slice.slice_minus1));
 	dst.push_back(
 	    fmt::format("\t clear_word0.word0               = 0x{:08x}\n", rt.clear_word0.word0));
 	dst.push_back(
 	    fmt::format("\t clear_word1.word1               = 0x{:08x}\n", rt.clear_word1.word1));
 	dst.push_back(
 	    fmt::format("\t dcc_addr.addr                   = 0x{:016x}\n", rt.dcc_addr.addr));
-	dst.push_back(fmt::format("\t size.width                      = 0x{:08x}\n", rt.size.width));
-	dst.push_back(fmt::format("\t size.height                     = 0x{:08x}\n", rt.size.height));
 
 	return dst;
 }
@@ -240,13 +205,8 @@ static bool RenderIsColorDimension(uint32_t dimension) {
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
 static void RtCheck(const HW::RenderTarget& rt) {
 	if (rt.base.addr != 0) {
-		// bool render_to_texture = (rt.attrib.tile_mode == 0x0d);
 		//  EXIT_NOT_IMPLEMENTED(rt.base_addr == 0);
 
-		EXIT_NOT_IMPLEMENTED(rt.pitch.pitch_div8_minus1 != 0);
-		EXIT_NOT_IMPLEMENTED(rt.pitch.fmask_pitch_div8_minus1 != 0);
-		EXIT_NOT_IMPLEMENTED(rt.slice.slice_div64_minus1 != 0);
-
 		EXIT_NOT_IMPLEMENTED(rt.view.base_array_slice_index > rt.view.last_array_slice_index);
 		if (rt.view.base_array_slice_index != 0x00000000 ||
 		    rt.view.last_array_slice_index != 0x00000000) {
@@ -266,11 +226,6 @@ static void RtCheck(const HW::RenderTarget& rt) {
 				logged = true;
 			}
 		}
-		const bool native_2x_one_frag =
-		    rt.info.fmask_compression_enable && !rt.info.fmask_data_compression_disable &&
-		    rt.info.cmask_fast_clear_enable && rt.fmask.addr != 0 && rt.cmask.addr != 0 &&
-		    rt.attrib.num_samples == 1 && rt.attrib.num_fragments == 1;
-		EXIT_NOT_IMPLEMENTED(rt.info.fmask_one_frag_mode && !native_2x_one_frag);
 		if (rt.info.fmask_compression_enable) {
 			EXIT_NOT_IMPLEMENTED(rt.attrib.num_samples == 0 && rt.attrib.num_fragments == 0);
 			static bool logged = false;
@@ -295,22 +250,6 @@ static void RtCheck(const HW::RenderTarget& rt) {
 				logged = true;
 			}
 		}
-		if (rt.info.cmask_is_linear != 0x00000000 || rt.info.cmask_addr_type != 0x00000000) {
-			static bool logged = false;
-			if (!logged) {
-				LOGF("RenderTarget: temporary: ignoring PS5 CMASK metadata fields "
-				     "is_linear=%u addr_type=%u\n",
-				     rt.info.cmask_is_linear, rt.info.cmask_addr_type);
-				logged = true;
-			}
-		}
-		if (rt.info.alt_tile_mode) {
-			static bool logged = false;
-			if (!logged) {
-				LOGF("RenderTarget: temporary: ignoring alternate tile mode flag\n");
-				logged = true;
-			}
-		}
 		if (rt.info.blend_bypass) {
 			static bool logged = false;
 			if (!logged) {
@@ -337,8 +276,6 @@ static void RtCheck(const HW::RenderTarget& rt) {
 				logged = true;
 			}
 		}
-		// EXIT_NOT_IMPLEMENTED(rt.tile_mode != 0x0000000a);
-		// EXIT_NOT_IMPLEMENTED(rt.fmask_tile_mode != 0x0000000a);
 		if (rt.attrib.num_samples != 0x00000000 || rt.attrib.num_fragments != 0x00000000) {
 			static bool logged = false;
 			if (!logged) {
@@ -373,34 +310,22 @@ static void RtCheck(const HW::RenderTarget& rt) {
 		if (!RenderIsColorDimension(rt.attrib3.dimension)) {
 			EXIT("unknown PS5 render-target dimension: 0x%08" PRIx32 "\n", rt.attrib3.dimension);
 		}
-		if (!rt.attrib3.cmask_pipe_aligned) {
+		if (!rt.attrib3.metadata_pipe_aligned) {
 			static bool logged = false;
 			if (!logged) {
-				LOGF("RenderTarget: temporary: accepting unaligned PS5 CMASK pipe flag\n");
-				logged = true;
-			}
-		}
-		if (!rt.attrib3.dcc_pipe_aligned) {
-			static bool logged = false;
-			if (!logged) {
-				LOGF("RenderTarget: temporary: accepting unaligned PS5 DCC pipe flag\n");
+				LOGF("RenderTarget: temporary: accepting unaligned PS5 metadata pipe flag\n");
 				logged = true;
 			}
 		}
+		EXIT_NOT_IMPLEMENTED(rt.attrib3.write_vrs_rate_hint_to_cmask);
 
 		// EXIT_NOT_IMPLEMENTED(rt.dcc_max_uncompressed_block_size != 0x00000002);
 		// EXIT_NOT_IMPLEMENTED(rt.dcc.max_compressed_block_size != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(rt.dcc.min_compressed_block_size != 0x00000000);
 		// EXIT_NOT_IMPLEMENTED(rt.dcc.color_transform != 0x00000000);
 		EXIT_NOT_IMPLEMENTED(rt.dcc.overwrite_combiner_disable != false);
-		// EXIT_NOT_IMPLEMENTED(rt.dcc.force_independent_blocks != false);
-		// EXIT_NOT_IMPLEMENTED(rt.dcc.independent_128b_blocks != false);
 		// EXIT_NOT_IMPLEMENTED(rt.dcc.data_write_on_dcc_clear_to_reg != false);
 		EXIT_NOT_IMPLEMENTED(rt.dcc.dcc_clear_key_enable != false);
-		if (rt.cmask.addr != 0x0000000000000000 || rt.cmask_slice.slice_minus1 != 0x00000000 ||
-		    rt.fmask.addr != 0x0000000000000000 ||
-		    (rt.fmask_slice.slice_minus1 != 0x00000000 &&
-		     rt.fmask_slice.slice_minus1 != rt.slice.slice_div64_minus1) ||
+		if (rt.cmask.addr != 0x0000000000000000 || rt.fmask.addr != 0x0000000000000000 ||
 		    rt.dcc_addr.addr != 0x0000000000000000) {
 			static bool logged = false;
 			if (!logged) {
@@ -410,9 +335,6 @@ static void RtCheck(const HW::RenderTarget& rt) {
 				logged = true;
 			}
 		}
-
-		EXIT_NOT_IMPLEMENTED(rt.size.width != 0);
-		EXIT_NOT_IMPLEMENTED(rt.size.height != 0);
 	}
 }
 
@@ -432,37 +354,20 @@ static void ZPrint(const char* func, const HW::DepthRenderTarget& z) {
 	     "\t stencil_info.htile_stencil_disabled   = %s\n"
 	     "\t stencil_info.expclear_enabled         = %s\n"
 	     "\t stencil_info.partially_resident       = %s\n"
-	     "\t depth_info.addr5_swizzle_mask         = 0x%08" PRIx32 "\n"
-	     "\t depth_info.array_mode                 = 0x%08" PRIx32 "\n"
-	     "\t depth_info.pipe_config                = 0x%08" PRIx32 "\n"
-	     "\t depth_info.bank_width                 = 0x%08" PRIx32 "\n"
-	     "\t depth_info.bank_height                = 0x%08" PRIx32 "\n"
-	     "\t depth_info.macro_tile_aspect          = 0x%08" PRIx32 "\n"
-	     "\t depth_info.num_banks                  = 0x%08" PRIx32 "\n"
 	     "\t depth_view.slice_start                = 0x%08" PRIx32 "\n"
 	     "\t depth_view.slice_max                  = 0x%08" PRIx32 "\n"
 	     "\t depth_view.current_mip_level          = 0x%02" PRIx8 "\n"
 	     "\t depth_view.depth_write_disable        = %s\n"
 	     "\t depth_view.stencil_write_disable      = %s\n"
-	     "\t htile_surface.linear                  = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.full_cache              = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.htile_uses_preload_win  = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.preload                 = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.prefetch_width          = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.prefetch_height         = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.dst_outside_zero_to_one = 0x%08" PRIx32 "\n"
 	     "\t z_read_base_addr                      = 0x%016" PRIx64 "\n"
 	     "\t stencil_read_base_addr                = 0x%016" PRIx64 "\n"
 	     "\t z_write_base_addr                     = 0x%016" PRIx64 "\n"
 	     "\t stencil_write_base_addr               = 0x%016" PRIx64 "\n"
-	     "\t pitch_div8_minus1                     = 0x%08" PRIx32 "\n"
-	     "\t height_div8_minus1                    = 0x%08" PRIx32 "\n"
-	     "\t slice_div64_minus1                    = 0x%08" PRIx32 "\n"
 	     "\t htile_data_base_addr                  = 0x%016" PRIx64 "\n"
-	     "\t width                                 = 0x%08" PRIx32 "\n"
-	     "\t height                                = 0x%08" PRIx32 "\n"
+	     "\t shading_rate_encoding                 = 0x%02" PRIx8 "\n"
 	     "\t size.x_max                            = 0x%04" PRIx16 "\n"
-	     "\t size.y_max                            = 0x%04" PRIx16 "\n",
+	     "\t size.y_max                            = 0x%04" PRIx16 "\n"
+	     "\t size.valid                            = %s\n",
 	     static_cast<uint32_t>(z.z_info.format), z.z_info.num_samples,
 	     static_cast<uint32_t>(z.z_info.texture_compatibility),
 	     z.z_info.htile_acceleration ? "true" : "false",
@@ -473,30 +378,27 @@ static void ZPrint(const char* func, const HW::DepthRenderTarget& z) {
 	     static_cast<uint32_t>(z.stencil_info.texture_compatibility),
 	     z.stencil_info.htile_stencil_disabled ? "true" : "false",
 	     z.stencil_info.expclear_enabled ? "true" : "false",
-	     z.stencil_info.partially_resident ? "true" : "false", z.depth_info.addr5_swizzle_mask,
-	     z.depth_info.array_mode, z.depth_info.pipe_config, z.depth_info.bank_width,
-	     z.depth_info.bank_height, z.depth_info.macro_tile_aspect, z.depth_info.num_banks,
-	     z.depth_view.slice_start, z.depth_view.slice_max, z.depth_view.current_mip_level,
+	     z.stencil_info.partially_resident ? "true" : "false", z.depth_view.slice_start,
+	     z.depth_view.slice_max, z.depth_view.current_mip_level,
 	     z.depth_view.depth_write_disable ? "true" : "false",
-	     z.depth_view.stencil_write_disable ? "true" : "false", z.htile_surface.linear,
-	     z.htile_surface.full_cache, z.htile_surface.htile_uses_preload_win,
-	     z.htile_surface.preload, z.htile_surface.prefetch_width, z.htile_surface.prefetch_height,
-	     z.htile_surface.dst_outside_zero_to_one, z.z_read_base_addr, z.stencil_read_base_addr,
-	     z.z_write_base_addr, z.stencil_write_base_addr, z.pitch_div8_minus1, z.height_div8_minus1,
-	     z.slice_div64_minus1, z.htile_data_base_addr, z.width, z.height, z.size.x_max,
-	     z.size.y_max);
+	     z.depth_view.stencil_write_disable ? "true" : "false", z.z_read_base_addr,
+	     z.stencil_read_base_addr, z.z_write_base_addr, z.stencil_write_base_addr,
+	     z.htile_data_base_addr, z.shading_rate_encoding, z.size.x_max, z.size.y_max,
+	     z.size.valid ? "true" : "false");
 }
 
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
 static void ZCheck(const HW::DepthRenderTarget& z, const HW::DepthControl& dc,
                    const HW::RenderControl& rc) {
-	const bool depth_active =
-	    dc.z_enable || dc.z_write_enable || dc.depth_bounds_enable || rc.depth_clear_enable;
-	const bool stencil_active = dc.stencil_enable || rc.stencil_clear_enable;
+	const bool depth_active = dc.z_enable || dc.z_write_enable || dc.depth_bounds_enable ||
+	                          rc.depth_clear_enable || rc.copy_depth_to_color;
+	const bool stencil_active =
+	    dc.stencil_enable || rc.stencil_clear_enable || rc.copy_stencil_to_color;
 	if (!depth_active && !stencil_active) {
 		return;
 	}
 
+	EXIT_NOT_IMPLEMENTED(rc.copy_depth_to_color || rc.copy_stencil_to_color);
 	EXIT_NOT_IMPLEMENTED(!z.z_info.HasValidTextureCompatibility());
 	EXIT_NOT_IMPLEMENTED(!z.stencil_info.HasValidTextureCompatibility());
 	if (z.z_info.format == Prospero::DepthFormat::kInvalid) {
@@ -533,22 +435,6 @@ static void ZCheck(const HW::DepthRenderTarget& z, const HW::DepthControl& dc,
 
 	if (z.z_info.format != Prospero::DepthFormat::kInvalid ||
 	    z.stencil_info.format != Prospero::StencilFormat::kInvalid) {
-
-		EXIT_NOT_IMPLEMENTED(z.depth_info.addr5_swizzle_mask != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.array_mode != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.pipe_config != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.bank_width != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.bank_height != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.macro_tile_aspect != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.num_banks != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.linear != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.full_cache != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.htile_uses_preload_win != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.preload != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.prefetch_width != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.prefetch_height != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.dst_outside_zero_to_one != 0x00000000);
-
 		if (z.depth_view.current_mip_level != 0x00000000) {
 			static std::atomic<uint32_t> log_count {0};
 			if (log_count.fetch_add(1, std::memory_order_relaxed) < 16) {
@@ -569,23 +455,8 @@ static void ZCheck(const HW::DepthRenderTarget& z, const HW::DepthControl& dc,
 		EXIT_NOT_IMPLEMENTED(!z.depth_view.stencil_write_disable &&
 		                     z.stencil_read_base_addr != z.stencil_write_base_addr);
 		EXIT_NOT_IMPLEMENTED(!z.depth_view.depth_write_disable && z.z_write_base_addr == 0);
-		// EXIT_NOT_IMPLEMENTED(z.pitch_div8_minus1 != 0x000000ff);
-		// EXIT_NOT_IMPLEMENTED(z.height_div8_minus1 != 0x0000008f);
-		// EXIT_NOT_IMPLEMENTED(z.slice_div64_minus1 != 0x00008fff);
 		// EXIT_NOT_IMPLEMENTED(z.htile_data_base_addr == 0);
-		// EXIT_NOT_IMPLEMENTED(z.width != 0x00000780);
-		// EXIT_NOT_IMPLEMENTED(z.height != 0x00000438);
-
-		EXIT_NOT_IMPLEMENTED(z.width != 0);
-		EXIT_NOT_IMPLEMENTED(z.height != 0);
-		if (z.size.x_max == 0 || z.size.y_max == 0) {
-			static std::atomic<uint32_t> log_count {0};
-			if (log_count.fetch_add(1) < 16) {
-				LOGF("DepthTarget: temporary: inferring PS5 default depth extent x_max=0x%04" PRIx16
-				     ", y_max=0x%04" PRIx16 "\n",
-				     z.size.x_max, z.size.y_max);
-			}
-		}
+		EXIT_NOT_IMPLEMENTED(!z.size.valid);
 	}
 }
 
@@ -617,7 +488,7 @@ static void ClipCheck(const HW::ClipControl& c) {
 	// dx_linear_attr_clip_enable preserves linear (noperspective) attributes at clip-generated
 	// vertices, which Vulkan provides as part of clipping and interpolation.
 	EXIT_NOT_IMPLEMENTED(c.user_clip_planes != 0 || c.user_clip_plane_mode != 0 ||
-	                     c.vertex_kill_any || c.user_clip_plane_negate_y || c.clip_disable ||
+	                     c.vertex_kill_any || c.user_clip_plane_negate_y ||
 	                     c.user_clip_plane_cull_only || c.cull_on_clipping_error_disable ||
 	                     c.force_viewport_index_from_vs_enable);
 }
@@ -768,8 +639,6 @@ static void DPrint(const char* func, const HW::DepthControl& c, const HW::Stenci
 	     "\t backface_enable      = %s\n"
 	     "\t stencilfunc          = %" PRIu8 "\n"
 	     "\t stencilfunc_bf       = %" PRIu8 "\n"
-	     "\t color_writes_on_depth_fail_enable  = %s\n"
-	     "\t color_writes_on_depth_pass_disable = %s\n"
 	     "\t stencil_fail         = %" PRIu8 "\n"
 	     "\t stencil_zpass        = %" PRIu8 "\n"
 	     "\t stencil_zfail        = %" PRIu8 "\n"
@@ -786,12 +655,11 @@ static void DPrint(const char* func, const HW::DepthControl& c, const HW::Stenci
 	     "\t stencil_opval_bf     = %" PRIu8 "\n",
 	     c.stencil_enable ? "true" : "false", c.z_enable ? "true" : "false",
 	     c.z_write_enable ? "true" : "false", c.depth_bounds_enable ? "true" : "false", c.zfunc,
-	     c.backface_enable ? "true" : "false", c.stencilfunc, c.stencilfunc_bf,
-	     c.color_writes_on_depth_fail_enable ? "true" : "false",
-	     c.color_writes_on_depth_pass_disable ? "true" : "false", s.stencil_fail, s.stencil_zpass,
-	     s.stencil_zfail, s.stencil_fail_bf, s.stencil_zpass_bf, s.stencil_zfail_bf,
-	     sm.stencil_testval, sm.stencil_mask, sm.stencil_writemask, sm.stencil_opval,
-	     sm.stencil_testval_bf, sm.stencil_mask_bf, sm.stencil_writemask_bf, sm.stencil_opval_bf);
+	     c.backface_enable ? "true" : "false", c.stencilfunc, c.stencilfunc_bf, s.stencil_fail,
+	     s.stencil_zpass, s.stencil_zfail, s.stencil_fail_bf, s.stencil_zpass_bf,
+	     s.stencil_zfail_bf, sm.stencil_testval, sm.stencil_mask, sm.stencil_writemask,
+	     sm.stencil_opval, sm.stencil_testval_bf, sm.stencil_mask_bf, sm.stencil_writemask_bf,
+	     sm.stencil_opval_bf);
 }
 
 static void DCheck(const HW::DepthControl& c, const HW::StencilControl& s,
@@ -803,8 +671,6 @@ static void DCheck(const HW::DepthControl& c, const HW::StencilControl& s,
 	// Back-face stencil state is handled separately when enabled.
 	// EXIT_NOT_IMPLEMENTED(c.stencilfunc != 0);
 	// EXIT_NOT_IMPLEMENTED(c.stencilfunc_bf != 0);
-	EXIT_NOT_IMPLEMENTED(c.color_writes_on_depth_fail_enable != false);
-	EXIT_NOT_IMPLEMENTED(c.color_writes_on_depth_pass_disable != false);
 	// EXIT_NOT_IMPLEMENTED(s.stencil_fail != 0);
 	// EXIT_NOT_IMPLEMENTED(s.stencil_zpass != 0);
 	// EXIT_NOT_IMPLEMENTED(s.stencil_zfail != 0);
@@ -836,7 +702,12 @@ static void EqaaPrint(const char* func, const HW::EqaaControl& c) {
 	     c.static_anchor_associations ? "true" : "false");
 }
 
-static void EqaaCheck(const HW::EqaaControl& c) {
+static void EqaaCheck(const HW::EqaaControl& c, const HW::AaConfig& cf) {
+	// The EQAA controls only take effect while multisampling is on, so a single-sample
+	// target leaves them inert and has nothing to warn about.
+	if (cf.msaa_num_samples == 0) {
+		return;
+	}
 	if (c.max_anchor_samples != 0 || c.ps_iter_samples != 0 || c.mask_export_num_samples != 0 ||
 	    c.alpha_to_mask_num_samples != 0 || c.high_quality_intersections ||
 	    c.incoherent_eqaa_reads || c.interpolate_comp_z || c.static_anchor_associations) {
@@ -1201,12 +1072,12 @@ void hw_check(const CommandBuffer& buffer) {
 	log_phase("blend");
 	BcCheck(bc, bclr, cc);
 	log_phase("eqaa");
-	EqaaCheck(eqaa);
+	EqaaCheck(eqaa, ac);
 	log_phase("aa");
 	AaCheck(aa, ac);
 	log_phase("done");
 
-	if (RenderTargetMaskHasBoundMrt(buffer)) {
+	if (graphics_debug_dump_enabled() && RenderTargetMaskHasBoundMrt(buffer)) {
 		LOGF("MRT render target mask: 0x%08" PRIx32 "\n", hw.GetRenderTargetMask());
 		for (uint32_t i = 0; i < 8; i++) {
 			const auto& mrt = hw.GetRenderTarget(i);

src/graphics/host_gpu/renderer/depthRenderTarget.cpp

@@ -108,9 +108,10 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 	const auto& sc          = hw.GetStencilControl();
 	const auto& sm          = hw.GetStencilMask();
 	const bool  has_stencil = z.stencil_info.format != Prospero::StencilFormat::kInvalid;
-	const bool  depth_active =
-	    dc.z_enable || dc.z_write_enable || dc.depth_bounds_enable || rc.depth_clear_enable;
-	const bool stencil_active = has_stencil && (dc.stencil_enable || rc.stencil_clear_enable);
+	const bool depth_active = dc.z_enable || dc.z_write_enable || dc.depth_bounds_enable ||
+	                          rc.depth_clear_enable || rc.copy_depth_to_color;
+	const bool stencil_active =
+	    has_stencil && (dc.stencil_enable || rc.stencil_clear_enable || rc.copy_stencil_to_color);
 	if (!depth_active && !stencil_active) {
 		return;
 	}
@@ -127,22 +128,14 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 	    !z.stencil_info.expclear_enabled && !z.stencil_info.partially_resident &&
 	    z.depth_view.slice_start == 0 && z.depth_view.slice_max == 0 &&
 	    z.depth_view.current_mip_level == 0 && !z.depth_view.depth_write_disable &&
-	    !z.depth_view.stencil_write_disable && z.depth_info.addr5_swizzle_mask == 0 &&
-	    z.depth_info.array_mode == 0 && z.depth_info.pipe_config == 0 &&
-	    z.depth_info.bank_width == 0 && z.depth_info.bank_height == 0 &&
-	    z.depth_info.macro_tile_aspect == 0 && z.depth_info.num_banks == 0 &&
-	    z.htile_surface.linear == 0 && z.htile_surface.full_cache == 0 &&
-	    z.htile_surface.htile_uses_preload_win == 0 && z.htile_surface.preload == 0 &&
-	    z.htile_surface.prefetch_width == 0 && z.htile_surface.prefetch_height == 0 &&
-	    z.htile_surface.dst_outside_zero_to_one == 0 && z.z_read_base_addr == 0 &&
-	    z.z_write_base_addr == 0 && z.stencil_read_base_addr == 0 &&
-	    z.stencil_write_base_addr == 0 && z.htile_data_base_addr == 0 &&
+	    !z.depth_view.stencil_write_disable && z.z_read_base_addr == 0 && z.z_write_base_addr == 0 &&
+	    z.stencil_read_base_addr == 0 && z.stencil_write_base_addr == 0 &&
+	    z.htile_data_base_addr == 0 &&
 	    // DB_DEPTH_SIZE_XY is independent state and may remain programmed after the attachment
 	    // formats and addresses are unbound. A zero encoding is the valid 1x1 value, so its
 	    // presence alone must not manufacture a depth attachment.
-	    !z.z_info.htile_acceleration && !z.width_height_valid && !z.pitch_height_valid &&
-	    z.size.x_max == 0 && z.size.y_max == 0 && z.pitch_div8_minus1 == 0 &&
-	    z.height_div8_minus1 == 0 && z.slice_div64_minus1 == 0 && z.width == 0 && z.height == 0;
+	    !z.z_info.htile_acceleration && z.shading_rate_encoding == 0 && z.size.x_max == 0 &&
+	    z.size.y_max == 0;
 	if (attachment_unbound) {
 		static std::atomic_bool logged = false;
 		if (!logged.exchange(true, std::memory_order_relaxed)) {
@@ -151,6 +144,8 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 		return;
 	}
 	const bool has_htile = z.z_info.htile_acceleration;
+	const bool unsupported_shading_rate_encoding =
+	    z.shading_rate_encoding > 1 || (z.shading_rate_encoding != 0 && !has_htile);
 	const auto samples   = render_sample_count(z.z_info.num_samples);
 	if (samples == 0) {
 		DepthFatal("unsupported depth fragment count: %u", z.z_info.num_samples);
@@ -165,17 +160,12 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 			DepthFatal("invalid depth view: base=%u last=%u", z.depth_view.slice_start,
 			           z.depth_view.slice_max);
 	}
-	if (rc.copy_centroid || rc.copy_sample != 0 || z.z_info.expclear_enabled ||
+	if (rc.copy_depth_to_color || rc.copy_stencil_to_color || rc.copy_centroid ||
+	    rc.copy_sample != 0 || z.z_info.expclear_enabled ||
 	    z.stencil_info.expclear_enabled || z.z_info.partially_resident ||
 	    z.stencil_info.partially_resident || z.z_info.max_mip_level != 0 ||
-	    z.depth_view.current_mip_level != 0 || z.depth_info.addr5_swizzle_mask != 0 ||
-	    z.depth_info.array_mode != 0 || z.depth_info.pipe_config != 0 ||
-	    z.depth_info.bank_width != 0 || z.depth_info.bank_height != 0 ||
-	    z.depth_info.macro_tile_aspect != 0 || z.depth_info.num_banks != 0 ||
-	    z.htile_surface.linear != 0 || z.htile_surface.full_cache != 0 ||
-	    z.htile_surface.htile_uses_preload_win != 0 || z.htile_surface.preload != 0 ||
-	    z.htile_surface.prefetch_width != 0 || z.htile_surface.prefetch_height != 0 ||
-	    z.htile_surface.dst_outside_zero_to_one != 0 || z.z_read_base_addr == 0 ||
+	    z.depth_view.current_mip_level != 0 || unsupported_shading_rate_encoding ||
+	    z.z_read_base_addr == 0 ||
 	    (!z.depth_view.depth_write_disable && z.z_write_base_addr != z.z_read_base_addr) ||
 	    (z.z_read_base_addr & 0xffffu) != 0 ||
 	    dc.zfunc > static_cast<uint8_t>(vk::CompareOp::eAlways)) {
@@ -189,10 +179,7 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 		    (z.stencil_read_base_addr & 0xffffu) != 0) {
 			DepthFatal("unsupported stencil attachment state");
 		}
-	} else if (z.stencil_read_base_addr != 0 || z.stencil_write_base_addr != 0 ||
-	           !htile_stencil_compat ||
-	           z.stencil_info.texture_compatibility !=
-	               Prospero::TextureCompatibleStencil::kDisable) {
+	} else if (z.stencil_read_base_addr != 0 || z.stencil_write_base_addr != 0) {
 		DepthFatal("stencil state without an active stencil attachment");
 	}
 	if (has_htile) {
@@ -203,18 +190,13 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 			DepthFatal("HTile clear tracking supports at most 32 slices");
 		}
 	}
-	const bool size_xy_valid = z.size.valid;
-	const bool wh_valid      = z.width_height_valid && z.width != 0 && z.height != 0;
-	if (!size_xy_valid && !wh_valid) {
+	if (!z.size.valid) {
 		DepthFatal("missing depth extent");
 	}
-	const uint32_t width  = size_xy_valid ? static_cast<uint32_t>(z.size.x_max) + 1u : z.width;
-	const uint32_t height = size_xy_valid ? static_cast<uint32_t>(z.size.y_max) + 1u : z.height;
-	if (width > 16384 || height > 16384 ||
-	    (size_xy_valid && wh_valid && (width != z.width || height != z.height)) ||
-	    (!z.pitch_height_valid &&
-	     (z.pitch_div8_minus1 != 0 || z.height_div8_minus1 != 0 || z.slice_div64_minus1 != 0))) {
-		DepthFatal("inconsistent depth extent or encoded layout");
+	const uint32_t width  = static_cast<uint32_t>(z.size.x_max) + 1u;
+	const uint32_t height = static_cast<uint32_t>(z.size.y_max) + 1u;
+	if (width > 16384 || height > 16384) {
+		DepthFatal("invalid depth extent");
 	}
 	const auto* policy = FindDepthFormatPolicy(z.z_info.format);
 	if (policy == nullptr) {
@@ -229,11 +211,6 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 	const auto     guest_format = policy->guest_format;
 	const uint32_t bytes        = policy->bytes_per_element;
 	const auto     pitch        = TileGetDepthPitch(width, bytes, z.z_info.num_samples);
-	if (z.pitch_height_valid && ((static_cast<uint64_t>(z.pitch_div8_minus1) + 1u) * 8u != pitch ||
-	                             (static_cast<uint64_t>(z.height_div8_minus1) + 1u) * 8u !=
-	                                 ((static_cast<uint64_t>(height) + 7u) & ~7ull))) {
-		DepthFatal("encoded depth pitch or height mismatch");
-	}
 	TileSizeAlign depth_size {};
 	TileSizeAlign stencil_size {};
 	TileSizeAlign htile_size {};
@@ -244,14 +221,6 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 	    (has_htile != (htile_size.align == 32768 && htile_size.size != 0))) {
 		DepthFatal("unsupported depth/stencil/HTile footprint");
 	}
-	if (z.pitch_height_valid &&
-	    (static_cast<uint64_t>(z.slice_div64_minus1) + 1u) * 64u != depth_size.size) {
-		DepthFatal("depth footprint mismatch: extent=%ux%u pitch=%u expected=0x%016" PRIx64
-		           " align=0x%016" PRIx64 " encoded_valid=%u encoded=0x%016" PRIx64,
-		           width, height, pitch, depth_size.size, depth_size.align,
-		           z.pitch_height_valid ? 1u : 0u,
-		           (static_cast<uint64_t>(z.slice_div64_minus1) + 1u) * 64u);
-	}
 	if (depth_size.size > UINT64_MAX / view.image_layers ||
 	    stencil_size.size > UINT64_MAX / view.image_layers ||
 	    htile_size.size > UINT64_MAX / view.image_layers) {
@@ -260,9 +229,9 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 	const auto depth_backing_size   = depth_size.size * view.image_layers;
 	const auto stencil_backing_size = stencil_size.size * view.image_layers;
 	const auto htile_backing_size   = htile_size.size * view.image_layers;
-	if (depth_backing_size > TRACKER_ADDRESS_SIZE - z.z_read_base_addr ||
-	    (has_stencil && stencil_backing_size > TRACKER_ADDRESS_SIZE - z.stencil_read_base_addr) ||
-	    (has_htile && htile_backing_size > TRACKER_ADDRESS_SIZE - z.htile_data_base_addr)) {
+	if (!GuestRange {z.z_read_base_addr, depth_backing_size}.Valid() ||
+	    (has_stencil && !GuestRange {z.stencil_read_base_addr, stencil_backing_size}.Valid()) ||
+	    (has_htile && !GuestRange {z.htile_data_base_addr, htile_backing_size}.Valid())) {
 		DepthFatal("layered depth backing range is invalid");
 	}
 	r.htile                   = has_htile;

src/graphics/host_gpu/renderer/depthRenderTarget.h

@@ -11,9 +11,11 @@
 
 namespace Libs::Graphics {
 
+// Without a stencil plane, Hi-Stencil fields are inactive. An active plane is compatible when
+// Hi-Stencil is disabled or HTile backing is present.
 inline constexpr bool depth_htile_stencil_acceleration_compatible(bool has_stencil, bool has_htile,
                                                                   bool htile_stencil_disabled) {
-	return htile_stencil_disabled || (has_stencil && has_htile);
+	return !has_stencil || htile_stencil_disabled || has_htile;
 }
 
 struct RenderDepthInfo {

src/graphics/host_gpu/renderer/image/image.cpp

@@ -59,24 +59,14 @@ namespace {
 	if (HasFormatFeature(properties, vk::FormatFeatureFlagBits::eColorAttachment)) {
 		usage |= vk::ImageUsageFlagBits::eColorAttachment;
 	}
-	if (info.samples == 1 &&
-	    HasFormatFeature(properties, vk::FormatFeatureFlagBits::eStorageImage)) {
+	if (info.samples == 1) {
 		usage |= vk::ImageUsageFlagBits::eStorage;
-	} else if (info.samples == 1) {
-		const auto compatible = SrgbStorageViewFormat(info.pixel_format);
-		if (compatible != vk::Format::eUndefined &&
-		    HasFormatFeature(graphics.GetFormatProperties(compatible),
-		                     vk::FormatFeatureFlagBits::eStorageImage)) {
-			usage |= vk::ImageUsageFlagBits::eStorage;
-		}
 	}
 	return usage;
 }
 
-void ValidateRange(GuestRange range, const char* name) {
-	if ((range.address == 0) != (range.size == 0) ||
-	    (range.address != 0 && (range.address >= TRACKER_ADDRESS_SIZE ||
-	                            range.size > TRACKER_ADDRESS_SIZE - range.address))) {
+void ValidateOptionalRange(GuestRange range, const char* name) {
+	if (!range.ValidOrEmpty()) {
 		EXIT("invalid %s image range: address=0x%016llx size=0x%016llx\n", name,
 		     static_cast<unsigned long long>(range.address),
 		     static_cast<unsigned long long>(range.size));
@@ -567,13 +557,13 @@ void Image::CopyMip(Image& source, uint32_t mip, uint32_t layer) {
 namespace ImageOps {
 
 void Validate(const ImageInfo& info) {
-	ValidateRange(info.data, "data");
-	ValidateRange(info.stencil, "stencil");
+	ValidateOptionalRange(info.data, "data");
+	ValidateOptionalRange(info.stencil, "stencil");
 
 	if (info.pixel_format == vk::Format::eUndefined) {
 		const bool metadata_empty =
-		    info.metadata.range.address == 0 && info.metadata.range.size == 0 &&
-		    info.metadata.kind == ImageMetadataKind::None && info.metadata.control == 0 &&
+		    info.metadata.range.Empty() && info.metadata.kind == ImageMetadataKind::None &&
+		    info.metadata.control == 0 &&
 		    info.metadata.compression == VideoOutCompression::Uncompressed &&
 		    !info.metadata.stencil_compressed;
 		if (info.data.Empty() || info.HasStencil() || !metadata_empty || info.extent.width == 0 ||
@@ -620,16 +610,14 @@ void Validate(const ImageInfo& info) {
 	}
 	switch (info.metadata.kind) {
 		case ImageMetadataKind::None:
-			if (info.metadata.range.address != 0 || info.metadata.range.size != 0 ||
-			    info.metadata.control != 0 ||
+			if (!info.metadata.range.Empty() || info.metadata.control != 0 ||
 			    info.metadata.compression != VideoOutCompression::Uncompressed ||
 			    info.metadata.stencil_compressed) {
 				EXIT("metadata-free image has metadata state\n");
 			}
 			break;
 		case ImageMetadataKind::Htile:
-			ValidateRange(info.metadata.range, "HTILE");
-			if (info.metadata.range.Empty() ||
+			if (!info.metadata.range.Valid() ||
 			    info.metadata.compression != VideoOutCompression::Uncompressed) {
 				EXIT("invalid HTILE metadata\n");
 			}

src/graphics/host_gpu/renderer/image/imageInfo.h

@@ -15,18 +15,6 @@
 
 namespace Libs::Graphics {
 
-struct GuestRange {
-	uint64_t address = 0;
-	uint64_t size    = 0;
-
-	[[nodiscard]] constexpr bool Empty() const noexcept { return address == 0 || size == 0; }
-	[[nodiscard]] constexpr bool Valid() const noexcept {
-		return !Empty() && address < TRACKER_ADDRESS_SIZE && size <= TRACKER_ADDRESS_SIZE - address;
-	}
-	[[nodiscard]] constexpr uint64_t End() const noexcept { return address + size; }
-	auto                             operator<=>(const GuestRange&) const = default;
-};
-
 enum class VideoOutCompression : uint8_t { Uncompressed, Dcc256_256_0, Dcc256_64_64, Unsupported };
 
 enum class ImageMetadataKind : uint8_t { None, Htile, Dcc };
@@ -443,6 +431,10 @@ IsSupportedDisplayRenderTargetTileMode(Prospero::TileMode tile_mode) noexcept {
 		return encoded <= 0.04045f ? encoded / 12.92f : std::pow((encoded + 0.055f) / 1.055f, 2.4f);
 	};
 	switch (format) {
+		// A single-plane float target carries its clear as raw float bits, the same encoding the
+		// depth decoder below uses. Without this the clear is discarded and the target keeps stale
+		// contents.
+		case vk::Format::eR32Sfloat: next.float32[0] = std::bit_cast<float>(packed); break;
 		case vk::Format::eR32Uint: next.uint32[0] = packed; break;
 		case vk::Format::eR32Sint: next.int32[0] = static_cast<int32_t>(packed); break;
 		case vk::Format::eR8G8B8A8Srgb:

src/graphics/host_gpu/renderer/image/imageView.h

@@ -103,11 +103,13 @@ SelectSampledDepthView(vk::Format image_format, vk::Format view_format, uint32_t
 
 [[nodiscard]] inline bool
 IsSupportedSampledDepthResource(const ShaderRecompiler::IR::ImageResource& resource) noexcept {
-	if (resource.kind != ShaderRecompiler::IR::ResourceKind::Image &&
-	    resource.kind != ShaderRecompiler::IR::ResourceKind::ImageUint) {
+	if (resource.resource_class != ShaderRecompiler::IR::ImageResourceClass::Sampled) {
 		return false;
 	}
-	if (resource.kind == ShaderRecompiler::IR::ResourceKind::ImageUint && resource.depth_compare) {
+	if (resource.numeric_class == Prospero::TextureNumericClass::Unsupported) {
+		return false;
+	}
+	if (resource.numeric_class != Prospero::TextureNumericClass::Float && resource.depth_compare) {
 		return false;
 	}
 	return (resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2D ||
@@ -128,8 +130,9 @@ inline void ValidateStorageColorView(vk::Format image_format, vk::Format view_fo
 
 [[nodiscard]] inline bool
 IsSupportedStorageImageResource(const ShaderRecompiler::IR::ImageResource& resource) noexcept {
-	return (resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImage ||
-	        resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint) &&
+	return resource.resource_class == ShaderRecompiler::IR::ImageResourceClass::Storage &&
+	       (resource.numeric_class == Prospero::TextureNumericClass::Float ||
+	        resource.numeric_class == Prospero::TextureNumericClass::Uint) &&
 	       (resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim1D ||
 	        resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim1DArray ||
 	        resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2D ||
@@ -141,19 +144,19 @@ IsSupportedStorageImageResource(const ShaderRecompiler::IR::ImageResource& resou
 	         resource.mip_count != 0u)) &&
 	       resource.written &&
 	       (!resource.atomic ||
-	        (resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint &&
-	         resource.read)) &&
+	        (resource.numeric_class == Prospero::TextureNumericClass::Uint && resource.read)) &&
 	       !resource.depth_compare;
 }
 
 inline void
 ValidateStorageImageResource(const ShaderRecompiler::IR::ImageResource& resource) noexcept {
 	if (!IsSupportedStorageImageResource(resource)) {
-		EXIT("unsupported storage color image resource: kind=%u dimension=%u mip=%u "
+		EXIT("unsupported storage color image resource: class=%u numeric=%u dimension=%u mip=%u "
 		     "read=%d written=%d atomic=%d depth_compare=%d\n",
-		     static_cast<uint32_t>(resource.kind), static_cast<uint32_t>(resource.dimension),
-		     static_cast<uint32_t>(resource.mip_mode), resource.read, resource.written,
-		     resource.atomic, resource.depth_compare);
+		     static_cast<uint32_t>(resource.resource_class),
+		     static_cast<uint32_t>(resource.numeric_class),
+		     static_cast<uint32_t>(resource.dimension), static_cast<uint32_t>(resource.mip_mode),
+		     resource.read, resource.written, resource.atomic, resource.depth_compare);
 	}
 }
 

src/graphics/host_gpu/renderer/pipeline/descriptors.cpp

@@ -48,60 +48,27 @@ namespace {
 
 using BindingKind = ShaderRecompiler::IR::DescriptorBindingKind;
 
-bool IsSampledImage(BindingKind kind) {
-	switch (kind) {
-		case BindingKind::Sampled1D:
-		case BindingKind::Sampled1DArray:
-		case BindingKind::Sampled2D:
-		case BindingKind::Sampled2DArray:
-		case BindingKind::Sampled2DMsaa:
-		case BindingKind::Sampled2DMsaaArray:
-		case BindingKind::Sampled3D:
-		case BindingKind::SampledUint1D:
-		case BindingKind::SampledUint1DArray:
-		case BindingKind::SampledUint2D:
-		case BindingKind::SampledUint2DArray:
-		case BindingKind::SampledUint2DMsaa:
-		case BindingKind::SampledUint2DMsaaArray:
-		case BindingKind::SampledUint3D: return true;
-		default: return false;
-	}
-}
-
-bool IsStorageImage(BindingKind kind) {
-	switch (kind) {
-		case BindingKind::Storage1D:
-		case BindingKind::Storage1DArray:
-		case BindingKind::Storage2D:
-		case BindingKind::Storage2DArray:
-		case BindingKind::Storage3D:
-		case BindingKind::StorageUint1D:
-		case BindingKind::StorageUint1DArray:
-		case BindingKind::StorageUint2D:
-		case BindingKind::StorageUint2DArray:
-		case BindingKind::StorageUint3D:
-		case BindingKind::StorageAtomic1D:
-		case BindingKind::StorageAtomic1DArray:
-		case BindingKind::StorageAtomic2D:
-		case BindingKind::StorageAtomic2DArray:
-		case BindingKind::StorageAtomic3D: return true;
-		default: return false;
-	}
-}
-
 } // namespace
 
 vk::DescriptorType NativeDescriptorType(BindingKind kind) {
-	if (kind == BindingKind::Samplers) {
-		return vk::DescriptorType::eSampler;
-	}
-	if (IsSampledImage(kind)) {
+	const auto image_class = ShaderRecompiler::IR::ImageBindingResourceClass(kind);
+	if (image_class == ShaderRecompiler::IR::ImageResourceClass::Sampled) {
 		return vk::DescriptorType::eSampledImage;
 	}
-	if (IsStorageImage(kind)) {
+	if (image_class == ShaderRecompiler::IR::ImageResourceClass::Storage) {
 		return vk::DescriptorType::eStorageImage;
 	}
-	return vk::DescriptorType::eStorageBuffer;
+	switch (kind) {
+		case BindingKind::Samplers: return vk::DescriptorType::eSampler;
+		case BindingKind::Buffers:
+		case BindingKind::Gds:
+		case BindingKind::BdaPagetable:
+		case BindingKind::FaultBuffer:
+		case BindingKind::FlattenedSrt:
+		case BindingKind::ShaderData: return vk::DescriptorType::eStorageBuffer;
+		case BindingKind::Count: EXIT("invalid native descriptor binding kind");
+	}
+	EXIT("invalid native descriptor binding kind");
 }
 
 uint32_t NativeDescriptorCount(const ShaderRecompiler::IR::DescriptorBinding& binding) {
@@ -141,12 +108,6 @@ static void BindNullStorageBuffer(RenderContext& context, BufferView& dst) {
 	dst.range  = 16;
 }
 
-static void CopyNativeDescriptor(const ShaderRecompiler::IR::DescriptorValue& source,
-                                 std::span<uint32_t>                          destination) {
-	EXIT_IF(source.dword_count != destination.size());
-	std::copy_n(source.dwords.begin(), destination.size(), destination.begin());
-}
-
 static Prospero::ImageType TextureType(const ShaderTextureResource& descriptor) {
 	const auto type = descriptor.Type();
 	return type == Prospero::ImageType::kCube ? Prospero::ImageType::kColor2DArray : type;
@@ -229,11 +190,10 @@ static bool IsSupportedSampledColorResource(const ShaderRecompiler::IR::ImageRes
 			break;
 		default: break;
 	}
-	const bool sampled_kind = resource.kind == ShaderRecompiler::IR::ResourceKind::Image ||
-	                          resource.kind == ShaderRecompiler::IR::ResourceKind::ImageUint;
-	return sampled_kind && supported_dimension &&
-	       resource.mip_mode == ShaderRecompiler::IR::ImageMipMode::None && resource.read &&
-	       !resource.written && !resource.atomic && !resource.depth_compare;
+	return resource.resource_class == ShaderRecompiler::IR::ImageResourceClass::Sampled &&
+	       resource.numeric_class != Prospero::TextureNumericClass::Unsupported &&
+	       supported_dimension && resource.mip_mode == ShaderRecompiler::IR::ImageMipMode::None &&
+	       resource.read && !resource.written && !resource.atomic && !resource.depth_compare;
 }
 
 TargetTextureViewInfo ResolveTargetTextureView(const ShaderRecompiler::IR::ImageResource& resource,
@@ -387,15 +347,17 @@ static void ValidateDepthTargetBinding(const ShaderRecompiler::IR::ImageResource
 	    TileGetTexturePitch(descriptor.Format(), static_cast<uint32_t>(descriptor.Width5()) + 1u,
 	                        descriptor.TileMode());
 	EXIT("unsupported sampled depth target: resource=%d descriptor=%d encoding=%d format=%d "
-	     "kind=%u dimension=%u mip_mode=%u read=%d written=%d atomic=%d compare=%d "
+	     "class=%u numeric=%u dimension=%u mip_mode=%u read=%d written=%d atomic=%d compare=%d "
 	     "guest_format=%u swizzle=0x%03x image_format=%d view_format=%d image_layers=%u "
 	     "descriptor_type=%u base_array=%u depth=%u descriptor_pitch=%u target_pitch=%u "
 	     "addr=0x%016" PRIx64 " size=0x%016" PRIx64
 	     " dwords=%08x,%08x,%08x,%08x,%08x,%08x,%08x,%08x\n",
-	     resource_ok, descriptor_ok, encoding_ok, format_ok, static_cast<uint32_t>(resource.kind),
-	     static_cast<uint32_t>(resource.dimension), static_cast<uint32_t>(resource.mip_mode),
-	     resource.read, resource.written, resource.atomic, resource.depth_compare,
-	     static_cast<uint32_t>(descriptor.Format()), descriptor.DstSelXYZW(),
+	     resource_ok, descriptor_ok, encoding_ok, format_ok,
+	     static_cast<uint32_t>(resource.resource_class),
+	     static_cast<uint32_t>(resource.numeric_class), static_cast<uint32_t>(resource.dimension),
+	     static_cast<uint32_t>(resource.mip_mode), resource.read, resource.written, resource.atomic,
+	     resource.depth_compare, static_cast<uint32_t>(descriptor.Format()),
+	     descriptor.DstSelXYZW(),
 	     image == nullptr ? static_cast<int>(vk::Format::eUndefined)
 	                      : static_cast<int>(image->info.pixel_format),
 	     static_cast<int>(view_format), image == nullptr ? 0u : image->info.resources.layers,
@@ -508,29 +470,33 @@ void ValidateStorageTexture(const ShaderRecompiler::IR::ImageResource& resource,
 	const bool resource_ok   = IsSupportedStorageImageResource(resource);
 	const bool descriptor_ok = IsSupportedStorageTextureDescriptor(resource, descriptor);
 	const bool encoding_ok   = IsSupportedStorageTextureEncoding(resource, descriptor);
-	const bool uint_resource =
-	    resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint;
+	const bool uint_resource    = resource.numeric_class == Prospero::TextureNumericClass::Uint;
 	const bool raw_sint_storage = format == Prospero::BufferFormat::k32SInt && uint_resource &&
 	                              resource.written && !resource.read && !resource.atomic;
+	const auto numeric_class = Prospero::SampledTextureNumericClass(format);
 	const bool format_ok =
-	    raw_sint_storage || (Prospero::IsSampledTextureFormat(format) &&
-	                         uint_resource == Prospero::IsUintTextureFormat(format) &&
-	                         (!resource.atomic || format == Prospero::BufferFormat::k32UInt));
+	    raw_sint_storage ||
+	    (numeric_class != Prospero::TextureNumericClass::Unsupported &&
+	     numeric_class != Prospero::TextureNumericClass::Sint &&
+	     uint_resource == (numeric_class == Prospero::TextureNumericClass::Uint) &&
+	     (!resource.atomic || format == Prospero::BufferFormat::k32UInt));
 	if (resource_ok && descriptor_ok && encoding_ok && format_ok && size != 0) {
 		return;
 	}
 	EXIT("unsupported storage texture: resource=%d descriptor=%d encoding=%d format=%d "
-	     "kind=%u dimension=%u mip_mode=%u atomic=%d compare=%d "
+	     "class=%u numeric=%u dimension=%u mip_mode=%u atomic=%d compare=%d "
 	     "base_level=%u last_level=%u max_mip=%u min_lod=%u base_array=%u bc=%u msaa=%d "
 	     "depth_tile_bpe=%u swizzle_ok=%d "
 	     "addr=0x%016" PRIx64 " size=0x%016" PRIx64
 	     " extent=%ux%ux%u type=%u format=%u tile=%u swizzle=0x%03x read=%d written=%d "
 	     "dwords=%08x,%08x,%08x,%08x,%08x,%08x,%08x,%08x\n",
-	     resource_ok, descriptor_ok, encoding_ok, format_ok, static_cast<uint32_t>(resource.kind),
-	     static_cast<uint32_t>(resource.dimension), static_cast<uint32_t>(resource.mip_mode),
-	     resource.atomic, resource.depth_compare, descriptor.BaseLevel(), descriptor.LastLevel(),
-	     descriptor.MaxMip(), descriptor.MinLod(), descriptor.BaseArray5(), descriptor.BCSwizzle(),
-	     descriptor.MsaaDepth(), Prospero::RenderTargetBytesPerElement(format),
+	     resource_ok, descriptor_ok, encoding_ok, format_ok,
+	     static_cast<uint32_t>(resource.resource_class),
+	     static_cast<uint32_t>(resource.numeric_class), static_cast<uint32_t>(resource.dimension),
+	     static_cast<uint32_t>(resource.mip_mode), resource.atomic, resource.depth_compare,
+	     descriptor.BaseLevel(), descriptor.LastLevel(), descriptor.MaxMip(), descriptor.MinLod(),
+	     descriptor.BaseArray5(), descriptor.BCSwizzle(), descriptor.MsaaDepth(),
+	     Prospero::RenderTargetBytesPerElement(format),
 	     IsValidImageSwizzle(descriptor.DstSelXYZW()), descriptor.Base40(), size,
 	     static_cast<uint32_t>(descriptor.Width5()) + 1u,
 	     static_cast<uint32_t>(descriptor.Height5()) + 1u,
@@ -547,10 +513,16 @@ struct NullImageSpec {
 };
 
 static NullImageSpec NullTextureSpec(const ShaderRecompiler::IR::ImageResource& resource) {
-	const bool uint_image = resource.kind == ShaderRecompiler::IR::ResourceKind::ImageUint ||
-	                        resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint;
-	return uint_image ? NullImageSpec {vk::Format::eR32Uint, Prospero::BufferFormat::k32UInt}
-	                  : NullImageSpec {vk::Format::eR32Sfloat, Prospero::BufferFormat::k32Float};
+	switch (resource.numeric_class) {
+		case Prospero::TextureNumericClass::Float:
+			return {vk::Format::eR32Sfloat, Prospero::BufferFormat::k32Float};
+		case Prospero::TextureNumericClass::Uint:
+			return {vk::Format::eR32Uint, Prospero::BufferFormat::k32UInt};
+		case Prospero::TextureNumericClass::Sint:
+			return {vk::Format::eR32Sint, Prospero::BufferFormat::k32SInt};
+		case Prospero::TextureNumericClass::Unsupported: break;
+	}
+	EXIT("null image has unsupported numeric class\n");
 }
 
 static TextureCache::ImageDesc NullTextureDesc(const ShaderRecompiler::IR::ImageResource& resource,
@@ -680,8 +652,7 @@ static ImageViewInfo TextureViewInfo(const ShaderRecompiler::IR::ImageResource&
 
 TextureBinding RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageResource&   resource,
                                               const ShaderRecompiler::IR::DescriptorValue& value) {
-	ShaderTextureResource descriptor;
-	CopyNativeDescriptor(value, descriptor.fields);
+	auto descriptor = DecodeNativeDescriptor<ShaderTextureResource>(value);
 	const bool storage = resource.written;
 	if (storage) {
 		ValidateStorageImageResource(resource);
@@ -718,15 +689,16 @@ TextureBinding RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageR
 	      !msaa_tile || (descriptor.MsaaDepth() && !depth_tile) ||
 	      (!msaa_array && (descriptor.Depth() != 0 || descriptor.BaseArray5() != 0))))) {
 		EXIT("unsupported texture mip view: base=%u last=%u levels=%u max=%u type=%u tile=%u "
-		     "kind=%u dimension=%u mip_mode=%u read=%d written=%d "
+		     "class=%u numeric=%u dimension=%u mip_mode=%u read=%d written=%d "
 		     "dwords=%08x,%08x,%08x,%08x,%08x,%08x,%08x,%08x\n",
 		     base_level, last_level, levels, descriptor.MaxMip(),
 		     static_cast<uint32_t>(descriptor.Type()), static_cast<uint32_t>(tile),
-		     static_cast<uint32_t>(resource.kind), static_cast<uint32_t>(resource.dimension),
-		     static_cast<uint32_t>(resource.mip_mode), resource.read, resource.written,
-		     descriptor.fields[0], descriptor.fields[1], descriptor.fields[2], descriptor.fields[3],
-		     descriptor.fields[4], descriptor.fields[5], descriptor.fields[6],
-		     descriptor.fields[7]);
+		     static_cast<uint32_t>(resource.resource_class),
+		     static_cast<uint32_t>(resource.numeric_class),
+		     static_cast<uint32_t>(resource.dimension), static_cast<uint32_t>(resource.mip_mode),
+		     resource.read, resource.written, descriptor.fields[0], descriptor.fields[1],
+		     descriptor.fields[2], descriptor.fields[3], descriptor.fields[4], descriptor.fields[5],
+		     descriptor.fields[6], descriptor.fields[7]);
 	}
 	const auto samples = multisampled ? 1u << last_level : 1u;
 	const auto view_levels =
@@ -737,15 +709,11 @@ TextureBinding RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageR
 	const bool shader_conversion =
 	    surface_format.conversion_format != Prospero::BufferFormat::kInvalid;
 	const bool sampled_numeric_class =
-	    storage || (Prospero::IsSampledTextureFormat(format) &&
-	                (resource.kind == ShaderRecompiler::IR::ResourceKind::ImageUint) ==
-	                    Prospero::IsUintTextureFormat(format));
-	if (!storage &&
-	    (resource.kind == ShaderRecompiler::IR::ResourceKind::Image ||
-	     resource.kind == ShaderRecompiler::IR::ResourceKind::ImageUint) &&
+	    storage || resource.numeric_class == Prospero::SampledTextureNumericClass(format);
+	if (!storage && resource.resource_class == ShaderRecompiler::IR::ImageResourceClass::Sampled &&
 	    !sampled_numeric_class) {
-		EXIT("sampled image numeric class mismatch: kind=%u format=%u addr=0x%016" PRIx64 "\n",
-		     static_cast<uint32_t>(resource.kind), static_cast<uint32_t>(format), address);
+		EXIT("sampled image numeric class mismatch: numeric=%u format=%u addr=0x%016" PRIx64 "\n",
+		     static_cast<uint32_t>(resource.numeric_class), static_cast<uint32_t>(format), address);
 	}
 
 	const bool    volume       = type == Prospero::ImageType::kColor3D;
@@ -827,17 +795,11 @@ TextureBinding RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageR
 }
 
 static vk::Sampler NativeSampler(RenderContext&                       context,
-                                 const ShaderRecompiler::IR::Program& program, uint32_t index,
+                                 const ShaderRecompiler::IR::CompiledShaderInfo& program,
+                                 uint32_t index,
                                  const ShaderRecompiler::IR::DescriptorValue& value) {
-	ShaderSamplerResource descriptor;
-	CopyNativeDescriptor(value, descriptor.fields);
-	const bool depth_compare = std::any_of(program.info.sampled_pairs.begin(),
-	                                       program.info.sampled_pairs.end(), [&](const auto& pair) {
-		                                       return pair.sampler == index &&
-		                                              pair.image < program.info.images.size() &&
-		                                              program.info.images[pair.image].depth_compare;
-	                                       });
-	if (!depth_compare) {
+	auto descriptor = DecodeNativeDescriptor<ShaderSamplerResource>(value);
+	if (!program.info.samplers[index].depth_compare) {
 		descriptor.fields[0] &= ~(0x7u << 12u);
 	}
 	if (program.info.samplers[index].force_point_filtering) {
@@ -894,15 +856,10 @@ PreparedBindings RenderExecutor::PrepareBindings(const ShaderStageRuntime& runti
 	KYTY_PROFILER_FUNCTION();
 	EXIT_IF(!runtime);
 	const auto& program  = *runtime.program;
-	const auto& snapshot = *runtime.resources;
-	std::string error;
-	if (!ShaderRecompiler::IR::ValidateResourceSpecialization(program, snapshot, &error)) {
-		EXIT("invalid native shader runtime snapshot: %s\n", error.c_str());
-	}
-
+	const auto& snapshot = runtime.resources;
 	PreparedBindings prepared;
-	prepared.program  = runtime.program.get();
-	prepared.snapshot = runtime.resources.get();
+	prepared.program  = runtime.program;
+	prepared.snapshot = &runtime.resources;
 	auto& descriptors = prepared.resources;
 	descriptors.buffers.reserve(program.info.buffers.size());
 	descriptors.images.reserve(program.info.images.size());
@@ -915,17 +872,11 @@ PreparedBindings RenderExecutor::PrepareBindings(const ShaderStageRuntime& runti
 	for (uint32_t i = 0; i < program.info.samplers.size(); i++) {
 		descriptors.samplers.push_back(NativeSampler(m_context, program, i, snapshot.samplers[i]));
 	}
-	if (ShaderRecompiler::IR::FindBinding(
-	        program.bindings, ShaderRecompiler::IR::DescriptorBindingKind::FlattenedSrt) !=
-	    nullptr) {
-		prepared.flattened_srt.assign(snapshot.flattened_srt.begin(), snapshot.flattened_srt.end());
-	}
-
-	prepared.user_data.reserve(program.bindings.ShaderDataDwords());
+	prepared.shader_data.reserve(program.bindings.ShaderDataDwords());
 	for (const auto reg: program.bindings.user_data_registers) {
-		prepared.user_data.push_back(snapshot.user_data[reg - program.user_data_base]);
+		prepared.shader_data.push_back(snapshot.user_data[reg - program.user_data_base]);
 	}
-	prepared.user_data.resize(program.bindings.ShaderDataDwords());
+	prepared.shader_data.resize(program.bindings.ShaderDataDwords());
 	if (ShaderRecompiler::IR::FindBinding(
 	        program.bindings, ShaderRecompiler::IR::DescriptorBindingKind::Gds) != nullptr) {
 		descriptors.gds.buffer = m_context.GetBufferCache().GetGdsBuffer()->Handle();
@@ -940,22 +891,21 @@ void RenderExecutor::FindBuffers(PreparedBindings& prepared) {
 	const auto& snapshot = *prepared.snapshot;
 	auto&       cache    = m_context.GetBufferCache();
 
-	prepared.buffer_ids.clear();
-	prepared.buffer_ids.reserve(program.info.buffers.size());
+	prepared.buffer_sources.clear();
+	prepared.buffer_sources.reserve(program.info.buffers.size());
 	for (uint32_t i = 0; i < program.info.buffers.size(); i++) {
-		ShaderBufferResource descriptor;
-		CopyNativeDescriptor(snapshot.buffers[i], descriptor.fields);
+		auto descriptor = DecodeNativeDescriptor<ShaderBufferResource>(snapshot.buffers[i]);
 		const auto address = descriptor.Base48();
 		const auto stride  = descriptor.Stride();
 		const auto records = descriptor.NumRecords();
 		EXIT_IF(stride != 0 && records > UINT64_MAX / stride);
 		const auto requested_size = stride != 0 ? static_cast<uint64_t>(stride) * records : records;
 		if (address == 0 || requested_size == 0) {
-			prepared.buffer_ids.emplace_back();
+			prepared.buffer_sources.emplace_back(descriptor, BufferId {});
 			continue;
 		}
 		const auto size = Libs::LibKernel::Memory::ClampRangeSize(address, requested_size);
-		prepared.buffer_ids.push_back(cache.FindBuffer(address, size));
+		prepared.buffer_sources.emplace_back(descriptor, cache.FindBuffer(address, size));
 	}
 
 }
@@ -967,32 +917,33 @@ void RenderExecutor::RebindBuffers(PreparedBindings& prepared) {
 	const auto& snapshot  = *prepared.snapshot;
 	auto&       resources = prepared.resources;
 	const auto& layout    = program.bindings;
-	EXIT_IF(prepared.buffer_ids.size() != program.info.buffers.size());
+	EXIT_IF(prepared.buffer_sources.size() != program.info.buffers.size());
 
 	resources.buffers.clear();
 	resources.buffers.reserve(program.info.buffers.size());
-	EXIT_IF(prepared.user_data.size() != layout.ShaderDataDwords());
-	std::fill(prepared.user_data.begin() + layout.memory_offset_dword, prepared.user_data.end(), 0);
+	EXIT_IF(prepared.shader_data.size() != layout.ShaderDataDwords());
+	std::fill(prepared.shader_data.begin() + layout.memory_offset_dword,
+	          prepared.shader_data.end(), 0);
 	auto pack_memory_offset = [&](uint32_t index, uint32_t offset) {
 		const auto dword = layout.memory_offset_dword + index / 4u;
 		const auto shift = (index % 4u) * 8u;
-		prepared.user_data[dword] |= offset << shift;
+		prepared.shader_data[dword] |= offset << shift;
 	};
 	for (uint32_t i = 0; i < program.info.buffers.size(); i++) {
-		ShaderBufferResource descriptor;
-		CopyNativeDescriptor(snapshot.buffers[i], descriptor.fields);
+		const auto& [descriptor, buffer_id] = prepared.buffer_sources[i];
 		uint32_t buffer_offset = 0;
 		resources.buffers.push_back(NativeStorageBuffer(m_context, descriptor,
 		                                                program.info.buffers[i], program.stage, i,
-		                                                buffer_offset, prepared.buffer_ids[i]));
+		                                                buffer_offset, buffer_id));
 		pack_memory_offset(i, buffer_offset);
 	}
-	if (!prepared.flattened_srt.empty()) {
-		resources.flattened_srt = NativeUpload(m_context, prepared.flattened_srt);
+	if (ShaderRecompiler::IR::FindBinding(
+	        layout, ShaderRecompiler::IR::DescriptorBindingKind::FlattenedSrt) != nullptr) {
+		resources.flattened_srt = NativeUpload(m_context, snapshot.flattened_srt);
 	}
 	if (ShaderRecompiler::IR::FindBinding(
-	        program.bindings, ShaderRecompiler::IR::DescriptorBindingKind::UserData) != nullptr) {
-		resources.user_data = NativeUpload(m_context, prepared.user_data);
+	        program.bindings, ShaderRecompiler::IR::DescriptorBindingKind::ShaderData) != nullptr) {
+		resources.shader_data = NativeUpload(m_context, prepared.shader_data);
 	}
 }
 
@@ -1081,14 +1032,13 @@ void RenderExecutor::CommitBindings(CommandBuffer&                     buffer,
 	auto   vk_buffer        = buffer.Handle();
 	size_t descriptor_count = 0;
 	size_t write_count      = 0;
+	ShaderRecompiler::IR::PushData push_data;
+	bool                           has_push_data = false;
 	constexpr auto GraphicsStages =
 	    vk::ShaderStageFlagBits::eVertex | vk::ShaderStageFlagBits::eFragment;
-	const auto push_constant_stages = pipeline_bind_point == vk::PipelineBindPoint::eGraphics
-	                                      ? vk::ShaderStageFlags {GraphicsStages}
-	                                      : vk::ShaderStageFlags {vk::ShaderStageFlagBits::eCompute};
 	for (const auto* prepared: prepared_bindings) {
 		EXIT_IF(prepared == nullptr || prepared->program == nullptr ||
-		        prepared->snapshot == nullptr || prepared->committed);
+		        prepared->snapshot == nullptr);
 		write_count += prepared->program->bindings.descriptors.size();
 		for (const auto& binding: prepared->program->bindings.descriptors) {
 			descriptor_count += NativeDescriptorCount(binding);
@@ -1102,7 +1052,6 @@ void RenderExecutor::CommitBindings(CommandBuffer&                     buffer,
 	m_descriptor_buffers.clear();
 	m_descriptor_images.clear();
 	m_descriptor_writes.clear();
-	m_push_constants.fill(0);
 	m_descriptor_buffers.reserve(descriptor_count);
 	m_descriptor_images.reserve(descriptor_count);
 	m_descriptor_writes.reserve(write_count);
@@ -1166,48 +1115,54 @@ void RenderExecutor::CommitBindings(CommandBuffer&                     buffer,
 			write.descriptorCount   = NativeDescriptorCount(binding);
 			const auto buffer_start = m_descriptor_buffers.size();
 			const auto image_start  = m_descriptor_images.size();
-			switch (binding.kind) {
-				case BindingKind::Buffers:
-					for (const auto resource: binding.resources) {
-						const auto& view = descriptors.buffers.at(resource);
-						EXIT_IF(view.buffer == nullptr);
-						m_descriptor_buffers.emplace_back(view.buffer, view.offset, view.range);
-					}
-					break;
-				case BindingKind::BdaPagetable:
-				case BindingKind::FaultBuffer: {
-					auto& cache = m_context.GetBufferCache();
-					const auto* bda_buffer = binding.kind == BindingKind::BdaPagetable
-					                             ? cache.GetBdaPageTableBuffer()
-					                             : cache.GetFaultBuffer();
-					m_descriptor_buffers.emplace_back(bda_buffer->Handle(), 0, bda_buffer->Size());
-					break;
+			if (ShaderRecompiler::IR::ImageBindingResourceClass(binding.kind) !=
+			    ShaderRecompiler::IR::ImageResourceClass::None) {
+				for (const auto resource: binding.resources) {
+					m_descriptor_images.push_back(MakeImageInfo(
+					    descriptors.images.at(resource), m_image_occurrences.at(resource)++));
 				}
-				case BindingKind::FlattenedSrt:
-				case BindingKind::UserData:
-				case BindingKind::Gds: {
-					const auto& view =
-					    binding.kind == BindingKind::FlattenedSrt ? descriptors.flattened_srt
-					    : binding.kind == BindingKind::UserData   ? descriptors.user_data
-					                                              : descriptors.gds;
-					EXIT_IF(view.buffer == nullptr);
-					m_descriptor_buffers.emplace_back(view.buffer, view.offset, view.range);
-					break;
+			} else {
+				switch (binding.kind) {
+					case BindingKind::Buffers:
+						for (const auto resource: binding.resources) {
+							const auto& view = descriptors.buffers.at(resource);
+							EXIT_IF(view.buffer == nullptr);
+							m_descriptor_buffers.emplace_back(view.buffer, view.offset, view.range);
+						}
+						break;
+					case BindingKind::BdaPagetable:
+					case BindingKind::FaultBuffer: {
+						auto&       cache      = m_context.GetBufferCache();
+						const auto* bda_buffer = binding.kind == BindingKind::BdaPagetable
+						                             ? cache.GetBdaPageTableBuffer()
+						                             : cache.GetFaultBuffer();
+						m_descriptor_buffers.emplace_back(bda_buffer->Handle(), 0,
+						                                  bda_buffer->Size());
+						break;
+					}
+					case BindingKind::FlattenedSrt:
+					case BindingKind::ShaderData:
+					case BindingKind::Gds: {
+						const BufferView* view = &descriptors.gds;
+						if (binding.kind == BindingKind::FlattenedSrt) {
+							view = &descriptors.flattened_srt;
+						} else if (binding.kind == BindingKind::ShaderData) {
+							view = &descriptors.shader_data;
+						}
+						EXIT_IF(view->buffer == nullptr);
+						m_descriptor_buffers.emplace_back(view->buffer, view->offset, view->range);
+						break;
+					}
+					case BindingKind::Samplers:
+						for (const auto resource: binding.resources) {
+							const auto sampler = descriptors.samplers.at(resource);
+							EXIT_IF(sampler == nullptr);
+							m_descriptor_images.emplace_back(sampler, nullptr,
+							                                 vk::ImageLayout::eUndefined);
+						}
+						break;
+					case BindingKind::Count: EXIT("invalid descriptor binding kind");
 				}
-				case BindingKind::Samplers:
-					for (const auto resource: binding.resources) {
-						const auto sampler = descriptors.samplers.at(resource);
-						EXIT_IF(sampler == nullptr);
-						m_descriptor_images.emplace_back(sampler, nullptr,
-						                                 vk::ImageLayout::eUndefined);
-					}
-					break;
-				default:
-					for (const auto resource: binding.resources) {
-						m_descriptor_images.push_back(MakeImageInfo(
-						    descriptors.images.at(resource), m_image_occurrences.at(resource)++));
-					}
-					break;
 			}
 			if (m_descriptor_buffers.size() != buffer_start) {
 				write.pBufferInfo = m_descriptor_buffers.data() + buffer_start;
@@ -1225,20 +1180,22 @@ void RenderExecutor::CommitBindings(CommandBuffer&                     buffer,
 			EXIT_IF(m_image_occurrences[i] != expected);
 		}
 
-		if (program.bindings.push_constant_size != 0) {
-			const auto offset = program.bindings.push_constant_offset;
-			EXIT_IF(offset % sizeof(uint32_t) != 0 ||
-			        program.bindings.push_constant_size !=
-			            prepared->user_data.size() * sizeof(uint32_t) ||
-			        offset + program.bindings.push_constant_size >
-			            ShaderRecompiler::IR::NativePushConstantSize);
-			std::copy(prepared->user_data.begin(), prepared->user_data.end(),
-			          m_push_constants.begin() + offset / sizeof(uint32_t));
+		const auto shader_data_dwords = program.bindings.ShaderDataDwords();
+		EXIT_IF(prepared->shader_data.size() != shader_data_dwords);
+		if (program.bindings.UsesPushData()) {
+			std::ranges::copy(prepared->shader_data,
+			                  push_data.dwords.begin() + program.bindings.push_data_start_dword);
+			has_push_data = true;
 		}
 	}
-	vk_buffer.pushConstants(pipeline.pipeline_layout, push_constant_stages, 0,
-	                        ShaderRecompiler::IR::NativePushConstantSize,
-	                        m_push_constants.data());
+
+	if (has_push_data) {
+		const auto stages = pipeline_bind_point == vk::PipelineBindPoint::eGraphics
+		                        ? vk::ShaderStageFlags {GraphicsStages}
+		                        : vk::ShaderStageFlags {vk::ShaderStageFlagBits::eCompute};
+		vk_buffer.pushConstants(pipeline.pipeline_layout, stages, 0, sizeof(push_data),
+		                        push_data.dwords.data());
+	}
 
 	if (!m_descriptor_writes.empty()) {
 		EXIT_IF(pipeline.descriptor_set_layout == nullptr);
@@ -1258,9 +1215,6 @@ void RenderExecutor::CommitBindings(CommandBuffer&                     buffer,
 			                             0, nullptr);
 		}
 	}
-	for (auto* prepared: prepared_bindings) {
-		prepared->committed = true;
-	}
 }
 
 } // namespace Libs::Graphics

src/graphics/host_gpu/renderer/pipeline/descriptors.h

@@ -40,17 +40,15 @@ struct NativeDescriptors {
 	std::vector<vk::Sampler>    samplers;
 	BufferView                  gds;
 	BufferView                  flattened_srt;
-	BufferView                  user_data;
+	BufferView                  shader_data;
 };
 
 struct PreparedBindings {
-	const ShaderRecompiler::IR::Program*          program  = nullptr;
+	const ShaderRecompiler::IR::CompiledShaderInfo* program  = nullptr;
 	const ShaderRecompiler::IR::ResourceSnapshot* snapshot = nullptr;
 	NativeDescriptors                             resources;
-	std::vector<BufferId>                         buffer_ids;
-	std::vector<uint32_t>                         flattened_srt;
-	std::vector<uint32_t>                         user_data;
-	bool                                          committed = false;
+	std::vector<std::pair<ShaderBufferResource, BufferId>> buffer_sources;
+	std::vector<uint32_t>                         shader_data;
 };
 
