agc: new abis

Extracted from 3db2b3c5c5e1a26a861df7ebcacd9ccb8c484420 in KytyPS5/KytyPS5#147.

.github/workflows/build.yml

@@ -83,7 +83,12 @@ jobs:
       - name: Build
         shell: cmd
         run: |
-          cmake --build _Build/windows --target launcher --parallel
+          cmake --build _Build/windows --target launcher audio_out2_port_tests virtual_memory_allocation_tests --parallel
+
+      - name: Test
+        shell: cmd
+        run: |
+          ctest --test-dir _Build/windows --output-on-failure -R "^(audio_out2_port|virtual_memory_allocation)$"
 
       - name: Install
         shell: cmd
@@ -153,7 +158,15 @@ jobs:
       - name: Build
         shell: bash
         run: |
-          cmake --build _Build/macos --target launcher --parallel
+          cmake --build _Build/macos \
+            --target launcher audio_out2_port_tests virtual_memory_allocation_tests \
+            --parallel
+
+      - name: Test
+        shell: bash
+        run: |
+          ctest --test-dir _Build/macos --output-on-failure \
+            -R '^(audio_out2_port|virtual_memory_allocation)$'
 
       - name: Install
         shell: bash
@@ -284,13 +297,14 @@ jobs:
         run: |
           cmake --build _Build/linux \
             --target launcher page_manager_tests memory_tracker_tests \
+              audio_out2_port_tests virtual_memory_allocation_tests \
             --parallel
 
       - name: Test
         shell: bash
         run: |
           ctest --test-dir _Build/linux --output-on-failure \
-            -R '^(page_manager|memory_tracker)$'
+            -R '^(audio_out2_port|page_manager|memory_tracker|virtual_memory_allocation)$'
 
       - name: Install
         shell: bash

README.md

@@ -27,8 +27,9 @@ Development is focused on compatibility and boot reliability.
 Windows is the primary platform and receives the most testing. Linux builds and runs; see
 [Building on Linux](#building-on-linux).
 
-macOS support is experimental. Compatibility with the same games on Windows and macOS has not yet
-been tested.
+macOS support is experimental. The emulator is built for x86-64 and runs on Apple Silicon under
+Rosetta 2, with Vulkan provided by MoltenVK. A small number of titles have been verified in-game
+on Apple Silicon hardware; see [Building on macOS](#building-on-macos).
 
 ## Bugs and Issues
 
@@ -114,9 +115,10 @@ the Vulkan/SPIR-V validation rules.
 
 ### System requirements
 
-- Windows 10 version 1803, or a current Linux distribution
-- A 64-bit x86 processor
-- A Vulkan 1.3-capable GPU with current drivers
+- Windows 10 version 1803, a current Linux distribution, or macOS on Apple Silicon
+- A 64-bit x86 processor (on macOS, an Apple Silicon processor with Rosetta 2)
+- A Vulkan 1.3-capable GPU with current drivers (on macOS, Vulkan is provided by the bundled
+  MoltenVK)
 
 ### Build requirements (Windows)
 
@@ -188,6 +190,56 @@ time.
 
 Note that the CMake source root is `src`, not the repository root.
 
+### Building on macOS
+
+macOS builds target x86-64 and run under Rosetta 2 on Apple Silicon, so the PS5's x86-64 game
+code executes through the same translation layer as the emulator itself. Prebuilt archives are
+attached to releases; the steps below are for building from source.
+
+Requirements:
+
+- An Apple Silicon Mac with Rosetta 2 installed (`softwareupdate --install-rosetta`)
+- Xcode (or the Command Line Tools)
+- Homebrew packages: `brew install cmake ninja glslang`
+- Qt 6 (Concurrent, Network, Widgets) with x86-64 support. The official Qt installation is
+  universal and works; Homebrew's Qt is arm64-only and will not link
+
+```bash
+git submodule update --init --recursive
+
+cmake -S src -B _Build/macos -G Ninja -DCMAKE_BUILD_TYPE=Release \
+  -DCMAKE_OSX_ARCHITECTURES=x86_64 \
+  -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ \
+  -DCMAKE_PREFIX_PATH="$Qt6_DIR"
+
+cmake --build _Build/macos --target launcher --parallel
+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.
+
+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:
+
+```bash
+codesign --force --sign - _Build/macos/install/libMoltenVK.dylib
+```
+
+Release archives already include a signed `libMoltenVK.dylib`.
+
+### Regression tests
+
+Build every regression executable and run the registered tests with:
+
+```powershell
+cmake --build _Build/windows --target kyty_tests
+ctest --test-dir _Build/windows --output-on-failure
+```
+
+Use `_Build/linux` instead of `_Build/windows` for a Linux build.
+
 ### Visual Studio Code
 
 A ready-made Visual Studio Code setup is included in [`.vscode`](.vscode). It configures CMake
@@ -233,6 +285,14 @@ 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:
+
+```bash
+cd _Build/macos/install
+SDL_VULKAN_LIBRARY="$PWD/libMoltenVK.dylib" ./kyty_emulator --game "/games/ExampleGame"
+```
+
 Run `kyty_emulator --help` to see the available graphics, logging, validation, profiling, and
 debugging options.
 

src/CMakeLists.txt

@@ -312,6 +312,19 @@ function(add_kyty_full_emulator_test target source)
 		target_link_libraries(${target} onecore)
 		add_custom_command(TARGET ${target} POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_if_different "${KYTY_THIRD_PARTY_DIR}/winpthread/bin/libwinpthread-1.dll" $<TARGET_FILE_DIR:${target}>/libwinpthread-1.dll)
 	endif()
+	# The macOS x86_64 guest address space needs its .zerofill segments anchored
+	# by linker flags, or the kernel kills the binary on load (posix_spawn EIO).
+	configure_macos_guest_address_space(${target})
+endfunction()
+
+function(configure_macos_guest_address_space target)
+	if(APPLE AND (CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" OR
+			(NOT CMAKE_OSX_ARCHITECTURES AND CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|AMD64)$")))
+		target_sources(${target} PRIVATE kernel/macosGuestAddressSpace.cpp)
+		target_compile_definitions(${target} PRIVATE KYTY_LINKED_GUEST_ADDRESS_SPACE=1)
+		target_link_options(${target} PRIVATE
+			-Wl,-ld_classic,-no_pie,-no_fixup_chains,-no_huge,-pagezero_size,0x40000,-segaddr,SYSTEM_MANAGED,0x40000,-segaddr,SYSTEM_RESERVED,0x7ffffc000,-segaddr,USER_AREA,0x7000000000,-image_base,0x700000000000)
+	endif()
 endfunction()
 
 add_kyty_full_emulator_test(shader_cfg_tests ../tests/shaderCfgTests.cpp)
@@ -319,6 +332,7 @@ add_kyty_full_emulator_test(shader_cfg_tests ../tests/shaderCfgTests.cpp)
 add_executable(scalar_provenance_tests EXCLUDE_FROM_ALL
 	../tests/ScalarProvenanceTests.cpp
 	graphics/host_gpu/hostMemory.cpp
+	graphics/shader/recompiler/ir/ReadLaneElimination.cpp
 	graphics/shader/recompiler/ir/ScalarProvenance.cpp
 	graphics/shader/recompiler/ir/SrtWalker.cpp
 )
@@ -338,7 +352,6 @@ add_executable(memory_tracker_tests EXCLUDE_FROM_ALL
 )
 target_link_libraries(memory_tracker_tests fmt::fmt common)
 target_include_directories(memory_tracker_tests PRIVATE ${inc_headers})
-target_compile_definitions(memory_tracker_tests PRIVATE KYTY_MEMORY_TRACKER_TESTS=1)
 
 add_executable(shader_vertex_metadata_tests EXCLUDE_FROM_ALL
 	../tests/ShaderVertexMetadataTests.cpp
@@ -381,6 +394,14 @@ add_executable(resource_mutex_tests EXCLUDE_FROM_ALL
 target_link_libraries(resource_mutex_tests common)
 target_include_directories(resource_mutex_tests PRIVATE ${inc_headers})
 
+add_executable(audio_out2_port_tests EXCLUDE_FROM_ALL
+	../tests/AudioOut2PortTests.cpp
+	libs/libAudio2.cpp
+	loader/timer.cpp
+)
+target_link_libraries(audio_out2_port_tests common fmt::fmt)
+target_include_directories(audio_out2_port_tests PRIVATE ${inc_headers})
+
 add_executable(event_queue_lifetime_tests EXCLUDE_FROM_ALL
 	../tests/EventQueueLifetimeTests.cpp
 	kernel/eventQueue.cpp
@@ -431,12 +452,20 @@ if(NOT KYTY_CLANG_CL)
 endif()
 
 if(BUILD_TESTING)
+	add_test(NAME shader_cfg COMMAND $<TARGET_FILE:shader_cfg_tests>)
+	add_test(NAME scalar_provenance COMMAND $<TARGET_FILE:scalar_provenance_tests>)
 	add_test(NAME image_page_table COMMAND $<TARGET_FILE:image_page_table_tests>)
 	add_test(NAME memory_tracker COMMAND $<TARGET_FILE:memory_tracker_tests>)
 	add_test(NAME page_manager COMMAND $<TARGET_FILE:page_manager_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_tracking COMMAND $<TARGET_FILE:resource_tracking_tests>)
 	add_test(NAME resource_mutex COMMAND $<TARGET_FILE:resource_mutex_tests>)
 	add_test(NAME event_queue_lifetime COMMAND $<TARGET_FILE:event_queue_lifetime_tests>)
+	add_test(NAME audio_out2_port COMMAND $<TARGET_FILE:audio_out2_port_tests>)
 	add_test(NAME shader_recompiler_compute COMMAND $<TARGET_FILE:shader_recompiler_compute_tests>)
+	add_test(NAME virtual_memory_allocation
+		COMMAND $<TARGET_FILE:virtual_memory_allocation_tests>)
 	add_test(NAME command_scheduler_timeline
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --scheduler-only)
 	add_test(NAME stream_buffer_ring
@@ -466,10 +495,26 @@ if(BUILD_TESTING)
 		add_test(NAME buffer_cache_ranges
 			COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --buffer-cache-range-only)
 	endif()
+
+	add_custom_target(kyty_tests DEPENDS
+		shader_cfg_tests
+		scalar_provenance_tests
+		image_page_table_tests
+		memory_tracker_tests
+		page_manager_tests
+		shader_vertex_metadata_tests
+		shader_stage_runtime_tests
+		resource_tracking_tests
+		resource_mutex_tests
+		event_queue_lifetime_tests
+		shader_recompiler_compute_tests
+		virtual_memory_allocation_tests
+	)
 endif()
 
 
 add_executable(kyty_emulator main.cpp ${kyty_emulator_src})
+configure_macos_guest_address_space(kyty_emulator)
 
 target_link_libraries(kyty_emulator ${kyty_emulator_link_libraries})
 if (WIN32)
@@ -498,7 +543,12 @@ set(KYTY_EMULATOR_MAP_LINK_PATH "${CMAKE_CURRENT_BINARY_DIR}/${KYTY_EMULATOR_MAP
 set(KYTY_EMULATOR_PDB_LINK_PATH "${CMAKE_CURRENT_BINARY_DIR}/kyty_emulator.pdb")
 
 if(KYTY_CLANG_CL)
-	set_target_properties(kyty_emulator PROPERTIES LINK_FLAGS "/DYNAMICBASE:NO /DEBUG:FULL /PDB:${KYTY_EMULATOR_PDB_LINK_PATH} /lldmap:${KYTY_EMULATOR_MAP_LINK_PATH}")
+	target_link_options(kyty_emulator PRIVATE
+		"/DYNAMICBASE:NO"
+		"/DEBUG:FULL"
+		"/PDB:${KYTY_EMULATOR_PDB_LINK_PATH}"
+		"/lldmap:${KYTY_EMULATOR_MAP_LINK_PATH}"
+	)
 	add_custom_command(TARGET kyty_emulator POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_if_different "${KYTY_THIRD_PARTY_DIR}/winpthread/bin/libwinpthread-1.dll" $<TARGET_FILE_DIR:kyty_emulator>/libwinpthread-1.dll)
 elseif(WIN32 OR LINUX)
 	set_target_properties(kyty_emulator PROPERTIES LINK_FLAGS "${KYTY_LD_OPTIONS} -Wl,-Map=${KYTY_EMULATOR_MAP_LINK_PATH}")

src/common/hostException.cpp

@@ -324,6 +324,10 @@ bool InstallHandler(Handler handler) {
 	sa.sa_sigaction = SignalHandler;
 	sa.sa_flags     = SA_SIGINFO;
 	sigemptyset(&sa.sa_mask);
+	// The guest signal-dispatch path (KernelRaiseException) interrupts threads with
+	// SIGUSR1; block it while a fault is being resolved so a stop-the-world request
+	// cannot preempt the handler between the protection fix and the retry.
+	sigaddset(&sa.sa_mask, SIGUSR1);
 
 	// macOS raises SIGBUS for protection faults on some paths and SIGSEGV on others;
 	// SIGILL covers instructions the host cannot execute (routed to the x64 emulator).

src/common/lruCache.h

@@ -49,8 +49,7 @@ public:
 
 	template <typename Function>
 	void ForEachItemBelow(Tick tick, Function&& function) {
-		constexpr bool ReturnsBool =
-		    std::is_same_v<std::invoke_result_t<Function, Object>, bool>;
+		constexpr bool ReturnsBool = std::is_same_v<std::invoke_result_t<Function, Object>, bool>;
 		for (auto* item = m_first; item != nullptr;) {
 			if (item->tick > tick) {
 				return;

src/common/platform/sysLinuxDbg.cpp

@@ -33,8 +33,7 @@ static bool OnOwnStack() {
 	}
 	void*      base = nullptr;
 	size_t     size = 0;
-	const bool ok =
-	    pthread_attr_getstack(&attr, &base, &size) == 0 && base != nullptr && size != 0;
+	const bool ok = pthread_attr_getstack(&attr, &base, &size) == 0 && base != nullptr && size != 0;
 	pthread_attr_destroy(&attr);
 	if (!ok) {
 		return false;

src/common/platform/sysLinuxFileIO.cpp

@@ -172,8 +172,7 @@ sys_file_t* SysFileCreate(const std::filesystem::path& file_name) {
 	return ret;
 }
 
-sys_file_t* SysFileOpenR(const std::filesystem::path& file_name,
-                         sys_file_cache_type_t cache_type) {
+sys_file_t* SysFileOpenR(const std::filesystem::path& file_name, sys_file_cache_type_t cache_type) {
 	auto* ret = new sys_file_t;
 
 	ret->type = SYS_FILE_FILE;
@@ -218,8 +217,7 @@ sys_file_t* SysFileCreate() {
 	return ret;
 }
 
-sys_file_t* SysFileOpenW(const std::filesystem::path& file_name,
-                         sys_file_cache_type_t cache_type) {
+sys_file_t* SysFileOpenW(const std::filesystem::path& file_name, sys_file_cache_type_t cache_type) {
 	auto* ret = new sys_file_t;
 
 	auto real_name     = get_internal_name(file_name);

src/common/platform/sysLinuxVirtual.cpp

@@ -136,8 +136,8 @@ static void* map_anonymous(uintptr_t addr, size_t size, int protect, int flags)
 			break;
 		}
 		const auto hint = (top - step) & ~(LOW_ARENA_GRAIN - 1);
-		void*      ptr  = mmap(reinterpret_cast<void*>(hint), size, protect,
-		                       flags | MAP_FIXED_NOREPLACE, -1, 0); // NOLINT
+		void* ptr = mmap(reinterpret_cast<void*>(hint), size, protect, flags | MAP_FIXED_NOREPLACE,
+		                 -1, 0); // NOLINT
 		if (ptr != MAP_FAILED) {
 			return ptr;
 		}
@@ -194,8 +194,8 @@ uint64_t SysVirtualAllocAligned(uint64_t address, uint64_t size, VirtualMemory::
 	if (ptr != MAP_FAILED && ((ret_addr & (alignment - 1)) != 0)) {
 		munmap(ptr, size);
 