 [[nodiscard]] vk::DescriptorType

src/graphics/host_gpu/renderer/pipeline/pipelineCache.cpp

@@ -1,6 +1,8 @@
 #include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
 
 #include "common/assert.h"
+#include "common/emulatorConfig.h"
+#include "common/file.h"
 #include "common/logging/log.h"
 #include "common/profiler.h"
 #include "graphics/guest_gpu/hardwareContext.h"
@@ -10,19 +12,68 @@
 #include "graphics/host_gpu/renderer/image/imageView.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
+#include "graphics/shader/recompiler/ShaderRecompiler.h"
 #include "graphics/shader/shaderCompiler.h"
+#include "kernel/memory.h"
+#include "kytyGitVersion.h"
+#include "loader/systemContent.h"
 
+#include <algorithm>
 #include <atomic>
+#include <cctype>
 #include <cstdio>
 #include <cstring>
+#include <fmt/format.h>
+#include <limits>
 #include <span>
+#include <spirv-tools/libspirv.hpp>
+#include <string_view>
+#include <tuple>
 #include <utility>
 #include <vector>
+#include <xxhash.h>
 
 namespace Libs::Graphics {
 
 namespace {
 
+std::string DriverCacheSignature(const vk::PhysicalDeviceProperties& properties) {
+	constexpr char hex[] = "0123456789abcdef";
+	std::string    uuid(VK_UUID_SIZE * 2, '0');
+	for (size_t i = 0; i < VK_UUID_SIZE; i++) {
+		uuid[i * 2]     = hex[properties.pipelineCacheUUID[i] >> 4u];
+		uuid[i * 2 + 1] = hex[properties.pipelineCacheUUID[i] & 0xfu];
+	}
+	return fmt::format("KytyPC1:{}:{:08x}:{:08x}:{:08x}:{}\n", KYTY_GIT_REVISION,
+	                   properties.vendorID, properties.deviceID, properties.driverVersion, uuid);
+}
+
+std::string PipelineCacheTitleId() {
+	std::string title_id;
+	if ((!Loader::SystemContentParamSfoGetString("TITLE_ID", &title_id) || title_id.empty()) &&
+	    (!Loader::SystemContentParamSfoGetString("CONTENT_ID", &title_id) || title_id.empty())) {
+		return {};
+	}
+	if (!std::ranges::all_of(title_id, [](unsigned char c) {
+		    return std::isalnum(c) != 0 || c == '-' || c == '_';
+	    })) {
+		return {};
+	}
+	return title_id;
+}
+
+template <typename... Args>
+void PipelineCacheLog(fmt::format_string<Args...> format, Args&&... args) {
+	auto message = fmt::format(format, std::forward<Args>(args)...);
+	message += '\n';
+	if (Log::GetDirection() != Log::Direction::Console) {
+		std::fwrite(message.data(), 1, message.size(), stdout);
+		std::fflush(stdout);
+	}
+	Log::Write(message);
+	Log::Flush();
+}
+
 void NormalizeStaticParamsForDynamicState(PipelineStaticParameters& static_params) {
 	static_params.viewport_scale[0]  = 0.5f;
 	static_params.viewport_scale[1]  = 0.5f;
@@ -37,42 +88,188 @@ void NormalizeStaticParamsForDynamicState(PipelineStaticParameters& static_param
 	static_params.scissor_ltrb[3] = 1;
 }
 
+bool ReadShaderGuestMemory(void*, uint64_t address, uint32_t* value) {
+	return value != nullptr &&
+	       Libs::LibKernel::Memory::TryReadGpuCleanBacking(address, value, sizeof(*value));
+}
+
+void DumpShaderSpirv(const char* stage_name, uint64_t shader_hash,
+                     const std::vector<uint32_t>& spirv) {
+	if (!Config::GraphicsDebugDumpEnabled()) {
+		return;
+	}
+	static std::atomic_int id = 0;
+	const auto path = Config::GetShaderLogFolder() / fmt::format("{:04d}_new_shader_{}_{:016x}.spv",
+	                                                             id++, stage_name, shader_hash);
+	Common::File::CreateDirectories(path.parent_path());
+	Common::File file(path);
+	if (file.IsInvalid()) {
+		const auto path_text = Common::PathToString(path);
+		LOGF_COLOR(Log::Color::BrightRed, "Can't create file: %s\n", path_text.c_str());
+		return;
+	}
+	file.Write(spirv.data(), spirv.size() * sizeof(uint32_t));
+}
+
+void DumpShaderOriginal(const char* stage_name, uint64_t shader_hash,
+                        std::span<const uint32_t> code, const std::string& decoded_dump) {
+	if (!Config::GraphicsDebugDumpEnabled()) {
+		return;
+	}
+	EXIT_IF(code.empty());
+	static std::atomic_int id = 0;
+	const auto base = Config::GetShaderLogFolder() / "original" /
+	                  fmt::format("{:04d}_new_shader_{}_{:016x}", id++, stage_name, shader_hash);
+	Common::File::CreateDirectories(base.parent_path());
+	for (const auto& [suffix, data, size]: {
+	         std::tuple {".bin", static_cast<const void*>(code.data()), code.size_bytes()},
+	         std::tuple {".rdna2", static_cast<const void*>(decoded_dump.data()),
+	                     decoded_dump.size()},
+	     }) {
+		if (size == 0) {
+			continue;
+		}
+		auto path = base;
+		path += suffix;
+		Common::File file(path);
+		if (file.IsInvalid()) {
+			const auto path_text = Common::PathToString(path);
+			LOGF_COLOR(Log::Color::BrightRed, "Can't create file: %s\n", path_text.c_str());
+		} else {
+			file.Write(data, size);
+		}
+	}
+}
+
+bool ValidateShaderSpirv(const char* label, uint64_t shader_hash,
+                         const std::vector<uint32_t>& spirv) {
+	if (!Config::ShaderValidationEnabled()) {
+		return true;
+	}
+	spvtools::SpirvTools tools(SPV_ENV_VULKAN_1_3);
+	std::string          messages;
+	tools.SetMessageConsumer([&messages](spv_message_level_t, const char*,
+	                                     const spv_position_t& position, const char* message) {
+		messages += fmt::format("{}: {} ({}) {}\n", static_cast<int>(position.line),
+		                        static_cast<int>(position.column), static_cast<int>(position.index),
+		                        message);
+	});
+	if (tools.Validate(spirv)) {
+		return true;
+	}
+	spvtools::SpirvTools disassembler(SPV_ENV_VULKAN_1_2);
+	std::string          text;
+	disassembler.Disassemble(spirv, &text,
+	                         static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_NO_HEADER) |
+	                             static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_FRIENDLY_NAMES) |
+	                             static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_COMMENT) |
+	                             static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_INDENT) |
+	                             static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_COLOR));
+	LOGF_COLOR(Log::Color::BrightRed, "%s SPIR-V validation failed hash=0x%016" PRIx64 ":\n%s",
+	           label, shader_hash, messages.c_str());
+	LOGF("%s\n", text.c_str());
+	return false;
+}
+
 } // namespace
 
 struct PipelineCache::ProgramCache {
-	struct SourceKey {
-		ShaderType stage = ShaderType::Unknown;
-		uint64_t   hash  = 0;
+	struct ProgramKey {
+		ShaderType            stage           = ShaderType::Unknown;
+		uint64_t              hash            = 0;
+		uint32_t              user_data_count = 0;
+		uint32_t              code_size       = 0;
+		std::vector<uint32_t> static_state;
 
-		bool operator==(const SourceKey&) const = default;
+		bool operator==(const ProgramKey&) const = default;
 	};
 
 	struct Permutation {
-		std::vector<uint32_t>                               static_key;
-		std::shared_ptr<const ShaderRecompiler::IR::Program> program;
-		ShaderProgram                                       handle;
+		ShaderRecompiler::IR::ResourceSpecialization specialization;
+		ShaderRecompiler::IR::CompiledShaderInfo     program;
+		ShaderProgram                                handle;
 	};
 
-	struct SourceKeyHash {
-		std::size_t operator()(const SourceKey& key) const {
+	struct SourceEntry {
+		explicit SourceEntry(ShaderRecompiler::IR::ResourcePlan plan)
+		    : resource_plan(std::move(plan)) {
+			permutations.reserve(8);
+		}
+
+		ShaderRecompiler::IR::ResourcePlan resource_plan;
+		std::vector<Permutation>           permutations;
+	};
+
+	struct ProgramKeyHash {
+		std::size_t operator()(const ProgramKey& key) const {
 			std::size_t hash = static_cast<std::size_t>(key.stage);
 			PipelineKeyHash::Mix(hash, static_cast<std::size_t>(key.hash));
 			if constexpr (sizeof(std::size_t) < sizeof(uint64_t)) {
 				PipelineKeyHash::Mix(hash, static_cast<std::size_t>(key.hash >> 32u));
 			}
+			PipelineKeyHash::Mix(hash, key.user_data_count);
+			PipelineKeyHash::Mix(hash, key.code_size);
+			PipelineKeyHash::Mix(hash, key.static_state.size());
+			// Bucket same-shape static variants by source. ProgramKey equality performs the one
+			// exact state comparison needed on a stable hit without hashing up to 429 words first.
 			return hash;
 		}
 	};
 
-	static uint64_t MixId(uint64_t hash, uint64_t value) {
-		return hash ^ (value + 0x9e3779b97f4a7c15ull + (hash << 6u) + (hash >> 2u));
+	static constexpr std::size_t MaxStaticKeyWords = 13 + ShaderVertexInputInfo::RES_MAX * 13;
+
+	template <ShaderType Stage>
+	Permutation CompilePermutation(const ShaderParams&                          params,
+	                               const ShaderRecompiler::CompileOptions&      options,
+	                               ShaderRecompiler::TranslateResult            translated,
+	                               ShaderRecompiler::IR::ResourceSpecialization specialization,
+	                               uint32_t push_data_start_dword) {
+		constexpr const char* stage_name = [] {
+			if constexpr (Stage == ShaderType::Vertex) {
+				return "vs";
+			} else if constexpr (Stage == ShaderType::Pixel) {
+				return "ps";
+			} else {
+				static_assert(Stage == ShaderType::Compute);
+				return "cs";
+			}
+		}();
+		auto result = ShaderRecompiler::CompileProgram(std::move(translated), options,
+		                                               specialization, push_data_start_dword);
+		DumpShaderOriginal(stage_name, options.shader_hash, params.code, result.decoded_dump);
+		if (!ValidateShaderSpirv(options.dump_label, options.shader_hash, result.spirv)) {
+			DumpShaderSpirv(stage_name, options.shader_hash, result.spirv);
+			EXIT("%s failed hash=0x%016" PRIx64 ": SPIR-V validation failed\n", options.dump_label,
+			     options.shader_hash);
+		}
+		DumpShaderSpirv(stage_name, options.shader_hash, result.spirv);
+
+		vk::ShaderModuleCreateInfo create_info {};
+		create_info.sType       = vk::StructureType::eShaderModuleCreateInfo;
+		create_info.codeSize    = result.spirv.size() * sizeof(uint32_t);
+		create_info.pCode       = result.spirv.data();
+		vk::ShaderModule module = nullptr;
+		RequireVulkanSuccess(device.createShaderModule(&create_info, nullptr, &module),
+		                     "create recompiled shader module");
+		EXIT_IF(module == nullptr);
+		if (options.dump_ir) {
+			if (!options.early_dump) {
+				LOGF("%s decoded RDNA2:\n%s", options.dump_label, result.decoded_dump.c_str());
+				LOGF("%s IR:\n%s", options.dump_label, result.ir_dump.c_str());
+			}
+			LOGF("%s SPIR-V words=%" PRIu64 " wave_size=%u\n", options.dump_label,
+			     static_cast<uint64_t>(result.spirv.size()), options.wave_size);
+		}
+		return {
+		    .specialization = std::move(specialization),
+		    .program        = std::move(result.program).TakeCompiledInfo(),
+		    .handle         = {.id = ++next_shader_id, .module = module},
+		};
 	}
 
-	static constexpr std::size_t MaxStaticKeyWords =
-	    5 + ShaderVertexInputInfo::RES_MAX * 17;
-
 	template <typename InputInfo>
-	ShaderProgram Get(const ShaderParams& params, InputInfo& input_info) {
+	ShaderProgram Get(const ShaderParams& params, InputInfo& input_info,
+	                  uint32_t& push_data_cursor) {
 		constexpr ShaderType stage = [] {
 			if constexpr (std::is_same_v<InputInfo, ShaderVertexInputInfo>) {
 				return ShaderType::Vertex;
@@ -84,44 +281,116 @@ struct PipelineCache::ProgramCache {
 			}
 		}();
 
-		BuildStageStaticKey(input_info, key_scratch);
-		auto& permutations = programs[{stage, params.hash}];
-		for (const auto& permutation: permutations) {
-			if (permutation.static_key == key_scratch &&
-			    MaterializeProgram(permutation.program, params, input_info)) {
-				return permutation.handle;
+		lookup_key.stage           = stage;
+		lookup_key.hash            = params.hash;
+		lookup_key.user_data_count = static_cast<uint32_t>(params.user_data.size());
+		lookup_key.code_size       = static_cast<uint32_t>(params.code.size());
+		BuildStageStaticKey(input_info, lookup_key.static_state);
+		auto                                         entry = programs.find(lookup_key);
+		ShaderRecompiler::IR::ResourceSnapshot       resources;
+		ShaderRecompiler::IR::ResourceSpecialization specialization;
+		const ShaderRecompiler::IR::SrtRuntime       runtime {
+		    .user_data                  = params.user_data,
+		    .shader_base                = params.Base(),
+		    .read_specialization_memory = ReadShaderGuestMemory,
+		};
+		if (entry != programs.end()) {
+			EXIT_IF(!ShaderRecompiler::IR::MaterializeResources(
+			    entry->second.resource_plan, runtime, resources, specialization));
+			if (const auto permutation = std::ranges::find_if(
+			        entry->second.permutations, [&](const Permutation& candidate) {
+				        const auto& layout = candidate.program.bindings;
+				        return layout.push_data_start_dword ==
+				                   ShaderRecompiler::IR::PushData::StartFor(
+				                       push_data_cursor, layout.ShaderDataDwords()) &&
+				               candidate.specialization == specialization;
+			        });
+			    permutation != entry->second.permutations.end()) {
+				input_info.stage = {.program   = &permutation->program,
+				                    .resources = std::move(resources)};
+				permutation->program.bindings.AdvancePushData(push_data_cursor);
+				return permutation->handle;
 			}
 		}
 
-		const auto module = CompileProgram(device, params, input_info);
-		EXIT_IF(module == nullptr || !input_info.stage);
-		uint64_t id = MixId(MixId(params.hash, static_cast<uint64_t>(stage)), permutations.size());
-		if (id == 0) {
-			id = 1;
+		ShaderStageInputInfo stage_input {};
+		if constexpr (std::is_same_v<InputInfo, ShaderVertexInputInfo>) {
+			stage_input.vertex = &input_info;
+		} else if constexpr (std::is_same_v<InputInfo, ShaderPixelInputInfo>) {
+			stage_input.pixel = &input_info;
+		} else {
+			stage_input.compute = &input_info;
 		}
-		const ShaderProgram handle {.id = id, .module = module};
-		permutations.push_back({key_scratch, input_info.stage.program, handle});
+		constexpr const char* label = [] {
+			if constexpr (stage == ShaderType::Vertex) {
+				return "ShaderRecompiler VS";
+			} else if constexpr (stage == ShaderType::Pixel) {
+				return "ShaderRecompiler PS";
+			} else {
+				return "ShaderRecompiler CS";
+			}
+		}();
+		ShaderRecompiler::CompileOptions options;
+		options.stage       = stage;
+		options.shader_hash = params.hash;
+		options.user_data   = params.user_data;
+		options.dump_ir     = Config::GetShaderLogDirection() != Config::ShaderLogDirection::Silent;
+		options.early_dump  = options.dump_ir;
+		options.dump_label  = label;
+		options.input_info  = stage_input;
+		if constexpr (stage == ShaderType::Vertex) {
+			options.user_data_base = 8;
+			options.scratch_dwords = input_info.scratch_size_dwords;
+		} else if constexpr (stage == ShaderType::Pixel) {
+			options.scratch_dwords = input_info.scratch_size_dwords;
+		} else {
+			options.scratch_dwords = input_info.scratch_size_dwords;
+			options.wave_size      = input_info.wave_size;
+		}
+		auto translated = ShaderRecompiler::TranslateProgram(params.code, options);
+		if (entry == programs.end()) {
+			auto resource_plan = ShaderRecompiler::IR::ExtractResourcePlan(translated.program);
+			EXIT_IF(!ShaderRecompiler::IR::MaterializeResources(resource_plan, runtime, resources,
+			                                                    specialization));
+			entry = programs.try_emplace(lookup_key, std::move(resource_plan)).first;
+		}
+		entry->second.permutations.push_back(CompilePermutation<stage>(
+		    params, options, std::move(translated), std::move(specialization), push_data_cursor));
+		const auto& permutation = entry->second.permutations.back();
+		input_info.stage = {.program = &permutation.program, .resources = std::move(resources)};
+		permutation.program.bindings.AdvancePushData(push_data_cursor);
 
 		std::printf("Num compiled %u shaders\n", ++num_compiled);
-		return handle;
+		return permutation.handle;
 	}
 
 	explicit ProgramCache(vk::Device device): device(device) {
-		key_scratch.reserve(MaxStaticKeyWords);
+		lookup_key.static_state.reserve(MaxStaticKeyWords);
+	}
+	~ProgramCache() {
+		for (const auto& [key, entry]: programs) {
+			(void)key;
+			for (const auto& permutation: entry.permutations) {
+				device.destroyShaderModule(permutation.handle.module, nullptr);
+			}
+		}
 	}
 
-	std::unordered_map<SourceKey, std::vector<Permutation>, SourceKeyHash> programs;
-	std::vector<uint32_t> key_scratch;
-	vk::Device device;
-	uint32_t   num_compiled = 0;
+	std::unordered_map<ProgramKey, SourceEntry, ProgramKeyHash> programs;
+	ProgramKey                                                  lookup_key;
+	vk::Device                                                  device;
+	uint32_t                                                    num_compiled   = 0;
+	uint64_t                                                    next_shader_id = 0;
 };
 
 PipelineCache::PipelineCache(GraphicContext& graphics)
     : m_graphics(graphics), m_program_cache(std::make_unique<ProgramCache>(graphics.device)) {
 	EXIT_NOT_IMPLEMENTED(!Common::Thread::IsMainThread());
+	InitializeDriverCache();
 }
 
 PipelineCache::~PipelineCache() {
+	Save();
 	auto destroy = [this](const auto& pipelines) {
 		for (const auto& [key, pipeline]: pipelines) {
 			(void)key;
@@ -132,39 +401,199 @@ PipelineCache::~PipelineCache() {
 	};
 	destroy(m_graphics_pipelines);
 	destroy(m_compute_pipelines);
-	for (const auto& [key, permutations]: m_program_cache->programs) {
-		(void)key;
-		for (const auto& permutation: permutations) {
-			m_graphics.device.destroyShaderModule(permutation.handle.module, nullptr);
-		}
+	if (m_driver_cache != nullptr) {
+		m_graphics.device.destroyPipelineCache(m_driver_cache, nullptr);
 	}
 }
 
-ShaderProgram PipelineCache::GetVertexProgram(const HW::VertexShaderInfo& regs,
-                                              const HW::ShaderRegisters&  sh,
-                                              ShaderVertexInputInfo&      input_info) {
-	const auto params = PrepareProgram(regs, sh, input_info);
-	Common::LockGuard lock(m_mutex);
-	return m_program_cache->Get(params, input_info);
+void PipelineCache::InitializeDriverCache() {
+	const auto title_id = PipelineCacheTitleId();
+	if (title_id.empty()) {
+		return;
+	}
+	if (KYTY_BUILD != KYTY_BUILD_RELEASE) {
+		PipelineCacheLog("Vulkan pipeline cache: disabled (non-Release build)");
+		return;
+	}
+	const std::string_view git_hash     = KYTY_GIT_HASH;
+	const std::string_view git_revision = KYTY_GIT_REVISION;
+	if (git_hash == "unknown" || git_revision == "unknown") {
+		PipelineCacheLog("Vulkan pipeline cache: disabled (unknown git revision)");
+		return;
+	}
+	if (git_hash.ends_with("-dirty")) {
+		PipelineCacheLog("Vulkan pipeline cache: disabled (dirty build)");
+		return;
+	}
+
+	m_driver_cache_path     = std::filesystem::path("_PipelineCache") / (title_id + ".bin");
+	const auto path         = Common::PathToString(m_driver_cache_path);
+	const bool cache_exists = Common::File::IsFileExisting(m_driver_cache_path);
+	if (cache_exists) {
+		PipelineCacheLog("Vulkan pipeline cache: loading {}", path);
+	} else {
+		PipelineCacheLog("Vulkan pipeline cache: initializing {}", path);
+	}
+	std::vector<uint8_t> initial_data;
+	if (cache_exists) {
+		Common::File file(m_driver_cache_path, Common::File::Mode::Read);
+		const auto   file_size = file.IsInvalid() ? 0 : file.Size();
+		const auto   signature = DriverCacheSignature(m_graphics.GetPhysicalDeviceProperties());
+		if (file_size >= signature.size() + sizeof(uint64_t) &&
+		    file_size <= std::numeric_limits<uint32_t>::max()) {
+			std::string cached_signature(signature.size(), '\0');
+			uint64_t    payload_hash = 0;
+			initial_data.resize(file_size - signature.size() - sizeof(payload_hash));
+			uint32_t signature_read = 0;
+			uint32_t hash_read      = 0;
+			uint32_t payload_read   = 0;
+			file.Read(cached_signature.data(), static_cast<uint32_t>(cached_signature.size()),
+			          &signature_read);
+			file.Read(&payload_hash, sizeof(payload_hash), &hash_read);
+			file.Read(initial_data.data(), static_cast<uint32_t>(initial_data.size()),
+			          &payload_read);
+			file.Close();
+			if (signature_read != cached_signature.size() || hash_read != sizeof(payload_hash) ||
+			    payload_read != initial_data.size() || cached_signature != signature ||
+			    XXH3_64bits(initial_data.data(), initial_data.size()) != payload_hash) {
+				initial_data.clear();
+				PipelineCacheLog(
+				    "Vulkan pipeline cache: invalidating {} (driver, emulator, or data mismatch)",
+				    path);
+			}
+		} else {
+			file.Close();
+			PipelineCacheLog("Vulkan pipeline cache: invalidating {} (invalid file size)", path);
+		}
+	}
+
+	vk::PipelineCacheCreateInfo create {};
+	create.sType           = vk::StructureType::ePipelineCacheCreateInfo;
+	create.initialDataSize = initial_data.size();
+	create.pInitialData    = initial_data.empty() ? nullptr : initial_data.data();
+	auto result = m_graphics.device.createPipelineCache(&create, nullptr, &m_driver_cache);
+	if (result != vk::Result::eSuccess && !initial_data.empty()) {
+		PipelineCacheLog("Vulkan pipeline cache: driver rejected {} ({}); starting empty", path,
+		                 VulkanToString(result));
+		initial_data.clear();
+		create.initialDataSize = 0;
+		create.pInitialData    = nullptr;
+		result = m_graphics.device.createPipelineCache(&create, nullptr, &m_driver_cache);
+	}
+	if (result != vk::Result::eSuccess) {
+		PipelineCacheLog("Vulkan pipeline cache: disabled ({})", VulkanToString(result));
+		m_driver_cache = nullptr;
+		return;
+	}
+	if (!initial_data.empty()) {
+		PipelineCacheLog("Vulkan pipeline cache: loaded {} bytes from {}", initial_data.size(),
+		                 path);
+	} else {
+		PipelineCacheLog("Vulkan pipeline cache: initialized empty");
+	}
 }
 
-ShaderProgram PipelineCache::GetPixelProgram(
-    const HW::PixelShaderInfo& regs, const HW::ShaderRegisters& sh,
-    const ShaderVertexInputInfo&                        vertex_info,
-    std::span<const Prospero::ColorComponentMapping, 8> target_export_mapping,
-    ShaderPixelInputInfo&                               input_info) {
-	const auto params = PrepareProgram(regs, sh, vertex_info, target_export_mapping, input_info);
+void PipelineCache::Save() {
 	Common::LockGuard lock(m_mutex);
-	return m_program_cache->Get(params, input_info);
+	if (m_driver_cache == nullptr) {
+		return;
+	}
+
+	size_t               size = 0;
+	vk::Result           result;
+	std::vector<uint8_t> payload;
+	for (uint32_t attempt = 0; attempt < 3; attempt++) {
+		size   = 0;
+		result = m_graphics.device.getPipelineCacheData(m_driver_cache, &size, nullptr);
+		if (result != vk::Result::eSuccess || size == 0 ||
+		    size > std::numeric_limits<uint32_t>::max()) {
+			break;
+		}
+		payload.resize(size);
+		result = m_graphics.device.getPipelineCacheData(m_driver_cache, &size, payload.data());
+		if (result != vk::Result::eIncomplete) {
+			break;
+		}
+	}
+	if (result != vk::Result::eSuccess || size == 0 ||
+	    size > std::numeric_limits<uint32_t>::max()) {
+		PipelineCacheLog("Vulkan pipeline cache: save failed ({}, {} bytes)",
+		                 VulkanToString(result), size);
+		return;
+	}
+	payload.resize(size);
+	auto       prefix       = DriverCacheSignature(m_graphics.GetPhysicalDeviceProperties());
+	const auto payload_hash = XXH3_64bits(payload.data(), payload.size());
+	prefix.append(reinterpret_cast<const char*>(&payload_hash), sizeof(payload_hash));
+	if (!Common::File::CreateDirectories(m_driver_cache_path.parent_path())) {
+		PipelineCacheLog("Vulkan pipeline cache: failed to create cache directory");
+		return;
+	}
+	auto temp_path = m_driver_cache_path;
+	temp_path += ".tmp";
+	Common::File file;
+	uint32_t     prefix_written  = 0;
+	uint32_t     payload_written = 0;
+	if (file.Create(temp_path)) {
+		file.Write(prefix.data(), static_cast<uint32_t>(prefix.size()), &prefix_written);
+		file.Write(payload.data(), static_cast<uint32_t>(payload.size()), &payload_written);
+	}
+	const bool flushed = !file.IsInvalid() && file.Flush();
+	file.Close();
+	if (prefix_written != prefix.size() || payload_written != payload.size() || !flushed ||
+	    !Common::File::RenameFile(temp_path, m_driver_cache_path)) {
+		PipelineCacheLog("Vulkan pipeline cache: failed to write {}",
+		                 Common::PathToString(m_driver_cache_path));
+		return;
+	}
+	PipelineCacheLog("Vulkan pipeline cache: saved {} bytes to {}", payload.size(),
+	                 Common::PathToString(m_driver_cache_path));
+	m_graphics.device.destroyPipelineCache(m_driver_cache, nullptr);
+	m_driver_cache = nullptr;
+}
+
+PipelineCache::GraphicsPrograms PipelineCache::GetGraphicsPrograms(
+    const HW::VertexShaderInfo& vertex_regs, const HW::PixelShaderInfo& pixel_regs,
+    const HW::ShaderRegisters& sh, const HW::Context& context,
+    std::span<const Prospero::ColorComponentMapping, 8> target_export_mapping,
+    bool pixel_active, ShaderVertexInputInfo& vertex_info, ShaderPixelInputInfo& pixel_info) {
+	const auto vertex_params = PrepareProgram(vertex_regs, sh, vertex_info);
+	ShaderParams pixel_params;
+	if (pixel_active) {
+		pixel_params = PrepareProgram(pixel_regs, sh, target_export_mapping, pixel_info);
+	}
+	if (context.GetClipControl().clip_disable) {
+		const auto& viewport = context.GetScreenViewport().viewports[0];
+		const auto& limits   = m_graphics.GetPhysicalDeviceProperties().limits;
+		auto&       clip     = vertex_info.clip_space;
+		clip.scale[0]        = viewport.xscale;
+		clip.scale[1]        = viewport.yscale;
+		clip.offset[0]       = viewport.xoffset;
+		clip.offset[1]       = viewport.yoffset;
+		clip.half_extent[0] =
+		    static_cast<float>(std::min(limits.maxViewportDimensions[0], 16384u)) * 0.5f;
+		clip.half_extent[1] =
+		    static_cast<float>(std::min(limits.maxViewportDimensions[1], 16384u)) * 0.5f;
+		clip.enabled = true;
+	}
+	Common::LockGuard lock(m_mutex);
+	uint32_t          push_data_cursor = 0;
+	GraphicsPrograms  result;
+	if (pixel_active) {
+		result.pixel = m_program_cache->Get(pixel_params, pixel_info, push_data_cursor);
+	}
+	result.vertex = m_program_cache->Get(vertex_params, vertex_info, push_data_cursor);
+	return result;
 }
 
 ShaderProgram PipelineCache::GetComputeProgram(const HW::ComputeShaderInfo& regs,
                                                const HW::ShaderRegisters&   sh,
                                                ShaderComputeInputInfo&      input_info) {
 	input_info.needs_lds_barriers = !m_graphics.compute_wave64_supported;
-	const auto params             = PrepareProgram(regs, sh, input_info);
+	const auto        params      = PrepareProgram(regs, sh, input_info);
 	Common::LockGuard lock(m_mutex);
-	return m_program_cache->Get(params, input_info);
+	uint32_t          push_data_cursor = 0;
+	return m_program_cache->Get(params, input_info, push_data_cursor);
 }
 
 bool PipelineStaticParameters::operator==(const PipelineStaticParameters& other) const noexcept {
@@ -188,8 +617,7 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 	Common::LockGuard lock(m_mutex);
 	auto&             ctx = command.GetRegisters();
 
-	const HW::BlendColor& bclr                                     = ctx.GetBlendColor();
-	uint32_t              color_mask[RENDER_COLOR_ATTACHMENTS_MAX] = {};
+	uint32_t color_mask[RENDER_COLOR_ATTACHMENTS_MAX] = {};
 	for (uint32_t i = 0; i < color_count; i++) {
 		color_mask[i] =
 		    (colors[i].image_id ? colors[i].export_mapping.ApplyMask(render_target_mask_slot(
@@ -293,11 +721,6 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 		static_params.blend_enable[i]         = bc.enable;
 		static_params.blend_bypass[i]         = rt.info.blend_bypass;
 	}
-	static_params.blend_color_red   = bclr.red;
-	static_params.blend_color_green = bclr.green;
-	static_params.blend_color_blue  = bclr.blue;
-	static_params.blend_color_alpha = bclr.alpha;
-
 	NormalizeStaticParamsForDynamicState(static_params);
 
 	GraphicsPipelineKey key {};
@@ -305,6 +728,30 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 	key.vs_shader_id  = p.vs_shader_id;
 	key.ps_shader_id  = p.ps_shader_id;
 	key.static_params = static_params;
+	EXIT_IF(vs_input_info.buffers_num < 0 ||
+	        vs_input_info.buffers_num > ShaderVertexInputInfo::RES_MAX ||
+	        vs_input_info.resources_num < 0 ||
+	        vs_input_info.resources_num > ShaderVertexInputInfo::RES_MAX);
+	key.vertex_input.binding_count   = static_cast<uint8_t>(vs_input_info.buffers_num);
+	key.vertex_input.attribute_count = static_cast<uint8_t>(vs_input_info.resources_num);
+	uint32_t attributes_num          = 0;
+	for (int binding = 0; binding < vs_input_info.buffers_num; binding++) {
+		const auto& buffer = vs_input_info.buffers[binding];
+		EXIT_IF(buffer.attr_num < 0 || buffer.attr_num > ShaderVertexInputBuffer::ATTR_MAX);
+		attributes_num += static_cast<uint32_t>(buffer.attr_num);
+		EXIT_IF(attributes_num > static_cast<uint32_t>(vs_input_info.resources_num));
+		key.vertex_input.bindings[binding] = {.stride   = buffer.stride,
+		                                      .instance = buffer.fetch_index != 0};
+		for (int attribute = 0; attribute < buffer.attr_num; attribute++) {
+			const auto index = buffer.attr_indices[attribute];
+			EXIT_IF(index < 0 || index >= vs_input_info.resources_num);
+			key.vertex_input.attributes[index] = {
+			    .offset  = buffer.attr_offsets[attribute],
+			    .binding = static_cast<uint8_t>(binding),
+			};
+		}
+	}
+	EXIT_IF(attributes_num != static_cast<uint32_t>(vs_input_info.resources_num));
 
 	if (auto iter = m_graphics_pipelines.find(key); iter != m_graphics_pipelines.end()) {
 		return *iter->second;
@@ -315,16 +762,16 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 		if (ps_active) {
 			ShaderDbgDumpInputInfo(*ps_input_info);
 		}
-		LOGF("PipelineTrace: shader modules VS=%" PRIu64 " module=%p PS=%" PRIu64
-		     " module=%p\n",
+		LOGF("PipelineTrace: shader modules VS=%" PRIu64 " module=%p PS=%" PRIu64 " module=%p\n",
 		     vs_id, static_cast<void*>(vertex_program.module), ps_id,
 		     static_cast<void*>(pixel_program.module));
 	}
 
 	auto cached = std::make_unique<GraphicsPipeline>(p);
 	LogPipelineTrace("CreatePipelineInternal begin", vs_id, ps_id);
-	CreatePipelineInternal(m_graphics, *cached, rendering, vs_input_info, vertex_program.module,
-	                       ps_input_info, pixel_program.module, static_params);
+	CreatePipelineInternal(m_graphics, *cached, rendering, key.vertex_input, vs_input_info,
+	                       vertex_program.module, ps_input_info, pixel_program.module,
+	                       static_params, m_driver_cache);
 	LogPipelineTrace("CreatePipelineInternal done", vs_id, ps_id);
 
 	EXIT_NOT_IMPLEMENTED(cached->pipeline == nullptr);
@@ -360,7 +807,7 @@ PipelineCache::CreateComputePipeline(ShaderComputeInputInfo& input_info,
 	}
 
 	auto cached = std::make_unique<ComputePipeline>(p);
-	CreatePipelineInternal(m_graphics, *cached, input_info, compute_program.module);
+	CreatePipelineInternal(m_graphics, *cached, input_info, compute_program.module, m_driver_cache);
 
 	EXIT_NOT_IMPLEMENTED(cached->pipeline == nullptr);
 	EXIT_NOT_IMPLEMENTED(cached->pipeline_layout == nullptr);

src/graphics/host_gpu/renderer/pipeline/pipelineCache.h

@@ -10,6 +10,7 @@
 #include "graphics/shader/shader.h"
 
 #include <cstddef>
+#include <filesystem>
 #include <memory>
 #include <span>
 #include <type_traits>
@@ -64,10 +65,6 @@ struct PipelineStaticParameters {
 	bool                       separate_alpha_blend[RENDER_COLOR_ATTACHMENTS_MAX] = {};
 	bool                       blend_enable[RENDER_COLOR_ATTACHMENTS_MAX]         = {};
 	bool                       blend_bypass[RENDER_COLOR_ATTACHMENTS_MAX]         = {};
-	float                      blend_color_red                                    = 0.0f;
-	float                      blend_color_green                                  = 0.0f;
-	float                      blend_color_blue                                   = 0.0f;
-	float                      blend_color_alpha                                  = 0.0f;
 
 	bool operator==(const PipelineStaticParameters& other) const noexcept;
 };
@@ -84,7 +81,7 @@ static_assert(sizeof(PipelineStaticParameters) ==
                   sizeof(bool) + sizeof(PipelineStencilStaticState) * 2 + sizeof(uint32_t) +
                   sizeof(uint32_t[RENDER_COLOR_ATTACHMENTS_MAX]) + sizeof(bool) * 3 +
                   sizeof(uint8_t[RENDER_COLOR_ATTACHMENTS_MAX]) * 6 +
-                  sizeof(bool[RENDER_COLOR_ATTACHMENTS_MAX]) * 3 + sizeof(float) * 4);
+                  sizeof(bool[RENDER_COLOR_ATTACHMENTS_MAX]) * 3);
 
 struct PipelineRenderingState {
 	std::array<vk::Format, RENDER_COLOR_ATTACHMENTS_MAX> color_formats {};
@@ -95,13 +92,31 @@ struct PipelineRenderingState {
 	bool operator==(const PipelineRenderingState&) const = default;
 };
 
+struct PipelineVertexInputState {
+	struct Binding {
+		uint32_t stride                           = 0;
+		bool     instance                         = false;
+		bool     operator==(const Binding&) const = default;
+	};
+	struct Attribute {
+		uint32_t offset                             = 0;
+		uint8_t  binding                            = 0;
+		bool     operator==(const Attribute&) const = default;
+	};
+
+	std::array<Binding, ShaderVertexInputInfo::RES_MAX>   bindings {};
+	std::array<Attribute, ShaderVertexInputInfo::RES_MAX> attributes {};
+	uint8_t                                               binding_count   = 0;
+	uint8_t                                               attribute_count = 0;
+
+	bool operator==(const PipelineVertexInputState&) const = default;
+};
+
 struct ShaderProgram {
 	uint64_t         id     = 0;
 	vk::ShaderModule module = nullptr;
 
-	explicit operator bool() const {
-		return id != 0 && module != nullptr;
-	}
+	explicit operator bool() const { return id != 0 && module != nullptr; }
 };
 
 class PipelineCache {
@@ -109,6 +124,7 @@ public:
 	explicit PipelineCache(GraphicContext& graphics);
 	~PipelineCache();
 	KYTY_CLASS_NO_COPY(PipelineCache);
+	void Save();
 
 	struct Pipeline {
 		vk::PipelineLayout      pipeline_layout       = nullptr;
@@ -126,24 +142,26 @@ public:
 		uint64_t cs_shader_id = 0;
 	};
 
-	ShaderProgram GetVertexProgram(const HW::VertexShaderInfo& regs,
-	                               const HW::ShaderRegisters&  sh,
-	                               ShaderVertexInputInfo&      input_info);
-	ShaderProgram
-	GetPixelProgram(const HW::PixelShaderInfo& regs, const HW::ShaderRegisters& sh,
-	                const ShaderVertexInputInfo&                        vertex_info,
-	                std::span<const Prospero::ColorComponentMapping, 8> target_export_mapping,
-	                ShaderPixelInputInfo&                               input_info);
-	ShaderProgram GetComputeProgram(const HW::ComputeShaderInfo& regs,
-	                                const HW::ShaderRegisters& sh,
-	                                ShaderComputeInputInfo& input_info);
+	struct GraphicsPrograms {
+		ShaderProgram vertex;
+		ShaderProgram pixel;
+	};
 
-	GraphicsPipeline& CreateGraphicsPipeline(
-	    std::span<const RenderColorInfo> colors, const RenderDepthInfo& depth,
-	    const ShaderVertexInputInfo& vs_input_info, CommandBuffer& command,
-	    const ShaderPixelInputInfo* ps_input_info, vk::PrimitiveTopology topology,
-	    bool primitive_restart_enable, const ShaderProgram& vertex_program,
-	    const ShaderProgram& pixel_program);
+	GraphicsPrograms GetGraphicsPrograms(
+	    const HW::VertexShaderInfo& vertex_regs, const HW::PixelShaderInfo& pixel_regs,
+	    const HW::ShaderRegisters& sh, const HW::Context& context,
+	    std::span<const Prospero::ColorComponentMapping, 8> target_export_mapping,
+	    bool pixel_active, ShaderVertexInputInfo& vertex_info, ShaderPixelInputInfo& pixel_info);
+	ShaderProgram GetComputeProgram(const HW::ComputeShaderInfo& regs,
+	                                const HW::ShaderRegisters&   sh,
+	                                ShaderComputeInputInfo&      input_info);
+
+	GraphicsPipeline&
+	CreateGraphicsPipeline(std::span<const RenderColorInfo> colors, const RenderDepthInfo& depth,
+	                       const ShaderVertexInputInfo& vs_input_info, CommandBuffer& command,
+	                       const ShaderPixelInputInfo* ps_input_info,
+	                       vk::PrimitiveTopology topology, bool primitive_restart_enable,
+	                       const ShaderProgram& vertex_program, const ShaderProgram& pixel_program);
 	ComputePipeline& CreateComputePipeline(ShaderComputeInputInfo& input_info,
 	                                       const ShaderProgram&    compute_program);
 
@@ -154,11 +172,13 @@ private:
 		PipelineRenderingState   rendering;
 		uint64_t                 vs_shader_id = 0;
 		uint64_t                 ps_shader_id = 0;
+		PipelineVertexInputState vertex_input;
 		PipelineStaticParameters static_params;
 
 		bool operator==(const GraphicsPipelineKey& other) const {
 			return rendering == other.rendering && vs_shader_id == other.vs_shader_id &&
-			       ps_shader_id == other.ps_shader_id && static_params == other.static_params;
+			       ps_shader_id == other.ps_shader_id && vertex_input == other.vertex_input &&
+			       static_params == other.static_params;
 		}
 	};
 
@@ -199,6 +219,16 @@ private:
 			PipelineKeyHash::MixRendering(hash, key.rendering);
 			PipelineKeyHash::Mix(hash, key.vs_shader_id);
 			PipelineKeyHash::Mix(hash, key.ps_shader_id);
+			PipelineKeyHash::Mix(hash, key.vertex_input.binding_count);
+			for (uint32_t i = 0; i < key.vertex_input.binding_count; i++) {
+				PipelineKeyHash::Mix(hash, key.vertex_input.bindings[i].stride);
+				PipelineKeyHash::Mix(hash, key.vertex_input.bindings[i].instance);
+			}
+			PipelineKeyHash::Mix(hash, key.vertex_input.attribute_count);
+			for (uint32_t i = 0; i < key.vertex_input.attribute_count; i++) {
+				PipelineKeyHash::Mix(hash, key.vertex_input.attributes[i].offset);
+				PipelineKeyHash::Mix(hash, key.vertex_input.attributes[i].binding);
+			}
 			PipelineKeyHash::MixStaticParams(hash, key.static_params);
 			return hash;
 		}
@@ -212,25 +242,30 @@ private:
 		}
 	};
 
-	GraphicContext&                m_graphics;
+	GraphicContext&               m_graphics;
 	std::unique_ptr<ProgramCache> m_program_cache;
+	vk::PipelineCache             m_driver_cache = nullptr;
+	std::filesystem::path         m_driver_cache_path;
 	std::unordered_map<GraphicsPipelineKey, std::unique_ptr<GraphicsPipeline>,
 	                   GraphicsPipelineKeyHash>
 	    m_graphics_pipelines;
 	std::unordered_map<ComputePipelineKey, std::unique_ptr<ComputePipeline>, ComputePipelineKeyHash>
 	              m_compute_pipelines;
 	Common::Mutex m_mutex;
+
+	void InitializeDriverCache();
 };
 
 void LogPipelineTrace(const char* phase, uint64_t vertex_program_id, uint64_t pixel_program_id);
 void CreatePipelineInternal(
     GraphicContext& graphics, PipelineCache::GraphicsPipeline& pipeline,
-    const PipelineRenderingState& rendering, const ShaderVertexInputInfo& vs_input_info,
-    vk::ShaderModule vertex_module, const ShaderPixelInputInfo* ps_input_info,
-	vk::ShaderModule pixel_module, const PipelineStaticParameters& static_params);
+    const PipelineRenderingState& rendering, const PipelineVertexInputState& vertex_input,
+    const ShaderVertexInputInfo& vs_input_info, vk::ShaderModule vertex_module,
+    const ShaderPixelInputInfo* ps_input_info, vk::ShaderModule pixel_module,
+    const PipelineStaticParameters& static_params, vk::PipelineCache driver_cache);
 void CreatePipelineInternal(GraphicContext& graphics, PipelineCache::ComputePipeline& pipeline,
                             const ShaderComputeInputInfo& input_info,
-                            vk::ShaderModule              compute_module);
+                            vk::ShaderModule compute_module, vk::PipelineCache driver_cache);
 
 } // namespace Libs::Graphics
 

src/graphics/host_gpu/renderer/pipeline/shaderResourceBarrier.cpp

@@ -69,7 +69,7 @@ vk::BufferMemoryBarrier MakeGdsDependency(vk::Buffer buffer) {
 }
 
 std::vector<ShaderBufferWriteRange>
-CollectShaderBufferWrites(const ShaderRecompiler::IR::Program&          program,
+CollectShaderBufferWrites(const ShaderRecompiler::IR::CompiledShaderInfo& program,
                           const ShaderRecompiler::IR::ResourceSnapshot& resources) {
 	EXIT_IF(resources.buffers.size() != program.info.buffers.size());
 	std::vector<ShaderBufferWriteRange> writes;
@@ -97,7 +97,7 @@ CollectShaderBufferWrites(const ShaderRecompiler::IR::Program&          program,
 
 bool HasShaderBufferWrites(const ShaderStageRuntime& runtime) {
 	EXIT_IF(!runtime);
-	return !CollectShaderBufferWrites(*runtime.program, *runtime.resources).empty();
+	return !CollectShaderBufferWrites(*runtime.program, runtime.resources).empty();
 }
 
 void ShaderAccessBarrier(vk::CommandBuffer vk_buffer, vk::PipelineStageFlags source_stages) {

src/graphics/host_gpu/renderer/pipeline/shaderResourceBarrier.h

@@ -23,7 +23,7 @@ VulkanMemoryBarrier     MakeShaderWriteHazardDependency();
 VulkanMemoryBarrier     MakeShaderWriteDependency();
 vk::BufferMemoryBarrier MakeGdsDependency(vk::Buffer buffer);
 std::vector<ShaderBufferWriteRange>
-CollectShaderBufferWrites(const ShaderRecompiler::IR::Program&          program,
+CollectShaderBufferWrites(const ShaderRecompiler::IR::CompiledShaderInfo& program,
                           const ShaderRecompiler::IR::ResourceSnapshot& resources);
 bool HasShaderBufferWrites(const ShaderStageRuntime& runtime);
 void ShaderAccessBarrier(vk::CommandBuffer vk_buffer, vk::PipelineStageFlags source_stages);

src/graphics/host_gpu/renderer/pipeline/shaders.cpp

@@ -390,7 +390,7 @@ static vk::BlendOp GetBlendOp(uint32_t op) {
 }
 
 static void AddLayoutBindings(std::vector<vk::DescriptorSetLayoutBinding>& descriptor_bindings,
-                              const ShaderRecompiler::IR::Program& program,
+                              const ShaderRecompiler::IR::CompiledShaderInfo& program,
                               vk::ShaderStageFlagBits              stage) {
 	for (const auto& binding: program.bindings.descriptors) {
 		descriptor_bindings.push_back(
@@ -421,9 +421,10 @@ static void CreateDescriptorLayout(GraphicContext& graphics, PipelineCache::Pipe
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
 void CreatePipelineInternal(
     GraphicContext& graphics, PipelineCache::GraphicsPipeline& pipeline,
-    const PipelineRenderingState& rendering, const ShaderVertexInputInfo& vs_input_info,
-    vk::ShaderModule vertex_module, const ShaderPixelInputInfo* ps_input_info,
-	vk::ShaderModule pixel_module, const PipelineStaticParameters& static_params) {
+    const PipelineRenderingState& rendering, const PipelineVertexInputState& vertex_input,
+    const ShaderVertexInputInfo& vs_input_info, vk::ShaderModule vertex_module,
+    const ShaderPixelInputInfo* ps_input_info, vk::ShaderModule pixel_module,
+    const PipelineStaticParameters& static_params, vk::PipelineCache driver_cache) {
 	const bool ps_active = ps_input_info != nullptr;
 	EXIT_IF(vertex_module == nullptr || (ps_active && pixel_module == nullptr));
 
@@ -504,135 +505,131 @@ void CreatePipelineInternal(
 		shader_stages[shader_stage_count++] = frag_shader_stage_info;
 	}
 
-	vk::VertexInputAttributeDescription input_attr[ShaderVertexInputInfo::RES_MAX];
-	vk::VertexInputBindingDescription   input_desc[ShaderVertexInputInfo::RES_MAX];
+	vk::VertexInputAttributeDescription input_attr[ShaderVertexInputInfo::RES_MAX] {};
+	vk::VertexInputBindingDescription   input_desc[ShaderVertexInputInfo::RES_MAX] {};
 
-	for (int bi = 0; bi < vs_input_info.buffers_num; bi++) {
-		const auto& b          = vs_input_info.buffers[bi];
-		input_desc[bi].binding = bi;
-		input_desc[bi].stride  = b.stride;
-		input_desc[bi].inputRate =
-		    (b.fetch_index == 0 ? vk::VertexInputRate::eVertex : vk::VertexInputRate::eInstance);
+	for (uint32_t binding = 0; binding < vertex_input.binding_count; binding++) {
+		input_desc[binding].binding   = binding;
+		input_desc[binding].stride    = vertex_input.bindings[binding].stride;
+		input_desc[binding].inputRate = vertex_input.bindings[binding].instance
+		                                    ? vk::VertexInputRate::eInstance
+		                                    : vk::VertexInputRate::eVertex;
+	}
+	for (uint32_t index = 0; index < vertex_input.attribute_count; index++) {
+		input_attr[index].binding  = vertex_input.attributes[index].binding;
+		input_attr[index].location = index;
+		input_attr[index].offset   = vertex_input.attributes[index].offset;
 
-		for (int ai = 0; ai < b.attr_num; ai++) {
-			auto index                 = b.attr_indices[ai];
-			input_attr[index].binding  = bi;
-			input_attr[index].location = index;
-			input_attr[index].offset   = b.attr_offsets[ai];
+		uint32_t   attr_size     = 4;
+		const auto registers_num = vs_input_info.resources_dst[index].registers_num;
+		const auto compiled_components =
+		    vs_input_info.stage.program->info.vertex_fetch_components[index];
+		const auto used_components =
+		    compiled_components > 0 ? static_cast<int>(compiled_components) : registers_num;
+		GetInputFormat(vs_input_info.resources[index], input_attr[index].format, attr_size,
+		               static_cast<uint32_t>(used_components));
 
-			uint32_t attr_size       = 4;
-			auto     registers_num   = vs_input_info.resources_dst[index].registers_num;
-			const auto compiled_components =
-			    vs_input_info.stage.program->info.vertex_fetch_components[index];
-			auto used_components =
-			    compiled_components > 0 ? static_cast<int>(compiled_components) : registers_num;
-			GetInputFormat(vs_input_info.resources[index], input_attr[index].format, attr_size,
-			               static_cast<uint32_t>(used_components));
-
-			if (graphics_debug_dump_enabled()) {
-				static std::atomic_uint log_count = 0;
-				const auto              log_id = log_count.fetch_add(1, std::memory_order_relaxed);
-				if (log_id < 128) {
-					LOGF("VertexInputState[%u]: attr=%u binding=%u offset=%u stride=%u fmt=%d "
-					     "src_fmt=%u dst=v%u regs=%u"
-					     " fetched_components=%u attr_size=%u swizzle=%u,%u,%u,%u\n",
-					     log_id, index, input_attr[index].binding, input_attr[index].offset,
-					     input_desc[bi].stride, static_cast<int>(input_attr[index].format),
-					     static_cast<uint32_t>(vs_input_info.resources[index].Format()),
-					     static_cast<uint32_t>(vs_input_info.resources_dst[index].register_start),
-					     static_cast<uint32_t>(registers_num),
-					     static_cast<uint32_t>(used_components), attr_size,
-					     static_cast<uint32_t>(vs_input_info.resources[index].DstSelX()),
-					     static_cast<uint32_t>(vs_input_info.resources[index].DstSelY()),
-					     static_cast<uint32_t>(vs_input_info.resources[index].DstSelZ()),
-					     static_cast<uint32_t>(vs_input_info.resources[index].DstSelW()));
-				}
+		if (graphics_debug_dump_enabled()) {
+			static std::atomic_uint log_count = 0;
+			const auto              log_id    = log_count.fetch_add(1, std::memory_order_relaxed);
+			if (log_id < 128) {
+				LOGF("VertexInputState[%u]: attr=%u binding=%u offset=%u stride=%u fmt=%d "
+				     "src_fmt=%u dst=v%u regs=%u"
+				     " fetched_components=%u attr_size=%u swizzle=%u,%u,%u,%u\n",
+				     log_id, index, input_attr[index].binding, input_attr[index].offset,
+				     input_desc[input_attr[index].binding].stride,
+				     static_cast<int>(input_attr[index].format),
+				     static_cast<uint32_t>(vs_input_info.resources[index].Format()),
+				     static_cast<uint32_t>(vs_input_info.resources_dst[index].register_start),
+				     static_cast<uint32_t>(registers_num), static_cast<uint32_t>(used_components),
+				     attr_size, static_cast<uint32_t>(vs_input_info.resources[index].DstSelX()),
+				     static_cast<uint32_t>(vs_input_info.resources[index].DstSelY()),
+				     static_cast<uint32_t>(vs_input_info.resources[index].DstSelZ()),
+				     static_cast<uint32_t>(vs_input_info.resources[index].DstSelW()));
 			}
+		}
 
-			if (vs_input_info.resources[index].OutOfBounds() != 0) {
-				static bool logged = false;
-				if (!logged) {
-					LOGF("VertexInput: temporary: accepting PS5 out-of-bounds behavior %" PRIu8
-					     "\n",
-					     vs_input_info.resources[index].OutOfBounds());
-					logged = true;
-				}
+		if (vs_input_info.resources[index].OutOfBounds() != 0) {
+			static bool logged = false;
+			if (!logged) {
+				LOGF("VertexInput: temporary: accepting PS5 out-of-bounds behavior %" PRIu8 "\n",
+				     vs_input_info.resources[index].OutOfBounds());
+				logged = true;
 			}
+		}
 
-			EXIT_NOT_IMPLEMENTED(vs_input_info.resources[index].AddTid());
-			EXIT_NOT_IMPLEMENTED(vs_input_info.resources[index].SwizzleEnabled());
+		EXIT_NOT_IMPLEMENTED(vs_input_info.resources[index].AddTid());
+		EXIT_NOT_IMPLEMENTED(vs_input_info.resources[index].SwizzleEnabled());
 
-			auto log_unsupported_vertex_swizzle = [index, attr_size, registers_num](
-			                                          uint32_t swizzle, uint32_t expected) {
-				static bool logged = false;
-				if (!logged) {
-					LOGF(
-					    "VertexInput: temporary: accepting unsupported dst swizzle at attr %" PRIu32
-					    " (attr_size=%" PRIu32 ", regs=%" PRIu32 ", swizzle=0x%03" PRIx32
-					    ", expected=0x%03" PRIx32 ")\n",
-					    static_cast<uint32_t>(index), attr_size, registers_num, swizzle, expected);
-					logged = true;
-				}
-			};
-
-			switch (registers_num) {
-				case 1: {
-					auto swizzle = vs_input_info.resources[index].DstSelX();
-					if (swizzle != DstSel(4)) {
-						log_unsupported_vertex_swizzle(swizzle, DstSel(4));
-					}
-					break;
-				}
-				case 2: {
-					auto swizzle  = vs_input_info.resources[index].DstSelXY();
-					auto expected = (attr_size == 1 ? DstSel(4, 0) : DstSel(4, 5));
-					if (swizzle != expected) {
-						log_unsupported_vertex_swizzle(swizzle, expected);
-					}
-					break;
-				}
-				case 3: {
-					auto swizzle  = vs_input_info.resources[index].DstSelXYZ();
-					auto expected = DstSel(4, 5, 6);
-					switch (attr_size) {
-						case 1: expected = DstSel(4, 0, 0); break;
-						case 2: expected = DstSel(4, 5, 0); break;
-						default: break;
-					}
-					if (swizzle != expected) {
-						log_unsupported_vertex_swizzle(swizzle, expected);
-					}
-					break;
-				}
-				case 4: {
-					auto swizzle   = vs_input_info.resources[index].DstSelXYZW();
-					auto expected  = DstSel(4, 5, 6, 7);
-					bool supported = false;
-					switch (attr_size) {
-						case 1:
-							expected  = DstSel(4, 0, 0, 1);
-							supported = (swizzle == expected);
-							break;
-						case 2:
-							expected  = DstSel(4, 5, 0, 1);
-							supported = (swizzle == expected);
-							break;
-						case 3:
-							expected  = DstSel(4, 5, 6, 1);
-							supported = (swizzle == expected || swizzle == DstSel(4, 5, 6, 0));
-							break;
-						default:
-							supported = (swizzle == expected || swizzle == DstSel(4, 5, 6, 1) ||
-							             swizzle == DstSel(4, 5, 6, 0));
-							break;
-					}
-					if (!supported) {
-						log_unsupported_vertex_swizzle(swizzle, expected);
-					}
-					break;
-				}
-				default: EXIT("invalid registers_num");
+		auto log_unsupported_vertex_swizzle = [index, attr_size, registers_num](uint32_t swizzle,
+		                                                                        uint32_t expected) {
+			static bool logged = false;
+			if (!logged) {
+				LOGF("VertexInput: temporary: accepting unsupported dst swizzle at attr %" PRIu32
+				     " (attr_size=%" PRIu32 ", regs=%" PRIu32 ", swizzle=0x%03" PRIx32
+				     ", expected=0x%03" PRIx32 ")\n",
+				     static_cast<uint32_t>(index), attr_size, registers_num, swizzle, expected);
+				logged = true;
 			}
+		};
+
+		switch (registers_num) {
+			case 1: {
+				auto swizzle = vs_input_info.resources[index].DstSelX();
+				if (swizzle != DstSel(4)) {
+					log_unsupported_vertex_swizzle(swizzle, DstSel(4));
+				}
+				break;
+			}
+			case 2: {
+				auto swizzle  = vs_input_info.resources[index].DstSelXY();
+				auto expected = (attr_size == 1 ? DstSel(4, 0) : DstSel(4, 5));
+				if (swizzle != expected) {
+					log_unsupported_vertex_swizzle(swizzle, expected);
+				}
+				break;
+			}
+			case 3: {
+				auto swizzle  = vs_input_info.resources[index].DstSelXYZ();
+				auto expected = DstSel(4, 5, 6);
+				switch (attr_size) {
+					case 1: expected = DstSel(4, 0, 0); break;
+					case 2: expected = DstSel(4, 5, 0); break;
+					default: break;
+				}
+				if (swizzle != expected) {
+					log_unsupported_vertex_swizzle(swizzle, expected);
+				}
+				break;
+			}
+			case 4: {
+				auto swizzle   = vs_input_info.resources[index].DstSelXYZW();
+				auto expected  = DstSel(4, 5, 6, 7);
+				bool supported = false;
+				switch (attr_size) {
+					case 1:
+						expected  = DstSel(4, 0, 0, 1);
+						supported = (swizzle == expected);
+						break;
+					case 2:
+						expected  = DstSel(4, 5, 0, 1);
+						supported = (swizzle == expected);
+						break;
+					case 3:
+						expected  = DstSel(4, 5, 6, 1);
+						supported = (swizzle == expected || swizzle == DstSel(4, 5, 6, 0));
+						break;
+					default:
+						supported = (swizzle == expected || swizzle == DstSel(4, 5, 6, 1) ||
+						             swizzle == DstSel(4, 5, 6, 0));
+						break;
+				}
+				if (!supported) {
+					log_unsupported_vertex_swizzle(swizzle, expected);
+				}
+				break;
+			}
+			default: EXIT("invalid registers_num");
 		}
 	}
 
@@ -640,9 +637,9 @@ void CreatePipelineInternal(
 	vertex_input_info.sType = vk::StructureType::ePipelineVertexInputStateCreateInfo;
 	vertex_input_info.pNext = nullptr;
 	vertex_input_info.flags = {};
-	vertex_input_info.vertexBindingDescriptionCount   = vs_input_info.buffers_num;
+	vertex_input_info.vertexBindingDescriptionCount   = vertex_input.binding_count;
 	vertex_input_info.pVertexBindingDescriptions      = input_desc;
-	vertex_input_info.vertexAttributeDescriptionCount = vs_input_info.resources_num;
+	vertex_input_info.vertexAttributeDescriptionCount = vertex_input.attribute_count;
 	vertex_input_info.pVertexAttributeDescriptions    = input_attr;
 
 	vk::PipelineInputAssemblyStateCreateInfo input_assembly {};
@@ -791,15 +788,11 @@ void CreatePipelineInternal(
 #else
 	color_blending.pNext = &color_write;
 #endif
-	color_blending.flags             = {};
-	color_blending.logicOpEnable     = VK_FALSE;
-	color_blending.logicOp           = vk::LogicOp::eCopy;
-	color_blending.attachmentCount   = static_params.color_count;
-	color_blending.pAttachments      = color_blend_attachment;
-	color_blending.blendConstants[0] = static_params.blend_color_red;
-	color_blending.blendConstants[1] = static_params.blend_color_green;
-	color_blending.blendConstants[2] = static_params.blend_color_blue;
-	color_blending.blendConstants[3] = static_params.blend_color_alpha;
+	color_blending.flags           = {};
+	color_blending.logicOpEnable   = VK_FALSE;
+	color_blending.logicOp         = vk::LogicOp::eCopy;
+	color_blending.attachmentCount = static_params.color_count;
+	color_blending.pAttachments    = color_blend_attachment;
 
 	EXIT_IF(!vs_input_info.stage);
 	std::vector<vk::DescriptorSetLayoutBinding> descriptor_bindings;
@@ -813,8 +806,8 @@ void CreatePipelineInternal(
 	CreateDescriptorLayout(graphics, pipeline, descriptor_bindings);
 	constexpr auto GraphicsStages =
 	    vk::ShaderStageFlagBits::eVertex | vk::ShaderStageFlagBits::eFragment;
-	const vk::PushConstantRange push_constants {
-	    GraphicsStages, 0, ShaderRecompiler::IR::NativePushConstantSize};
+	const vk::PushConstantRange push_constants {GraphicsStages, 0,
+	                                            ShaderRecompiler::IR::NativePushConstantSize};
 
 	vk::PipelineLayoutCreateInfo pipeline_layout_info {};
 	pipeline_layout_info.sType                  = vk::StructureType::ePipelineLayoutCreateInfo;
@@ -876,6 +869,7 @@ void CreatePipelineInternal(
 	    vk::DynamicState::eStencilCompareMask,
 	    vk::DynamicState::eStencilReference,
 	    vk::DynamicState::eStencilWriteMask,
+	    vk::DynamicState::eBlendConstants,
 #if !defined(__APPLE__)
 	    // Keep last so depth-only pipelines can omit this dynamic state.
 	    vk::DynamicState::eColorWriteEnableEXT, // unsupported by MoltenVK; static mask instead
@@ -940,7 +934,7 @@ void CreatePipelineInternal(
 		     viewport.y, viewport.width, viewport.height, scissor.offset.x, scissor.offset.y,
 		     scissor.extent.width, scissor.extent.height);
 	}
-	result = graphics.device.createGraphicsPipelines(nullptr, 1, &pipeline_info, nullptr,
+	result = graphics.device.createGraphicsPipelines(driver_cache, 1, &pipeline_info, nullptr,
 	                                                 &pipeline.pipeline);
 	if (graphics_debug_dump_enabled()) {
 		LOGF("PipelineTrace: vkCreateGraphicsPipelines done result=%s pipeline=%p\n",
@@ -961,7 +955,7 @@ void CreatePipelineInternal(
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
 void CreatePipelineInternal(GraphicContext& graphics, PipelineCache::ComputePipeline& pipeline,
                             const ShaderComputeInputInfo& input_info,
-                            vk::ShaderModule              compute_module) {
+	vk::ShaderModule compute_module, vk::PipelineCache driver_cache) {
 	EXIT_IF(compute_module == nullptr);
 
 	vk::PipelineShaderStageCreateInfo                     comp_shader_stage_info {};
@@ -1024,7 +1018,8 @@ void CreatePipelineInternal(GraphicContext& graphics, PipelineCache::ComputePipe
 
 	LOGF("PipelineTrace: vkCreateComputePipelines begin layout=%p\n",
 	     static_cast<void*>(pipeline.pipeline_layout));
-	result = graphics.device.createComputePipelines(nullptr, 1, &info, nullptr, &pipeline.pipeline);
+	result = graphics.device.createComputePipelines(driver_cache, 1, &info, nullptr,
+	                                                &pipeline.pipeline);
 	LOGF("PipelineTrace: vkCreateComputePipelines done result=%s pipeline=%p\n",
 	     VulkanToString(result).c_str(), static_cast<void*>(pipeline.pipeline));
 	EXIT_NOT_IMPLEMENTED(result != vk::Result::eSuccess);

src/graphics/host_gpu/renderer/render.h

@@ -48,6 +48,31 @@ enum class CommandBufferDebugOp : uint32_t {
 	Unknown,
 };
 
+enum class DrawOffsetSource : uint8_t {
+	DrawState,
+	IndirectArgs,
+};
+
+struct DrawIndexArgs {
+	uint32_t         index_count                = 0;
+	const void*      index_addr                 = nullptr;
+	uint32_t         instance_count             = 0;
+	uint32_t         index_type_and_size        = 0;
+	int32_t          base_vertex                = 0;
+	uint32_t         first_instance             = 0;
+	DrawOffsetSource offset_source              = DrawOffsetSource::DrawState;
+	uint32_t         render_target_slice_offset = 0;
+};
+
+struct DrawAutoArgs {
+	uint32_t         vertex_count               = 0;
+	uint32_t         instance_count             = 0;
+	uint32_t         first_vertex               = 0;
+	uint32_t         first_instance             = 0;
+	DrawOffsetSource offset_source              = DrawOffsetSource::DrawState;
+	uint32_t         render_target_slice_offset = 0;
+};
+
 struct SubmitInfo {
 	static constexpr uint32_t MaxSemaphores = 3;
 
@@ -129,13 +154,6 @@ public:
 	explicit RenderExecutor(RenderContext& context): m_context(context) {}
 	KYTY_CLASS_NO_COPY(RenderExecutor);
 
-	void DrawIndex(uint64_t submit_id, CommandBuffer& buffer, uint32_t index_type_and_size,
-	               uint32_t index_count, const void* index_addr, uint32_t flags, uint32_t type,
-	               uint32_t instance_count = 1, uint32_t render_target_slice_offset = 0,
-	               int32_t vertex_offset_add = 0, uint32_t first_instance = 0);
-	void DrawAuto(uint64_t submit_id, CommandBuffer& buffer, uint32_t index_count, uint32_t flags,
-	              uint32_t render_target_slice_offset = 0, uint32_t instance_count = 1,
-	              uint32_t first_vertex = 0, uint32_t first_instance = 0);
 	void DispatchDirect(uint64_t submit_id, CommandBuffer& buffer, uint32_t thread_group_x,
 	                    uint32_t thread_group_y, uint32_t thread_group_z, uint32_t mode);
 
@@ -148,6 +166,9 @@ public:
 	                    std::span<PreparedBindings* const> bindings);
 
 private:
+	void DrawIndex(uint64_t submit_id, CommandBuffer& buffer, const DrawIndexArgs& args);
+	void DrawAuto(uint64_t submit_id, CommandBuffer& buffer, const DrawAutoArgs& args);
+
 	struct GraphicsBindings {
 		PreparedBindings                vertex;
 		std::optional<PreparedBindings> pixel;
@@ -190,9 +211,8 @@ private:
 	std::vector<vk::DescriptorImageInfo>  m_descriptor_images;
 	std::vector<vk::WriteDescriptorSet>   m_descriptor_writes;
 	std::vector<uint32_t>                 m_image_occurrences;
-	std::array<uint32_t, ShaderRecompiler::IR::NativePushConstantSize / sizeof(uint32_t)>
-	    m_push_constants {};
 
+	friend class CommandProcessor;
 	friend struct RenderExecutorTestAccess;
 };
 

src/graphics/host_gpu/renderer/renderCompute.cpp

@@ -46,7 +46,7 @@ static uint64_t BufferDescriptorSize(const ShaderBufferResource& descriptor) {
 bool RenderExecutor::TryConsumeComputeMetaClear(const ShaderComputeInputInfo& input,
                                                 const CommandBuffer&          buffer) {
 	const auto& program   = *input.stage.program;
-	const auto& resources = *input.stage.resources;
+	const auto& resources = input.stage.resources;
 	if (resources.buffers.size() != program.info.buffers.size()) {
 		EXIT("compute runtime buffer count does not match shader metadata\n");
 	}
@@ -54,7 +54,9 @@ bool RenderExecutor::TryConsumeComputeMetaClear(const ShaderComputeInputInfo& in
 	for (uint32_t i = 0; i < program.info.buffers.size(); i++) {
 		const auto& resource   = program.info.buffers[i];
 		const auto  descriptor = DecodeNativeDescriptor<ShaderBufferResource>(resources.buffers[i]);
-		if (!resource.written && cache.IsMeta(descriptor.Base48())) {
+		// A metadata resource that is also read is not proven to be a full overwrite. Execute it
+		// conservatively instead of replacing the dispatch with a coarse full-surface clear.
+		if (cache.IsMeta(descriptor.Base48()) && (!resource.written || resource.read)) {
 			return false;
 		}
 	}
@@ -79,7 +81,7 @@ bool ResolveComputeImageClear(const ShaderComputeInputInfo& input, uint32_t grou
                               ShaderBufferResource& resolved_descriptor, uint32_t& resolved_clear,
                               uint64_t& resolved_size) {
 	const auto& program   = *input.stage.program;
-	const auto& resources = *input.stage.resources;
+	const auto& resources = input.stage.resources;
 	if (program.info.buffers.size() != 1 || resources.buffers.size() != 1 ||
 	    !program.info.images.empty() || !program.info.samplers.empty() || program.info.uses_dma ||
 	    !resources.images.empty() || !resources.samplers.empty()) {
@@ -222,7 +224,7 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 	const bool     large_workgroup =
 	    (input_info.threads_num[0] * input_info.threads_num[1] * input_info.threads_num[2] >= 512);
 	const auto& program   = *input_info.stage.program;
-	const auto& resources = *input_info.stage.resources;
+	const auto& resources = input_info.stage.resources;
 	if (TryConsumeComputeMetaClear(input_info, buffer)) {
 		ResetBindings();
 		return;
@@ -234,8 +236,7 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 	}
 	const auto sampled_images = std::count_if(
 	    program.info.images.begin(), program.info.images.end(), [](const auto& image) {
-		    return image.kind == ShaderRecompiler::IR::ResourceKind::Image ||
-		           image.kind == ShaderRecompiler::IR::ResourceKind::ImageUint;
+		    return image.resource_class == ShaderRecompiler::IR::ImageResourceClass::Sampled;
 	    });
 	const bool                   has_sampler = !program.info.samplers.empty();
 	static std::atomic<uint32_t> dispatch_log_count {0};
@@ -248,7 +249,10 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 		     thread_group_z, mode, input_info.threads_num[0], input_info.threads_num[1],
 		     input_info.threads_num[2], program.info.buffers.size(), program.info.images.size(),
 		     sampled_images, program.info.images.size() - sampled_images,
-		     program.info.samplers.size(), program.bindings.push_constant_size);
+		     program.info.samplers.size(),
+		     program.bindings.UsesPushData()
+		         ? static_cast<uint32_t>(sizeof(ShaderRecompiler::IR::PushData))
+		         : 0u);
 		for (uint32_t i = 0; i < program.info.buffers.size(); i++) {
 			const auto& buffer = program.info.buffers[i];
 			const auto  r      = DecodeNativeDescriptor<ShaderBufferResource>(resources.buffers[i]);
@@ -263,8 +267,7 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 			LOGF("  CS texture[%u]: source=%u usage=%s sampled=%s addr=0x%010" PRIx64
 			     " type=%u fmt=%u extent=%ux%u depth=%u levels=%u tile=%u\n",
 			     i, image.source, image.written ? "read-write" : "read-only",
-			     (image.kind == ShaderRecompiler::IR::ResourceKind::Image ||
-			      image.kind == ShaderRecompiler::IR::ResourceKind::ImageUint)
+			     image.resource_class == ShaderRecompiler::IR::ImageResourceClass::Sampled
 			         ? "true"
 			         : "false",
 			     r.Base40(), static_cast<uint32_t>(r.Type()), static_cast<uint32_t>(r.Format()),
@@ -345,8 +348,8 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 	    std::any_of(program.info.images.begin(), program.info.images.end(),
 	                [](const auto& image) {
 		                return image.written &&
-		                       (image.kind == ShaderRecompiler::IR::ResourceKind::StorageImage ||
-		                        image.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint);
+		                       image.resource_class ==
+		                           ShaderRecompiler::IR::ImageResourceClass::Storage;
 	                }) ||
 	    has_storage_writes;
 	if (has_storage_writes) {

src/graphics/host_gpu/renderer/renderDraw.cpp

@@ -52,7 +52,7 @@ int32_t ResolveVertexOffset(uint32_t index_offset, const ShaderVertexInputInfo&
 
 	EXIT_IF(!vs_input_info.stage);
 	const auto& program   = *vs_input_info.stage.program;
-	const auto& resources = *vs_input_info.stage.resources;
+	const auto& resources = vs_input_info.stage.resources;
 	if (program.info.vertex_offset_sgpr >= static_cast<int32_t>(program.user_data_base)) {
 		const auto index =
 		    static_cast<uint32_t>(program.info.vertex_offset_sgpr) - program.user_data_base;
@@ -64,6 +64,25 @@ int32_t ResolveVertexOffset(uint32_t index_offset, const ShaderVertexInputInfo&
 	return 0;
 }
 
+uint32_t ResolveInstanceOffset(const ShaderVertexInputInfo& vs_input_info) {
+	if (!vs_input_info.fetch_embedded) {
+		return 0;
+	}
+
+	EXIT_IF(!vs_input_info.stage);
+	const auto& program   = *vs_input_info.stage.program;
+	const auto& resources = vs_input_info.stage.resources;
+	if (program.info.instance_offset_sgpr >= static_cast<int32_t>(program.user_data_base)) {
+		const auto index =
+		    static_cast<uint32_t>(program.info.instance_offset_sgpr) - program.user_data_base;
+		if (index < resources.user_data.size()) {
+			return resources.user_data[index];
+		}
+	}
+
+	return 0;
+}
+
 static std::atomic<uint32_t> g_draw_state_log_count   = 0;
 static std::atomic<uint32_t> g_draw_input_log_count   = 0;
 static std::atomic<uint32_t> g_mrt_state_log_count    = 0;
@@ -98,10 +117,6 @@ static const char* RenderColorTypeName(RenderColorType type) {
 	}
 }
 
-static bool IsDualSourceBlendFactor(uint32_t factor) {
-	return factor >= 0x0fu && factor <= 0x12u;
-}
-
 static void LogFramebufferSkip(const char* draw_name, const RenderColorInfo& color,
                                const RenderDepthInfo& depth, const CommandBuffer& buffer,
                                uint32_t index_count, uint32_t flags) {
@@ -134,12 +149,6 @@ static void LogMrtState(const char* draw_name, const CommandBuffer& buffer,
 	const auto  cb_shader_mask = sh_regs.m_cbShaderMask;
 	const auto& bc0            = ctx.GetBlendControl(0);
 
-	bool interesting = rt_mask != 0x0f || (cb_shader_mask & ~0x0fu) != 0 ||
-	                   IsDualSourceBlendFactor(bc0.color_srcblend) ||
-	                   IsDualSourceBlendFactor(bc0.color_destblend) ||
-	                   (bc0.separate_alpha_blend && (IsDualSourceBlendFactor(bc0.alpha_srcblend) ||
-	                                                 IsDualSourceBlendFactor(bc0.alpha_destblend)));
-
 	auto log_id = g_mrt_state_log_count.fetch_add(1);
 	if (log_id >= 32) {
 		return;
@@ -199,8 +208,7 @@ static void LogDrawTargetState(const char* draw_name, const RenderColorInfo& col
 	const auto& ps_resources   = ps_input_info.stage.program->info;
 	const auto  sampled_images = std::count_if(
 	    ps_resources.images.begin(), ps_resources.images.end(), [](const auto& image) {
-		    return image.kind == ShaderRecompiler::IR::ResourceKind::Image ||
-		           image.kind == ShaderRecompiler::IR::ResourceKind::ImageUint;
+		    return image.resource_class == ShaderRecompiler::IR::ImageResourceClass::Sampled;
 	    });
 
 	vk::Extent2D extent = color.image_id ? color.extent : vk::Extent2D {};
@@ -331,10 +339,18 @@ static void SetGraphicsDynamicParams(const CommandBuffer& buffer, vk::CommandBuf
 	const auto final_scissor = calc_final_scissor(vp, ctx.GetScanModeControl(), framebuffer_extent);
 
 	vk::Viewport viewport {};
-	viewport.x        = vp.viewports[0].xoffset - vp.viewports[0].xscale;
-	viewport.y        = vp.viewports[0].yoffset - vp.viewports[0].yscale;
-	viewport.width    = vp.viewports[0].xscale * 2.0f;
-	viewport.height   = vp.viewports[0].yscale * 2.0f;
+	if (ctx.GetClipControl().clip_disable) {
+		const auto& limits = buffer.GetGraphics().GetPhysicalDeviceProperties().limits;
+		viewport.x         = 0.0f;
+		viewport.y         = 0.0f;
+		viewport.width     = static_cast<float>(std::min(limits.maxViewportDimensions[0], 16384u));
+		viewport.height    = static_cast<float>(std::min(limits.maxViewportDimensions[1], 16384u));
+	} else {
+		viewport.x      = vp.viewports[0].xoffset - vp.viewports[0].xscale;
+		viewport.y      = vp.viewports[0].yoffset - vp.viewports[0].yscale;
+		viewport.width  = vp.viewports[0].xscale * 2.0f;
+		viewport.height = vp.viewports[0].yscale * 2.0f;
+	}
 	viewport.minDepth = vp.viewports[0].zoffset;
 	viewport.maxDepth = vp.viewports[0].zscale + vp.viewports[0].zoffset;
 	vk_buffer.setViewport(0, 1, &viewport);
@@ -357,6 +373,9 @@ static void SetGraphicsDynamicParams(const CommandBuffer& buffer, vk::CommandBuf
 		line_width = 1.0f;
 	}
 	vk_buffer.setLineWidth(line_width);
+	const auto&      blend = ctx.GetBlendColor();
+	const std::array blend_constants {blend.red, blend.green, blend.blue, blend.alpha};
+	vk_buffer.setBlendConstants(blend_constants.data());
 
 	const auto& mode              = ctx.GetModeControl();
 	const auto& poly_offset       = ctx.GetPolyOffset();
@@ -411,7 +430,7 @@ static void SetGraphicsDynamicParams(const CommandBuffer& buffer, vk::CommandBuf
 static bool DrawHasValidVertexShader(const HW::Shader& sh_ctx) {
 
 	const auto& vs = sh_ctx.GetVs();
-	return vs.gs_regs.chksum != 0 && ShaderAddressValid(vs.es_regs.data_addr);
+	return ShaderAddressValid(vs.es_regs.data_addr);
 }
 
 static bool PixelShaderHasDepthOrCoverageSideEffects(const HW::ShaderRegisters& sh_regs) {
@@ -443,22 +462,20 @@ static bool ShouldSkipGeShader(const CommandBuffer& buffer) {
 	};
 
 	const bool ps5_ngg_vertex_path = stages == 0x02002000 && vertex_info.es_regs.data_addr != 0 &&
-	                                 vertex_info.gs_regs.chksum != 0 &&
 	                                 sh_regs.m_vgtGsMaxVertOut == 0x00000000 &&
 	                                 is_known_gs_out_prim_type(sh_regs.m_vgtGsOutPrimType);
 
 	const bool unsupported_stage_mask = (stages != 0 && stages != 0x02002000);
 	const bool unsupported_gs_stage = (vertex_info.es_regs.data_addr != 0 &&
 	                                   vertex_info.gs_regs.data_addr != 0 && !ps5_ngg_vertex_path);
-	const bool ge_group_size =
-	    ge_cntl.primitive_group_size > 0x0040 || ge_cntl.vertex_group_size > 0x0040;
+	// GE_CNTL group sizes control guest scheduling and do not constrain the host vertex path.
 	const bool ge_shader_regs =
 	    (sh_regs.m_geNggSubgrpCntl != 0x00000000 && sh_regs.m_geNggSubgrpCntl != 0x00000001) ||
 	    sh_regs.m_vgtGsMaxVertOut != 0x00000000 ||
 	    !is_known_gs_out_prim_type(sh_regs.m_vgtGsOutPrimType) ||
 	    sh_regs.m_geMaxOutputPerSubgroup > 0x00000040;
 
-	if (unsupported_stage_mask || unsupported_gs_stage || ge_group_size || ge_shader_regs) {
+	if (unsupported_stage_mask || unsupported_gs_stage || ge_shader_regs) {
 		static std::once_flag warning_once;
 		std::call_once(warning_once, [] {
 			std::printf("Warning: game uses unsupported graphics pipelines; some draw calls were "
@@ -490,15 +507,13 @@ struct DrawRenderState {
 	RenderState           rendering;
 	ShaderVertexInputInfo vs_input_info;
 	ShaderPixelInputInfo  ps_input_info;
-	ShaderProgram         vertex_program;
-	ShaderProgram         pixel_program;
+	PipelineCache::GraphicsPrograms programs;
 };
 
 struct DrawCallInfo {
 	const char*          name           = nullptr;
 	CommandBufferDebugOp debug_op       = CommandBufferDebugOp::DrawIndex;
 	uint32_t             index_count    = 0;
-	uint32_t             flags          = 0;
 	uint32_t             instance_count = 0;
 	uint32_t             first_instance = 0;
 };
@@ -740,6 +755,7 @@ struct DrawEmitInfo {
 	bool     indexed       = false;
 	int32_t  vertex_offset = 0;
 	uint32_t first_vertex  = 0;
+	uint32_t first_instance = 0;
 };
 
 struct DrawIndexBufferSource {
@@ -873,8 +889,8 @@ static void SetDrawDebugPhase(CommandBuffer& buffer, uint64_t submit_id, const D
                               uint32_t phase) {
 	EXIT_IF(draw.name == nullptr);
 
-	buffer.SetDebugInfo(static_cast<uint32_t>(draw.debug_op), submit_id, phase, draw.index_count,
-	                    draw.flags, draw.instance_count, draw.first_instance);
+	buffer.SetDebugInfo(static_cast<uint32_t>(draw.debug_op), submit_id, phase, draw.index_count, 0,
+	                    draw.instance_count, draw.first_instance);
 }
 
 static bool GetDrawTopology(const HW::UserConfig& ucfg, bool auto_draw,
@@ -986,7 +1002,7 @@ bool RenderExecutor::PrepareDrawRenderState(uint64_t submit_id, CommandBuffer& b
 	                         static_cast<bool>(state.depth_info.image_id));
 	if (state.color_count == 0 && !with_depth && !state.ps_active) {
 		LogFramebufferSkip(draw.name, state.color_info[0], state.depth_info, buffer,
-		                   draw.index_count, draw.flags);
+		                   draw.index_count, 0);
 		return false;
 	}
 
@@ -1003,30 +1019,20 @@ static void RefreshShaders(CommandBuffer& buffer, const DrawCallInfo& draw, bool
 	const auto& pixel_shader_info  = sh_ctx.GetPs();
 	const auto& shader_regs        = ctx.GetShaderRegisters();
 
-	state.vertex_program = {};
-	state.pixel_program  = {};
-	state.ps_input_info  = {};
+	state.programs      = {};
+	state.ps_input_info = {};
 	std::array<Prospero::ColorComponentMapping, RENDER_COLOR_ATTACHMENTS_MAX>
 	    target_export_mapping {};
 	for (uint32_t i = 0; i < state.color_count; i++) {
 		target_export_mapping[state.color_info[i].target_slot] = state.color_info[i].export_mapping;
 	}
-	if (log_phases) {
-		LogDrawPhase(draw.name, "GetVertexProgram");
-	}
 	auto& pipeline_cache = buffer.GetContext().GetPipelineCache();
-	state.vertex_program =
-	    pipeline_cache.GetVertexProgram(vertex_shader_info, shader_regs, state.vs_input_info);
-
-	if (!state.ps_active) {
-		return;
-	}
 	if (log_phases) {
-		LogDrawPhase(draw.name, "GetPixelProgram");
+		LogDrawPhase(draw.name, "GetGraphicsPrograms");
 	}
-	state.pixel_program =
-	    pipeline_cache.GetPixelProgram(pixel_shader_info, shader_regs, state.vs_input_info,
-	                                   target_export_mapping, state.ps_input_info);
+	state.programs = pipeline_cache.GetGraphicsPrograms(
+	    vertex_shader_info, pixel_shader_info, shader_regs, ctx, target_export_mapping,
+	    state.ps_active, state.vs_input_info, state.ps_input_info);
 }
 
 static PreparedVertexBuffers PrepareVertexBuffers(uint64_t submit_id, CommandBuffer& buffer,
@@ -1105,7 +1111,7 @@ static void LogDrawStateIfNeeded(const CommandBuffer& buffer, const DrawCallInfo
 
 	if (state.ps_active) {
 		LogDrawTargetState(draw.name, state.color_info[0], state.depth_info, buffer,
-		                   state.ps_input_info, draw.index_count, draw.flags);
+		                   state.ps_input_info, draw.index_count, 0);
 	}
 	LogDrawInputState(buffer, state.color_info[0], state.vs_input_info, index_type_and_size,
 	                  draw.index_count, index_addr);
@@ -1127,10 +1133,10 @@ static void EmitDrawPrimitives(const HW::UserConfig& ucfg, vk::CommandBuffer vk_
 		case Prospero::PrimitiveType::kRectList:
 			if (emit.indexed) {
 				vk_buffer.drawIndexed(draw.index_count, draw.instance_count, 0, emit.vertex_offset,
-				                      draw.first_instance);
+				                      emit.first_instance);
 			} else {
 				vk_buffer.draw(draw.index_count, draw.instance_count, emit.first_vertex,
-				               draw.first_instance);
+				               emit.first_instance);
 			}
 			break;
 		case Prospero::PrimitiveType::kRectListLegacy:
@@ -1139,17 +1145,17 @@ static void EmitDrawPrimitives(const HW::UserConfig& ucfg, vk::CommandBuffer vk_
 			}
 			// Sarah
 			EXIT_NOT_IMPLEMENTED(!(draw.index_count == 3 && vs_input_info.buffers_num == 0));
-			vk_buffer.draw(4, draw.instance_count, emit.first_vertex, draw.first_instance);
+			vk_buffer.draw(4, draw.instance_count, emit.first_vertex, emit.first_instance);
 			break;
 		case Prospero::PrimitiveType::kQuadListLegacy:
 			EXIT_NOT_IMPLEMENTED((draw.index_count & 0x3u) != 0);
 			for (uint32_t i = 0; i < draw.index_count; i += 4) {
 				if (emit.indexed) {
 					vk_buffer.drawIndexed(4, draw.instance_count, i, emit.vertex_offset,
-					                      draw.first_instance);
+					                      emit.first_instance);
 				} else {
 					vk_buffer.draw(4, draw.instance_count, i + emit.first_vertex,
-					               draw.first_instance);
+					               emit.first_instance);
 				}
 			}
 			break;
@@ -1180,7 +1186,7 @@ void RenderExecutor::ExecutePreparedDraw(uint64_t submit_id, CommandBuffer& buff
 	auto& pipeline = m_context.GetPipelineCache().CreateGraphicsPipeline(
 	    std::span {state.color_info, state.color_count}, state.depth_info, state.vs_input_info, buffer,
 	    state.ps_active ? &state.ps_input_info : nullptr, topology, primitive_restart_enable,
-	    state.vertex_program, state.pixel_program);
+	    state.programs.vertex, state.programs.pixel);
 
 	// Resource preparation above may synchronously finish and restart the scheduler. From this
 	// point onward, every operation targets the current command buffer and cannot touch guest
@@ -1242,23 +1248,21 @@ void RenderExecutor::ExecutePreparedDraw(uint64_t submit_id, CommandBuffer& buff
 }
 
 void RenderExecutor::DrawIndex(uint64_t submit_id, CommandBuffer& buffer,
-                               uint32_t index_type_and_size, uint32_t index_count,
-                               const void* index_addr, uint32_t flags, uint32_t type,
-                               uint32_t instance_count, uint32_t render_target_slice_offset,
-                               int32_t vertex_offset_add, uint32_t first_instance) {
+                               const DrawIndexArgs& args) {
 	KYTY_PROFILER_FUNCTION();
 
 	EXIT_IF(buffer.IsInvalid());
+	EXIT_IF(args.offset_source == DrawOffsetSource::DrawState && args.first_instance != 0);
 	m_context.GetCommandScheduler().PopPendingOperations();
 	auto& ucfg   = buffer.GetUserConfig();
 	auto& sh_ctx = buffer.GetShaders();
 
 	buffer.SetDebugInfo(static_cast<uint32_t>(CommandBufferDebugOp::DrawIndex), submit_id,
-	                    index_count, flags, type, instance_count,
-	                    reinterpret_cast<uint64_t>(index_addr));
+	                    args.index_count, 0, 1, args.instance_count,
+	                    reinterpret_cast<uint64_t>(args.index_addr));
 
 	Common::LockGuard lock(m_context.GetMutex());
-	if (index_count == 0 || instance_count == 0) {
+	if (args.index_count == 0 || args.instance_count == 0) {
 		return;
 	}
 
@@ -1284,15 +1288,12 @@ void RenderExecutor::DrawIndex(uint64_t submit_id, CommandBuffer& buffer,
 		     "\t index_type_and_size = 0x%08" PRIx32 "\n"
 		     "\t index_count         = 0x%08" PRIx32 "\n"
 		     "\t index_addr          = 0x%016" PRIx64 "\n"
-		     "\t flags               = 0x%08" PRIx32 "\n"
-		     "\t type                = 0x%08" PRIx32 "\n"
 		     "\t instance_count      = 0x%08" PRIx32 "\n"
-		     "\t rt_slice_offset     = 0x%08" PRIx32 "\n"
-		     "\t vertex_offset_add   = 0x%08" PRIx32 "\n"
+		     "\t base_vertex         = 0x%08" PRIx32 "\n"
 		     "\t first_instance      = 0x%08" PRIx32 "\n",
-		     index_type_and_size, index_count, reinterpret_cast<uint64_t>(index_addr), flags, type,
-		     instance_count, render_target_slice_offset, static_cast<uint32_t>(vertex_offset_add),
-		     first_instance);
+		     args.index_type_and_size, args.index_count,
+		     reinterpret_cast<uint64_t>(args.index_addr), args.instance_count,
+		     static_cast<uint32_t>(args.base_vertex), args.first_instance);
 	}
 
 	uc_check(ucfg);
@@ -1307,44 +1308,42 @@ void RenderExecutor::DrawIndex(uint64_t submit_id, CommandBuffer& buffer,
 	vk::IndexType index_type           = vk::IndexType::eUint16;
 	uint64_t      index_size           = 0;
 	bool          expand_index8_to_u16 = false;
-	const bool primitive_restart = ResolvePrimitiveRestart(buffer, topology, index_type_and_size);
+	const bool    primitive_restart =
+	    ResolvePrimitiveRestart(buffer, topology, args.index_type_and_size);
 
-	switch (static_cast<Prospero::IndexType>(index_type_and_size)) {
+	switch (static_cast<Prospero::IndexType>(args.index_type_and_size)) {
 		case Prospero::IndexType::kIndex16:
 			index_type = vk::IndexType::eUint16;
-			index_size = 2 * static_cast<uint64_t>(index_count);
+			index_size = 2 * static_cast<uint64_t>(args.index_count);
 			break;
 		case Prospero::IndexType::kIndex32:
 			index_type = vk::IndexType::eUint32;
-			index_size = 4 * static_cast<uint64_t>(index_count);
+			index_size = 4 * static_cast<uint64_t>(args.index_count);
 			break;
 		// Some games use it - need vulkan extension
 		case Prospero::IndexType::kIndex8:
 			index_type           = vk::IndexType::eUint16;
-			index_size           = static_cast<uint64_t>(index_count);
+			index_size           = static_cast<uint64_t>(args.index_count);
 			expand_index8_to_u16 = true;
 			break;
-		default: EXIT("unknown index_type_and_size: %u\n", index_type_and_size);
+		default: EXIT("unknown index_type_and_size: %u\n", args.index_type_and_size);
 	}
 
-	EXIT_NOT_IMPLEMENTED(flags != 0);
-	EXIT_NOT_IMPLEMENTED(type != 1);
-	const DrawCallInfo    draw {"DrawIndex",    CommandBufferDebugOp::DrawIndex,
-	                            index_count,    flags,
-	                            instance_count, first_instance};
+	const DrawCallInfo    draw {"DrawIndex", CommandBufferDebugOp::DrawIndex, args.index_count,
+	                            args.instance_count, args.first_instance};
 	std::vector<uint16_t> expanded_indices;
 	if (expand_index8_to_u16) {
-		EXIT_NOT_IMPLEMENTED(index_addr == nullptr);
-		const auto* src = static_cast<const uint8_t*>(index_addr);
-		expanded_indices.resize(index_count);
-		for (uint32_t i = 0; i < index_count; i++) {
+		EXIT_NOT_IMPLEMENTED(args.index_addr == nullptr);
+		const auto* src = static_cast<const uint8_t*>(args.index_addr);
+		expanded_indices.resize(args.index_count);
+		for (uint32_t i = 0; i < args.index_count; i++) {
 			expanded_indices[i] = primitive_restart && src[i] == 0xffu ? 0xffffu : src[i];
 		}
 	}
 
 	DrawIndexBufferSource index_source {};
 	index_source.enabled = true;
-	index_source.address = reinterpret_cast<uint64_t>(index_addr);
+	index_source.address = reinterpret_cast<uint64_t>(args.index_addr);
 	index_source.host_data =
 	    expanded_indices.empty() ? nullptr : static_cast<const void*>(expanded_indices.data());
 	index_source.size =
@@ -1352,21 +1351,28 @@ void RenderExecutor::DrawIndex(uint64_t submit_id, CommandBuffer& buffer,
 	index_source.type = index_type;
 
 	DrawRenderState state {};
-	if (!PrepareDrawRenderState(submit_id, buffer, draw, render_target_slice_offset, true, state)) {
+	if (!PrepareDrawRenderState(submit_id, buffer, draw, args.render_target_slice_offset, true,
+	                            state)) {
 		ResetBindings();
 		return;
 	}
 
 	RefreshShaders(buffer, draw, true, state);
 
-	LogDrawStateIfNeeded(buffer, draw, state, true, false, index_type_and_size, index_addr);
+	LogDrawStateIfNeeded(buffer, draw, state, true, false, args.index_type_and_size,
+	                     args.index_addr);
 
+	const bool indirect = args.offset_source == DrawOffsetSource::IndirectArgs;
 	const auto vertex_offset =
-	    ResolveVertexOffset(ucfg.GetIndexOffset(), state.vs_input_info) + vertex_offset_add;
+	    indirect
+	        ? args.base_vertex
+	        : ResolveVertexOffset(ucfg.GetIndexOffset(), state.vs_input_info) + args.base_vertex;
 
 	DrawEmitInfo emit {};
 	emit.indexed       = true;
 	emit.vertex_offset = vertex_offset;
+	emit.first_instance =
+	    indirect ? args.first_instance : ResolveInstanceOffset(state.vs_input_info);
 
 	ExecutePreparedDraw(submit_id, buffer, draw, state, topology, emit, index_source,
 	                    primitive_restart, true, true, false);
@@ -1374,22 +1380,21 @@ void RenderExecutor::DrawIndex(uint64_t submit_id, CommandBuffer& buffer,
 }
 
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
-void RenderExecutor::DrawAuto(uint64_t submit_id, CommandBuffer& buffer, uint32_t index_count,
-                              uint32_t flags, uint32_t render_target_slice_offset,
-                              uint32_t instance_count, uint32_t first_vertex,
-                              uint32_t first_instance) {
+void RenderExecutor::DrawAuto(uint64_t submit_id, CommandBuffer& buffer, const DrawAutoArgs& args) {
 	KYTY_PROFILER_FUNCTION();
 
 	EXIT_IF(buffer.IsInvalid());
+	EXIT_IF(args.offset_source == DrawOffsetSource::DrawState && args.first_instance != 0);
 	m_context.GetCommandScheduler().PopPendingOperations();
 	auto& ucfg   = buffer.GetUserConfig();
 	auto& sh_ctx = buffer.GetShaders();
 
 	buffer.SetDebugInfo(static_cast<uint32_t>(CommandBufferDebugOp::DrawIndexAuto), submit_id,
-	                    index_count, flags, first_vertex, instance_count, first_instance);
+	                    args.vertex_count, 0, args.first_vertex, args.instance_count,
+	                    args.first_instance);
 
 	Common::LockGuard lock(m_context.GetMutex());
-	if (index_count == 0 || instance_count == 0) {
+	if (args.vertex_count == 0 || args.instance_count == 0) {
 		return;
 	}
 
@@ -1412,27 +1417,22 @@ void RenderExecutor::DrawAuto(uint64_t submit_id, CommandBuffer& buffer, uint32_
 		hw_print(buffer);
 
 		LOGF("GraphicsRenderDrawIndexAuto():Parameters:\n"
-		     "\t index_count         = 0x%08" PRIx32 "\n"
-		     "\t flags               = 0x%08" PRIx32 "\n"
-		     "\t rt_slice_offset     = 0x%08" PRIx32 "\n"
+		     "\t vertex_count        = 0x%08" PRIx32 "\n"
 		     "\t instance_count      = 0x%08" PRIx32 "\n"
 		     "\t first_vertex        = 0x%08" PRIx32 "\n"
 		     "\t first_instance      = 0x%08" PRIx32 "\n",
-		     index_count, flags, render_target_slice_offset, instance_count, first_vertex,
-		     first_instance);
+		     args.vertex_count, args.instance_count, args.first_vertex, args.first_instance);
 	}
 
 	uc_check(ucfg);
 
 	hw_check(buffer);
 
-	EXIT_NOT_IMPLEMENTED(flags != 0);
 	const DrawCallInfo draw {"DrawIndexAuto", CommandBufferDebugOp::DrawIndexAuto,
-	                         index_count,     flags,
-	                         instance_count,  first_instance};
+	                         args.vertex_count, args.instance_count, args.first_instance};
 
 	DrawRenderState state {};
-	if (!PrepareDrawRenderState(submit_id, buffer, draw, render_target_slice_offset, false,
+	if (!PrepareDrawRenderState(submit_id, buffer, draw, args.render_target_slice_offset, false,
 	                            state)) {
 		ResetBindings();
 		return;
@@ -1447,11 +1447,12 @@ void RenderExecutor::DrawAuto(uint64_t submit_id, CommandBuffer& buffer, uint32_
 
 	const bool rect_list = topology == vk::PrimitiveTopology::ePatchList;
 	if (rect_list && state.vs_input_info.buffers_num == 0 &&
-	    state.vs_input_info.param_export_mask == 0 && state.ps_input_info.input_num != 0) {
+	    state.vs_input_info.stage.program->param_export_mask == 0 &&
+	    state.ps_input_info.input_num != 0) {
 		if (graphics_debug_dump_enabled()) {
 			LOGF("DrawIndexAuto: skipping rect-list draw with no VS param exports and PS inputs: "
 			     "ps_inputs=%u ps=0x%016" PRIx64 " es=0x%016" PRIx64 " gs=0x%016" PRIx64 "\n",
-			     state.ps_input_info.input_num, sh_ctx.GetPs().ps_regs.chksum,
+			     state.ps_input_info.input_num, sh_ctx.GetPs().ps_regs.data_addr,
 			     sh_ctx.GetVs().es_regs.data_addr, sh_ctx.GetVs().gs_regs.data_addr);
 		}
 		ResetBindings();
@@ -1462,10 +1463,15 @@ void RenderExecutor::DrawAuto(uint64_t submit_id, CommandBuffer& buffer, uint32_
 	                     ucfg.GetPrimType() == Prospero::PrimitiveType::kRectListLegacy, 0,
 	                     nullptr);
 
-	const auto   vertex_offset = ResolveVertexOffset(ucfg.GetIndexOffset(), state.vs_input_info) +
-	                             static_cast<int32_t>(first_vertex);
+	const bool indirect = args.offset_source == DrawOffsetSource::IndirectArgs;
+	const auto vertex_offset =
+	    indirect ? static_cast<int32_t>(args.first_vertex)
+	             : ResolveVertexOffset(ucfg.GetIndexOffset(), state.vs_input_info) +
+	                   static_cast<int32_t>(args.first_vertex);
 	DrawEmitInfo emit {};
 	emit.first_vertex = static_cast<uint32_t>(vertex_offset);
+	emit.first_instance =
+	    indirect ? args.first_instance : ResolveInstanceOffset(state.vs_input_info);
 
 	DrawIndexBufferSource index_source {};
 	ExecutePreparedDraw(submit_id, buffer, draw, state, topology, emit, index_source, false, false,

src/graphics/host_gpu/renderer/renderDraw.h

@@ -7,8 +7,9 @@ namespace Libs::Graphics {
 
 struct ShaderVertexInputInfo;
 
-[[nodiscard]] int32_t ResolveVertexOffset(uint32_t                     index_offset,
-                                          const ShaderVertexInputInfo& vs_input_info);
+[[nodiscard]] int32_t  ResolveVertexOffset(uint32_t                     index_offset,
+                                           const ShaderVertexInputInfo& vs_input_info);
+[[nodiscard]] uint32_t ResolveInstanceOffset(const ShaderVertexInputInfo& vs_input_info);
 
 } // namespace Libs::Graphics
 

src/graphics/presentation/imeOverlay.cpp

@@ -209,7 +209,7 @@ void RefreshVisibility() {
 	}
 	const bool was_controller = std::exchange(g_controller_captured, capture_controller);
 	if (capture_controller || was_controller) {
-		Controller::ControllerResetInputState();
+		Controller::ResetInputState();
 	}
 	g_active = visible;
 
@@ -404,7 +404,7 @@ void ShutdownImeInput() {
 		g_input_lifecycle_active = false;
 		g_active                 = false;
 		if (std::exchange(g_controller_captured, false)) {
-			Controller::ControllerResetInputState();
+			Controller::ResetInputState();
 		}
 		g_visibility_event.store(static_cast<Uint32>(-1), std::memory_order_release);
 	}

src/graphics/presentation/window/hostInput.cpp

@@ -1,18 +1,24 @@
 #include "graphics/presentation/window/hostInput.h"
 
+#include "SDL_error.h"
+#include "SDL_events.h"
 #include "SDL_keyboard.h"
 #include "SDL_keycode.h"
 #include "SDL_mouse.h"
+#include "SDL_timer.h"
+#include "SDL_video.h"
 #include "common/assert.h"
 #include "common/emulatorConfig.h"
+#include "common/logging/log.h"
 #include "libs/controller.h"
 
 #include <algorithm>
 #include <array>
+#include <cmath>
 #include <cstddef>
+#include <cstdlib>
 #include <string>
 #include <string_view>
-#include <utility>
 
 namespace Libs::Graphics {
 
@@ -23,6 +29,7 @@ struct ControlInfo {
 	uint32_t         button   = 0;
 	Controller::Axis axis     = Controller::Axis::AxisMax;
 	bool             positive = false;
+	float            touch_x  = 0.0f;
 };
 
 static constexpr std::array CONTROL_INFO = {
@@ -41,7 +48,10 @@ static constexpr std::array CONTROL_INFO = {
     ControlInfo {"Circle", Controller::PAD_BUTTON_CIRCLE},
     ControlInfo {"Cross", Controller::PAD_BUTTON_CROSS},
     ControlInfo {"Square", Controller::PAD_BUTTON_SQUARE},
-    ControlInfo {"TouchPad", Controller::PAD_BUTTON_TOUCH_PAD},
+    ControlInfo {"TouchPad", Controller::PAD_BUTTON_TOUCH_PAD, Controller::Axis::AxisMax, false,
+                 0.25f},
+    ControlInfo {"TouchPadRight", Controller::PAD_BUTTON_TOUCH_PAD, Controller::Axis::AxisMax,
+                 false, 0.75f},
     ControlInfo {"LeftStickLeft", 0, Controller::Axis::LeftX, false},
     ControlInfo {"LeftStickRight", 0, Controller::Axis::LeftX, true},
     ControlInfo {"LeftStickUp", 0, Controller::Axis::LeftY, false},
@@ -60,17 +70,23 @@ struct Binding {
 	std::size_t control      = INVALID_CONTROL;
 };
 
+constexpr int              MOUSE_POLL_INTERVAL_MS = 33;
+constexpr std::string_view MOUSE_SENSITIVITY      = "MouseSensitivity=";
+
+struct MouseJoystickState {
+	bool     enabled   = false;
+	bool     output    = false;
+	uint64_t next_poll = 0;
+};
+
+MouseJoystickState g_mouse;
+
 std::size_t ControlFromName(std::string_view name) {
 	const auto info = std::find_if(CONTROL_INFO.begin(), CONTROL_INFO.end(),
 	                               [name](const auto& item) { return item.name == name; });
 	return static_cast<std::size_t>(std::distance(CONTROL_INFO.begin(), info));
 }
 
-const ControlInfo& Info(std::size_t control) {
-	EXIT_IF(control >= CONTROL_INFO.size());
-	return CONTROL_INFO[control];
-}
-
 SDL_Keycode NormalizeKey(SDL_Keycode key) {
 	switch (key) {
 		case SDLK_RSHIFT: return SDLK_LSHIFT;
@@ -103,10 +119,17 @@ bool Conflicts(const Binding& first, const Binding& second) {
 
 class InputMap {
 public:
-	InputMap(): m_custom(!Config::GetKeymap().empty()) {
+	InputMap() {
 		for (const auto& value: Config::GetKeymap()) {
 			const std::string_view entry = value;
-			const auto             split = entry.find('=');
+			if (entry.starts_with(MOUSE_SENSITIVITY)) {
+				const float sensitivity =
+				    std::strtof(value.c_str() + MOUSE_SENSITIVITY.size(), nullptr);
+				m_mouse_sensitivity =
+				    std::clamp(std::isfinite(sensitivity) ? sensitivity : 1.0f, 0.1f, 5.0f);
+				continue;
+			}
+			const auto split = entry.find('=');
 
 			Binding binding;
 			if (split != std::string_view::npos) {
@@ -118,9 +141,9 @@ public:
 				}
 			}
 
-			const bool reserved =
-			    binding.key == SDLK_ESCAPE || binding.key == SDLK_SPACE || binding.key == SDLK_F1 ||
-			    binding.key == SDLK_F11;
+			const bool reserved = binding.key == SDLK_ESCAPE || binding.key == SDLK_SPACE ||
+			                      binding.key == SDLK_F1 || binding.key == SDLK_F7 ||
+			                      binding.key == SDLK_F11;
 			if (binding.control == INVALID_CONTROL || reserved ||
 			    (binding.key == SDLK_UNKNOWN && binding.mouse_button == 0)) {
 				EXIT("Invalid input mapping: %s\n", value.c_str());
@@ -128,7 +151,8 @@ public:
 			Add(binding);
 		}
 	}
-	[[nodiscard]] bool Custom() const { return m_custom; }
+	[[nodiscard]] bool  Custom() const { return m_size != 0; }
+	[[nodiscard]] float MouseSensitivity() const { return m_mouse_sensitivity; }
 
 	[[nodiscard]] std::size_t FindKey(int key_code) const {
 		key_code = NormalizeKey(static_cast<SDL_Keycode>(key_code));
@@ -159,8 +183,8 @@ private:
 	}
 
 	std::array<Binding, CONTROL_INFO.size()> m_bindings {};
-	std::size_t                              m_size = 0;
-	bool                                     m_custom;
+	std::size_t                              m_size              = 0;
+	float                                    m_mouse_sensitivity = 1.0f;
 };
 
 const InputMap& GetInputMap() {
@@ -170,16 +194,20 @@ const InputMap& GetInputMap() {
 
 void SetButton(uint32_t button, bool down) {
 	if (button == Controller::PAD_BUTTON_L2) {
-		Controller::ControllerAxis(Controller::HOST_INPUT_CONTROLLER_ID,
-		                           Controller::Axis::TriggerLeft, down ? 255 : 0);
+		Controller::SetAxis(Controller::HOST_INPUT_CONTROLLER_ID, Controller::Axis::TriggerLeft,
+		                    down ? 255 : 0);
 	} else if (button == Controller::PAD_BUTTON_R2) {
-		Controller::ControllerAxis(Controller::HOST_INPUT_CONTROLLER_ID,
-		                           Controller::Axis::TriggerRight, down ? 255 : 0);
+		Controller::SetAxis(Controller::HOST_INPUT_CONTROLLER_ID, Controller::Axis::TriggerRight,
+		                    down ? 255 : 0);
 	} else if (button != 0) {
-		Controller::ControllerButton(Controller::HOST_INPUT_CONTROLLER_ID, button, down);
+		Controller::SetButton(Controller::HOST_INPUT_CONTROLLER_ID, button, down);
 	}
 }
 
+void SetTouchPad(float x, bool down) {
+	Controller::SetTouchPad(Controller::HOST_INPUT_CONTROLLER_ID, 0, down, x, 0.5f);
+}
+
 uint32_t DefaultKeyboardButton(int key_code) {
 	switch (NormalizeKey(static_cast<SDL_Keycode>(key_code))) {
 		case SDLK_UP: return Controller::PAD_BUTTON_UP;
@@ -196,8 +224,6 @@ uint32_t DefaultKeyboardButton(int key_code) {
 		case SDLK_LCTRL: return Controller::PAD_BUTTON_R3;
 		case SDLK_RETURN:
 		case SDLK_RETURN2: return Controller::PAD_BUTTON_OPTIONS;
-		case SDLK_BACKSPACE:
-		case SDLK_TAB: return Controller::PAD_BUTTON_TOUCH_PAD;
 		default: return 0;
 	}
 }
@@ -210,8 +236,13 @@ struct StickKeys {
 };
 
 void SetStickAxis(Controller::Axis axis, bool negative, bool positive) {
-	const int value = negative == positive ? 128 : negative ? 0 : 255;
-	Controller::ControllerAxis(Controller::HOST_INPUT_CONTROLLER_ID, axis, value);
+	int value = 128;
+	if (negative && !positive) {
+		value = 0;
+	} else if (positive && !negative) {
+		value = 255;
+	}
+	Controller::SetAxis(Controller::HOST_INPUT_CONTROLLER_ID, axis, value);
 }
 
 void SetControl(std::size_t control, bool down) {
@@ -219,7 +250,11 @@ void SetControl(std::size_t control, bool down) {
 		return;
 	}
 
-	const auto& info = Info(control);
+	const auto& info = CONTROL_INFO[control];
+	if (info.button == Controller::PAD_BUTTON_TOUCH_PAD) {
+		SetTouchPad(info.touch_x, down);
+		return;
+	}
 	if (info.button != 0) {
 		SetButton(info.button, down);
 		return;
@@ -246,6 +281,8 @@ void DefaultKeyboardInput(int key_code, bool down) {
 	static StickKeys right;
 
 	switch (NormalizeKey(static_cast<SDL_Keycode>(key_code))) {
+		case SDLK_BACKSPACE: SetTouchPad(0.25f, down); return;
+		case SDLK_TAB: SetTouchPad(0.75f, down); return;
 		case SDLK_a:
 			left.left = down;
 			SetStickAxis(Controller::Axis::LeftX, left.left, left.right);
@@ -281,6 +318,33 @@ void DefaultKeyboardInput(int key_code, bool down) {
 		default: SetButton(DefaultKeyboardButton(key_code), down); return;
 	}
 }
+
+void MouseToJoystick(int delta_x, int delta_y) {
+	const double distance = std::hypot(delta_x, delta_y);
+	const double scale =
+	    std::clamp(distance * GetInputMap().MouseSensitivity() + 16.0, 64.0, 128.0) / distance;
+	const auto map_axis = [scale](int delta) {
+		return std::clamp(128 + static_cast<int>(std::lround(delta * scale)), 0, 255);
+	};
+	Controller::SetRightStick(Controller::HOST_INPUT_CONTROLLER_ID, map_axis(delta_x),
+	                          map_axis(delta_y));
+}
+
+void CenterMouseStick() {
+	if (!g_mouse.output) {
+		return;
+	}
+	Controller::SetRightStick(Controller::HOST_INPUT_CONTROLLER_ID, 128, 128);
+	g_mouse.output = false;
+}
+
+bool SetRelativeMouseMode(bool enabled) {
+	if (SDL_SetRelativeMouseMode(enabled ? SDL_TRUE : SDL_FALSE) == 0) {
+		return true;
+	}
+	LOGF("Mouse-to-joystick relative mode failed: %s\n", SDL_GetError());
+	return false;
+}
 } // namespace
 
 void HostInputInit() {
@@ -303,4 +367,63 @@ void HostInputMouseButton(uint8_t mouse_button, bool down) {
 	}
 }
 
+void HostInputToggleMouseToJoystick() {
+	if (g_mouse.enabled) {
+		SetRelativeMouseMode(false);
+		CenterMouseStick();
+		g_mouse = {};
+		LOGF("Mouse to right stick: disabled\n");
+		return;
+	}
+
+	if (!SetRelativeMouseMode(true)) {
+		return;
+	}
+	int ignored_x = 0;
+	int ignored_y = 0;
+	SDL_GetRelativeMouseState(&ignored_x, &ignored_y);
+	g_mouse.enabled   = true;
+	g_mouse.next_poll = SDL_GetTicks64() + MOUSE_POLL_INTERVAL_MS;
+	LOGF("Mouse to right stick: enabled (F7 to release)\n");
+}
+
+int PollMouse(uint64_t now_ms) {
+	if (now_ms < g_mouse.next_poll) {
+		return static_cast<int>(g_mouse.next_poll - now_ms);
+	}
+	g_mouse.next_poll = now_ms + MOUSE_POLL_INTERVAL_MS;
+
+	int delta_x = 0;
+	int delta_y = 0;
+	SDL_GetRelativeMouseState(&delta_x, &delta_y);
+	if (delta_x == 0 && delta_y == 0) {
+		CenterMouseStick();
+		return MOUSE_POLL_INTERVAL_MS;
+	}
+
+	MouseToJoystick(delta_x, delta_y);
+	g_mouse.output = true;
+	return MOUSE_POLL_INTERVAL_MS;
+}
+
+bool HostInputWaitEvent(SDL_Event* event) {
+	if (!g_mouse.enabled || SDL_GetKeyboardFocus() == nullptr) {
+		CenterMouseStick();
+		if (SDL_WaitEvent(event) == 0) {
+			EXIT("%s\n", SDL_GetError());
+		}
+		return true;
+	}
+
+	const int timeout_ms = PollMouse(SDL_GetTicks64());
+	SDL_ClearError();
+	if (SDL_WaitEventTimeout(event, timeout_ms) != 0) {
+		return true;
+	}
+	if (SDL_GetError()[0] != '\0') {
+		EXIT("%s\n", SDL_GetError());
+	}
+	return false;
+}
+
 } // namespace Libs::Graphics

src/graphics/presentation/window/hostInput.h

@@ -3,11 +3,15 @@
 
 #include <cstdint>
 
+union SDL_Event;
+
 namespace Libs::Graphics {
 
-void HostInputInit();
-void HostInputKey(int key_code, bool down);
-void HostInputMouseButton(uint8_t mouse_button, bool down);
+void               HostInputInit();
+void               HostInputKey(int key_code, bool down);
+void               HostInputMouseButton(uint8_t mouse_button, bool down);
+void               HostInputToggleMouseToJoystick();
+[[nodiscard]] bool HostInputWaitEvent(SDL_Event* event);
 
 } // namespace Libs::Graphics
 

src/graphics/presentation/window/swapchain.cpp

@@ -1,30 +1,9 @@
-#include "SDL.h"
-#include "SDL_error.h"
-#include "SDL_events.h"
-#include "SDL_gamecontroller.h"
-#include "SDL_hints.h"
-#include "SDL_joystick.h"
-#include "SDL_keyboard.h"
-#include "SDL_keycode.h"
-#include "SDL_mouse.h"
-#include "SDL_pixels.h"
-#include "SDL_rwops.h"
-#include "SDL_stdinc.h"
-#include "SDL_surface.h"
-#include "SDL_thread.h"
-#include "SDL_touch.h"
-#include "SDL_video.h"
-#include "SDL_vulkan.h"
 #include "common/assert.h"
 #include "common/common.h"
 #include "common/emulatorConfig.h"
-#include "common/file.h"
 #include "common/logging/log.h"
 #include "common/profiler.h"
-#include "common/stringUtils.h"
-#include "common/systemInfo.h"
 #include "common/threads.h"
-#include "common/timer.h"
 #include "graphics/host_gpu/graphicContext.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
@@ -34,19 +13,11 @@
 #include "graphics/presentation/presenter.h"
 #include "graphics/presentation/videoOut.h"
 #include "graphics/presentation/window/windowInternal.h"
-#include "libs/controller.h"
-#include "loader/systemContent.h"
 
 #include <algorithm>
-#include <array>
-#include <cstdio>
-#include <cstdlib>
-#include <cstring>
 #include <deque>
-#include <fmt/format.h>
 #include <limits>
 #include <memory>
-#include <string>
 #include <vector>
 #include <vulkan/vk_platform.h>
 
@@ -334,7 +305,6 @@ public:
 
 private:
 	void Destroy();
-	void RefreshSurfaceSize();
 
 	WindowContext&              m_window;
 	vk::SwapchainKHR            m_handle = nullptr;
@@ -396,12 +366,12 @@ struct Presenter::Impl {
 
 void Swapchain::Create() {
 	auto& graphics = m_window.graphic_ctx;
-	EXIT_IF(graphics.screen_width == 0);
-	EXIT_IF(graphics.screen_height == 0);
 	EXIT_IF(graphics.device == nullptr);
 	EXIT_IF(m_window.surface == nullptr);
 
 	Common::LockGuard lock(m_window.mutex);
+	EXIT_IF(graphics.screen_width == 0);
+	EXIT_IF(graphics.screen_height == 0);
 	const auto&       surface = m_window.surface_capabilities;
 	EXIT_NOT_IMPLEMENTED(surface.formats.empty());
 
@@ -459,7 +429,21 @@ void Swapchain::Create() {
 	create_info.imageSharingMode = vk::SharingMode::eExclusive;
 	create_info.preTransform     = transform;
 	create_info.compositeAlpha   = composite;
-	create_info.presentMode      = vk::PresentModeKHR::eFifo;
+	switch (Config::GetPresentMode()) {
+		case Config::PresentMode::Mailbox:
+			create_info.presentMode = vk::PresentModeKHR::eMailbox;
+			break;
+		case Config::PresentMode::Immediate:
+			create_info.presentMode = vk::PresentModeKHR::eImmediate;
+			break;
+		case Config::PresentMode::Fifo:
+		default: create_info.presentMode = vk::PresentModeKHR::eFifo; break;
+	}
+	if (std::find(surface.present_modes.begin(), surface.present_modes.end(),
+	              create_info.presentMode) == surface.present_modes.end()) {
+		LOGF("warning: requested present mode is unavailable; falling back to Fifo\n");
+		create_info.presentMode = vk::PresentModeKHR::eFifo;
+	}
 	create_info.clipped          = VK_TRUE;
 	RequireVulkanSuccess(graphics.device.createSwapchainKHR(&create_info, nullptr, &m_handle),
 	                     "vkCreateSwapchainKHR");
@@ -554,30 +538,18 @@ void Swapchain::Destroy() {
 	m_render_complete.clear();
 }
 
-void Swapchain::RefreshSurfaceSize() {
-	int width  = 0;
-	int height = 0;
-	SDL_Vulkan_GetDrawableSize(m_window.window, &width, &height);
-	if (width > 0 && height > 0) {
-		m_window.graphic_ctx.screen_width  = static_cast<uint32_t>(width);
-		m_window.graphic_ctx.screen_height = static_cast<uint32_t>(height);
-	}
-
-	m_window.RefreshSurfaceCapabilities();
-}
-
 void Swapchain::Recreate(bool surface_lost) {
 	Destroy();
 	if (surface_lost) {
 #if defined(__APPLE__)
 		// Surface recreation goes through SDL_Vulkan_CreateSurface, which touches the
 		// window's view/layer and must run on the main thread on macOS.
-		m_window.RunOnMainThread([this] { m_window.RecreateSurface(); }, true);
+		m_window.RunOnMainThread([this] { m_window.RecreateSurface(); });
 #else
 		m_window.RecreateSurface();
 #endif
 	}
-	RefreshSurfaceSize();
+	m_window.RefreshSurfaceCapabilities();
 	Create();
 }
 
@@ -803,29 +775,6 @@ void Presenter::Present(Frame& frame, bool reuse) {
 	KYTY_PROFILER_FUNCTION();
 	m_impl->frames.ValidateForPresent(&frame, reuse);
 
-	auto& window = m_impl->window;
-	if (window.window_hidden) {
-#if defined(__APPLE__)
-		// AppKit traps if a window is shown off the main thread; marshal and wait so the
-		// swapchain below is recreated against a visible window.
-		window.RunOnMainThread(
-		    [&window] {
-			    window.UpdateIcon();
-			    SDL_ShowWindow(window.window);
-			    SDL_RaiseWindow(window.window);
-		    },
-		    true);
-#else
-		window.UpdateIcon();
-
-		SDL_ShowWindow(window.window);
-		SDL_RaiseWindow(window.window);
-#endif
-
-		window.window_hidden = false;
-		m_impl->RecoverSwapchain(Swapchain::Status::Recreate);
-	}
-
 	const auto ime_visual = GetImeVisualState();
 	auto&      swapchain  = m_impl->swapchain;
 	for (uint32_t attempt = 0; attempt < 2; attempt++) {
@@ -848,7 +797,7 @@ void Presenter::Present(Frame& frame, bool reuse) {
 		}
 
 		m_impl->presented_ime_revision.store(ime_visual.revision, std::memory_order_release);
-		window.UpdateTitle();
+		m_impl->window.UpdateTitle();
 		m_impl->frames.Release(&frame, true);
 		return;
 	}

src/graphics/presentation/window/vulkanWindow.cpp

@@ -22,7 +22,6 @@
 #include "common/logging/log.h"
 #include "common/profiler.h"
 #include "common/stringUtils.h"
-#include "common/systemInfo.h"
 #include "common/threads.h"
 #include "common/timer.h"
 #include "graphics/host_gpu/graphicContext.h"
@@ -64,6 +63,17 @@ struct VulkanExtensions {
 	std::vector<vk::LayerProperties>     available_layers;
 };
 
+vk::PhysicalDeviceVulkan12Features WindowContext::RequiredVulkan12Features() noexcept {
+	vk::PhysicalDeviceVulkan12Features features {};
+	features.sType                    = vk::StructureType::ePhysicalDeviceVulkan12Features;
+	features.samplerMirrorClampToEdge = VK_TRUE;
+	features.timelineSemaphore        = VK_TRUE;
+	features.shaderOutputLayer        = VK_TRUE;
+	features.bufferDeviceAddress      = VK_TRUE;
+	features.shaderBufferInt64Atomics = VK_TRUE;
+	return features;
+}
+
 vk::PhysicalDeviceVulkan13Features WindowContext::RequiredVulkan13Features() noexcept {
 	vk::PhysicalDeviceVulkan13Features features {};
 	features.sType            = vk::StructureType::ePhysicalDeviceVulkan13Features;
@@ -158,6 +168,10 @@ static void VulkanFindPhysicalDevice(vk::Instance instance, vk::SurfaceKHR surfa
 	    });
 	EXIT_NOT_IMPLEMENTED(devices.empty());
 
+	if (Config::GetGpuIndex() >= 0) {
+		devices = {devices[Config::GetGpuIndex()]};
+	}
+
 	vk::PhysicalDevice  best_device       = nullptr;
 	uint32_t            best_queue_family = static_cast<uint32_t>(-1);
 	SurfaceCapabilities best_capabilities;
@@ -214,6 +228,7 @@ static void VulkanFindPhysicalDevice(vk::Instance instance, vk::SurfaceKHR surfa
 		device_features2.pNext = &features13;
 
 		device.getFeatures2(&device_features2);
+		const auto required_features12 = WindowContext::RequiredVulkan12Features();
 		const auto required_features13 = WindowContext::RequiredVulkan13Features();
 
 		const auto queue_family = VulkanFindQueueFamily(device, surface);
@@ -246,14 +261,31 @@ static void VulkanFindPhysicalDevice(vk::Instance instance, vk::SurfaceKHR surfa
 		}
 #endif
 
-		if (features12.samplerMirrorClampToEdge != VK_TRUE) {
+		if (required_features12.samplerMirrorClampToEdge == VK_TRUE &&
+		    features12.samplerMirrorClampToEdge != VK_TRUE) {
 			LOGF("samplerMirrorClampToEdge is not supported\n");
 			skip_device = true;
 		}
-		if (features12.timelineSemaphore != VK_TRUE) {
+		if (required_features12.timelineSemaphore == VK_TRUE &&
+		    features12.timelineSemaphore != VK_TRUE) {
 			LOGF("timelineSemaphore is not supported\n");
 			skip_device = true;
 		}
+		if (required_features12.shaderOutputLayer == VK_TRUE &&
+		    features12.shaderOutputLayer != VK_TRUE) {
+			LOGF("shaderOutputLayer is not supported\n");
+			skip_device = true;
+		}
+		if (required_features12.bufferDeviceAddress == VK_TRUE &&
+		    features12.bufferDeviceAddress != VK_TRUE) {
+			LOGF("bufferDeviceAddress is not supported\n");
+			skip_device = true;
+		}
+		if (required_features12.shaderBufferInt64Atomics == VK_TRUE &&
+		    features12.shaderBufferInt64Atomics != VK_TRUE) {
+			LOGF("shaderBufferInt64Atomics is not supported\n");
+			skip_device = true;
+		}
 		if (features13.robustImageAccess != VK_TRUE) {
 			LOGF("robustImageAccess is not supported\n");
 			skip_device = true;
@@ -288,11 +320,6 @@ static void VulkanFindPhysicalDevice(vk::Instance instance, vk::SurfaceKHR surfa
 			LOGF("largePoints is not supported\n");
 			skip_device = true;
 		}
-		if (features12.shaderOutputLayer != VK_TRUE) {
-			LOGF("shaderOutputLayer is not supported\n");
-			skip_device = true;
-		}
-
 		if (device_features2.features.fragmentStoresAndAtomics != VK_TRUE) {
 			LOGF("fragmentStoresAndAtomics is not supported\n");
 			skip_device = true;
@@ -544,10 +571,8 @@ static vk::Device VulkanCreateDevice(vk::PhysicalDevice physical_device, const V
 #endif
 	depth_clip_control.depthClipControl = VK_TRUE;
 
-	vk::PhysicalDeviceVulkan12Features features12 {};
-	features12.sType                    = vk::StructureType::ePhysicalDeviceVulkan12Features;
-	features12.pNext                    = &depth_clip_control;
-	features12.samplerMirrorClampToEdge = VK_TRUE;
+	auto features12  = WindowContext::RequiredVulkan12Features();
+	features12.pNext = &depth_clip_control;
 
 	vk::PhysicalDeviceVulkan13Features supported_features13 {};
 	supported_features13.sType = vk::StructureType::ePhysicalDeviceVulkan13Features;
@@ -583,8 +608,12 @@ static vk::Device VulkanCreateDevice(vk::PhysicalDevice physical_device, const V
 	supported_features2.sType = vk::StructureType::ePhysicalDeviceFeatures2;
 	supported_features2.pNext = &supported_features13;
 	physical_device.getFeatures2(&supported_features2);
+	const auto required_features12 = WindowContext::RequiredVulkan12Features();
 	const auto required_features13 = WindowContext::RequiredVulkan13Features();
-	EXIT_NOT_IMPLEMENTED(supported_features12.timelineSemaphore != VK_TRUE);
+	EXIT_NOT_IMPLEMENTED(required_features12.samplerMirrorClampToEdge == VK_TRUE &&
+	                     supported_features12.samplerMirrorClampToEdge != VK_TRUE);
+	EXIT_NOT_IMPLEMENTED(required_features12.timelineSemaphore == VK_TRUE &&
+	                     supported_features12.timelineSemaphore != VK_TRUE);
 	EXIT_NOT_IMPLEMENTED(required_features13.dynamicRendering == VK_TRUE &&
 	                     supported_features13.dynamicRendering != VK_TRUE);
 	EXIT_NOT_IMPLEMENTED(required_features13.synchronization2 == VK_TRUE &&
@@ -596,15 +625,15 @@ static vk::Device VulkanCreateDevice(vk::PhysicalDevice physical_device, const V
 	EXIT_NOT_IMPLEMENTED(supported_features2.features.largePoints != VK_TRUE);
 	EXIT_NOT_IMPLEMENTED(supported_features2.features.shaderInt64 != VK_TRUE);
 	EXIT_NOT_IMPLEMENTED(supported_features2.features.vertexPipelineStoresAndAtomics != VK_TRUE);
-	EXIT_NOT_IMPLEMENTED(supported_features12.shaderOutputLayer != VK_TRUE);
-	EXIT_NOT_IMPLEMENTED(supported_features12.bufferDeviceAddress != VK_TRUE);
+	EXIT_NOT_IMPLEMENTED(required_features12.shaderOutputLayer == VK_TRUE &&
+	                     supported_features12.shaderOutputLayer != VK_TRUE);
+	EXIT_NOT_IMPLEMENTED(required_features12.bufferDeviceAddress == VK_TRUE &&
+	                     supported_features12.bufferDeviceAddress != VK_TRUE);
+	EXIT_NOT_IMPLEMENTED(required_features12.shaderBufferInt64Atomics == VK_TRUE &&
+	                     supported_features12.shaderBufferInt64Atomics != VK_TRUE);
 #if !defined(__APPLE__)
 	EXIT_NOT_IMPLEMENTED(supported_fragment_barycentric.fragmentShaderBarycentric != VK_TRUE);
 #endif
-	features12.timelineSemaphore = VK_TRUE;
-	features12.shaderOutputLayer = VK_TRUE;
-	features12.bufferDeviceAddress = VK_TRUE;
-
 	vk::PhysicalDeviceFeatures device_features {};
 	device_features.fragmentStoresAndAtomics = VK_TRUE;
 	device_features.samplerAnisotropy        = VK_TRUE;
@@ -943,6 +972,18 @@ void WindowContext::CreateVulkan() {
 	inst_info.sType                   = vk::StructureType::eInstanceCreateInfo;
 	inst_info.pNext                   = (r.enable_validation_layers ? &dbg_create_info : nullptr);
 	inst_info.flags                   = {};
+#if defined(__APPLE__)
+	// MoltenVK requires VK_KHR_portability_enumeration + flag to surface
+	// portability devices. Without this, enumeratePhysicalDevices hides the
+	// MoltenVK adapter on some driver versions.
+	if (HasExtension(r.available_extensions, VK_KHR_PORTABILITY_ENUMERATION_EXTENSION_NAME)) {
+		if (!HasExtension(r.required_extensions, VK_KHR_PORTABILITY_ENUMERATION_EXTENSION_NAME)) {
+			r.required_extensions.push_back(VK_KHR_PORTABILITY_ENUMERATION_EXTENSION_NAME);
+		}
+		inst_info.flags |= vk::InstanceCreateFlagBits::eEnumeratePortabilityKHR;
+		LOGF("Vulkan instance: enabled %s\n", VK_KHR_PORTABILITY_ENUMERATION_EXTENSION_NAME);
+	}
+#endif
 	inst_info.pApplicationInfo        = &app_info;
 	inst_info.enabledExtensionCount   = static_cast<uint32_t>(r.required_extensions.size());
 	inst_info.ppEnabledExtensionNames = r.required_extensions.data();
@@ -1037,10 +1078,6 @@ void WindowContext::CreateVulkan() {
 		}
 	}
 
-	memcpy(device_name, device_properties.deviceName, sizeof(device_name));
-	std::snprintf(processor_name, sizeof(processor_name), "%s",
-	              Common::GetSystemInfo().ProcessorName.c_str());
-
 	VulkanInitSubgroupSizeControl(graphic_ctx.physical_device, graphic_ctx);
 
 	graphic_ctx.device = VulkanCreateDevice(graphic_ctx.physical_device, r, queue_family,
@@ -1079,7 +1116,6 @@ void WindowContext::RecreateSurface() {
 		EXIT("Could not recreate the Vulkan surface: %s\n", SDL_GetError());
 	}
 	surface = native_surface;
-	RefreshSurfaceCapabilities();
 }
 
 WindowContext::~WindowContext() {

src/graphics/presentation/window/window.cpp

@@ -3,6 +3,7 @@
 #include "SDL.h"
 #include "SDL_error.h"
 #include "SDL_events.h"
+#include "SDL_filesystem.h"
 #include "SDL_gamecontroller.h"
 #include "SDL_hints.h"
 #include "SDL_joystick.h"
@@ -23,12 +24,12 @@
 #include "common/file.h"
 #include "common/logging/log.h"
 #include "common/profiler.h"
-#include "common/stringUtils.h"
 #include "common/systemInfo.h"
 #include "common/threads.h"
 #include "common/timer.h"
 #include "graphics/host_gpu/graphicContext.h"
 #include "graphics/host_gpu/renderer/render.h"
+#include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/vma.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 #include "graphics/presentation/imeOverlay.h"
@@ -39,10 +40,8 @@
 #include "libs/controller.h"
 #include "loader/systemContent.h"
 
-#include <algorithm>
-#include <cstdio>
 #include <cstdlib>
-#include <cstring>
+#include <fmt/format.h>
 #include <memory>
 #include <string>
 #include <vector>
@@ -52,8 +51,6 @@
 #define STBI_NO_SIMD
 #include "stb_image.h"
 
-#include <fmt/format.h>
-
 // IWYU pragma: no_include <intrin.h>
 
 #define KYTY_ENABLE_DEBUG_PRINTF
@@ -79,7 +76,6 @@ static uint32_t ControllerButtonToPadButton(int button) {
 		case SDL_CONTROLLER_BUTTON_B: return Controller::PAD_BUTTON_CIRCLE;
 		case SDL_CONTROLLER_BUTTON_X: return Controller::PAD_BUTTON_SQUARE;
 		case SDL_CONTROLLER_BUTTON_Y: return Controller::PAD_BUTTON_TRIANGLE;
-		case SDL_CONTROLLER_BUTTON_BACK: return Controller::PAD_BUTTON_TOUCH_PAD;
 		case SDL_CONTROLLER_BUTTON_START: return Controller::PAD_BUTTON_OPTIONS;
 		case SDL_CONTROLLER_BUTTON_LEFTSTICK: return Controller::PAD_BUTTON_L3;
 		case SDL_CONTROLLER_BUTTON_RIGHTSTICK: return Controller::PAD_BUTTON_R3;
@@ -89,6 +85,7 @@ static uint32_t ControllerButtonToPadButton(int button) {
 		case SDL_CONTROLLER_BUTTON_DPAD_DOWN: return Controller::PAD_BUTTON_DOWN;
 		case SDL_CONTROLLER_BUTTON_DPAD_LEFT: return Controller::PAD_BUTTON_LEFT;
 		case SDL_CONTROLLER_BUTTON_DPAD_RIGHT: return Controller::PAD_BUTTON_RIGHT;
+		case SDL_CONTROLLER_BUTTON_TOUCHPAD: return Controller::PAD_BUTTON_TOUCH_PAD;
 		default: return 0;
 	}
 }
@@ -197,8 +194,6 @@ std::unique_ptr<WindowContext> g_window;
 } // namespace
 
 constexpr const char* KYTY_SDL_WINDOW_CAPTION = "Game";
-constexpr uint32_t    KYTY_SDL_WINDOW_FLAGS =
-    (static_cast<uint32_t>(SDL_WINDOW_HIDDEN) | static_cast<uint32_t>(SDL_WINDOW_VULKAN));
 constexpr int KYTY_SDL_WINDOWPOS_CENTERED = SDL_WINDOWPOS_CENTERED; /*NOLINT(hicpp-signed-bitwise)*/
 
 static void SetPause(WindowLoopState& game, bool flag) {
@@ -247,7 +242,6 @@ static void GameEventKeyboard(WindowLoopState& game, const EventKeyboard& key) {
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS || KYTY_PLATFORM == KYTY_PLATFORM_LINUX
 	if (key.down) {
 		switch (key.key_code) {
-			case SDLK_ESCAPE: game.need_exit = true; break;
 			case SDLK_SPACE: SetPause(game, !game.paused.load(std::memory_order_acquire)); break;
 			case SDLK_F1:
 				if (!key.repeat) {
@@ -357,26 +351,26 @@ static void GameEventController([[maybe_unused]] const EventController& f) {
 		auto* pad = SDL_GameControllerOpen(f.id);
 		EXIT_NOT_IMPLEMENTED(pad == nullptr);
 		int id = SDL_JoystickInstanceID(SDL_GameControllerGetJoystick(pad));
-		Controller::ControllerConnect(id);
+		Controller::Connect(id);
 	}
 
 	if (f.removed) {
-		Controller::ControllerDisconnect(f.id);
+		Controller::Disconnect(f.id);
 		SDL_GameControllerClose(SDL_GameControllerFromInstanceID(f.id));
 	}
 
 	if (f.down || f.up) {
 		const auto button = ControllerButtonToPadButton(f.button);
 		if (button != 0) {
-			Controller::ControllerButton(f.id, button, f.down);
+			Controller::SetButton(f.id, button, f.down);
 		}
 	}
 
 	if (f.axis) {
 		const auto axis = ControllerAxisFromSdl(f.axis_id);
 		if (axis != Controller::Axis::AxisMax) {
-			Controller::ControllerAxis(f.id, axis,
-			                           ControllerAxisValueFromSdl(f.axis_id, f.axis_value));
+			Controller::SetAxis(f.id, axis,
+			                    ControllerAxisValueFromSdl(f.axis_id, f.axis_value));
 		}
 	}
 }
@@ -407,6 +401,7 @@ static void GameEventDidEnterForeground(WindowLoopState& game) {
 
 void WindowContext::Resize(uint32_t new_width, uint32_t new_height) {
 	EXIT_IF(new_width == 0 || new_height == 0);
+	Common::LockGuard lock(mutex);
 	graphic_ctx.screen_width  = new_width;
 	graphic_ctx.screen_height = new_height;
 }
@@ -511,6 +506,13 @@ void WindowContext::ProcessEvent(double time_s) {
 	auto& game  = loop;
 	auto* event = &game.event;
 	EXIT_IF(SDL_GetEventState(SDL_DISPLAYEVENT) != SDL_ENABLE);
+	if ((event->type == SDL_KEYDOWN || event->type == SDL_KEYUP) &&
+	    event->key.keysym.sym == SDLK_F7) {
+		if (event->type == SDL_KEYDOWN && event->key.repeat == 0) {
+			HostInputToggleMouseToJoystick();
+		}
+		return;
+	}
 	if (ProcessImeInput(*event)) {
 		return;
 	}
@@ -702,6 +704,16 @@ void WindowContext::ProcessEvent(double time_s) {
 			break;
 		}
 
+		case SDL_CONTROLLERTOUCHPADDOWN:
+		case SDL_CONTROLLERTOUCHPADMOTION:
+		case SDL_CONTROLLERTOUCHPADUP:
+			if (event->ctouchpad.touchpad == 0) {
+				Controller::SetTouchPad(event->ctouchpad.which, event->ctouchpad.finger,
+				                        event->ctouchpad.type != SDL_CONTROLLERTOUCHPADUP,
+				                        event->ctouchpad.x, event->ctouchpad.y);
+			}
+			break;
+
 		case SDL_CONTROLLERDEVICEADDED:
 		case SDL_CONTROLLERDEVICEREMOVED:
 		case SDL_CONTROLLERDEVICEREMAPPED: {
@@ -728,8 +740,7 @@ void WindowContext::ProcessEvent(double time_s) {
 	}
 }
 
-#if defined(__APPLE__)
-void WindowContext::RunOnMainThread(std::function<void()> task, bool wait) {
+void WindowContext::RunOnMainThread(std::function<void()> task) {
 	if (Common::Thread::IsMainThread()) {
 		task();
 		return;
@@ -747,9 +758,6 @@ void WindowContext::RunOnMainThread(std::function<void()> task, bool wait) {
 	event.type = SDL_USEREVENT;
 	SDL_PushEvent(&event);
 
-	if (!wait) {
-		return;
-	}
 	Common::LockGuard lock(main_task_mutex);
 	while (main_tasks_run < ticket) {
 		main_task_done.Wait(&main_task_mutex);
@@ -772,7 +780,6 @@ void WindowContext::DrainMainThreadTasks() {
 	main_tasks_run += tasks.size();
 	main_task_done.SignalAll();
 }
-#endif
 
 void WindowContext::Run() {
 	Common::Timer timer;
@@ -783,9 +790,7 @@ void WindowContext::Run() {
 	loop.paused.store(false, std::memory_order_release);
 
 	while (!loop.need_exit) {
-#if defined(__APPLE__)
 		DrainMainThreadTasks();
-#endif
 		if (loop.paused.load(std::memory_order_acquire)) {
 			if (!timer.IsPaused()) {
 				timer.Pause();
@@ -794,8 +799,8 @@ void WindowContext::Run() {
 			timer.Resume();
 		}
 
-		if (SDL_WaitEvent(&loop.event) == 0) {
-			EXIT("%s\n", SDL_GetError());
+		if (!HostInputWaitEvent(&loop.event)) {
+			continue;
 		}
 		ProcessEvent(timer.GetTimeS());
 	}
@@ -809,6 +814,8 @@ static void WindowCreate(WindowContext& context) {
 	int width  = static_cast<int>(context.graphic_ctx.screen_width);
 	int height = static_cast<int>(context.graphic_ctx.screen_height);
 
+	SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_PS4_RUMBLE, "1");
+	SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_PS5_RUMBLE, "1");
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
 	SDL_SetHint(SDL_HINT_WINDOWS_DPI_SCALING, "0");
 #endif
@@ -821,11 +828,26 @@ static void WindowCreate(WindowContext& context) {
 
 	LOGF("WindowCreate(): width = %d, height = %d\n", width, height);
 
-	uint32_t window_flags = KYTY_SDL_WINDOW_FLAGS;
-	if (Config::FullscreenEnabled()) {
-		window_flags |= static_cast<uint32_t>(SDL_WINDOW_FULLSCREEN_DESKTOP);
-	}
+	uint32_t window_flags = WindowContext::InitialWindowFlags(Config::FullscreenEnabled());
 #if defined(__APPLE__)
+	// SDL loads Vulkan while creating a Vulkan window, so select the bundled
+	// MoltenVK loader before calling SDL_CreateWindow. Keep an explicit user
+	// override, and fall back to SDL's normal loader search when no bundle is present.
+	if (std::getenv("SDL_VULKAN_LIBRARY") == nullptr) {
+		if (char* base_path = SDL_GetBasePath(); base_path != nullptr) {
+			const std::string base_path_str = base_path;
+			SDL_free(base_path);
+			std::string moltenvk_path = base_path_str + "libMoltenVK.dylib";
+			if (!Common::File::IsFileExisting(moltenvk_path)) {
+				moltenvk_path = base_path_str + "../Frameworks/libMoltenVK.dylib";
+			}
+			if (Common::File::IsFileExisting(moltenvk_path) &&
+			    SDL_setenv("SDL_VULKAN_LIBRARY", moltenvk_path.c_str(), 0) == 0) {
+				LOGF("Vulkan loader: %s\n", moltenvk_path.c_str());
+			}
+		}
+	}
+
 	// macOS 26 window chrome (CoreUI asset decode, SwiftUI titlebar) has been observed
 	// throwing NSExceptions under Rosetta during the first CATransaction commit. A
 	// borderless window skips that machinery entirely.
@@ -836,13 +858,20 @@ static void WindowCreate(WindowContext& context) {
 	context.window = SDL_CreateWindow(KYTY_SDL_WINDOW_CAPTION, KYTY_SDL_WINDOWPOS_CENTERED,
 	                                  KYTY_SDL_WINDOWPOS_CENTERED, width, height, window_flags);
 
-	context.window_hidden = true;
-
 	if (context.window == nullptr) {
 		EXIT("%s\n", SDL_GetError());
 	}
 