-		ptr      = map_anonymous(addr, size + alignment, protect,
-		                         MAP_PRIVATE | MAP_ANON | MAP_NORESERVE);
+		ptr =
+		    map_anonymous(addr, size + alignment, protect, MAP_PRIVATE | MAP_ANON | MAP_NORESERVE);
 		ret_addr = reinterpret_cast<uintptr_t>(ptr);
 		if (ptr != MAP_FAILED) {
 #if defined(__APPLE__)

src/common/threads.cpp

@@ -5,9 +5,9 @@
 
 #include <algorithm>
 #include <atomic>
+#include <cerrno>
 #include <chrono>             // IWYU pragma: keep
 #include <condition_variable> // IWYU pragma: keep
-#include <cerrno>
 #include <mutex>
 #include <vector>
 

src/common/uniqueFunction.h

@@ -20,9 +20,7 @@ class UniqueFunction {
 	public:
 		explicit Callable(Function function): m_function(std::move(function)) {}
 
-		Result Invoke(Args&&... args) override {
-			return m_function(std::forward<Args>(args)...);
-		}
+		Result Invoke(Args&&... args) override { return m_function(std::forward<Args>(args)...); }
 
 	private:
 		Function m_function;

src/common/virtualMemory.cpp

@@ -58,25 +58,6 @@ bool FlushInstructionCache(uint64_t address, uint64_t size) {
 	return SysVirtualFlushInstructionCache(address, size);
 }
 
-bool PatchReplace(uint64_t vaddr, uint64_t value) {
-	Mode old_mode {};
-	Protect(vaddr, 8, Mode::ReadWrite, &old_mode);
-
-	auto* ptr = reinterpret_cast<uint64_t*>(vaddr);
-
-	bool ret = (*ptr != value);
-
-	*ptr = value;
-
-	Protect(vaddr, 8, old_mode);
-
-	if (IsExecute(old_mode)) {
-		FlushInstructionCache(vaddr, 8);
-	}
-
-	return ret;
-}
-
 } // namespace VirtualMemory
 
 } // namespace Common

src/common/virtualMemory.h

@@ -37,7 +37,6 @@ bool     Free(uint64_t address);
 bool     FreeRange(uint64_t address, uint64_t size);
 bool     Protect(uint64_t address, uint64_t size, Mode mode, Mode* old_mode = nullptr);
 bool     FlushInstructionCache(uint64_t address, uint64_t size);
-bool     PatchReplace(uint64_t vaddr, uint64_t value);
 
 } // namespace VirtualMemory
 

src/emulator.cpp

@@ -105,7 +105,7 @@ static void ClearDebugTextureFolder() {
 	}
 }
 
-static void Init(const Config::ConfigOptions& cfg) {
+static void Init(const Config::ConfigOptions& cfg, const std::filesystem::path& param_json) {
 	EXIT_IF(!Common::Thread::IsMainThread());
 
 	auto* slist = Common::SubsystemsList::Instance();
@@ -127,12 +127,21 @@ static void Init(const Config::ConfigOptions& cfg) {
 	slist->InitAll(true);
 
 	Config::Load(cfg);
+	slist->Add(log, {core, config});
+	slist->InitAll(true);
+
+	if (Common::File::IsFileExisting(param_json)) {
+		Loader::SystemContentLoadParamSfo(param_json);
+		if (const auto flexible_memory_size = Loader::SystemContentGetFlexibleMemorySize();
+		    flexible_memory_size != 0) {
+			Libs::LibKernel::Memory::SetFlexibleMemorySize(flexible_memory_size);
+		}
+	}
 
 	slist->Add(audio, {core, log, pthread, memory});
 	slist->Add(controller, {core, log, config});
 	slist->Add(file_system, {core, log, pthread});
 	slist->Add(graphics, {core, log, pthread, memory, config, profiler, controller});
-	slist->Add(log, {core, config});
 	slist->Add(memory, {core, log});
 	slist->Add(network, {core, log, pthread});
 	slist->Add(profiler, {core, config});
@@ -180,7 +189,8 @@ void Run(const RunOptions& options) {
 		EXIT("ELF is required\n");
 	}
 
-	Init(options.config);
+	const auto param_json = options.app0_dir / "sce_sys" / "param.json";
+	Init(options.config, param_json);
 
 	ClearDebugTextureFolder();
 
@@ -192,15 +202,6 @@ void Run(const RunOptions& options) {
 	Libs::LibKernel::FileSystem::Mount(options.app0_dir, "/app0");
 	Libs::LibKernel::FileSystem::Mount(options.app0_dir, "/hostapp");
 
-	auto param_json = options.app0_dir / "sce_sys" / "param.json";
-	if (Common::File::IsFileExisting(param_json)) {
-		Loader::SystemContentLoadParamSfo(param_json);
-		if (auto flexible_memory_size = Loader::SystemContentGetFlexibleMemorySize();
-		    flexible_memory_size != 0) {
-			Libs::LibKernel::Memory::SetFlexibleMemorySize(flexible_memory_size);
-		}
-	}
-
 	MountSandboxDirs();
 
 	auto* rt = Common::Singleton<Loader::RuntimeLinker>::Instance();

src/graphics/guest_gpu/command_processor/pm4Handlers.cpp

@@ -1926,8 +1926,14 @@ KYTY_CP_OP_PARSER(CpOpCopyData) {
 	const uint32_t num_bytes           = ((control >> 16u) & 0x1u) != 0 ? 8u : 4u;
 	const uint64_t src                 = buffer[1] | (static_cast<uint64_t>(buffer[2]) << 32u);
 	const uint64_t dst                 = buffer[3] | (static_cast<uint64_t>(buffer[4]) << 32u);
-	if (src_sel == (9u << 1u)) {
-		if (dst_sel != (2u << 1u) || dst == 0 || (dst & (num_bytes - 1u)) != 0) {
+	uint32_t       reference_clock_dst = 0;
+	switch (src_sel) {
+		case 9u: reference_clock_dst = 2u; break;
+		case 18u: reference_clock_dst = 4u; break;
+		default: break;
+	}
+	if (reference_clock_dst != 0) {
+		if (dst_sel != reference_clock_dst || dst == 0 || (dst & (num_bytes - 1u)) != 0) {
 			EXIT("unsupported reference-clock copyData, src_sel=0x%02" PRIx32
 			     " dst_sel=0x%02" PRIx32 " dst=0x%016" PRIx64 " size=%u\n",
 			     src_sel, dst_sel, dst, num_bytes);
@@ -3390,6 +3396,12 @@ void GraphicsInitJmpTablesCxIndirect() {
 	g_hw_ctx_indirect_func[Pm4::DB_COUNT_CONTROL] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
 		HwCtxIgnoreDepthMetadataRegister(cmd_offset, value);
 	};
+	for (auto cmd_offset = Pm4::DB_SRESULTS_COMPARE_STATE0;
+	     cmd_offset <= Pm4::DB_SRESULTS_COMPARE_STATE1; cmd_offset++) {
+		g_hw_ctx_indirect_func[cmd_offset] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
+			HwCtxIgnoreDepthMetadataRegister(cmd_offset, value);
+		};
+	}
 	g_hw_ctx_indirect_func[Pm4::DB_RENDER_OVERRIDE] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
 		HwCtxIgnoreDepthMetadataRegister(cmd_offset, value);
 	};

src/graphics/guest_gpu/gpu_defs.h

@@ -72,6 +72,7 @@ enum class ChannelLayout : uint32_t {
 	k32_32       = 11,
 	k16_16_16_16 = 12,
 	k32_32_32_32 = 14,
+	k5_6_5       = 16,
 	k5_5_5_1     = 17,
 	k4_4_4_4     = 19,
 	kBc1         = 35,
@@ -374,6 +375,8 @@ enum class BufferFormat : uint32_t {
 	k32_32_32_32UInt    = 75,
 	k32_32_32_32SInt    = 76,
 	k32_32_32_32Float   = 77,
+	k8Srgb              = 128,
+	k8_8Srgb            = 129,
 	k8_8_8_8Srgb        = 130,
 	k9_9_9_5Float       = 132,
 	k5_6_5UNorm         = 133,

src/graphics/guest_gpu/gpu_format.cpp

@@ -39,6 +39,7 @@ constexpr FormatInfo kFormatInfo[] = {
     {GpuEnumValue(BufferFormat::k16_16Float), 4, 0, 4, true, false},
     {GpuEnumValue(BufferFormat::k11_11_10Float), 4, 0, 4, true, false},
     {GpuEnumValue(BufferFormat::k10_10_10_2UNorm), 4, 0, 4, true, false},
+    {GpuEnumValue(BufferFormat::k10_10_10_2UInt), 4, 0, 4, true, true},
     {GpuEnumValue(BufferFormat::k8_8_8_8UNorm), 4, 0, 4, true, false},
     {GpuEnumValue(BufferFormat::k8_8_8_8SNorm), 4, 0, 4, true, false},
     {GpuEnumValue(BufferFormat::k8_8_8_8UInt), 4, 0, 4, true, true},
@@ -57,6 +58,8 @@ constexpr FormatInfo kFormatInfo[] = {
     {GpuEnumValue(BufferFormat::k32_32_32_32UInt), 16, 0, 16, true, true},
     {GpuEnumValue(BufferFormat::k32_32_32_32SInt), 16, 0, 16, false, false},
     {GpuEnumValue(BufferFormat::k32_32_32_32Float), 16, 0, 16, true, false},
+    {GpuEnumValue(BufferFormat::k8Srgb), 1, 0, 0, true, false},
+    {GpuEnumValue(BufferFormat::k8_8Srgb), 2, 0, 0, true, false},
     {GpuEnumValue(BufferFormat::k8_8_8_8Srgb), 4, 0, 4, true, false},
     {GpuEnumValue(BufferFormat::k9_9_9_5Float), 4, 0, 0, true, false},
     {GpuEnumValue(BufferFormat::k5_6_5UNorm), 2, 0, 2, true, false},

src/graphics/guest_gpu/graphicsRun.cpp

@@ -410,8 +410,13 @@ void CommandProcessor::WriteData(uint32_t* dst, const uint32_t* src, uint32_t dw
 	const uint32_t increment     = (write_control >> 16u) & 0x1u;
 	const uint32_t write_confirm = (write_control >> 20u) & 0x1u;
 
-	if (dst_sel != 0 && dst_sel != 2 && dst_sel != 4 && dst_sel != 5) {
-		EXIT("unsupported writeData destination selector 0x%02" PRIx32 "\n", dst_sel);
+	switch (dst_sel) {
+		case 0:
+		case 2:
+		case 4:
+		case 5:
+		case 6: break;
+		default: EXIT("unsupported writeData destination selector 0x%02" PRIx32 "\n", dst_sel);
 	}
 	EXIT_NOT_IMPLEMENTED(increment != 0);
 
@@ -962,9 +967,8 @@ 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);
+	m_renderer.GetRenderExecutor().DrawIndex(m_submit_id, CurrentBuffer(), m_index_type_and_size,
+	                                         index_count, index_addr, flags, 1, m_num_instances);
 }
 
 void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiator, bool indexed) {
@@ -1190,8 +1194,8 @@ void CommandProcessor::DispatchDirect(uint32_t thread_group_x, uint32_t thread_g
 			}
 		}
 
-		m_renderer.GetRenderExecutor().DispatchDirect(
-		    m_submit_id, CurrentBuffer(), thread_group_x, thread_group_y, thread_group_z, mode);
+		m_renderer.GetRenderExecutor().DispatchDirect(m_submit_id, CurrentBuffer(), thread_group_x,
+		                                              thread_group_y, thread_group_z, mode);
 	}
 
 	constexpr uint32_t DispatchInitiatorUseThreadDimensions = 1u << 5u;
@@ -1237,9 +1241,9 @@ void CommandProcessor::DrawIndexAuto(uint32_t index_count, uint32_t flags,
                                      uint32_t first_vertex, uint32_t first_instance) {
 	CheckBuffer();
 
-	m_renderer.GetRenderExecutor().DrawAuto(
-	    m_submit_id, CurrentBuffer(), index_count, flags, render_target_slice_offset,
-	    instance_count, first_vertex, first_instance);
+	m_renderer.GetRenderExecutor().DrawAuto(m_submit_id, CurrentBuffer(), index_count, flags,
+	                                        render_target_slice_offset, instance_count,
+	                                        first_vertex, first_instance);
 }
 
 void CommandProcessor::WaitFlipDone(uint32_t video_out_handle, uint32_t display_buffer_index) {
@@ -1317,8 +1321,8 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 				if (eop_event_type == 0x2f && cache_action == 0x00 && event_index == 0x06) {
 					auto* dst = static_cast<uint32_t*>(dst_gpu_addr);
 					SynchronizeGpu();
-					Sync::ReadGds(m_renderer.GetBufferCache().GetGdsBuffer(), dst,
-					              value & 0xffffu, value >> 16u);
+					Sync::ReadGds(m_renderer.GetBufferCache().GetGdsBuffer(), dst, value & 0xffffu,
+					              value >> 16u);
 					Sync::WriteAtEndOfPipeGds32(m_submit_id, CurrentBuffer(), dst, value & 0xffffu,
 					                            value >> 16u);
 					return;
@@ -1486,8 +1490,7 @@ void CommandProcessor::EmitGlobalBarrier() {
 	barrier.srcStageMask  = vk::PipelineStageFlagBits2::eAllCommands;
 	barrier.srcAccessMask = vk::AccessFlagBits2::eMemoryWrite;
 	barrier.dstStageMask  = vk::PipelineStageFlagBits2::eAllCommands;
-	barrier.dstAccessMask =
-	    vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite;
+	barrier.dstAccessMask = vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite;
 
 	vk::DependencyInfo dependency {};
 	dependency.memoryBarrierCount = 1;

src/graphics/guest_gpu/pm4.cpp

@@ -110,7 +110,6 @@ void DumpPm4PacketStream(Common::File* file, uint32_t* cmd_buffer, uint32_t star
 	auto* cmd = cmd_buffer + start_dw;
 	auto  dw  = num_dw;
 	while (dw != 0) {
-		EXIT_NOT_IMPLEMENTED(dw < 2);
 		EXIT_NOT_IMPLEMENTED(dw > num_dw);
 
 		auto cmd_id = *cmd++;
@@ -120,6 +119,9 @@ void DumpPm4PacketStream(Common::File* file, uint32_t* cmd_buffer, uint32_t star
 		uint32_t len = 0;
 
 		const auto packet_type = static_cast<PacketType>(cmd_id >> 30u);
+		// Type-2 packets are header-only padding; every other packet type requires a body.
+		EXIT_NOT_IMPLEMENTED(dw < 2 && packet_type != PacketType::Type2);
+
 		switch (packet_type) {
 			case PacketType::Type3: {
 				const bool    sh_gx = (cmd_id & 0x2u) == 0;

src/graphics/host_gpu/memoryTracker.cpp

@@ -4,16 +4,6 @@
 
 namespace Libs::Graphics {
 
-#if defined(KYTY_MEMORY_TRACKER_TESTS)
-namespace {
-std::atomic<MemoryTracker::UnmapContentionHook> g_unmap_contention_hook {nullptr};
-}
-
-void MemoryTracker::SetUnmapContentionHook(UnmapContentionHook hook) noexcept {
-	g_unmap_contention_hook.store(hook, std::memory_order_release);
-}
-#endif
-
 static_assert(std::atomic<void*>::is_always_lock_free);
 
 MemoryTracker::MemoryTracker(PageManager& page_manager, PageWatchMode gpu_watch_mode)
@@ -31,6 +21,7 @@ MemoryTracker::MemoryTracker(PageManager& page_manager, PageWatchMode gpu_watch_
 
 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 ||
@@ -68,6 +59,7 @@ 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 ||
@@ -94,7 +86,6 @@ RegionManager* MemoryTracker::GetOrCreateRegion(uint64_t index) {
 bool MemoryTracker::IsRegionCpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
 	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
 	return Iterate<true>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
 		return manager->IsModified<DirtySource::Cpu>(offset, bytes);
@@ -104,7 +95,6 @@ bool MemoryTracker::IsRegionCpuModified(uint64_t vaddr, uint64_t size) {
 bool MemoryTracker::IsRegionGpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
 	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
 	return Iterate<false>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
 		return manager->IsModified<DirtySource::Gpu>(offset, bytes);
@@ -114,7 +104,6 @@ bool MemoryTracker::IsRegionGpuModified(uint64_t vaddr, uint64_t size) {
 void MemoryTracker::MarkRegionAsCpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
 	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
 	Iterate<true>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
 		const auto       changed =
@@ -126,7 +115,6 @@ void MemoryTracker::MarkRegionAsCpuModified(uint64_t vaddr, uint64_t size) {
 void MemoryTracker::MarkRegionAsGpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
 	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
 	Iterate<true>(vaddr, size, [this](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
 		const auto       changed =
@@ -138,7 +126,6 @@ void MemoryTracker::MarkRegionAsGpuModified(uint64_t vaddr, uint64_t size) {
 void MemoryTracker::UnmarkRegionAsGpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
 	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
 	Iterate<true>(vaddr, size, [this](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
 		if (!manager->IsFullyModified<DirtySource::Gpu>(offset, bytes)) {
@@ -151,8 +138,6 @@ void MemoryTracker::UnmarkRegionAsGpuModified(uint64_t vaddr, uint64_t size) {
 }
 
 void MemoryTracker::UntrackMemoryLocked(uint64_t vaddr, uint64_t size) {
-	RequireMapped(vaddr, size);
-
 	std::vector<RegionManager*> managers;
 	managers.reserve((vaddr % TRACKER_REGION_SIZE + size + TRACKER_REGION_SIZE - 1) /
 	                 TRACKER_REGION_SIZE);
@@ -174,7 +159,6 @@ void MemoryTracker::UntrackMemoryLocked(uint64_t vaddr, uint64_t size) {
 		const auto changed =
 		    manager->ChangeState<DirtySource::Cpu, true>(manager->GetCpuAddr() + offset, bytes);
 		manager->ApplyProtection(changed, false);
-		manager->Untrack(manager->GetCpuAddr() + offset, bytes);
 	});
 	locks.clear();
 }
@@ -185,109 +169,4 @@ void MemoryTracker::UntrackMemory(uint64_t vaddr, uint64_t size) {
 	UntrackMemoryLocked(vaddr, size);
 }
 
-void MemoryTracker::UnmapMemory(uint64_t vaddr, uint64_t size) {
-	CheckNotInUploadCallback();
-	std::unique_lock access(m_access_mutex, std::try_to_lock);
-	if (!access.owns_lock()) {
-#if defined(KYTY_MEMORY_TRACKER_TESTS)
-		if (const auto hook = g_unmap_contention_hook.load(std::memory_order_acquire);
-		    hook != nullptr) {
-			hook();
-		}
-#endif
-		access.lock();
-	}
-	UntrackMemoryLocked(vaddr, size);
-	m_page_manager.OnGpuUnmap(vaddr, size);
-}
-
-bool MemoryTracker::InvalidateRegion(uint64_t vaddr, uint64_t size, PageFaultPhase phase) noexcept {
-	switch (phase) {
-		case PageFaultPhase::Release: return true;
-		case PageFaultPhase::Invalidate: {
-			const auto action = BeginCpuFault(vaddr, size);
-			switch (action) {
-				case CpuFaultAction::Untracked: return false;
-				case CpuFaultAction::Continue: return true;
-				case CpuFaultAction::Download:
-					EXIT("generic region invalidation cannot download GPU-dirty memory\n");
-			}
-		}
-		case PageFaultPhase::Complete:
-			return CompleteCpuFault(vaddr, size, PageFaultAccess::Write, false);
-	}
-	EXIT("unsupported region invalidation phase\n");
-}
-
-bool MemoryTracker::InvalidateVirtualGpuWrite(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-                                              PageFaultPhase phase) noexcept {
-	switch (phase) {
-		case PageFaultPhase::Release: return true;
-		case PageFaultPhase::Invalidate: {
-			const bool gpu_modified = Iterate<false>(
-			    vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-				    std::scoped_lock lock(manager->lock);
-				    return manager->IsModified<DirtySource::Gpu>(offset, bytes);
-			    });
-			if (!gpu_modified) {
-				return false;
-			}
-			const auto action = BeginCpuFault(vaddr, size);
-			if (access != PageFaultAccess::Write || action != CpuFaultAction::Download) {
-				EXIT("virtual GPU write fault requires write access to GPU-dirty memory\n");
-			}
-			return true;
-		}
-		case PageFaultPhase::Complete: {
-			if (access != PageFaultAccess::Write) {
-				EXIT("virtual GPU write completion requires write access\n");
-			}
-			bool completed = false;
-			Iterate<false>(
-			    vaddr, size, [&completed](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-				    std::scoped_lock lock(manager->lock);
-				    if (completed) {
-					    EXIT("virtual GPU write fault spans multiple tracked regions\n");
-				    }
-				    completed =
-				        manager->CompleteVirtualGpuWrite(manager->GetCpuAddr() + offset, bytes);
-			    });
-			return completed;
-		}
-	}
-	EXIT("unsupported virtual GPU write invalidation phase\n");
-}
-
-CpuFaultAction MemoryTracker::BeginCpuFault(uint64_t vaddr, uint64_t size,
-                                            PageFaultAccess access) noexcept {
-	CheckNotInUploadCallback();
-	CpuFaultAction action = CpuFaultAction::Untracked;
-	Iterate<false>(
-	    vaddr, size, [&action, access](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-		    std::scoped_lock lock(manager->lock);
-		    if (action != CpuFaultAction::Untracked) {
-			    EXIT("CPU fault spans multiple tracked regions\n");
-		    }
-		    action = manager->BeginCpuFault(manager->GetCpuAddr() + offset, bytes, access);
-	    });
-	return action;
-}
-
-bool MemoryTracker::CompleteCpuFault(uint64_t vaddr, uint64_t size, PageFaultAccess access,
-                                     bool downloaded) noexcept {
-	CheckNotInUploadCallback();
-	bool found = false;
-	Iterate<false>(
-	    vaddr, size,
-	    [&found, access, downloaded](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-		    std::scoped_lock lock(manager->lock);
-		    if (found) {
-			    EXIT("CPU fault completion spans multiple tracked regions\n");
-		    }
-		    found = manager->CompleteCpuFault(manager->GetCpuAddr() + offset, bytes, access,
-		                                      downloaded);
-	    });
-	return found;
-}
-
 } // namespace Libs::Graphics

src/graphics/host_gpu/memoryTracker.h

@@ -30,14 +30,6 @@ public:
 	void               MarkRegionAsGpuModified(uint64_t vaddr, uint64_t size);
 	void               UnmarkRegionAsGpuModified(uint64_t vaddr, uint64_t size);
 	void               UntrackMemory(uint64_t vaddr, uint64_t size);
-	void               UnmapMemory(uint64_t vaddr, uint64_t size);
-	[[nodiscard]] CpuFaultAction
-	                   BeginCpuFault(uint64_t vaddr, uint64_t size,
-	                                 PageFaultAccess access = PageFaultAccess::Write) noexcept;
-	[[nodiscard]] bool CompleteCpuFault(uint64_t vaddr, uint64_t size, PageFaultAccess access,
-	                                    bool downloaded) noexcept;
-	[[nodiscard]] bool InvalidateRegion(uint64_t vaddr, uint64_t size,
-	                                    PageFaultPhase phase) noexcept;
 	template <typename Flush>
 	void InvalidateRegion(uint64_t vaddr, uint64_t size, Flush&& on_flush) {
 		static_assert(std::is_invocable_v<Flush&>);
@@ -79,12 +71,15 @@ public:
 			EXIT("memory invalidation retained GPU-owned pages\n");
 		}
 	}
-	[[nodiscard]] bool InvalidateVirtualGpuWrite(PageFaultAccess access, uint64_t vaddr,
-	                                             uint64_t size, PageFaultPhase phase) noexcept;
+#if KYTY_BUILD == KYTY_BUILD_DEBUG
 	void ValidateGpuDirtyPages(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
 	                           const char* operation) const noexcept;
 	void ValidateGpuDirtyOwnership(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
 	                               const char* operation);
+#else
+	void ValidateGpuDirtyPages(const RangeSet&, uint64_t, uint64_t, const char*) const noexcept {}
+	void ValidateGpuDirtyOwnership(const RangeSet&, uint64_t, uint64_t, const char*) {}
+#endif
 
 	template <bool clear, typename Preflight, typename Func>
 	void ForEachDownloadRange(uint64_t vaddr, uint64_t size, Preflight&& preflight, Func&& func) {
@@ -92,7 +87,6 @@ public:
 		static_assert(std::is_nothrow_invocable_v<Func&, uint64_t, uint64_t>);
 		CheckNotInUploadCallback();
 		std::lock_guard             access(m_access_mutex);
-		RequireMapped(vaddr, size);
 		std::vector<RegionManager*> managers;
 		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t, uint64_t) {
 			managers.push_back(manager);
@@ -104,9 +98,6 @@ public:
 		}
 		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 			const auto address = manager->GetCpuAddr() + offset;
-			if (manager->HasPendingFault(address, bytes)) {
-				EXIT("GPU download synchronization raced a pending CPU fault\n");
-			}
 			manager->template ForEachModifiedRange<DirtySource::Gpu, false>(address, bytes,
 			                                                                preflight);
 		});
@@ -132,11 +123,6 @@ public:
 		    vaddr, size, [](uint64_t, uint64_t) noexcept {}, std::forward<Func>(func));
 	}
 
-#if defined(KYTY_MEMORY_TRACKER_TESTS)
-	using UnmapContentionHook = void (*)() noexcept;
-	static void SetUnmapContentionHook(UnmapContentionHook hook) noexcept;
-#endif
-
 	template <typename RangeFunc, typename UploadFunc>
 	void ForEachUploadRange(uint64_t vaddr, uint64_t size, bool is_written, RangeFunc&& range_func,
 	                        UploadFunc&& upload_func) {
@@ -144,12 +130,10 @@ public:
 		static_assert(std::is_nothrow_invocable_v<UploadFunc&>);
 		CheckNotInUploadCallback();
 		std::unique_lock access(m_access_mutex);
-		RequireMapped(vaddr, size);
 		Iterate<true>(vaddr, size, [](RegionManager*, uint64_t, uint64_t) {});
 		const auto* previous_upload_owner = std::exchange(s_upload_owner, this);
 		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 			manager->lock.lock();
-			manager->Track(manager->GetCpuAddr() + offset, bytes);
 			manager->ForEachModifiedRange<DirtySource::Cpu, true>(manager->GetCpuAddr() + offset,
 			                                                      bytes, range_func);
 			if (!is_written) {
@@ -211,14 +195,6 @@ private:
 
 	static void    ValidateRange(uint64_t vaddr, uint64_t size);
 	void           UntrackMemoryLocked(uint64_t vaddr, uint64_t size);
-	void        RequireMapped(uint64_t vaddr, uint64_t size) const {
-		ValidateRange(vaddr, size);
-		if (!m_page_manager.IsMapped(vaddr, size)) {
-			EXIT("memory tracker range [0x%llx, 0x%llx) is not mapped\n",
-			     static_cast<unsigned long long>(vaddr),
-			     static_cast<unsigned long long>(vaddr + size));
-		}
-	}
 	RegionManager* GetOrCreateRegion(uint64_t index);
 
 	std::unique_ptr<std::atomic<RegionManager*>[]> m_regions;

src/graphics/host_gpu/pageManager.cpp

@@ -1,6 +1,7 @@
 #include "graphics/host_gpu/pageManager.h"
 
 #include "graphics/host_gpu/regionDefinitions.h"
+#include "kernel/memory.h"
 
 #include <algorithm>
 #include <array>
@@ -21,16 +22,11 @@
 #undef min
 #undef max
 #elif defined(__APPLE__)
-#include <mach/mach.h>
-#include <mach/mach_vm.h>
 #include <pthread.h>
 #include <sys/mman.h>
 #include <unistd.h>
 #else
-#include <cerrno>
-#include <cstring>
 #include <execinfo.h>
-#include <fcntl.h>
 #include <sys/mman.h>
 #include <sys/syscall.h>
 #include <unistd.h>
@@ -57,47 +53,8 @@ constexpr uint64_t REGION_PAGES = REGION_SIZE / PAGE_SIZE;
 constexpr uint32_t NO_ACCESS_PROTECTION  = PAGE_NOACCESS;
 constexpr uint32_t READ_ONLY_PROTECTION  = PAGE_READONLY;
 constexpr uint32_t READ_WRITE_PROTECTION = PAGE_READWRITE;
-
-#if defined(__APPLE__)
-// Map the tracker's Win32-style protection tags to POSIX mprotect flags.
-static int PageProtToPosix(uint32_t protection) {
-	switch (protection) {
-		case PAGE_NOACCESS: return PROT_NONE;
-		case PAGE_READONLY: return PROT_READ;
-		case PAGE_READWRITE: return PROT_READ | PROT_WRITE;
-		default: return PROT_NONE;
-	}
-}
-
-// Query the current protection of the page containing vaddr via the Mach VM map and
-// collapse it to the tracker's read/write tags (execute is irrelevant to write tracking).
-static uint32_t MachQueryPageProt(uint64_t vaddr) {
-	auto                           region_addr = static_cast<mach_vm_address_t>(vaddr);
-	mach_vm_size_t                 region_size = 0;
-	vm_region_basic_info_data_64_t info {};
-	mach_msg_type_number_t         count       = VM_REGION_BASIC_INFO_COUNT_64;
-	mach_port_t                    object_name = MACH_PORT_NULL;
-
-	kern_return_t kr =
-	    mach_vm_region(mach_task_self(), &region_addr, &region_size, VM_REGION_BASIC_INFO_64,
-	                   reinterpret_cast<vm_region_info_t>(&info), &count, &object_name);
-	if (kr != KERN_SUCCESS || region_addr > vaddr) {
-		return PAGE_NOACCESS; // no region covering vaddr
-	}
-	if ((info.protection & VM_PROT_WRITE) != 0) {
-		return PAGE_READWRITE;
-	}
-	if ((info.protection & VM_PROT_READ) != 0) {
-		return PAGE_READONLY;
-	}
-	return PAGE_NOACCESS;
-}
-#elif defined(__linux__)
 // Zero is the unknown protection sentinel.
 constexpr uint32_t UNKNOWN_PROTECTION = 0;
-#endif
-
-thread_local bool g_in_fault_resolution = false;
 
 [[noreturn]] void FailFast(const char* reason = nullptr) noexcept {
 	std::fputs("PageManager fail-fast: ", stderr);
@@ -136,6 +93,15 @@ thread_local bool g_in_fault_resolution = false;
 	std::_Exit(322);
 }
 
+Common::VirtualMemory::Mode ToMemoryMode(uint32_t protection) {
+	switch (protection) {
+		case NO_ACCESS_PROTECTION: return Common::VirtualMemory::Mode::NoAccess;
+		case READ_ONLY_PROTECTION: return Common::VirtualMemory::Mode::Read;
+		case READ_WRITE_PROTECTION: return Common::VirtualMemory::Mode::ReadWrite;
+		default: Fatal("unmappable protection 0x%08" PRIx32, protection);
+	}
+}
+
 uint32_t CurrentThread() noexcept {
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
 	return GetCurrentThreadId();
@@ -155,130 +121,6 @@ uint32_t CurrentThread() noexcept {
 #endif
 }
 