 	SDL_SetWindowResizable(context.window, SDL_FALSE);
+	context.UpdateIcon();
+}
+
+uint32_t WindowContext::InitialWindowFlags(bool fullscreen) noexcept {
+	auto flags = static_cast<uint32_t>(SDL_WINDOW_VULKAN);
+	if (fullscreen) {
+		flags |= static_cast<uint32_t>(SDL_WINDOW_FULLSCREEN_DESKTOP);
+	}
+	return flags;
 }
 
 Presenter& WindowInit(uint32_t width, uint32_t height) {
@@ -866,10 +895,12 @@ void WindowRun() {
 	EXIT_IF(g_window == nullptr);
 
 	g_window->Run();
+	g_window->render_context->GetPipelineCache().Save();
 }
 
 void WindowShutdown() {
 	if (g_window != nullptr) {
+		Controller::EmergencyShutdown();
 		g_window.reset();
 	}
 }
@@ -959,6 +990,7 @@ void WindowContext::UpdateTitle() {
 	    Loader::SystemContentParamSfoGetString("TITLE_ID", title_id, sizeof(title_id));
 	static bool has_app_ver =
 	    Loader::SystemContentParamSfoGetString("APP_VER", app_ver, sizeof(app_ver));
+	static const std::string processor_name = Common::GetSystemInfo().ProcessorName;
 	static uint64_t fps_start   = Common::Timer::QueryPerformanceCounter();
 	static uint64_t frame_num   = 0;
 	static uint64_t fps_frames  = 0;
@@ -983,19 +1015,14 @@ void WindowContext::UpdateTitle() {
 		fps_frames  = 0;
 	}
 
-	auto fps = fmt::format(
+	const auto* device_name = graphic_ctx.GetPhysicalDeviceProperties().deviceName.data();
+	auto text = fmt::format(
 	    "[{} | {}] {}{}{}{}{}{}[{}] [{}], frame: {}, fps: {:f}", KYTY_BUILD_LABEL, build_type,
 	    (has_title ? title : ""), (has_title ? ", " : ""), (has_title_id ? title_id : ""),
 	    (has_title_id ? ", " : ""), (has_app_ver ? app_ver : ""), (has_app_ver ? " " : ""),
 	    device_name, processor_name, frame_num, current_fps);
 
-#if defined(__APPLE__)
-	// AppKit traps on title changes off the main thread; fire-and-forget keeps present pacing.
-	RunOnMainThread([this, fps = std::move(fps)] { SDL_SetWindowTitle(window, fps.c_str()); },
-	                false);
-#else
-	SDL_SetWindowTitle(window, fps.c_str());
-#endif
+	RunOnMainThread([this, text = std::move(text)] { SDL_SetWindowTitle(window, text.c_str()); });
 }
 
 } // namespace Libs::Graphics

src/graphics/presentation/window/windowInternal.h

@@ -9,13 +9,10 @@
 
 #include <atomic>
 #include <cstdint>
+#include <functional>
 #include <memory>
 #include <vector>
 
-#if defined(__APPLE__)
-#include <functional>
-#endif
-
 namespace Libs::Graphics {
 
 class Presenter;
@@ -38,7 +35,9 @@ struct WindowContext {
 	~WindowContext();
 	KYTY_CLASS_NO_COPY(WindowContext);
 
+	[[nodiscard]] static vk::PhysicalDeviceVulkan12Features RequiredVulkan12Features() noexcept;
 	[[nodiscard]] static vk::PhysicalDeviceVulkan13Features RequiredVulkan13Features() noexcept;
+	[[nodiscard]] static uint32_t InitialWindowFlags(bool fullscreen) noexcept;
 	void                                                    CreateVulkan();
 	void                                                    RecreateSurface();
 	void                                                    RefreshSurfaceCapabilities();
@@ -49,36 +48,25 @@ struct WindowContext {
 	void ProcessDisplayEvent(const SDL_DisplayEvent& event);
 	void ProcessEvent(double time_seconds);
 	void Run();
-
-#if defined(__APPLE__)
-	// AppKit only allows window operations (show, icon, title, view/layer changes) on the
-	// main thread; the present thread marshals them through the SDL main loop with these.
-	// wait=true blocks until the task has run on the main thread.
-	void RunOnMainThread(std::function<void()> task, bool wait);
+	// SDL window operations must complete on the main thread.
+	void RunOnMainThread(std::function<void()> task);
 	void DrainMainThreadTasks();
-#endif
 
 	GraphicContext                 graphic_ctx;
 	SDL_Window*                    window        = nullptr;
-	bool                           window_hidden = true;
 	vk::SurfaceKHR                 surface       = nullptr;
 	SurfaceCapabilities            surface_capabilities;
 	std::unique_ptr<RenderContext> render_context;
 	std::unique_ptr<Presenter>     presenter;
 	WindowLoopState                loop;
 
-	char device_name[VK_MAX_PHYSICAL_DEVICE_NAME_SIZE] = {0};
-	char processor_name[64]                            = {0};
-
 	Common::Mutex mutex;
 
-#if defined(__APPLE__)
 	Common::Mutex                      main_task_mutex;
 	Common::CondVar                    main_task_done;
 	std::vector<std::function<void()>> main_tasks;            // guarded by main_task_mutex
 	uint64_t                           main_tasks_queued = 0; // guarded by main_task_mutex
 	uint64_t                           main_tasks_run    = 0; // guarded by main_task_mutex
-#endif
 };
 
 } // namespace Libs::Graphics

src/graphics/shader/recompiler/ShaderRecompiler.cpp

@@ -30,21 +30,13 @@ namespace Libs::Graphics::ShaderRecompiler {
 
 namespace {
 
-bool ReadZeroMemory(void*, uint64_t, uint32_t* value) {
-	if (value == nullptr) {
-		return false;
-	}
-	*value = 0;
-	return true;
-}
-
 const char* GetDumpLabel(const CompileOptions& options) {
 	return options.dump_label != nullptr ? options.dump_label : "ShaderRecompiler";
 }
 
-std::string MakeIrDump(const CFG::Graph& cfg, const IR::Program& ir) {
+std::string MakeIrDump(std::string_view cfg, const IR::Program& ir) {
 	std::string dump = "CFG:\n";
-	dump += CFG::GraphToString(cfg);
+	dump += cfg;
 	dump += "\nIR:\n";
 	dump += fmt::format("mode={} scratch_dwords={}\n",
 	                    ir.dispatcher_fallback ? "dispatcher" : "structured", ir.scratch_dwords);
@@ -242,8 +234,10 @@ EmbeddedFetchData DetectEmbeddedVertexFetch(const Decoder::Program&      decoded
                                             uint32_t wave_size) {
 	EmbeddedFetchData data;
 	data.loads.reserve(input_info->resources_num);
-	int32_t offset_candidate = -1;
-	bool    offset_conflict  = false;
+	int32_t vertex_offset_candidate   = -1;
+	int32_t instance_offset_candidate = -1;
+	bool    vertex_offset_conflict    = false;
+	bool    instance_offset_conflict  = false;
 
 	const int shift_regs = 8;
 	const int attrib_reg = input_info->fetch_attrib_reg + shift_regs;
@@ -271,26 +265,34 @@ EmbeddedFetchData DetectEmbeddedVertexFetch(const Decoder::Program&      decoded
 
 	for (const auto& inst: decoded.instructions) {
 		// Fetch shaders accumulate the draw's vertex offset in v0. The PS5 NGG ABI
-		// seeds S_NGG_VERTEX_INDEX in v5 and applies the same offset there before fetching.
+		// seeds S_NGG_VERTEX_INDEX in v5 and S_NGG_INSTANCE_INDEX in v8, then applies the
+		// corresponding direct-draw offsets before fetching.
 		const bool vertex_index_accumulator =
 		    IsDecodedVgpr(inst.dst) &&
 		    (inst.dst.reg == 0 || (user_data_base == 8 && inst.dst.reg == 5));
+		const bool instance_index_accumulator =
+		    IsDecodedVgpr(inst.dst) && (inst.dst.reg == (user_data_base == 8 ? 8u : 3u));
 		uint32_t   sad_zero = 0;
-		const bool vertex_offset_add =
-		    vertex_index_accumulator && IsDecodedSgpr(inst.src0) &&
+		const bool index_offset_add =
+		    (vertex_index_accumulator || instance_index_accumulator) &&
+		    IsDecodedSgpr(inst.src0) &&
 		    ((inst.opcode == Decoder::Opcode::V_ADD_I32 && IsDecodedVgpr(inst.src1) &&
 		      inst.src1.reg == inst.dst.reg) ||
-		     (user_data_base == 8 && inst.dst.reg == 5 &&
+		     (user_data_base == 8 && (inst.dst.reg == 5 || inst.dst.reg == 8) &&
 		      inst.opcode == Decoder::Opcode::V_SAD_U32 && IsDecodedVgpr(inst.src2) &&
 		      inst.src2.reg == inst.dst.reg &&
 		      TryDecodedOperandConstant(sgprs, inst.src1, sad_zero) && sad_zero == 0));
-		if (data.loads.empty() && vertex_offset_add) {
+		if (data.loads.empty() && index_offset_add) {
 			const auto reg = DecodedSgprReg(inst.src0);
 			if (reg >= user_data_base && reg - user_data_base < user_data_count) {
-				if (offset_candidate >= 0 && offset_candidate != static_cast<int32_t>(reg)) {
-					offset_conflict = true;
+				auto& candidate = vertex_index_accumulator ? vertex_offset_candidate
+				                                           : instance_offset_candidate;
+				auto& conflict  = vertex_index_accumulator ? vertex_offset_conflict
+				                                           : instance_offset_conflict;
+				if (candidate >= 0 && candidate != static_cast<int32_t>(reg)) {
+					conflict = true;
 				} else {
-					offset_candidate = static_cast<int32_t>(reg);
+					candidate = static_cast<int32_t>(reg);
 				}
 			}
 		}
@@ -449,8 +451,13 @@ EmbeddedFetchData DetectEmbeddedVertexFetch(const Decoder::Program&      decoded
 						load.attrib_id    = buffer.attrib_id;
 						load.components   = DecodedDstSize(inst);
 						load.prolog_loads = buffer.prolog_loads;
-						if (data.loads.empty() && !offset_conflict) {
-							data.vertex_offset_sgpr = offset_candidate;
+						if (data.loads.empty()) {
+							if (!vertex_offset_conflict) {
+								data.vertex_offset_sgpr = vertex_offset_candidate;
+							}
+							if (!instance_offset_conflict) {
+								data.instance_offset_sgpr = instance_offset_candidate;
+							}
 						}
 						data.loads.push_back(load);
 					}
@@ -472,20 +479,14 @@ EmbeddedFetchData DetectEmbeddedVertexFetch(const Decoder::Program&      decoded
 
 } // namespace
 
-bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
-                  CompileResult& result, std::string* error) {
+TranslateResult TranslateProgram(std::span<const uint32_t> code, const CompileOptions& options) {
 	if (code.empty()) {
-		if (error != nullptr) {
-			*error = "invalid shader recompiler input";
-		}
-		return false;
+		EXIT("shader recompiler input is empty\n");
 	}
 	if (options.stage != ShaderType::Compute && options.stage != ShaderType::Vertex &&
 	    options.stage != ShaderType::Pixel) {
-		if (error != nullptr) {
-			*error = "shader recompiler supports compute, vertex, and pixel stages";
-		}
-		return false;
+		EXIT("shader recompiler received unsupported stage %u\n",
+		     static_cast<unsigned>(options.stage));
 	}
 
 	const auto compile_begin = std::chrono::steady_clock::now();
@@ -500,9 +501,7 @@ bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
 	     static_cast<uint64_t>(code.size()));
 
 	Decoder::Program decoded;
-	if (!Decoder::DecodeProgram(code, decoded, error)) {
-		return false;
-	}
+	Decoder::DecodeProgram(code, decoded);
 	LOGF("%s phase end: stage=%s hash=0x%016" PRIx64 " decode instructions=%" PRIu64
 	     " elapsed_ms=%" PRIu64 "\n",
 	     GetDumpLabel(options), StageName(options.stage), options.shader_hash,
@@ -516,12 +515,9 @@ bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
 		}
 	}
 
-	CFG::Graph cfg;
 	LOGF("%s phase begin: stage=%s hash=0x%016" PRIx64 " CFG BuildGraph\n", GetDumpLabel(options),
 	     StageName(options.stage), options.shader_hash);
-	if (!CFG::BuildGraph(decoded, cfg, error)) {
-		return false;
-	}
+	auto cfg = CFG::BuildGraph(decoded);
 	LOGF("%s phase end: stage=%s hash=0x%016" PRIx64 " CFG BuildGraph blocks=%" PRIu64
 	     " loops=%" PRIu64 " back_edges=%" PRIu64 " elapsed_ms=%" PRIu64 "\n",
 	     GetDumpLabel(options), StageName(options.stage), options.shader_hash,
@@ -534,13 +530,12 @@ bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
 		dispatcher_reason   = cfg.unsupported_reason;
 		LogDispatcherFallback(options, cfg, "build", dispatcher_reason);
 	} else {
-		std::string structure_error;
-		const auto  unstructured_cfg = cfg;
+		const auto unstructured_cfg = cfg;
 		LOGF("%s phase begin: stage=%s hash=0x%016" PRIx64 " CFG Structurize\n",
 		     GetDumpLabel(options), StageName(options.stage), options.shader_hash);
-		if (!CFG::Structurize(cfg, &structure_error)) {
+		if (!CFG::Structurize(cfg)) {
 			dispatcher_fallback      = true;
-			dispatcher_reason        = structure_error;
+			dispatcher_reason        = cfg.unsupported_reason;
 			const auto failure_kind  = cfg.failure_kind;
 			const auto failure_block = cfg.failure_block;
 			LogDispatcherFallback(options, cfg, "structurize", dispatcher_reason);
@@ -548,7 +543,7 @@ bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
 			cfg.unsupported        = true;
 			cfg.failure_kind       = failure_kind;
 			cfg.failure_block      = failure_block;
-			cfg.unsupported_reason = structure_error;
+			cfg.unsupported_reason = dispatcher_reason;
 		} else {
 			LOGF("%s structured CFG success: blocks=%" PRIu64 "\n", GetDumpLabel(options),
 			     static_cast<uint64_t>(cfg.blocks.size()));
@@ -560,20 +555,26 @@ bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
 		     static_cast<uint64_t>(cfg.natural_loops.size()), phase_ms());
 	}
 
-	IR::Program ir;
 	const ShaderVertexInputInfo*  vertex  = nullptr;
 	const ShaderPixelInputInfo*   pixel   = nullptr;
 	const ShaderComputeInputInfo* compute = nullptr;
 	switch (options.stage) {
-		case ShaderType::Vertex: vertex = options.input_info.vertex; break;
-		case ShaderType::Pixel: pixel = options.input_info.pixel; break;
-		case ShaderType::Compute: compute = options.input_info.compute; break;
+		case ShaderType::Vertex:
+			vertex = options.input_info.vertex;
+			break;
+		case ShaderType::Pixel:
+			pixel = options.input_info.pixel;
+			break;
+		case ShaderType::Compute:
+			compute = options.input_info.compute;
+			break;
 		default: break;
 	}
 	EmbeddedFetchData embedded_fetch;
 	if (options.stage == ShaderType::Vertex && vertex->fetch_embedded) {
 		embedded_fetch = DetectEmbeddedVertexFetch(decoded, vertex, options.user_data_base,
-		                                           options.user_data_count, options.wave_size);
+		                                           static_cast<uint32_t>(options.user_data.size()),
+		                                           options.wave_size);
 		if (!embedded_fetch.loads.empty()) {
 			LOGF("%s embedded vertex fetch plan: detected=%" PRIu64 "\n", GetDumpLabel(options),
 			     static_cast<uint64_t>(embedded_fetch.loads.size()));
@@ -584,7 +585,7 @@ bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
 	    .wave_size           = options.wave_size,
 	    .shader_hash         = options.shader_hash,
 	    .user_data_base      = options.user_data_base,
-	    .user_data_count     = options.user_data_count,
+	    .user_data_count     = static_cast<uint32_t>(options.user_data.size()),
 	    .scratch_dwords      = options.scratch_dwords,
 	    .dispatcher_fallback = dispatcher_fallback,
 	    .cfg_failure_kind    = cfg.failure_kind,
@@ -597,9 +598,7 @@ bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
 	};
 	LOGF("%s phase begin: stage=%s hash=0x%016" PRIx64 " IR TranslateProgram\n",
 	     GetDumpLabel(options), StageName(options.stage), options.shader_hash);
-	if (!Frontend::TranslateProgram(decoded, cfg, translate_options, ir, error)) {
-		return false;
-	}
+	auto ir = Frontend::TranslateProgram(decoded, cfg, translate_options);
 	LOGF("%s phase end: stage=%s hash=0x%016" PRIx64 " IR TranslateProgram blocks=%" PRIu64
 	     " elapsed_ms=%" PRIu64 "\n",
 	     GetDumpLabel(options), StageName(options.stage), options.shader_hash,
@@ -626,104 +625,74 @@ bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
 			     lds_barriers.inserted_barriers);
 		}
 	}
-	std::string srt_error;
-	if (!IR::BuildSrtPlan(ir, &srt_error)) {
-		LOGF("%s SRT planning failed: %s\n", GetDumpLabel(options), srt_error.c_str());
-		if (error != nullptr) {
-			*error = std::move(srt_error);
-		}
-		return false;
-	}
+	IR::BuildSrtPlan(ir);
 	IR::EliminateDeadCode(ir.blocks);
-	if (!IR::TrackResources(ir, error)) {
-		return false;
-	}
+	IR::TrackResources(ir);
 	IR::EliminateDeadCode(ir.blocks);
 	if (options.stage == ShaderType::Vertex) {
-		ir.info.vertex_offset_sgpr = embedded_fetch.vertex_offset_sgpr;
+		ir.info.vertex_offset_sgpr   = embedded_fetch.vertex_offset_sgpr;
+		ir.info.instance_offset_sgpr = embedded_fetch.instance_offset_sgpr;
 	}
 
-	IR::ResourceSnapshot resources;
-	if (options.resource_snapshot != nullptr) {
-		resources = *options.resource_snapshot;
-	} else {
-		IR::SrtRuntime           runtime;
-		std::array<uint32_t, 64> zero_user_data {};
-		if (options.user_data == nullptr) {
-			for (uint32_t i = 2; i < zero_user_data.size(); i += 4) {
-				zero_user_data[i] = UINT32_MAX;
-			}
-		}
-		const auto* user_data =
-		    options.user_data != nullptr ? options.user_data : zero_user_data.data();
-		runtime.user_data   = std::span<const uint32_t>(user_data, options.user_data_count);
-		runtime.shader_base = options.shader_base != 0 ? options.shader_base
-		                                               : reinterpret_cast<uint64_t>(code.data());
-		runtime.read_memory = options.read_memory;
-		runtime.read_specialization_memory = options.read_specialization_memory;
-		if (runtime.read_memory == nullptr && options.user_data == nullptr) {
-			runtime.read_memory = ReadZeroMemory;
-		}
-		runtime.userdata         = options.read_memory_data;
-		if (!IR::MaterializeResources(ir, runtime, resources, error)) {
-			return false;
-		}
+	TranslateResult result;
+	result.program = std::move(ir);
+	if (options.dump_ir) {
+		result.decoded_dump = std::move(decoded_dump);
+		result.cfg_dump     = CFG::GraphToString(cfg);
 	}
-	if (!IR::SpecializeResources(ir, resources, error)) {
-		return false;
+	return result;
+}
+
+CompileResult CompileProgram(TranslateResult translated, const CompileOptions& options,
+                             const IR::ResourceSpecialization& specialization,
+                             uint32_t push_data_start_dword) {
+	const auto emit_begin = std::chrono::steady_clock::now();
+	auto& ir = translated.program;
+	IR::ApplyResourceSpecialization(ir, specialization);
+
+	const ShaderVertexInputInfo*  vertex  = nullptr;
+	const ShaderPixelInputInfo*   pixel   = nullptr;
+	const ShaderComputeInputInfo* compute = nullptr;
+	switch (options.stage) {
+		case ShaderType::Vertex: vertex = options.input_info.vertex; break;
+		case ShaderType::Pixel: pixel = options.input_info.pixel; break;
+		case ShaderType::Compute: compute = options.input_info.compute; break;
+		default: EXIT("invalid shader stage\n");
 	}
 
 	IR::ShaderInfoOptions info_options;
 	info_options.vertex  = vertex;
 	info_options.pixel   = pixel;
 	info_options.compute = compute;
-	if (!IR::CollectShaderInfo(ir, info_options, error)) {
-		return false;
-	}
-	if (!IR::AllocateBindings(ir, options.push_constant_offset, error)) {
-		return false;
-	}
-	if (!Spirv::AnalyzeProgramRequirements(ir, error)) {
-		return false;
-	}
+	IR::CollectShaderInfo(ir, info_options);
+	IR::AllocateBindings(ir, push_data_start_dword);
+	Spirv::AnalyzeProgramRequirements(ir);
 	std::string ir_dump;
 	if (options.dump_ir) {
-		ir_dump = MakeIrDump(cfg, ir);
+		ir_dump = MakeIrDump(translated.cfg_dump, ir);
 		if (options.early_dump) {
 			LOGF("%s native IR and bindings (early):\n%s", GetDumpLabel(options), ir_dump.c_str());
 		}
 	}
 
-	std::vector<uint32_t> spirv;
-	std::string           emit_error;
 	LOGF("%s phase begin: stage=%s hash=0x%016" PRIx64 " SPIR-V EmitProgram\n",
 	     GetDumpLabel(options), StageName(options.stage), options.shader_hash);
-	if (!Spirv::EmitProgram(ir, resources, options.input_info, spirv, &emit_error)) {
-		LOGF("%s typed SPIR-V emit failed: %s\n", GetDumpLabel(options), emit_error.c_str());
-		if (dispatcher_fallback && error != nullptr) {
-			*error = fmt::format("dispatcher fallback failed after {}: {}",
-			                     dispatcher_reason.c_str(), emit_error.c_str());
-			LOGF("%s dispatcher fallback emit failed: %s\n", GetDumpLabel(options), error->c_str());
-		} else if (error != nullptr) {
-			*error = emit_error;
-		}
-		return false;
-	}
+	auto spirv = Spirv::EmitProgram(ir, options.input_info);
 	LOGF("%s phase end: stage=%s hash=0x%016" PRIx64 " SPIR-V EmitProgram words=%" PRIu64
 	     " elapsed_ms=%" PRIu64 "\n",
 	     GetDumpLabel(options), StageName(options.stage), options.shader_hash,
-	     static_cast<uint64_t>(spirv.size()), phase_ms());
-	result.spirv     = std::move(spirv);
-	result.program   = std::move(ir);
-	result.resources = std::move(resources);
+	     static_cast<uint64_t>(spirv.size()),
+	     static_cast<uint64_t>(std::chrono::duration_cast<std::chrono::milliseconds>(
+	                               std::chrono::steady_clock::now() - emit_begin)
+	                               .count()));
+	CompileResult result;
+	result.spirv   = std::move(spirv);
+	result.program = std::move(ir);
 	if (options.dump_ir) {
-		result.decoded_dump = std::move(decoded_dump);
+		result.decoded_dump = std::move(translated.decoded_dump);
 		result.ir_dump      = std::move(ir_dump);
-	} else {
-		result.decoded_dump.clear();
-		result.ir_dump.clear();
 	}
-	return true;
+	return result;
 }
 
 } // namespace Libs::Graphics::ShaderRecompiler

src/graphics/shader/recompiler/ShaderRecompiler.h

@@ -15,32 +15,34 @@ struct CompileOptions {
 	ShaderType                  stage           = ShaderType::Compute;
 	uint32_t                    wave_size       = 64;
 	uint32_t                    user_data_base  = 0;
-	uint32_t                    user_data_count = 64;
 	uint32_t                    scratch_dwords  = 0;
 	uint64_t                    shader_hash     = 0;
-	uint64_t                    shader_base     = 0;
-	uint32_t                    push_constant_offset      = 0;
 	bool                        dump_ir                    = true;
 	bool                        early_dump                 = false;
 	const char*                 dump_label                 = nullptr;
-	const uint32_t*             user_data                  = nullptr;
-	IR::SrtMemoryReader         read_memory                = nullptr;
-	IR::SrtMemoryReader         read_specialization_memory = nullptr;
-	void*                       read_memory_data           = nullptr;
-	const IR::ResourceSnapshot* resource_snapshot          = nullptr;
+	std::span<const uint32_t>   user_data;
 	ShaderStageInputInfo        input_info;
 };
 
-struct CompileResult {
-	std::vector<uint32_t> spirv;
-	std::string           decoded_dump;
-	std::string           ir_dump;
-	IR::Program           program;
-	IR::ResourceSnapshot  resources;
+struct TranslateResult {
+	IR::Program program;
+	std::string decoded_dump;
+	std::string cfg_dump;
 };
 
-bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
-                  CompileResult& result, std::string* error);
+struct CompileResult {
+	std::vector<uint32_t>  spirv;
+	std::string            decoded_dump;
+	std::string            ir_dump;
+	IR::Program            program;
+};
+
+[[nodiscard]] TranslateResult TranslateProgram(std::span<const uint32_t> code,
+                                               const CompileOptions& options);
+[[nodiscard]] CompileResult CompileProgram(TranslateResult translated,
+                                           const CompileOptions& options,
+                                           const IR::ResourceSpecialization& specialization,
+	                                       uint32_t push_data_start_dword = 0);
 
 } // namespace Libs::Graphics::ShaderRecompiler
 

src/graphics/shader/recompiler/backend/spirv/SpirvEmitter.cpp

@@ -1,19 +1,23 @@
 #include "graphics/shader/recompiler/backend/spirv/SpirvEmitter.h"
 
+#include "common/assert.h"
 #include "graphics/shader/recompiler/backend/spirv/spirvEmitterInternal.h"
 #include "graphics/shader/recompiler/ir/ShaderIR.h"
-#include "graphics/shader/shader.h"
 
 #include <algorithm>
 #include <array>
 
 namespace Libs::Graphics::ShaderRecompiler::Spirv {
 
-using ShaderError::Fail;
-
 namespace {
 
-bool ValidateNativeProgram(const IR::Program& program, std::string* error) {
+[[noreturn]] void Fail(const IR::Program& program, const char* reason) {
+	EXIT("SPIR-V validation failed: hash=0x%016" PRIx64 " stage=%u reason=%s\n",
+	     program.shader_hash, static_cast<unsigned>(program.stage), reason);
+	std::abort();
+}
+
+void ValidateNativeProgram(const IR::Program& program) {
 	using Kind                                             = IR::DescriptorBindingKind;
 	constexpr auto                               KindCount = static_cast<size_t>(Kind::Count);
 	std::array<std::vector<uint32_t>, KindCount> expected;
@@ -36,13 +40,13 @@ bool ValidateNativeProgram(const IR::Program& program, std::string* error) {
 	for (uint32_t i = 0; i < program.info.images.size(); i++) {
 		const auto kind = IR::DescriptorBindingForImage(program.info.images[i]);
 		if (!kind.has_value()) {
-			return Fail(error, "native shader plan has an invalid image class");
+			Fail(program, "native shader plan has an invalid image class");
 		}
 		present[static_cast<size_t>(*kind)] = true;
 		const auto dynamic = program.info.images[i].mip_mode == IR::ImageMipMode::DynamicStorage;
 		const auto count   = dynamic ? program.info.images[i].mip_count : 1u;
 		if (count == 0u || (!dynamic && program.info.images[i].mip_count != 1u)) {
-			return Fail(error, "native shader plan has an invalid image mip descriptor count");
+			Fail(program, "native shader plan has an invalid image mip descriptor count");
 		}
 		expected[static_cast<size_t>(*kind)].insert(expected[static_cast<size_t>(*kind)].end(),
 		                                            count, i);
@@ -58,11 +62,11 @@ bool ValidateNativeProgram(const IR::Program& program, std::string* error) {
 			}
 			const auto index = inst.Flags<IR::MemoryFlags>().index;
 			if (index >= program.memory_info.size()) {
-				return Fail(error, "shared operation has invalid memory metadata");
+				Fail(program, "shared operation has invalid memory metadata");
 			}
 			const auto kind = program.memory_info[index].kind;
 			if (kind != IR::ResourceKind::Lds && kind != IR::ResourceKind::Gds) {
-				return Fail(error, "shared operation has invalid resource kind");
+				Fail(program, "shared operation has invalid resource kind");
 			}
 			uses_gds |= kind == IR::ResourceKind::Gds;
 		}
@@ -77,13 +81,13 @@ bool ValidateNativeProgram(const IR::Program& program, std::string* error) {
 	const bool uses_flattened_runtime =
 	    !program.srt_reads.empty() ||
 	     std::ranges::any_of(program.info.images, [](const IR::ImageResource& image) {
-		     return image.indirect_mapping_capacity != 0u;
+		     return image.indirect_search_iterations != 0u;
 	     });
 	if (uses_flattened_runtime) {
 		Expect(Kind::FlattenedSrt);
 	}
-	if (program.bindings.ShaderDataDwords() != 0 && program.bindings.push_constant_size == 0) {
-		Expect(Kind::UserData);
+	if (program.bindings.ShaderDataDwords() != 0 && !program.bindings.UsesPushData()) {
+		Expect(Kind::ShaderData);
 	}
 
 	std::array<bool, KindCount> seen {};
@@ -91,30 +95,28 @@ bool ValidateNativeProgram(const IR::Program& program, std::string* error) {
 		const auto kind = static_cast<size_t>(binding.kind);
 		if (kind >= KindCount || seen[kind] || !present[kind] ||
 		    binding.resources != expected[kind]) {
-			return Fail(error, "native descriptor groups do not match shader topology");
+			Fail(program, "native descriptor groups do not match shader topology");
 		}
 		seen[kind] = true;
 	}
 	for (size_t i = 0; i < KindCount; i++) {
 		if (present[i] != seen[i]) {
-			return Fail(error, "native shader plan is missing a required descriptor group");
+			Fail(program, "native shader plan is missing a required descriptor group");
 		}
 	}
-	const auto has_user_storage = present[static_cast<size_t>(Kind::UserData)];
-	if (program.bindings.push_constant_offset % sizeof(uint32_t) != 0 ||
-	    program.bindings.push_constant_offset + program.bindings.push_constant_size >
-	        IR::NativePushConstantSize ||
+	const auto has_shader_data_storage = present[static_cast<size_t>(Kind::ShaderData)];
+	const auto shader_data_dwords = program.bindings.ShaderDataDwords();
+	if ((program.bindings.UsesPushData() &&
+	     !IR::PushData::CanFit(program.bindings.push_data_start_dword, shader_data_dwords)) ||
 	    program.bindings.memory_offset_dword != program.bindings.user_data_registers.size() ||
 	    program.bindings.memory_offset_count != program.info.buffers.size() ||
-	    (!has_user_storage &&
-	     program.bindings.push_constant_size != program.bindings.ShaderDataDwords() * 4u) ||
-	    (has_user_storage && program.bindings.push_constant_size != 0) ||
+	    has_shader_data_storage != (shader_data_dwords != 0 && !program.bindings.UsesPushData()) ||
 	    !std::is_sorted(program.bindings.user_data_registers.begin(),
 	                    program.bindings.user_data_registers.end()) ||
 	    std::adjacent_find(program.bindings.user_data_registers.begin(),
 	                       program.bindings.user_data_registers.end()) !=
 	        program.bindings.user_data_registers.end()) {
-		return Fail(error, "native user-data layout is inconsistent");
+		Fail(program, "native shader-data layout is inconsistent");
 	}
 
 	const auto planning_only_handle = [&](const IR::Inst& handle) {
@@ -139,7 +141,7 @@ bool ValidateNativeProgram(const IR::Program& program, std::string* error) {
 						break;
 					}
 					if (dense >= program.info.buffers.size()) {
-						return Fail(error, "typed buffer handle has an invalid dense resource");
+						Fail(program, "typed buffer handle has an invalid dense resource");
 					}
 					break;
 				case IR::ValueOpcode::GetAddressResource:
@@ -147,29 +149,29 @@ bool ValidateNativeProgram(const IR::Program& program, std::string* error) {
 						break;
 					}
 					if (inst.NumArgs() != 2 || !program.info.uses_dma) {
-						return Fail(error, "typed address handle has invalid DMA metadata");
+						Fail(program, "typed address handle has invalid DMA metadata");
 					}
 					break;
 				case IR::ValueOpcode::GetScratchResource:
 					if (inst.NumArgs() != 0 || program.scratch_dwords == 0) {
-						return Fail(error, "typed scratch handle has invalid shader metadata");
+						Fail(program, "typed scratch handle has invalid shader metadata");
 					}
 					break;
 				case IR::ValueOpcode::GetImageResource:
 					if (dense >= program.info.images.size()) {
-						return Fail(error, "typed image handle has an invalid dense resource");
+						Fail(program, "typed image handle has an invalid dense resource");
 					}
 					break;
 				case IR::ValueOpcode::GetSamplerResource:
 					if (dense >= program.info.samplers.size()) {
-						return Fail(error, "typed sampler handle has an invalid dense resource");
+						Fail(program, "typed sampler handle has an invalid dense resource");
 					}
 					break;
 				case IR::ValueOpcode::ReadConst: {
 					const auto slot = inst.Arg(1).Resolve();
 					if (!slot.IsImmediate() || slot.GetType() != IR::Type::U32 ||
 					    slot.U32() >= program.srt_reads.size()) {
-						return Fail(error, "flattened SRT read has an invalid dense slot");
+						Fail(program, "flattened SRT read has an invalid dense slot");
 					}
 					break;
 				}
@@ -177,45 +179,48 @@ bool ValidateNativeProgram(const IR::Program& program, std::string* error) {
 			}
 		}
 	}
-	return true;
 }
 
 } // namespace
 
-bool AnalyzeProgramRequirements(IR::Program& program, std::string* error) {
+void AnalyzeProgramRequirements(IR::Program& program) {
 	program.spirv_requirements.reset();
 	IR::SpirvRequirements requirements {};
 	const auto MarkBallot = [&] { requirements.subgroup_ballot = true; };
 	for (const auto* block: program.blocks) {
 		for (const auto& inst: *block) {
+			if (IR::BufferAccessOf(inst.GetOpcode()) == IR::BufferAccess::Atomic &&
+			    inst.GetType() == IR::Type::U64) {
+				requirements.buffer_int64_atomics = true;
+			}
 			const auto address_access = IR::AddressOpcodeInfoOf(inst.GetOpcode()).access;
 			if (address_access != IR::AddressAccess::None) {
 				const auto memory_index = inst.Flags<IR::MemoryFlags>().index;
 				if (memory_index >= program.memory_info.size()) {
-					return Fail(error, "address operation has invalid memory metadata");
+					Fail(program, "address operation has invalid memory metadata");
 				}
 				if (program.memory_info[memory_index].kind == IR::ResourceKind::Scratch) {
 					if (program.scratch_dwords == 0) {
-						return Fail(error, "scratch operation has no per-thread storage");
+						Fail(program, "scratch operation has no per-thread storage");
 					}
 					requirements.function_scratch = true;
 				} else if (address_access == IR::AddressAccess::Write) {
-					return Fail(error, "writable FLAT/GLOBAL addresses require GPU ownership tracking");
+					Fail(program, "writable FLAT/GLOBAL addresses require GPU ownership tracking");
 				}
 			}
 			if (IR::BufferAccessOf(inst.GetOpcode()) != IR::BufferAccess::None) {
 				const auto memory_index = inst.Flags<IR::MemoryFlags>().index;
 				if (memory_index >= program.memory_info.size()) {
-					return Fail(error, "buffer operation has invalid memory metadata");
+					Fail(program, "buffer operation has invalid memory metadata");
 				}
 				const auto& memory = program.memory_info[memory_index];
 				if (memory.kind == IR::ResourceKind::Buffer) {
 					if (memory.resource >= program.info.buffers.size()) {
-						return Fail(error, "buffer operation has invalid resource metadata");
+						Fail(program, "buffer operation has invalid resource metadata");
 					}
 					if ((program.info.buffers[memory.resource].packed_stride & (1u << 20u)) != 0u) {
 						if (program.stage != ShaderType::Compute) {
-							return Fail(error, "buffer ADD_TID is only valid for compute shaders");
+							Fail(program, "buffer ADD_TID is only valid for compute shaders");
 						}
 						requirements.subgroup_local_invocation_id = true;
 					}
@@ -225,11 +230,11 @@ bool AnalyzeProgramRequirements(IR::Program& program, std::string* error) {
 			if (shared_access != IR::SharedAccess::None) {
 				const auto index = inst.Flags<IR::MemoryFlags>().index;
 				if (index >= program.memory_info.size()) {
-					return Fail(error, "shared operation has invalid memory metadata");
+					Fail(program, "shared operation has invalid memory metadata");
 				}
 				const auto kind = program.memory_info[index].kind;
 				if (kind != IR::ResourceKind::Lds && kind != IR::ResourceKind::Gds) {
-					return Fail(error, "shared operation has invalid resource kind");
+					Fail(program, "shared operation has invalid resource kind");
 				}
 				if (program.stage != ShaderType::Compute && kind == IR::ResourceKind::Lds) {
 					requirements.function_lds = true;
@@ -266,7 +271,8 @@ bool AnalyzeProgramRequirements(IR::Program& program, std::string* error) {
 					requirements.subgroup_local_invocation_id = true;
 					break;
 				}
-				case IR::ValueOpcode::SwizzleU32: {
+				case IR::ValueOpcode::SwizzleU32:
+				case IR::ValueOpcode::BpermuteU32: {
 					MarkBallot();
 					requirements.subgroup_shuffle             = true;
 					requirements.subgroup_local_invocation_id = true;
@@ -282,7 +288,7 @@ bool AnalyzeProgramRequirements(IR::Program& program, std::string* error) {
 				case IR::ValueOpcode::SetAttribute: {
 					const auto index = inst.Flags<IR::ExportFlags>().index;
 					if (index >= program.export_info.size()) {
-						return Fail(error, "attribute export has invalid metadata");
+						Fail(program, "attribute export has invalid metadata");
 					}
 					if (program.stage == ShaderType::Pixel &&
 					    program.export_info[index].vm) {
@@ -295,38 +301,24 @@ bool AnalyzeProgramRequirements(IR::Program& program, std::string* error) {
 		}
 	}
 	program.spirv_requirements.emplace(requirements);
-	return true;
 }
 
-bool EmitProgram(const IR::Program& program, const IR::ResourceSnapshot& resources,
-                 ShaderStageInputInfo input_info, std::vector<uint32_t>& spirv,
-                 std::string* error) {
+std::vector<uint32_t> EmitProgram(const IR::Program& program,
+                                  ShaderStageInputInfo input_info) {
 	using namespace Emitter;
 
 	if (program.stage != ShaderType::Compute && program.stage != ShaderType::Vertex &&
 	    program.stage != ShaderType::Pixel) {
-		SetError(error, "binary SPIR-V emitter supports compute, vertex, and pixel shaders");
-		return false;
+		Fail(program, "binary SPIR-V emitter supports compute, vertex, and pixel shaders");
 	}
 	if (!program.srt_plan_complete || !program.resource_tracking_complete ||
 	    !program.shader_info_complete || !program.binding_layout_complete ||
 	    !program.spirv_requirements.has_value()) {
-		SetError(error, "SPIR-V emitter requires a fully planned native shader program");
-		return false;
+		Fail(program, "SPIR-V emitter requires a fully planned native shader program");
 	}
-	if (!IR::ValidateResourceSnapshot(program, resources, error)) {
-		return false;
-	}
-	if (!IR::ValidateResourceSpecialization(program, resources, error)) {
-		return false;
-	}
-	if (!ValidateNativeProgram(program, error)) {
-		return false;
-	}
-	if (!IR::ValidateProgram(program, true, error)) {
-		return false;
-	}
-	EmitterState state(program, resources, input_info);
+	ValidateNativeProgram(program);
+	IR::ValidateProgram(program, true);
+	EmitterState state(program, input_info);
 	state.stage     = program.stage;
 	state.wave_size = program.wave_size;
 	state.inputs.reserve(program.info.inputs.size());
@@ -336,17 +328,13 @@ bool EmitProgram(const IR::Program& program, const IR::ResourceSnapshot& resourc
 	AllocateInputVariables(state);
 	AllocateOutputVariables(state);
 	DefineModule(state);
-	if (!EmitProgram(state, program, error)) {
-		return false;
-	}
+	EmitProgram(state, program);
 
 	auto binary = state.builder.Build();
 	if (binary.empty()) {
-		SetError(error, "SPIR-V builder returned an empty module");
-		return false;
+		Fail(program, "SPIR-V builder returned an empty module");
 	}
-	spirv = std::move(binary);
-	return true;
+	return binary;
 }
 
 } // namespace Libs::Graphics::ShaderRecompiler::Spirv

src/graphics/shader/recompiler/backend/spirv/SpirvEmitter.h

@@ -3,16 +3,16 @@
 
 #include "common/common.h"
 #include "common/stringUtils.h"
-#include "graphics/shader/recompiler/ir/passes/ResourceMaterialization.h"
+#include "graphics/shader/recompiler/ir/ShaderIR.h"
 
 #include <vector>
 
 namespace Libs::Graphics::ShaderRecompiler::Spirv {
 
-bool AnalyzeProgramRequirements(IR::Program& program, std::string* error);
+void AnalyzeProgramRequirements(IR::Program& program);
 
-bool EmitProgram(const IR::Program& program, const IR::ResourceSnapshot& resources,
-                 ShaderStageInputInfo input_info, std::vector<uint32_t>& spirv, std::string* error);
+std::vector<uint32_t> EmitProgram(const IR::Program& program,
+                                  ShaderStageInputInfo input_info);
 
 } // namespace Libs::Graphics::ShaderRecompiler::Spirv
 

src/graphics/shader/recompiler/backend/spirv/spirvEmitterAnalysis.cpp

@@ -37,12 +37,6 @@ bool PixelParameterIsCustom(const EmitterState& state, uint32_t attr) {
 	       ShaderPixelParameterIsCustom(*state.input_info.pixel, attr);
 }
 
-void SetError(std::string* error, const char* message) {
-	if (error != nullptr) {
-		*error = message;
-	}
-}
-
 bool HasOutput(const std::vector<OutputBinding>& outputs, IR::StageOutputKind kind,
                uint32_t index) {
 	return std::any_of(outputs.begin(), outputs.end(), [kind, index](const OutputBinding& binding) {
@@ -92,8 +86,8 @@ uint32_t          ConstantU32(EmitterState& state, uint32_t value);
 	std::abort();
 }
 
-DescriptorResourceBinding ResourceForDescriptor(const EmitterState&       state,
-                                                IR::DescriptorBindingKind kind, uint32_t resource) {
+uint32_t ResourceForDescriptor(const EmitterState& state, IR::DescriptorBindingKind kind,
+                               uint32_t resource) {
 	const auto* descriptor = IR::FindBinding(state.program.bindings, kind);
 	if (descriptor == nullptr) {
 		ExitDescriptorBindingFailure(state, kind, resource, "descriptor group was not allocated");
@@ -104,7 +98,7 @@ DescriptorResourceBinding ResourceForDescriptor(const EmitterState&       state,
 		ExitDescriptorBindingFailure(state, kind, resource,
 		                             "resource is absent from descriptor group");
 	}
-	return {descriptor, static_cast<uint32_t>(found - descriptor->resources.begin())};
+	return static_cast<uint32_t>(found - descriptor->resources.begin());
 }
 
 uint32_t DescriptorElementPointer(EmitterState& state, uint32_t result_ptr_type,
@@ -120,68 +114,56 @@ uint32_t DescriptorElementPointer(EmitterState& state, uint32_t result_ptr_type,
 	return pointer;
 }
 
-ImageViewKind ImageViewKindFromDimension(Decoder::ImageDimension dimension) {
-	switch (dimension) {
-		case Decoder::ImageDimension::Dim1D: return ImageViewKind::Dim1D;
-		case Decoder::ImageDimension::Dim1DArray: return ImageViewKind::Dim1DArray;
-		case Decoder::ImageDimension::Dim2DArray: return ImageViewKind::Dim2DArray;
-		case Decoder::ImageDimension::Dim3D: return ImageViewKind::Dim3D;
-		case Decoder::ImageDimension::Dim2DMsaa: return ImageViewKind::Dim2DMsaa;
-		case Decoder::ImageDimension::Dim2DMsaaArray: return ImageViewKind::Dim2DMsaaArray;
-		default: return ImageViewKind::Dim2D;
+const ImageDimensionInfo& ImageDimensionInfoFor(ImageDimension dimension) {
+	for (const auto& info: ImageDimensions) {
+		if (info.dimension == dimension) {
+			return info;
+		}
 	}
+	EXIT("SPIR-V image dimension %u is invalid\n", static_cast<uint32_t>(dimension));
+	std::abort();
 }
 
-ImageViewKind SampledImageViewKind(const EmitterState& state, const IR::MemoryInfo& mem,
-                                   uint32_t use_pc) {
-	(void)state;
-	(void)use_pc;
-	return ImageViewKindFromDimension(mem.image_dimension);
-}
-
-ImageViewKind StorageImageViewKind(const IR::MemoryInfo& mem) {
-	return ImageViewKindFromDimension(mem.image_dimension);
-}
-
-uint32_t ImageViewCoordinateComponents(ImageViewKind view) {
-	switch (view) {
-		case ImageViewKind::Dim1D: return 1u;
-		case ImageViewKind::Dim1DArray:
-		case ImageViewKind::Dim2D: return 2u;
-		case ImageViewKind::Dim2DArray:
-		case ImageViewKind::Dim2DMsaaArray:
-		case ImageViewKind::Dim3D: return 3u;
-		case ImageViewKind::Dim2DMsaa: return 2u;
-		default: return 0u;
+uint32_t ImageScalarType(EmitterState& state, Prospero::TextureNumericClass numeric_class) {
+	switch (numeric_class) {
+		case Prospero::TextureNumericClass::Float: return TypeF32(state);
+		case Prospero::TextureNumericClass::Uint: return TypeU32(state);
+		case Prospero::TextureNumericClass::Sint: return TypeI32(state);
+		case Prospero::TextureNumericClass::Unsupported: break;
 	}
+	EXIT("invalid image numeric class");
 }
 
-uint32_t ImageViewSpatialComponents(ImageViewKind view) {
-	switch (view) {
-		case ImageViewKind::Dim1D:
-		case ImageViewKind::Dim1DArray: return 1u;
-		case ImageViewKind::Dim2D:
-		case ImageViewKind::Dim2DArray:
-		case ImageViewKind::Dim2DMsaa:
-		case ImageViewKind::Dim2DMsaaArray: return 2u;
-		case ImageViewKind::Dim3D: return 3u;
-		default: return 0u;
+uint32_t ImageVectorType(EmitterState& state, Prospero::TextureNumericClass numeric_class,
+                         uint32_t components) {
+	return state.builder.Type(OpTypeVector, {ImageScalarType(state, numeric_class), components});
+}
+
+uint32_t ImageType(EmitterState& state, const IR::ImageResource& image) {
+	uint32_t sampled = 0;
+	uint32_t format  = ImageFormatUnknown;
+	if (image.resource_class == IR::ImageResourceClass::Sampled) {
+		EXIT_IF(image.atomic);
+		sampled = 1;
+	} else if (image.resource_class == IR::ImageResourceClass::Storage) {
+		EXIT_IF(image.numeric_class == Prospero::TextureNumericClass::Sint ||
+		        image.numeric_class == Prospero::TextureNumericClass::Unsupported);
+		sampled = 2;
+		if (image.atomic) {
+			EXIT_IF(image.numeric_class != Prospero::TextureNumericClass::Uint);
+			format = ImageFormatR32ui;
+		}
+	} else {
+		EXIT("invalid image resource class");
 	}
-}
-
-uint32_t ImageViewImageType(EmitterState& state, ImageViewKind view, bool integer) {
-	const auto component = integer ? TypeU32(state) : TypeF32(state);
+	const auto& info = ImageDimensionInfoFor(image.dimension);
 	return state.builder.Type(OpTypeImage,
-	                          {component, ImageSpirvDimension(view), 0, ImageSpirvArrayed(view),
-	                           ImageSpirvMultisampled(view), 1, ImageFormatUnknown});
+	                          {ImageScalarType(state, image.numeric_class), info.spirv_dimension, 0,
+	                           info.arrayed, info.multisampled, sampled, format});
 }
 
-uint32_t ImageViewSampledImageType(EmitterState& state, ImageViewKind view, bool integer) {
-	return state.builder.Type(OpTypeSampledImage, {ImageViewImageType(state, view, integer)});
-}
-
-uint32_t ImageViewSizeType(EmitterState& state, ImageViewKind view) {
-	switch (ImageViewCoordinateComponents(view)) {
+uint32_t ImageViewSizeType(EmitterState& state, ImageDimension dimension) {
+	switch (ImageDimensionInfoFor(dimension).coordinate_components) {
 		case 1u: return TypeU32(state);
 		case 2u: return TypeU32Vector(state, 2);
 		case 3u: return TypeU32Vector(state, 3);
@@ -189,128 +171,89 @@ uint32_t ImageViewSizeType(EmitterState& state, ImageViewKind view) {
 	}
 }
 
-uint32_t StorageImageType(EmitterState& state, StorageImageClass image_class, ImageViewKind view) {
-	uint32_t component = TypeF32(state);
-	uint32_t format    = ImageFormatUnknown;
-	if (image_class == StorageImageClass::FormatlessUint) {
-		component = TypeU32(state);
-	} else if (image_class == StorageImageClass::AtomicUint) {
-		component = TypeU32(state);
-		format    = ImageFormatR32ui;
-	}
-	return state.builder.Type(OpTypeImage, {component, ImageSpirvDimension(view), 0,
-	                                        ImageSpirvArrayed(view), 0, 2, format});
-}
-
-uint32_t StorageImagePointerType(EmitterState& state, StorageImageClass image_class,
-                                 ImageViewKind view) {
-	return state.builder.Type(
-	    OpTypePointer, {StorageClassUniformConstant, StorageImageType(state, image_class, view)});
-}
-
-StorageImageClass StorageImageClassForResource(const EmitterState& state, uint32_t resource) {
-	const auto& image      = state.program.info.images.at(resource);
-	const bool  uint_image = image.kind == IR::ResourceKind::StorageImageUint;
-	EXIT_IF(image.kind != IR::ResourceKind::StorageImage && !uint_image);
-	EXIT_IF(image.atomic && !uint_image);
-	return StorageImageClassFor(uint_image, image.atomic);
-}
-
-uint32_t StorageImageVariable(const EmitterState& state, StorageImageClass image_class,
-                              ImageViewKind view) {
-	return state.storage_image_variables[StorageImageIndex(image_class, view)];
-}
-
-uint32_t LoadSampledImageDescriptor(EmitterState& state, const IR::MemoryInfo& mem, uint32_t use_pc,
-                                    ImageViewKind view) {
-	(void)use_pc;
-	const bool integer      = mem.kind == IR::ResourceKind::ImageUint;
-	const auto kind         = SampledBindingKind(integer, view);
-	const auto binding      = ResourceForDescriptor(state, kind, mem.resource);
-	const auto variable     = state.sampled_image_variables[SampledImageIndex(integer, view)];
+uint32_t LoadSampledImageDescriptor(EmitterState& state, uint32_t resource) {
+	const auto& image_resource = state.program.info.images.at(resource);
+	EXIT_IF(image_resource.resource_class != IR::ImageResourceClass::Sampled);
+	const auto kind = IR::DescriptorBindingForImage(image_resource);
+	EXIT_IF(!kind.has_value());
+	const auto array_index  = ResourceForDescriptor(state, *kind, resource);
+	const auto variable     = state.image_variables[IR::ImageBindingIndex(*kind)];
 	const auto pointer_type = state.builder.Type(
-	    OpTypePointer, {StorageClassUniformConstant, ImageViewImageType(state, view, integer)});
+	    OpTypePointer, {StorageClassUniformConstant, ImageType(state, image_resource)});
 	const auto pointer =
-	    DescriptorElementPointer(state, pointer_type, variable, binding.array_index, kind,
-	                             mem.resource, "sampled image descriptor array was not emitted");
+	    DescriptorElementPointer(state, pointer_type, variable, array_index, *kind, resource,
+	                             "sampled image descriptor array was not emitted");
 	const auto image = state.builder.AllocateId();
-	state.builder.AddFunction({OpLoad, ImageViewImageType(state, view, integer), image, pointer});
+	state.builder.AddFunction({OpLoad, ImageType(state, image_resource), image, pointer});
 	return image;
 }
 
-uint32_t LoadSamplerDescriptor(EmitterState& state, uint32_t sampler, uint32_t use_pc) {
-	(void)use_pc;
-	const auto binding = ResourceForDescriptor(state, IR::DescriptorBindingKind::Samplers, sampler);
+uint32_t LoadSamplerDescriptor(EmitterState& state, uint32_t sampler) {
+	const auto array_index =
+	    ResourceForDescriptor(state, IR::DescriptorBindingKind::Samplers, sampler);
 	const auto sampler_type = state.builder.Type(OpTypeSampler);
 	const auto pointer_type =
 	    state.builder.Type(OpTypePointer, {StorageClassUniformConstant, sampler_type});
 	const auto pointer = DescriptorElementPointer(
-	    state, pointer_type, state.sampler_variable, binding.array_index,
+	    state, pointer_type, state.sampler_variable, array_index,
 	    IR::DescriptorBindingKind::Samplers, sampler, "sampler descriptor array was not emitted");
 	const auto sampler_id = state.builder.AllocateId();
 	state.builder.AddFunction({OpLoad, sampler_type, sampler_id, pointer});
 	return sampler_id;
 }
 
-uint32_t MakeSampledImage(EmitterState& state, const IR::MemoryInfo& mem, uint32_t use_pc,
-                          ImageViewKind view) {
-	const auto image   = LoadSampledImageDescriptor(state, mem, use_pc, view);
-	const auto sampler = LoadSamplerDescriptor(state, mem.sampler, use_pc);
-	if (image == 0 || sampler == 0) {
-		ExitDescriptorBindingFailure(
-		    state, SampledBindingKind(mem.kind == IR::ResourceKind::ImageUint, view), mem.resource,
-		    "sampled image or sampler descriptor load failed");
-	}
-	const auto sampled_image = state.builder.AllocateId();
-	state.builder.AddFunction(
-	    {OpSampledImage,
-	     ImageViewSampledImageType(state, view, mem.kind == IR::ResourceKind::ImageUint),
-	     sampled_image, image, sampler});
+uint32_t MakeSampledImage(EmitterState& state, uint32_t resource, uint32_t sampler) {
+	const auto& image_resource = state.program.info.images.at(resource);
+	const auto  image          = LoadSampledImageDescriptor(state, resource);
+	const auto  sampler_id     = LoadSamplerDescriptor(state, sampler);
+	const auto  sampled_image = state.builder.AllocateId();
+	const auto  sampled_type  =
+	    state.builder.Type(OpTypeSampledImage, {ImageType(state, image_resource)});
+	state.builder.AddFunction({OpSampledImage, sampled_type, sampled_image, image, sampler_id});
 	return sampled_image;
 }
 
-uint32_t MakeSampledImage(EmitterState& state, const IR::MemoryInfo& mem, uint32_t use_pc,
-                          ImageViewKind view, uint32_t image_resource) {
-	auto selected     = mem;
-	selected.resource = image_resource;
-	return MakeSampledImage(state, selected, use_pc, view);
-}
-
-uint32_t StorageImageDescriptorPointer(EmitterState& state, uint32_t resource, ImageViewKind view) {
-	const auto image_class = StorageImageClassForResource(state, resource);
-	const auto kind        = StorageBindingKind(image_class, view);
-	const auto binding  = ResourceForDescriptor(state, kind, resource);
-	const auto ptr_type    = StorageImagePointerType(state, image_class, view);
-	const auto variable    = StorageImageVariable(state, image_class, view);
-	return DescriptorElementPointer(state, ptr_type, variable, binding.array_index, kind, resource,
+uint32_t StorageImageDescriptorPointer(EmitterState& state, uint32_t resource) {
+	const auto& image = state.program.info.images.at(resource);
+	EXIT_IF(image.resource_class != IR::ImageResourceClass::Storage);
+	const auto kind = IR::DescriptorBindingForImage(image);
+	EXIT_IF(!kind.has_value());
+	const auto array_index = ResourceForDescriptor(state, *kind, resource);
+	const auto pointer_type =
+	    state.builder.Type(OpTypePointer, {StorageClassUniformConstant, ImageType(state, image)});
+	const auto variable = state.image_variables[IR::ImageBindingIndex(*kind)];
+	return DescriptorElementPointer(state, pointer_type, variable, array_index, *kind, resource,
 	                                "storage image descriptor array was not emitted");
 }
 
-void EmitStorageImageWrite(EmitterState& state, uint32_t resource, ImageViewKind view,
-                           uint32_t mip_lod, uint32_t coord, uint32_t texel) {
-	const auto image_class = StorageImageClassForResource(state, resource);
-	if (image_class != StorageImageClass::AtomicUint) {
+void EmitStorageImageWrite(EmitterState& state, uint32_t resource, uint32_t mip_lod, uint32_t coord,
+                           uint32_t texel) {
+	const auto& image = state.program.info.images.at(resource);
+	EXIT_IF(image.resource_class != IR::ImageResourceClass::Storage);
+	if (!image.atomic) {
 		state.builder.RequireCapability(CapabilityStorageImageWriteWithoutFormat);
 	}
-	const auto  kind    = StorageBindingKind(image_class, view);
-	const auto  binding = ResourceForDescriptor(state, kind, resource);
-	const auto& image   = state.program.info.images.at(resource);
-	const auto  LoadAt  = [&](uint32_t array_index) {
-		const auto pointer = DescriptorElementPointer(
-		    state, StorageImagePointerType(state, image_class, view),
-		    StorageImageVariable(state, image_class, view), array_index, kind, resource,
-		    "storage image descriptor array was not emitted");
+	const auto kind = IR::DescriptorBindingForImage(image);
+	EXIT_IF(!kind.has_value());
+	const auto array_index = ResourceForDescriptor(state, *kind, resource);
+	const auto image_type  = ImageType(state, image);
+	const auto pointer_type =
+	    state.builder.Type(OpTypePointer, {StorageClassUniformConstant, image_type});
+	const auto variable = state.image_variables[IR::ImageBindingIndex(*kind)];
+	const auto LoadAt   = [&](uint32_t array_index) {
+		const auto pointer =
+		    DescriptorElementPointer(state, pointer_type, variable, array_index, *kind, resource,
+		                             "storage image descriptor array was not emitted");
 		const auto descriptor = state.builder.AllocateId();
-		state.builder.AddFunction(
-		    {OpLoad, StorageImageType(state, image_class, view), descriptor, pointer});
+		state.builder.AddFunction({OpLoad, image_type, descriptor, pointer});
 		return descriptor;
 	};
 	if (image.mip_mode != IR::ImageMipMode::DynamicStorage) {
-		state.builder.AddFunction({OpImageWrite, LoadAt(binding.array_index), coord, texel});
+		state.builder.AddFunction({OpImageWrite, LoadAt(array_index), coord, texel});
 		return;
 	}
 	if (image.mip_count == 0u) {
-		ExitDescriptorBindingFailure(state, kind, resource,
+		ExitDescriptorBindingFailure(state, *kind, resource,
 		                             "dynamic storage image has no mip descriptors");
 	}
 
@@ -326,7 +269,7 @@ void EmitStorageImageWrite(EmitterState& state, uint32_t resource, ImageViewKind
 	state.builder.AddFunction(words);
 	for (uint32_t mip = 0; mip < image.mip_count; mip++) {
 		EmitLabel(state, labels[mip]);
-		state.builder.AddFunction({OpImageWrite, LoadAt(binding.array_index + mip), coord, texel});
+		state.builder.AddFunction({OpImageWrite, LoadAt(array_index + mip), coord, texel});
 		state.builder.AddFunction({OpBranch, merge_label});
 	}
 	EmitLabel(state, merge_label);

src/graphics/shader/recompiler/backend/spirv/spirvEmitterFlow.cpp

@@ -408,6 +408,41 @@ void EmitAuxPositionExport(ValueEmitContext& ctx, uint32_t data, const IR::Expor
 	}
 }
 
+uint32_t ConvertClipCoordinate(EmitterState& state, uint32_t coordinate, float scale,
+                               float offset, float half_extent) {
+	const auto window  = state.builder.AllocateId();
+	const auto biased  = state.builder.AllocateId();
+	const auto divided = state.builder.AllocateId();
+	const auto ndc     = state.builder.AllocateId();
+	state.builder.AddFunction(
+	    {OpFMul, TypeF32(state), window, coordinate, ConstantF32Value(state, scale)});
+	state.builder.AddFunction(
+	    {OpFAdd, TypeF32(state), biased, window, ConstantF32Value(state, offset)});
+	state.builder.AddFunction(
+	    {OpFDiv, TypeF32(state), divided, biased, ConstantF32Value(state, half_extent)});
+	state.builder.AddFunction(
+	    {OpFSub, TypeF32(state), ndc, divided, ConstantF32Value(state, 1.0f)});
+	return ndc;
+}
+
+uint32_t ConvertPositionToClipSpace(EmitterState& state, uint32_t position) {
+	const auto& transform = state.input_info.vertex->clip_space;
+	uint32_t    components[4] {};
+	for (uint32_t i = 0; i < 4; i++) {
+		components[i] = state.builder.AllocateId();
+		state.builder.AddFunction(
+		    {OpCompositeExtract, TypeF32(state), components[i], position, i});
+	}
+	components[0] = ConvertClipCoordinate(state, components[0], transform.scale[0],
+	                                      transform.offset[0], transform.half_extent[0]);
+	components[1] = ConvertClipCoordinate(state, components[1], transform.scale[1],
+	                                      transform.offset[1], transform.half_extent[1]);
+	const auto converted = state.builder.AllocateId();
+	state.builder.AddFunction({OpCompositeConstruct, TypeF32Vector(state, 4), converted,
+	                           components[0], components[1], components[2], components[3]});
+	return converted;
+}
+
 void EmitExport(ValueEmitContext& ctx, const IR::Inst& inst) {
 	auto&       state = ctx.state;
 	const auto& exp   = ctx.Export(inst);
@@ -467,6 +502,9 @@ void EmitExport(ValueEmitContext& ctx, const IR::Inst& inst) {
 			}
 		}
 		if (exp.kind == IR::ExportTargetKind::Position) {
+			if (state.stage == ShaderType::Vertex && state.input_info.vertex->clip_space.enabled) {
+				value = ConvertPositionToClipSpace(state, value);
+			}
 			const auto pointer = state.builder.AllocateId();
 			state.builder.AddFunction(
 			    {OpAccessChain, TypePointer(state, StorageClassOutput, TypeF32Vector(state, 4)),

src/graphics/shader/recompiler/backend/spirv/spirvEmitterHelpers.cpp

@@ -110,6 +110,12 @@ DppTargetLane EmitDppTargetLane(EmitterState& state, uint32_t control) {
 	if (control == 0x141u) {
 		return EmitDppMirrorTargetLane(state, subid, true);
 	}
+	if (control >= 0x160u && control <= 0x16fu) {
+		const auto target = state.builder.AllocateId();
+		state.builder.AddFunction(
+		    {OpBitwiseXor, TypeU32(state), target, subid, ConstantU32(state, control & 0xfu)});
+		return {target, EmitTrueBool(state)};
+	}
 	return {subid, EmitTrueBool(state)};
 }
 

src/graphics/shader/recompiler/backend/spirv/spirvEmitterImage.cpp

@@ -103,20 +103,21 @@ uint32_t AddressF32(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::
 	           : Unary(ctx.state, OpBitcast, TypeF32(ctx.state), value);
 }
 
-ImageSampleLayout Layout(const IR::MemoryInfo& mem, ImageViewKind view) {
+ImageSampleLayout Layout(const IR::MemoryInfo& mem, ImageDimension dimension) {
 	ImageSampleLayout layout;
 	uint32_t          cursor = 0;
+	const auto&       info   = ImageDimensionInfoFor(dimension);
 	if (HasFlag(mem, Decoder::ImageSampleFlagOffset)) layout.offset = cursor++;
 	if (HasFlag(mem, Decoder::ImageSampleFlagBias)) layout.bias = cursor++;
 	if (HasFlag(mem, Decoder::ImageSampleFlagCompare)) layout.dref = cursor++;
 	if (HasFlag(mem, Decoder::ImageSampleFlagDerivative)) {
 		layout.grad_x = cursor;
-		cursor += ImageViewSpatialComponents(view);
+		cursor += info.spatial_components;
 		layout.grad_y = cursor;
-		cursor += ImageViewSpatialComponents(view);
+		cursor += info.spatial_components;
 	}
 	layout.coord = cursor;
-	cursor += ImageViewCoordinateComponents(view);
+	cursor += info.coordinate_components;
 	if (HasFlag(mem, Decoder::ImageSampleFlagLod)) layout.lod = cursor++;
 	return layout;
 }
@@ -144,11 +145,12 @@ uint32_t CubeLayer(EmitterState& state, uint32_t value) {
 
 uint32_t CoordF32(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::Inst& address,
                   uint32_t first, uint32_t components) {
+	const bool cube = ctx.state.program.info.images.at(mem.resource).cube;
 	auto x = AddressF32(ctx, mem, address, first);
 	if (components == 1u) return x;
 	auto y = mem.image_address_components > first + 1u ? AddressF32(ctx, mem, address, first + 1u)
 	                                                   : ZeroF32(ctx.state);
-	if (mem.image_cube) {
+	if (cube) {
 		x = CubeAxis(ctx.state, x);
 		y = CubeAxis(ctx.state, y);
 	}
@@ -157,7 +159,7 @@ uint32_t CoordF32(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::In
 		auto z = mem.image_address_components > first + 2u
 		             ? AddressF32(ctx, mem, address, first + 2u)
 		             : ZeroF32(ctx.state);
-		if (mem.image_cube) z = CubeLayer(ctx.state, z);
+		if (cube) z = CubeLayer(ctx.state, z);
 		ctx.state.builder.AddFunction(
 		    {OpCompositeConstruct, TypeF32Vector(ctx.state, 3), result, x, y, z});
 	} else {
@@ -168,8 +170,8 @@ uint32_t CoordF32(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::In
 }
 
 uint32_t CoordU32(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::Inst& address,
-                  ImageViewKind view) {
-	const auto components = ImageViewCoordinateComponents(view);
+                  ImageDimension dimension) {
+	const auto components = ImageDimensionInfoFor(dimension).coordinate_components;
 	const auto x          = AddressU32(ctx, mem, address, 0);
 	if (components == 1u) return x;
 	const auto y      = mem.image_address_components > 1u ? AddressU32(ctx, mem, address, 1)
@@ -188,8 +190,8 @@ uint32_t CoordU32(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::In
 }
 
 uint32_t LodU32(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::Inst& address,
-                ImageViewKind view) {
-	const auto component = ImageViewCoordinateComponents(view);
+                ImageDimension dimension) {
+	const auto component = ImageDimensionInfoFor(dimension).coordinate_components;
 	return mem.image_has_mip && mem.image_address_components > component
 	           ? AddressU32(ctx, mem, address, component)
 	           : ConstantU32(ctx.state, 0);
@@ -199,8 +201,33 @@ uint32_t FloatBits(ValueEmitContext& ctx, uint32_t value) {
 	return Unary(ctx.state, OpBitcast, TypeU32(ctx.state), value);
 }
 
-uint32_t ResultVector(ValueEmitContext& ctx, uint32_t value, bool integer, bool dref,
+uint32_t SampledComponentBits(ValueEmitContext& ctx, uint32_t value,
+                              Prospero::TextureNumericClass numeric_class) {
+	if (numeric_class == Prospero::TextureNumericClass::Uint) {
+		return value;
+	}
+	return Unary(ctx.state, OpBitcast, TypeU32(ctx.state), value);
+}
+
+uint32_t SampledComponentZero(EmitterState& state, Prospero::TextureNumericClass numeric_class) {
+	switch (numeric_class) {
+		case Prospero::TextureNumericClass::Float: return ZeroF32(state);
+		case Prospero::TextureNumericClass::Uint: return ConstantU32(state, 0);
+		case Prospero::TextureNumericClass::Sint: return ConstantI32(state, 0);
+		case Prospero::TextureNumericClass::Unsupported: break;
+	}
+	EXIT("invalid sampled image numeric class");
+}
+
+uint32_t ResultVector(ValueEmitContext& ctx, uint32_t value,
+                      Prospero::TextureNumericClass numeric_class, bool dref,
                       const IR::MemoryInfo& mem, bool gather = false) {
+	auto value_class = numeric_class;
+	if (dref) {
+		value_class = Prospero::TextureNumericClass::Float;
+	}
+	const bool integer = value_class == Prospero::TextureNumericClass::Uint ||
+	                     value_class == Prospero::TextureNumericClass::Sint;
 	if (mem.data_bits == 16u) {
 		uint32_t   packed[4] = {ConstantU32(ctx.state, 0), ConstantU32(ctx.state, 0),
 		                        ConstantU32(ctx.state, 0), ConstantU32(ctx.state, 0)};
@@ -209,25 +236,27 @@ uint32_t ResultVector(ValueEmitContext& ctx, uint32_t value, bool integer, bool
 			const auto component = gather ? index : DmaskComponent(mem.dmask, index);
 			const auto result    = ctx.state.builder.AllocateId();
 			ctx.state.builder.AddFunction({OpCompositeExtract,
-			                               integer ? TypeU32(ctx.state) : TypeF32(ctx.state),
-			                               result, value, component});
+			                               ImageScalarType(ctx.state, value_class), result, value,
+			                               component});
 			return result;
 		};
 		for (uint32_t word = 0; word < mem.data_dwords; word++) {
 			const auto low_index  = word * 2u;
 			const auto high_index = low_index + 1u;
 			const auto low        = scalar(low_index);
-			const auto high = high_index < mem.component_count
-			                      ? scalar(high_index)
-			                      : (integer ? ConstantU32(ctx.state, 0) : ZeroF32(ctx.state));
+			const auto high       = high_index < mem.component_count
+			                            ? scalar(high_index)
+			                            : SampledComponentZero(ctx.state, value_class);
 			if (integer) {
-				const auto mask = ConstantU32(ctx.state, 0xffffu);
-				packed[word] =
-				    Binary(ctx.state, OpBitwiseOr, TypeU32(ctx.state),
-				           Binary(ctx.state, OpBitwiseAnd, TypeU32(ctx.state), low, mask),
-				           Binary(ctx.state, OpShiftLeftLogical, TypeU32(ctx.state),
-				                  Binary(ctx.state, OpBitwiseAnd, TypeU32(ctx.state), high, mask),
-				                  ConstantU32(ctx.state, 16u)));
+				const auto low_bits  = SampledComponentBits(ctx, low, value_class);
+				const auto high_bits = SampledComponentBits(ctx, high, value_class);
+				const auto mask      = ConstantU32(ctx.state, 0xffffu);
+				packed[word]         = Binary(
+				    ctx.state, OpBitwiseOr, TypeU32(ctx.state),
+				    Binary(ctx.state, OpBitwiseAnd, TypeU32(ctx.state), low_bits, mask),
+				    Binary(ctx.state, OpShiftLeftLogical, TypeU32(ctx.state),
+				           Binary(ctx.state, OpBitwiseAnd, TypeU32(ctx.state), high_bits, mask),
+				           ConstantU32(ctx.state, 16u)));
 			} else {
 				const auto pair = ctx.state.builder.AllocateId();
 				ctx.state.builder.AddFunction(
@@ -249,10 +278,9 @@ uint32_t ResultVector(ValueEmitContext& ctx, uint32_t value, bool integer, bool
 			continue;
 		}
 		const auto scalar = ctx.state.builder.AllocateId();
-		ctx.state.builder.AddFunction({OpCompositeExtract,
-		                               integer ? TypeU32(ctx.state) : TypeF32(ctx.state), scalar,
-		                               value, index});
-		component[index] = integer ? scalar : FloatBits(ctx, scalar);
+		ctx.state.builder.AddFunction(
+		    {OpCompositeExtract, ImageScalarType(ctx.state, value_class), scalar, value, index});
+		component[index] = SampledComponentBits(ctx, scalar, value_class);
 	}
 	const auto result = ctx.state.builder.AllocateId();
 	ctx.state.builder.AddFunction({OpCompositeConstruct, TypeU32Vector(ctx.state, 4), result,
@@ -260,20 +288,20 @@ uint32_t ResultVector(ValueEmitContext& ctx, uint32_t value, bool integer, bool
 	return result;
 }
 