-#if defined(__linux__)
-int ToHostProtection(uint32_t protection) {
-	switch (protection) {
-		case NO_ACCESS_PROTECTION: return PROT_NONE;
-		case READ_ONLY_PROTECTION: return PROT_READ;
-		case READ_WRITE_PROTECTION: return PROT_READ | PROT_WRITE;
-		default: Fatal("unmappable protection 0x%08" PRIx32, protection);
-	}
-}
-
-struct HostMapping {
-	uint64_t end        = 0;
-	uint32_t protection = UNKNOWN_PROTECTION;
-};
-
-// Async-signal-safe lookup in the address-ordered /proc/self/maps.
-HostMapping QueryHostMapping(uint64_t vaddr) noexcept {
-	int fd = ::open("/proc/self/maps", O_RDONLY | O_CLOEXEC); // NOLINT
-	if (fd < 0) {
-		return {};
-	}
-
-	enum class Field { Start, End, Perms, Rest };
-
-	HostMapping result {};
-	auto        field      = Field::Start;
-	uint64_t    start      = 0;
-	uint64_t    end        = 0;
-	char        perms[4]   = {};
-	uint32_t    perms_len  = 0;
-	bool        line_valid = true;
-
-	char buffer[8192];
-
-	for (bool done = false; !done;) {
-		const auto got = ::read(fd, buffer, sizeof(buffer));
-		if (got < 0) {
-			if (errno == EINTR) {
-				continue;
-			}
-			break;
-		}
-		if (got == 0) {
-			break;
-		}
-
-		for (ssize_t i = 0; i < got && !done; i++) {
-			const char c = buffer[i];
-
-			if (c == '\n') {
-				field      = Field::Start;
-				start      = 0;
-				end        = 0;
-				perms_len  = 0;
-				line_valid = true;
-				continue;
-			}
-
-			if (!line_valid) {
-				continue;
-			}
-
-			switch (field) {
-				case Field::Start:
-				case Field::End: {
-					uint64_t digit = 0;
-					if (c >= '0' && c <= '9') {
-						digit = static_cast<uint64_t>(c - '0');
-					} else if (c >= 'a' && c <= 'f') {
-						digit = static_cast<uint64_t>(c - 'a') + 10;
-					} else if (c == '-' && field == Field::Start) {
-						field = Field::End;
-						break;
-					} else if (c == ' ' && field == Field::End) {
-						field     = Field::Perms;
-						perms_len = 0;
-						break;
-					} else {
-						line_valid = false;
-						break;
-					}
-
-					auto& value = (field == Field::Start ? start : end);
-					value       = (value << 4u) | digit;
-					break;
-				}
-
-				case Field::Perms: {
-					if (c != ' ') {
-						if (perms_len < sizeof(perms)) {
-							perms[perms_len] = c;
-						}
-						perms_len++;
-						break;
-					}
-
-					if (vaddr < start) {
-						done = true;
-					} else if (vaddr < end && perms_len >= 2) {
-						result.end        = end;
-						result.protection = perms[1] == 'w'   ? READ_WRITE_PROTECTION
-						                    : perms[0] == 'r' ? READ_ONLY_PROTECTION
-						                                      : NO_ACCESS_PROTECTION;
-						done              = true;
-					} else {
-						field = Field::Rest;
-					}
-					break;
-				}
-
-				case Field::Rest: break;
-			}
-		}
-	}
-
-	::close(fd);
-	return result;
-}
-
-uint32_t QueryHostProtection(uint64_t vaddr) noexcept {
-	return QueryHostMapping(vaddr).protection;
-}
-#endif
-
 class SpinGuard final {
 public:
 	explicit SpinGuard(std::atomic_flag& lock): m_lock(lock) {
@@ -313,21 +155,13 @@ uint64_t PageEnd(uint64_t vaddr, uint64_t size) {
 struct PageManager::Impl {
 	struct PageState {
 		std::atomic_flag lock                = ATOMIC_FLAG_INIT;
-		uint32_t         mappings            = 0;
-		uint32_t         gpu_read_mappings   = 0;
-		uint32_t         gpu_write_mappings  = 0;
 		uint32_t         write_watchers      = 0;
 		uint32_t         access_watchers     = 0;
 		uint32_t         original_protection = 0;
 		uint32_t         backing_writer      = 0;
-#if defined(__linux__)
 		// Shadow the protection applied through Protect().
 		uint32_t current_protection = UNKNOWN_PROTECTION;
-#endif
 		bool     resolving          = false;
-		bool resolving_read_write = false;
-		bool late_read_pending    = false;
-		bool late_write_pending   = false;
 	};
 
 	struct Region {
@@ -354,10 +188,7 @@ struct PageManager::Impl {
 		std::span<PageState*> m_pages;
 	};
 
-	Impl(PageFaultHandler handler, void* context): fault_handler(handler), fault_context(context) {
-		if (fault_handler == nullptr) {
-			Fatal("null fault handler");
-		}
+	Impl() {
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
 		SYSTEM_INFO info {};
 		GetSystemInfo(&info);
@@ -386,9 +217,8 @@ struct PageManager::Impl {
 		for (const auto& region: region_storage) {
 			for (auto& page: region->pages) {
 				SpinGuard lock(page.lock);
-				if (page.mappings != 0 || page.gpu_read_mappings != 0 ||
-				    page.gpu_write_mappings != 0 || page.write_watchers != 0 ||
-				    page.access_watchers != 0 || page.backing_writer != 0 || page.resolving) {
+				if (page.write_watchers != 0 || page.access_watchers != 0 ||
+				    page.backing_writer != 0 || page.resolving) {
 					FailFast("PageManager destroyed with live page state");
 				}
 			}
@@ -430,300 +260,59 @@ struct PageManager::Impl {
 		return page.original_protection;
 	}
 
-	static void PublishDelayedFaults(PageState& page, uint32_t old_protection,
-	                                 uint32_t new_protection) {
-		if (old_protection == NO_ACCESS_PROTECTION && new_protection != NO_ACCESS_PROTECTION) {
-			page.late_read_pending = true;
-		}
-		if ((old_protection == NO_ACCESS_PROTECTION || old_protection == READ_ONLY_PROTECTION) &&
-		    new_protection == READ_WRITE_PROTECTION) {
-			page.late_write_pending = true;
-		}
-	}
-
-	static void ValidateInitialProtection(std::span<PageState*> pages, uint64_t vaddr) {
-		const auto end = vaddr + pages.size() * PAGE_SIZE;
-#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
-		for (auto address = vaddr; address < end;) {
-			MEMORY_BASIC_INFORMATION info {};
-			if (VirtualQuery(reinterpret_cast<const void*>(static_cast<uintptr_t>(address)), &info,
-			                 sizeof(info)) == 0 ||
-			    info.State != MEM_COMMIT || info.Protect != PAGE_READWRITE) {
-				Fatal("basic path requires PAGE_READWRITE at 0x%016" PRIx64 " (state=0x%08" PRIx32
-				      ", protection=0x%08" PRIx32 ")",
-				      address, static_cast<uint32_t>(info.State),
-				      static_cast<uint32_t>(info.Protect));
-			}
-			const auto region_end = reinterpret_cast<uint64_t>(info.BaseAddress) + info.RegionSize;
-			if (region_end <= address) {
-				Fatal("VirtualQuery returned an invalid region at 0x%016" PRIx64, address);
-			}
-			address = std::min(end, region_end);
-		}
-#elif defined(__APPLE__)
-		for (auto address = vaddr; address < end; address += PAGE_SIZE) {
-			const uint32_t protection = MachQueryPageProt(address);
-			if (protection != PAGE_READWRITE) {
-				Fatal("basic path requires PAGE_READWRITE at 0x%016" PRIx64
-				      " (protection=0x%08" PRIx32 ")",
-				      address, protection);
-			}
-		}
-#else
-		for (auto address = vaddr; address < end;) {
-			const auto mapping = QueryHostMapping(address);
-			if (mapping.protection != READ_WRITE_PROTECTION || mapping.end <= address) {
-				Fatal("basic path requires a read/write mapping at 0x%016" PRIx64
-				      " (protection=0x%08" PRIx32 ")",
-				      address, mapping.protection);
-			}
-			address = std::min(end, mapping.end);
-		}
-		for (auto* page: pages) {
-			page->current_protection = READ_WRITE_PROTECTION;
-		}
-#endif
+	static void InitializeProtection(std::span<PageState*> pages) {
 		for (auto* page: pages) {
 			page->original_protection = READ_WRITE_PROTECTION;
+			page->current_protection  = READ_WRITE_PROTECTION;
 		}
 	}
 
-	static bool AllowsAccess([[maybe_unused]] const PageState& page, uint64_t vaddr,
-	                         PageFaultAccess access) noexcept {
-#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
-		MEMORY_BASIC_INFORMATION info {};
-		if (VirtualQuery(reinterpret_cast<const void*>(static_cast<uintptr_t>(vaddr)), &info,
-		                 sizeof(info)) == 0 ||
-		    info.State != MEM_COMMIT) {
-			return false;
-		}
-		switch (access) {
-			case PageFaultAccess::Read:
-				return info.Protect == PAGE_READONLY || info.Protect == PAGE_READWRITE;
-			case PageFaultAccess::Write: return info.Protect == PAGE_READWRITE;
-			default: return false;
-		}
-#elif defined(__APPLE__)
-		const uint32_t protection = MachQueryPageProt(vaddr);
-		switch (access) {
-			case PageFaultAccess::Read:
-				return protection == PAGE_READONLY || protection == PAGE_READWRITE;
-			case PageFaultAccess::Write: return protection == PAGE_READWRITE;
-			default: return false;
-		}
-#else
-		const auto permitted = [](uint32_t protection, PageFaultAccess wanted) {
-			switch (wanted) {
-				case PageFaultAccess::Read:
-					return protection == READ_ONLY_PROTECTION ||
-					       protection == READ_WRITE_PROTECTION;
-				case PageFaultAccess::Write: return protection == READ_WRITE_PROTECTION;
-				default: return false;
-			}
-		};
-
-		if (!permitted(page.current_protection, access)) {
-			return false;
-		}
-		return permitted(QueryHostProtection(vaddr), access);
-#endif
-	}
-
-	static void ProtectRange(std::span<PageState*> pages, uint64_t vaddr, uint32_t protection,
-	                         std::span<const uint32_t> expected_old, bool fault_path) noexcept {
+	void ProtectRange(std::span<PageState*> pages, uint64_t vaddr, uint32_t protection,
+	                  std::span<const uint32_t> expected_old) noexcept {
 		const auto size = pages.size() * PAGE_SIZE;
 		if (pages.size() != expected_old.size()) {
 			FailFast("protection range state size mismatch");
 		}
-#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
-		struct HostRange {
-			uint64_t begin = 0;
-			uint64_t end   = 0;
-		};
-		std::vector<HostRange> host_ranges;
-		const auto             end = vaddr + size;
-		for (auto address = vaddr; address < end;) {
-			MEMORY_BASIC_INFORMATION info {};
-			if (VirtualQuery(reinterpret_cast<const void*>(static_cast<uintptr_t>(address)), &info,
-			                 sizeof(info)) == 0 ||
-			    info.State != MEM_COMMIT) {
-				if (fault_path) {
-					FailFast("VirtualProtect fault transition did not match expected protection");
-				}
-				Fatal("invalid protection transition at 0x%016" PRIx64 ", state=0x%08" PRIx32
-				      ", new=0x%08" PRIx32,
-				      address, static_cast<uint32_t>(info.State), protection);
-			}
-			const auto region_end = reinterpret_cast<uint64_t>(info.BaseAddress) + info.RegionSize;
-			const auto query_end  = std::min(end, region_end);
-			if (query_end <= address) {
-				if (fault_path) {
-					FailFast("VirtualQuery returned an invalid fault transition region");
-				}
-				Fatal("VirtualQuery returned an invalid region at 0x%016" PRIx64, address);
-			}
-			const auto first_page = static_cast<size_t>((address - vaddr) / PAGE_SIZE);
-			const auto last_page =
-			    static_cast<size_t>((query_end - vaddr + PAGE_SIZE - 1) / PAGE_SIZE);
-			for (auto page = first_page; page < last_page; page++) {
-				if (info.Protect != expected_old[page]) {
-					if (fault_path) {
-						FailFast(
-						    "VirtualProtect fault transition did not match expected protection");
-					}
-					Fatal("invalid protection transition at 0x%016" PRIx64 ", actual=0x%08" PRIx32
-					      ", expected=0x%08" PRIx32 ", new=0x%08" PRIx32,
-					      vaddr + page * PAGE_SIZE, static_cast<uint32_t>(info.Protect),
-					      expected_old[page], protection);
-				}
-			}
-			const auto allocation = reinterpret_cast<uint64_t>(info.AllocationBase);
-			if (host_ranges.empty() || allocation != host_ranges.back().begin) {
-				host_ranges.push_back({allocation, query_end});
-			} else {
-				host_ranges.back().end = query_end;
-			}
-			address = query_end;
-		}
-		for (auto range: host_ranges) {
-			range.begin               = std::max(range.begin, vaddr);
-			DWORD      old_protection = 0;
-			const auto first_page     = static_cast<size_t>((range.begin - vaddr) / PAGE_SIZE);
-			if (VirtualProtect(reinterpret_cast<void*>(static_cast<uintptr_t>(range.begin)),
-			                   range.end - range.begin, protection, &old_protection) == 0 ||
-			    old_protection != expected_old[first_page]) {
-				if (fault_path) {
-					FailFast("VirtualProtect fault transition did not match expected protection");
-				}
-				Fatal("invalid protection transition at 0x%016" PRIx64 ", old=0x%08" PRIx32
-				      ", expected=0x%08" PRIx32 ", new=0x%08" PRIx32,
-				      range.begin, static_cast<uint32_t>(old_protection), expected_old[first_page],
-				      protection);
-			}
-		}
-#elif defined(__APPLE__)
-		// mprotect cannot report the previous protection, so the expected_old comparison
-		// is dropped; the tracker is the sole mutator of these pages and drives the
-		// transition from its own shadow state.
-		(void)expected_old;
-		if (mprotect(reinterpret_cast<void*>(static_cast<uintptr_t>(vaddr)), size,
-		             PageProtToPosix(protection)) != 0) {
-			if (fault_path) {
-				FailFast("mprotect fault transition failed");
-			}
-			Fatal("mprotect failed at 0x%016" PRIx64 ", new=0x%08" PRIx32, vaddr, protection);
-		}
-#else
 		for (size_t i = 0; i < pages.size(); i++) {
 			const auto actual = pages[i]->current_protection;
 			if (actual != UNKNOWN_PROTECTION && actual != expected_old[i]) {
-				if (fault_path) {
-					FailFast("mprotect fault transition did not match expected protection");
-				}
 				Fatal("invalid protection transition at 0x%016" PRIx64 ", old=0x%08" PRIx32
 				      ", expected=0x%08" PRIx32 ", new=0x%08" PRIx32,
 				      vaddr + i * PAGE_SIZE, actual, expected_old[i], protection);
 			}
 		}
-		if (::mprotect(reinterpret_cast<void*>(static_cast<uintptr_t>(vaddr)), size,
-		               ToHostProtection(protection)) != 0) {
-			if (fault_path) {
-				FailFast("mprotect failed on the fault path");
-			}
-			Fatal("mprotect failed at 0x%016" PRIx64 ", new=0x%08" PRIx32 " (%s)", vaddr,
-			      protection, std::strerror(errno));
+		if (!Libs::LibKernel::Memory::ProtectGuestHostMemory(vaddr, size,
+		                                                     ToMemoryMode(protection))) {
+			Fatal("address-space protection failed at 0x%016" PRIx64 ", new=0x%08" PRIx32, vaddr,
+			      protection);
 		}
 		for (auto* page: pages) {
 			page->current_protection = protection;
 		}
-#endif
 	}
 
-	static void Protect(PageState& page, uint64_t vaddr, uint32_t protection, uint32_t expected_old,
-	                    bool fault_path) noexcept {
+	void Protect(PageState& page, uint64_t vaddr, uint32_t protection,
+	             uint32_t expected_old) noexcept {
 		PageState* pages[]    = {&page};
 		uint32_t   expected[] = {expected_old};
-		ProtectRange(pages, vaddr, protection, expected, fault_path);
+		ProtectRange(pages, vaddr, protection, expected);
 	}
 
 	std::unique_ptr<std::atomic<Region*>[]> regions;
 	std::vector<std::unique_ptr<Region>>    region_storage;
 	std::mutex                              region_mutex;
-	PageFaultHandler                        fault_handler = nullptr;
-	void*                                   fault_context = nullptr;
 };
 
 static_assert(std::atomic<void*>::is_always_lock_free);
 
-PageManager::PageManager(PageFaultHandler fault_handler, void* fault_context)
-    : m_impl(std::make_unique<Impl>(fault_handler, fault_context)) {}
+PageManager::PageManager(): m_impl(std::make_unique<Impl>()) {}
 
 PageManager::~PageManager() = default;
 
 uint64_t PageManager::GetPageSize() const {
-	if (g_in_fault_resolution) {
-		FailFast("nested page fault while resolving a watched page");
-	}
 	return PAGE_SIZE;
 }
 
-bool PageManager::IsTracked(uint64_t vaddr) const noexcept {
-	if (g_in_fault_resolution) {
-		FailFast("IsTracked called during fault resolution");
-	}
-	auto* region = m_impl->FindRegion(vaddr);
-	if (region == nullptr) {
-		return false;
-	}
-	auto&     page = m_impl->GetPage(*region, vaddr);
-	SpinGuard lock(page.lock);
-	return page.write_watchers != 0 || page.access_watchers != 0;
-}
-
-bool PageManager::IsMapped(uint64_t vaddr, uint64_t size) const noexcept {
-	if (vaddr == 0 || size == 0 || vaddr >= ADDRESS_SIZE || size > ADDRESS_SIZE - vaddr) {
-		return false;
-	}
-	const auto end = PageStart(vaddr + size - 1) + PAGE_SIZE;
-	for (auto page_vaddr = PageStart(vaddr); page_vaddr < end; page_vaddr += PAGE_SIZE) {
-		auto* region = m_impl->FindRegion(page_vaddr);
-		if (region == nullptr) {
-			return false;
-		}
-		auto&     page = m_impl->GetPage(*region, page_vaddr);
-		SpinGuard lock(page.lock);
-		if (page.mappings == 0) {
-			return false;
-		}
-	}
-	return true;
-}
-
-bool PageManager::HasGpuAccess(uint64_t vaddr, uint64_t size, GpuAccess access) const noexcept {
-	if (access != GpuAccess::Read && access != GpuAccess::Write && access != GpuAccess::ReadWrite) {
-		FailFast("HasGpuAccess received an invalid GPU access mode");
-	}
-	const bool need_read  = access == GpuAccess::Read || access == GpuAccess::ReadWrite;
-	const bool need_write = access == GpuAccess::Write || access == GpuAccess::ReadWrite;
-	if (vaddr == 0 || size == 0 || vaddr >= ADDRESS_SIZE || size > ADDRESS_SIZE - vaddr) {
-		return false;
-	}
-	const auto end = PageEnd(vaddr, size);
-	for (auto addr = PageStart(vaddr); addr < end; addr += PAGE_SIZE) {
-		auto* region = m_impl->FindRegion(addr);
-		if (region == nullptr) {
-			return false;
-		}
-		auto&     page = m_impl->GetPage(*region, addr);
-		SpinGuard lock(page.lock);
-		if ((need_read && page.gpu_read_mappings == 0) ||
-		    (need_write && page.gpu_write_mappings == 0)) {
-			return false;
-		}
-	}
-	return true;
-}
-
 void PageManager::UpdatePageWatchers(bool track, uint64_t vaddr, uint64_t size,
                                      PageWatchMode mode) {
 	if (mode != PageWatchMode::Write && mode != PageWatchMode::ReadWrite) {
@@ -754,9 +343,6 @@ void PageManager::UpdatePageWatchers(bool track, uint64_t vaddr, uint64_t size,
 			if (page.resolving && track) {
 				FailFast("new page watcher raced active fault resolution");
 			}
-			if (page.mappings == 0) {
-				Fatal("watching unmapped page 0x%016" PRIx64, address);
-			}
 			auto& watchers =
 			    (mode == PageWatchMode::ReadWrite ? page.access_watchers : page.write_watchers);
 			if (track) {
@@ -784,8 +370,7 @@ void PageManager::UpdatePageWatchers(bool track, uint64_t vaddr, uint64_t size,
 					last++;
 				}
 				if (first != last) {
-					Impl::ValidateInitialProtection(std::span {pages}.subspan(first, last - first),
-					                                chunk_begin + first * PAGE_SIZE);
+					Impl::InitializeProtection(std::span {pages}.subspan(first, last - first));
 				}
 				first = last;
 			}
@@ -831,99 +416,25 @@ void PageManager::UpdatePageWatchers(bool track, uint64_t vaddr, uint64_t size,
 					last = current + 1;
 				}
 			}
-			Impl::ProtectRange(std::span {pages}.subspan(first, last - first),
+			m_impl->ProtectRange(std::span {pages}.subspan(first, last - first),
 			                     chunk_begin + first * PAGE_SIZE, protection,
-			                   std::span {old_protections}.subspan(first, last - first), false);
+			                     std::span {old_protections}.subspan(first, last - first));
 			first = current;
 		}
 
-		for (size_t i = 0; i < page_count; i++) {
-			auto&      page       = *pages[i];
-			const auto protection = new_protections[i];
-			if (track) {
-				switch (protection) {
-					case NO_ACCESS_PROTECTION:
-						page.late_read_pending  = false;
-						page.late_write_pending = false;
-						break;
-					case READ_ONLY_PROTECTION: page.late_write_pending = false; break;
-					default: break;
-				}
-			} else if (page.backing_writer == 0) {
-				Impl::PublishDelayedFaults(page, old_protections[i], protection);
-				if (page.write_watchers == 0 && page.access_watchers == 0) {
-					page.original_protection = 0;
-				}
+		for (auto* page: pages) {
+			if (!track && page->backing_writer == 0 && page->write_watchers == 0 &&
+			    page->access_watchers == 0) {
+				page->original_protection = 0;
 			}
 		}
 		chunk_begin = chunk_end;
 	}
 }
 
-void PageManager::OnGpuMap(uint64_t vaddr, uint64_t size, GpuAccess access) {
-	if (g_in_fault_resolution) {
-		FailFast("GPU mapping changed during fault resolution");
-	}
-	if (access != GpuAccess::Read && access != GpuAccess::Write && access != GpuAccess::ReadWrite) {
-		FailFast("GPU map received an invalid access mode");
-	}
-	const bool gpu_read  = access == GpuAccess::Read || access == GpuAccess::ReadWrite;
-	const bool gpu_write = access == GpuAccess::Write || access == GpuAccess::ReadWrite;
-	const auto end       = PageEnd(vaddr, size);
-	for (auto addr = PageStart(vaddr); addr < end; addr += PAGE_SIZE) {
-		auto&     page = m_impl->GetPage(*m_impl->GetOrCreateRegion(addr), addr);
-		SpinGuard lock(page.lock);
-		if (page.resolving || page.mappings == std::numeric_limits<uint32_t>::max() ||
-		    (gpu_read && page.gpu_read_mappings == std::numeric_limits<uint32_t>::max()) ||
-		    (gpu_write && page.gpu_write_mappings == std::numeric_limits<uint32_t>::max())) {
-			Fatal("invalid map state at 0x%016" PRIx64, addr);
-		}
-		page.mappings++;
-		page.gpu_read_mappings += gpu_read ? 1u : 0u;
-		page.gpu_write_mappings += gpu_write ? 1u : 0u;
-#if defined(__linux__)
-		// New guest mappings start read/write.
-		if (page.current_protection == UNKNOWN_PROTECTION) {
-			page.current_protection = READ_WRITE_PROTECTION;
-		}
-#endif
-	}
-}
+void PageManager::OnGpuMap(uint64_t, uint64_t) {}
 
-void PageManager::OnGpuUnmap(uint64_t vaddr, uint64_t size, GpuAccess access) {
-	if (g_in_fault_resolution) {
-		FailFast("GPU unmapping changed during fault resolution");
-	}
-	if (access != GpuAccess::Read && access != GpuAccess::Write && access != GpuAccess::ReadWrite) {
-		FailFast("GPU unmap received an invalid access mode");
-	}
-	const bool gpu_read  = access == GpuAccess::Read || access == GpuAccess::ReadWrite;
-	const bool gpu_write = access == GpuAccess::Write || access == GpuAccess::ReadWrite;
-	const auto end       = PageEnd(vaddr, size);
-	for (auto page_vaddr = PageStart(vaddr); page_vaddr < end; page_vaddr += PAGE_SIZE) {
-		auto* region = m_impl->FindRegion(page_vaddr);
-		if (region == nullptr) {
-			Fatal("unmapping unknown page 0x%016" PRIx64, page_vaddr);
-		}
-		auto&     page = m_impl->GetPage(*region, page_vaddr);
-		SpinGuard lock(page.lock);
-		if (page.resolving || page.mappings == 0 || (gpu_read && page.gpu_read_mappings == 0) ||
-		    (gpu_write && page.gpu_write_mappings == 0) ||
-		    (page.mappings == 1 && (page.write_watchers != 0 || page.access_watchers != 0))) {
-			Fatal("invalid unmap state at 0x%016" PRIx64, page_vaddr);
-		}
-		page.mappings--;
-		page.gpu_read_mappings -= gpu_read ? 1u : 0u;
-		page.gpu_write_mappings -= gpu_write ? 1u : 0u;
-		if (page.mappings == 0) {
-			if (page.gpu_read_mappings != 0 || page.gpu_write_mappings != 0) {
-				FailFast("GPU unmap left nonzero GPU mapping counts");
-			}
-			page.late_read_pending  = false;
-			page.late_write_pending = false;
-		}
-	}
-}
+void PageManager::OnGpuUnmap(uint64_t, uint64_t) {}
 
 PageManager::BackingWrite::BackingWrite(PageManager& manager, uint64_t vaddr,
                                         uint64_t size) noexcept
@@ -967,9 +478,6 @@ PageManager::ReserveBackingWrites(std::span<const RangeSet::Range> ranges) {
 }
 
 void PageManager::BeginBackingWrite(uint64_t vaddr, uint64_t size) noexcept {
-	if (g_in_fault_resolution) {
-		FailFast("backing write began during fault resolution");
-	}
 	const auto end    = PageEnd(vaddr, size);
 	const auto writer = CurrentThread();
 	for (auto address = PageStart(vaddr); address < end; address += PAGE_SIZE) {
@@ -979,20 +487,15 @@ void PageManager::BeginBackingWrite(uint64_t vaddr, uint64_t size) noexcept {
 		}
 		auto&     page = m_impl->GetPage(*region, address);
 		SpinGuard lock(page.lock);
-		if (page.mappings == 0 || page.resolving || page.backing_writer != 0 ||
-		    page.access_watchers == 0) {
+		if (page.resolving || page.backing_writer != 0 || page.access_watchers == 0) {
 			Fatal("backing write races page resolution at 0x%016" PRIx64, address);
 		}
 		page.resolving      = true;
-		page.resolving_read_write = true;
 		page.backing_writer = writer;
 	}
 }
 
 void PageManager::EndBackingWrite(uint64_t vaddr, uint64_t size) noexcept {
-	if (g_in_fault_resolution) {
-		FailFast("backing write ended during fault resolution");
-	}
 	const auto end    = PageEnd(vaddr, size);
 	const auto writer = CurrentThread();
 	for (auto address = PageStart(vaddr); address < end; address += PAGE_SIZE) {
@@ -1008,126 +511,14 @@ void PageManager::EndBackingWrite(uint64_t vaddr, uint64_t size) noexcept {
 		const auto old_protection = NO_ACCESS_PROTECTION;
 		const auto new_protection = Impl::WatcherProtection(page);
 		if (new_protection != old_protection) {
-			Impl::Protect(page, address, new_protection, old_protection, false);
+			m_impl->Protect(page, address, new_protection, old_protection);
 		}
-		Impl::PublishDelayedFaults(page, old_protection, new_protection);
 		if (page.write_watchers == 0 && page.access_watchers == 0) {
 			page.original_protection = 0;
 		}
 		page.backing_writer = 0;
 		page.resolving      = false;
-		page.resolving_read_write = false;
 	}
 }
 
-bool PageManager::HandleFault(PageFaultAccess access, uint64_t fault_vaddr) noexcept {
-	if (g_in_fault_resolution) {
-		FailFast("nested HandleFault call");
-	}
-	auto* region = m_impl->FindRegion(fault_vaddr);
-	if (region == nullptr) {
-		return false;
-	}
-	auto& page   = m_impl->GetPage(*region, fault_vaddr);
-	bool  waited = false;
-	while (true) {
-		SpinGuard lock(page.lock);
-		if (access == PageFaultAccess::Read && page.late_read_pending &&
-		    Impl::AllowsAccess(page, fault_vaddr, access)) {
-			page.late_read_pending = false;
-			return true;
-		}
-		if (access == PageFaultAccess::Write && page.late_write_pending &&
-		    Impl::AllowsAccess(page, fault_vaddr, access)) {
-			page.late_write_pending = false;
-			return true;
-		}
-		if (page.resolving) {
-			if (page.backing_writer == CurrentThread()) {
-				FailFast("backing writer faulted on its own reserved page");
-			}
-			if ((!page.resolving_read_write && access != PageFaultAccess::Write) ||
-			    (page.resolving_read_write && access != PageFaultAccess::Read &&
-			     access != PageFaultAccess::Write)) {
-				FailFast("fault access is incompatible with the active resolver");
-			}
-			waited = true;
-			continue;
-		}
-		if (page.write_watchers == 0 && page.access_watchers == 0) {
-			if (access != PageFaultAccess::Read && access != PageFaultAccess::Write) {
-				return false;
-			}
-			bool&      pending = (access == PageFaultAccess::Read ? page.late_read_pending
-			                                                      : page.late_write_pending);
-			const bool allowed = Impl::AllowsAccess(page, fault_vaddr, access);
-			pending            = false;
-			if (waited && !allowed) {
-				FailFast("page remained inaccessible after waiting for its resolver");
-			}
-			// More than one CPU can fault before a protection transition becomes visible. The first
-			// delayed fault consumes the hint bit; later faults must also resume once the mapped
-			// page already permits the requested access. A genuinely read-only/no-access page still
-			// falls through to the guest exception path.
-			return allowed;
-		}
-		if ((access != PageFaultAccess::Read && access != PageFaultAccess::Write) ||
-		    (access == PageFaultAccess::Read && page.access_watchers == 0)) {
-			FailFast("fault access is incompatible with active page watchers");
-		}
-		page.resolving            = true;
-		page.resolving_read_write = page.access_watchers != 0;
-		break;
-	}
-	g_in_fault_resolution = true;
-	const bool handled    = m_impl->fault_handler(m_impl->fault_context, access, fault_vaddr, 1,
-	                                              PageFaultPhase::Invalidate);
-	g_in_fault_resolution = false;
-	{
-		SpinGuard lock(page.lock);
-		if (!handled || !page.resolving) {
-			FailFast("fault invalidation did not preserve the resolving state");
-		}
-	}
-	g_in_fault_resolution = true;
-	const bool completed  = m_impl->fault_handler(m_impl->fault_context, access, fault_vaddr, 1,
-	                                              PageFaultPhase::Complete);
-	g_in_fault_resolution = false;
-	{
-		SpinGuard lock(page.lock);
-		if (!completed || !page.resolving) {
-			FailFast("fault completion did not preserve the resolving state");
-		}
-		if (page.write_watchers != 0 || page.access_watchers != 0) {
-			const auto old_protection  = Impl::WatcherProtection(page);
-			const bool read_only_fault = access == PageFaultAccess::Read;
-			if (read_only_fault && page.access_watchers == 0) {
-				FailFast("read fault completed without a read/write watcher");
-			}
-			page.access_watchers = 0;
-			if (!read_only_fault) {
-				page.write_watchers = 0;
-			}
-			const auto restored_protection = Impl::WatcherProtection(page);
-			Impl::Protect(page, PageStart(fault_vaddr), restored_protection, old_protection, true);
-			if (page.write_watchers == 0) {
-				page.original_protection = 0;
-			}
-			Impl::PublishDelayedFaults(page, old_protection, restored_protection);
-		} else if (!Impl::AllowsAccess(page, fault_vaddr, access)) {
-			FailFast("fault completion left the page inaccessible");
-		}
-		page.resolving            = false;
-		page.resolving_read_write = false;
-	}
-	g_in_fault_resolution = true;
-	const bool released   = m_impl->fault_handler(m_impl->fault_context, access, fault_vaddr, 1,
-	                                              PageFaultPhase::Release);
-	g_in_fault_resolution = false;
-	if (!released) {
-		FailFast("fault release callback failed");
-	}
-	return true;
-}
-
 } // namespace Libs::Graphics

src/graphics/host_gpu/pageManager.h

@@ -11,12 +11,7 @@
 namespace Libs::Graphics {
 
 enum class PageFaultAccess { Read, Write, Execute, Unknown };
-enum class PageFaultPhase { Invalidate, Complete, Release };
 enum class PageWatchMode { Write, ReadWrite };
-enum class GpuAccess { Read, Write, ReadWrite };
-
-using PageFaultHandler = bool (*)(void* context, PageFaultAccess access, uint64_t vaddr,
-                                  uint64_t size, PageFaultPhase phase) noexcept;
 
 class PageManager final {
 public:
@@ -32,23 +27,19 @@ public:
 		uint64_t     m_size  = 0;
 	};
 
-	PageManager(PageFaultHandler fault_handler, void* fault_context);
+	PageManager();
 	// The owner must stop all PageManager callers before destruction.
 	~PageManager();
 
 	KYTY_CLASS_NO_COPY(PageManager);
 