-uint32_t QueryDimensions(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem,
+uint32_t QueryDimensions(ValueEmitContext& ctx, const IR::MemoryInfo& mem,
                          const IR::Inst& address) {
-	const auto pc    = inst.Flags<IR::MemoryFlags>().pc;
-	const auto view  = SampledImageViewKind(ctx.state, mem, pc);
-	const auto image = LoadSampledImageDescriptor(ctx.state, mem, pc, view);
-	const auto size  = ctx.state.builder.AllocateId();
-	if (ImageSpirvMultisampled(view) != 0u) {
+	const auto  dimension = ctx.state.program.info.images.at(mem.resource).dimension;
+	const auto& info      = ImageDimensionInfoFor(dimension);
+	const auto  image     = LoadSampledImageDescriptor(ctx.state, mem.resource);
+	const auto  size      = ctx.state.builder.AllocateId();
+	if (info.multisampled != 0u) {
 		ctx.state.builder.AddFunction(
-		    {OpImageQuerySize, ImageViewSizeType(ctx.state, view), size, image});
+		    {OpImageQuerySize, ImageViewSizeType(ctx.state, dimension), size, image});
 	} else {
-		ctx.state.builder.AddFunction({OpImageQuerySizeLod, ImageViewSizeType(ctx.state, view),
+		ctx.state.builder.AddFunction({OpImageQuerySizeLod, ImageViewSizeType(ctx.state, dimension),
 		                               size, image, AddressU32(ctx, mem, address, 0)});
 	}
-	const auto components = ImageViewCoordinateComponents(view);
+	const auto components = info.coordinate_components;
 	uint32_t   result[4]  = {ConstantU32(ctx.state, 0), ConstantU32(ctx.state, 0),
 	                         ConstantU32(ctx.state, 0), ConstantU32(ctx.state, 0)};
 	if (components == 1u) {
@@ -285,7 +313,7 @@ uint32_t QueryDimensions(ValueEmitContext& ctx, const IR::Inst& inst, const IR::
 			    {OpCompositeExtract, TypeU32(ctx.state), result[index], size, index});
 		}
 	}
-	if (ImageSpirvMultisampled(view) == 0u) {
+	if (info.multisampled == 0u) {
 		result[3] = ctx.state.builder.AllocateId();
 		ctx.state.builder.AddFunction({OpImageQueryLevels, TypeU32(ctx.state), result[3], image});
 	}
@@ -296,8 +324,8 @@ uint32_t QueryDimensions(ValueEmitContext& ctx, const IR::Inst& inst, const IR::
 }
 
 uint32_t PackedOffset(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::Inst& address,
-                      const ImageSampleLayout& layout, ImageViewKind view) {
-	const auto components = ImageViewSpatialComponents(view);
+                      const ImageSampleLayout& layout, ImageDimension dimension) {
+	const auto components = ImageDimensionInfoFor(dimension).spatial_components;
 	const auto zero       = ConstantI32(ctx.state, 0);
 	if (layout.offset == NoImageComponent || mem.image_address_components <= layout.offset) {
 		if (components == 1u) return zero;
@@ -332,8 +360,8 @@ uint32_t PackedOffset(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR
 	return result;
 }
 
-uint32_t HorizontalOffsets(EmitterState& state, ImageViewKind view) {
-	const auto components   = ImageViewSpatialComponents(view);
+uint32_t HorizontalOffsets(EmitterState& state, ImageDimension dimension) {
+	const auto components   = ImageDimensionInfoFor(dimension).spatial_components;
 	const auto count        = ConstantU32(state, 4);
 	const auto element_type = components == 1u ? TypeI32(state) : TypeI32Vector(state, 2);
 	const auto array_type   = state.builder.Type(OpTypeArray, {element_type, count});
@@ -363,8 +391,7 @@ Format::BufferFormatInfo ImageConversionFormat(const EmitterState&   state,
 	const auto format = state.program.info.images[mem.resource].conversion_format;
 	if (format == Prospero::BufferFormat::kInvalid) return {};
 	const auto info = Format::GetFormatInfo(format);
-	EXIT_IF(!Prospero::IsSampledTextureFormat(format) ||
-	        !Prospero::IsUintTextureFormat(format) ||
+	EXIT_IF(Prospero::SampledTextureNumericClass(format) != Prospero::TextureNumericClass::Uint ||
 	        Prospero::RemapTextureFormat(format) == format ||
 	        info.type != Format::ComponentType::Uint || !info.packed_bitfield ||
 	        info.byte_size != sizeof(uint32_t) || info.component_count == 0u ||
@@ -439,6 +466,49 @@ uint32_t UnpackImageGather(ValueEmitContext& ctx, const IR::MemoryInfo& mem, uin
 	return result;
 }
 
+uint32_t EmitOneDimensionalGatherLz(ValueEmitContext& ctx, const IR::MemoryInfo& mem,
+                                    uint32_t coord, Prospero::TextureNumericClass numeric_class) {
+	auto& state = ctx.state;
+	state.builder.RequireCapability(CapabilityImageQuery);
+	const auto image = LoadSampledImageDescriptor(state, mem.resource);
+	const auto width = state.builder.AllocateId();
+	state.builder.AddFunction(
+	    {OpImageQuerySizeLod, TypeU32(state), width, image, ConstantU32(state, 0)});
+	const auto width_f32 = state.builder.AllocateId();
+	state.builder.AddFunction({OpConvertUToF, TypeF32(state), width_f32, width});
+	const auto left = state.builder.AllocateId();
+	state.builder.AddFunction({OpExtInst, TypeF32(state), left, GlslStd450(state), GlslFloor,
+	                           Binary(state, OpFSub, TypeF32(state),
+	                                  Binary(state, OpFMul, TypeF32(state), coord, width_f32),
+	                                  ConstantF32(state, 0x3f000000u))});
+
+	const auto sampled     = MakeSampledImage(state, mem.resource, mem.sampler);
+	const auto vector_type = ImageVectorType(state, numeric_class, 4);
+	const auto scalar_type = ImageScalarType(state, numeric_class);
+	const auto component =
+	    ImageConversionFormat(state, mem).format == Prospero::BufferFormat::kInvalid
+	        ? ImageGatherComponent(mem.dmask)
+	        : 0u;
+	uint32_t values[2] {};
+	for (uint32_t index = 0; index < 2u; index++) {
+		const auto sample_coord =
+		    Binary(state, OpFDiv, TypeF32(state),
+		           Binary(state, OpFAdd, TypeF32(state), left,
+		                  ConstantF32(state, index == 0u ? 0x3f000000u : 0x3fc00000u)),
+		           width_f32);
+		const auto texel = state.builder.AllocateId();
+		state.builder.AddFunction({OpImageSampleExplicitLod, vector_type, texel, sampled,
+		                           sample_coord, ImageOperandsLodMask, ZeroF32(state)});
+		values[index] = state.builder.AllocateId();
+		state.builder.AddFunction(
+		    {OpCompositeExtract, scalar_type, values[index], texel, component});
+	}
+	const auto result = state.builder.AllocateId();
+	state.builder.AddFunction(
+	    {OpCompositeConstruct, vector_type, result, values[0], values[1], values[1], values[0]});
+	return result;
+}
+
 uint32_t PackImageTexel(ValueEmitContext& ctx, const IR::MemoryInfo& mem, uint32_t texel) {
 	const auto info = ImageConversionFormat(ctx.state, mem);
 	if (info.format == Prospero::BufferFormat::kInvalid) return texel;
@@ -506,6 +576,7 @@ uint32_t StoreTexel(ValueEmitContext& ctx, const IR::MemoryInfo& mem, uint32_t d
 
 uint32_t ImageAtomicOpcode(IR::ValueOpcode opcode) {
 	switch (opcode) {
+		case IR::ValueOpcode::ImageAtomicSwap32: return OpAtomicExchange;
 		case IR::ValueOpcode::ImageAtomicIAdd32: return OpAtomicIAdd;
 		case IR::ValueOpcode::ImageAtomicUMin32: return OpAtomicUMin;
 		case IR::ValueOpcode::ImageAtomicUMax32: return OpAtomicUMax;
@@ -526,24 +597,24 @@ bool EmitValueImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 	}
 	auto&       state     = ctx.state;
 	const auto& mem       = ctx.Memory(inst);
-	const auto  pc        = inst.Flags<IR::MemoryFlags>().pc;
 	const auto  image_arg = inst.Arg(0);
 	ctx.ResourceIndex(image_arg, IR::ValueOpcode::GetImageResource);
+	const auto& image   = state.program.info.images.at(mem.resource);
 	const auto* address = ctx.ImageAddress(inst.Arg(image_info.needs_sampler ? 2 : 1));
 	if (address == nullptr) return true;
 	if (op == IR::ValueOpcode::ImageQueryDimensions) {
 		state.builder.RequireCapability(CapabilityImageQuery);
-		ctx.Define(inst, QueryDimensions(ctx, inst, mem, *address));
+		ctx.Define(inst, QueryDimensions(ctx, mem, *address));
 		return true;
 	}
 	if (op == IR::ValueOpcode::ImageQueryLod) {
 		state.builder.RequireCapability(CapabilityImageQuery);
-		const auto view    = SampledImageViewKind(state, mem, pc);
-		const auto sampled = MakeSampledImage(state, mem, pc, view);
-		const auto lod     = state.builder.AllocateId();
+		const auto dimension = image.dimension;
+		const auto sampled   = MakeSampledImage(state, mem.resource, mem.sampler);
+		const auto lod       = state.builder.AllocateId();
 		state.builder.AddFunction(
 		    {OpImageQueryLod, TypeF32Vector(state, 2), lod, sampled,
-		     CoordF32(ctx, mem, *address, 0, ImageViewSpatialComponents(view))});
+		     CoordF32(ctx, mem, *address, 0, ImageDimensionInfoFor(dimension).spatial_components)});
 		uint32_t values[4] = {ConstantU32(state, 0), ConstantU32(state, 0), ConstantU32(state, 0),
 		                      ConstantU32(state, 0)};
 		for (uint32_t index = 0; index < 2u; index++) {
@@ -558,127 +629,153 @@ bool EmitValueImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 		return true;
 	}
 	if (op == IR::ValueOpcode::ImageRead) {
-		const auto view      = SampledImageViewKind(state, mem, pc);
-		const bool integer   = mem.kind == IR::ResourceKind::ImageUint;
-		const auto condition = ctx.Arg(inst, 2);
+		const auto  dimension      = image.dimension;
+		const auto& dimension_info = ImageDimensionInfoFor(dimension);
+		const auto  numeric_class  = image.numeric_class;
+		const auto  condition      = ctx.Arg(inst, 2);
 		ctx.Define(
 		    inst,
 		    EmitValueOrDefaultIfCondition(
 		        state, condition, TypeU32Vector(state, 4), ConstantU32CompositeZero(state, 4),
 		        [&]() {
-			        const auto image = LoadSampledImageDescriptor(state, mem, pc, view);
-			        const auto color = state.builder.AllocateId();
-			        const auto coord = CoordU32(ctx, mem, *address, view);
-			        if (ImageSpirvMultisampled(view) != 0u) {
+			        const auto descriptor = LoadSampledImageDescriptor(state, mem.resource);
+			        const auto color      = state.builder.AllocateId();
+			        const auto coord      = CoordU32(ctx, mem, *address, dimension);
+			        if (dimension_info.multisampled != 0u) {
 				        state.builder.AddFunction(
-				            {OpImageFetch,
-				             integer ? TypeU32Vector(state, 4) : TypeF32Vector(state, 4), color,
-				             image, coord, ImageOperandsSampleMask,
-				             AddressU32(ctx, mem, *address, ImageViewCoordinateComponents(view))});
+				            {OpImageFetch, ImageVectorType(state, numeric_class, 4), color,
+				             descriptor, coord, ImageOperandsSampleMask,
+				             AddressU32(ctx, mem, *address, dimension_info.coordinate_components)});
 			        } else {
-				        state.builder.AddFunction(
-				            {OpImageFetch,
-				             integer ? TypeU32Vector(state, 4) : TypeF32Vector(state, 4), color,
-				             image, coord, ImageOperandsLodMask, LodU32(ctx, mem, *address, view)});
+				        state.builder.AddFunction({OpImageFetch,
+				                                   ImageVectorType(state, numeric_class, 4), color,
+				                                   descriptor, coord, ImageOperandsLodMask,
+				                                   LodU32(ctx, mem, *address, dimension)});
 			        }
-			        return ResultVector(ctx, UnpackImageTexel(ctx, mem, color), integer, false,
-			                            mem);
+			        return ResultVector(ctx, UnpackImageTexel(ctx, mem, color), numeric_class,
+			                            false, mem);
 		        }));
 		return true;
 	}
 	if (op == IR::ValueOpcode::ImageWrite) {
-		const auto uint_image = mem.kind == IR::ResourceKind::StorageImageUint;
-		const auto view       = StorageImageViewKind(mem);
+		const bool uint_image = image.numeric_class == Prospero::TextureNumericClass::Uint;
+		const auto dimension  = image.dimension;
 		EmitIfCondition(state, ctx.Arg(inst, 3), [&]() {
 			const auto mip_lod =
 			    state.program.info.images[mem.resource].mip_mode == IR::ImageMipMode::DynamicStorage
-			        ? LodU32(ctx, mem, *address, view)
+			        ? LodU32(ctx, mem, *address, dimension)
 			        : 0u;
-			const auto coord = CoordU32(ctx, mem, *address, view);
+			const auto coord = CoordU32(ctx, mem, *address, dimension);
 			const auto texel = StoreTexel(ctx, mem, ctx.Arg(inst, 2), uint_image);
-			EmitStorageImageWrite(state, mem.resource, view, mip_lod, coord, texel);
+			EmitStorageImageWrite(state, mem.resource, mip_lod, coord, texel);
 		});
 		return true;
 	}
 	if (op == IR::ValueOpcode::ImageSampleRaw || op == IR::ValueOpcode::ImageGatherRaw) {
-		const auto view    = SampledImageViewKind(state, mem, pc);
-		const auto layout  = Layout(mem, view);
-		const bool integer = mem.kind == IR::ResourceKind::ImageUint;
-		const bool dref    = HasFlag(mem, Decoder::ImageSampleFlagCompare);
+		const auto  dimension      = image.dimension;
+		const auto& dimension_info = ImageDimensionInfoFor(dimension);
+		const auto  layout         = Layout(mem, dimension);
+		const auto  numeric_class  = image.numeric_class;
+		const bool  dref           = HasFlag(mem, Decoder::ImageSampleFlagCompare);
 		if (dref && state.program.info.images[mem.resource].conversion_format !=
 		                Prospero::BufferFormat::kInvalid) {
 			ctx.Fail(inst, "uses depth comparison with a packed integer image");
 			return true;
 		}
 		const auto coord =
-		    CoordF32(ctx, mem, *address, layout.coord, ImageViewCoordinateComponents(view));
+		    CoordF32(ctx, mem, *address, layout.coord, dimension_info.coordinate_components);
 		if (op == IR::ValueOpcode::ImageGatherRaw) {
-			const auto            sampled = MakeSampledImage(state, mem, pc, view);
+			if (dimension == ImageDimension::Dim1D) {
+				if (dref || !HasFlag(mem, Decoder::ImageSampleFlagLevelZero) ||
+				    HasFlag(mem, Decoder::ImageSampleFlagOffset) ||
+				    HasFlag(mem, Decoder::ImageSampleFlagGatherHorizontal)) {
+					ctx.Fail(inst, "has an unsupported 1D gather variant");
+					return true;
+				}
+				const auto sample = EmitOneDimensionalGatherLz(ctx, mem, coord, numeric_class);
+				ctx.Define(inst, ResultVector(ctx, UnpackImageGather(ctx, mem, sample),
+				                              numeric_class, false, mem, true));
+				return true;
+			}
+			if (dimension == ImageDimension::Dim1DArray) {
+				ctx.Fail(inst, "has an unsupported 1D-array gather");
+				return true;
+			}
+			const auto            sampled = MakeSampledImage(state, mem.resource, mem.sampler);
 			const auto            sample  = state.builder.AllocateId();
-			std::vector<uint32_t> words =
-			    dref ? std::vector<uint32_t> {OpImageDrefGather,
-			                                  TypeF32Vector(state, 4),
-			                                  sample,
-			                                  sampled,
-			                                  coord,
-			                                  layout.dref != NoImageComponent
-			                                      ? AddressF32(ctx, mem, *address, layout.dref)
-			                                      : ZeroF32(state)}
-			         : std::vector<uint32_t> {
-			               OpImageGather,
-			               integer ? TypeU32Vector(state, 4) : TypeF32Vector(state, 4),
-			               sample,
-			               sampled,
-			               coord,
-			               ConstantU32(state, ImageConversionFormat(state, mem).format ==
-			                                          Prospero::BufferFormat::kInvalid
-			                                      ? ImageGatherComponent(mem.dmask)
-			                                      : 0u)};
+			std::vector<uint32_t> words;
+			if (dref) {
+				auto dref_value = ZeroF32(state);
+				if (layout.dref != NoImageComponent) {
+					dref_value = AddressF32(ctx, mem, *address, layout.dref);
+				}
+				words = {OpImageDrefGather, TypeF32Vector(state, 4), sample, sampled, coord,
+				         dref_value};
+			} else {
+				uint32_t component = 0;
+				if (ImageConversionFormat(state, mem).format == Prospero::BufferFormat::kInvalid) {
+					component = ImageGatherComponent(mem.dmask);
+				}
+				words = {OpImageGather, ImageVectorType(state, numeric_class, 4),
+				         sample,        sampled,
+				         coord,         ConstantU32(state, component)};
+			}
 			if (HasFlag(mem, Decoder::ImageSampleFlagGatherHorizontal)) {
 				words.push_back(ImageOperandsConstOffsetsMask);
-				words.push_back(HorizontalOffsets(state, view));
+				words.push_back(HorizontalOffsets(state, dimension));
 			} else if (layout.offset != NoImageComponent) {
 				words.push_back(ImageOperandsOffsetMask);
-				words.push_back(PackedOffset(ctx, mem, *address, layout, view));
+				words.push_back(PackedOffset(ctx, mem, *address, layout, dimension));
 			}
 			state.builder.AddFunction(words);
+			auto result_numeric_class = numeric_class;
+			if (dref) {
+				result_numeric_class = Prospero::TextureNumericClass::Float;
+			}
 			ctx.Define(inst, ResultVector(ctx, UnpackImageGather(ctx, mem, sample),
-			                              integer && !dref, false, mem, true));
+			                              result_numeric_class, false, mem, true));
 			return true;
 		}
 		const bool explicit_lod = HasFlag(mem, Decoder::ImageSampleFlagDerivative) ||
 		                          HasFlag(mem, Decoder::ImageSampleFlagLod) ||
 		                          HasFlag(mem, Decoder::ImageSampleFlagLevelZero) ||
 		                          state.stage != ShaderType::Pixel;
-		const auto opcode = explicit_lod
-		                        ? (dref ? OpImageSampleDrefExplicitLod : OpImageSampleExplicitLod)
-		                        : (dref ? OpImageSampleDrefImplicitLod : OpImageSampleImplicitLod);
-		const auto result_type =
-		    dref ? TypeF32(state) : (integer ? TypeU32Vector(state, 4) : TypeF32Vector(state, 4));
-		const auto dref_value =
-		    dref ? (layout.dref != NoImageComponent ? AddressF32(ctx, mem, *address, layout.dref)
-		                                            : ZeroF32(state))
-		         : 0u;
+		uint32_t opcode = OpImageSampleImplicitLod;
+		if (explicit_lod) {
+			opcode = dref ? OpImageSampleDrefExplicitLod : OpImageSampleExplicitLod;
+		} else if (dref) {
+			opcode = OpImageSampleDrefImplicitLod;
+		}
+		uint32_t result_type = ImageVectorType(state, numeric_class, 4);
+		uint32_t dref_value  = 0;
+		if (dref) {
+			result_type = TypeF32(state);
+			dref_value  = ZeroF32(state);
+			if (layout.dref != NoImageComponent) {
+				dref_value = AddressF32(ctx, mem, *address, layout.dref);
+			}
+		}
 		uint32_t              operand_mask = 0;
 		std::vector<uint32_t> operands;
 		if (HasFlag(mem, Decoder::ImageSampleFlagDerivative)) {
 			operand_mask |= ImageOperandsGradMask;
 			operands.push_back(
-			    CoordF32(ctx, mem, *address, layout.grad_x, ImageViewSpatialComponents(view)));
+			    CoordF32(ctx, mem, *address, layout.grad_x, dimension_info.spatial_components));
 			operands.push_back(
-			    CoordF32(ctx, mem, *address, layout.grad_y, ImageViewSpatialComponents(view)));
+			    CoordF32(ctx, mem, *address, layout.grad_y, dimension_info.spatial_components));
 		} else if (explicit_lod) {
 			operand_mask |= ImageOperandsLodMask;
-			operands.push_back(HasFlag(mem, Decoder::ImageSampleFlagLod) &&
-			                           layout.lod != NoImageComponent
-			                       ? AddressF32(ctx, mem, *address, layout.lod)
-			                       : ZeroF32(state));
+			auto lod = ZeroF32(state);
+			if (HasFlag(mem, Decoder::ImageSampleFlagLod) && layout.lod != NoImageComponent) {
+				lod = AddressF32(ctx, mem, *address, layout.lod);
+			}
+			operands.push_back(lod);
 		} else if (layout.bias != NoImageComponent) {
 			operand_mask |= ImageOperandsBiasMask;
 			operands.push_back(AddressF32(ctx, mem, *address, layout.bias));
 		}
 		const auto EmitSample = [&](uint32_t resource) {
-			const auto            sampled = MakeSampledImage(state, mem, pc, view, resource);
+			const auto            sampled = MakeSampledImage(state, resource, mem.sampler);
 			const auto            sample  = state.builder.AllocateId();
 			std::vector<uint32_t> words {opcode, result_type, sample, sampled, coord};
 			if (dref) {
@@ -691,11 +788,13 @@ bool EmitValueImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 			state.builder.AddFunction(words);
 			return sample;
 		};
-		const auto& image = state.program.info.images[mem.resource];
 		if (image.indirect_root != mem.resource) {
 			const auto sample = EmitSample(mem.resource);
-			ctx.Define(inst, ResultVector(ctx, dref ? sample : UnpackImageTexel(ctx, mem, sample),
-			                              integer, dref, mem));
+			auto       result = sample;
+			if (!dref) {
+				result = UnpackImageTexel(ctx, mem, sample);
+			}
+			ctx.Define(inst, ResultVector(ctx, result, numeric_class, dref, mem));
 			return true;
 		}
 		const auto* handle = image_arg.ResolveInstruction();
@@ -708,7 +807,7 @@ bool EmitValueImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 			return true;
 		}
 		const auto key = ctx.Def(handle->Arg(source->indirect_image->key_arg));
-		if (state.flattened_srt_variable == 0 || image.indirect_mapping_capacity == 0u ||
+		if (state.flattened_srt_variable == 0 || image.indirect_search_iterations == 0u ||
 		    image.indirect_resources.size() < 2u) {
 			ctx.Fail(inst, "has no indirect image runtime mapping");
 			return true;
@@ -726,7 +825,7 @@ bool EmitValueImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 		auto       low      = ConstantU32(state, 0u);
 		auto       high     = LoadMapping(mapping);
 		auto       selected = ConstantU32(state, 0u);
-		for (uint32_t iteration = 0; iteration < std::bit_width(image.indirect_mapping_capacity);
+		for (uint32_t iteration = 0; iteration < image.indirect_search_iterations;
 		     iteration++) {
 			const auto active = Binary(state, OpULessThan, TypeBool(state), low, high);
 			const auto mid =
@@ -786,22 +885,24 @@ bool EmitValueImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 		}
 		EmitLabel(state, merge_label);
 		state.builder.AddFunction(phi_words);
-		ctx.Define(inst,
-		           ResultVector(ctx, dref ? phi_words[2] : UnpackImageTexel(ctx, mem, phi_words[2]),
-		                        integer, dref, mem));
+		auto result = phi_words[2];
+		if (!dref) {
+			result = UnpackImageTexel(ctx, mem, result);
+		}
+		ctx.Define(inst, ResultVector(ctx, result, numeric_class, dref, mem));
 		return true;
 	}
 	const auto atomic_opcode = ImageAtomicOpcode(op);
 	if (atomic_opcode != 0u) {
-		const auto view = StorageImageViewKind(mem);
+		const auto dimension = image.dimension;
 		ctx.Define(inst, EmitValueOrZeroIfCondition(state, ctx.Arg(inst, 3), [&]() {
 			           const auto pointer = state.builder.AllocateId();
 			           const auto pointer_type =
 			               state.builder.Type(OpTypePointer, {StorageClassImage, TypeU32(state)});
 			           state.builder.AddFunction(
 			               {OpImageTexelPointer, pointer_type, pointer,
-			                StorageImageDescriptorPointer(state, mem.resource, view),
-			                CoordU32(ctx, mem, *address, view), ConstantU32(state, 0)});
+			                StorageImageDescriptorPointer(state, mem.resource),
+			                CoordU32(ctx, mem, *address, dimension), ConstantU32(state, 0)});
 			           const auto old = state.builder.AllocateId();
 			           state.builder.AddFunction({atomic_opcode, TypeU32(state), old, pointer,
 			                                      ConstantU32(state, ScopeDevice),

src/graphics/shader/recompiler/backend/spirv/spirvEmitterInternal.h

@@ -6,7 +6,6 @@
 #include "graphics/shader/recompiler/BufferFormat.h"
 #include "graphics/shader/recompiler/backend/spirv/SpirvBuilder.h"
 #include "graphics/shader/recompiler/ir/ShaderIR.h"
-#include "graphics/shader/recompiler/ir/ShaderIR.h"
 #include "graphics/shader/recompiler/ir/passes/BindingLayout.h"
 #include "graphics/shader/recompiler/ir/passes/ResourceMaterialization.h"
 
@@ -40,6 +39,7 @@ enum : uint32_t {
 	MemoryModelGLSL450                       = 1,
 	CapabilityShader                         = 1,
 	CapabilityInt64                          = 11,
+	CapabilityInt64Atomics                   = 12,
 	CapabilityImageGatherExtended            = 25,
 	CapabilityClipDistance                   = 32,
 	CapabilityCullDistance                   = 33,
@@ -74,6 +74,7 @@ enum : uint32_t {
 	DecorationBuiltIn       = 11,
 	DecorationNoPerspective = 13,
 	DecorationFlat          = 14,
+	DecorationAliased       = 20,
 	DecorationLocation      = 30,
 	DecorationArrayStride   = 6,
 	DecorationBinding       = 33,
@@ -129,6 +130,7 @@ enum : uint32_t {
 	MemorySemanticsAcquireRelease  = 0x00000008u,
 	MemorySemanticsUniformMemory   = 0x00000040u,
 	MemorySemanticsWorkgroupMemory = 0x00000100u,
+	MemorySemanticsImageMemory     = 0x00000800u,
 };
 
 enum : uint32_t {
@@ -318,61 +320,54 @@ struct OutputBinding {
 	std::string         debug_name;
 };
 
-struct DescriptorResourceBinding {
-	const IR::DescriptorBinding* descriptor  = nullptr;
-	uint32_t                     array_index = 0;
+using ImageDimension = Decoder::ImageDimension;
+
+struct ImageDimensionInfo {
+	ImageDimension dimension;
+	uint32_t       spirv_dimension;
+	uint32_t       coordinate_components;
+	uint32_t       spatial_components;
+	uint32_t       arrayed;
+	uint32_t       multisampled;
 };
 
-enum class ImageViewKind {
-	Dim1D,
-	Dim1DArray,
-	Dim2D,
-	Dim2DArray,
-	Dim3D,
-	Dim2DMsaa,
-	Dim2DMsaaArray,
-	Count,
-};
+constexpr std::array<ImageDimensionInfo, 7> ImageDimensions {{
+    {ImageDimension::Dim1D, Dim1D, 1, 1, 0, 0},
+    {ImageDimension::Dim1DArray, Dim1D, 2, 1, 1, 0},
+    {ImageDimension::Dim2D, Dim2D, 2, 2, 0, 0},
+    {ImageDimension::Dim2DArray, Dim2D, 3, 2, 1, 0},
+    {ImageDimension::Dim3D, Dim3D, 3, 3, 0, 0},
+    {ImageDimension::Dim2DMsaa, Dim2D, 2, 2, 0, 1},
+    {ImageDimension::Dim2DMsaaArray, Dim2D, 3, 2, 1, 1},
+}};
 
-constexpr uint32_t SampledImageViewKindCount = static_cast<uint32_t>(ImageViewKind::Count);
-constexpr uint32_t StorageImageViewKindCount = static_cast<uint32_t>(ImageViewKind::Dim2DMsaa);
-
-enum class StorageImageClass : uint32_t {
-	FormatlessFloat,
-	FormatlessUint,
-	AtomicUint,
-	Count,
-};
+const ImageDimensionInfo& ImageDimensionInfoFor(ImageDimension dimension);
 
 struct EmitterState {
-	EmitterState(const IR::Program& program_, const IR::ResourceSnapshot& resources_,
-	             ShaderStageInputInfo input_info_)
-	    : program(program_), resources(resources_), input_info(input_info_),
+	EmitterState(const IR::Program& program_, ShaderStageInputInfo input_info_)
+	    : program(program_), input_info(input_info_),
 	      requirements(*program_.spirv_requirements) {}
 
-	Builder                      builder;
-	const IR::Program&           program;
-	const IR::ResourceSnapshot&  resources;
-	ShaderStageInputInfo         input_info;
-	const IR::SpirvRequirements& requirements;
-	ShaderType                   stage                   = ShaderType::Unknown;
-	uint32_t                     wave_size               = 64;
-	uint32_t                     storage_buffer_variable = 0;
-	std::array<uint32_t, IR::ShaderInfo::MaxBuffers>      memory_byte_offsets {};
-	uint32_t                                             bda_pagetable_variable = 0;
-	uint32_t                                             fault_buffer_variable  = 0;
-	uint32_t                                             bda_pointer_function   = 0;
-	uint32_t                                             gds_variable            = 0;
-	uint32_t                                             gds_length              = 0;
-	uint32_t                                             push_constant_variable  = 0;
-	uint32_t                                             vsharp_storage_variable = 0;
-	uint32_t                                             flattened_srt_variable  = 0;
-	uint32_t                                             lds_variable            = 0;
-	uint32_t                                             scratch_variable        = 0;
-	std::array<uint32_t, 2u * SampledImageViewKindCount> sampled_image_variables {};
-	std::array<uint32_t,
-	           static_cast<uint32_t>(StorageImageClass::Count) * StorageImageViewKindCount>
-	                           storage_image_variables {};
+	Builder                                          builder;
+	const IR::Program&                               program;
+	ShaderStageInputInfo                             input_info;
+	const IR::SpirvRequirements&                     requirements;
+	ShaderType                                       stage                   = ShaderType::Unknown;
+	uint32_t                                         wave_size               = 64;
+	uint32_t                                         storage_buffer_variable = 0;
+	uint32_t                                         storage_buffer_u64_variable = 0;
+	std::array<uint32_t, IR::ShaderInfo::MaxBuffers> memory_byte_offsets {};
+	uint32_t                                         bda_pagetable_variable  = 0;
+	uint32_t                                         fault_buffer_variable   = 0;
+	uint32_t                                         bda_pointer_function    = 0;
+	uint32_t                                         gds_variable            = 0;
+	uint32_t                                         gds_length              = 0;
+	uint32_t                                         push_constant_variable  = 0;
+	uint32_t                                         shader_data_storage_variable = 0;
+	uint32_t                                         flattened_srt_variable  = 0;
+	uint32_t                                         lds_variable            = 0;
+	uint32_t                                         scratch_variable        = 0;
+	std::array<uint32_t, IR::ImageBindingCount>      image_variables {};
 	uint32_t                   sampler_variable                      = 0;
 	uint32_t                   main_func                             = 0;
 	uint32_t                   entry_label                           = 0;
@@ -399,6 +394,7 @@ uint32_t TypeBoolVector(EmitterState& state, uint32_t components);
 uint32_t TypeU32(EmitterState& state);
 uint32_t TypeU64(EmitterState& state);
 uint32_t TypeDeviceAddress(EmitterState& state);
+uint32_t TypeScalarU64(EmitterState& state);
 uint32_t TypeU32Pair(EmitterState& state);
 uint32_t TypeI32(EmitterState& state);
 uint32_t TypeI32Pair(EmitterState& state);
@@ -412,6 +408,8 @@ uint32_t TypePointer(EmitterState& state, uint32_t storage_class, uint32_t point
 uint32_t TypeFunction(EmitterState& state);
 uint32_t TypeStorageBufferPointer(EmitterState& state);
 uint32_t TypeStorageBufferElementPointer(EmitterState& state);
+uint32_t TypeStorageBufferU64Pointer(EmitterState& state);
+uint32_t TypeStorageBufferU64ElementPointer(EmitterState& state);
 uint32_t TypeDeviceAddressStoragePointer(EmitterState& state);
 uint32_t TypePhysicalU32Pointer(EmitterState& state);
 uint32_t TypePushConstantElementPointer(EmitterState& state);
@@ -439,18 +437,17 @@ struct ValueEmitContext {
 	const IR::MemoryInfo& Memory(const IR::Inst& inst) const;
 	const IR::ExportInfo& Export(const IR::Inst& inst) const;
 	uint32_t              Label(const IR::Block* block) const;
-	void                  Fail(const IR::Inst& inst, const char* reason);
+	[[noreturn]] void     Fail(const char* reason) const;
+	[[noreturn]] void     Fail(const IR::Inst& inst, const char* reason) const;
 
 	EmitterState&                                                      state;
-	const IR::Program&                                            program;
+	const IR::Program&                                                 program;
 	std::unordered_map<const IR::Inst*, uint32_t>                      definitions;
 	std::unordered_map<const IR::Block*, uint32_t>                     labels;
 	const std::unordered_map<const IR::Inst*, uint32_t>*               dispatcher_spills = nullptr;
 	std::unordered_map<const IR::Inst*, std::pair<uint32_t, uint32_t>> dispatcher_block_loads;
 	const IR::Block*                                                   current_block = nullptr;
 	uint32_t                                                           scratch_u32_variable = 0;
-	bool                                                               failed               = false;
-	std::string                                                        error;
 };
 
 enum class VertexInputScalarKind { Float, Sint, Uint };
@@ -472,119 +469,6 @@ struct ImageSampleLayout {
 	uint32_t grad_y = NoImageComponent;
 };
 
-constexpr uint32_t SampledImageIndex(bool integer, ImageViewKind view) {
-	return static_cast<uint32_t>(view) + (integer ? SampledImageViewKindCount : 0u);
-}
-
-constexpr StorageImageClass StorageImageClassFor(bool uint_image, bool atomic) {
-	if (atomic) {
-		return StorageImageClass::AtomicUint;
-	}
-	if (uint_image) {
-		return StorageImageClass::FormatlessUint;
-	}
-	return StorageImageClass::FormatlessFloat;
-}
-
-constexpr uint32_t StorageImageIndex(StorageImageClass image_class, ImageViewKind view) {
-	return static_cast<uint32_t>(view) +
-	       static_cast<uint32_t>(image_class) * StorageImageViewKindCount;
-}
-
-constexpr IR::DescriptorBindingKind SampledBindingKind(bool integer, ImageViewKind view) {
-	if (integer) {
-		switch (view) {
-			case ImageViewKind::Dim1D: return IR::DescriptorBindingKind::SampledUint1D;
-			case ImageViewKind::Dim1DArray: return IR::DescriptorBindingKind::SampledUint1DArray;
-			case ImageViewKind::Dim2D: return IR::DescriptorBindingKind::SampledUint2D;
-			case ImageViewKind::Dim2DArray: return IR::DescriptorBindingKind::SampledUint2DArray;
-			case ImageViewKind::Dim3D: return IR::DescriptorBindingKind::SampledUint3D;
-			case ImageViewKind::Dim2DMsaa: return IR::DescriptorBindingKind::SampledUint2DMsaa;
-			case ImageViewKind::Dim2DMsaaArray:
-				return IR::DescriptorBindingKind::SampledUint2DMsaaArray;
-			default: break;
-		}
-	}
-	switch (view) {
-		case ImageViewKind::Dim1D: return IR::DescriptorBindingKind::Sampled1D;
-		case ImageViewKind::Dim1DArray: return IR::DescriptorBindingKind::Sampled1DArray;
-		case ImageViewKind::Dim2D: return IR::DescriptorBindingKind::Sampled2D;
-		case ImageViewKind::Dim2DArray: return IR::DescriptorBindingKind::Sampled2DArray;
-		case ImageViewKind::Dim3D: return IR::DescriptorBindingKind::Sampled3D;
-		case ImageViewKind::Dim2DMsaa: return IR::DescriptorBindingKind::Sampled2DMsaa;
-		case ImageViewKind::Dim2DMsaaArray: return IR::DescriptorBindingKind::Sampled2DMsaaArray;
-		default: break;
-	}
-	return IR::DescriptorBindingKind::Count;
-}
-
-constexpr IR::DescriptorBindingKind StorageBindingKind(StorageImageClass image_class,
-                                                       ImageViewKind     view) {
-	switch (image_class) {
-		case StorageImageClass::FormatlessFloat:
-			switch (view) {
-				case ImageViewKind::Dim1D: return IR::DescriptorBindingKind::Storage1D;
-				case ImageViewKind::Dim1DArray: return IR::DescriptorBindingKind::Storage1DArray;
-				case ImageViewKind::Dim2D: return IR::DescriptorBindingKind::Storage2D;
-				case ImageViewKind::Dim2DArray: return IR::DescriptorBindingKind::Storage2DArray;
-				case ImageViewKind::Dim3D: return IR::DescriptorBindingKind::Storage3D;
-				default: break;
-			}
-			break;
-		case StorageImageClass::FormatlessUint:
-			switch (view) {
-				case ImageViewKind::Dim1D: return IR::DescriptorBindingKind::StorageUint1D;
-				case ImageViewKind::Dim1DArray:
-					return IR::DescriptorBindingKind::StorageUint1DArray;
-				case ImageViewKind::Dim2D: return IR::DescriptorBindingKind::StorageUint2D;
-				case ImageViewKind::Dim2DArray:
-					return IR::DescriptorBindingKind::StorageUint2DArray;
-				case ImageViewKind::Dim3D: return IR::DescriptorBindingKind::StorageUint3D;
-				default: break;
-			}
-			break;
-		case StorageImageClass::AtomicUint:
-			switch (view) {
-				case ImageViewKind::Dim1D: return IR::DescriptorBindingKind::StorageAtomic1D;
-				case ImageViewKind::Dim1DArray:
-					return IR::DescriptorBindingKind::StorageAtomic1DArray;
-				case ImageViewKind::Dim2D: return IR::DescriptorBindingKind::StorageAtomic2D;
-				case ImageViewKind::Dim2DArray:
-					return IR::DescriptorBindingKind::StorageAtomic2DArray;
-				case ImageViewKind::Dim3D: return IR::DescriptorBindingKind::StorageAtomic3D;
-				default: break;
-			}
-			break;
-		case StorageImageClass::Count: break;
-	}
-	return IR::DescriptorBindingKind::Count;
-}
-
-constexpr uint32_t ImageSpirvDimension(ImageViewKind view) {
-	switch (view) {
-		case ImageViewKind::Dim1D:
-		case ImageViewKind::Dim1DArray: return Dim1D;
-		case ImageViewKind::Dim2D:
-		case ImageViewKind::Dim2DArray:
-		case ImageViewKind::Dim2DMsaa:
-		case ImageViewKind::Dim2DMsaaArray:
-		case ImageViewKind::Count: return Dim2D;
-		case ImageViewKind::Dim3D: return Dim3D;
-	}
-	return Dim2D;
-}
-
-constexpr uint32_t ImageSpirvArrayed(ImageViewKind view) {
-	return view == ImageViewKind::Dim1DArray || view == ImageViewKind::Dim2DArray ||
-	               view == ImageViewKind::Dim2DMsaaArray
-	           ? 1u
-	           : 0u;
-}
-
-constexpr uint32_t ImageSpirvMultisampled(ImageViewKind view) {
-	return view == ImageViewKind::Dim2DMsaa || view == ImageViewKind::Dim2DMsaaArray ? 1u : 0u;
-}
-
 struct F32Class {
 	uint32_t bits = 0;
 	uint32_t nan  = 0;
@@ -613,8 +497,6 @@ uint32_t VertexParameterVectorOrScalarType(EmitterState& state, VertexInputScala
 uint32_t VertexParameterInputPointerType(EmitterState& state, VertexInputScalarKind kind,
                                          uint32_t components);
 
-void SetError(std::string* error, const char* message);
-
 bool HasOutput(const std::vector<OutputBinding>& outputs, IR::StageOutputKind kind, uint32_t index);
 
 void CopyProgramInputsAndOutputs(EmitterState& state, const IR::Program& program);
@@ -629,48 +511,33 @@ uint32_t EmitSubgroupLocalInvocationId(EmitterState& state);
                                                IR::DescriptorBindingKind kind, uint32_t resource,
                                                const char* reason);
 
-DescriptorResourceBinding ResourceForDescriptor(const EmitterState&       state,
-                                                IR::DescriptorBindingKind kind, uint32_t resource);
+uint32_t ResourceForDescriptor(const EmitterState& state, IR::DescriptorBindingKind kind,
+                               uint32_t resource);
 
 uint32_t DescriptorElementPointer(EmitterState& state, uint32_t result_ptr_type,
                                   uint32_t variable_id, uint32_t array_index,
                                   IR::DescriptorBindingKind kind, uint32_t resource,
                                   const char* variable_name);
 
-ImageViewKind SampledImageViewKind(const EmitterState& state, const IR::MemoryInfo& mem,
-                                   uint32_t use_pc);
+uint32_t ImageScalarType(EmitterState& state, Prospero::TextureNumericClass numeric_class);
 
-uint32_t ImageViewCoordinateComponents(ImageViewKind view);
+uint32_t ImageVectorType(EmitterState& state, Prospero::TextureNumericClass numeric_class,
+                         uint32_t components);
 
-uint32_t ImageViewSpatialComponents(ImageViewKind view);
+uint32_t ImageType(EmitterState& state, const IR::ImageResource& image);
 
-uint32_t ImageViewImageType(EmitterState& state, ImageViewKind view, bool integer);
+uint32_t ImageViewSizeType(EmitterState& state, ImageDimension dimension);
 
-uint32_t ImageViewSampledImageType(EmitterState& state, ImageViewKind view, bool integer);
+uint32_t LoadSampledImageDescriptor(EmitterState& state, uint32_t resource);
 
-uint32_t ImageViewSizeType(EmitterState& state, ImageViewKind view);
+uint32_t LoadSamplerDescriptor(EmitterState& state, uint32_t sampler);
 
-uint32_t StorageImageType(EmitterState& state, StorageImageClass image_class, ImageViewKind view);
+uint32_t MakeSampledImage(EmitterState& state, uint32_t resource, uint32_t sampler);
 
-uint32_t StorageImagePointerType(EmitterState& state, StorageImageClass image_class,
-                                 ImageViewKind view);
+uint32_t StorageImageDescriptorPointer(EmitterState& state, uint32_t resource);
 
-uint32_t LoadSampledImageDescriptor(EmitterState& state, const IR::MemoryInfo& mem, uint32_t use_pc,
-                                    ImageViewKind view);
-
-uint32_t LoadSamplerDescriptor(EmitterState& state, uint32_t sampler, uint32_t use_pc);
-
-uint32_t MakeSampledImage(EmitterState& state, const IR::MemoryInfo& mem, uint32_t use_pc,
-                          ImageViewKind view);
-uint32_t MakeSampledImage(EmitterState& state, const IR::MemoryInfo& mem, uint32_t use_pc,
-                          ImageViewKind view, uint32_t image_resource);
-
-ImageViewKind StorageImageViewKind(const IR::MemoryInfo& mem);
-
-uint32_t StorageImageDescriptorPointer(EmitterState& state, uint32_t resource, ImageViewKind view);
-
-void EmitStorageImageWrite(EmitterState& state, uint32_t resource, ImageViewKind view,
-                           uint32_t mip_lod, uint32_t coord, uint32_t texel);
+void EmitStorageImageWrite(EmitterState& state, uint32_t resource, uint32_t mip_lod, uint32_t coord,
+                           uint32_t texel);
 
 uint32_t ExecutionModelForStage(ShaderType stage);
 
@@ -758,11 +625,17 @@ struct MemoryResourceAccess {
 	uint32_t         object_pointer   = 0;
 	uint32_t         length           = 0;
 	uint32_t         index_offset     = 0;
+	uint32_t         byte_offset      = 0;
 	bool             add_index_offset = false;
 };
 
 MemoryResourceAccess PrepareMemoryResourceAccess(EmitterState& state, const IR::MemoryInfo& mem);
 
+MemoryResourceAccess PrepareStorageBufferResourceAccess(EmitterState& state,
+                                                         const IR::MemoryInfo& mem,
+                                                         uint32_t variable,
+                                                         uint32_t pointer_type);
+
 uint32_t EmitMemoryElementIndex(EmitterState& state, const MemoryResourceAccess& access,
                                 uint32_t raw_index);
 
@@ -772,6 +645,10 @@ uint32_t EmitMemoryElementInBounds(EmitterState& state, const MemoryResourceAcce
 uint32_t EmitMemoryElementPointer(EmitterState& state, const MemoryResourceAccess& access,
                                   uint32_t index);
 
+uint32_t EmitStorageBufferElementPointer(EmitterState& state,
+                                         const MemoryResourceAccess& access, uint32_t index,
+                                         uint32_t pointer_type);
+
 uint32_t EmitTBufferBitcastF32ToU32(EmitterState& state, uint32_t value);
 
 uint32_t EmitTBufferBitcastU32ToF32(EmitterState& state, uint32_t value);
@@ -795,6 +672,9 @@ uint32_t NormalizeFormatComponent(EmitterState& state, const Format::BufferForma
 
 void EmitDeviceAtomicMemoryBarrier(EmitterState& state);
 
+uint32_t EmitFloatAtomicReplacement(EmitterState& state, uint32_t old, uint32_t source,
+                                    bool max_value);
+
 uint32_t EmitDsSwizzleTargetLane(EmitterState& state, uint32_t subid, uint32_t control);
 
 uint32_t EmitSelectValueU32(EmitterState& state, uint32_t cond, uint32_t true_value,
@@ -864,7 +744,7 @@ bool EmitValueMemory(ValueEmitContext& ctx, const IR::Inst& inst);
 
 bool EmitValueImage(ValueEmitContext& ctx, const IR::Inst& inst);
 
-bool EmitProgram(EmitterState& state, const IR::Program& program, std::string* error);
+void EmitProgram(EmitterState& state, const IR::Program& program);
 
 void DefineGetBdaPointer(EmitterState& state);
 
@@ -910,6 +790,47 @@ uint32_t EmitValueOrZeroIfCondition(EmitterState& state, uint32_t condition, Fn&
 	                                     std::forward<Fn>(fn));
 }
 
+template <typename Fn>
+uint32_t AtomicUpdate(EmitterState& state, uint32_t pointer, IR::ResourceKind kind, Fn&& desired) {
+	const auto scope = kind == IR::ResourceKind::Lds ? ScopeWorkgroup : ScopeDevice;
+	const auto memory = [&] {
+		switch (kind) {
+			case IR::ResourceKind::Lds: return MemorySemanticsWorkgroupMemory;
+			case IR::ResourceKind::Image: return MemorySemanticsImageMemory;
+			default: return MemorySemanticsUniformMemory;
+		}
+	}();
+	const auto preheader = state.builder.AllocateId();
+	const auto header    = state.builder.AllocateId();
+	const auto cont      = state.builder.AllocateId();
+	const auto merge     = state.builder.AllocateId();
+	const auto initial   = state.builder.AllocateId();
+	const auto observed  = state.builder.AllocateId();
+	const auto exchanged = state.builder.AllocateId();
+	state.builder.AddFunction({OpBranch, preheader});
+	EmitLabel(state, preheader);
+	state.builder.AddFunction({OpAtomicLoad, TypeU32(state), initial, pointer,
+	                           ConstantU32(state, scope), ConstantU32(state, MemorySemanticsNone)});
+	state.builder.AddFunction({OpBranch, header});
+	EmitLabel(state, header);
+	state.builder.AddFunction(
+	    {OpPhi, TypeU32(state), observed, initial, preheader, exchanged, cont});
+	const auto next = desired(observed);
+	state.builder.AddFunction({OpAtomicCompareExchange, TypeU32(state), exchanged, pointer,
+	                           ConstantU32(state, scope), ConstantU32(state, MemorySemanticsNone),
+	                           ConstantU32(state, MemorySemanticsNone), next, observed});
+	const auto success = state.builder.AllocateId();
+	state.builder.AddFunction({OpIEqual, TypeBool(state), success, exchanged, observed});
+	state.builder.AddFunction({OpLoopMerge, merge, cont, LoopControlNone});
+	state.builder.AddFunction({OpBranchConditional, success, merge, cont});
+	EmitLabel(state, cont);
+	state.builder.AddFunction({OpBranch, header});
+	EmitLabel(state, merge);
+	state.builder.AddFunction({OpMemoryBarrier, ConstantU32(state, scope),
+	                           ConstantU32(state, MemorySemanticsAcquireRelease | memory)});
+	return observed;
+}
+
 } // namespace Libs::Graphics::ShaderRecompiler::Spirv::Emitter
 
 #endif /* EMULATOR_INCLUDE_EMULATOR_GRAPHICS_SHADER_RECOMPILER_SPIRVEMITTER_INTERNAL_H_ */

src/graphics/shader/recompiler/backend/spirv/spirvEmitterMemory.cpp

@@ -30,6 +30,19 @@ uint32_t AndCondition(EmitterState& state, uint32_t lhs, uint32_t rhs) {
 	return Binary(state, OpLogicalAnd, TypeBool(state), lhs, rhs);
 }
 
+uint32_t EmitDsMaskedLaneRead(EmitterState& state, uint32_t source, uint32_t target,
+                              uint32_t exec) {
+	const auto shuffled = state.builder.AllocateId();
+	state.builder.AddFunction({OpGroupNonUniformShuffle, TypeU32(state), shuffled,
+	                           ConstantU32(state, ScopeSubgroup), source, target});
+	const auto source_exec = state.builder.AllocateId();
+	state.builder.AddFunction({OpGroupNonUniformShuffle, TypeBool(state), source_exec,
+	                           ConstantU32(state, ScopeSubgroup), exec, target});
+	const auto source_active =
+	    AndCondition(state, source_exec, EmitSubgroupLaneActiveBool(state, target));
+	return Select(state, TypeU32(state), source_active, shuffled, ConstantU32(state, 0));
+}
+
 uint32_t BufferByteAddress(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem,
                            uint32_t index, uint32_t offset, uint32_t soffset) {
 	auto&          state   = ctx.state;
@@ -468,41 +481,6 @@ uint32_t FormattedLoad(ValueEmitContext& ctx, const IR::Inst& inst, const IR::Me
 	});
 }
 
-template <typename Fn>
-uint32_t AtomicUpdate(EmitterState& state, uint32_t pointer, IR::ResourceKind kind, Fn&& desired) {
-	const auto scope     = kind == IR::ResourceKind::Lds ? ScopeWorkgroup : ScopeDevice;
-	const auto memory    = kind == IR::ResourceKind::Lds ? MemorySemanticsWorkgroupMemory
-	                                                     : MemorySemanticsUniformMemory;
-	const auto preheader = state.builder.AllocateId();
-	const auto header    = state.builder.AllocateId();
-	const auto cont      = state.builder.AllocateId();
-	const auto merge     = state.builder.AllocateId();
-	const auto initial   = state.builder.AllocateId();
-	const auto observed  = state.builder.AllocateId();
-	const auto exchanged = state.builder.AllocateId();
-	state.builder.AddFunction({OpBranch, preheader});
-	EmitLabel(state, preheader);
-	state.builder.AddFunction({OpAtomicLoad, TypeU32(state), initial, pointer,
-	                           ConstantU32(state, scope), ConstantU32(state, MemorySemanticsNone)});
-	state.builder.AddFunction({OpBranch, header});
-	EmitLabel(state, header);
-	state.builder.AddFunction(
-	    {OpPhi, TypeU32(state), observed, initial, preheader, exchanged, cont});
-	const auto next = desired(observed);
-	state.builder.AddFunction({OpAtomicCompareExchange, TypeU32(state), exchanged, pointer,
-	                           ConstantU32(state, scope), ConstantU32(state, MemorySemanticsNone),
-	                           ConstantU32(state, MemorySemanticsNone), next, observed});
-	const auto success = Binary(state, OpIEqual, TypeBool(state), exchanged, observed);
-	state.builder.AddFunction({OpLoopMerge, merge, cont, LoopControlNone});
-	state.builder.AddFunction({OpBranchConditional, success, merge, cont});
-	EmitLabel(state, cont);
-	state.builder.AddFunction({OpBranch, header});
-	EmitLabel(state, merge);
-	state.builder.AddFunction({OpMemoryBarrier, ConstantU32(state, scope),
-	                           ConstantU32(state, MemorySemanticsAcquireRelease | memory)});
-	return observed;
-}
-
 void StoreSubwordInBounds(ValueEmitContext& ctx, const IR::MemoryInfo& mem,
                           const MemoryResourceAccess& resource, uint32_t address, uint32_t index,
                           uint32_t bits, uint32_t data) {
@@ -601,9 +579,11 @@ void FormattedStore(ValueEmitContext& ctx, const IR::Inst& inst, const IR::Memor
 	});
 }
 
-uint32_t AtomicOpcode(IR::ValueOpcode opcode) {
+uint32_t SpirvAtomicOpcode(IR::ValueOpcode opcode) {
 	switch (opcode) {
+		case IR::ValueOpcode::BufferAtomicCmpSwap32: return OpAtomicCompareExchange;
 		case IR::ValueOpcode::BufferAtomicSwap32:
+		case IR::ValueOpcode::BufferAtomicSwap64:
 		case IR::ValueOpcode::SharedAtomicSwap32: return OpAtomicExchange;
 		case IR::ValueOpcode::BufferAtomicIAdd32:
 		case IR::ValueOpcode::SharedAtomicIAdd32: return OpAtomicIAdd;
@@ -620,6 +600,7 @@ uint32_t AtomicOpcode(IR::ValueOpcode opcode) {
 		case IR::ValueOpcode::BufferAtomicAnd32:
 		case IR::ValueOpcode::SharedAtomicAnd32: return OpAtomicAnd;
 		case IR::ValueOpcode::BufferAtomicOr32:
+		case IR::ValueOpcode::BufferAtomicOr64:
 		case IR::ValueOpcode::SharedAtomicOr32: return OpAtomicOr;
 		case IR::ValueOpcode::BufferAtomicXor32:
 		case IR::ValueOpcode::SharedAtomicXor32: return OpAtomicXor;
@@ -627,23 +608,37 @@ uint32_t AtomicOpcode(IR::ValueOpcode opcode) {
 	}
 }
 
-uint32_t EmitAtomic(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem,
-                    bool return_value) {
+uint32_t EmitAtomicOperation(ValueEmitContext& ctx, const IR::Inst& inst, uint32_t pointer,
+                             uint32_t scope) {
+	const auto old = ctx.state.builder.AllocateId();
+	if (inst.GetOpcode() == IR::ValueOpcode::BufferAtomicCmpSwap32) {
+		const auto desired    = ctx.Arg(inst, inst.NumArgs() - 3);
+		const auto comparator = ctx.Arg(inst, inst.NumArgs() - 2);
+		ctx.state.builder.AddFunction(
+		    {OpAtomicCompareExchange, TypeU32(ctx.state), old, pointer, ConstantU32(ctx.state, scope),
+		     ConstantU32(ctx.state, MemorySemanticsNone),
+		     ConstantU32(ctx.state, MemorySemanticsNone), desired, comparator});
+	} else {
+		const auto value = ctx.Arg(inst, inst.NumArgs() - 2);
+		ctx.state.builder.AddFunction(
+		    {SpirvAtomicOpcode(inst.GetOpcode()), TypeU32(ctx.state), old, pointer,
+		     ConstantU32(ctx.state, scope), ConstantU32(ctx.state, MemorySemanticsNone), value});
+	}
+	return old;
+}
+
+uint32_t EmitAtomic(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem) {
 	const auto result =
 	    EmitValueOrZeroIfCondition(ctx.state, ctx.Arg(inst, inst.NumArgs() - 1), [&]() {
 		    const auto access = PrepareMemoryElement(ctx, mem, DwordIndex(ctx, inst, mem));
 		    return EmitValueOrZeroIfCondition(
 		        ctx.state, EmitMemoryElementInBounds(ctx.state, access.resource, access.index),
 		        [&]() {
-			        const auto value = ctx.Arg(inst, inst.NumArgs() - 2);
-			        const auto old   = ctx.state.builder.AllocateId();
 			        const auto scope =
 			            mem.kind == IR::ResourceKind::Lds ? ScopeWorkgroup : ScopeDevice;
-			        ctx.state.builder.AddFunction(
-			            {AtomicOpcode(inst.GetOpcode()), TypeU32(ctx.state), old,
-			             EmitMemoryElementPointer(ctx.state, access.resource, access.index),
-			             ConstantU32(ctx.state, scope), ConstantU32(ctx.state, MemorySemanticsNone),
-			             value});
+			        const auto pointer =
+			            EmitMemoryElementPointer(ctx.state, access.resource, access.index);
+			        const auto old = EmitAtomicOperation(ctx, inst, pointer, scope);
 			        if (mem.kind == IR::ResourceKind::Lds) {
 				        const auto semantics =
 				            MemorySemanticsAcquireRelease | MemorySemanticsWorkgroupMemory;
@@ -656,40 +651,36 @@ uint32_t EmitAtomic(ValueEmitContext& ctx, const IR::Inst& inst, const IR::Memor
 			        return old;
 		        });
 	    });
-	return return_value ? result : ConstantU32(ctx.state, 0);
+	return result;
 }
 
-uint32_t FloatAtomicReplacement(EmitterState& state, uint32_t old, uint32_t source,
-                                bool max_value) {
-	struct OrderedBits {
-		uint32_t nan;
-		uint32_t zero;
-		uint32_t key;
-	};
-	const auto classify = [&](uint32_t bits) {
-		const auto value    = Unary(state, OpBitcast, TypeF32(state), bits);
-		const auto cls      = EmitClassifyF32(state, value);
-		const auto negative = Binary(
-		    state, OpINotEqual, TypeBool(state),
-		    Binary(state, OpBitwiseAnd, TypeU32(state), bits, ConstantU32(state, 0x80000000u)),
-		    ConstantU32(state, 0));
-		const auto negative_key = Unary(state, OpNot, TypeU32(state), bits);
-		const auto positive_key =
-		    Binary(state, OpBitwiseXor, TypeU32(state), bits, ConstantU32(state, 0x80000000u));
-		return OrderedBits {cls.nan, cls.zero,
-		                    Select(state, TypeU32(state), negative, negative_key, positive_key)};
-	};
-	const auto source_class = classify(source);
-	const auto old_class    = classify(old);
-	const auto unordered =
-	    Binary(state, OpLogicalOr, TypeBool(state),
-	           Binary(state, OpLogicalOr, TypeBool(state), source_class.nan, old_class.nan),
-	           Binary(state, OpLogicalAnd, TypeBool(state), source_class.zero, old_class.zero));
-	const auto ordered = Unary(state, OpLogicalNot, TypeBool(state), unordered);
-	const auto compare = Binary(state, max_value ? OpUGreaterThan : OpULessThan, TypeBool(state),
-	                            source_class.key, old_class.key);
-	return Select(state, TypeU32(state),
-	              Binary(state, OpLogicalAnd, TypeBool(state), ordered, compare), source, old);
+uint32_t EmitBufferAtomic64(ValueEmitContext& ctx, const IR::Inst& inst,
+                            const IR::MemoryInfo& mem) {
+	auto& state = ctx.state;
+	return EmitValueOrDefaultIfCondition(
+	    state, ctx.Arg(inst, inst.NumArgs() - 1), TypeU64(state), ConstantU64(state, 0), [&]() {
+		    const auto resource = PrepareStorageBufferResourceAccess(
+		        state, mem, state.storage_buffer_u64_variable, TypeStorageBufferU64Pointer(state));
+		    const auto byte_address = Binary(state, OpIAdd, TypeU32(state),
+		                                     ByteAddress(ctx, inst, mem), resource.byte_offset);
+		    const auto index = Binary(state, OpShiftRightLogical, TypeU32(state), byte_address,
+		                              ConstantU32(state, 3u));
+		    return EmitValueOrDefaultIfCondition(
+		        state, EmitMemoryElementInBounds(state, resource, index), TypeU64(state),
+		        ConstantU64(state, 0), [&]() {
+			        const auto value = Unary(state, OpBitcast, TypeScalarU64(state),
+			                                 ctx.Arg(inst, inst.NumArgs() - 2));
+			        const auto old   = state.builder.AllocateId();
+			        state.builder.AddFunction(
+			            {SpirvAtomicOpcode(inst.GetOpcode()), TypeScalarU64(state), old,
+			             EmitStorageBufferElementPointer(
+			                 state, resource, index, TypeStorageBufferU64ElementPointer(state)),
+			             ConstantU32(state, ScopeDevice),
+			             ConstantU32(state, MemorySemanticsNone), value});
+			        EmitDeviceAtomicMemoryBarrier(state);
+			        return Unary(state, OpBitcast, TypeU64(state), old);
+		        });
+	    });
 }
 
 uint32_t FloatAtomic(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem,
@@ -702,7 +693,7 @@ uint32_t FloatAtomic(ValueEmitContext& ctx, const IR::Inst& inst, const IR::Memo
 			        EmitMemoryElementPointer(ctx.state, access.resource, access.index);
 			    const auto source = ctx.Arg(inst, inst.NumArgs() - 2);
 			    return AtomicUpdate(ctx.state, pointer, mem.kind, [&](uint32_t old) {
-				    return FloatAtomicReplacement(ctx.state, old, source, max_value);
+				    return EmitFloatAtomicReplacement(ctx.state, old, source, max_value);
 			    });
 		    });
 	});
@@ -1067,12 +1058,15 @@ bool EmitValueMemory(ValueEmitContext& ctx, const IR::Inst& inst) {
 	const auto op                = inst.GetOpcode();
 	const auto buffer_components = IR::BufferComponentCount(op);
 	if (buffer_components > 1u) {
-		if (IR::BufferAccessOf(op) == IR::BufferAccess::Read) {
+		const auto access = IR::BufferAccessOf(op);
+		if (access == IR::BufferAccess::Read) {
 			ctx.Define(inst, LoadWideBuffer(ctx, inst, buffer_components));
-		} else {
+			return true;
+		}
+		if (access == IR::BufferAccess::Write) {
 			StoreWideBuffer(ctx, inst, buffer_components);
+			return true;
 		}
-		return true;
 	}
 	const auto shared_components = IR::SharedComponentCount(op);
 	if (shared_components > 1u) {
@@ -1163,9 +1157,13 @@ bool EmitValueMemory(ValueEmitContext& ctx, const IR::Inst& inst) {
 			StoreWord(ctx, inst, mem);
 		return true;
 	}
-	const auto atomic_opcode = AtomicOpcode(op);
+	const auto atomic_opcode = SpirvAtomicOpcode(op);
+	if (atomic_opcode != 0 && inst.GetType() == IR::Type::U64) {
+		ctx.Define(inst, EmitBufferAtomic64(ctx, inst, ctx.Memory(inst)));
+		return true;
+	}
 	if (atomic_opcode != 0) {
-		ctx.Define(inst, EmitAtomic(ctx, inst, ctx.Memory(inst), true));
+		ctx.Define(inst, EmitAtomic(ctx, inst, ctx.Memory(inst)));
 		return true;
 	}
 	if (op == IR::ValueOpcode::BufferAtomicFMin32 || op == IR::ValueOpcode::BufferAtomicFMax32) {
@@ -1181,24 +1179,27 @@ bool EmitValueMemory(ValueEmitContext& ctx, const IR::Inst& inst) {
 		ctx.Define(inst, AppendConsume(ctx, inst, op == IR::ValueOpcode::DataAppend));
 		return true;
 	}
-	if (op == IR::ValueOpcode::SwizzleU32) {
-		ctx.state.builder.AddFunction({OpStore, ctx.scratch_u32_variable, ctx.Arg(inst, 0)});
-		const auto source = ctx.state.builder.AllocateId();
-		ctx.state.builder.AddFunction(
-		    {OpLoad, TypeU32(ctx.state), source, ctx.scratch_u32_variable});
-		const auto target =
-		    EmitDsSwizzleTargetLane(state, EmitSubgroupLocalInvocationId(state),
-		                            inst.Arg(1).IsImmediate() ? inst.Arg(1).U32() : 0);
-		const auto shuffled = state.builder.AllocateId();
-		state.builder.AddFunction({OpGroupNonUniformShuffle, TypeU32(state), shuffled,
-		                           ConstantU32(state, ScopeSubgroup), source, target});
-		const auto source_exec = state.builder.AllocateId();
-		state.builder.AddFunction({OpGroupNonUniformShuffle, TypeBool(state), source_exec,
-		                           ConstantU32(state, ScopeSubgroup), ctx.Arg(inst, 2), target});
-		const auto source_active =
-		    AndCondition(state, source_exec, EmitSubgroupLaneActiveBool(state, target));
-		ctx.Define(inst,
-		           Select(state, TypeU32(state), source_active, shuffled, ConstantU32(state, 0)));
+	if (op == IR::ValueOpcode::SwizzleU32 || op == IR::ValueOpcode::BpermuteU32) {
+		uint32_t source = ctx.Arg(inst, 0);
+		uint32_t target = 0;
+		if (op == IR::ValueOpcode::SwizzleU32) {
+			state.builder.AddFunction({OpStore, ctx.scratch_u32_variable, source});
+			source = state.builder.AllocateId();
+			state.builder.AddFunction({OpLoad, TypeU32(state), source, ctx.scratch_u32_variable});
+			target = EmitDsSwizzleTargetLane(state, EmitSubgroupLocalInvocationId(state),
+			                                 inst.Arg(1).IsImmediate() ? inst.Arg(1).U32() : 0);
+		} else {
+			const auto index = Binary(
+			    state, OpBitwiseAnd, TypeU32(state),
+			    Binary(state, OpShiftRightLogical, TypeU32(state), ctx.Arg(inst, 1),
+			           ConstantU32(state, 2)),
+			    ConstantU32(state, 31));
+			const auto base =
+			    Binary(state, OpBitwiseAnd, TypeU32(state), EmitSubgroupLocalInvocationId(state),
+			           ConstantU32(state, ~31u));
+			target = Binary(state, OpBitwiseOr, TypeU32(state), base, index);
+		}
+		ctx.Define(inst, EmitDsMaskedLaneRead(state, source, target, ctx.Arg(inst, 2)));
 		return true;
 	}
 	return false;

src/graphics/shader/recompiler/backend/spirv/spirvEmitterMemoryHelpers.cpp

@@ -4,6 +4,7 @@ namespace Libs::Graphics::ShaderRecompiler::Spirv::Emitter {
 
 uint32_t EmitShaderDataDwordLoad(EmitterState& state, uint32_t dword_index) {
 	if (state.push_constant_variable != 0) {
+		dword_index += state.program.bindings.push_data_start_dword;
 		const auto pointer = state.builder.AllocateId();
 		const auto value   = state.builder.AllocateId();
 		state.builder.AddFunction({OpAccessChain, TypePushConstantElementPointer(state), pointer,
@@ -12,11 +13,11 @@ uint32_t EmitShaderDataDwordLoad(EmitterState& state, uint32_t dword_index) {
 		state.builder.AddFunction({OpLoad, TypeU32(state), value, pointer});
 		return value;
 	}
-	if (state.vsharp_storage_variable != 0) {
+	if (state.shader_data_storage_variable != 0) {
 		const auto pointer = state.builder.AllocateId();
 		const auto value   = state.builder.AllocateId();
 		state.builder.AddFunction({OpAccessChain, TypeStorageBufferElementPointer(state), pointer,
-		                           state.vsharp_storage_variable, ConstantU32(state, 0),
+		                           state.shader_data_storage_variable, ConstantU32(state, 0),
 		                           ConstantU32(state, dword_index)});
 		state.builder.AddFunction({OpLoad, TypeU32(state), value, pointer});
 		return value;
@@ -80,6 +81,27 @@ static void EnsureLdsStorage(EmitterState& state) {
 	state.builder.AddName(state.lds_variable, "lds_dwords");
 }
 
+MemoryResourceAccess PrepareStorageBufferResourceAccess(EmitterState& state,
+                                                         const IR::MemoryInfo& mem,
+                                                         uint32_t variable,
+                                                         uint32_t pointer_type) {
+	if (variable == 0) {
+		ExitDescriptorBindingFailure(state, IR::DescriptorBindingKind::Buffers, mem.resource,
+		                             "storage buffer descriptor array was not emitted");
+	}
+	const auto array_index =
+	    ResourceForDescriptor(state, IR::DescriptorBindingKind::Buffers, mem.resource);
+	MemoryResourceAccess access {.kind = mem.kind};
+	access.object_pointer = state.builder.AllocateId();
+	state.builder.AddFunction({OpAccessChain, pointer_type, access.object_pointer, variable,
+	                           ConstantU32(state, array_index)});
+	access.byte_offset = state.memory_byte_offsets[array_index];
+	access.length      = state.builder.AllocateId();
+	state.builder.AddFunction(
+	    {OpArrayLength, TypeU32(state), access.length, access.object_pointer, 0});
+	return access;
+}
+
 MemoryResourceAccess PrepareMemoryResourceAccess(EmitterState& state, const IR::MemoryInfo& mem) {
 	MemoryResourceAccess access {.kind = mem.kind};
 	switch (mem.kind) {
@@ -112,22 +134,13 @@ MemoryResourceAccess PrepareMemoryResourceAccess(EmitterState& state, const IR::
 			EXIT("physical address memory must use the BDA emitter\n");
 		case IR::ResourceKind::ScalarBuffer:
 		case IR::ResourceKind::Buffer: {
-			if (state.storage_buffer_variable == 0) {
-				ExitDescriptorBindingFailure(state, IR::DescriptorBindingKind::Buffers,
-				                             mem.resource,
-				                             "storage buffer descriptor array was not emitted");
-			}
-			const auto binding =
-			    ResourceForDescriptor(state, IR::DescriptorBindingKind::Buffers, mem.resource);
-			access.object_pointer = state.builder.AllocateId();
-			state.builder.AddFunction({OpAccessChain, TypeStorageBufferPointer(state),
-			                           access.object_pointer, state.storage_buffer_variable,
-			                           ConstantU32(state, binding.array_index)});
-			access.index_offset     = EmitBinaryU32(state, OpShiftRightLogical,
-			                                        state.memory_byte_offsets[binding.array_index],
-			                                        ConstantU32(state, 2u));
+			access = PrepareStorageBufferResourceAccess(
+			    state, mem, state.storage_buffer_variable, TypeStorageBufferPointer(state));
+			access.index_offset =
+			    EmitBinaryU32(state, OpShiftRightLogical, access.byte_offset,
+			                  ConstantU32(state, 2u));
 			access.add_index_offset = true;
-			break;
+			return access;
 		}
 		default: EXIT("unsupported memory resource kind: %u\n", static_cast<unsigned>(mem.kind));
 	}
@@ -151,17 +164,25 @@ uint32_t EmitMemoryElementInBounds(EmitterState& state, const MemoryResourceAcce
 
 uint32_t EmitMemoryElementPointer(EmitterState& state, const MemoryResourceAccess& access,
                                   uint32_t index) {
-	const auto pointer = state.builder.AllocateId();
 	if (access.kind == IR::ResourceKind::Lds || access.kind == IR::ResourceKind::Scratch) {
+		const auto pointer = state.builder.AllocateId();
 		const auto storage_class = access.kind == IR::ResourceKind::Scratch ? StorageClassFunction
 		                           : state.stage == ShaderType::Compute     ? StorageClassWorkgroup
 		                                                                    : StorageClassFunction;
 		state.builder.AddFunction({OpAccessChain, TypeU32ElementPointer(state, storage_class),
 		                           pointer, access.object_pointer, index});
-	} else {
-		state.builder.AddFunction({OpAccessChain, TypeStorageBufferElementPointer(state), pointer,
-		                           access.object_pointer, ConstantU32(state, 0), index});
+		return pointer;
 	}
+	return EmitStorageBufferElementPointer(state, access, index,
+	                                       TypeStorageBufferElementPointer(state));
+}
+
+uint32_t EmitStorageBufferElementPointer(EmitterState& state,
+                                         const MemoryResourceAccess& access, uint32_t index,
+                                         uint32_t pointer_type) {
+	const auto pointer = state.builder.AllocateId();
+	state.builder.AddFunction({OpAccessChain, pointer_type, pointer, access.object_pointer,
+	                           ConstantU32(state, 0), index});
 	return pointer;
 }
 