 	[[nodiscard]] uint64_t GetPageSize() const;
-	[[nodiscard]] bool     IsTracked(uint64_t vaddr) const noexcept;
-	[[nodiscard]] bool     IsMapped(uint64_t vaddr, uint64_t size) const noexcept;
-	[[nodiscard]] bool HasGpuAccess(uint64_t vaddr, uint64_t size, GpuAccess access) const noexcept;
 
 	void UpdatePageWatchers(bool track, uint64_t vaddr, uint64_t size,
 	                        PageWatchMode mode = PageWatchMode::Write);
-	void OnGpuMap(uint64_t vaddr, uint64_t size, GpuAccess access = GpuAccess::ReadWrite);
-	void OnGpuUnmap(uint64_t vaddr, uint64_t size, GpuAccess access = GpuAccess::ReadWrite);
+	void OnGpuMap(uint64_t vaddr, uint64_t size);
+	void OnGpuUnmap(uint64_t vaddr, uint64_t size);
 
-	[[nodiscard]] bool HandleFault(PageFaultAccess access, uint64_t fault_vaddr) noexcept;
 	[[nodiscard]] std::vector<std::unique_ptr<BackingWrite>>
 	ReserveBackingWrites(std::span<const RangeSet::Range> ranges);
 

src/graphics/host_gpu/regionManager.h

@@ -25,8 +25,6 @@
 
 namespace Libs::Graphics {
 
-enum class CpuFaultAction { Untracked, Continue, Download };
-
 class TrackingSpinLock final {
 public:
 	void lock() noexcept {
@@ -85,18 +83,6 @@ public:
 	KYTY_CLASS_NO_COPY(RegionManager);
 
 	[[nodiscard]] uint64_t GetCpuAddr() const { return m_cpu_addr; }
-	void                   Track(uint64_t vaddr, uint64_t size) {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			m_tracked.set(page);
-		}
-	}
-	void Untrack(uint64_t vaddr, uint64_t size) {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			m_tracked.reset(page);
-		}
-	}
 	template <DirtySource source>
 	[[nodiscard]] bool IsModified(uint64_t offset, uint64_t size) const {
 		const auto [start, end] = GetPageRange(m_cpu_addr + offset, size);
@@ -126,15 +112,15 @@ public:
 		const auto [start, end] = GetPageRange(vaddr, size);
 		if constexpr (source == DirtySource::Cpu && enable) {
 			for (auto page = start; page < end; page++) {
-				if (m_gpu_dirty.test(page) || m_fault_pending.test(page)) {
-					EXIT("CPU dirty state conflicts with GPU dirty or pending fault state\n");
+				if (m_gpu_dirty.test(page)) {
+					EXIT("CPU dirty state conflicts with GPU dirty state\n");
 				}
 			}
 		}
 		if constexpr (source == DirtySource::Gpu && enable) {
 			for (auto page = start; page < end; page++) {
-				if (m_cpu_dirty.test(page) || m_fault_pending.test(page)) {
-					EXIT("GPU dirty state conflicts with CPU dirty or pending fault state\n");
+				if (m_cpu_dirty.test(page)) {
+					EXIT("GPU dirty state conflicts with CPU dirty state\n");
 				}
 			}
 		}
@@ -151,110 +137,10 @@ public:
 		return changed;
 	}
 
-	[[nodiscard]] CpuFaultAction BeginCpuFault(uint64_t vaddr, uint64_t size,
-	                                           PageFaultAccess access = PageFaultAccess::Write) {
-		if (access != PageFaultAccess::Read && access != PageFaultAccess::Write) {
-			EXIT("unsupported CPU fault access while beginning ownership transfer\n");
-		}
-		const auto [start, end] = GetPageRange(vaddr, size);
-		const bool tracked      = m_tracked.test(start);
-		for (auto page = start; page < end; page++) {
-			if (m_tracked.test(page) != tracked) {
-				EXIT("CPU fault spans mixed tracked and untracked pages\n");
-			}
-			if (m_fault_pending.test(page)) {
-				return CpuFaultAction::Untracked;
-			}
-			if (m_cpu_dirty.test(page) != m_writable.test(page) ||
-			    (m_gpu_dirty.test(page) && (m_cpu_dirty.test(page) || m_writable.test(page)))) {
-				EXIT("inconsistent CPU fault page state\n");
-			}
-		}
-		if (!tracked) {
-			return CpuFaultAction::Untracked;
-		}
-		bool gpu_dirty = m_gpu_dirty.test(start);
-		bool writable  = m_writable.test(start);
-		for (auto page = start + 1; page < end; page++) {
-			if (m_gpu_dirty.test(page) != gpu_dirty || m_writable.test(page) != writable) {
-				EXIT("CPU fault spans pages with incompatible dirty or writable state\n");
-			}
-		}
-		for (auto page = start; page < end; page++) {
-			if (!gpu_dirty && access == PageFaultAccess::Write) {
-				m_cpu_dirty.set(page);
-				m_writable.set(page);
-			}
-			m_fault_pending.set(page);
-		}
-		return gpu_dirty ? CpuFaultAction::Download : CpuFaultAction::Continue;
-	}
-
-	[[nodiscard]] bool CompleteCpuFault(uint64_t vaddr, uint64_t size, PageFaultAccess access,
-	                                    bool downloaded) {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			if (!m_fault_pending.test(page)) {
-				return false;
-			}
-		}
-		for (auto page = start; page < end; page++) {
-			const bool gpu_dirty = m_gpu_dirty.test(page);
-			if (gpu_dirty != downloaded) {
-				EXIT("CPU fault download result disagrees with GPU dirty state\n");
-			}
-			if (gpu_dirty) {
-				m_gpu_dirty.reset(page);
-				switch (access) {
-					case PageFaultAccess::Read: break;
-					case PageFaultAccess::Write:
-						m_cpu_dirty.set(page);
-						m_writable.set(page);
-						break;
-					default: EXIT("unsupported CPU fault access after GPU download\n");
-				}
-			}
-			m_fault_pending.reset(page);
-		}
-		return true;
-	}
-
-	[[nodiscard]] bool HasPendingFault(uint64_t vaddr, uint64_t size) const {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			if (m_fault_pending.test(page)) {
-				return true;
-			}
-		}
-		return false;
-	}
-
-	[[nodiscard]] bool CompleteVirtualGpuWrite(uint64_t vaddr, uint64_t size) {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			if (!m_fault_pending.test(page)) {
-				return false;
-			}
-			if (!m_gpu_dirty.test(page)) {
-				EXIT("virtual GPU write completion found a non-GPU-dirty page\n");
-			}
-		}
-		for (auto page = start; page < end; page++) {
-			m_gpu_dirty.reset(page);
-			m_cpu_dirty.set(page);
-			m_writable.set(page);
-			m_fault_pending.reset(page);
-		}
-		return true;
-	}
-
 	template <DirtySource source, bool clear, typename Func>
 	RegionBits ForEachModifiedRange(uint64_t vaddr, uint64_t size, Func&& func) {
 		const auto [start, end] = GetPageRange(vaddr, size);
 		auto mask               = GetBits<source>();
-		if constexpr (source == DirtySource::Cpu) {
-			mask &= ~m_fault_pending;
-		}
 		for (auto page = 0u; page < start; page++) {
 			mask.reset(page);
 		}
@@ -351,8 +237,6 @@ private:
 	RegionBits   m_cpu_dirty;
 	RegionBits   m_gpu_dirty;
 	RegionBits   m_writable;
-	RegionBits   m_fault_pending;
-	RegionBits   m_tracked;
 };
 
 } // namespace Libs::Graphics

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

@@ -123,24 +123,6 @@ struct BufferCache::RetiredBuffer {
 	std::shared_ptr<Buffer> owner;
 };
 
-struct BufferCache::FaultReadback {
-	PageFaultAccess            access = PageFaultAccess::Unknown;
-	uint64_t                   vaddr  = 0;
-	uint64_t                   size   = 0;
-	std::vector<DownloadRange> ranges;
-	bool                       installed = false;
-
-	[[nodiscard]] bool Active() const noexcept { return !ranges.empty(); }
-
-	void Reset() {
-		access    = PageFaultAccess::Unknown;
-		vaddr     = 0;
-		size      = 0;
-		installed = false;
-		ranges.clear();
-	}
-};
-
 struct BufferCache::PendingBackingPublication {
 	uint64_t address = 0;
 	uint64_t size    = 0;
@@ -253,7 +235,7 @@ BufferCache::BufferCache(GraphicContext& graphics, CommandScheduler& scheduler,
                          ResourceMutex& resource_mutex)
     : m_graphics(graphics), m_scheduler(scheduler),
       m_gds_buffer(graphics, scheduler, MemoryUsage::Stream, 0, AllFlags, GdsBufferSize),
-      m_fault_readback(std::make_unique<FaultReadback>()), m_memory_tracker(page_manager),
+      m_memory_tracker(page_manager),
       m_staging_buffer(graphics, scheduler, MemoryUsage::Upload, 512 * MiB),
       m_stream_buffer(graphics, scheduler, MemoryUsage::Stream, 64 * MiB),
       m_download_buffer(graphics, scheduler, MemoryUsage::Download, 32 * MiB),
@@ -277,9 +259,6 @@ BufferCache::BufferCache(GraphicContext& graphics, CommandScheduler& scheduler,
 }
 
 BufferCache::~BufferCache() {
-	if (m_fault_readback->Active()) {
-		EXIT("BufferCache: destroyed with an active fault readback\n");
-	}
 	if (!m_gpu_modified_ranges.Empty()) {
 		EXIT("BufferCache: destroyed with pending GPU-modified ranges\n");
 	}
@@ -447,93 +426,6 @@ void BufferCache::ReadMemory(uint64_t vaddr, uint64_t size) {
 	}
 }
 
-bool BufferCache::InvalidateMemory(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-                                   PageFaultPhase phase) noexcept {
-	const auto page = vaddr & ~(TRACKER_PAGE_SIZE - 1);
-	if (size == 0 || size > page + TRACKER_PAGE_SIZE - vaddr) {
-		EXIT("BufferCache: invalid page-fault range\n");
-	}
-
-	if (phase == PageFaultPhase::Complete) {
-		FaultSafeCacheLock lock(this, m_mutex);
-		auto&              fault = *m_fault_readback;
-		if (!fault.Active()) {
-			return m_memory_tracker.CompleteCpuFault(vaddr, size, access, false);
-		}
-		if (fault.access != access || fault.vaddr != vaddr || fault.size != size ||
-		    fault.installed) {
-			EXIT("BufferCache: mismatched fault readback completion\n");
-		}
-		PublishDownloads(fault.ranges);
-		if (!m_memory_tracker.CompleteCpuFault(vaddr, size, access, true)) {
-			EXIT("BufferCache: failed to complete downloaded CPU fault\n");
-		}
-		fault.installed = true;
-		return true;
-	}
-
-	if (phase == PageFaultPhase::Release) {
-		FaultSafeCacheLock lock(this, m_mutex);
-		auto&              fault = *m_fault_readback;
-		if (fault.Active()) {
-			if (fault.access != access || fault.vaddr != vaddr || fault.size != size ||
-			    !fault.installed) {
-				EXIT("BufferCache: mismatched fault readback release\n");
-			}
-			for (const auto& range: fault.ranges) {
-				m_gpu_modified_ranges.Subtract(range.address, range.size);
-			}
-			fault.Reset();
-		}
-		return true;
-	}
-
-	if (phase != PageFaultPhase::Invalidate) {
-		EXIT("BufferCache: unsupported page-fault phase\n");
-	}
-
-	const auto action = m_memory_tracker.BeginCpuFault(vaddr, size, access);
-	if (action != CpuFaultAction::Download) {
-		return action == CpuFaultAction::Continue;
-	}
-
-	auto&                     fault = *m_fault_readback;
-	std::vector<DownloadCopy> copies;
-	{
-		FaultSafeCacheLock lock(this, m_mutex);
-		if (fault.Active()) {
-			EXIT("BufferCache: nested fault readback\n");
-		}
-		fault.access = access;
-		fault.vaddr  = vaddr;
-		fault.size   = size;
-
-		m_gpu_modified_ranges.ForEachIntersection(
-		    page, TRACKER_PAGE_SIZE, [&](RangeSet::Range range) {
-			    auto owner = m_buffers.upper_bound(range.address);
-			    if (owner == m_buffers.begin()) {
-				    EXIT("BufferCache: fault readback has no buffer owner\n");
-			    }
-			    --owner;
-			    auto& cached = *owner->second;
-			    if (!cached.buffer->IsInBounds(range.address, range.size)) {
-				    EXIT("BufferCache: fault readback is outside its buffer owner\n");
-			    }
-			    copies.push_back({cached.buffer, cached.buffer->Offset(range.address),
-			                      range.address, range.size});
-		    });
-		if (copies.empty()) {
-			EXIT("BufferCache: GPU-dirty fault page has no dirty byte ranges\n");
-		}
-	}
-	fault.ranges = RecordDownloads(copies);
-	if (!fault.Active()) {
-		EXIT("BufferCache: GPU-dirty fault page has no dirty byte ranges\n");
-	}
-	m_scheduler.FinishCurrent();
-	return true;
-}
-
 void BufferCache::UnmapMemory(uint64_t vaddr, uint64_t size) {
 	if (vaddr == 0 || size == 0 || size > UINT64_MAX - vaddr) {
 		EXIT("BufferCache: invalid unmap range\n");
@@ -714,7 +606,6 @@ BufferBinding BufferCache::ObtainBuffer(CommandBuffer& command, uint64_t vaddr,
 	if (command.IsInvalid() || command.IsExecute()) {
 		EXIT("BufferCache: buffer request requires a recording command buffer\n");
 	}
-	ValidateGpuAccess(vaddr, size, is_read, is_written);
 	std::lock_guard transaction(m_resource_mutex);
 	(void)SynchronizeBacking(vaddr, size);
 
@@ -999,7 +890,6 @@ void BufferCache::FillBuffer(uint64_t vaddr, uint64_t size, uint32_t value, bool
 	if (vaddr == 0) {
 		EXIT("BufferCache: invalid fill memory address\n");
 	}
-	ValidateGpuAccess(vaddr, size, false, true);
 	(void)m_texture_cache.ClearMeta(vaddr);
 	{
 		std::lock_guard transaction(m_resource_mutex);
@@ -1041,12 +931,6 @@ void BufferCache::CopyBuffer(uint64_t dst_vaddr, uint64_t src_vaddr, uint64_t si
 	    (src_gds && (src_vaddr > m_gds_buffer.Size() || size > m_gds_buffer.Size() - src_vaddr))) {
 		EXIT("BufferCache: invalid or overlapping copy range\n");
 	}
-	if (src_memory) {
-		ValidateGpuAccess(src_vaddr, size, true, false);
-	}
-	if (dst_memory) {
-		ValidateGpuAccess(dst_vaddr, size, false, true);
-	}
 	if (src_memory || dst_memory) {
 		std::lock_guard transaction(m_resource_mutex);
 		if (src_memory) {
@@ -1203,26 +1087,13 @@ void BufferCache::PublishImageBuffer(uint64_t vaddr, uint64_t size) {
 	owner->second->tick_accessed_last = m_gc_tick;
 }
 
-void BufferCache::ValidateGpuAccess(uint64_t vaddr, uint64_t size, bool is_read,
-                                    bool is_written) const {
-	if ((!is_read && !is_written) || vaddr == 0 || size == 0 || size > UINT64_MAX - vaddr) {
-		EXIT("BufferCache: invalid GPU access request\n");
-	}
-	if (is_read && !m_page_manager.HasGpuAccess(vaddr, size, GpuAccess::Read)) {
-		EXIT("BufferCache: GPU-read access denied\n");
-	}
-	if (is_written && !m_page_manager.HasGpuAccess(vaddr, size, GpuAccess::Write)) {
-		EXIT("BufferCache: GPU-write access denied\n");
-	}
-}
-
 void BufferCache::RunGarbageCollector() {
 	std::lock_guard transaction(m_resource_mutex);
 	const auto      tick = m_gc_tick++;
 	if (m_graphics.CanReportMemoryUsage()) {
 		m_total_used_memory = m_graphics.GetDeviceMemoryUsage();
 	}
-	if (m_total_used_memory < m_trigger_gc_memory || m_fault_readback->Active()) {
+	if (m_total_used_memory < m_trigger_gc_memory) {
 		return;
 	}
 

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

@@ -47,8 +47,6 @@ public:
 	~BufferCache();
 	KYTY_CLASS_NO_COPY(BufferCache);
 
-	[[nodiscard]] bool InvalidateMemory(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-	                                    PageFaultPhase phase) noexcept;
 	void                        InvalidateMemory(uint64_t vaddr, uint64_t size);
 	void                        ReadMemory(uint64_t vaddr, uint64_t size);
 	void                        UnmapMemory(uint64_t vaddr, uint64_t size);
@@ -77,7 +75,6 @@ public:
 	void               CompleteBackingPublication(uint64_t vaddr, uint64_t size, uint64_t tick);
 	[[nodiscard]] bool SynchronizeBacking(uint64_t vaddr, uint64_t size);
 	void               PublishImageBuffer(uint64_t vaddr, uint64_t size);
-	void ValidateGpuAccess(uint64_t vaddr, uint64_t size, bool is_read, bool is_written) const;
 	void               RunGarbageCollector();
 
 private:
@@ -91,7 +88,6 @@ private:
 	struct DownloadCopy;
 	struct DownloadRange;
 	struct RetiredBuffer;
-	struct FaultReadback;
 	struct PendingBackingPublication;
 	static constexpr uint64_t               DOWNLOAD_ALIGNMENT = 64;
 	[[nodiscard]] static uint64_t           AlignDown(uint64_t value) noexcept;
@@ -121,7 +117,6 @@ private:
 	Common::Mutex                                     m_mutex;
 	std::shared_ptr<Buffer>                           m_null_buffer;
 	std::map<uint64_t, std::unique_ptr<CachedBuffer>> m_buffers;
-	std::unique_ptr<FaultReadback>                    m_fault_readback;
 	RangeSet                                          m_gpu_modified_ranges;
 	RangeSet                                          m_image_invalidated_ranges;
 	std::mutex                                        m_publication_mutex;

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

@@ -4,37 +4,14 @@
 #include "graphics/guest_gpu/command_processor/commandProcessor.h"
 #include "graphics/guest_gpu/graphicsRun.h"
 #include "graphics/host_gpu/renderer/commandScheduler.h"
-
 namespace Libs::Graphics {
 
 GpuResourceManager::GpuResourceManager(GraphicContext& graphics, CommandScheduler& scheduler)
-    : m_page_manager(FaultThunk, this),
-      m_buffer_cache(graphics, scheduler, m_page_manager, m_texture_cache, m_resource_mutex),
+    : m_buffer_cache(graphics, scheduler, m_page_manager, m_texture_cache, m_resource_mutex),
       m_texture_cache(graphics, scheduler, m_page_manager, m_buffer_cache, m_resource_mutex) {}
 
 GpuResourceManager::~GpuResourceManager() = default;
 
-bool GpuResourceManager::FaultThunk(void* context, PageFaultAccess access, uint64_t vaddr,
-                                    uint64_t size, PageFaultPhase phase) noexcept {
-	return static_cast<GpuResourceManager*>(context)->InvalidateMemory(access, vaddr, size, phase);
-}
-
-bool GpuResourceManager::InvalidateMemory(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-                                          PageFaultPhase phase) noexcept {
-	// Let the authoritative image materialize first. A clean overlapping buffer marks a write
-	// fault CPU-dirty when it begins ownership transfer; doing that before image preflight would
-	// make the image appear to race a real CPU write. Completion and release retain buffer-first
-	// ordering so its pending fault is gone before TextureCache publishes the downloaded backing.
-	if (phase == PageFaultPhase::Invalidate) {
-		const bool image_handled  = m_texture_cache.InvalidateMemory(access, vaddr, size, phase);
-		const bool buffer_handled = m_buffer_cache.InvalidateMemory(access, vaddr, size, phase);
-		return buffer_handled || image_handled;
-	}
-	const bool buffer_handled = m_buffer_cache.InvalidateMemory(access, vaddr, size, phase);
-	const bool image_handled  = m_texture_cache.InvalidateMemory(access, vaddr, size, phase);
-	return buffer_handled || image_handled;
-}
-
 bool GpuResourceManager::HandleFault(PageFaultAccess access, uint64_t fault_vaddr) noexcept {
 	constexpr uint64_t fault_size = 8;
 	if (!IsMapped(fault_vaddr, fault_size)) {
@@ -115,22 +92,19 @@ bool GpuResourceManager::IsMapped(uint64_t vaddr, uint64_t size) const noexcept
 	return m_mapped_ranges.Contains(vaddr, size);
 }
 
-void GpuResourceManager::MapMemory(uint64_t vaddr, uint64_t size, GpuAccess access) {
+void GpuResourceManager::MapMemory(uint64_t vaddr, uint64_t size) {
 	{
 		std::lock_guard lock(m_mapped_ranges_mutex);
 		m_mapped_ranges.Add(vaddr, size);
 	}
-	m_page_manager.OnGpuMap(vaddr, size, access);
+	m_page_manager.OnGpuMap(vaddr, size);
 }
 
-void GpuResourceManager::UnmapMemory(uint64_t vaddr, uint64_t size, GpuAccess access) {
-	if (!IsMapped(vaddr, size)) {
-		EXIT("cannot unmap an unmapped GPU resource range\n");
-	}
-	const auto unmap = [this, vaddr, size, access] {
-		m_texture_cache.UnmapMemory(vaddr, size);
+void GpuResourceManager::UnmapMemory(uint64_t vaddr, uint64_t size) {
+	const auto unmap = [this, vaddr, size] {
 		m_buffer_cache.UnmapMemory(vaddr, size);
-		m_page_manager.OnGpuUnmap(vaddr, size, access);
+		m_texture_cache.UnmapMemory(vaddr, size);
+		m_page_manager.OnGpuUnmap(vaddr, size);
 		std::lock_guard lock(m_mapped_ranges_mutex);
 		m_mapped_ranges.Subtract(vaddr, size);
 	};

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

@@ -29,16 +29,11 @@ public:
 	[[nodiscard]] bool HandleFault(PageFaultAccess access, uint64_t fault_vaddr) noexcept;
 	[[nodiscard]] bool InvalidateMemory(uint64_t vaddr, uint64_t size);
 	[[nodiscard]] bool IsMapped(uint64_t vaddr, uint64_t size) const noexcept;
-	void               MapMemory(uint64_t vaddr, uint64_t size, GpuAccess access);
-	void               UnmapMemory(uint64_t vaddr, uint64_t size, GpuAccess access);
+	void               MapMemory(uint64_t vaddr, uint64_t size);
+	void               UnmapMemory(uint64_t vaddr, uint64_t size);
 	void               RunGarbageCollector();
 
 private:
-	static bool FaultThunk(void* context, PageFaultAccess access, uint64_t vaddr, uint64_t size,
-	                       PageFaultPhase phase) noexcept;
-	[[nodiscard]] bool InvalidateMemory(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-	                                    PageFaultPhase phase) noexcept;
-
 	PageManager               m_page_manager;
 	ResourceMutex             m_resource_mutex;
 	BufferCache               m_buffer_cache;

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

@@ -56,7 +56,10 @@ vk::Sampler SamplerCache::GetSampler(const ShaderSamplerResource& r) {
 			case Prospero::SamplerAnisoRatio::kFour: aniso_ratio = 4.0f; break;
 			case Prospero::SamplerAnisoRatio::kEight: aniso_ratio = 8.0f; break;
 			case Prospero::SamplerAnisoRatio::kSixteen: aniso_ratio = 16.0f; break;
-			default: EXIT("unknown ratio: %d\n", static_cast<int>(r.MaxAnisoRatio()));
+			default:
+				EXIT("unknown ratio: %d dwords=%08x,%08x,%08x,%08x\n",
+				     static_cast<int>(r.MaxAnisoRatio()), r.fields[0], r.fields[1], r.fields[2],
+				     r.fields[3]);
 		}
 	}
 

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

@@ -135,8 +135,8 @@ void Buffer::Write(uint64_t offset, const void* source, uint64_t size) {
 void Buffer::Flush(uint64_t offset, uint64_t size) {
 	EXIT_IF(m_mapped.empty() || offset > m_size || size > m_size - offset);
 	if (!m_is_coherent && size != 0) {
-		const auto result = vmaFlushAllocation(m_graphics->allocator, m_buffer->memory.allocation,
-		                                      offset, size);
+		const auto result =
+		    vmaFlushAllocation(m_graphics->allocator, m_buffer->memory.allocation, offset, size);
 		EXIT_NOT_IMPLEMENTED(static_cast<vk::Result>(result) != vk::Result::eSuccess);
 	}
 }
@@ -144,8 +144,8 @@ void Buffer::Flush(uint64_t offset, uint64_t size) {
 vk::BufferMemoryBarrier Buffer::Barrier(uint64_t offset, uint64_t size, vk::AccessFlags source,
                                         vk::AccessFlags destination) const {
 	if (Handle() == nullptr || size == 0 || offset > m_size || size > m_size - offset) {
-		EXIT("Buffer: invalid DMA barrier, handle=%p offset=0x%016" PRIx64
-		     " size=0x%016" PRIx64 " capacity=0x%016" PRIx64 "\n",
+		EXIT("Buffer: invalid DMA barrier, handle=%p offset=0x%016" PRIx64 " size=0x%016" PRIx64
+		     " capacity=0x%016" PRIx64 "\n",
 		     static_cast<const void*>(Handle()), offset, size, m_size);
 	}
 	vk::BufferMemoryBarrier barrier {};
@@ -175,8 +175,7 @@ void Buffer::CopyFrom(CommandBuffer& command, const Buffer& source, uint64_t sou
 	command.EndRendering();
 	const vk::BufferMemoryBarrier before[] = {
 	    source.Barrier(source_offset, size, source_before, vk::AccessFlagBits::eTransferRead),
-	    Barrier(destination_offset, size, destination_before,
-	            vk::AccessFlagBits::eTransferWrite),
+	    Barrier(destination_offset, size, destination_before, vk::AccessFlagBits::eTransferWrite),
 	};
 	const auto host_access  = vk::AccessFlagBits::eHostRead | vk::AccessFlagBits::eHostWrite;
 	auto       before_stage = vk::PipelineStageFlags {vk::PipelineStageFlagBits::eAllCommands};
@@ -214,8 +213,7 @@ void Buffer::Fill(uint64_t offset, uint64_t size, uint32_t value) {
 	                       vk::PipelineStageFlagBits::eTransfer, vk::DependencyFlagBits::eByRegion,
 	                       0, nullptr, 1, &before, 0, nullptr);
 	native.fillBuffer(Handle(), offset, size, value);
-	const auto after =
-	    Barrier(offset, size, vk::AccessFlagBits::eTransferWrite,
+	const auto after = Barrier(offset, size, vk::AccessFlagBits::eTransferWrite,
 	                           vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite);
 	native.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer,
 	                       vk::PipelineStageFlagBits::eAllCommands,

src/graphics/host_gpu/renderer/cache/streamBuffer.h

@@ -54,16 +54,15 @@ public:
 	[[nodiscard]] bool IsInBounds(uint64_t address, uint64_t size) const noexcept;
 	void               Write(uint64_t offset, const void* source, uint64_t size);
 	void               Flush(uint64_t offset, uint64_t size);
-	void               CopyFrom(
-	                  CommandBuffer& command, const Buffer& source, uint64_t source_offset,
+	void CopyFrom(CommandBuffer& command, const Buffer& source, uint64_t source_offset,
 	              uint64_t destination_offset, uint64_t size,
 	              vk::AccessFlags source_before      = vk::AccessFlagBits::eMemoryWrite,
-	                  vk::AccessFlags destination_before =
-	                      vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite,
-	                  vk::AccessFlags source_after =
-	                      vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite,
-	                  vk::AccessFlags destination_after =
-	                      vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite);
+	              vk::AccessFlags destination_before = vk::AccessFlagBits::eMemoryRead |
+	                                                   vk::AccessFlagBits::eMemoryWrite,
+	              vk::AccessFlags source_after       = vk::AccessFlagBits::eMemoryRead |
+	                                                   vk::AccessFlagBits::eMemoryWrite,
+	              vk::AccessFlags destination_after  = vk::AccessFlagBits::eMemoryRead |
+	                                                   vk::AccessFlagBits::eMemoryWrite);
 	void Fill(uint64_t offset, uint64_t size, uint32_t value);
 
 protected:

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

@@ -1377,7 +1377,6 @@ bool TextureCache::ClearImageFromBuffer(CommandBuffer& command, uint64_t address
 	if (command.IsInvalid() || !GuestRange {address, size}.Valid()) {
 		EXIT("TextureCache: invalid image clear\n");
 	}
-	m_buffer_cache.ValidateGpuAccess(address, size, false, true);
 	std::lock_guard      transaction(m_resource_mutex);
 	CacheLock            lock(*this, m_lock);
 	ImageId              selected {};
@@ -1827,37 +1826,6 @@ bool TextureCache::TouchMeta(uint64_t address, uint32_t slice, bool is_clear) {
 	return true;
 }
 
-bool TextureCache::InvalidateMemory(PageFaultAccess access, uint64_t address, uint64_t size,
-                                    PageFaultPhase phase) noexcept {
-	if ((access != PageFaultAccess::Read && access != PageFaultAccess::Write) ||
-	    !GuestRange {address, size}.Valid()) {
-		return false;
-	}
-	if (access == PageFaultAccess::Read) {
-		return false;
-	}
-	if (phase == PageFaultPhase::Invalidate) {
-		CacheLock  lock(*this, m_lock);
-		const bool tracked =
-		    std::ranges::any_of(FindImagesInRegion(address, size, true), [&](ImageId id) {
-			    const auto owner = ResolveOwner(id);
-			    return owner != nullptr && !owner->depth_id && owner->IsTracked();
-		    });
-		if (tracked) {
-			InvalidateCpuAliases(address, size);
-		}
-		return tracked;
-	}
-	if (phase != PageFaultPhase::Complete && phase != PageFaultPhase::Release) {
-		return false;
-	}
-	CacheLock lock(*this, m_lock);
-	return std::ranges::any_of(FindImagesInRegion(address, size, true), [&](ImageId id) {
-		const auto owner = ResolveOwner(id);
-		return owner != nullptr && !owner->depth_id;
-	});
-}
-
 void TextureCache::UnmapMemory(uint64_t address, uint64_t size) {
 	if (!GuestRange {address, size}.Valid()) {
 		EXIT("TextureCache: invalid unmap range\n");

src/graphics/host_gpu/renderer/cache/textureCache.h

@@ -76,8 +76,6 @@ public:
 	[[nodiscard]] bool ClearMeta(uint64_t address);
 	[[nodiscard]] bool TouchMeta(uint64_t address, uint32_t slice, bool is_clear);
 
-	[[nodiscard]] bool InvalidateMemory(PageFaultAccess access, uint64_t address, uint64_t size,
-	                                    PageFaultPhase phase) noexcept;
 	void UnmapMemory(uint64_t address, uint64_t size);
 	void ProcessDownloadImages();
 	void RunGarbageCollector();

src/graphics/host_gpu/renderer/colorRenderTarget.cpp

@@ -7,8 +7,8 @@
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/guest_gpu/tile.h"
 #include "graphics/host_gpu/graphicContext.h"
-#include "graphics/host_gpu/renderer/image/textureCommon.h"
 #include "graphics/host_gpu/renderer/debug.h"
+#include "graphics/host_gpu/renderer/image/textureCommon.h"
 #include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
@@ -79,10 +79,8 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	const auto view = ResolveTargetViewInfo(
 	    rt.view.base_array_slice_index, rt.view.last_array_slice_index, render_target_slice_offset);
 	switch (view.type) {
-		case TargetViewType::Image2D: break;
-		case TargetViewType::Image2DArray:
-			EXIT("layered render-target views are unsupported: base=%u count=%u\n", view.base_layer,
-			     view.layer_count);
+		case TargetViewType::Image2D:
+		case TargetViewType::Image2DArray: break;
 		case TargetViewType::Unsupported:
 			EXIT("invalid render-target view: base=%u last=%u draw_offset=%u\n",
 			     rt.view.base_array_slice_index, rt.view.last_array_slice_index,

src/graphics/host_gpu/renderer/colorRenderTarget.h

@@ -2,8 +2,8 @@
 #define EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_COLORRENDERTARGET_H_
 