@@ -304,6 +325,39 @@ void EmitDeviceAtomicMemoryBarrier(EmitterState& state) {
 	    {OpMemoryBarrier, ConstantU32(state, ScopeDevice), ConstantU32(state, semantics)});
 }
 
+uint32_t EmitFloatAtomicReplacement(EmitterState& state, uint32_t old, uint32_t source,
+                                    bool max_value) {
+	struct OrderedBits {
+		uint32_t nan;
+		uint32_t zero;
+		uint32_t key;
+	};
+	const auto classify = [&](uint32_t bits) {
+		const auto value = EmitBitcastU32ToF32(state, bits);
+		const auto cls   = EmitClassifyF32(state, value);
+		const auto negative = EmitCompareU32Constant(
+		    state, OpINotEqual, EmitAndConstant(state, bits, 0x80000000u), 0u);
+		const auto negative_key = state.builder.AllocateId();
+		state.builder.AddFunction({OpNot, TypeU32(state), negative_key, bits});
+		const auto positive_key = EmitBinaryU32(
+		    state, OpBitwiseXor, bits, ConstantU32(state, 0x80000000u));
+		return OrderedBits {
+		    cls.nan, cls.zero,
+		    EmitSelectValueU32(state, negative, negative_key, positive_key)};
+	};
+	const auto source_class = classify(source);
+	const auto old_class    = classify(old);
+	const auto unordered = EmitLogicalOrBool(
+	    state, EmitLogicalOrBool(state, source_class.nan, old_class.nan),
+	    EmitLogicalAndBool(state, source_class.zero, old_class.zero));
+	const auto compare = state.builder.AllocateId();
+	state.builder.AddFunction({max_value ? OpUGreaterThan : OpULessThan, TypeBool(state), compare,
+	                           source_class.key, old_class.key});
+	return EmitSelectValueU32(
+	    state, EmitLogicalAndBool(state, EmitLogicalNotBool(state, unordered), compare), source,
+	    old);
+}
+
 uint32_t EmitDsSwizzleTargetLane(EmitterState& state, uint32_t subid, uint32_t control) {
 	if ((control & 0xc000u) == 0xc000u) {
 		const uint32_t mask         = control & 0x1fu;

src/graphics/shader/recompiler/backend/spirv/spirvEmitterModule.cpp

@@ -29,6 +29,10 @@ uint32_t TypeU64(EmitterState& state) {
 }
 
 uint32_t TypeDeviceAddress(EmitterState& state) {
+	return TypeScalarU64(state);
+}
+
+uint32_t TypeScalarU64(EmitterState& state) {
 	return state.builder.Type(OpTypeInt, {64, 0});
 }
 
@@ -94,6 +98,25 @@ uint32_t TypeStorageBufferElementPointer(EmitterState& state) {
 	return TypePointer(state, StorageClassStorageBuffer, TypeU32(state));
 }
 
+uint32_t StorageU64RuntimeArrayType(EmitterState& state) {
+	return state.builder.DecoratedType(OpTypeRuntimeArray, {TypeScalarU64(state)},
+	                                   {{OpDecorate, {DecorationArrayStride, sizeof(uint64_t)}}});
+}
+
+uint32_t StorageBufferU64Type(EmitterState& state) {
+	return state.builder.DecoratedType(
+	    OpTypeStruct, {StorageU64RuntimeArrayType(state)},
+	    {{OpMemberDecorate, {0, DecorationOffset, 0}}, {OpDecorate, {DecorationBlock}}});
+}
+
+uint32_t TypeStorageBufferU64Pointer(EmitterState& state) {
+	return TypePointer(state, StorageClassStorageBuffer, StorageBufferU64Type(state));
+}
+
+uint32_t TypeStorageBufferU64ElementPointer(EmitterState& state) {
+	return TypePointer(state, StorageClassStorageBuffer, TypeScalarU64(state));
+}
+
 uint32_t TypeDeviceAddressStoragePointer(EmitterState& state) {
 	return TypePointer(state, StorageClassStorageBuffer, TypeDeviceAddress(state));
 }
@@ -119,8 +142,7 @@ uint32_t TypeU32ElementPointer(EmitterState& state, uint32_t storage_class) {
 namespace {
 
 uint32_t PushConstantArrayType(EmitterState& state) {
-	const auto count =
-	    ConstantU32(state, state.program.bindings.push_constant_size / sizeof(uint32_t));
+	const auto count = ConstantU32(state, IR::PushData::DwordCount);
 	return state.builder.DecoratedType(OpTypeArray, {TypeU32(state), count},
 	                                   {{OpDecorate, {DecorationArrayStride, sizeof(uint32_t)}}});
 }
@@ -128,7 +150,7 @@ uint32_t PushConstantArrayType(EmitterState& state) {
 uint32_t PushConstantBlockType(EmitterState& state) {
 	return state.builder.DecoratedType(
 	    OpTypeStruct, {PushConstantArrayType(state)},
-	    {{OpMemberDecorate, {0, DecorationOffset, state.program.bindings.push_constant_offset}},
+	    {{OpMemberDecorate, {0, DecorationOffset, 0}},
 	     {OpDecorate, {DecorationBlock}}});
 }
 
@@ -163,6 +185,13 @@ void DefineDescriptorVariables(EmitterState& state) {
 		const auto pointer_type = TypePointer(state, StorageClassStorageBuffer, array_type);
 		state.storage_buffer_variable =
 		    state.builder.DefineGlobalVariable(pointer_type, StorageClassStorageBuffer);
+		if (state.requirements.buffer_int64_atomics) {
+			const auto u64_array_type =
+			    state.builder.Type(OpTypeArray, {StorageBufferU64Type(state), count});
+			state.storage_buffer_u64_variable = state.builder.DefineGlobalVariable(
+			    TypePointer(state, StorageClassStorageBuffer, u64_array_type),
+			    StorageClassStorageBuffer);
+		}
 	}
 	if (DescriptorBinding(state, IR::DescriptorBindingKind::BdaPagetable) != nullptr) {
 		state.bda_pagetable_variable = state.builder.DefineGlobalVariable(
@@ -173,52 +202,37 @@ void DefineDescriptorVariables(EmitterState& state) {
 		state.fault_buffer_variable = state.builder.DefineGlobalVariable(
 		    TypeStorageBufferPointer(state), StorageClassStorageBuffer);
 	}
-	if (state.program.bindings.push_constant_size != 0) {
+	if (state.program.bindings.UsesPushData()) {
 		const auto pointer_type =
 		    TypePointer(state, StorageClassPushConstant, PushConstantBlockType(state));
 		state.push_constant_variable =
 		    state.builder.DefineGlobalVariable(pointer_type, StorageClassPushConstant);
 	}
-	if (DescriptorBinding(state, IR::DescriptorBindingKind::UserData) != nullptr) {
-		state.vsharp_storage_variable = state.builder.DefineGlobalVariable(
+	if (DescriptorBinding(state, IR::DescriptorBindingKind::ShaderData) != nullptr) {
+		state.shader_data_storage_variable = state.builder.DefineGlobalVariable(
 		    TypeStorageBufferPointer(state), StorageClassStorageBuffer);
 	}
 	if (DescriptorBinding(state, IR::DescriptorBindingKind::FlattenedSrt) != nullptr) {
 		state.flattened_srt_variable = state.builder.DefineGlobalVariable(
 		    TypeStorageBufferPointer(state), StorageClassStorageBuffer);
 	}
-	for (uint32_t i = 0; i < state.sampled_image_variables.size(); i++) {
-		const auto view    = static_cast<ImageViewKind>(i % SampledImageViewKindCount);
-		const bool integer = i >= SampledImageViewKindCount;
-		const auto kind    = SampledBindingKind(integer, view);
-		if (DescriptorBinding(state, kind) == nullptr) {
+	for (const auto& binding: state.program.bindings.descriptors) {
+		if (IR::ImageBindingResourceClass(binding.kind) == IR::ImageResourceClass::None) {
 			continue;
 		}
-		const auto count        = ConstantU32(state, DescriptorCount(state, kind));
-		const auto image_type   = ImageViewImageType(state, view, integer);
-		const auto array_type   = state.builder.Type(OpTypeArray, {image_type, count});
-		const auto pointer_type = TypePointer(state, StorageClassUniformConstant, array_type);
-		state.sampled_image_variables[i] =
+		const auto& image        = state.program.info.images.at(binding.resources.front());
+		const auto  count        = ConstantU32(state, DescriptorCount(state, binding.kind));
+		const auto  image_type   = ImageType(state, image);
+		const auto  array_type   = state.builder.Type(OpTypeArray, {image_type, count});
+		const auto  pointer_type = TypePointer(state, StorageClassUniformConstant, array_type);
+		state.image_variables[IR::ImageBindingIndex(binding.kind)] =
 		    state.builder.DefineGlobalVariable(pointer_type, StorageClassUniformConstant);
-		if (view == ImageViewKind::Dim1D || view == ImageViewKind::Dim1DArray) {
-			state.builder.RequireCapability(CapabilitySampled1D);
-		}
-	}
-	for (uint32_t i = 0; i < state.storage_image_variables.size(); i++) {
-		const auto view        = static_cast<ImageViewKind>(i % StorageImageViewKindCount);
-		const auto image_class = static_cast<StorageImageClass>(i / StorageImageViewKindCount);
-		const auto kind        = StorageBindingKind(image_class, view);
-		if (DescriptorBinding(state, kind) == nullptr) {
-			continue;
-		}
-		const auto count        = ConstantU32(state, DescriptorCount(state, kind));
-		const auto image_type   = StorageImageType(state, image_class, view);
-		const auto array_type   = state.builder.Type(OpTypeArray, {image_type, count});
-		const auto pointer_type = TypePointer(state, StorageClassUniformConstant, array_type);
-		state.storage_image_variables[i] =
-		    state.builder.DefineGlobalVariable(pointer_type, StorageClassUniformConstant);
-		if (view == ImageViewKind::Dim1D || view == ImageViewKind::Dim1DArray) {
-			state.builder.RequireCapability(CapabilityImage1D);
+		if (image.dimension == ImageDimension::Dim1D ||
+		    image.dimension == ImageDimension::Dim1DArray) {
+			const auto capability = image.resource_class == IR::ImageResourceClass::Sampled
+			                            ? CapabilitySampled1D
+			                            : CapabilityImage1D;
+			state.builder.RequireCapability(capability);
 		}
 	}
 	if (DescriptorBinding(state, IR::DescriptorBindingKind::Samplers) != nullptr) {
@@ -586,6 +600,14 @@ void AddDescriptorAnnotationsAndNames(EmitterState& state) {
 	if (state.storage_buffer_variable != 0) {
 		Decorate(state.storage_buffer_variable, "buffers", IR::DescriptorBindingKind::Buffers);
 	}
+	if (state.storage_buffer_u64_variable != 0) {
+		Decorate(state.storage_buffer_u64_variable, "buffers_u64",
+		         IR::DescriptorBindingKind::Buffers);
+		state.builder.AddAnnotation(
+		    {OpDecorate, state.storage_buffer_variable, DecorationAliased});
+		state.builder.AddAnnotation(
+		    {OpDecorate, state.storage_buffer_u64_variable, DecorationAliased});
+	}
 	if (state.bda_pagetable_variable != 0) {
 		Decorate(state.bda_pagetable_variable, "bda_pagetable",
 		         IR::DescriptorBindingKind::BdaPagetable);
@@ -594,45 +616,13 @@ void AddDescriptorAnnotationsAndNames(EmitterState& state) {
 		Decorate(state.fault_buffer_variable, "fault_buffer",
 		         IR::DescriptorBindingKind::FaultBuffer);
 	}
-	constexpr const char* SampledNames[] = {"sampled_1d",
-	                                        "sampled_1d_array",
-	                                        "sampled_2d",
-	                                        "sampled_2d_array",
-	                                        "sampled_3d",
-	                                        "sampled_2d_msaa",
-	                                        "sampled_2d_msaa_array",
-	                                        "sampled_uint_1d",
-	                                        "sampled_uint_1d_array",
-	                                        "sampled_uint_2d",
-	                                        "sampled_uint_2d_array",
-	                                        "sampled_uint_3d",
-	                                        "sampled_uint_2d_msaa",
-	                                        "sampled_uint_2d_msaa_array"};
-	for (uint32_t i = 0; i < state.sampled_image_variables.size(); i++) {
-		const auto view = static_cast<ImageViewKind>(i % SampledImageViewKindCount);
-		Decorate(state.sampled_image_variables[i], SampledNames[i],
-		         SampledBindingKind(i >= SampledImageViewKindCount, view));
-	}
-	constexpr const char* StorageNames[] = {"storage_1d",
-	                                        "storage_1d_array",
-	                                        "storage_2d",
-	                                        "storage_2d_array",
-	                                        "storage_3d",
-	                                        "storage_uint_1d",
-	                                        "storage_uint_1d_array",
-	                                        "storage_uint_2d",
-	                                        "storage_uint_2d_array",
-	                                        "storage_uint_3d",
-	                                        "storage_atomic_1d",
-	                                        "storage_atomic_1d_array",
-	                                        "storage_atomic_2d",
-	                                        "storage_atomic_2d_array",
-	                                        "storage_atomic_3d"};
-	for (uint32_t i = 0; i < state.storage_image_variables.size(); i++) {
-		const auto view        = static_cast<ImageViewKind>(i % StorageImageViewKindCount);
-		const auto image_class = static_cast<StorageImageClass>(i / StorageImageViewKindCount);
-		Decorate(state.storage_image_variables[i], StorageNames[i],
-		         StorageBindingKind(image_class, view));
+	for (const auto& binding: state.program.bindings.descriptors) {
+		if (IR::ImageBindingResourceClass(binding.kind) == IR::ImageResourceClass::None) {
+			continue;
+		}
+		const auto name = "image_" + std::to_string(static_cast<uint32_t>(binding.kind));
+		Decorate(state.image_variables[IR::ImageBindingIndex(binding.kind)], name.c_str(),
+		         binding.kind);
 	}
 	if (state.sampler_variable != 0) {
 		Decorate(state.sampler_variable, "samplers", IR::DescriptorBindingKind::Samplers);
@@ -651,9 +641,9 @@ void AddVsharpAnnotationsAndNames(EmitterState& state) {
 		state.builder.AddName(PushConstantBlockType(state), "BufferResource");
 		state.builder.AddName(state.push_constant_variable, "vsharp");
 	}
-	if (state.vsharp_storage_variable != 0) {
-		DecorateDescriptor(state, state.vsharp_storage_variable, "user_data",
-		                   IR::DescriptorBindingKind::UserData);
+	if (state.shader_data_storage_variable != 0) {
+		DecorateDescriptor(state, state.shader_data_storage_variable, "shader_data",
+		                   IR::DescriptorBindingKind::ShaderData);
 	}
 }
 
@@ -675,6 +665,10 @@ void DefineModule(EmitterState& state) {
 		state.builder.RequireCapability(CapabilityPhysicalStorageBufferAddresses);
 		state.builder.RequireExtension("SPV_KHR_physical_storage_buffer");
 	}
+	if (state.requirements.buffer_int64_atomics) {
+		state.builder.RequireCapability(CapabilityInt64);
+		state.builder.RequireCapability(CapabilityInt64Atomics);
+	}
 	if (state.clip_distance_variable != 0) {
 		state.builder.RequireCapability(CapabilityClipDistance);
 	}

src/graphics/shader/recompiler/backend/spirv/spirvEmitterProgram.cpp

@@ -1,8 +1,9 @@
 #include "graphics/shader/recompiler/backend/spirv/spirvEmitterInternal.h"
 
+#include "common/assert.h"
+
 #include <algorithm>
 #include <bit>
-#include <fmt/format.h>
 #include <functional>
 #include <optional>
 #include <unordered_set>
@@ -196,57 +197,50 @@ uint32_t EmitDispatcherNextPc(ValueEmitContext& ctx, const DispatcherFunctionSta
 	}
 }
 
-bool EmitDirectInstruction(ValueEmitContext& ctx, const IR::Inst& inst) {
+void EmitDirectInstruction(ValueEmitContext& ctx, const IR::Inst& inst) {
 	if (EmitValueFlow(ctx, inst) || EmitValueAlu(ctx, inst) || EmitValueMemory(ctx, inst) ||
 	    EmitValueImage(ctx, inst)) {
-		return true;
+		return;
 	}
 	ctx.Fail(inst, "has no direct SPIR-V emitter");
-	return false;
 }
 
-bool EmitStructuredInstruction(ValueEmitContext& ctx, StructuredFunctionState& structured,
+void EmitStructuredInstruction(ValueEmitContext& ctx, StructuredFunctionState& structured,
                                const IR::Inst& inst) {
 	if (inst.GetOpcode() == IR::ValueOpcode::Phi) {
 		const auto type = ctx.TypeId(inst.GetType());
 		if (type == 0 || inst.NumArgs() == 0) {
 			ctx.Fail(inst, "has no native SPIR-V representation");
-			return false;
 		}
 		for (size_t index = 0; index < inst.NumArgs(); index++) {
 			const auto* predecessor = inst.PhiBlock(index);
 			if (predecessor == nullptr || !ctx.labels.contains(predecessor)) {
 				ctx.Fail(inst, "has a predecessor outside the structured function");
-				return false;
 			}
 		}
 		structured.deferred_phis.push_back(
 		    {ctx.state.builder.AddDeferredPhi(type, ctx.Result(inst), inst.NumArgs()), &inst});
-		return true;
+		return;
 	}
-	return EmitDirectInstruction(ctx, inst);
+	EmitDirectInstruction(ctx, inst);
 }
 
-bool EmitDispatcherInstruction(ValueEmitContext& ctx, const DispatcherFunctionState& dispatcher,
+void EmitDispatcherInstruction(ValueEmitContext& ctx, const DispatcherFunctionState& dispatcher,
                                const IR::Inst& inst) {
 	if (inst.GetOpcode() == IR::ValueOpcode::Phi) {
 		const auto type = ctx.TypeId(inst.GetType());
 		if (type == 0) {
 			ctx.Fail(inst, "cannot be loaded by the dispatcher");
-			return false;
 		}
 		ctx.state.builder.AddFunction(
 		    {OpLoad, type, ctx.Result(inst), dispatcher.spills.at(&inst)});
-		return true;
-	}
-	if (!EmitDirectInstruction(ctx, inst)) {
-		return false;
+		return;
 	}
+	EmitDirectInstruction(ctx, inst);
 	if (const auto found = dispatcher.spills.find(&inst); found != dispatcher.spills.end()) {
 		ctx.state.builder.AddFunction(
 		    {OpStore, found->second, ctx.Def(IR::Value(const_cast<IR::Inst*>(&inst)))});
 	}
-	return true;
 }
 
 template <typename EmitInstruction>
@@ -258,14 +252,11 @@ void EmitBlock(ValueEmitContext& ctx, const IR::Block* block, EmitInstruction&&
 		if (inst.GetOpcode() == IR::ValueOpcode::Phi) {
 			if (emitted_non_phi) {
 				ctx.Fail(inst, "appears after a non-Phi instruction");
-				return;
 			}
 		} else {
 			emitted_non_phi = true;
 		}
-		if (!emit_instruction(inst)) {
-			return;
-		}
+		emit_instruction(inst);
 	}
 }
 
@@ -275,15 +266,10 @@ void PatchStructuredPhis(ValueEmitContext& ctx, StructuredFunctionState& structu
 			const auto* predecessor = deferred.instruction->PhiBlock(index);
 			const auto  found       = structured.block_exit_labels.find(predecessor);
 			if (found == structured.block_exit_labels.end()) {
-				ctx.failed = true;
-				ctx.error  = "typed Phi predecessor was not emitted";
-				return;
+				ctx.Fail(*deferred.instruction, "has a predecessor that was not emitted");
 			}
 			ctx.state.builder.PatchDeferredPhi(
 			    deferred.phi, index, ctx.Def(deferred.instruction->Arg(index)), found->second);
-			if (ctx.failed) {
-				return;
-			}
 		}
 	}
 }
@@ -291,20 +277,15 @@ void PatchStructuredPhis(ValueEmitContext& ctx, StructuredFunctionState& structu
 void EmitStructuredFunction(ValueEmitContext& ctx) {
 	StructuredFunctionState structured;
 	ctx.state.builder.AddFunction({OpBranch, ctx.Label(ctx.program.blocks.front())});
-	for (size_t index = 0; index < ctx.program.blocks.size() && !ctx.failed; index++) {
+	for (size_t index = 0; index < ctx.program.blocks.size(); index++) {
 		const auto* block = ctx.program.blocks[index];
 		EmitBlock(ctx, block, [&](const IR::Inst& inst) {
-			return EmitStructuredInstruction(ctx, structured, inst);
+			EmitStructuredInstruction(ctx, structured, inst);
 		});
-		if (ctx.failed) {
-			break;
-		}
 		structured.block_exit_labels.emplace(block, ctx.state.current_label);
 		EmitStructuredTerminator(ctx, block, ctx.program.block_info[index]);
 	}
-	if (!ctx.failed) {
-		PatchStructuredPhis(ctx, structured);
-	}
+	PatchStructuredPhis(ctx, structured);
 }
 
 void EmitDispatcherFunction(ValueEmitContext& ctx, const DispatcherFunctionState& dispatcher) {
@@ -312,11 +293,8 @@ void EmitDispatcherFunction(ValueEmitContext& ctx, const DispatcherFunctionState
 	const auto* entry = ctx.program.blocks.front();
 	state.builder.AddFunction({OpBranch, ctx.Label(entry)});
 	EmitBlock(ctx, entry, [&](const IR::Inst& inst) {
-		return EmitDispatcherInstruction(ctx, dispatcher, inst);
+		EmitDispatcherInstruction(ctx, dispatcher, inst);
 	});
-	if (ctx.failed) {
-		return;
-	}
 	const auto initial_pc =
 	    EmitDispatcherNextPc(ctx, dispatcher, entry, ctx.program.block_info.front());
 	const auto initial_parent = state.current_label;
@@ -347,22 +325,16 @@ void EmitDispatcherFunction(ValueEmitContext& ctx, const DispatcherFunctionState
 	std::vector<uint32_t> next_pc_words {OpPhi, TypeU32(state), next_pc,
 	                                     ConstantU32(state, UINT32_MAX), dispatcher.select_label};
 
-	for (size_t index = 1; index < ctx.program.blocks.size() && !ctx.failed; index++) {
+	for (size_t index = 1; index < ctx.program.blocks.size(); index++) {
 		EmitBlock(ctx, ctx.program.blocks[index], [&](const IR::Inst& inst) {
-			return EmitDispatcherInstruction(ctx, dispatcher, inst);
+			EmitDispatcherInstruction(ctx, dispatcher, inst);
 		});
-		if (ctx.failed) {
-			break;
-		}
 		const auto selected = EmitDispatcherNextPc(ctx, dispatcher, ctx.program.blocks[index],
 		                                           ctx.program.block_info[index]);
 		next_pc_words.push_back(selected);
 		next_pc_words.push_back(state.current_label);
 		state.builder.AddFunction({OpBranch, dispatcher.after_switch_label});
 	}
-	if (ctx.failed) {
-		return;
-	}
 	EmitLabel(state, dispatcher.after_switch_label);
 	state.builder.AddFunction(next_pc_words);
 	state.builder.AddFunction({OpBranch, dispatcher.continue_label});
@@ -408,9 +380,7 @@ uint32_t ValueEmitContext::Def(IR::Value value) {
 	}
 	const auto* inst = value.ResolveInstruction();
 	if (inst == nullptr) {
-		failed = true;
-		error  = "direct SPIR-V emitter received a non-value argument";
-		return ConstantU32(state, 0);
+		Fail("direct SPIR-V emitter received a non-value argument");
 	}
 	if (dispatcher_spills != nullptr && current_block != nullptr &&
 	    inst->Parent() != current_block) {
@@ -469,9 +439,7 @@ uint32_t ValueEmitContext::Define(const IR::Inst& inst, uint32_t value) {
 uint32_t ValueEmitContext::ResourceIndex(IR::Value value, IR::ValueOpcode opcode) {
 	const auto* inst = value.ResolveInstruction();
 	if (inst == nullptr || inst->GetOpcode() != opcode) {
-		failed = true;
-		error  = "typed resource handle has the wrong producer";
-		return 0;
+		Fail("typed resource handle has the wrong producer");
 	}
 	return inst->Flags<uint32_t>();
 }
@@ -479,9 +447,7 @@ uint32_t ValueEmitContext::ResourceIndex(IR::Value value, IR::ValueOpcode opcode
 const IR::Inst* ValueEmitContext::ImageAddress(IR::Value value) {
 	const auto* inst = value.ResolveInstruction();
 	if (inst == nullptr || inst->GetOpcode() != IR::ValueOpcode::MakeImageAddress) {
-		failed = true;
-		error  = "typed image address was not constructed by MakeImageAddress";
-		return nullptr;
+		Fail("typed image address was not constructed by MakeImageAddress");
 	}
 	return inst;
 }
@@ -498,12 +464,20 @@ uint32_t ValueEmitContext::Label(const IR::Block* block) const {
 	return labels.at(block);
 }
 
-void ValueEmitContext::Fail(const IR::Inst& inst, const char* reason) {
-	failed = true;
-	error  = fmt::format("typed opcode {} {}", IR::ValueOpcodeName(inst.GetOpcode()), reason);
+[[noreturn]] void ValueEmitContext::Fail(const char* reason) const {
+	EXIT("SPIR-V emission failed: hash=0x%016" PRIx64 " stage=%u reason=%s\n",
+	     program.shader_hash, static_cast<unsigned>(program.stage), reason);
+	std::abort();
 }
 
-bool EmitProgram(EmitterState& state, const IR::Program& program, std::string* error) {
+[[noreturn]] void ValueEmitContext::Fail(const IR::Inst& inst, const char* reason) const {
+	EXIT("SPIR-V emission failed: hash=0x%016" PRIx64 " stage=%u opcode=%s reason=%s\n",
+	     program.shader_hash, static_cast<unsigned>(program.stage),
+	     IR::ValueOpcodeName(inst.GetOpcode()), reason);
+	std::abort();
+}
+
+void EmitProgram(EmitterState& state, const IR::Program& program) {
 	ValueEmitContext                       ctx(state, program);
 	std::optional<DispatcherFunctionState> dispatcher;
 	if (state.stage == ShaderType::Pixel && state.requirements.pixel_valid_mask) {
@@ -562,10 +536,6 @@ bool EmitProgram(EmitterState& state, const IR::Program& program, std::string* e
 		dispatch.merge_label        = state.builder.AllocateId();
 		ctx.dispatcher_spills       = &dispatch.spills;
 	}
-	if (ctx.failed) {
-		SetError(error, ctx.error.c_str());
-		return false;
-	}
 	DefineGetBdaPointer(state);
 	for (const auto* block: program.blocks) {
 		if (std::ranges::any_of(*block, [](const IR::Inst& inst) {
@@ -630,11 +600,6 @@ bool EmitProgram(EmitterState& state, const IR::Program& program, std::string* e
 		EmitStructuredFunction(ctx);
 	}
 	state.builder.AddFunction({OpFunctionEnd});
-	if (ctx.failed) {
-		SetError(error, ctx.error.c_str());
-		return false;
-	}
-	return true;
 }
 
 } // namespace Libs::Graphics::ShaderRecompiler::Spirv::Emitter

src/graphics/shader/recompiler/frontend/cfg/ShaderCFG.cpp

@@ -1,5 +1,7 @@
 #include "graphics/shader/recompiler/frontend/cfg/ShaderCFG.h"
 
+#include "common/assert.h"
+
 #include <algorithm>
 #include <fmt/format.h>
 #include <iterator>
@@ -24,19 +26,18 @@ struct SetpcTargetInfo {
 	std::vector<uint32_t> selector_target_pcs;
 };
 
-void SetError(std::string* error, const std::string& message) {
-	if (error != nullptr) {
-		*error = message;
-	}
-}
-
-void SetFailure(Graph& graph, FailureKind kind, uint32_t block_id, const std::string& message,
-                std::string* error) {
+void SetFailure(Graph& graph, FailureKind kind, uint32_t block_id, const std::string& message) {
 	graph.unsupported        = true;
 	graph.failure_kind       = kind;
 	graph.failure_block      = block_id;
 	graph.unsupported_reason = message;
-	SetError(error, message);
+}
+
+[[noreturn]] void ExitBuildFailure(Graph& graph, FailureKind kind, uint32_t block_id,
+                                   const std::string& message) {
+	SetFailure(graph, kind, block_id, message);
+	EXIT("shader CFG build failed: %s", message.c_str());
+	std::abort();
 }
 
 uint32_t InstructionEndPc(const Instruction& inst) {
@@ -1267,7 +1268,7 @@ bool MergeLeavesContainingLoop(const Graph& graph, uint32_t header, uint32_t mer
 	return false;
 }
 
-bool CanonicalizeNaturalLoops(Graph& graph, std::string* error) {
+bool CanonicalizeNaturalLoops(Graph& graph) {
 	const auto rewrite_budget = graph.blocks.size() * 2u + 16u;
 	for (size_t rewrite = 0; rewrite < rewrite_budget; rewrite++) {
 		bool changed = false;
@@ -1322,19 +1323,18 @@ bool CanonicalizeNaturalLoops(Graph& graph, std::string* error) {
 	}
 
 	SetFailure(graph, FailureKind::StructuredControlFlow, graph.entry_block,
-	           "CFG loop canonicalization exceeded rewrite budget", error);
+	           "CFG loop canonicalization exceeded rewrite budget");
 	return false;
 }
 
-bool IsolateSemanticLoopHeaders(Graph& graph, std::string* error) {
+bool IsolateSemanticLoopHeaders(Graph& graph) {
 	const auto isolation_budget = graph.natural_loops.size() + 1u;
 	for (size_t isolation = 0; isolation < isolation_budget; isolation++) {
-		const auto loop = std::find_if(graph.natural_loops.begin(), graph.natural_loops.end(),
-		                               [&](const auto& value) {
-			                               const auto* header = graph.FindBlock(value.header);
-			                               return header != nullptr &&
-			                                      header->inst_begin != header->inst_end;
-		                               });
+		const auto loop = std::find_if(
+		    graph.natural_loops.begin(), graph.natural_loops.end(), [&](const auto& value) {
+			    const auto* header = graph.FindBlock(value.header);
+			    return header != nullptr && header->inst_begin != header->inst_end;
+		    });
 		if (loop == graph.natural_loops.end()) {
 			return true;
 		}
@@ -1346,14 +1346,13 @@ bool IsolateSemanticLoopHeaders(Graph& graph, std::string* error) {
 		// displacing OpLoopMerge into a helper-created block.
 		if (!IsolateSemanticLoopHeader(graph, loop->header)) {
 			SetFailure(graph, FailureKind::StructuredControlFlow, loop->header,
-			           fmt::format("failed to isolate semantic loop header {}", loop->header),
-			           error);
+			           fmt::format("failed to isolate semantic loop header {}", loop->header));
 			return false;
 		}
 	}
 
 	SetFailure(graph, FailureKind::StructuredControlFlow, graph.entry_block,
-	           "CFG semantic loop-header isolation exceeded rewrite budget", error);
+	           "CFG semantic loop-header isolation exceeded rewrite budget");
 	return false;
 }
 
@@ -1453,7 +1452,7 @@ std::vector<uint32_t> SelectionRegion(const Graph& graph, const BasicBlock& head
 	return region;
 }
 
-bool SplitOneSelectionMerge(Graph& graph, std::string* error) {
+bool SplitOneSelectionMerge(Graph& graph) {
 	std::vector<uint32_t> loop_headers;
 	loop_headers.reserve(graph.natural_loops.size());
 	for (const auto& loop: graph.natural_loops) {
@@ -1501,8 +1500,7 @@ bool SplitOneSelectionMerge(Graph& graph, std::string* error) {
 			    graph, FailureKind::StructuredControlFlow, block_id,
 			    fmt::format("selection header block {} has externally entered region block {}; "
 			                "semantic block cloning is disabled",
-			                block_id, *external),
-			    error);
+			                block_id, *external));
 			return false;
 		}
 		const auto construct_blocks = DominatedBlocks(graph, block_id, merge);
@@ -1514,11 +1512,11 @@ bool SplitOneSelectionMerge(Graph& graph, std::string* error) {
 	return false;
 }
 
-bool SplitSharedMergeBlocks(Graph& graph, std::string* error) {
+bool SplitSharedMergeBlocks(Graph& graph) {
 	const auto original_block_count = static_cast<uint32_t>(graph.blocks.size());
 	const auto split_budget         = std::max<uint32_t>(16u, original_block_count * 4u);
 	for (uint32_t splits = 0; splits < split_budget; splits++) {
-		if (!SplitOneLoopMerge(graph) && !SplitOneSelectionMerge(graph, error)) {
+		if (!SplitOneLoopMerge(graph) && !SplitOneSelectionMerge(graph)) {
 			return !graph.unsupported;
 		}
 		RebuildPredecessors(graph);
@@ -1528,8 +1526,7 @@ bool SplitSharedMergeBlocks(Graph& graph, std::string* error) {
 	           fmt::format("CFG shared merge splitting exceeded budget: original_blocks={} "
 	                       "current_blocks={} split_budget={}",
 	                       original_block_count, static_cast<uint64_t>(graph.blocks.size()),
-	                       split_budget),
-	           error);
+	                       split_budget));
 	return false;
 }
 
@@ -1890,12 +1887,11 @@ bool RouteSharedSelectionArm(Graph& graph, uint32_t route_variable) {
 
 } // namespace
 
-bool BuildGraph(const Decoder::Program& program, Graph& graph, std::string* error) {
-	graph = {};
-
+Graph BuildGraph(const Decoder::Program& program) {
+	Graph graph;
 	if (program.instructions.empty()) {
-		SetError(error, "cannot build CFG for empty shader");
-		return false;
+		ExitBuildFailure(graph, FailureKind::InvalidLabel, UINT32_MAX,
+		                 "cannot build CFG for empty shader");
 	}
 
 	const auto first_pc = program.instructions.front().pc;
@@ -1905,11 +1901,10 @@ bool BuildGraph(const Decoder::Program& program, Graph& graph, std::string* erro
 	for (const auto& inst: program.instructions) {
 		instruction_pcs.insert(inst.pc);
 		if (inst.opcode == Opcode::UNSUPPORTED) {
-			SetFailure(graph, FailureKind::UnsupportedInstruction, UINT32_MAX,
-			           fmt::format("unsupported decoded instruction in CFG at pc 0x{:08x}: {}",
-			                       inst.pc, Decoder::InstructionToString(inst).c_str()),
-			           error);
-			return false;
+			ExitBuildFailure(
+			    graph, FailureKind::UnsupportedInstruction, UINT32_MAX,
+			    fmt::format("unsupported decoded instruction in CFG at pc 0x{:08x}: {}", inst.pc,
+			                Decoder::InstructionToString(inst).c_str()));
 		}
 	}
 
@@ -1923,11 +1918,9 @@ bool BuildGraph(const Decoder::Program& program, Graph& graph, std::string* erro
 		const auto  next_pc = InstructionEndPc(inst);
 		if (IsBranch(inst.opcode)) {
 			if (!IsValidTarget(inst.branch_target, instruction_pcs, first_pc, end_pc)) {
-				SetFailure(graph, FailureKind::InvalidBranchTarget, UINT32_MAX,
-				           fmt::format("branch at pc 0x{:08x} targets invalid pc 0x{:08x}", inst.pc,
-				                       inst.branch_target),
-				           error);
-				return false;
+				ExitBuildFailure(graph, FailureKind::InvalidBranchTarget, UINT32_MAX,
+				                 fmt::format("branch at pc 0x{:08x} targets invalid pc 0x{:08x}",
+				                             inst.pc, inst.branch_target));
 			}
 			labels.insert(inst.branch_target);
 			if (next_pc <= end_pc) {
@@ -1936,21 +1929,19 @@ bool BuildGraph(const Decoder::Program& program, Graph& graph, std::string* erro
 		} else if (inst.opcode == Opcode::S_SETPC_B64) {
 			SetpcTargetInfo target_info;
 			if (!ResolveSetpcTargets(program, i, target_info)) {
-				SetFailure(graph, FailureKind::InvalidBranchTarget, UINT32_MAX,
-				           fmt::format("unsupported dynamic S_SETPC_B64 at pc 0x{:08x}", inst.pc),
-				           error);
-				return false;
+				ExitBuildFailure(
+				    graph, FailureKind::InvalidBranchTarget, UINT32_MAX,
+				    fmt::format("unsupported dynamic S_SETPC_B64 at pc 0x{:08x}", inst.pc));
 			}
 			const auto& target_pcs = target_info.indirect
 			                             ? target_info.target_pcs
 			                             : std::vector<uint32_t> {target_info.target};
 			for (const auto target: target_pcs) {
 				if (!IsValidTarget(target, instruction_pcs, first_pc, end_pc)) {
-					SetFailure(graph, FailureKind::InvalidBranchTarget, UINT32_MAX,
-					           fmt::format("S_SETPC_B64 at pc 0x{:08x} targets invalid pc 0x{:08x}",
-					                       inst.pc, target),
-					           error);
-					return false;
+					ExitBuildFailure(
+					    graph, FailureKind::InvalidBranchTarget, UINT32_MAX,
+					    fmt::format("S_SETPC_B64 at pc 0x{:08x} targets invalid pc 0x{:08x}",
+					                inst.pc, target));
 				}
 				labels.insert(target);
 			}
@@ -1971,10 +1962,9 @@ bool BuildGraph(const Decoder::Program& program, Graph& graph, std::string* erro
 			continue;
 		}
 		if (start != end_pc && !instruction_pcs.contains(start)) {
-			SetFailure(graph, FailureKind::InvalidLabel, UINT32_MAX,
-			           fmt::format("CFG label does not start on an instruction: 0x{:08x}", start),
-			           error);
-			return false;
+			ExitBuildFailure(
+			    graph, FailureKind::InvalidLabel, UINT32_MAX,
+			    fmt::format("CFG label does not start on an instruction: 0x{:08x}", start));
 		}
 
 		BasicBlock block;
@@ -2046,11 +2036,10 @@ bool BuildGraph(const Decoder::Program& program, Graph& graph, std::string* erro
 			block.terminator.true_block = pc_to_block.at(last.branch_target);
 			const auto fallthrough      = pc_to_block.find(next_pc);
 			if (fallthrough == pc_to_block.end()) {
-				SetFailure(graph, FailureKind::MissingFallthrough, block.id,
-				           fmt::format("conditional branch at pc 0x{:08x} has no fallthrough block",
-				                       last.pc),
-				           error);
-				return false;
+				ExitBuildFailure(
+				    graph, FailureKind::MissingFallthrough, block.id,
+				    fmt::format("conditional branch at pc 0x{:08x} has no fallthrough block",
+				                last.pc));
 			}
 			block.terminator.false_block = fallthrough->second;
 		} else {
@@ -2131,27 +2120,26 @@ bool BuildGraph(const Decoder::Program& program, Graph& graph, std::string* erro
 		graph.unsupported = false;
 	}
 
-	return true;
+	return graph;
 }
 
 namespace {
 
-bool StructurizeImpl(Graph& graph, std::string* error) {
+bool StructurizeImpl(Graph& graph) {
 	if (graph.unsupported || graph.irreducible) {
 		if (graph.unsupported_reason.empty()) {
 			graph.unsupported_reason = "unsupported CFG";
 		}
-		SetError(error, graph.unsupported_reason);
 		return false;
 	}
 
-	if (!CanonicalizeNaturalLoops(graph, error)) {
+	if (!CanonicalizeNaturalLoops(graph)) {
 		return false;
 	}
-	if (!SplitSharedMergeBlocks(graph, error)) {
+	if (!SplitSharedMergeBlocks(graph)) {
 		return false;
 	}
-	if (!IsolateSemanticLoopHeaders(graph, error)) {
+	if (!IsolateSemanticLoopHeaders(graph)) {
 		return false;
 	}
 	ClearStructuredTerminators(graph);
@@ -2164,8 +2152,7 @@ bool StructurizeImpl(Graph& graph, std::string* error) {
 			    graph, FailureKind::StructuredControlFlow, header,
 			    fmt::format(
 			        "duplicate structured merge block {} for header {} (already used by header {})",
-			        merge, header, it->second),
-			    error);
+			        merge, header, it->second));
 			return false;
 		}
 		return true;
@@ -2176,16 +2163,14 @@ bool StructurizeImpl(Graph& graph, std::string* error) {
 		if (header == nullptr || loop.merge == UINT32_MAX || loop.continue_block == UINT32_MAX) {
 			SetFailure(
 			    graph, FailureKind::StructuredControlFlow, loop.header,
-			    fmt::format("loop at block {} has no structured merge/continue", loop.header),
-			    error);
+			    fmt::format("loop at block {} has no structured merge/continue", loop.header));
 			return false;
 		}
 		if (header->inst_begin != header->inst_end ||
 		    header->terminator.kind != TerminatorKind::Branch) {
-			SetFailure(graph, FailureKind::StructuredControlFlow, loop.header,
-			           fmt::format("loop header {} is not a dedicated empty control block",
-			                       loop.header),
-			           error);
+			SetFailure(
+			    graph, FailureKind::StructuredControlFlow, loop.header,
+			    fmt::format("loop header {} is not a dedicated empty control block", loop.header));
 			return false;
 		}
 		if (!reserve_merge_block(loop.header, loop.merge)) {
@@ -2208,8 +2193,7 @@ bool StructurizeImpl(Graph& graph, std::string* error) {
 		const auto merge = FindSelectionMerge(graph, block);
 		if (merge == UINT32_MAX) {
 			SetFailure(graph, FailureKind::StructuredControlFlow, block.id,
-			           fmt::format("conditional block {} has no structured merge", block.id),
-			           error);
+			           fmt::format("conditional block {} has no structured merge", block.id));
 			return false;
 		}
 
@@ -2224,16 +2208,15 @@ bool StructurizeImpl(Graph& graph, std::string* error) {
 
 } // namespace
 
-bool Structurize(Graph& graph, std::string* error) {
+bool Structurize(Graph& graph) {
 	Graph original = graph;
-	if (StructurizeImpl(graph, error)) {
+	if (StructurizeImpl(graph)) {
 		return true;
 	}
 
-	Graph       failed_graph = std::move(graph);
-	std::string failed_error = error != nullptr ? *error : std::string {};
-	graph                    = std::move(original);
-	const auto route_budget  = static_cast<uint32_t>(graph.blocks.size());
+	Graph failed_graph      = std::move(graph);
+	graph                   = std::move(original);
+	const auto route_budget = static_cast<uint32_t>(graph.blocks.size());
 	// Apply one route at a time and retry. Eagerly routing every matching diamond can
 	// rewrite unrelated selections that were already structurally valid.
 	for (uint32_t route_variable = 0; route_variable < route_budget; route_variable++) {
@@ -2241,18 +2224,12 @@ bool Structurize(Graph& graph, std::string* error) {
 			break;
 		}
 		Graph routed = graph;
-		if (error != nullptr) {
-			error->clear();
-		}
-		if (StructurizeImpl(routed, error)) {
+		if (StructurizeImpl(routed)) {
 			graph = std::move(routed);
 			return true;
 		}
 	}
 	graph = std::move(failed_graph);
-	if (error != nullptr) {
-		*error = std::move(failed_error);
-	}
 	return false;
 }
 

src/graphics/shader/recompiler/frontend/cfg/ShaderCFG.h

@@ -108,8 +108,8 @@ struct Graph {
 	uint32_t          FindNearestCommonPostDominator(uint32_t block_a, uint32_t block_b) const;
 };
 
-bool        BuildGraph(const Decoder::Program& program, Graph& graph, std::string* error);
-bool        Structurize(Graph& graph, std::string* error);
+Graph       BuildGraph(const Decoder::Program& program);
+bool        Structurize(Graph& graph);
 std::string BranchConditionToString(BranchCondition condition);
 std::string FailureKindToString(FailureKind kind);
 std::string GraphToString(const Graph& graph);

src/graphics/shader/recompiler/frontend/decode/ImageOps.cpp

@@ -202,6 +202,7 @@ constexpr MimgGatherInfo MIMG_GATHER_OPCODE_LIST[] = {
 };
 
 constexpr MimgAtomicInfo MIMG_ATOMIC_OPCODE_LIST[] = {
+    {0x0fu, "image_atomic_swap", Opcode::IMAGE_ATOMIC_SWAP},
     {0x11u, "image_atomic_add", Opcode::IMAGE_ATOMIC_ADD},
     {0x15u, "image_atomic_umin", Opcode::IMAGE_ATOMIC_UMIN},
     {0x17u, "image_atomic_umax", Opcode::IMAGE_ATOMIC_UMAX},
@@ -357,8 +358,8 @@ void DecodeMimg(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 
 	DecodeVectorGpr(vdata, inst.dst);
 	DecodeVectorGpr(vaddr, inst.src0);
-	DecodeScalarSource(srsrc * 4u, pc, inst.src1, nullptr);
-	DecodeScalarSource(ssamp * 4u, pc, inst.src2, nullptr);
+	DecodeScalarSource(srsrc * 4u, pc, inst.src1);
+	DecodeScalarSource(ssamp * 4u, pc, inst.src2);
 	inst.src_count = 3;
 }
 

src/graphics/shader/recompiler/frontend/decode/MemoryOps.cpp

@@ -47,6 +47,7 @@ constexpr MemoryOpcodeInfo MUBUF_OPCODE_LIST[] = {
     {0x1eu, Opcode::BUFFER_STORE_DWORDX4, 4, 32},
     {0x1fu, Opcode::BUFFER_STORE_DWORDX3, 3, 32},
     {0x30u, Opcode::BUFFER_ATOMIC_SWAP, 1, 32},
+    {0x31u, Opcode::BUFFER_ATOMIC_CMPSWAP, 1, 32},
     {0x32u, Opcode::BUFFER_ATOMIC_ADD, 1, 32},
     {0x33u, Opcode::BUFFER_ATOMIC_SUB, 1, 32},
     {0x35u, Opcode::BUFFER_ATOMIC_SMIN, 1, 32},
@@ -58,6 +59,8 @@ constexpr MemoryOpcodeInfo MUBUF_OPCODE_LIST[] = {
     {0x3bu, Opcode::BUFFER_ATOMIC_XOR, 1, 32},
     {0x3fu, Opcode::BUFFER_ATOMIC_FMIN, 1, 32},
     {0x40u, Opcode::BUFFER_ATOMIC_FMAX, 1, 32},
+    {0x50u, Opcode::BUFFER_ATOMIC_SWAP_X2, 2, 32},
+    {0x5au, Opcode::BUFFER_ATOMIC_OR_X2, 2, 32},
 };
 
 constexpr MemoryOpcodeInfo MTBUF_OPCODE_LIST[] = {
@@ -103,8 +106,11 @@ constexpr MemoryOpcodeInfo DS_OPCODE_LIST[] = {
     {0x4du, Opcode::DS_WRITE_B64, 2, 32},        {0x4eu, Opcode::DS_WRITE2_B64, 4, 32},
     {0x4fu, Opcode::DS_WRITE2ST64_B64, 4, 32},   {0x76u, Opcode::DS_READ_B64, 2, 32},
     {0x77u, Opcode::DS_READ2_B64, 4, 32},        {0x78u, Opcode::DS_READ2ST64_B64, 4, 32},
+    {0xa1u, Opcode::DS_WRITE_B16_D16_HI, 1, 16},
     {0xa6u, Opcode::DS_READ_U16_D16, 1, 16},
+    {0xa7u, Opcode::DS_READ_U16_D16_HI, 1, 16},
     {0xb0u, Opcode::DS_WRITE_ADDTID_B32, 1, 32}, {0xb1u, Opcode::DS_READ_ADDTID_B32, 1, 32},
+    {0xb3u, Opcode::DS_BPERMUTE_B32, 1, 32},
     {0xdeu, Opcode::DS_WRITE_B96, 3, 32},        {0xdfu, Opcode::DS_WRITE_B128, 4, 32},
     {0xfeu, Opcode::DS_READ_B96, 3, 32},         {0xffu, Opcode::DS_READ_B128, 4, 32},
 };
@@ -147,6 +153,7 @@ bool IsDsWriteOpcode(Opcode opcode) {
 	switch (opcode) {
 		case Opcode::DS_WRITE_B8:
 		case Opcode::DS_WRITE_B16:
+		case Opcode::DS_WRITE_B16_D16_HI:
 		case Opcode::DS_WRITE2_B32:
 		case Opcode::DS_WRITE2ST64_B32:
 		case Opcode::DS_WRITE2_B64:
@@ -192,6 +199,7 @@ uint32_t DsSourceCount(Opcode opcode) {
 		case Opcode::DS_WRITE2ST64_B64:
 		case Opcode::DS_MIN_F32:
 		case Opcode::DS_MAX_F32: return 3u;
+		case Opcode::DS_BPERMUTE_B32: return 2u;
 		case Opcode::DS_READ_ADDTID_B32:
 		case Opcode::DS_CONSUME:
 		case Opcode::DS_APPEND: return 0u;
@@ -236,11 +244,11 @@ void DecodeSmem(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 		MarkMemoryUnsupported(inst, Family::SMEM, opcode, "SMEM opcode is not implemented");
 	}
 
-	DecodeScalarDestination(sdst, pc, inst.dst, nullptr);
+	DecodeScalarDestination(sdst, pc, inst.dst);
 	// SMEM encodes SBASE in SGPR pairs. Scalar-buffer loads still use the same
 	// pair index; their descriptor operand consumes four SGPRs from that base.
-	DecodeScalarSource(sbase * 2u, pc, inst.src0, nullptr);
-	DecodeScalarSource(soffset, pc, inst.src1, nullptr);
+	DecodeScalarSource(sbase * 2u, pc, inst.src0);
+	DecodeScalarSource(soffset, pc, inst.src1);
 	inst.src_count = 2;
 }
 
@@ -273,8 +281,8 @@ void DecodeMubuf(uint32_t pc, std::span<const uint32_t> code, uint32_t word_inde
 
 	DecodeVectorGpr(vdata, inst.dst);
 	DecodeVectorGpr(vaddr, inst.src0);
-	DecodeScalarSource(srsrc * 4u, pc, inst.src1, nullptr);
-	DecodeScalarSource(soffset, pc, inst.src2, nullptr);
+	DecodeScalarSource(srsrc * 4u, pc, inst.src1);
+	DecodeScalarSource(soffset, pc, inst.src2);
 	inst.src_count = 3;
 }
 
@@ -311,8 +319,8 @@ void DecodeMtbuf(uint32_t pc, std::span<const uint32_t> code, uint32_t word_inde
 
 	DecodeVectorGpr(vdata, inst.dst);
 	DecodeVectorGpr(vaddr, inst.src0);
-	DecodeScalarSource(srsrc * 4u, pc, inst.src1, nullptr);
-	DecodeScalarSource(soffset, pc, inst.src2, nullptr);
+	DecodeScalarSource(srsrc * 4u, pc, inst.src1);
+	DecodeScalarSource(soffset, pc, inst.src2);
 	inst.src_count = 3;
 }
 
@@ -359,7 +367,7 @@ void DecodeFlat(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 		DecodeVectorGpr(addr + 1u, inst.src1);
 		inst.src_count = 2;
 	} else {
-		DecodeScalarSource(saddr, pc, inst.src1, nullptr);
+		DecodeScalarSource(saddr, pc, inst.src1);
 		inst.src_count = 2;
 	}
 }
@@ -392,9 +400,10 @@ void DecodeDs(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
 		SetUnsupported(inst, Family::DS, opcode, "DS swizzle FFT mode is not implemented");
 	}
 	if (inst.gds &&
-	    (inst.opcode == Opcode::DS_SWIZZLE_B32 || inst.opcode == Opcode::DS_WRITE_ADDTID_B32 ||
+	    (inst.opcode == Opcode::DS_SWIZZLE_B32 || inst.opcode == Opcode::DS_BPERMUTE_B32 ||
+	     inst.opcode == Opcode::DS_WRITE_ADDTID_B32 ||
 	     inst.opcode == Opcode::DS_READ_ADDTID_B32)) {
-		SetUnsupported(inst, Family::DS, opcode, "DS swizzle/addtid is available only for LDS");
+		SetUnsupported(inst, Family::DS, opcode, "DS lane operation is available only for LDS");
 	}
 	if (inst.opcode == Opcode::DS_WRITE_ADDTID_B32 && data1 != 0u) {
 		SetUnsupported(inst, Family::DS, opcode,
@@ -421,13 +430,17 @@ void DecodeDs(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
 	}
 
 	DecodeVectorGpr(vdst, inst.dst);
-	if (inst.opcode == Opcode::DS_READ_U16_D16) {
+	if (inst.opcode == Opcode::DS_READ_U16_D16 ||
+	    inst.opcode == Opcode::DS_READ_U16_D16_HI) {
 		// Native D16 reads update only the selected destination half. Reuse the partial
-		// destination representation so typed IR translation preserves the upper word.
-		inst.dst.sdwa_sel = 4u;
+		// destination representation so typed IR translation preserves the other half.
+		inst.dst.sdwa_sel = inst.opcode == Opcode::DS_READ_U16_D16 ? 4u : 5u;
 	}
 	DecodeVectorGpr(addr, inst.src0);
 	DecodeVectorGpr(data0, inst.src1);
+	if (inst.opcode == Opcode::DS_WRITE_B16_D16_HI) {
+		inst.src1.sdwa_sel = 5u;
+	}
 	DecodeVectorGpr(data1, inst.src2);
 	inst.src_count = DsSourceCount(inst.opcode);
 }

src/graphics/shader/recompiler/frontend/decode/ScalarAluOps.cpp

@@ -27,7 +27,8 @@ constexpr OpcodeMap SOP2_OPCODE_LIST[] = {
     {0x22u, Opcode::S_ASHR_I32},        {0x24u, Opcode::S_BFM_B32},
     {0x25u, Opcode::S_BFM_B64},         {0x26u, Opcode::S_MUL_I32},
     {0x27u, Opcode::S_BFE_U32},         {0x28u, Opcode::S_BFE_I32},
-    {0x29u, Opcode::S_BFE_U64},         {0x2eu, Opcode::S_LSHL1_ADD_U32},
+    {0x29u, Opcode::S_BFE_U64},         {0x2cu, Opcode::S_ABSDIFF_I32},
+    {0x2eu, Opcode::S_LSHL1_ADD_U32},
     {0x2fu, Opcode::S_LSHL2_ADD_U32},   {0x30u, Opcode::S_LSHL3_ADD_U32},
     {0x31u, Opcode::S_LSHL4_ADD_U32},   {0x32u, Opcode::S_PACK_LL_B32_B16},
     {0x33u, Opcode::S_PACK_LH_B32_B16}, {0x34u, Opcode::S_PACK_HH_B32_B16},
@@ -45,9 +46,12 @@ constexpr OpcodeMap SOP1_OPCODE_LIST[] = {
     {0x10u, Opcode::S_BCNT1_I32_B64},
     {0x13u, Opcode::S_FF1_I32_B32},
     {0x14u, Opcode::S_FF1_I32_B64},
+    {0x15u, Opcode::S_FLBIT_I32_B32},
     {0x16u, Opcode::S_FLBIT_I32_B64},
     {0x1bu, Opcode::S_BITSET0_B32},
+    {0x1cu, Opcode::S_BITSET0_B64},
     {0x1du, Opcode::S_BITSET1_B32},
+    {0x1eu, Opcode::S_BITSET1_B64},
     {0x1fu, Opcode::S_GETPC_B64},
     {0x20u, Opcode::S_SETPC_B64},
     {0x24u, Opcode::S_AND_SAVEEXEC_B64},
@@ -92,6 +96,7 @@ constexpr OpcodeMap SOPP_OPCODE_LIST[] = {
     {0x0au, Opcode::S_BARRIER},
     {0x0cu, Opcode::S_WAITCNT},
     {0x0eu, Opcode::S_SLEEP},
+    {0x0fu, Opcode::S_SETPRIO},
     {0x10u, Opcode::S_SENDMSG},
     {0x12u, Opcode::S_TRAP},
     {0x16u, Opcode::S_TTRACEDATA},
@@ -105,21 +110,18 @@ constexpr auto SOPK_OPS = Detail::MakeOpcodeTable<0x20>(SOPK_OPCODE_LIST);
 constexpr auto SOPC_OPS = Detail::MakeOpcodeTable<0x80>(SOPC_OPCODE_LIST);
 constexpr auto SOPP_OPS = Detail::MakeOpcodeTable<0x80>(SOPP_OPCODE_LIST);
 
-bool DecodeBinarySources(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
-                         Instruction& inst, uint32_t ssrc0, uint32_t ssrc1, std::string* error) {
-	if (!DecodeScalarSource(ssrc0, pc, inst.src0, error) ||
-	    !DecodeScalarSource(ssrc1, pc, inst.src1, error)) {
-		return false;
-	}
+void DecodeBinarySources(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                         Instruction& inst, uint32_t ssrc0, uint32_t ssrc1) {
+	DecodeScalarSource(ssrc0, pc, inst.src0);
+	DecodeScalarSource(ssrc1, pc, inst.src1);
 	inst.src_count = 2;
 	ReadLiteralOperands(code, word_index, inst);
-	return true;
 }
 
 } // namespace
 
-bool DecodeSop1(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error) {
+void DecodeSop1(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst) {
 	const uint32_t word   = code[word_index];
 	const uint32_t opcode = (word >> 8u) & 0xffu;
 	const uint32_t ssrc0  = word & 0xffu;
@@ -134,35 +136,31 @@ bool DecodeSop1(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 
 	if (inst.opcode == Opcode::UNSUPPORTED) {
 		SetUnsupported(inst, Family::SOP1, opcode, "SOP1 opcode is not implemented");
-		return true;
+		return;
 	}
 
 	switch (inst.opcode) {
 		case Opcode::S_GETPC_B64:
 			inst.src_count = 0;
-			return DecodeScalarDestination(sdst, pc, inst.dst, error);
+			DecodeScalarDestination(sdst, pc, inst.dst);
+			return;
 		case Opcode::S_SETPC_B64:
 			inst.src_count = 1;
 			inst.dst.kind  = OperandKind::Null;
-			if (!DecodeScalarSource(ssrc0, pc, inst.src0, error)) {
-				return false;
-			}
+			DecodeScalarSource(ssrc0, pc, inst.src0);
 			ReadLiteralOperands(code, word_index, inst);
-			return true;
+			return;
 		default: break;
 	}
 
-	if (!DecodeScalarSource(ssrc0, pc, inst.src0, error) ||
-	    !DecodeScalarDestination(sdst, pc, inst.dst, error)) {
-		return false;
-	}
+	DecodeScalarSource(ssrc0, pc, inst.src0);
+	DecodeScalarDestination(sdst, pc, inst.dst);
 	inst.src_count = 1;
 	ReadLiteralOperands(code, word_index, inst);
-	return true;
 }
 
-bool DecodeSop2(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error) {
+void DecodeSop2(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst) {
 	const uint32_t word   = code[word_index];
 	const uint32_t opcode = (word >> 23u) & 0x7fu;
 	const uint32_t ssrc1  = (word >> 8u) & 0xffu;
@@ -178,17 +176,15 @@ bool DecodeSop2(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 
 	if (inst.opcode == Opcode::UNSUPPORTED) {
 		SetUnsupported(inst, Family::SOP2, opcode, "SOP2 opcode is not implemented");
-		return true;
+		return;
 	}
 
-	if (!DecodeScalarDestination(sdst, pc, inst.dst, error)) {
-		return false;
-	}
-	return DecodeBinarySources(pc, code, word_index, inst, ssrc0, ssrc1, error);
+	DecodeScalarDestination(sdst, pc, inst.dst);
+	DecodeBinarySources(pc, code, word_index, inst, ssrc0, ssrc1);
 }
 
-bool DecodeSopk(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error) {
+void DecodeSopk(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst) {
 	const uint32_t word   = code[word_index];
 	const uint32_t opcode = (word >> 23u) & 0x1fu;
 	const uint32_t sdst   = (word >> 16u) & 0x7fu;
@@ -207,18 +203,18 @@ bool DecodeSopk(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 
 	if (inst.opcode == Opcode::UNSUPPORTED) {
 		SetUnsupported(inst, Family::SOPK, opcode, "SOPK opcode is not implemented");
-		return true;
+		return;
 	}
 
 	switch (inst.opcode) {
-		case Opcode::S_MOVK_I32: return DecodeScalarDestination(sdst, pc, inst.dst, error);
+		case Opcode::S_MOVK_I32: DecodeScalarDestination(sdst, pc, inst.dst); return;
 		case Opcode::S_WAITCNT: {
 			const uint32_t waitcnt = word & 0xffffu;
 			inst.dst.kind          = OperandKind::Null;
 			inst.src0.signed_val   = static_cast<int32_t>(waitcnt);
 			inst.src0.value        = waitcnt;
 			inst.src_count         = 1;
-			return true;
+			return;
 		}
 		case Opcode::S_SETREG_B32:
 			inst.dst.kind        = OperandKind::Null;
@@ -226,26 +222,25 @@ bool DecodeSopk(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 			inst.src1.value      = word & 0xffffu;
 			inst.src1.signed_val = static_cast<int32_t>(imm);
 			inst.src_count       = 2;
-			return DecodeScalarSource(sdst, pc, inst.src0, error);
+			DecodeScalarSource(sdst, pc, inst.src0);
+			return;
 		default: break;
 	}
 
 	inst.src1 = inst.src0;
-	if (!DecodeScalarSource(sdst, pc, inst.src0, error)) {
-		return false;
-	}
+	DecodeScalarSource(sdst, pc, inst.src0);
 	if (inst.opcode == Opcode::S_ADD_I32 || inst.opcode == Opcode::S_MULK_I32) {
 		inst.src_count = 2;
-		return DecodeScalarDestination(sdst, pc, inst.dst, error);
+		DecodeScalarDestination(sdst, pc, inst.dst);
+		return;
 	}
 
 	inst.dst.kind  = OperandKind::Scc;
 	inst.src_count = 2;
-	return true;
 }
 
-bool DecodeSopc(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error) {
+void DecodeSopc(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst) {
 	const uint32_t word   = code[word_index];
 	const uint32_t ssrc1  = (word >> 8u) & 0xffu;
 	const uint32_t ssrc0  = word & 0xffu;
@@ -261,10 +256,10 @@ bool DecodeSopc(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 
 	if (inst.opcode == Opcode::UNSUPPORTED) {
 		SetUnsupported(inst, Family::SOPC, opcode, "SOPC opcode is not implemented");
-		return true;
+		return;
 	}
 
-	return DecodeBinarySources(pc, code, word_index, inst, ssrc0, ssrc1, error);
+	DecodeBinarySources(pc, code, word_index, inst, ssrc0, ssrc1);
 }
 
 void DecodeSopp(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
@@ -285,6 +280,7 @@ void DecodeSopp(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 	                           : static_cast<int32_t>(static_cast<int16_t>(simm));
 	inst.src_count = (inst.opcode == Opcode::S_NOP || inst.opcode == Opcode::S_WAITCNT ||
 	                  inst.opcode == Opcode::S_WAITCNT_DEPCTR || inst.opcode == Opcode::S_SLEEP ||
+	                  inst.opcode == Opcode::S_SETPRIO ||
 	                  inst.opcode == Opcode::S_SENDMSG || inst.opcode == Opcode::S_TRAP ||
 	                  inst.opcode == Opcode::S_TTRACEDATA || inst.opcode == Opcode::S_INST_PREFETCH)
 	                     ? 1

src/graphics/shader/recompiler/frontend/decode/ScalarAluOps.h

@@ -5,14 +5,14 @@
 
 namespace Libs::Graphics::ShaderRecompiler::Decoder {
 
-bool DecodeSop1(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error);
-bool DecodeSop2(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error);
-bool DecodeSopk(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error);
-bool DecodeSopc(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error);
+void DecodeSop1(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst);
+void DecodeSop2(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst);
+void DecodeSopk(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst);
+void DecodeSopc(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst);
 void DecodeSopp(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
                 Instruction& inst);
 

src/graphics/shader/recompiler/frontend/decode/ShaderDecoder.cpp

@@ -1,5 +1,6 @@
 #include "graphics/shader/recompiler/frontend/decode/ShaderDecoder.h"
 
+#include "common/assert.h"
 #include "graphics/shader/recompiler/frontend/decode/ExportOps.h"
 #include "graphics/shader/recompiler/frontend/decode/ImageOps.h"
 #include "graphics/shader/recompiler/frontend/decode/MemoryOps.h"
@@ -17,12 +18,6 @@ uint32_t FloatBits(float value) {
 	return std::bit_cast<uint32_t>(value);
 }
 
-void SetError(std::string* error, const char* message) {
-	if (error != nullptr) {
-		*error = message;
-	}
-}
-
 bool HasLiteral(const Instruction& inst) {
 	return inst.src0.kind == OperandKind::LiteralConstant ||
 	       inst.src1.kind == OperandKind::LiteralConstant ||
@@ -216,86 +211,76 @@ const char* ImageDimensionToString(ImageDimension dimension) {
 	}
 }
 
-bool DecodeScalarSource(uint32_t code, uint32_t pc, Operand& operand, std::string* error) {
+void DecodeScalarSource(uint32_t code, uint32_t pc, Operand& operand) {
 	operand = {};
 
 	if (code <= 105u) {
 		operand.kind = OperandKind::Sgpr;
 		operand.reg  = code;
-		return true;
+		return;
 	}
 	if (code >= 128u && code <= 192u) {
 		operand.kind       = OperandKind::IntegerInlineConstant;
 		operand.signed_val = static_cast<int32_t>(code - 128u);
 		operand.value      = static_cast<uint32_t>(operand.signed_val);
-		return true;
+		return;
 	}
 	if (code >= 193u && code <= 208u) {
 		operand.kind       = OperandKind::IntegerInlineConstant;
 		operand.signed_val = 192 - static_cast<int32_t>(code);
 		operand.value      = static_cast<uint32_t>(operand.signed_val);
-		return true;
+		return;
 	}
 	if (code >= 240u && code <= 247u) {
 		constexpr float values[] = {0.5f, -0.5f, 1.0f, -1.0f, 2.0f, -2.0f, 4.0f, -4.0f};
 		operand.kind             = OperandKind::FloatInlineConstant;
 		operand.float_val        = values[code - 240u];
 		operand.value            = FloatBits(operand.float_val);
-		return true;
+		return;
 	}
 	if (code >= 256u && code <= 511u) {
 		DecodeVectorGpr(code - 256u, operand);
-		return true;
+		return;
 	}
 
 	switch (code) {
-		case 106u: operand.kind = OperandKind::VccLo; return true;
-		case 107u: operand.kind = OperandKind::VccHi; return true;
-		case 124u: operand.kind = OperandKind::M0; return true;
-		case 125u: operand.kind = OperandKind::Null; return true;
-		case 126u: operand.kind = OperandKind::ExecLo; return true;
-		case 127u: operand.kind = OperandKind::ExecHi; return true;
-		case 239u: operand.kind = OperandKind::PopsExitingWaveId; return true;
+		case 106u: operand.kind = OperandKind::VccLo; return;
+		case 107u: operand.kind = OperandKind::VccHi; return;
+		case 124u: operand.kind = OperandKind::M0; return;
+		case 125u: operand.kind = OperandKind::Null; return;
+		case 126u: operand.kind = OperandKind::ExecLo; return;
+		case 127u: operand.kind = OperandKind::ExecHi; return;
+		case 239u: operand.kind = OperandKind::PopsExitingWaveId; return;
 		case 248u:
 			operand.kind      = OperandKind::FloatInlineConstant;
 			operand.float_val = 0.15915494309189535f;
 			operand.value     = FloatBits(operand.float_val);
-			return true;
-		case 251u: operand.kind = OperandKind::VccZ; return true;
-		case 252u: operand.kind = OperandKind::ExecZ; return true;
-		case 253u: operand.kind = OperandKind::Scc; return true;
-		case 255u: operand.kind = OperandKind::LiteralConstant; return true;
-		default:
-			if (error != nullptr) {
-				*error = fmt::format("unsupported scalar source operand 0x{:08x} at pc 0x{:08x}",
-				                     code, pc);
-			}
-			return false;
+			return;
+		case 251u: operand.kind = OperandKind::VccZ; return;
+		case 252u: operand.kind = OperandKind::ExecZ; return;
+		case 253u: operand.kind = OperandKind::Scc; return;
+		case 255u: operand.kind = OperandKind::LiteralConstant; return;
+		default: EXIT("unsupported scalar source operand 0x%08x at pc 0x%08x", code, pc);
 	}
 }
 
-bool DecodeScalarDestination(uint32_t code, uint32_t pc, Operand& operand, std::string* error) {
+void DecodeScalarDestination(uint32_t code, uint32_t pc, Operand& operand) {
 	operand = {};
 
 	if (code <= 105u) {
 		operand.kind = OperandKind::Sgpr;
 		operand.reg  = code;
-		return true;
+		return;
 	}
 
 	switch (code) {
-		case 106u: operand.kind = OperandKind::VccLo; return true;
-		case 107u: operand.kind = OperandKind::VccHi; return true;
-		case 124u: operand.kind = OperandKind::M0; return true;
-		case 125u: operand.kind = OperandKind::Null; return true;
-		case 126u: operand.kind = OperandKind::ExecLo; return true;
-		case 127u: operand.kind = OperandKind::ExecHi; return true;
-		default:
-			if (error != nullptr) {
-				*error = fmt::format(
-				    "unsupported scalar destination operand 0x{:08x} at pc 0x{:08x}", code, pc);
-			}
-			return false;
+		case 106u: operand.kind = OperandKind::VccLo; return;
+		case 107u: operand.kind = OperandKind::VccHi; return;
+		case 124u: operand.kind = OperandKind::M0; return;
+		case 125u: operand.kind = OperandKind::Null; return;
+		case 126u: operand.kind = OperandKind::ExecLo; return;
+		case 127u: operand.kind = OperandKind::ExecHi; return;
+		default: EXIT("unsupported scalar destination operand 0x%08x at pc 0x%08x", code, pc);
 	}
 }
 
@@ -368,39 +353,33 @@ Family GetInstructionFamily(uint32_t word) {
 	}
 }
 
-bool DecodeInstruction(std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                       std::string* error) {
+void DecodeInstruction(std::span<const uint32_t> code, uint32_t word_index, Instruction& inst) {
 	const uint32_t pc = word_index * sizeof(uint32_t);
 	switch (GetInstructionFamily(code[word_index])) {
-		case Family::SOP1: return DecodeSop1(pc, code, word_index, inst, error);
-		case Family::SOP2: return DecodeSop2(pc, code, word_index, inst, error);
-		case Family::SOPK: return DecodeSopk(pc, code, word_index, inst, error);
-		case Family::SOPC: return DecodeSopc(pc, code, word_index, inst, error);
-		case Family::SOPP: DecodeSopp(pc, code, word_index, inst); return true;
-		case Family::VOP1: return DecodeVop1(pc, code, word_index, inst, error);
-		case Family::VOP2: return DecodeVop2(pc, code, word_index, inst, error);
-		case Family::VOP3: return DecodeVop3(pc, code, word_index, inst, error);
-		case Family::VOP3P: return DecodeVop3p(pc, code, word_index, inst, error);
-		case Family::VOPC: return DecodeVopc(pc, code, word_index, inst, error);
-		case Family::VINTRP: DecodeVintrp(pc, code, word_index, inst); return true;
-		case Family::SMEM: DecodeSmem(pc, code, word_index, inst); return true;
-		case Family::MUBUF: DecodeMubuf(pc, code, word_index, inst); return true;
-		case Family::MTBUF: DecodeMtbuf(pc, code, word_index, inst); return true;
-		case Family::FLAT: DecodeFlat(pc, code, word_index, inst); return true;
-		case Family::DS: DecodeDs(pc, code, word_index, inst); return true;
-		case Family::MIMG: DecodeMimg(pc, code, word_index, inst); return true;
-		case Family::EXP: DecodeExp(pc, code, word_index, inst); return true;
+		case Family::SOP1: DecodeSop1(pc, code, word_index, inst); return;
+		case Family::SOP2: DecodeSop2(pc, code, word_index, inst); return;
+		case Family::SOPK: DecodeSopk(pc, code, word_index, inst); return;
+		case Family::SOPC: DecodeSopc(pc, code, word_index, inst); return;
+		case Family::SOPP: DecodeSopp(pc, code, word_index, inst); return;
+		case Family::VOP1: DecodeVop1(pc, code, word_index, inst); return;
+		case Family::VOP2: DecodeVop2(pc, code, word_index, inst); return;
+		case Family::VOP3: DecodeVop3(pc, code, word_index, inst); return;
+		case Family::VOP3P: DecodeVop3p(pc, code, word_index, inst); return;
+		case Family::VOPC: DecodeVopc(pc, code, word_index, inst); return;
+		case Family::VINTRP: DecodeVintrp(pc, code, word_index, inst); return;
+		case Family::SMEM: DecodeSmem(pc, code, word_index, inst); return;
+		case Family::MUBUF: DecodeMubuf(pc, code, word_index, inst); return;
+		case Family::MTBUF: DecodeMtbuf(pc, code, word_index, inst); return;
+		case Family::FLAT: DecodeFlat(pc, code, word_index, inst); return;
+		case Family::DS: DecodeDs(pc, code, word_index, inst); return;
+		case Family::MIMG: DecodeMimg(pc, code, word_index, inst); return;
+		case Family::EXP: DecodeExp(pc, code, word_index, inst); return;
 		default:
-			if (error != nullptr) {
-				*error =
-				    fmt::format("unknown RDNA2 instruction family at pc 0x{:08x}, raw=0x{:08x}", pc,
-				                code[word_index]);
-			}
-			return false;
+			EXIT("unknown RDNA2 instruction family at pc 0x%08x, raw=0x%08x", pc, code[word_index]);
 	}
 }
 
-bool DecodeProgram(std::span<const uint32_t> code, Program& program, std::string* error) {
+void DecodeProgram(std::span<const uint32_t> code, Program& program) {
 	program.instructions.clear();
 	program.instructions.reserve(code.size());
 	program.code = code;
@@ -408,10 +387,7 @@ bool DecodeProgram(std::span<const uint32_t> code, Program& program, std::string
 	std::vector<bool> branch_targets;
 	for (uint32_t word_index = 0; word_index < code.size();) {
 		program.instructions.emplace_back();
-		if (!DecodeInstruction(code, word_index, program.instructions.back(), error)) {
-			program.instructions.pop_back();
-			return false;
-		}
+		DecodeInstruction(code, word_index, program.instructions.back());
 
 		const auto& inst = program.instructions.back();
 		word_index += inst.word_count;
@@ -425,12 +401,11 @@ bool DecodeProgram(std::span<const uint32_t> code, Program& program, std::string
 		}
 		if (inst.opcode == Opcode::S_ENDPGM &&
 		    (word_index >= code.size() || branch_targets.empty() || !branch_targets[word_index])) {
-			return true;
+			return;
 		}
 	}
 
-	SetError(error, "shader decode reached the code boundary before S_ENDPGM");
-	return false;
+	EXIT("shader decode reached the code boundary before S_ENDPGM");
 }
 
 std::string OperandToString(const Operand& operand) {
@@ -501,8 +476,11 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::S_ABS_I32:
 		case Opcode::S_BREV_B32:
 		case Opcode::S_BCNT1_I32_B32:
+		case Opcode::S_FLBIT_I32_B32:
 		case Opcode::S_FF1_I32_B32:
 		case Opcode::S_FF1_I32_B64:
+		case Opcode::S_BITSET0_B64:
+		case Opcode::S_BITSET1_B64:
 		case Opcode::S_NOT_B64:
 		case Opcode::S_WQM_B64:
 		case Opcode::S_QUADMASK_B64:
@@ -529,6 +507,7 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::S_WAITCNT:
 		case Opcode::S_WAITCNT_DEPCTR:
 		case Opcode::S_SLEEP:
+		case Opcode::S_SETPRIO:
 		case Opcode::S_TRAP:
 		case Opcode::S_SENDMSG:
 		case Opcode::S_TTRACEDATA:
@@ -556,6 +535,7 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::IMAGE_SAMPLE:
 		case Opcode::IMAGE_STORE:
 		case Opcode::IMAGE_STORE_MIP:
+		case Opcode::IMAGE_ATOMIC_SWAP:
 		case Opcode::IMAGE_ATOMIC_ADD:
 		case Opcode::IMAGE_ATOMIC_UMIN:
 		case Opcode::IMAGE_ATOMIC_UMAX:
@@ -612,6 +592,8 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::TBUFFER_STORE_FORMAT_XYZ:
 		case Opcode::TBUFFER_STORE_FORMAT_XYZW:
 		case Opcode::BUFFER_ATOMIC_SWAP:
+		case Opcode::BUFFER_ATOMIC_CMPSWAP:
+		case Opcode::BUFFER_ATOMIC_SWAP_X2:
 		case Opcode::BUFFER_ATOMIC_ADD:
 		case Opcode::BUFFER_ATOMIC_SUB:
 		case Opcode::BUFFER_ATOMIC_SMIN:
@@ -620,6 +602,7 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::BUFFER_ATOMIC_UMAX:
 		case Opcode::BUFFER_ATOMIC_AND:
 		case Opcode::BUFFER_ATOMIC_OR:
+		case Opcode::BUFFER_ATOMIC_OR_X2:
 		case Opcode::BUFFER_ATOMIC_XOR:
 		case Opcode::BUFFER_ATOMIC_FMIN:
 		case Opcode::BUFFER_ATOMIC_FMAX:
@@ -661,6 +644,7 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::DS_MIN_F32:
 		case Opcode::DS_MAX_F32:
 		case Opcode::DS_SWIZZLE_B32:
+		case Opcode::DS_BPERMUTE_B32:
 		case Opcode::DS_READ_I8:
 		case Opcode::DS_READ_U8:
 		case Opcode::DS_READ_I16:
@@ -675,6 +659,7 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::DS_READ_B128:
 		case Opcode::DS_WRITE_B8:
 		case Opcode::DS_WRITE_B16:
+		case Opcode::DS_WRITE_B16_D16_HI:
 		case Opcode::DS_WRITE2_B32:
 		case Opcode::DS_WRITE2ST64_B32:
 		case Opcode::DS_WRITE2_B64:

src/graphics/shader/recompiler/frontend/decode/ShaderDecoder.h

@@ -40,11 +40,13 @@ enum class Opcode {
 	S_MOV_B64,
 	S_MOVK_I32,
 	S_ABS_I32,
+	S_ABSDIFF_I32,
 	S_BREV_B32,
 	S_BCNT1_I32_B32,
 	S_BCNT1_I32_B64,
 	S_FF1_I32_B32,
 	S_FF1_I32_B64,
+	S_FLBIT_I32_B32,
 	S_FLBIT_I32_B64,
 	S_BITREPLICATE_B64_B32,
 	S_QUADMASK_B64,
@@ -68,6 +70,8 @@ enum class Opcode {
 	S_BITCMP1_B32,
 	S_BITSET0_B32,
 	S_BITSET1_B32,
+	S_BITSET0_B64,
+	S_BITSET1_B64,
 	S_MIN_I32,
 	S_MAX_I32,
 	S_MIN_U32,
@@ -107,6 +111,7 @@ enum class Opcode {
 	S_BFM_B64,
 	S_CSELECT_B32,
 	S_CSELECT_B64,
+	S_SETPRIO,
 	S_PACK_LL_B32_B16,
 	S_PACK_LH_B32_B16,
 	S_PACK_HH_B32_B16,
@@ -179,12 +184,15 @@ enum class Opcode {
 	V_CEIL_F16,
 	V_TRUNC_F16,
 	V_RNDNE_F16,
+	V_SIN_F16,
+	V_COS_F16,
 	V_SIN_F32,
 	V_COS_F32,
 	V_NOT_B32,
 	V_BFREV_B32,
 	V_FFBH_U32,
 	V_FFBL_B32,
+	V_FFBH_I32,
 	V_ADD_F32,
 	V_SUB_F32,
 	V_SUBREV_F32,
@@ -278,6 +286,7 @@ enum class Opcode {
 	V_MAD_MIXHI_F16,
 	V_ADD_NC_U32,
 	V_ADDC_U32,
+	V_SUB_CO_CI_U32,
 	V_SUBREV_CO_CI_U32,
 	V_SUB_NC_U32,
 	V_SUBREV_NC_U32,
@@ -348,6 +357,7 @@ enum class Opcode {
 	V_CMP_GE_I32,
 	V_CMP_T_I32,
 	V_CMP_CLASS_F32,
+	V_CMPX_CLASS_F32,
 	V_CMP_LT_I16,
 	V_CMP_EQ_I16,
 	V_CMP_LE_I16,
@@ -391,6 +401,8 @@ enum class Opcode {
 	V_CMP_GE_U32,
 	V_CMP_T_U32,
 	V_CMP_EQ_I64,
+	V_CMP_LT_U64,
+	V_CMP_EQ_U64,
 	V_CMP_GT_U64,
 	V_CMP_NE_U64,
 	V_CMPX_NE_I64,
@@ -443,6 +455,8 @@ enum class Opcode {
 	TBUFFER_STORE_FORMAT_XYZ,
 	TBUFFER_STORE_FORMAT_XYZW,
 	BUFFER_ATOMIC_SWAP,
+	BUFFER_ATOMIC_CMPSWAP,
+	BUFFER_ATOMIC_SWAP_X2,
 	BUFFER_ATOMIC_ADD,
 	BUFFER_ATOMIC_SUB,
 	BUFFER_ATOMIC_SMIN,
@@ -451,6 +465,7 @@ enum class Opcode {
 	BUFFER_ATOMIC_UMAX,
 	BUFFER_ATOMIC_AND,
 	BUFFER_ATOMIC_OR,
+	BUFFER_ATOMIC_OR_X2,
 	BUFFER_ATOMIC_XOR,
 	BUFFER_ATOMIC_FMIN,
 	BUFFER_ATOMIC_FMAX,
@@ -490,6 +505,7 @@ enum class Opcode {
 	DS_MIN_F32,
 	DS_MAX_F32,
 	DS_SWIZZLE_B32,
+	DS_BPERMUTE_B32,
 	DS_CONSUME,
 	DS_APPEND,
 	DS_READ_I8,
@@ -497,6 +513,7 @@ enum class Opcode {
 	DS_READ_I16,
 	DS_READ_U16,
 	DS_READ_U16_D16,
+	DS_READ_U16_D16_HI,
 	DS_READ2_B32,
 	DS_READ2ST64_B32,
 	DS_READ_B32,
@@ -507,6 +524,7 @@ enum class Opcode {
 	DS_READ_B128,
 	DS_WRITE_B8,
 	DS_WRITE_B16,
+	DS_WRITE_B16_D16_HI,
 	DS_WRITE2_B32,
 	DS_WRITE2ST64_B32,
 	DS_WRITE2_B64,
@@ -523,6 +541,7 @@ enum class Opcode {
 	IMAGE_LOAD_MIP,
 	IMAGE_STORE,
 	IMAGE_STORE_MIP,
+	IMAGE_ATOMIC_SWAP,
 	IMAGE_ATOMIC_ADD,
 	IMAGE_ATOMIC_UMIN,
 	IMAGE_ATOMIC_UMAX,
@@ -624,6 +643,7 @@ struct Operand {
 	uint32_t dpp_ctrl           = 0;
 	uint32_t dpp_row_mask       = 0xf;
 	uint32_t dpp_bank_mask      = 0xf;
+	bool     explicit_sdwa_dst  = false;
 	bool     sdwa_sext          = false;
 	bool     dpp_fetch_inactive = false;
 	bool     dpp_bound_ctrl     = false;
@@ -695,12 +715,11 @@ struct Program {
 // Code spans are trusted to contain complete instructions, valid branch targets, and 32-bit PCs.
 Family GetInstructionFamily(uint32_t word);
 // The output object must be freshly initialized.
-bool DecodeInstruction(std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                       std::string* error);
-bool DecodeProgram(std::span<const uint32_t> code, Program& program, std::string* error);
+void DecodeInstruction(std::span<const uint32_t> code, uint32_t word_index, Instruction& inst);
+void DecodeProgram(std::span<const uint32_t> code, Program& program);
 
-bool DecodeScalarSource(uint32_t code, uint32_t pc, Operand& operand, std::string* error);
-bool DecodeScalarDestination(uint32_t code, uint32_t pc, Operand& operand, std::string* error);
+void DecodeScalarSource(uint32_t code, uint32_t pc, Operand& operand);
+void DecodeScalarDestination(uint32_t code, uint32_t pc, Operand& operand);
 void DecodeVectorGpr(uint32_t reg, Operand& operand);
 void ReadLiteralOperands(std::span<const uint32_t> code, uint32_t word_index, Instruction& inst);
 void SetRawWords(Instruction& inst, std::span<const uint32_t> code, uint32_t word_index,

src/graphics/shader/recompiler/frontend/decode/VectorAluOps.h

@@ -5,16 +5,16 @@
 
 namespace Libs::Graphics::ShaderRecompiler::Decoder {
 
-bool DecodeVop2(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error);
-bool DecodeVop1(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error);
-bool DecodeVopc(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error);
-bool DecodeVop3(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index, Instruction& inst,
-                std::string* error);
-bool DecodeVop3p(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
-                 Instruction& inst, std::string* error);
+void DecodeVop2(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst);
+void DecodeVop1(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst);
+void DecodeVopc(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst);
+void DecodeVop3(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                Instruction& inst);
+void DecodeVop3p(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
+                 Instruction& inst);
 void DecodeVintrp(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
                   Instruction& inst);
 

src/graphics/shader/recompiler/frontend/translate/Attribute.cpp

@@ -1,9 +1,8 @@
+#include "common/assert.h"
 #include "graphics/shader/recompiler/frontend/translate/Translator.h"
-#include "graphics/shader/shader.h"
 
 #include <algorithm>
 #include <array>
-#include <fmt/format.h>
 
 namespace Libs::Graphics::ShaderRecompiler::Frontend {
 
@@ -47,14 +46,13 @@ IR::ExportFlags Translator::AddExportInfo(const Decoder::Instruction& inst) {
 	return {.index = index, .pc = inst.pc};
 }
 
-bool Translator::TranslateEmbeddedFetch(const Decoder::Instruction& inst, uint32_t attribute,
+void Translator::TranslateEmbeddedFetch(const Decoder::Instruction& inst, uint32_t attribute,
                                         uint32_t component_count) {
 	for (uint32_t component = 0; component < component_count; component++) {
 		const auto value =
 		    ir.Emit(IR::ValueOpcode::GetAttribute, {IR::Value(attribute), IR::Value(component)});
 		WriteOperand(OffsetOperand(inst.dst, component), value);
 	}
-	return true;
 }
 
 void Translator::V_INTERP_P1_F32() {}
@@ -65,25 +63,20 @@ void Translator::V_INTERP_P2_F32(const Decoder::Instruction& inst) {
 	WriteOperand(inst.dst, value);
 }
 
-bool Translator::V_INTERP_MOV_F32(const Decoder::Instruction& inst, std::string* error) {
+void Translator::V_INTERP_MOV_F32(const Decoder::Instruction& inst) {
 	if (inst.src0.value >= 3u) {
-		return ShaderError::Fail(error,
-		                         fmt::format("v_interp_mov_f32 mode {} is reserved at pc 0x{:08x}",
-		                                     inst.src0.value, inst.pc));
+		EXIT("v_interp_mov_f32 mode %u is reserved at pc 0x%08x", inst.src0.value, inst.pc);
 	}
 	const auto value = ir.Emit(
 	    IR::ValueOpcode::GetInterpolationParameter,
 	    {IR::Value(inst.src1.value), IR::Value(inst.src2.value), IR::Value(inst.src0.value)});
 	WriteOperand(inst.dst, value);
-	return true;
 }
 
-bool Translator::EXP(const Decoder::Instruction& inst, std::string* error) {
+void Translator::EXP(const Decoder::Instruction& inst) {
 	uint32_t index = 0;
 	if (ExportTargetKindFromTarget(inst.exp.target, index) == IR::ExportTargetKind::Unknown) {
-		return ShaderError::Fail(error,
-		                         fmt::format("unsupported EXP target 0x{:02x} at pc 0x{:08x}",
-		                                     inst.exp.target, inst.pc));
+		EXIT("unsupported EXP target 0x%02x at pc 0x%08x", inst.exp.target, inst.pc);
 	}
 	std::array<IR::Value, 4> components {IR::Value(0u), IR::Value(0u), IR::Value(0u),
 	                                     IR::Value(0u)};
@@ -93,14 +86,13 @@ bool Translator::EXP(const Decoder::Instruction& inst, std::string* error) {
 	const auto data = ir.Emit(IR::ValueOpcode::CompositeConstructU32x4,
 	                          {components[0], components[1], components[2], components[3]});
 	ir.Emit(IR::ValueOpcode::SetAttribute, {data, ir.GetExec()}, AddExportInfo(inst));
-	return true;
 }
 
-bool Translator::EmitInterpolation(const Decoder::Instruction& inst, std::string* error) {
+bool Translator::EmitInterpolation(const Decoder::Instruction& inst) {
 	switch (inst.opcode) {
 		case Decoder::Opcode::V_INTERP_P1_F32: V_INTERP_P1_F32(); return true;
 		case Decoder::Opcode::V_INTERP_P2_F32: V_INTERP_P2_F32(inst); return true;
-		case Decoder::Opcode::V_INTERP_MOV_F32: return V_INTERP_MOV_F32(inst, error);
+		case Decoder::Opcode::V_INTERP_MOV_F32: V_INTERP_MOV_F32(inst); return true;
 		default: return false;
 	}
 }

src/graphics/shader/recompiler/frontend/translate/Compare.cpp

@@ -55,11 +55,11 @@ void Translator::EmitFloatOrderedCompare(const Decoder::Instruction& inst, bool
 	EmitCompareResult(inst, ordered ? ir.LogicalNot(unordered) : unordered, false, false);
 }
 
-void Translator::V_CMP_CLASS_F32(const Decoder::Instruction& inst) {
+void Translator::EmitFloatClassCompare(const Decoder::Instruction& inst, bool cmpx) {
 	const auto value = ReadOperand(inst.src0, IR::Type::F32);
 	const auto mask  = ReadOperand(inst.src1, IR::Type::U32);
 	EmitCompareResult(inst, IR::U1(ir.Emit(IR::ValueOpcode::FPCmpClass32, {value, mask})), false,
-	                  false);
+	                  cmpx);
 }
 
 } // namespace Libs::Graphics::ShaderRecompiler::Frontend

src/graphics/shader/recompiler/frontend/translate/Control.cpp

@@ -1,3 +1,4 @@
+#include "common/assert.h"
 #include "graphics/shader/recompiler/frontend/translate/Translator.h"
 
 namespace Libs::Graphics::ShaderRecompiler::Frontend {
@@ -75,9 +76,9 @@ void Translator::SUB_U32(const Decoder::Instruction& inst, bool vector, bool rev
 	}
 }
 
-void Translator::SUBB_U32(const Decoder::Instruction& inst, bool vector) {
-	const auto lhs       = ReadU32(vector ? inst.src1 : inst.src0);
-	const auto rhs       = ReadU32(vector ? inst.src0 : inst.src1);
+void Translator::SUBB_U32(const Decoder::Instruction& inst, bool vector, bool reverse) {
+	const auto lhs       = ReadU32(reverse ? inst.src1 : inst.src0);
+	const auto rhs       = ReadU32(reverse ? inst.src0 : inst.src1);
 	const auto borrow_in = vector ? ConditionBit(inst.src2)
 	                              : ConditionBit(ConditionOperand(Decoder::OperandKind::Scc));
 	const auto partial   = ir.ISub(lhs, rhs);
@@ -93,6 +94,13 @@ void Translator::SUBB_U32(const Decoder::Instruction& inst, bool vector) {
 	}
 }
 
+void Translator::S_ABSDIFF_I32(const Decoder::Instruction& inst) {
+	const auto difference = ir.ISub(ReadU32(inst.src0), ReadU32(inst.src1));
+	const auto result     = IR::U32(ir.Emit(IR::ValueOpcode::IAbs32, {difference}));
+	WriteOperand(DestinationOperand(inst), result);
+	ir.SetScc(ir.INotEqual(result, IR::U32(IR::Value(0u))));
+}
+
 void Translator::S_ADD_SUB_I32(const Decoder::Instruction& inst, bool subtract) {
 	const auto lhs      = ReadU32(inst.src0);
 	const auto rhs      = ReadU32(inst.src1);
@@ -153,12 +161,12 @@ void Translator::S_INST_PREFETCH() {
 	ir.Emit(IR::ValueOpcode::InstPrefetch);
 }
 
-bool Translator::S_GETPC_B64(const Decoder::Instruction& inst, std::string* error) {
+void Translator::S_GETPC_B64(const Decoder::Instruction& inst) {
 	const auto base    = IR::U64(ir.Emit(IR::ValueOpcode::GetShaderBase));
 	const auto address = IR::U64(
 	    ir.Emit(IR::ValueOpcode::IAdd64, {base, IR::Value(static_cast<uint64_t>(inst.pc) + 4u)}));
 	if (inst.dst.kind == Decoder::OperandKind::Null) {
-		return true;
+		return;
 	}
 	auto high = PlainOperand(inst.dst);
 	switch (inst.dst.kind) {
@@ -169,25 +177,19 @@ bool Translator::S_GETPC_B64(const Decoder::Instruction& inst, std::string* erro
 				high.kind = Decoder::OperandKind::VccLo;
 				high.reg  = 0u;
 			} else {
-				if (error != nullptr) {
-					*error = "S_GETPC_B64 destination does not name a valid scalar pair";
-				}
-				return false;
+				EXIT("S_GETPC_B64 destination does not name a valid scalar pair at pc 0x%08x",
+				     inst.pc);
 			}
 			break;
 		case Decoder::OperandKind::VccLo: high.kind = Decoder::OperandKind::VccHi; break;
 		case Decoder::OperandKind::M0: high.kind = Decoder::OperandKind::Null; break;
 		case Decoder::OperandKind::ExecLo: high.kind = Decoder::OperandKind::ExecHi; break;
 		default:
-			if (error != nullptr) {
-				*error = "S_GETPC_B64 destination does not name a valid scalar pair";
-			}
-			return false;
+			EXIT("S_GETPC_B64 destination does not name a valid scalar pair at pc 0x%08x", inst.pc);
 	}
 	const auto words = ExtractU64(address);
 	WriteOperand(inst.dst, words[0]);
 	WriteOperand(high, words[1]);
-	return true;
 }
 
 void Translator::S_CSELECT_B32(const Decoder::Instruction& inst) {
@@ -274,21 +276,15 @@ void Translator::S_WQM_B64(const Decoder::Instruction& inst) {
 	ir.SetScc(result);
 }
 
-bool Translator::V_MOVRELS_B32(const Decoder::Instruction& inst, std::string* error) {
+void Translator::V_MOVRELS_B32(const Decoder::Instruction& inst) {
 	if (inst.dst.kind != Decoder::OperandKind::Vgpr ||
 	    inst.src0.kind != Decoder::OperandKind::Vgpr) {
-		if (error != nullptr) {
-			*error = "V_MOVRELS_B32 requires VGPR source and destination";
-		}
-		return false;
+		EXIT("V_MOVRELS_B32 requires VGPR source and destination at pc 0x%08x", inst.pc);
 	}
 	if (inst.dst.sdwa_sel != 6u || inst.dst.omod != 0u || inst.dst.clamp ||
 	    inst.src0.sdwa_sel != 6u || inst.src0.sdwa_sext || inst.src0.negate || inst.src0.absolute ||
 	    inst.src0.dpp) {
-		if (error != nullptr) {
-			*error = "V_MOVRELS_B32 modifiers are not implemented";
-		}
-		return false;
+		EXIT("V_MOVRELS_B32 modifiers are not implemented at pc 0x%08x", inst.pc);
 	}
 	const auto base     = inst.src0.reg;
 	const auto m0       = ir.BitwiseAnd(ReadU32(ConditionOperand(Decoder::OperandKind::M0)),
@@ -299,23 +295,16 @@ bool Translator::V_MOVRELS_B32(const Decoder::Instruction& inst, std::string* er
 		selected = ir.Select(match, ir.GetVectorReg(static_cast<IR::VectorReg>(index)), selected);
 	}
 	WriteOperand(DestinationOperand(inst), selected);
-	return true;
 }
 
-bool Translator::V_MOVRELD_B32(const Decoder::Instruction& inst, std::string* error) {
+void Translator::V_MOVRELD_B32(const Decoder::Instruction& inst) {
 	if (inst.dst.kind != Decoder::OperandKind::Vgpr) {
-		if (error != nullptr) {
-			*error = "V_MOVRELD_B32 requires VGPR destination";
-		}
-		return false;
+		EXIT("V_MOVRELD_B32 requires VGPR destination at pc 0x%08x", inst.pc);
 	}
 	if (inst.dst.sdwa_sel != 6u || inst.dst.omod != 0u || inst.dst.clamp ||
 	    inst.src0.sdwa_sel != 6u || inst.src0.sdwa_sext || inst.src0.negate || inst.src0.absolute ||
 	    inst.src0.dpp) {
-		if (error != nullptr) {
-			*error = "V_MOVRELD_B32 modifiers are not implemented";
-		}
-		return false;
+		EXIT("V_MOVRELD_B32 modifiers are not implemented at pc 0x%08x", inst.pc);
 	}
 	const auto base  = inst.dst.reg;
 	const auto value = ReadU32(inst.src0);
@@ -327,7 +316,6 @@ bool Translator::V_MOVRELD_B32(const Decoder::Instruction& inst, std::string* er
 		const auto write = ir.LogicalAnd(ir.GetExec(), match);
 		ir.SetVectorReg(reg, ir.Select(write, value, ir.GetVectorReg(reg)));
 	}
-	return true;
 }
 
 void Translator::V_READFIRSTLANE_B32(const Decoder::Instruction& inst) {

src/graphics/shader/recompiler/frontend/translate/Convert.cpp

@@ -141,8 +141,10 @@ void Translator::V_CVT_OFF_F32_I4(const Decoder::Instruction& inst) {
 }
 
 void Translator::V_CVT_PKRTZ_F16_F32(const Decoder::Instruction& inst) {
-	const auto lhs = ReadOperand(SourceAt(inst, 0), IR::Type::F32);
-	const auto rhs = ReadOperand(SourceAt(inst, 1), IR::Type::F32);
+	const auto lhs = ApplyF32ResultModifiers(
+	    inst.dst, IR::F32(ReadOperand(SourceAt(inst, 0), IR::Type::F32)));
+	const auto rhs = ApplyF32ResultModifiers(
+	    inst.dst, IR::F32(ReadOperand(SourceAt(inst, 1), IR::Type::F32)));
 	WriteOperand(DestinationOperand(inst), ir.Emit(IR::ValueOpcode::PackFloat2x16Rtz, {lhs, rhs}));
 }
 

src/graphics/shader/recompiler/frontend/translate/Dispatch.cpp

@@ -1,23 +1,18 @@
+#include "common/assert.h"
 #include "graphics/shader/recompiler/frontend/translate/Translator.h"
 
 namespace Libs::Graphics::ShaderRecompiler::Frontend {
 
-bool Translator::TranslateInstruction(const Decoder::Instruction& inst, std::string* error) {
+void Translator::TranslateInstruction(const Decoder::Instruction& inst) {
 	current_opcode = inst.opcode;
 	current_pc     = inst.pc;
 
 	switch (inst.opcode) {
 		case Decoder::Opcode::UNKNOWN:
 		case Decoder::Opcode::COUNT:
-			if (error != nullptr) {
-				*error = "decoded opcode has no IR translation";
-			}
-			return false;
+			EXIT("decoded opcode has no IR translation at pc 0x%08x", inst.pc);
 		case Decoder::Opcode::UNSUPPORTED:
-			if (error != nullptr) {
-				*error = "unsupported decoded instruction: " + Decoder::InstructionToString(inst);
-			}
-			return false;
+			EXIT("unsupported decoded instruction: %s", Decoder::InstructionToString(inst).c_str());
 		default: break;
 	}
 
@@ -27,27 +22,30 @@ bool Translator::TranslateInstruction(const Decoder::Instruction& inst, std::str
 		case Decoder::Family::SOP2:
 		case Decoder::Family::SOPK:
 		case Decoder::Family::SOPC:
-		case Decoder::Family::SOPP: translated = EmitScalar(inst, error); break;
+		case Decoder::Family::SOPP: translated = EmitScalar(inst); break;
 		case Decoder::Family::VOP1:
 		case Decoder::Family::VOP2:
 		case Decoder::Family::VOP3:
 		case Decoder::Family::VOP3P:
-		case Decoder::Family::VOPC: translated = EmitVector(inst, error); break;
+		case Decoder::Family::VOPC: translated = EmitVector(inst); break;
 		case Decoder::Family::SMEM:
 		case Decoder::Family::MUBUF:
 		case Decoder::Family::MTBUF:
 		case Decoder::Family::FLAT:
 		case Decoder::Family::DS:
-		case Decoder::Family::MIMG: translated = EmitMemory(inst, error); break;
-		case Decoder::Family::VINTRP: translated = EmitInterpolation(inst, error); break;
-		case Decoder::Family::EXP: translated = EXP(inst, error); break;
+		case Decoder::Family::MIMG: translated = EmitMemory(inst); break;
+		case Decoder::Family::VINTRP: translated = EmitInterpolation(inst); break;
+		case Decoder::Family::EXP:
+			EXP(inst);
+			translated = true;
+			break;
 		default: break;
 	}
 
-	if (!translated && error != nullptr && error->empty()) {
-		*error = "decoded opcode has no IR translation";
+	if (!translated) {
+		EXIT("opcode %s at pc 0x%08x has no IR translation",
+		     Decoder::InstructionToString(inst).c_str(), inst.pc);
 	}
-	return translated;
 }
 
 } // namespace Libs::Graphics::ShaderRecompiler::Frontend

src/graphics/shader/recompiler/frontend/translate/Float.cpp

@@ -63,6 +63,42 @@ bool Translator::Float16Unary(const Decoder::Instruction& inst, IR::ValueOpcode
 	return true;
 }
 
+bool Translator::Float16Trig(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
+	const auto argument = ReadF16AsF32(inst.src0);
+	const auto magnitude =
+	    ir.BitwiseAnd(ir.BitCastU32(argument), IR::U32(IR::Value(0x7fffffffu)));
+	auto result = IR::F32(ir.Emit(opcode, {argument}));
+	if (opcode == IR::ValueOpcode::FPSin) {
+		const auto fraction = IR::F32(ir.Emit(IR::ValueOpcode::FPFract32, {argument}));
+		const auto whole = IR::U1(
+		    ir.Emit(IR::ValueOpcode::FPOrdEqual32, {fraction, IR::Value::F32(0.0f)}));
+		const auto half = IR::U1(
+		    ir.Emit(IR::ValueOpcode::FPOrdEqual32, {fraction, IR::Value::F32(0.5f)}));
+		const auto nonzero = ir.INotEqual(magnitude, IR::U32(IR::Value(0u)));
+		const auto cardinal = ir.LogicalOr(ir.LogicalAnd(whole, nonzero), half);
+		result = IR::F32(ir.Emit(
+		    IR::ValueOpcode::SelectF32, {cardinal, IR::Value::F32(0.0f), result}));
+	} else {
+		// The architectural quarter-cycle result is exactly zero. Evaluating cos(pi/2) with a
+		// rounded F32 PI can otherwise produce a half-precision subnormal instead.
+		const auto fraction = IR::F32(ir.Emit(IR::ValueOpcode::FPFract32, {argument}));
+		const auto quarter = IR::U1(
+		    ir.Emit(IR::ValueOpcode::FPOrdEqual32, {fraction, IR::Value::F32(0.25f)}));
+		const auto three_quarters = IR::U1(
+		    ir.Emit(IR::ValueOpcode::FPOrdEqual32, {fraction, IR::Value::F32(0.75f)}));
+		result = IR::F32(
+		    ir.Emit(IR::ValueOpcode::SelectF32,
+		            {ir.LogicalOr(quarter, three_quarters), IR::Value::F32(0.0f), result}));
+	}
+
+	const auto infinite = ir.IEqual(magnitude, IR::U32(IR::Value(0x7f800000u)));
+	result              = ApplyF32ResultModifiers(inst.dst, result);
+	const auto bits    = PackHalf2x16(result, IR::F32(IR::Value::F32(0.0f)));
+	const auto invalid = IR::U32(IR::Value(inst.dst.clamp ? 0u : 0xfe00u));
+	WriteU16(DestinationOperand(inst), ir.Select(infinite, invalid, bits));
+	return true;
+}
+
 bool Translator::Float16Binary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                                bool reverse) {
 	const auto lhs = ReadF16AsF32(reverse ? inst.src1 : inst.src0);

src/graphics/shader/recompiler/frontend/translate/Integer.cpp

@@ -224,11 +224,14 @@ bool Translator::S_FF1_I32_B64(const Decoder::Instruction& inst) {
 	return true;
 }
 
-bool Translator::V_FFBH_U32(const Decoder::Instruction& inst) {
-	const auto source   = ReadU32(inst.src0);
-	const auto msb      = IR::U32(ir.Emit(IR::ValueOpcode::FindUMsb32, {source}));
+bool Translator::V_FFBH_32(const Decoder::Instruction& inst, bool sign) {
+	const auto source = ReadU32(inst.src0);
+	const auto value  = sign ? ir.BitwiseXor(
+	                              source, ir.ShiftRightArithmetic(source, IR::U32(IR::Value(31u))))
+	                         : source;
+	const auto msb      = IR::U32(ir.Emit(IR::ValueOpcode::FindUMsb32, {value}));
 	const auto position = ir.ISub(IR::U32(IR::Value(31u)), msb);
-	const auto result   = ir.Select(ir.INotEqual(source, IR::U32(IR::Value(0u))), position,
+	const auto result   = ir.Select(ir.INotEqual(value, IR::U32(IR::Value(0u))), position,
 	                                IR::U32(IR::Value(0xffffffffu)));
 	WriteOperand(DestinationOperand(inst), result);
 	return true;
@@ -309,6 +312,22 @@ bool Translator::S_BITSET_B32(const Decoder::Instruction& inst, bool set) {
 	return true;
 }
 
+bool Translator::S_BITSET_B64(const Decoder::Instruction& inst, bool set) {
+	const auto offset    = ir.BitwiseAnd(ReadU32(inst.src0), IR::U32(IR::Value(63u)));
+	const auto word_bit  = ir.BitwiseAnd(offset, IR::U32(IR::Value(31u)));
+	const auto bit       = ir.ShiftLeftLogical(IR::U32(IR::Value(1u)), word_bit);
+	const auto old       = ReadU32Pair(inst.dst);
+	const auto low_value = set ? ir.BitwiseOr(old[0], bit)
+	                           : ir.BitwiseAnd(old[0], ir.BitwiseNot(bit));
+	const auto high_value = set ? ir.BitwiseOr(old[1], bit)
+	                            : ir.BitwiseAnd(old[1], ir.BitwiseNot(bit));
+	const auto high = IR::U1(ir.Emit(IR::ValueOpcode::UGreaterThanEqual32,
+	                                {offset, IR::Value(32u)}));
+	WriteU32Pair(inst.dst,
+	             {ir.Select(high, old[0], low_value), ir.Select(high, high_value, old[1])});
+	return true;
+}
+
 bool Translator::V_BCNT_U32_B32(const Decoder::Instruction& inst) {
 	const auto count = IR::U32(ir.Emit(IR::ValueOpcode::BitCount32, {ReadU32(inst.src0)}));
 	WriteOperand(DestinationOperand(inst), ir.IAdd(count, ReadU32(inst.src1)));

src/graphics/shader/recompiler/frontend/translate/Memory.cpp

@@ -1,9 +1,7 @@
-#include "common/magicEnum.h"
 #include "graphics/shader/recompiler/frontend/translate/Translator.h"
 
 #include <algorithm>
 #include <array>
-#include <fmt/format.h>
 
 namespace Libs::Graphics::ShaderRecompiler::Frontend {
 
@@ -37,6 +35,7 @@ Decoder::Operand OffsetDecodedRegister(const Decoder::Operand& operand, uint32_t
 	result.dpp_ctrl           = 0;
 	result.dpp_row_mask       = 0xf;
 	result.dpp_bank_mask      = 0xf;
+	result.explicit_sdwa_dst  = false;
 	result.dpp_fetch_inactive = false;
 	result.dpp_bound_ctrl     = false;
 	result.dpp                = false;
@@ -85,18 +84,7 @@ ResourceKind MemoryKind(const Decoder::Instruction& decoded) {
 		case Decoder::Family::MTBUF: return ResourceKind::Buffer;
 		case Decoder::Family::FLAT: return FlatSegmentResourceKind(decoded.memory_segment);
 		case Decoder::Family::DS: return decoded.gds ? ResourceKind::Gds : ResourceKind::Lds;
-		case Decoder::Family::MIMG:
-			switch (decoded.opcode) {
-				case Decoder::Opcode::IMAGE_STORE:
-				case Decoder::Opcode::IMAGE_STORE_MIP: return ResourceKind::StorageImage;
-				case Decoder::Opcode::IMAGE_ATOMIC_ADD:
-				case Decoder::Opcode::IMAGE_ATOMIC_UMIN:
-				case Decoder::Opcode::IMAGE_ATOMIC_UMAX:
-				case Decoder::Opcode::IMAGE_ATOMIC_AND:
-				case Decoder::Opcode::IMAGE_ATOMIC_OR:
-				case Decoder::Opcode::IMAGE_ATOMIC_XOR: return ResourceKind::StorageImageUint;
-				default: return ResourceKind::Image;
-			}
+		case Decoder::Family::MIMG: return ResourceKind::Image;
 		default: return ResourceKind::None;
 	}
 }
@@ -154,10 +142,6 @@ IR::MemoryInfo MemoryInfoFromDecoded(const Decoder::Instruction& decoded) {
 	           decoded.opcode == Decoder::Opcode::DS_WRITE2ST64_B32) {
 		memory.data_dwords = 2u;
 	}
-	if (memory.kind == ResourceKind::StorageImage ||
-	    memory.kind == ResourceKind::StorageImageUint) {
-		memory.sampler = 0;
-	}
 	return memory;
 }
 
@@ -239,9 +223,12 @@ Decoder::Operand MemorySourceAt(const Decoder::Instruction& decoded, uint32_t in
 	if (decoded.family == Decoder::Family::MIMG) {
 		const bool store_or_atomic = decoded.opcode == Decoder::Opcode::IMAGE_STORE ||
 		                             decoded.opcode == Decoder::Opcode::IMAGE_STORE_MIP ||
-		                             (decoded.opcode >= Decoder::Opcode::IMAGE_ATOMIC_ADD &&
+		                             (decoded.opcode >= Decoder::Opcode::IMAGE_ATOMIC_SWAP &&
 		                              decoded.opcode <= Decoder::Opcode::IMAGE_ATOMIC_XOR);
-		return store_or_atomic ? (index == 0u ? decoded.dst : decoded.src0) : decoded.src0;
+		if (store_or_atomic) {
+			return index == 0u ? decoded.dst : decoded.src0;
+		}
+		return decoded.src0;
 	}
 	if (decoded.family == Decoder::Family::DS) {
 		switch (decoded.opcode) {
@@ -257,6 +244,7 @@ Decoder::Operand MemorySourceAt(const Decoder::Instruction& decoded, uint32_t in
 				return index == 0u ? decoded.src1 : index == 1u ? decoded.src0 : decoded.src2;
 			case Decoder::Opcode::DS_WRITE_B8:
 			case Decoder::Opcode::DS_WRITE_B16:
+			case Decoder::Opcode::DS_WRITE_B16_D16_HI:
 			case Decoder::Opcode::DS_WRITE_B32:
 			case Decoder::Opcode::DS_WRITE_B64:
 			case Decoder::Opcode::DS_WRITE_B96:
@@ -534,10 +522,24 @@ bool Translator::BUFFER_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode
 	const auto memory   = MemoryInfoFromDecoded(inst);
 	const auto resource = GetBufferResource(memory);
 	const auto address  = ReadBufferAddress(inst, 1);
-	const auto result   = ir.Emit(opcode,
-	                              {resource, address.index, address.offset, address.soffset,
-	                               ReadU32(MemorySourceAt(inst, 0)), ir.GetExec()},
-	                              AddMemoryInfo(memory, inst.pc));
+	const auto data_src = MemorySourceAt(inst, 0);
+	const auto flags    = AddMemoryInfo(memory, inst.pc);
+	IR::Value  result;
+	if (opcode == IR::ValueOpcode::BufferAtomicCmpSwap32) {
+		const auto desired    = ReadU32(data_src);
+		const auto comparator = ReadU32(OffsetOperand(data_src, 1u));
+		result = ir.Emit(opcode,
+		                 {resource, address.index, address.offset, address.soffset, desired,
+		                  comparator, ir.GetExec()},
+		                 flags);
+	} else {
+		const IR::Value value =
+		    memory.data_dwords == 2u ? IR::Value(ReadU64(data_src)) : IR::Value(ReadU32(data_src));
+		result = ir.Emit(opcode,
+		                 {resource, address.index, address.offset, address.soffset, value,
+		                  ir.GetExec()},
+		                 flags);
+	}
 	if (inst.glc) {
 		WriteOperand(inst.dst, result);
 	}
@@ -877,7 +879,14 @@ bool Translator::DS_SWIZZLE_B32(const Decoder::Instruction& inst) {
 	return true;
 }
 
-bool Translator::EmitMemory(const Decoder::Instruction& inst, std::string* error) {
+bool Translator::DS_BPERMUTE_B32(const Decoder::Instruction& inst) {
+	const auto address = ir.IAdd(ReadU32(inst.src0), IR::U32(IR::Value(inst.offset)));
+	WriteOperand(inst.dst, ir.Emit(IR::ValueOpcode::BpermuteU32,
+	                               {ReadU32(inst.src1), address, ir.GetExec()}));
+	return true;
+}
+
+bool Translator::EmitMemory(const Decoder::Instruction& inst) {
 	switch (inst.opcode) {
 		case Decoder::Opcode::S_LOAD_DWORD:
 		case Decoder::Opcode::S_LOAD_DWORDX2:
@@ -924,6 +933,10 @@ bool Translator::EmitMemory(const Decoder::Instruction& inst, std::string* error
 
 		case Decoder::Opcode::BUFFER_ATOMIC_SWAP:
 			return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicSwap32);
+		case Decoder::Opcode::BUFFER_ATOMIC_CMPSWAP:
+			return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicCmpSwap32);
+		case Decoder::Opcode::BUFFER_ATOMIC_SWAP_X2:
+			return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicSwap64);
 		case Decoder::Opcode::BUFFER_ATOMIC_ADD:
 			return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicIAdd32);
 		case Decoder::Opcode::BUFFER_ATOMIC_SUB:
@@ -940,6 +953,8 @@ bool Translator::EmitMemory(const Decoder::Instruction& inst, std::string* error
 			return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicAnd32);
 		case Decoder::Opcode::BUFFER_ATOMIC_OR:
 			return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicOr32);
+		case Decoder::Opcode::BUFFER_ATOMIC_OR_X2:
+			return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicOr64);
 		case Decoder::Opcode::BUFFER_ATOMIC_XOR:
 			return BUFFER_ATOMIC(inst, IR::ValueOpcode::BufferAtomicXor32);
 		case Decoder::Opcode::BUFFER_ATOMIC_FMIN:
@@ -986,6 +1001,8 @@ bool Translator::EmitMemory(const Decoder::Instruction& inst, std::string* error
 		case Decoder::Opcode::DS_WRXCHG_RTN_B32:
 			return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicSwap32, true);
 
+		case Decoder::Opcode::IMAGE_ATOMIC_SWAP:
+			return IMAGE_ATOMIC(inst, IR::ValueOpcode::ImageAtomicSwap32);
 		case Decoder::Opcode::IMAGE_ATOMIC_ADD:
 			return IMAGE_ATOMIC(inst, IR::ValueOpcode::ImageAtomicIAdd32);
 		case Decoder::Opcode::IMAGE_ATOMIC_UMIN:
@@ -1035,6 +1052,7 @@ bool Translator::EmitMemory(const Decoder::Instruction& inst, std::string* error
 		case Decoder::Opcode::DS_MAX_F32:
 			return DS_MINMAX_F32(inst, IR::ValueOpcode::SharedAtomicFMax32);
 		case Decoder::Opcode::DS_SWIZZLE_B32: return DS_SWIZZLE_B32(inst);
+		case Decoder::Opcode::DS_BPERMUTE_B32: return DS_BPERMUTE_B32(inst);
 		case Decoder::Opcode::DS_CONSUME:
 			return DS_APPEND_CONSUME(inst, IR::ValueOpcode::DataConsume);
 		case Decoder::Opcode::DS_APPEND:
@@ -1050,6 +1068,7 @@ bool Translator::EmitMemory(const Decoder::Instruction& inst, std::string* error
 		case Decoder::Opcode::DS_READ_I16:
 		case Decoder::Opcode::DS_READ_U16:
 		case Decoder::Opcode::DS_READ_U16_D16:
+		case Decoder::Opcode::DS_READ_U16_D16_HI:
 		case Decoder::Opcode::DS_READ_B32:
 		case Decoder::Opcode::DS_READ_B64:
 		case Decoder::Opcode::DS_READ_B96:
@@ -1060,16 +1079,12 @@ bool Translator::EmitMemory(const Decoder::Instruction& inst, std::string* error
 		case Decoder::Opcode::DS_WRITE2ST64_B64: return DS_WRITE2(inst);
 		case Decoder::Opcode::DS_WRITE_B8:
 		case Decoder::Opcode::DS_WRITE_B16:
+		case Decoder::Opcode::DS_WRITE_B16_D16_HI:
 		case Decoder::Opcode::DS_WRITE_B32:
 		case Decoder::Opcode::DS_WRITE_B64:
 		case Decoder::Opcode::DS_WRITE_B96:
 		case Decoder::Opcode::DS_WRITE_B128: return DS_WRITE(inst);
-		default:
-			if (error != nullptr) {
-				*error = fmt::format("memory-family opcode has no specialized IR translation: {}",
-				                     magic_enum::enum_name(inst.opcode));
-			}
-			return false;
+		default: return false;
 	}
 }
 

src/graphics/shader/recompiler/frontend/translate/Scalar.cpp

@@ -2,14 +2,14 @@
 
 namespace Libs::Graphics::ShaderRecompiler::Frontend {
 
-bool Translator::EmitScalar(const Decoder::Instruction& inst, std::string* error) {
+bool Translator::EmitScalar(const Decoder::Instruction& inst) {
 	using O = Decoder::Opcode;
 	switch (inst.opcode) {
 		case O::S_MOV_B32:
 		case O::S_MOVK_I32: MOV_B32(inst, false); return true;
 		case O::S_MOV_B64: S_MOV_B64(inst); return true;
 		case O::S_WQM_B64: S_WQM_B64(inst); return true;
-		case O::S_GETPC_B64: return S_GETPC_B64(inst, error);
+		case O::S_GETPC_B64: S_GETPC_B64(inst); return true;
 		case O::S_SETPC_B64: return true;
 		case O::S_CSELECT_B32: S_CSELECT_B32(inst); return true;
 		case O::S_CSELECT_B64: S_CSELECT_B64(inst); return true;
@@ -34,7 +34,8 @@ bool Translator::EmitScalar(const Decoder::Instruction& inst, std::string* error
 		case O::S_ADD_U32: ADD_U32(inst, false, false); return true;
 		case O::S_ADDC_U32: ADD_U32(inst, false, true); return true;
 		case O::S_SUB_U32: SUB_U32(inst, false, false); return true;
-		case O::S_SUBB_U32: SUBB_U32(inst, false); return true;
+		case O::S_SUBB_U32: SUBB_U32(inst, false, false); return true;
+		case O::S_ABSDIFF_I32: S_ABSDIFF_I32(inst); return true;
 		case O::S_ADD_I32: S_ADD_SUB_I32(inst, false); return true;
 		case O::S_SUB_I32: S_ADD_SUB_I32(inst, true); return true;
 		case O::S_LSHL1_ADD_U32: S_LSHL_ADD_U32(inst, 1u); return true;
@@ -181,10 +182,13 @@ bool Translator::EmitScalar(const Decoder::Instruction& inst, std::string* error
 		case O::S_XNOR_B32:
 			return ComposedIntegerBinary(inst, IR::ValueOpcode::BitwiseXor32, false, true, true);
 		case O::S_FF1_I32_B64: return S_FF1_I32_B64(inst);
+		case O::S_FLBIT_I32_B32: return V_FFBH_32(inst, false);
 		case O::S_FLBIT_I32_B64: return S_FLBIT_I32_B64(inst);
 
 		case O::S_BITSET0_B32: return S_BITSET_B32(inst, false);
 		case O::S_BITSET1_B32: return S_BITSET_B32(inst, true);
+		case O::S_BITSET0_B64: return S_BITSET_B64(inst, false);
+		case O::S_BITSET1_B64: return S_BITSET_B64(inst, true);
 		case O::S_BITREPLICATE_B64_B32: return S_BITREPLICATE_B64_B32(inst);
 		case O::S_QUADMASK_B64: return S_QUADMASK_B64(inst);
 		case O::S_BFM_B32: return BFM_B32(inst);
@@ -200,6 +204,7 @@ bool Translator::EmitScalar(const Decoder::Instruction& inst, std::string* error
 
 		case O::S_NOP:
 		case O::S_SLEEP:
+		case O::S_SETPRIO:
 		case O::S_TRAP: EmitControlNop(); return true;
 		case O::S_WAITCNT_DEPCTR: EmitWaitcnt(); return true;
 		case O::S_BARRIER: S_BARRIER(); return true;

src/graphics/shader/recompiler/frontend/translate/Translate.cpp

@@ -1,16 +1,14 @@
+#include "common/assert.h"
 #include "graphics/shader/recompiler/frontend/translate/Translator.h"
 #include "graphics/shader/shader.h"
 
 #include <algorithm>
 #include <array>
-#include <fmt/format.h>
 #include <unordered_map>
 #include <utility>
 
 namespace Libs::Graphics::ShaderRecompiler::Frontend {
 
-using ShaderError::Fail;
-
 Decoder::Operand Translator::SourceAt(const Decoder::Instruction& inst, uint32_t index) {
 	switch (index) {
 		case 0: return inst.src0;
@@ -104,6 +102,7 @@ Decoder::Operand Translator::PlainOperand(const Decoder::Operand& operand) {
 	result.dpp_ctrl           = 0;
 	result.dpp_row_mask       = 0xf;
 	result.dpp_bank_mask      = 0xf;
+	result.explicit_sdwa_dst  = false;
 	result.dpp_fetch_inactive = false;
 	result.dpp_bound_ctrl     = false;
 	result.dpp                = false;
@@ -392,12 +391,12 @@ void Translator::WriteOperand(const Decoder::Operand& operand, IR::Value value)
 		}
 	}
 	if (type == IR::Type::U16) {
-		WriteRawU32(operand, IR::U32(ir.Emit(IR::ValueOpcode::ConvertU32U16, {IR::U16(value)})));
+		Write16Bits(operand, IR::U32(ir.Emit(IR::ValueOpcode::ConvertU32U16, {IR::U16(value)})));
 		return;
 	}
 	if (type == IR::Type::F16) {
 		const auto bits = IR::U16(ir.Emit(IR::ValueOpcode::BitCastU16F16, {value}));
-		WriteRawU32(operand, IR::U32(ir.Emit(IR::ValueOpcode::ConvertU32U16, {bits})));
+		Write16Bits(operand, IR::U32(ir.Emit(IR::ValueOpcode::ConvertU32U16, {bits})));
 		return;
 	}
 	if (type == IR::Type::U64) {
@@ -417,38 +416,32 @@ IR::U32 Translator::PackHalf2x16(IR::F32 low, IR::F32 high) {
 	return IR::U32(ir.Emit(IR::ValueOpcode::PackHalf2x16, {pair}));
 }
 
+void Translator::Write16Bits(const Decoder::Operand& operand, IR::U32 value) {
+	auto destination = operand;
+	// Native GFX10 16-bit results preserve the unselected half. Plain/DPP destinations use the
+	// low half; native VOP3 may already select either half. Explicit SDWA retains encoded DST_U.
+	if (!destination.explicit_sdwa_dst) {
+		if (destination.sdwa_sel == 6u) {
+			destination.sdwa_sel = 4u;
+		}
+		destination.sdwa_dst_unused = 2u;
+	}
+	destination.omod   = 0u;
+	destination.op_sel = false;
+	destination.clamp  = false;
+	WriteRawU32(destination, ir.BitwiseAnd(value, IR::U32(IR::Value(0xffffu))));
+}
+
 void Translator::WriteF16(const Decoder::Operand& operand, IR::F32 value) {
 	value           = ApplyF32ResultModifiers(operand, value);
 	const auto half = IR::F16(ir.Emit(IR::ValueOpcode::ConvertF16F32, {value}));
 	const auto bits = IR::U32(ir.Emit(IR::ValueOpcode::ConvertU32U16,
 	                                  {IR::U16(ir.Emit(IR::ValueOpcode::BitCastU16F16, {half}))}));
-	auto       merged = bits;
-	if (operand.sdwa_sel == 4u || operand.sdwa_sel == 5u) {
-		merged = IR::U32(ir.Emit(IR::ValueOpcode::BitFieldInsert,
-		                         {ReadRawU32(PlainOperand(operand)), bits,
-		                          IR::Value(operand.sdwa_sel == 5u ? 16u : 0u), IR::Value(16u)}));
-	}
-	auto raw     = operand;
-	raw.sdwa_sel = 6u;
-	raw.omod     = 0u;
-	raw.op_sel   = false;
-	raw.clamp    = false;
-	WriteRawU32(raw, merged);
+	Write16Bits(operand, bits);
 }
 
 void Translator::WriteU16(const Decoder::Operand& operand, IR::U32 value) {
-	auto merged = ir.BitwiseAnd(value, IR::U32(IR::Value(0xffffu)));
-	if (operand.sdwa_sel == 4u || operand.sdwa_sel == 5u) {
-		merged = IR::U32(ir.Emit(IR::ValueOpcode::BitFieldInsert,
-		                         {ReadRawU32(PlainOperand(operand)), merged,
-		                          IR::Value(operand.sdwa_sel == 5u ? 16u : 0u), IR::Value(16u)}));
-	}
-	auto raw     = operand;
-	raw.sdwa_sel = 6u;
-	raw.omod     = 0u;
-	raw.op_sel   = false;
-	raw.clamp    = false;
-	WriteRawU32(raw, merged);
+	Write16Bits(operand, value);
 }
 
 IR::U32 Translator::ReadU32(const Decoder::Operand& operand) {
@@ -745,11 +738,10 @@ void Translator::WriteCompareResult(const Decoder::Operand& operand, IR::U1 valu
 	WriteMask(operand, ir.LogicalAnd(ir.GetExec(), value));
 }
 
-bool Translator::AddBranchCondition(const CFG::BasicBlock& source, IR::BlockInfo& info,
-                                    std::string* error) {
+void Translator::AddBranchCondition(const CFG::BasicBlock& source, IR::BlockInfo& info) {
 	if (source.terminator.goto_value >= 0) {
 		if (source.terminator.goto_variable == UINT32_MAX) {
-			return Fail(error, fmt::format("block {} sets an invalid goto variable", source.id));
+			EXIT("block %u sets an invalid goto variable", source.id);
 		}
 		ir.SetGotoVariable(source.terminator.goto_variable,
 		                   IR::U1(IR::Value(source.terminator.goto_value != 0)));
@@ -761,13 +753,13 @@ bool Translator::AddBranchCondition(const CFG::BasicBlock& source, IR::BlockInfo
 			info.indirect_target =
 			    ir.GetScalarReg(static_cast<IR::ScalarReg>(source.terminator.indirect_pc_sgpr));
 		} else {
-			return Fail(error, fmt::format("block {} has no indirect branch selector", source.id));
+			EXIT("block %u has no indirect branch selector", source.id);
 		}
 		ir.Emit(IR::ValueOpcode::ReferenceU32, {info.indirect_target});
-		return true;
+		return;
 	}
 	if (source.terminator.kind != CFG::TerminatorKind::ConditionalBranch) {
-		return true;
+		return;
 	}
 	// EXEC and VCC are invocation-local Boolean masks. Branching on that Boolean lets inactive
 	// invocations leave the region without reconstructing a host-subgroup mask.
@@ -782,17 +774,15 @@ bool Translator::AddBranchCondition(const CFG::BasicBlock& source, IR::BlockInfo
 		case CFG::BranchCondition::ExecNonZero: condition = ir.GetExec(); break;
 		case CFG::BranchCondition::GotoVariable:
 			if (source.terminator.goto_variable == UINT32_MAX) {
-				return Fail(error,
-				            fmt::format("block {} reads an invalid goto variable", source.id));
+				EXIT("block %u reads an invalid goto variable", source.id);
 			}
 			condition = ir.GetGotoVariable(source.terminator.goto_variable);
 			break;
 		case CFG::BranchCondition::Unknown:
-			return Fail(error, fmt::format("block {} has an unknown branch condition", source.id));
+			EXIT("block %u has an unknown branch condition", source.id);
 	}
 	info.condition = condition;
 	ir.Emit(IR::ValueOpcode::Reference, {condition});
-	return true;
 }
 
 namespace {
@@ -914,47 +904,42 @@ void IncludeInstructionVectorRegisters(const Decoder::Instruction& inst, uint32_
 	}
 }
 
-bool ValidateTranslateOptions(const TranslateOptions& options, std::string* error) {
+void ValidateTranslateOptions(const TranslateOptions& options) {
 	if (options.wave_size != 32u && options.wave_size != 64u) {
-		return Fail(error, "shader translation requires wave32 or wave64");
+		EXIT("shader translation requires wave32 or wave64, got %u", options.wave_size);
 	}
 	switch (options.stage) {
 		case ShaderType::Vertex:
-			return options.vertex != nullptr ||
-			       Fail(error, "vertex shader translation has no vertex input metadata");
+			if (options.vertex == nullptr) {
+				EXIT("vertex shader translation has no vertex input metadata");
+			}
+			return;
 		case ShaderType::Pixel:
-			return options.pixel != nullptr ||
-			       Fail(error, "pixel shader translation has no pixel input metadata");
+			if (options.pixel == nullptr) {
+				EXIT("pixel shader translation has no pixel input metadata");
+			}
+			return;
 		case ShaderType::Compute:
-			return options.compute != nullptr ||
-			       Fail(error, "compute shader translation has no compute input metadata");
-		default: return Fail(error, "shader translation has an unsupported stage");
+			if (options.compute == nullptr) {
+				EXIT("compute shader translation has no compute input metadata");
+			}
+			return;
+		default:
+			EXIT("shader translation has unsupported stage %u",
+			     static_cast<uint32_t>(options.stage));
 	}
 }
 
 } // namespace
 
-bool TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
-                      const TranslateOptions& options, IR::Program& result, std::string* error) {
-	result = {};
-	struct ResetOnFailure {
-		IR::Program& program;
-		bool         complete = false;
-
-		~ResetOnFailure() {
-			if (!complete) {
-				program = {};
-			}
-		}
-	} reset_on_failure {result};
-
-	if (!ValidateTranslateOptions(options, error)) {
-		return false;
-	}
+IR::Program TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
+                             const TranslateOptions& options) {
+	ValidateTranslateOptions(options);
 	if (cfg.blocks.empty()) {
-		return Fail(error, "cannot translate an empty CFG");
+		EXIT("cannot translate an empty CFG");
 	}
 
+	IR::Program result;
 	result.stage               = options.stage;
 	result.wave_size           = options.wave_size;
 	result.shader_hash         = options.shader_hash;
@@ -965,15 +950,16 @@ bool TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
 	result.cfg_failure_kind    = options.cfg_failure_kind;
 	result.fallback_reason     = options.fallback_reason;
 	if (options.embedded_fetch != nullptr) {
-		result.info.vertex_offset_sgpr = options.embedded_fetch->vertex_offset_sgpr;
+		result.info.vertex_offset_sgpr   = options.embedded_fetch->vertex_offset_sgpr;
+		result.info.instance_offset_sgpr = options.embedded_fetch->instance_offset_sgpr;
 	}
 
 	uint32_t vector_limit = 1u;
 	for (const auto& cfg_block: cfg.blocks) {
 		for (uint32_t index = cfg_block.inst_begin; index < cfg_block.inst_end; index++) {
 			if (index >= decoded.instructions.size()) {
-				return Fail(error,
-				            "CFG block references an instruction outside the decoded program");
+				EXIT("CFG block %u references instruction %u outside decoded program of size %zu",
+				     cfg_block.id, index, decoded.instructions.size());
 			}
 			const auto& instruction = decoded.instructions[index];
 			if (IsCodeTableLoad(cfg, instruction.pc)) {
@@ -990,7 +976,7 @@ bool TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
 		                    return block.id;
 	                    })->id;
 	if (max_id == UINT32_MAX) {
-		return Fail(error, "cannot allocate a typed entry block id");
+		EXIT("cannot allocate a typed entry block id");
 	}
 	CFG::Terminator terminator;
 	terminator.kind       = CFG::TerminatorKind::Branch;
@@ -1004,7 +990,7 @@ bool TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
 	block_indices.reserve(cfg.blocks.size());
 	for (const auto& source_block: cfg.blocks) {
 		if (!block_indices.emplace(source_block.id, result.blocks.size()).second) {
-			return Fail(error, fmt::format("CFG contains duplicate block id {}", source_block.id));
+			EXIT("CFG contains duplicate block id %u", source_block.id);
 		}
 		result.block_storage.push_back(std::make_unique<IR::Block>());
 		result.blocks.push_back(result.block_storage.back().get());
@@ -1016,8 +1002,7 @@ bool TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
 		for (const auto successor: source_block.successors) {
 			const auto target = block_indices.find(successor);
 			if (target == block_indices.end()) {
-				return Fail(error, fmt::format("CFG block {} has unknown successor {}",
-				                               source_block.id, successor));
+				EXIT("CFG block %u has unknown successor %u", source_block.id, successor);
 			}
 			result.blocks[source_index]->AddBranch(result.blocks[target->second]);
 		}
@@ -1115,7 +1100,7 @@ bool TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
 		}
 	}
 	for (const auto& cfg_block: cfg.blocks) {
-		const auto         typed_index = block_indices.at(cfg_block.id);
+		const auto typed_index = block_indices.at(cfg_block.id);
 		Translator translator(result, result.blocks[typed_index], vector_limit, options.wave_size);
 		for (uint32_t index = cfg_block.inst_begin; index < cfg_block.inst_end; index++) {
 			const auto& instruction = decoded.instructions[index];
@@ -1131,44 +1116,27 @@ bool TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
 			    instruction.data_dwords == embedded->components &&
 			    instruction.dst.kind == Decoder::OperandKind::Vgpr) {
 				if (options.vertex == nullptr) {
-					return Fail(error, "embedded vertex fetch plan has no vertex input metadata");
+					EXIT("embedded vertex fetch plan has no vertex input metadata");
 				}
 				const auto resource = ResolveEmbeddedFetchResource(*options.vertex, *embedded);
 				if (resource < 0 || resource >= options.vertex->resources_num) {
-					return Fail(
-					    error,
-					    fmt::format(
-					        "embedded vertex fetch at 0x{:08x} has no resource for attribute {}",
-					        instruction.pc, embedded->attrib_id));
+					EXIT("embedded vertex fetch at 0x%08x has no resource for attribute %d",
+					     instruction.pc, embedded->attrib_id);
 				}
 				result.info.vertex_fetch_components[static_cast<size_t>(resource)] =
 				    static_cast<uint8_t>(std::max<uint32_t>(
 				        result.info.vertex_fetch_components[static_cast<size_t>(resource)],
 				        embedded->components));
-				if (!translator.TranslateEmbeddedFetch(instruction, static_cast<uint32_t>(resource),
-				                                       embedded->components)) {
-					return false;
-				}
+				translator.TranslateEmbeddedFetch(instruction, static_cast<uint32_t>(resource),
+				                                  embedded->components);
 				continue;
 			}
-			if (!translator.TranslateInstruction(instruction, error)) {
-				if (error == nullptr || error->empty()) {
-					return Fail(error, fmt::format("opcode {} at 0x{:08x} has no IR translation",
-					                               magic_enum::enum_name(instruction.opcode),
-					                               instruction.pc));
-				}
-				return false;
-			}
-		}
-		if (!translator.AddBranchCondition(cfg_block, result.block_info[typed_index], error)) {
-			return false;
+			translator.TranslateInstruction(instruction);
 		}
+		translator.AddBranchCondition(cfg_block, result.block_info[typed_index]);
 	}
-	if (!IR::ValidateProgram(result, false, error)) {
-		return false;
-	}
-	reset_on_failure.complete = true;
-	return true;
+	IR::ValidateProgram(result, false);
+	return result;
 }
 
 } // namespace Libs::Graphics::ShaderRecompiler::Frontend

src/graphics/shader/recompiler/frontend/translate/Translate.h

@@ -19,7 +19,8 @@ struct EmbeddedFetchLoad {
 
 struct EmbeddedFetchPlan {
 	std::vector<EmbeddedFetchLoad> loads;
-	int32_t                        vertex_offset_sgpr = -1;
+	int32_t                        vertex_offset_sgpr   = -1;
+	int32_t                        instance_offset_sgpr = -1;
 };
 
 struct TranslateOptions {
@@ -38,7 +39,7 @@ struct TranslateOptions {
 	const EmbeddedFetchPlan*      embedded_fetch = nullptr;
 };
 
-bool TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
-                      const TranslateOptions& options, IR::Program& result, std::string* error);
+IR::Program TranslateProgram(const Decoder::Program& decoded, const CFG::Graph& cfg,
+                             const TranslateOptions& options);
 
 } // namespace Libs::Graphics::ShaderRecompiler::Frontend

src/graphics/shader/recompiler/frontend/translate/Translator.h

@@ -13,10 +13,10 @@ public:
 	    : program(program), ir(block), current_vector_limit(vector_limit),
 	      current_wave_size(wave_size) {}
 
-	bool TranslateInstruction(const Decoder::Instruction& inst, std::string* error);
-	bool TranslateEmbeddedFetch(const Decoder::Instruction& inst, uint32_t attribute,
+	void TranslateInstruction(const Decoder::Instruction& inst);
+	void TranslateEmbeddedFetch(const Decoder::Instruction& inst, uint32_t attribute,
 	                            uint32_t component_count);
-	bool AddBranchCondition(const CFG::BasicBlock& source, IR::BlockInfo& info, std::string* error);
+	void AddBranchCondition(const CFG::BasicBlock& source, IR::BlockInfo& info);
 
 private:
 	Decoder::Operand SourceAt(const Decoder::Instruction& inst, uint32_t index);
@@ -34,6 +34,7 @@ private:
 	IR::F32                ApplyF32ResultModifiers(const Decoder::Operand& operand, IR::F32 value);
 	void                   WriteOperand(const Decoder::Operand& operand, IR::Value value);
 	IR::U32                PackHalf2x16(IR::F32 low, IR::F32 high);
+	void                   Write16Bits(const Decoder::Operand& operand, IR::U32 value);
 	void                   WriteF16(const Decoder::Operand& operand, IR::F32 value);
 	void                   WriteU16(const Decoder::Operand& operand, IR::U32 value);
 	IR::U32                ReadU32(const Decoder::Operand& operand);
@@ -109,6 +110,7 @@ private:
 	bool DS_APPEND_CONSUME(const Decoder::Instruction& inst, IR::ValueOpcode opcode);
 	bool DS_ADDTID(const Decoder::Instruction& inst, bool write);
 	bool DS_SWIZZLE_B32(const Decoder::Instruction& inst);
+	bool DS_BPERMUTE_B32(const Decoder::Instruction& inst);
 
 	IR::F32 SelectF32(IR::U1 condition, IR::F32 true_value, IR::F32 false_value);
 	IR::U32 ConvertF32ToU32Saturated(IR::F32 value, float upper_bound, float safe_upper,
@@ -126,7 +128,7 @@ private:
 	void EmitFloatCompare(const Decoder::Instruction& inst, IR::ValueOpcode opcode, bool half,
 	                      bool cmpx);
 	void EmitFloatOrderedCompare(const Decoder::Instruction& inst, bool ordered);
-	void V_CMP_CLASS_F32(const Decoder::Instruction& inst);
+	void EmitFloatClassCompare(const Decoder::Instruction& inst, bool cmpx);
 	void V_CVT_F32_UBYTE(const Decoder::Instruction& inst, uint32_t byte_index);
 	void V_CVT_F32_U32(const Decoder::Instruction& inst);
 	void V_CVT_F32_I32(const Decoder::Instruction& inst);
@@ -148,6 +150,7 @@ private:
 	                   bool quiet_snan);
 	bool Float16Unary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
 	                  bool invalid_negative);
+	bool Float16Trig(const Decoder::Instruction& inst, IR::ValueOpcode opcode);
 	bool Float16Binary(const Decoder::Instruction& inst, IR::ValueOpcode opcode, bool reverse);
 	bool Float16Ternary(const Decoder::Instruction& inst, IR::ValueOpcode opcode, bool accumulator,
 	                    bool mix);
@@ -172,28 +175,29 @@ private:
 	bool PackedInteger16MinMax(const Decoder::Instruction& inst, IR::ValueOpcode opcode, bool sign);
 	bool S_U64_MASK(const Decoder::Instruction& inst, IR::ValueOpcode logical_opcode,
 	                IR::ValueOpcode bit_opcode, bool negate_rhs, bool negate_result, bool unary);
-	IR::U1  U64MaskBinary(const Decoder::Instruction& inst, IR::ValueOpcode opcode, bool negate_rhs,
-	                      bool negate_result);
-	bool    SimpleInteger(const Decoder::Instruction& inst, IR::ValueOpcode opcode, IR::Type type,
-	                      bool reverse, bool mask_shift_count, bool update_scc);
-	bool    ComposedIntegerBinary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
-	                              bool negate_rhs, bool negate_result, bool update_scc);
-	bool    V_AND_OR_B32(const Decoder::Instruction& inst);
-	bool    V_OR3_B32(const Decoder::Instruction& inst);
-	bool    V_XOR3_B32(const Decoder::Instruction& inst);
-	bool    S_FF1_I32_B64(const Decoder::Instruction& inst);
-	bool    V_FFBH_U32(const Decoder::Instruction& inst);
-	bool    S_FLBIT_I32_B64(const Decoder::Instruction& inst);
-	bool    Integer24(const Decoder::Instruction& inst, bool sign, bool addend);
-	bool    V_MAD_U64_U32(const Decoder::Instruction& inst);
-	bool    V_SAD_U32(const Decoder::Instruction& inst);
-	bool    V_ADD3_U32(const Decoder::Instruction& inst);
-	bool    S_BITSET_B32(const Decoder::Instruction& inst, bool set);
-	bool    V_BCNT_U32_B32(const Decoder::Instruction& inst);
-	bool    V_MBCNT_U32_B32(const Decoder::Instruction& inst, bool low);
-	bool    S_BITREPLICATE_B64_B32(const Decoder::Instruction& inst);
-	bool    S_QUADMASK_B64(const Decoder::Instruction& inst);
-	bool    BFM_B32(const Decoder::Instruction& inst);
+	IR::U1 U64MaskBinary(const Decoder::Instruction& inst, IR::ValueOpcode opcode, bool negate_rhs,
+	                     bool negate_result);
+	bool SimpleInteger(const Decoder::Instruction& inst, IR::ValueOpcode opcode, IR::Type type,
+	                   bool reverse, bool mask_shift_count, bool update_scc);
+	bool ComposedIntegerBinary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+	                           bool negate_rhs, bool negate_result, bool update_scc);
+	bool V_AND_OR_B32(const Decoder::Instruction& inst);
+	bool V_OR3_B32(const Decoder::Instruction& inst);
+	bool V_XOR3_B32(const Decoder::Instruction& inst);
+	bool S_FF1_I32_B64(const Decoder::Instruction& inst);
+	bool V_FFBH_32(const Decoder::Instruction& inst, bool sign);
+	bool S_FLBIT_I32_B64(const Decoder::Instruction& inst);
+	bool Integer24(const Decoder::Instruction& inst, bool sign, bool addend);
+	bool V_MAD_U64_U32(const Decoder::Instruction& inst);
+	bool V_SAD_U32(const Decoder::Instruction& inst);
+	bool V_ADD3_U32(const Decoder::Instruction& inst);
+	bool S_BITSET_B32(const Decoder::Instruction& inst, bool set);
+	bool S_BITSET_B64(const Decoder::Instruction& inst, bool set);
+	bool V_BCNT_U32_B32(const Decoder::Instruction& inst);
+	bool V_MBCNT_U32_B32(const Decoder::Instruction& inst, bool low);
+	bool S_BITREPLICATE_B64_B32(const Decoder::Instruction& inst);
+	bool S_QUADMASK_B64(const Decoder::Instruction& inst);
+	bool BFM_B32(const Decoder::Instruction& inst);
 	IR::U32 RightMask32(IR::U32 count);
 	IR::U64 RightMask64(IR::U32 count);
 	bool    S_BFM_B64(const Decoder::Instruction& inst);
@@ -215,7 +219,8 @@ private:
 	                bool negate_source, bool write_64);
 	void ADD_U32(const Decoder::Instruction& inst, bool vector, bool use_carry_in);
 	void SUB_U32(const Decoder::Instruction& inst, bool vector, bool reverse);
-	void SUBB_U32(const Decoder::Instruction& inst, bool vector);
+	void SUBB_U32(const Decoder::Instruction& inst, bool vector, bool reverse);
+	void S_ABSDIFF_I32(const Decoder::Instruction& inst);
 	void S_ADD_SUB_I32(const Decoder::Instruction& inst, bool subtract);
 	void S_LSHL_ADD_U32(const Decoder::Instruction& inst, uint32_t shift_amount);
 	void ScalarMinMax32(const Decoder::Instruction& inst, IR::ValueOpcode value_opcode,
@@ -226,27 +231,27 @@ private:
 	void S_SENDMSG();
 	void S_TTRACEDATA();
 	void S_INST_PREFETCH();
-	bool S_GETPC_B64(const Decoder::Instruction& inst, std::string* error);
+	void S_GETPC_B64(const Decoder::Instruction& inst);
 	void S_CSELECT_B32(const Decoder::Instruction& inst);
 	void S_CSELECT_B64(const Decoder::Instruction& inst);
 	void MOV_B32(const Decoder::Instruction& inst, bool apply_float_modifiers);
 	void S_MOV_B64(const Decoder::Instruction& inst);
 	void S_WQM_B64(const Decoder::Instruction& inst);
-	bool V_MOVRELS_B32(const Decoder::Instruction& inst, std::string* error);
-	bool V_MOVRELD_B32(const Decoder::Instruction& inst, std::string* error);
+	void V_MOVRELS_B32(const Decoder::Instruction& inst);
+	void V_MOVRELD_B32(const Decoder::Instruction& inst);
 	void V_READFIRSTLANE_B32(const Decoder::Instruction& inst);
 	void V_READLANE_B32(const Decoder::Instruction& inst);
 	void V_WRITELANE_B32(const Decoder::Instruction& inst);
 	void V_PERMLANE16_B32(const Decoder::Instruction& inst, bool x16);
 	void V_INTERP_P1_F32();
 	void V_INTERP_P2_F32(const Decoder::Instruction& inst);
-	bool V_INTERP_MOV_F32(const Decoder::Instruction& inst, std::string* error);
-	bool EXP(const Decoder::Instruction& inst, std::string* error);
+	void V_INTERP_MOV_F32(const Decoder::Instruction& inst);
+	void EXP(const Decoder::Instruction& inst);
 
-	bool EmitScalar(const Decoder::Instruction& inst, std::string* error);
-	bool EmitVector(const Decoder::Instruction& inst, std::string* error);
-	bool EmitInterpolation(const Decoder::Instruction& inst, std::string* error);
-	bool EmitMemory(const Decoder::Instruction& inst, std::string* error);
+	bool EmitScalar(const Decoder::Instruction& inst);
+	bool EmitVector(const Decoder::Instruction& inst);
+	bool EmitInterpolation(const Decoder::Instruction& inst);
+	bool EmitMemory(const Decoder::Instruction& inst);
 
 	IR::Program&    program;
 	IR::IREmitter   ir;

src/graphics/shader/recompiler/frontend/translate/Vector.cpp

@@ -2,7 +2,7 @@
 
 namespace Libs::Graphics::ShaderRecompiler::Frontend {
 
-bool Translator::EmitVector(const Decoder::Instruction& inst, std::string* error) {
+bool Translator::EmitVector(const Decoder::Instruction& inst) {
 	using O = Decoder::Opcode;
 	switch (inst.opcode) {
 		case O::V_NOP: return true;
@@ -10,11 +10,12 @@ bool Translator::EmitVector(const Decoder::Instruction& inst, std::string* error
 		case O::V_ADDC_U32: ADD_U32(inst, true, true); return true;
 		case O::V_SUB_I32: SUB_U32(inst, true, false); return true;
 		case O::V_SUBREV_I32: SUB_U32(inst, true, true); return true;
-		case O::V_SUBREV_CO_CI_U32: SUBB_U32(inst, true); return true;
+		case O::V_SUB_CO_CI_U32: SUBB_U32(inst, true, false); return true;
+		case O::V_SUBREV_CO_CI_U32: SUBB_U32(inst, true, true); return true;
 
 		case O::V_MOV_B32: MOV_B32(inst, true); return true;
-		case O::V_MOVRELS_B32: return V_MOVRELS_B32(inst, error);
-		case O::V_MOVRELD_B32: return V_MOVRELD_B32(inst, error);
+		case O::V_MOVRELS_B32: V_MOVRELS_B32(inst); return true;
+		case O::V_MOVRELD_B32: V_MOVRELD_B32(inst); return true;
 		case O::V_READFIRSTLANE_B32: V_READFIRSTLANE_B32(inst); return true;
 		case O::V_READLANE_B32: V_READLANE_B32(inst); return true;
 		case O::V_WRITELANE_B32: V_WRITELANE_B32(inst); return true;
@@ -96,8 +97,12 @@ bool Translator::EmitVector(const Decoder::Instruction& inst, std::string* error
 			EmitIntegerCompare(inst, IR::ValueOpcode::SLessThanEqual32, IR::Type::U32, false, true);
 			return true;
 		case O::V_CMP_EQ_I64:
+		case O::V_CMP_EQ_U64:
 			EmitIntegerCompare(inst, IR::ValueOpcode::IEqual64, IR::Type::U64, false, false);
 			return true;
+		case O::V_CMP_LT_U64:
+			EmitIntegerCompare(inst, IR::ValueOpcode::ULessThan64, IR::Type::U64, false, false);
+			return true;
 		case O::V_CMP_GT_U64:
 			EmitIntegerCompare(inst, IR::ValueOpcode::UGreaterThan64, IR::Type::U64, false, false);
 			return true;
@@ -270,7 +275,8 @@ bool Translator::EmitVector(const Decoder::Instruction& inst, std::string* error
 			return true;
 		case O::V_CMP_O_F32: EmitFloatOrderedCompare(inst, true); return true;
 		case O::V_CMP_U_F32: EmitFloatOrderedCompare(inst, false); return true;
-		case O::V_CMP_CLASS_F32: V_CMP_CLASS_F32(inst); return true;
+		case O::V_CMP_CLASS_F32: EmitFloatClassCompare(inst, false); return true;
+		case O::V_CMPX_CLASS_F32: EmitFloatClassCompare(inst, true); return true;
 
 		case O::V_CVT_F32_UBYTE0: V_CVT_F32_UBYTE(inst, 0); return true;
 		case O::V_CVT_F32_UBYTE1: V_CVT_F32_UBYTE(inst, 1); return true;
@@ -364,6 +370,8 @@ bool Translator::EmitVector(const Decoder::Instruction& inst, std::string* error
 		case O::V_CEIL_F16: return Float16Unary(inst, IR::ValueOpcode::FPCeil32, false);
 		case O::V_TRUNC_F16: return Float16Unary(inst, IR::ValueOpcode::FPTrunc32, false);
 		case O::V_RNDNE_F16: return Float16Unary(inst, IR::ValueOpcode::FPRoundEven32, false);
+		case O::V_SIN_F16: return Float16Trig(inst, IR::ValueOpcode::FPSin);
+		case O::V_COS_F16: return Float16Trig(inst, IR::ValueOpcode::FPCos);
 		case O::V_MIN3_F16: return Float16Ternary(inst, IR::ValueOpcode::FPMinTri32, false, false);
 		case O::V_MAX3_F16: return Float16Ternary(inst, IR::ValueOpcode::FPMaxTri32, false, false);
 		case O::V_MED3_F16: return Float16Ternary(inst, IR::ValueOpcode::FPMedTri32, false, false);
@@ -490,7 +498,8 @@ bool Translator::EmitVector(const Decoder::Instruction& inst, std::string* error
 		case O::V_AND_OR_B32: return V_AND_OR_B32(inst);
 		case O::V_OR3_B32: return V_OR3_B32(inst);
 		case O::V_XOR3_B32: return V_XOR3_B32(inst);
-		case O::V_FFBH_U32: return V_FFBH_U32(inst);
+		case O::V_FFBH_U32: return V_FFBH_32(inst, false);
+		case O::V_FFBH_I32: return V_FFBH_32(inst, true);
 
 		case O::V_MAD_I32_I24: return Integer24(inst, true, true);
 		case O::V_MAD_U32_U24: return Integer24(inst, false, true);

src/graphics/shader/recompiler/ir/Program.cpp

@@ -1,7 +1,6 @@
+#include "common/assert.h"
 #include "graphics/shader/recompiler/ir/ShaderIR.h"
 
-#include "graphics/shader/shader.h"
-
 #include <fmt/format.h>
 #include <map>
 #include <new>
@@ -11,10 +10,13 @@
 
 namespace Libs::Graphics::ShaderRecompiler::IR {
 
-using ShaderError::Fail;
-
 namespace {
 
+[[noreturn]] void Fail(std::string_view message) {
+	EXIT("shader IR validation failed: %s", std::string(message).c_str());
+	std::abort();
+}
+
 bool IsRegisterStatePseudo(ValueOpcode opcode) {
 	switch (opcode) {
 		case ValueOpcode::GetThreadBitScalarRegister:
@@ -52,7 +54,7 @@ bool IsRuntimeRead(ValueOpcode opcode) {
 	       AddressOpcodeInfoOf(opcode).access == AddressAccess::Read;
 }
 
-bool EquivalentValue(const Program& program, Value left, Value right,
+bool EquivalentValue(const ResourcePlan& program, Value left, Value right,
                      std::vector<std::pair<const Inst*, const Inst*>>& visited) {
 	left  = left.Resolve();
 	right = right.Resolve();
@@ -95,6 +97,20 @@ bool EquivalentValue(const Program& program, Value left, Value right,
 
 } // namespace
 
+ResourcePlan::~ResourcePlan() {
+	for (auto& inst: value_storage) {
+		inst.Invalidate();
+	}
+}
+
+ResourcePlan& ResourcePlan::operator=(ResourcePlan&& other) noexcept {
+	if (this != &other) {
+		this->~ResourcePlan();
+		new (this) ResourcePlan(std::move(other));
+	}
+	return *this;
+}
+
 Program::~Program() {
 	// Values may cross block boundaries. Detach all arguments before any block starts destroying
 	// its instruction storage so reverse-use links always point to live definitions.
@@ -113,12 +129,31 @@ Program& Program::operator=(Program&& other) noexcept {
 	return *this;
 }
 
-bool EquivalentValue(const Program& program, Value left, Value right) {
+CompiledShaderInfo Program::TakeCompiledInfo() && {
+	CompiledShaderInfo result {
+	    .stage           = stage,
+	    .shader_hash     = shader_hash,
+	    .wave_size       = wave_size,
+	    .user_data_base  = user_data_base,
+	    .user_data_count = user_data_count,
+	    .scratch_dwords  = scratch_dwords,
+	    .info            = std::move(info),
+	    .bindings        = std::move(bindings),
+	};
+	for (const auto& output: result.info.outputs) {
+		if (output.kind == StageOutputKind::Parameter && output.index < 32) {
+			result.param_export_mask |= 1u << output.index;
+		}
+	}
+	return result;
+}
+
+bool EquivalentValue(const ResourcePlan& program, Value left, Value right) {
 	std::vector<std::pair<const Inst*, const Inst*>> visited;
 	return EquivalentValue(program, left, right, visited);
 }
 
-Value ResolveInvariantPhi(const Program& program, Value value) {
+Value ResolveInvariantPhi(const ResourcePlan& program, Value value) {
 	value            = value.Resolve();
 	const auto* root = value.TryInstruction();
 	if (root == nullptr || root->GetOpcode() != ValueOpcode::Phi) {
@@ -149,10 +184,10 @@ Value ResolveInvariantPhi(const Program& program, Value value) {
 	return invariant;
 }
 
-bool ValidateProgram(const Program& program, bool require_ssa, std::string* error) {
+void ValidateProgram(const Program& program, bool require_ssa) {
 	if (program.blocks.size() != program.block_info.size() ||
 	    program.blocks.size() != program.block_storage.size()) {
-		return Fail(error, "value IR block storage is inconsistent");
+		return Fail("value IR block storage is inconsistent");
 	}
 	std::unordered_set<const Block*>           blocks;
 	std::unordered_map<const Block*, size_t>   block_indices;
@@ -162,26 +197,26 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 		const auto* block = program.blocks[block_index];
 		if (block == nullptr || program.block_storage[block_index] == nullptr ||
 		    block != program.block_storage[block_index].get()) {
-			return Fail(error, "value IR block pointer is inconsistent");
+			return Fail("value IR block pointer is inconsistent");
 		}
 		if (!blocks.insert(block).second) {
-			return Fail(error, "value IR block pointer is duplicated");
+			return Fail("value IR block pointer is duplicated");
 		}
 		block_indices.emplace(block, block_index);
 		if (program.block_info[block_index].id == UINT32_MAX) {
-			return Fail(error, "value IR block uses the reserved exit id");
+			return Fail("value IR block uses the reserved exit id");
 		}
 		if (!blocks_by_id.emplace(program.block_info[block_index].id, block).second) {
-			return Fail(error, "value IR block id is duplicated");
+			return Fail("value IR block id is duplicated");
 		}
 		for (const auto& inst: *block) {
 			if (!instructions.insert(&inst).second) {
-				return Fail(error, "value IR instruction is duplicated");
+				return Fail("value IR instruction is duplicated");
 			}
 		}
 	}
 	if (!program.blocks.empty() && !program.blocks.front()->ImmPredecessors().empty()) {
-		return Fail(error, "value IR entry block has a predecessor");
+		return Fail("value IR entry block has a predecessor");
 	}
 
 	for (size_t block_index = 0; block_index < program.blocks.size(); block_index++) {
@@ -189,28 +224,28 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 		std::unordered_set<const Block*> predecessors;
 		for (const auto* predecessor: block->ImmPredecessors()) {
 			if (predecessor == nullptr || !blocks.contains(predecessor)) {
-				return Fail(error, "value IR block has a foreign predecessor");
+				return Fail("value IR block has a foreign predecessor");
 			}
 			if (!predecessors.insert(predecessor).second) {
-				return Fail(error, "value IR block predecessor is duplicated");
+				return Fail("value IR block predecessor is duplicated");
 			}
 			if (std::ranges::find(predecessor->ImmSuccessors(), block) ==
 			    predecessor->ImmSuccessors().end()) {
-				return Fail(error, "value IR predecessor edge is not reciprocal");
+				return Fail("value IR predecessor edge is not reciprocal");
 			}
 		}
 
 		std::unordered_set<const Block*> successors;
 		for (const auto* successor: block->ImmSuccessors()) {
 			if (successor == nullptr || !blocks.contains(successor)) {
-				return Fail(error, "value IR block has a foreign successor");
+				return Fail("value IR block has a foreign successor");
 			}
 			if (!successors.insert(successor).second) {
-				return Fail(error, "value IR block successor is duplicated");
+				return Fail("value IR block successor is duplicated");
 			}
 			if (std::ranges::find(successor->ImmPredecessors(), block) ==
 			    successor->ImmPredecessors().end()) {
-				return Fail(error, "value IR successor edge is not reciprocal");
+				return Fail("value IR successor edge is not reciprocal");
 			}
 		}
 
@@ -234,41 +269,40 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 		switch (terminator.kind) {
 			case CFG::TerminatorKind::Branch:
 				if (!add_target(terminator.true_block)) {
-					return Fail(error, "value IR branch target is missing");
+					return Fail("value IR branch target is missing");
 				}
 				break;
 			case CFG::TerminatorKind::ConditionalBranch:
 				if (!add_target(terminator.true_block) || !add_target(terminator.false_block)) {
-					return Fail(error, "value IR conditional branch target is missing");
+					return Fail("value IR conditional branch target is missing");
 				}
 				if (!validate_control_value(program.block_info[block_index].condition, Type::U1)) {
-					return Fail(error, "value IR conditional branch condition is invalid");
+					return Fail("value IR conditional branch condition is invalid");
 				}
 				break;
 			case CFG::TerminatorKind::IndirectBranch: {
 				if (!validate_control_value(program.block_info[block_index].indirect_target,
 				                            Type::U32)) {
-					return Fail(error, "value IR indirect branch selector is invalid");
+					return Fail("value IR indirect branch selector is invalid");
 				}
 				std::unordered_set<uint32_t> indirect_targets;
 				for (const auto target: terminator.indirect_targets) {
 					if (!indirect_targets.insert(target).second) {
-						return Fail(error, "value IR indirect branch target is duplicated");
+						return Fail("value IR indirect branch target is duplicated");
 					}
 					if (!add_target(target)) {
-						return Fail(error, "value IR indirect branch target is missing");
+						return Fail("value IR indirect branch target is missing");
 					}
 				}
 				if (terminator.indirect_selector_values.size() !=
 				    terminator.indirect_selector_targets.size()) {
-					return Fail(error, "value IR indirect selector table is inconsistent");
+					return Fail("value IR indirect selector table is inconsistent");
 				}
 				for (const auto target: terminator.indirect_selector_targets) {
 					const auto found = blocks_by_id.find(target);
 					if (found == blocks_by_id.end() ||
 					    !expected_successors.contains(found->second)) {
-						return Fail(error,
-						            "value IR indirect selector target is not a CFG successor");
+						return Fail("value IR indirect selector target is not a CFG successor");
 					}
 				}
 				break;
@@ -280,73 +314,73 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 		     !blocks_by_id.contains(terminator.merge_block)) ||
 		    (terminator.continue_block != UINT32_MAX &&
 		     !blocks_by_id.contains(terminator.continue_block))) {
-			return Fail(error, "value IR structured control target is missing");
+			return Fail("value IR structured control target is missing");
 		}
 		if (successors != expected_successors) {
-			return Fail(error, "value IR terminator and successor graph disagree");
+			return Fail("value IR terminator and successor graph disagree");
 		}
 
 		bool saw_non_phi = false;
 		for (const auto& inst: *block) {
 			if (inst.Parent() != block) {
-				return Fail(error, "value IR instruction has the wrong parent block");
+				return Fail("value IR instruction has the wrong parent block");
 			}
 			if (inst.GetOpcode() == ValueOpcode::Phi) {
 				if (saw_non_phi) {
-					return Fail(error, "value IR Phi appears after a non-Phi instruction");
+					return Fail("value IR Phi appears after a non-Phi instruction");
 				}
 				if (inst.NumPhiBlocks() != inst.NumArgs()) {
-					return Fail(error, "value IR Phi parent table is inconsistent");
+					return Fail("value IR Phi parent table is inconsistent");
 				}
 				if (inst.NumArgs() == 0) {
-					return Fail(error, "value IR Phi has no incoming values");
+					return Fail("value IR Phi has no incoming values");
 				}
 				std::unordered_set<const Block*> incoming_blocks;
 				for (size_t arg_index = 0; arg_index < inst.NumArgs(); arg_index++) {
 					const auto* predecessor = inst.PhiBlock(arg_index);
 					if (predecessor == nullptr || !blocks.contains(predecessor) ||
 					    !predecessors.contains(predecessor)) {
-						return Fail(error, "value IR Phi has a foreign or non-predecessor parent");
+						return Fail("value IR Phi has a foreign or non-predecessor parent");
 					}
 					if (!incoming_blocks.insert(predecessor).second) {
-						return Fail(error, "value IR Phi parent is duplicated");
+						return Fail("value IR Phi parent is duplicated");
 					}
 					if (inst.Arg(arg_index).GetType() != inst.GetType()) {
-						return Fail(error, "value IR Phi incoming type does not match its result");
+						return Fail("value IR Phi incoming type does not match its result");
 					}
 				}
 				if (incoming_blocks != predecessors) {
-					return Fail(error, "value IR Phi does not cover every predecessor");
+					return Fail("value IR Phi does not cover every predecessor");
 				}
 			} else {
 				saw_non_phi = true;
 			}
 			if (require_ssa && IsRegisterStatePseudo(inst.GetOpcode())) {
-				return Fail(error, fmt::format("register-state pseudo {} survived SSA rewrite",
-				                               ValueOpcodeName(inst.GetOpcode())));
+				return Fail(fmt::format("register-state pseudo {} survived SSA rewrite",
+				                        ValueOpcodeName(inst.GetOpcode())));
 			}
 			if (inst.GetOpcode() != ValueOpcode::Phi && inst.GetOpcode() != ValueOpcode::Identity &&
 			    inst.GetType() == Type::Opaque) {
-				return Fail(error, fmt::format("untyped opcode {} survived translation",
-				                               ValueOpcodeName(inst.GetOpcode())));
+				return Fail(fmt::format("untyped opcode {} survived translation",
+				                        ValueOpcodeName(inst.GetOpcode())));
 			}
 			const bool fixed_signature =
 			    inst.GetOpcode() != ValueOpcode::Phi && inst.GetOpcode() != ValueOpcode::Identity;
 			if (fixed_signature && inst.NumArgs() != NumArgsOf(inst.GetOpcode())) {
-				return Fail(error, fmt::format("{} has {} arguments, expected {}",
-				                               ValueOpcodeName(inst.GetOpcode()), inst.NumArgs(),
-				                               NumArgsOf(inst.GetOpcode())));
+				return Fail(fmt::format("{} has {} arguments, expected {}",
+				                        ValueOpcodeName(inst.GetOpcode()), inst.NumArgs(),
+				                        NumArgsOf(inst.GetOpcode())));
 			}
 			if (inst.GetOpcode() == ValueOpcode::ReadConstBuffer) {
 				const auto memory_index = inst.Flags<MemoryFlags>().index;
 				if (memory_index >= program.memory_info.size()) {
-					return Fail(error, fmt::format("{} has an invalid memory-info index",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has an invalid memory-info index",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				const auto& memory = program.memory_info[memory_index];
 				if (memory.kind != ResourceKind::ScalarBuffer) {
-					return Fail(error, fmt::format("{} has an invalid scalar-memory resource kind",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has an invalid scalar-memory resource kind",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				const bool valid_group_width =
 				    memory.component_count == 1u || memory.component_count == 2u ||
@@ -354,23 +388,23 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 				    memory.component_count == 16u;
 				if (memory.data_bits != 32u || memory.data_dwords != 1u || !valid_group_width ||
 				    memory.component_index >= memory.component_count) {
-					return Fail(error, fmt::format("{} has inconsistent scalar-memory metadata",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has inconsistent scalar-memory metadata",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 			}
 			const auto address_info = AddressOpcodeInfoOf(inst.GetOpcode());
 			if (address_info.access != AddressAccess::None) {
 				const auto memory_index = inst.Flags<MemoryFlags>().index;
 				if (memory_index >= program.memory_info.size()) {
-					return Fail(error, fmt::format("{} has an invalid memory-info index",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has an invalid memory-info index",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				const auto& memory = program.memory_info[memory_index];
 				if (!IsAddressResourceKind(memory.kind) ||
 				    (memory.kind == ResourceKind::ScalarAddress &&
 				     inst.GetOpcode() != ValueOpcode::LoadAddressU32)) {
-					return Fail(error, fmt::format("{} has an invalid address resource kind",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has an invalid address resource kind",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				const bool scalar_address = memory.kind == ResourceKind::ScalarAddress;
 				const bool valid_group_width =
@@ -382,50 +416,50 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 				if (memory.data_bits != address_info.data_bits || memory.data_dwords != 1u ||
 				    !valid_group_width || memory.component_index >= memory.component_count ||
 				    memory.sampler != 0u) {
-					return Fail(error, fmt::format("{} has inconsistent address-memory metadata",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has inconsistent address-memory metadata",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 			}
 			const auto buffer_components = BufferComponentCount(inst.GetOpcode());
 			if (buffer_components != 0u) {
 				const auto memory_index = inst.Flags<MemoryFlags>().index;
 				if (memory_index >= program.memory_info.size()) {
-					return Fail(error, fmt::format("{} has an invalid memory-info index",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has an invalid memory-info index",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				const auto& memory = program.memory_info[memory_index];
 				if (memory.kind != ResourceKind::Buffer &&
 				    memory.kind != ResourceKind::ScalarBuffer) {
-					return Fail(error, fmt::format("{} has a non-buffer resource kind",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has a non-buffer resource kind",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				if (buffer_components > 1u &&
 				    (memory.kind != ResourceKind::Buffer || memory.data_bits != 32u ||
 				     memory.data_dwords != buffer_components || memory.component_index != 0u)) {
-					return Fail(error, fmt::format("{} has inconsistent native-wide metadata",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has inconsistent native-wide metadata",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				if (buffer_components == 1u &&
 				    (inst.GetOpcode() == ValueOpcode::LoadBufferU32 ||
 				     inst.GetOpcode() == ValueOpcode::StoreBufferU32) &&
 				    memory.data_dwords != 1u) {
-					return Fail(error, fmt::format("{} retains scalar-sibling width metadata",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} retains scalar-sibling width metadata",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 			}
 			const auto shared_components = SharedComponentCount(inst.GetOpcode());
 			if (shared_components != 0u) {
 				const auto memory_index = inst.Flags<MemoryFlags>().index;
 				if (memory_index >= program.memory_info.size()) {
-					return Fail(error, fmt::format("{} has an invalid memory-info index",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has an invalid memory-info index",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				const auto& memory = program.memory_info[memory_index];
 				if ((memory.kind != ResourceKind::Lds && memory.kind != ResourceKind::Gds) ||
 				    memory.resource != 0u || memory.sampler != 0u || memory.component_count == 0u ||
 				    memory.component_index >= memory.component_count) {
-					return Fail(error, fmt::format("{} has invalid shared-memory metadata",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has invalid shared-memory metadata",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				uint32_t expected_bits = 32u;
 				if (inst.GetOpcode() == ValueOpcode::LoadSharedU8 ||
@@ -437,26 +471,27 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 				}
 				if (memory.data_bits != expected_bits || memory.data_dwords != shared_components ||
 				    (shared_components > 1u && memory.component_index != 0u)) {
-					return Fail(error, fmt::format("{} has inconsistent shared-memory width",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has inconsistent shared-memory width",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 			}
 			const auto image_info = ImageOpcodeInfoOf(inst.GetOpcode());
 			if (image_info.access != ImageAccess::None) {
 				const auto memory_index = inst.Flags<MemoryFlags>().index;
 				if (memory_index >= program.memory_info.size()) {
-					return Fail(error, fmt::format("{} has an invalid memory-info index",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(fmt::format("{} has an invalid memory-info index",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 				const auto& memory = program.memory_info[memory_index];
-				if (!ImageResourceKindMatches(memory.kind, image_info.resource_class)) {
-					return Fail(error, fmt::format("{} has invalid image-memory metadata",
-					                               ValueOpcodeName(inst.GetOpcode())));
+				if (memory.kind != ResourceKind::Image ||
+				    image_info.resource_class == ImageResourceClass::None) {
+					return Fail(fmt::format("{} has invalid image-memory metadata",
+					                        ValueOpcodeName(inst.GetOpcode())));
 				}
 			}
 			if (inst.GetOpcode() == ValueOpcode::SetAttribute &&
 			    inst.Flags<ExportFlags>().index >= program.export_info.size()) {
-				return Fail(error, "SetAttribute has an invalid export-info index");
+				return Fail("SetAttribute has an invalid export-info index");
 			}
 			uint32_t composite_components = 0u;
 			switch (inst.GetOpcode()) {
@@ -469,25 +504,24 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 			if (composite_components != 0u &&
 			    (!inst.Arg(1).IsImmediate() || inst.Arg(1).GetType() != Type::U32 ||
 			     inst.Arg(1).U32() >= composite_components)) {
-				return Fail(error, fmt::format("{} has an invalid component index",
-				                               ValueOpcodeName(inst.GetOpcode())));
+				return Fail(fmt::format("{} has an invalid component index",
+				                        ValueOpcodeName(inst.GetOpcode())));
 			}
 			for (size_t arg_index = 0; arg_index < inst.NumArgs(); arg_index++) {
 				const auto arg = inst.Arg(arg_index);
 				if (arg.IsEmpty()) {
-					return Fail(error, fmt::format("{} has an empty argument",
-					                               ValueOpcodeName(inst.GetOpcode())));
+					return Fail(
+					    fmt::format("{} has an empty argument", ValueOpcodeName(inst.GetOpcode())));
 				}
 				if (fixed_signature && arg.GetType() != ArgTypeOf(inst.GetOpcode(), arg_index)) {
-					return Fail(error,
-					            fmt::format("{} argument {} has type {}, expected {}",
+					return Fail(fmt::format("{} argument {} has type {}, expected {}",
 					                        ValueOpcodeName(inst.GetOpcode()), arg_index,
 					                        TypeName(arg.GetType()),
 					                        TypeName(ArgTypeOf(inst.GetOpcode(), arg_index))));
 				}
 				if (const auto* definition = arg.TryInstruction();
 				    definition != nullptr && !instructions.contains(definition)) {
-					return Fail(error, "value IR argument has a foreign definition");
+					return Fail("value IR argument has a foreign definition");
 				}
 				if (const auto* definition = arg.TryInstruction(); definition != nullptr) {
 					const auto& uses = definition->Uses();
@@ -495,8 +529,7 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 						return candidate.user == &inst && candidate.operand == arg_index;
 					});
 					if (use == uses.end()) {
-						return Fail(error,
-						            fmt::format("{} argument {} is absent from {} reverse uses",
+						return Fail(fmt::format("{} argument {} is absent from {} reverse uses",
 						                        ValueOpcodeName(inst.GetOpcode()), arg_index,
 						                        ValueOpcodeName(definition->GetOpcode())));
 					}
@@ -506,7 +539,7 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 	}
 
 	if (program.blocks.empty()) {
-		return true;
+		return;
 	}
 
 	std::vector<bool>   reachable(program.blocks.size(), false);
@@ -523,7 +556,7 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 		}
 	}
 	if (!std::ranges::all_of(reachable, [](bool value) { return value; })) {
-		return Fail(error, "value IR contains an unreachable block");
+		return Fail("value IR contains an unreachable block");
 	}
 
 	std::vector<std::vector<bool>> dominators(program.blocks.size(),
@@ -577,28 +610,25 @@ bool ValidateProgram(const Program& program, bool require_ssa, std::string* erro
 					const auto* predecessor = inst.PhiBlock(arg_index);
 					if (definition->Parent() != predecessor &&
 					    !dominates(definition->Parent(), predecessor)) {
-						return Fail(error,
-						            "value IR Phi incoming definition does not dominate its edge");
+						return Fail("value IR Phi incoming definition does not dominate its edge");
 					}
 				} else if (definition->Parent() == block) {
 					if (instruction_positions.at(definition) >= instruction_positions.at(&inst)) {
-						return Fail(error,
-						            "value IR instruction uses a same-block definition before it");
+						return Fail("value IR instruction uses a same-block definition before it");
 					}
 				} else if (!dominates(definition->Parent(), block)) {
-					return Fail(error, "value IR instruction definition does not dominate its use");
+					return Fail("value IR instruction definition does not dominate its use");
 				}
 			}
 		}
 		const auto& info = program.block_info[block_index];
 		if (!info.condition.IsEmpty() && !control_dominates(info.condition, block)) {
-			return Fail(error, "value IR branch condition definition does not dominate its use");
+			return Fail("value IR branch condition definition does not dominate its use");
 		}
 		if (!info.indirect_target.IsEmpty() && !control_dominates(info.indirect_target, block)) {
-			return Fail(error, "value IR indirect selector definition does not dominate its use");
+			return Fail("value IR indirect selector definition does not dominate its use");
 		}
 	}
-	return true;
 }
 
 void ResolveControlFlowIdentities(Program& program) {

src/graphics/shader/recompiler/ir/ResourceSnapshot.h

@@ -0,0 +1,29 @@
+#ifndef EMULATOR_INCLUDE_EMULATOR_GRAPHICS_SHADER_RECOMPILER_RESOURCESNAPSHOT_H_
+#define EMULATOR_INCLUDE_EMULATOR_GRAPHICS_SHADER_RECOMPILER_RESOURCESNAPSHOT_H_
+
+#include <array>
+#include <cstdint>
+#include <vector>
+
+namespace Libs::Graphics::ShaderRecompiler::IR {
+
+struct DescriptorValue {
+	std::array<uint32_t, 8> dwords      = {};
+	uint32_t                dword_count = 0;
+
+	bool operator==(const DescriptorValue& other) const {
+		return dword_count == other.dword_count && dwords == other.dwords;
+	}
+};
+
+struct ResourceSnapshot {
+	std::vector<DescriptorValue> buffers;
+	std::vector<DescriptorValue> images;
+	std::vector<DescriptorValue> samplers;
+	std::vector<uint32_t>        flattened_srt;
+	std::vector<uint32_t>        user_data;
+};
+
+} // namespace Libs::Graphics::ShaderRecompiler::IR
+
+#endif // EMULATOR_INCLUDE_EMULATOR_GRAPHICS_SHADER_RECOMPILER_RESOURCESNAPSHOT_H_

src/graphics/shader/recompiler/ir/ShaderIR.h

@@ -4,14 +4,17 @@
 #include "common/common.h"
 #include "common/stringUtils.h"
 #include "graphics/guest_gpu/gpu_defs.h"
+#include "graphics/guest_gpu/gpu_format.h"
 #include "graphics/shader/recompiler/frontend/cfg/ShaderCFG.h"
 #include "graphics/shader/recompiler/frontend/decode/ShaderDecoder.h"
 #include "graphics/shader/recompiler/ir/Block.h"
+#include "graphics/shader/recompiler/ir/ResourceSnapshot.h"
 #include "graphics/shader/recompiler/ir/opcodes/ValueOpcodes.h"
 #include "graphics/shader/shader.h"
 
 #include <array>
 #include <bit>
+#include <list>
 #include <memory>
 #include <optional>
 #include <string_view>
@@ -30,9 +33,6 @@ enum class ResourceKind {
 	Lds,
 	Gds,
 	Image,
-	ImageUint,
-	StorageImage,
-	StorageImageUint,
 	Sampler
 };
 
@@ -41,19 +41,6 @@ enum class ResourceKind {
 	       kind == ResourceKind::Global || kind == ResourceKind::Scratch;
 }
 
-[[nodiscard]] constexpr bool ImageResourceKindMatches(ResourceKind       kind,
-                                                      ImageResourceClass resource_class) {
-	switch (resource_class) {
-		case ImageResourceClass::Sampled:
-			return kind == ResourceKind::Image || kind == ResourceKind::ImageUint;
-		case ImageResourceClass::Storage:
-			return kind == ResourceKind::StorageImage || kind == ResourceKind::StorageImageUint;
-		case ImageResourceClass::StorageUint: return kind == ResourceKind::StorageImageUint;
-		case ImageResourceClass::None: return false;
-	}
-	return false;
-}
-
 struct MemoryInfo {
 	ResourceKind            kind                     = ResourceKind::None;
 	uint32_t                resource                 = 0;
@@ -78,7 +65,6 @@ struct MemoryInfo {
 	bool                    typed                                                 = false;
 	bool                    formatted                                             = false;
 	bool                    image_has_mip                                         = false;
-	bool                    image_cube                                            = false;
 	bool                    image_r128                                            = false;
 	bool                    glc                                                   = false;
 	bool                    slc                                                   = false;
@@ -103,15 +89,6 @@ struct ExportInfo {
 	bool operator==(const ExportInfo& other) const = default;
 };
 
-struct DescriptorValue {
-	std::array<uint32_t, 8> dwords      = {};
-	uint32_t                dword_count = 0;
-
-	bool operator==(const DescriptorValue& other) const {
-		return dword_count == other.dword_count && dwords == other.dwords;
-	}
-};
-
 struct BufferResource {
 	static constexpr uint32_t NoImageAlias = UINT32_MAX;
 
@@ -138,24 +115,25 @@ constexpr uint32_t ShaderImageIdentitySwizzle = 0x00000facu;
 struct ImageResource {
 	static constexpr uint32_t NoIndirectImage = UINT32_MAX;
 
-	uint32_t                source                    = 0;
-	uint32_t                first_use_pc              = 0;
-	ResourceKind            kind                      = ResourceKind::None;
-	Decoder::ImageDimension dimension                 = Decoder::ImageDimension::Unknown;
-	ImageMipMode            mip_mode                  = ImageMipMode::None;
-	uint32_t                mip_count                 = 1;
-	Prospero::BufferFormat  conversion_format         = Prospero::BufferFormat::kInvalid;
-	uint32_t                shader_swizzle            = ShaderImageIdentitySwizzle;
-	bool                    read                      = false;
-	bool                    written                   = false;
-	bool                    atomic                    = false;
-	bool                    depth_compare             = false;
-	bool                    cube                      = false;
-	bool                    r128                      = false;
-	uint32_t                indirect_root             = NoIndirectImage;
-	uint32_t                indirect_mapping_offset   = 0;
-	uint32_t                indirect_mapping_capacity = 0;
-	std::vector<uint32_t>   indirect_resources;
+	uint32_t                      source            = 0;
+	uint32_t                      first_use_pc      = 0;
+	ImageResourceClass            resource_class    = ImageResourceClass::None;
+	Prospero::TextureNumericClass numeric_class     = Prospero::TextureNumericClass::Unsupported;
+	Decoder::ImageDimension       dimension         = Decoder::ImageDimension::Unknown;
+	ImageMipMode                  mip_mode          = ImageMipMode::None;
+	uint32_t                      mip_count         = 1;
+	Prospero::BufferFormat        conversion_format = Prospero::BufferFormat::kInvalid;
+	uint32_t                      shader_swizzle    = ShaderImageIdentitySwizzle;
+	bool                          read              = false;
+	bool                          written           = false;
+	bool                          atomic            = false;
+	bool                          depth_compare     = false;
+	bool                          cube              = false;
+	bool                          r128              = false;
+	uint32_t                      indirect_root     = NoIndirectImage;
+	uint32_t                      indirect_mapping_offset   = 0;
+	uint32_t                      indirect_search_iterations = 0;
+	std::vector<uint32_t>         indirect_resources;
 
 	bool operator==(const ImageResource& other) const = default;
 };
@@ -164,6 +142,7 @@ struct SamplerResource {
 	uint32_t source                = 0;
 	uint32_t first_use_pc          = 0;
 	bool     force_point_filtering = false;
+	bool     depth_compare         = false;
 
 	bool operator==(const SamplerResource& other) const = default;
 };
@@ -205,7 +184,6 @@ enum class StageOutputKind {
 struct PositionExportComponent {
 	uint32_t clip_distance = UINT32_MAX;
 	uint32_t cull_distance = UINT32_MAX;
-	bool     valid          = false;
 	bool     point_size     = false;
 	bool     layer          = false;
 	bool     viewport       = false;
@@ -234,7 +212,6 @@ inline PositionExportComponent DecodePositionExportComponent(uint32_t control,
 		return result;
 	}
 
-	result.valid = true;
 	if (vector == 0) {
 		result.point_size = component == 0 && (control & (1u << 16u)) != 0;
 		result.layer      = component == 2 && (control & (1u << 18u)) != 0;
@@ -274,124 +251,126 @@ struct StageOutput {
 	bool operator==(const StageOutput& other) const = default;
 };
 
-enum class DescriptorBindingKind {
-	Buffers,
-	Sampled1D,
-	Sampled1DArray,
-	Sampled2D,
-	Sampled2DArray,
-	Sampled2DMsaa,
-	Sampled2DMsaaArray,
-	Sampled3D,
-	SampledUint1D,
-	SampledUint1DArray,
-	SampledUint2D,
-	SampledUint2DArray,
-	SampledUint2DMsaa,
-	SampledUint2DMsaaArray,
-	SampledUint3D,
-	Storage1D,
-	Storage1DArray,
-	Storage2D,
-	Storage2DArray,
-	Storage3D,
-	StorageUint1D,
-	StorageUint1DArray,
-	StorageUint2D,
-	StorageUint2DArray,
-	StorageUint3D,
-	StorageAtomic1D,
-	StorageAtomic1DArray,
-	StorageAtomic2D,
-	StorageAtomic2DArray,
-	StorageAtomic3D,
-	Samplers,
+inline constexpr uint32_t FirstImageBinding        = 1u;
+inline constexpr uint32_t FirstStorageImageBinding = 22u;
+inline constexpr uint32_t ImageBindingCount        = 36u;
+
+enum class DescriptorBindingKind : uint32_t {
+	Buffers  = 0u,
+	Samplers = FirstImageBinding + ImageBindingCount,
 	Gds,
 	BdaPagetable,
 	FaultBuffer,
 	FlattenedSrt,
-	UserData,
+	ShaderData,
 	Count,
 };
 
-constexpr uint32_t NativePushConstantSize = 128;
+static_assert(static_cast<uint32_t>(DescriptorBindingKind::Samplers) == 37u);
+static_assert(static_cast<uint32_t>(DescriptorBindingKind::Count) == 43u);
+
+struct PushData {
+	static constexpr uint32_t DwordCount = 32;
+	static constexpr uint32_t NoStart    = UINT32_MAX;
+	std::array<uint32_t, DwordCount> dwords {};
+
+	[[nodiscard]] static constexpr bool CanFit(uint32_t start, uint32_t size) {
+		return size != 0 && start <= DwordCount && size <= DwordCount - start;
+	}
+	[[nodiscard]] static constexpr uint32_t StartFor(uint32_t cursor, uint32_t size) {
+		return CanFit(cursor, size) ? cursor : NoStart;
+	}
+};
+
+static_assert(sizeof(PushData) == 128);
+constexpr uint32_t NativePushConstantSize = sizeof(PushData);
 
 [[nodiscard]] constexpr uint32_t NativeBinding(ShaderType stage, DescriptorBindingKind kind) {
 	return static_cast<uint32_t>(kind) +
 	       (stage == ShaderType::Pixel ? static_cast<uint32_t>(DescriptorBindingKind::Count) : 0u);
 }
 
-[[nodiscard]] inline std::optional<DescriptorBindingKind>
-DescriptorBindingForImage(const ImageResource& image) {
-	using Dimension = Decoder::ImageDimension;
-	using Kind      = DescriptorBindingKind;
+[[nodiscard]] constexpr ImageResourceClass ImageBindingResourceClass(DescriptorBindingKind kind) {
+	const auto value = static_cast<uint32_t>(kind);
+	if (value >= FirstImageBinding && value < FirstStorageImageBinding) {
+		return ImageResourceClass::Sampled;
+	}
+	if (value >= FirstStorageImageBinding &&
+	    value < static_cast<uint32_t>(DescriptorBindingKind::Samplers)) {
+		return ImageResourceClass::Storage;
+	}
+	return ImageResourceClass::None;
+}
 
-	switch (image.kind) {
-		case ResourceKind::Image:
-			switch (image.dimension) {
-				case Dimension::Dim1D: return Kind::Sampled1D;
-				case Dimension::Dim1DArray: return Kind::Sampled1DArray;
-				case Dimension::Dim2D: return Kind::Sampled2D;
-				case Dimension::Dim2DArray: return Kind::Sampled2DArray;
-				case Dimension::Dim2DMsaa: return Kind::Sampled2DMsaa;
-				case Dimension::Dim2DMsaaArray: return Kind::Sampled2DMsaaArray;
-				case Dimension::Dim3D: return Kind::Sampled3D;
-				default: return std::nullopt;
-			}
-		case ResourceKind::ImageUint:
-			switch (image.dimension) {
-				case Dimension::Dim1D: return Kind::SampledUint1D;
-				case Dimension::Dim1DArray: return Kind::SampledUint1DArray;
-				case Dimension::Dim2D: return Kind::SampledUint2D;
-				case Dimension::Dim2DArray: return Kind::SampledUint2DArray;
-				case Dimension::Dim2DMsaa: return Kind::SampledUint2DMsaa;
-				case Dimension::Dim2DMsaaArray: return Kind::SampledUint2DMsaaArray;
-				case Dimension::Dim3D: return Kind::SampledUint3D;
-				default: return std::nullopt;
-			}
-		case ResourceKind::StorageImage:
-			if (image.atomic) {
+[[nodiscard]] constexpr uint32_t ImageBindingIndex(DescriptorBindingKind kind) {
+	return static_cast<uint32_t>(kind) - FirstImageBinding;
+}
+
+[[nodiscard]] constexpr std::optional<DescriptorBindingKind>
+DescriptorBindingForImage(const ImageResource& image) {
+	constexpr uint32_t SampledFloatBinding = 1u;
+	constexpr uint32_t SampledUintBinding  = 8u;
+	constexpr uint32_t SampledSintBinding  = 15u;
+	constexpr uint32_t StorageFloatBinding = 22u;
+	constexpr uint32_t StorageUintBinding  = 27u;
+	constexpr uint32_t AtomicUintBinding   = 32u;
+
+	uint32_t base    = 0;
+	bool     sampled = false;
+	if (image.resource_class == ImageResourceClass::Sampled) {
+		if (image.atomic) {
+			return std::nullopt;
+		}
+		sampled = true;
+		switch (image.numeric_class) {
+			case Prospero::TextureNumericClass::Float: base = SampledFloatBinding; break;
+			case Prospero::TextureNumericClass::Uint: base = SampledUintBinding; break;
+			case Prospero::TextureNumericClass::Sint: base = SampledSintBinding; break;
+			case Prospero::TextureNumericClass::Unsupported: return std::nullopt;
+			default: return std::nullopt;
+		}
+	} else if (image.resource_class == ImageResourceClass::Storage) {
+		if (image.atomic) {
+			if (image.numeric_class != Prospero::TextureNumericClass::Uint) {
 				return std::nullopt;
 			}
-			switch (image.dimension) {
-				case Dimension::Dim1D: return Kind::Storage1D;
-				case Dimension::Dim1DArray: return Kind::Storage1DArray;
-				case Dimension::Dim2D: return Kind::Storage2D;
-				case Dimension::Dim2DArray: return Kind::Storage2DArray;
-				case Dimension::Dim3D: return Kind::Storage3D;
+			base = AtomicUintBinding;
+		} else {
+			switch (image.numeric_class) {
+				case Prospero::TextureNumericClass::Float: base = StorageFloatBinding; break;
+				case Prospero::TextureNumericClass::Uint: base = StorageUintBinding; break;
+				case Prospero::TextureNumericClass::Sint:
+				case Prospero::TextureNumericClass::Unsupported: return std::nullopt;
 				default: return std::nullopt;
 			}
-		case ResourceKind::StorageImageUint:
-			switch (image.dimension) {
-				case Dimension::Dim1D:
-					if (image.atomic) {
-						return Kind::StorageAtomic1D;
-					}
-					return Kind::StorageUint1D;
-				case Dimension::Dim1DArray:
-					if (image.atomic) {
-						return Kind::StorageAtomic1DArray;
-					}
-					return Kind::StorageUint1DArray;
-				case Dimension::Dim2D:
-					if (image.atomic) {
-						return Kind::StorageAtomic2D;
-					}
-					return Kind::StorageUint2D;
-				case Dimension::Dim2DArray:
-					if (image.atomic) {
-						return Kind::StorageAtomic2DArray;
-					}
-					return Kind::StorageUint2DArray;
-				case Dimension::Dim3D:
-					if (image.atomic) {
-						return Kind::StorageAtomic3D;
-					}
-					return Kind::StorageUint3D;
-				default: return std::nullopt;
+		}
+	} else {
+		return std::nullopt;
+	}
+
+	uint32_t dimension = 0;
+	switch (image.dimension) {
+		case Decoder::ImageDimension::Dim1D: break;
+		case Decoder::ImageDimension::Dim1DArray: dimension = 1u; break;
+		case Decoder::ImageDimension::Dim2D: dimension = 2u; break;
+		case Decoder::ImageDimension::Dim2DArray: dimension = 3u; break;
+		case Decoder::ImageDimension::Dim2DMsaa:
+			if (!sampled) {
+				return std::nullopt;
 			}
+			dimension = 4u;
+			break;
+		case Decoder::ImageDimension::Dim2DMsaaArray:
+			if (!sampled) {
+				return std::nullopt;
+			}
+			dimension = 5u;
+			break;
+		case Decoder::ImageDimension::Dim3D: dimension = sampled ? 6u : 4u; break;
+		case Decoder::ImageDimension::Unknown: return std::nullopt;
 		default: return std::nullopt;
 	}
+	return static_cast<DescriptorBindingKind>(base + dimension);
 }
 
 struct DescriptorBinding {
@@ -402,8 +381,7 @@ struct DescriptorBinding {
 };
 
 struct BindingLayout {
-	uint32_t                       push_constant_offset = 0;
-	uint32_t                       push_constant_size  = 0;
+	uint32_t                       push_data_start_dword = PushData::NoStart;
 	uint32_t                       memory_offset_dword = 0;
 	uint32_t                       memory_offset_count = 0;
 	std::vector<uint32_t>          user_data_registers;
@@ -412,6 +390,14 @@ struct BindingLayout {
 	[[nodiscard]] uint32_t ShaderDataDwords() const {
 		return memory_offset_dword + (memory_offset_count + 3u) / 4u;
 	}
+	[[nodiscard]] bool UsesPushData() const {
+		return push_data_start_dword != PushData::NoStart;
+	}
+	void AdvancePushData(uint32_t& cursor) const {
+		if (UsesPushData()) {
+			cursor = push_data_start_dword + ShaderDataDwords();
+		}
+	}
 
 	bool operator==(const BindingLayout& other) const = default;
 };
@@ -430,6 +416,7 @@ struct ShaderInfo {
 	std::vector<StageOutput>         outputs;
 	std::array<uint8_t, 32>          vertex_fetch_components {};
 	int32_t                          vertex_offset_sgpr = -1;
+	int32_t                          instance_offset_sgpr = -1;
 	bool                             has_bitwise_xor    = false;
 	bool                             uses_dma           = false;
 
@@ -445,6 +432,7 @@ struct SpirvRequirements {
 	bool function_lds                 = false;
 	bool function_scratch             = false;
 	bool pixel_valid_mask             = false;
+	bool buffer_int64_atomics         = false;
 };
 
 struct BlockInfo {
@@ -477,12 +465,51 @@ struct DescriptorSource {
 struct SrtRead {
 	Value    value;
 	uint32_t flat_offset = 0;
-	uint32_t use_pc      = 0;
 
 	bool operator==(const SrtRead& other) const = default;
 };
 
-struct Program {
+// Stable shader metadata consumed by the renderer after native IR has been discarded.
+struct CompiledShaderInfo {
+	ShaderType                    stage               = ShaderType::Unknown;
+	uint64_t                      shader_hash         = 0;
+	uint32_t                      wave_size           = 64;
+	uint32_t                      user_data_base      = 0;
+	uint32_t                      user_data_count     = 64;
+	uint32_t                      scratch_dwords      = 0;
+	uint32_t                      param_export_mask   = 0;
+	ShaderInfo                    info;
+	BindingLayout                 bindings;
+};
+
+// Immutable runtime resource analysis retained by the shader cache. It owns only the native
+// value graph reachable from descriptors/SRT reads, rather than the translated shader CFG.
+struct ResourcePlan {
+	ResourcePlan() = default;
+	~ResourcePlan();
+
+	ResourcePlan(const ResourcePlan&)            = delete;
+	ResourcePlan& operator=(const ResourcePlan&) = delete;
+	ResourcePlan(ResourcePlan&&) noexcept         = default;
+	ResourcePlan& operator=(ResourcePlan&& other) noexcept;
+
+	ShaderType                    stage           = ShaderType::Unknown;
+	uint64_t                      shader_hash     = 0;
+	uint32_t                      user_data_base  = 0;
+	uint32_t                      user_data_count = 64;
+	std::list<Inst>                     value_storage;
+	std::vector<MemoryInfo>             memory_info;
+	std::vector<DescriptorSource>       descriptor_sources;
+	std::vector<uint32_t>               materialization_sources;
+	std::vector<SrtRead>                srt_reads;
+	std::vector<uint8_t>                clean_flat_slots;
+	bool                                requires_specialization_memory = false;
+	bool                                srt_plan_complete          = false;
+	bool                                resource_tracking_complete = false;
+	ShaderInfo                          info;
+};
+
+struct Program: ResourcePlan {
 	Program() = default;
 	~Program();
 
@@ -490,31 +517,22 @@ struct Program {
 	Program& operator=(const Program&) = delete;
 	Program(Program&&) noexcept         = default;
 	Program& operator=(Program&& other) noexcept;
+	CompiledShaderInfo TakeCompiledInfo() &&;
 
-	ShaderType                    stage               = ShaderType::Unknown;
-	uint64_t                      shader_hash         = 0;
-	uint32_t                      wave_size           = 64;
-	uint32_t                      user_data_base      = 0;
-	uint32_t                      user_data_count     = 64;
-	uint32_t                      scratch_dwords      = 0;
+	std::vector<std::unique_ptr<Block>> block_storage;
+	BlockList                           blocks;
+	uint32_t                      wave_size      = 64;
+	uint32_t                      scratch_dwords = 0;
 	bool                          dispatcher_fallback = false;
 	CFG::FailureKind              cfg_failure_kind    = CFG::FailureKind::None;
 	std::string                   fallback_reason;
-	std::vector<std::unique_ptr<Block>> block_storage;
-	BlockList                           blocks;
-	std::vector<BlockInfo>              block_info;
+	std::vector<BlockInfo>        block_info;
 	// Decoded MIMG/VMEM metadata carries details such as RDNA2 NSA address registers and
 	// storage-image swizzles. Typed memory instructions carry a dense index into these shader-local
 	// tables until those fields are consumed by emission.
-	std::vector<MemoryInfo>       memory_info;
 	std::vector<ExportInfo>       export_info;
-	std::vector<DescriptorSource> descriptor_sources;
-	std::vector<SrtRead>          srt_reads;
 	std::vector<Value>            dynamic_reads;
-	bool                          srt_plan_complete = false;
-	ShaderInfo                    info;
-	bool                          resource_tracking_complete = false;
-	bool                          shader_info_complete       = false;
+	bool                          shader_info_complete = false;
 	BindingLayout                 bindings;
 	bool                          binding_layout_complete = false;
 
@@ -523,10 +541,10 @@ struct Program {
 
 std::string ProgramToString(const Program& program);
 
-bool  ValidateProgram(const Program& program, bool require_ssa, std::string* error);
+void  ValidateProgram(const Program& program, bool require_ssa);
 void  ResolveControlFlowIdentities(Program& program);
-bool  EquivalentValue(const Program& program, Value left, Value right);
-Value ResolveInvariantPhi(const Program& program, Value value);
+bool  EquivalentValue(const ResourcePlan& program, Value left, Value right);
+Value ResolveInvariantPhi(const ResourcePlan& program, Value value);
 
 } // namespace Libs::Graphics::ShaderRecompiler::IR
 

src/graphics/shader/recompiler/ir/opcodes/ValueOpcodes.cpp

@@ -111,6 +111,8 @@ BufferAccess BufferAccessOf(ValueOpcode opcode) {
 		case ValueOpcode::StoreBufferU32x3:
 		case ValueOpcode::StoreBufferU32x4: return BufferAccess::Write;
 		case ValueOpcode::BufferAtomicSwap32:
+		case ValueOpcode::BufferAtomicCmpSwap32:
+		case ValueOpcode::BufferAtomicSwap64:
 		case ValueOpcode::BufferAtomicIAdd32:
 		case ValueOpcode::BufferAtomicISub32:
 		case ValueOpcode::BufferAtomicSMin32:
@@ -119,6 +121,7 @@ BufferAccess BufferAccessOf(ValueOpcode opcode) {
 		case ValueOpcode::BufferAtomicUMax32:
 		case ValueOpcode::BufferAtomicAnd32:
 		case ValueOpcode::BufferAtomicOr32:
+		case ValueOpcode::BufferAtomicOr64:
 		case ValueOpcode::BufferAtomicXor32:
 		case ValueOpcode::BufferAtomicFMin32:
 		case ValueOpcode::BufferAtomicFMax32: return BufferAccess::Atomic;
@@ -128,6 +131,8 @@ BufferAccess BufferAccessOf(ValueOpcode opcode) {
 
 uint32_t BufferComponentCount(ValueOpcode opcode) {
 	switch (opcode) {
+		case ValueOpcode::BufferAtomicSwap64:
+		case ValueOpcode::BufferAtomicOr64:
 		case ValueOpcode::LoadBufferU32x2:
 		case ValueOpcode::StoreBufferU32x2: return 2u;
 		case ValueOpcode::LoadBufferU32x3:
@@ -204,13 +209,14 @@ ImageOpcodeInfo ImageOpcodeInfoOf(ValueOpcode opcode) {
 			return {ImageAccess::Read, ImageResourceClass::Sampled, true};
 		case ValueOpcode::ImageWrite:
 			return {ImageAccess::Write, ImageResourceClass::Storage, false};
+		case ValueOpcode::ImageAtomicSwap32:
 		case ValueOpcode::ImageAtomicIAdd32:
 		case ValueOpcode::ImageAtomicUMin32:
 		case ValueOpcode::ImageAtomicUMax32:
 		case ValueOpcode::ImageAtomicAnd32:
 		case ValueOpcode::ImageAtomicOr32:
 		case ValueOpcode::ImageAtomicXor32:
-			return {ImageAccess::Atomic, ImageResourceClass::StorageUint, false};
+			return {ImageAccess::Atomic, ImageResourceClass::Storage, false};
 		default: return {};
 	}
 }

src/graphics/shader/recompiler/ir/opcodes/ValueOpcodes.h

@@ -20,7 +20,7 @@ enum class BufferAccess { None, Read, Write, Atomic };
 enum class SharedAccess { None, Read, Write, Atomic, Append, Consume };
 enum class AddressAccess { None, Read, Write };
 enum class ImageAccess { None, Read, Write, Atomic };
-enum class ImageResourceClass { None, Sampled, Storage, StorageUint };
+enum class ImageResourceClass { None, Sampled, Storage };
 
 struct AddressOpcodeInfo {
 	AddressAccess access    = AddressAccess::None;

src/graphics/shader/recompiler/ir/opcodes/ValueOpcodes.inc

@@ -188,6 +188,7 @@ VALUE_OPCODE(ReadFirstLane, U32, U32, U1)
 VALUE_OPCODE(ReadLane, U32, U32, U32)
 VALUE_OPCODE(WriteLane, U32, U32, U32, U32)
 VALUE_OPCODE(Permlane16U32, U32, U32, U32, U32, U1)
+VALUE_OPCODE(BpermuteU32, U32, U32, U32, U1)
 
 // Typed shader resources and memory operations. Descriptor state is represented by SSA values;
 // dense bindings are assigned by resource tracking.
@@ -221,6 +222,8 @@ VALUE_OPCODE(StoreBufferU32x2, Void, BufferResource, U32, U32, U32, U32x2, U1)
 VALUE_OPCODE(StoreBufferU32x3, Void, BufferResource, U32, U32, U32, U32x3, U1)
 VALUE_OPCODE(StoreBufferU32x4, Void, BufferResource, U32, U32, U32, U32x4, U1)
 VALUE_OPCODE(BufferAtomicSwap32, U32, BufferResource, U32, U32, U32, U32, U1)
+VALUE_OPCODE(BufferAtomicCmpSwap32, U32, BufferResource, U32, U32, U32, U32, U32, U1)
+VALUE_OPCODE(BufferAtomicSwap64, U64, BufferResource, U32, U32, U32, U64, U1)
 VALUE_OPCODE(BufferAtomicIAdd32, U32, BufferResource, U32, U32, U32, U32, U1)
 VALUE_OPCODE(BufferAtomicISub32, U32, BufferResource, U32, U32, U32, U32, U1)
 VALUE_OPCODE(BufferAtomicSMin32, U32, BufferResource, U32, U32, U32, U32, U1)
@@ -229,6 +232,7 @@ VALUE_OPCODE(BufferAtomicSMax32, U32, BufferResource, U32, U32, U32, U32, U1)
 VALUE_OPCODE(BufferAtomicUMax32, U32, BufferResource, U32, U32, U32, U32, U1)
 VALUE_OPCODE(BufferAtomicAnd32, U32, BufferResource, U32, U32, U32, U32, U1)
 VALUE_OPCODE(BufferAtomicOr32, U32, BufferResource, U32, U32, U32, U32, U1)
+VALUE_OPCODE(BufferAtomicOr64, U64, BufferResource, U32, U32, U32, U64, U1)
 VALUE_OPCODE(BufferAtomicXor32, U32, BufferResource, U32, U32, U32, U32, U1)
 VALUE_OPCODE(BufferAtomicFMin32, U32, BufferResource, U32, U32, U32, U32, U1)
 VALUE_OPCODE(BufferAtomicFMax32, U32, BufferResource, U32, U32, U32, U32, U1)
@@ -267,6 +271,7 @@ VALUE_OPCODE(ImageRead, U32x4, ImageResource, ImageAddress, U1)
 VALUE_OPCODE(ImageWrite, Void, ImageResource, ImageAddress, U32x4, U1)
 VALUE_OPCODE(ImageSampleRaw, U32x4, ImageResource, SamplerResource, ImageAddress)
 VALUE_OPCODE(ImageGatherRaw, U32x4, ImageResource, SamplerResource, ImageAddress)
+VALUE_OPCODE(ImageAtomicSwap32, U32, ImageResource, ImageAddress, U32, U1)
 VALUE_OPCODE(ImageAtomicIAdd32, U32, ImageResource, ImageAddress, U32, U1)
 VALUE_OPCODE(ImageAtomicUMin32, U32, ImageResource, ImageAddress, U32, U1)
 VALUE_OPCODE(ImageAtomicUMax32, U32, ImageResource, ImageAddress, U32, U1)

src/graphics/shader/recompiler/ir/passes/BindingLayout.cpp

@@ -1,5 +1,6 @@
 #include "graphics/shader/recompiler/ir/passes/BindingLayout.h"
 
+#include "common/assert.h"
 #include "graphics/shader/recompiler/ir/ShaderIR.h"
 
 #include <algorithm>
@@ -9,12 +10,12 @@
 namespace Libs::Graphics::ShaderRecompiler::IR {
 namespace {
 
-constexpr auto     FirstImageBinding     = DescriptorBindingKind::Sampled1D;
-constexpr uint32_t ImageBindingCount =
-    static_cast<uint32_t>(DescriptorBindingKind::StorageAtomic3D) -
-    static_cast<uint32_t>(FirstImageBinding) + 1u;
+[[noreturn]] void BindingFail(const char* message) {
+	EXIT("shader binding layout failed: %s", message);
+	std::abort();
+}
 
-bool CollectUserData(const Program& program, std::vector<uint32_t>& result) {
+std::vector<uint32_t> CollectUserData(const Program& program) {
 	std::array<bool, NumScalarRegs> registers {};
 	for (const auto* block: program.blocks) {
 		for (const auto& inst: *block) {
@@ -22,22 +23,22 @@ bool CollectUserData(const Program& program, std::vector<uint32_t>& result) {
 				continue;
 			}
 			if (inst.Arg(0).GetType() != Type::ScalarReg) {
-				return false;
+				BindingFail("typed shader contains an invalid user-data register");
 			}
 			const auto index = RegIndex(inst.Arg(0).ScalarRegister());
 			if (index >= NumScalarRegs) {
-				return false;
+				BindingFail("typed shader contains an invalid user-data register");
 			}
 			registers[index] = true;
 		}
 	}
-	result.clear();
+	std::vector<uint32_t> result;
 	for (uint32_t index = 0; index < registers.size(); index++) {
 		if (registers[index]) {
 			result.push_back(index);
 		}
 	}
-	return true;
+	return result;
 }
 
 void AddBinding(BindingLayout& layout, DescriptorBindingKind kind,
@@ -45,8 +46,8 @@ void AddBinding(BindingLayout& layout, DescriptorBindingKind kind,
 	layout.descriptors.push_back({kind, std::move(resources)});
 }
 
-bool UsesGds(const Program& program, bool& uses_gds) {
-	uses_gds = false;
+bool UsesGds(const Program& program) {
+	bool uses_gds = false;
 	for (const auto* block: program.blocks) {
 		for (const auto& inst: *block) {
 			if (SharedAccessOf(inst.GetOpcode()) == SharedAccess::None) {
@@ -54,46 +55,32 @@ bool UsesGds(const Program& program, bool& uses_gds) {
 			}
 			const auto index = inst.Flags<MemoryFlags>().index;
 			if (index >= program.memory_info.size()) {
-				return false;
+				BindingFail("typed shader contains invalid shared-memory metadata");
 			}
 			const auto kind = program.memory_info[index].kind;
 			if (kind != ResourceKind::Lds && kind != ResourceKind::Gds) {
-				return false;
+				BindingFail("typed shader contains invalid shared-memory metadata");
 			}
 			uses_gds |= kind == ResourceKind::Gds;
 		}
 	}
-	return true;
+	return uses_gds;
 }
 
 } // namespace
 
-bool AllocateBindings(Program& program, uint32_t push_constant_offset, std::string* error) {
+void AllocateBindings(Program& program, uint32_t push_data_start_dword) {
 	if (!program.shader_info_complete || program.binding_layout_complete) {
-		if (error != nullptr) {
-			*error = !program.shader_info_complete ? "shader info is not ready"
-			                                       : "binding layout already allocated";
-		}
-		return false;
+		EXIT("shader binding layout failed: %s", !program.shader_info_complete
+		                                             ? "shader info is not ready"
+		                                             : "binding layout already allocated");
 	}
-	if (push_constant_offset > NativePushConstantSize ||
-	    push_constant_offset % sizeof(uint32_t) != 0) {
-		if (error != nullptr) {
-			*error = "push-constant offset exceeds the Vulkan minimum guarantee or is unaligned";
-		}
-		return false;
-	}
-
 	BindingLayout next;
-	next.push_constant_offset = push_constant_offset;
-	if (!CollectUserData(program, next.user_data_registers)) {
-		if (error != nullptr) {
-			*error = "typed shader contains an invalid user-data register";
-		}
-		return false;
-	}
+	next.user_data_registers = CollectUserData(program);
 	next.memory_offset_dword = static_cast<uint32_t>(next.user_data_registers.size());
 	next.memory_offset_count = static_cast<uint32_t>(program.info.buffers.size());
+	next.push_data_start_dword =
+	    PushData::StartFor(push_data_start_dword, next.ShaderDataDwords());
 
 	if (!program.info.buffers.empty()) {
 		std::vector<uint32_t> resources(program.info.buffers.size());
@@ -107,35 +94,25 @@ bool AllocateBindings(Program& program, uint32_t push_constant_offset, std::stri
 	for (uint32_t i = 0; i < program.info.images.size(); i++) {
 		const auto kind = DescriptorBindingForImage(program.info.images[i]);
 		if (!kind.has_value()) {
-			if (error != nullptr) {
-				*error = "shader info contains an invalid image binding class";
-			}
-			return false;
+			EXIT("shader binding layout failed: image %u has an invalid binding class", i);
 		}
-		const auto group = static_cast<uint32_t>(*kind) - static_cast<uint32_t>(FirstImageBinding);
+		const auto group = ImageBindingIndex(*kind);
 		if (group >= image_groups.size()) {
-			if (error != nullptr) {
-				*error = "shader info contains an unmapped image binding class";
-			}
-			return false;
+			EXIT("shader binding layout failed: image %u has an unmapped binding class", i);
 		}
 		auto&      resources = image_groups[group];
 		const auto dynamic   = program.info.images[i].mip_mode == ImageMipMode::DynamicStorage;
 		const auto count     = dynamic ? program.info.images[i].mip_count : 1u;
 		if (count == 0u || (!dynamic && program.info.images[i].mip_count != 1u)) {
-			if (error != nullptr) {
-				*error = "image has an invalid specialized mip descriptor count";
-			}
-			return false;
+			EXIT("shader binding layout failed: image %u has invalid specialized mip count %u", i,
+			     program.info.images[i].mip_count);
 		}
 		resources.insert(resources.end(), count, i);
 	}
 	for (uint32_t i = 0; i < image_groups.size(); i++) {
 		if (!image_groups[i].empty()) {
-			AddBinding(
-			    next,
-			    static_cast<DescriptorBindingKind>(static_cast<uint32_t>(FirstImageBinding) + i),
-			    std::move(image_groups[i]));
+			AddBinding(next, static_cast<DescriptorBindingKind>(FirstImageBinding + i),
+			           std::move(image_groups[i]));
 		}
 	}
 
@@ -146,14 +123,7 @@ bool AllocateBindings(Program& program, uint32_t push_constant_offset, std::stri
 		}
 		AddBinding(next, DescriptorBindingKind::Samplers, std::move(resources));
 	}
-	bool uses_gds = false;
-	if (!UsesGds(program, uses_gds)) {
-		if (error != nullptr) {
-			*error = "typed shader contains invalid shared-memory metadata";
-		}
-		return false;
-	}
-	if (uses_gds) {
+	if (UsesGds(program)) {
 		AddBinding(next, DescriptorBindingKind::Gds);
 	}
 	if (program.info.uses_dma) {
@@ -163,22 +133,18 @@ bool AllocateBindings(Program& program, uint32_t push_constant_offset, std::stri
 	const bool uses_flattened_runtime =
 	    !program.srt_reads.empty() ||
 	    std::ranges::any_of(program.info.images, [](const ImageResource& image) {
-		    return image.indirect_mapping_capacity != 0u;
+		    return image.indirect_search_iterations != 0u;
 	    });
 	if (uses_flattened_runtime) {
 		AddBinding(next, DescriptorBindingKind::FlattenedSrt);
 	}
 
-	const auto push_dwords = (NativePushConstantSize - push_constant_offset) / sizeof(uint32_t);
-	if (next.ShaderDataDwords() <= push_dwords) {
-		next.push_constant_size = next.ShaderDataDwords() * sizeof(uint32_t);
-	} else {
-		AddBinding(next, DescriptorBindingKind::UserData);
+	if (next.ShaderDataDwords() != 0 && !next.UsesPushData()) {
+		AddBinding(next, DescriptorBindingKind::ShaderData);
 	}
 
 	program.bindings                = std::move(next);
 	program.binding_layout_complete = true;
-	return true;
 }
 
 const DescriptorBinding* FindBinding(const BindingLayout& layout, DescriptorBindingKind kind) {

src/graphics/shader/recompiler/ir/passes/BindingLayout.h

@@ -5,7 +5,7 @@
 
 namespace Libs::Graphics::ShaderRecompiler::IR {
 
-bool AllocateBindings(Program& program, uint32_t push_constant_offset, std::string* error);
+void AllocateBindings(Program& program, uint32_t push_data_start_dword = 0);
 
 const DescriptorBinding* FindBinding(const BindingLayout& layout, DescriptorBindingKind kind);