 #include "graphics/guest_gpu/gpu_defs.h"
-#include "graphics/host_gpu/renderer/renderTarget.h"
 #include "graphics/host_gpu/renderer/cache/textureCache.h"
+#include "graphics/host_gpu/renderer/renderTarget.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 
 #include <cstdint>

src/graphics/host_gpu/renderer/commandScheduler.cpp

@@ -331,8 +331,7 @@ void CommandScheduler::WaitPriorityOperations(uint64_t tick) {
 	EXIT_IF(g_deferred_callback_scheduler == this);
 	std::unique_lock lock(m_operation_mutex);
 	m_operation_available.wait(lock, [this, tick] {
-		const bool active_before_or_at =
-		    m_priority_active && m_priority_active_tick <= tick;
+		const bool active_before_or_at = m_priority_active && m_priority_active_tick <= tick;
 		const bool queued_before_or_at =
 		    !m_priority_operations.empty() && m_priority_operations.front().tick <= tick;
 		return !active_before_or_at && !queued_before_or_at;

src/graphics/host_gpu/renderer/context.cpp

@@ -8,8 +8,8 @@
 #include "graphics/host_gpu/renderer/colorRenderTarget.h"
 #include "graphics/host_gpu/renderer/debug.h"
 #include "graphics/host_gpu/renderer/depthRenderTarget.h"
-#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/image/imageView.h"
+#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/vma.h"
@@ -270,8 +270,8 @@ void CommandBuffer::BeginRendering(const RenderState& state) const {
 		colors[i].sType        = vk::StructureType::eRenderingAttachmentInfo;
 		colors[i].imageView    = attachment.image_view;
 		colors[i].imageLayout  = attachment.image_layout;
-		colors[i].loadOp = attachment.is_clear ? vk::AttachmentLoadOp::eClear
-		                                      : vk::AttachmentLoadOp::eLoad;
+		colors[i].loadOp =
+		    attachment.is_clear ? vk::AttachmentLoadOp::eClear : vk::AttachmentLoadOp::eLoad;
 		colors[i].storeOp                 = vk::AttachmentStoreOp::eStore;
 		colors[i].clearValue.color.uint32 = attachment.clear_value;
 	}
@@ -281,8 +281,8 @@ void CommandBuffer::BeginRendering(const RenderState& state) const {
 	depth.sType       = vk::StructureType::eRenderingAttachmentInfo;
 	depth.imageView   = depth_stencil.image_view;
 	depth.imageLayout = depth_stencil.image_layout;
-	depth.loadOp = depth_stencil.depth_clear ? vk::AttachmentLoadOp::eClear
-	                                        : vk::AttachmentLoadOp::eLoad;
+	depth.loadOp =
+	    depth_stencil.depth_clear ? vk::AttachmentLoadOp::eClear : vk::AttachmentLoadOp::eLoad;
 	depth.storeOp                       = vk::AttachmentStoreOp::eStore;
 	depth.clearValue.depthStencil.depth = std::bit_cast<float>(depth_stencil.clear_value[0]);
 
@@ -290,8 +290,8 @@ void CommandBuffer::BeginRendering(const RenderState& state) const {
 	stencil.sType       = vk::StructureType::eRenderingAttachmentInfo;
 	stencil.imageView   = depth_stencil.image_view;
 	stencil.imageLayout = depth_stencil.image_layout;
-	stencil.loadOp = depth_stencil.stencil_clear ? vk::AttachmentLoadOp::eClear
-	                                            : vk::AttachmentLoadOp::eLoad;
+	stencil.loadOp =
+	    depth_stencil.stencil_clear ? vk::AttachmentLoadOp::eClear : vk::AttachmentLoadOp::eLoad;
 	stencil.storeOp                         = vk::AttachmentStoreOp::eStore;
 	stencil.clearValue.depthStencil.stencil = depth_stencil.clear_value[1];
 

src/graphics/host_gpu/renderer/debug.cpp

@@ -497,7 +497,15 @@ static void ZPrint(const char* func, const HW::DepthRenderTarget& z) {
 }
 
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
-static void ZCheck(const HW::DepthRenderTarget& z) {
+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;
+	if (!depth_active && !stencil_active) {
+		return;
+	}
+
 	EXIT_NOT_IMPLEMENTED(!z.z_info.HasValidTextureCompatibility());
 	EXIT_NOT_IMPLEMENTED(!z.stencil_info.HasValidTextureCompatibility());
 	if (z.z_info.format == 0) {
@@ -548,14 +556,6 @@ static void ZCheck(const HW::DepthRenderTarget& z) {
 		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.slice_start != 0x00000000 || z.depth_view.slice_max != 0x00000000) {
-			static std::atomic<uint32_t> log_count {0};
-			if (log_count.fetch_add(1, std::memory_order_relaxed) < 16) {
-				LOGF("DepthTarget: temporary: ignoring PS5 array slice view start=0x%08" PRIx32
-				     ", max=0x%08" PRIx32 "\n",
-				     z.depth_view.slice_start, z.depth_view.slice_max);
-			}
-		}
 		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) {
@@ -1214,7 +1214,7 @@ void hw_check(const RenderCommandBuffer& buffer) {
 	log_phase("vp");
 	VpCheck(vp, smc);
 	log_phase("z");
-	ZCheck(z);
+	ZCheck(z, d, rc);
 	log_phase("clip");
 	ClipCheck(c);
 	log_phase("rc");

src/graphics/host_gpu/renderer/depthRenderTarget.cpp

@@ -10,10 +10,10 @@
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/guest_gpu/tile.h"
 #include "graphics/host_gpu/graphicContext.h"
-#include "graphics/host_gpu/renderer/image/textureCommon.h"
 #include "graphics/host_gpu/renderer/debug.h"
-#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/image/imageView.h"
+#include "graphics/host_gpu/renderer/image/textureCommon.h"
+#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/vulkanCommon.h"
@@ -150,10 +150,8 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, RenderCommandB
 	    has_stencil, has_htile, z.stencil_info.htile_stencil_disabled);
 	const auto view = ResolveTargetViewInfo(z.depth_view.slice_start, z.depth_view.slice_max);
 	switch (view.type) {
-		case TargetViewType::Image2D: break;
-		case TargetViewType::Image2DArray:
-			DepthFatal("layered depth views are unsupported: base=%u count=%u", view.base_layer,
-			           view.layer_count);
+		case TargetViewType::Image2D:
+		case TargetViewType::Image2DArray: break;
 		case TargetViewType::Unsupported:
 			DepthFatal("invalid depth view: base=%u last=%u", z.depth_view.slice_start,
 			           z.depth_view.slice_max);

src/graphics/host_gpu/renderer/depthRenderTarget.h

@@ -2,9 +2,9 @@
 #define EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_DEPTHRENDERTARGET_H_
 
 #include "common/assert.h"
+#include "graphics/host_gpu/renderer/cache/textureCache.h"
 #include "graphics/host_gpu/renderer/image/imageView.h"
 #include "graphics/host_gpu/renderer/renderTarget.h"
-#include "graphics/host_gpu/renderer/cache/textureCache.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 
 #include <cstdint>

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

@@ -182,8 +182,8 @@ void BlitHelper::ReinterpretColorAsMsDepth(Image& source, Image& destination) {
 	auto  command        = command_buffer.Handle();
 	source.Transit(vk::ImageLayout::eShaderReadOnlyOptimal, vk::AccessFlagBits2::eShaderRead, {},
 	               command);
-	destination.Transit(ColorToMsDepthLayout,
-	                    vk::AccessFlagBits2::eDepthStencilAttachmentWrite, {}, command);
+	destination.Transit(ColorToMsDepthLayout, vk::AccessFlagBits2::eDepthStencilAttachmentWrite, {},
+	                    command);
 
 	vk::RenderingAttachmentInfo depth_attachment {};
 	depth_attachment.sType                   = vk::StructureType::eRenderingAttachmentInfo;

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

@@ -21,8 +21,7 @@ struct GuestRange {
 
 	[[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;
+		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;
@@ -352,8 +351,7 @@ inline bool ImageInfo::IsDepth() const noexcept {
 	}
 	const auto transfer_bytes = DepthAspectTransferBytes(info.pixel_format);
 	return transfer_bytes == info.bytes_per_block ||
-	       (info.bytes_per_block == sizeof(uint16_t) &&
-	        transfer_bytes == sizeof(uint32_t));
+	       (info.bytes_per_block == sizeof(uint16_t) && transfer_bytes == sizeof(uint32_t));
 }
 
 [[nodiscard]] inline VideoOutCompression
@@ -472,16 +470,11 @@ IsSupportedDisplayRenderTargetTileMode(uint32_t tile_mode) noexcept {
 	const auto unorm8 = [](uint32_t value) { return static_cast<float>(value & 0xffu) / 255.0f; };
 	const auto srgb8  = [](uint32_t value) {
 		const auto encoded = static_cast<float>(value & 0xffu) / 255.0f;
-		return encoded <= 0.04045f ? encoded / 12.92f
-		                           : std::pow((encoded + 0.055f) / 1.055f, 2.4f);
+		return encoded <= 0.04045f ? encoded / 12.92f : std::pow((encoded + 0.055f) / 1.055f, 2.4f);
 	};
 	switch (format) {
-		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::eR32Uint: next.uint32[0] = packed; break;
+		case vk::Format::eR32Sint: next.int32[0] = static_cast<int32_t>(packed); break;
 		case vk::Format::eR8G8B8A8Srgb:
 			next.float32[0] = srgb8(packed);
 			next.float32[1] = srgb8(packed >> 8u);

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

@@ -70,15 +70,14 @@ namespace {
 			}
 		case vk::ImageType::e3D:
 			switch (info.type) {
-				case vk::ImageViewType::e3D:
-					return info.base_layer == 0 && info.layer_count == 1;
+				case vk::ImageViewType::e3D: return info.base_layer == 0 && info.layer_count == 1;
 				case vk::ImageViewType::e2D:
-					return static_cast<bool>(
-					           image.flags & vk::ImageCreateFlagBits::e2DArrayCompatible) &&
+					return static_cast<bool>(image.flags &
+					                         vk::ImageCreateFlagBits::e2DArrayCompatible) &&
 					       info.level_count == 1 && info.layer_count == 1;
 				case vk::ImageViewType::e2DArray:
-					return static_cast<bool>(
-					           image.flags & vk::ImageCreateFlagBits::e2DArrayCompatible) &&
+					return static_cast<bool>(image.flags &
+					                         vk::ImageCreateFlagBits::e2DArrayCompatible) &&
 					       info.level_count == 1;
 				default: return false;
 			}
@@ -327,8 +326,7 @@ bool FormatsCompatible(vk::Format base, vk::Format view) noexcept {
 vk::ImageView Image::FindView(const ImageViewInfo& view_info) {
 	const auto& image      = backing;
 	auto        normalized = view_info;
-	const bool  is_storage =
-	    static_cast<bool>(normalized.usage & vk::ImageUsageFlagBits::eStorage);
+	const bool  is_storage = static_cast<bool>(normalized.usage & vk::ImageUsageFlagBits::eStorage);
 	normalized.aspect      = FullAspectMask(image.format);
 	if (normalized.aspect & vk::ImageAspectFlagBits::eDepth &&
 	    IsDepthViewFormat(normalized.format)) {
@@ -340,12 +338,11 @@ vk::ImageView Image::FindView(const ImageViewInfo& view_info) {
 		normalized.format = image.format;
 		normalized.aspect = vk::ImageAspectFlagBits::eStencil;
 	}
-	normalized.usage =
-	    is_storage ? vk::ImageUsageFlagBits::eStorage : vk::ImageUsageFlags {};
+	normalized.usage = is_storage ? vk::ImageUsageFlagBits::eStorage : vk::ImageUsageFlags {};
 	const bool format_compatible = normalized.format != vk::Format::eUndefined &&
 	                               IsCompatibleViewFormat(image.format, normalized.format);
-	const bool slice_view = image.image_type == vk::ImageType::e3D &&
-	                        (normalized.type == vk::ImageViewType::e2D ||
+	const bool slice_view =
+	    image.image_type == vk::ImageType::e3D && (normalized.type == vk::ImageViewType::e2D ||
 	                                               normalized.type == vk::ImageViewType::e2DArray);
 	const bool levels_valid = normalized.level_count != 0 &&
 	                          normalized.base_level < image.mip_levels &&
@@ -353,15 +350,14 @@ vk::ImageView Image::FindView(const ImageViewInfo& view_info) {
 	const auto view_layers  = slice_view && levels_valid
 	                              ? std::max(image.extent.depth >> normalized.base_level, 1u)
 	                              : image.layers;
-	const bool ranges_valid = levels_valid &&
-	                          normalized.layer_count != 0 && normalized.base_layer < view_layers &&
+	const bool ranges_valid = levels_valid && normalized.layer_count != 0 &&
+	                          normalized.base_layer < view_layers &&
 	                          normalized.layer_count <= view_layers - normalized.base_layer;
 	const bool mapping_valid =
 	    IsComponentSwizzle(normalized.mapping.r) && IsComponentSwizzle(normalized.mapping.g) &&
 	    IsComponentSwizzle(normalized.mapping.b) && IsComponentSwizzle(normalized.mapping.a);
 	if (image.image == nullptr || !format_compatible || !ranges_valid || !mapping_valid ||
-	    !IsValidViewType(image, normalized) ||
-	    !IsValidAspect(image, normalized.aspect)) {
+	    !IsValidViewType(image, normalized) || !IsValidAspect(image, normalized.aspect)) {
 		EXIT("invalid image view: image_format=%d view_format=%d type=%d aspect=0x%x "
 		     "mip=%u+%u layer=%u+%u usage=0x%x image_levels=%u image_layers=%u\n",
 		     static_cast<int>(image.format), static_cast<int>(normalized.format),

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

@@ -88,7 +88,9 @@ SelectSampledDepthView(vk::Format image_format, vk::Format view_format, uint32_t
 IsSupportedSampledDepthResource(const ShaderRecompiler::IR::ImageResource& resource) noexcept {
 	return resource.kind == ShaderRecompiler::IR::ResourceKind::Image &&
 	       (resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2D ||
-	        resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2DArray) &&
+	        resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2DArray ||
+	        resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaa ||
+	        resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaaArray) &&
 	       resource.mip_mode == ShaderRecompiler::IR::ImageMipMode::None && resource.read &&
 	       !resource.written && !resource.atomic;
 }

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

@@ -70,6 +70,10 @@ constexpr RenderTargetFormatMapping kRenderTargetFormats[] = {
      Prospero::ChannelType::kFloat,
      Prospero::ChannelOrder::kStandard,
      {vk::Format::eB10G11R11UfloatPack32, 4}},
+    {Prospero::ChannelLayout::k5_6_5,
+     Prospero::ChannelType::kUNorm,
+     Prospero::ChannelOrder::kStandard,
+     {vk::Format::eB5G6R5UnormPack16, 2}},
     {Prospero::ChannelLayout::k16,
      Prospero::ChannelType::kUNorm,
      Prospero::ChannelOrder::kStandard,
@@ -397,10 +401,10 @@ TextureUploadLayout TextureCalcUploadLayout(uint32_t fmt, uint64_t width, uint64
 	return layout;
 }
 
-std::vector<vk::BufferImageCopy>
-TextureBuildImageCopies(const TextureUploadLayout& layout, uint32_t width, uint32_t height,
-                        uint32_t depth, uint64_t levels, bool array_texture,
-                        bool volume_texture) {
+std::vector<vk::BufferImageCopy> TextureBuildImageCopies(const TextureUploadLayout& layout,
+                                                         uint32_t width, uint32_t height,
+                                                         uint32_t depth, uint64_t levels,
+                                                         bool array_texture, bool volume_texture) {
 	uint32_t mip_width  = width;
 	uint32_t mip_height = height;
 	uint32_t mip_pitch  = volume_texture && static_cast<Prospero::TileMode>(layout.tile) !=
@@ -418,10 +422,9 @@ TextureBuildImageCopies(const TextureUploadLayout& layout, uint32_t width, uint3
 		for (uint32_t z = 0; z < mip_depth; z++) {
 			const auto          slice_offset = z * layout.slice_stride;
 			vk::BufferImageCopy region {};
-			region.bufferOffset =
-			    layout.level_sizes[i].offset + slice_offset;
-			region.imageSubresource = {vk::ImageAspectFlagBits::eColor, i,
-			                           array_texture ? z : 0, 1};
+			region.bufferOffset     = layout.level_sizes[i].offset + slice_offset;
+			region.imageSubresource = {vk::ImageAspectFlagBits::eColor, i, array_texture ? z : 0,
+			                           1};
 			region.imageOffset.z    = volume_texture ? static_cast<int>(z) : 0;
 			region.imageExtent      = {mip_width, mip_height, 1};
 			const bool linear =
@@ -434,8 +437,7 @@ TextureBuildImageCopies(const TextureUploadLayout& layout, uint32_t width, uint3
 					return ((value + block - 1u) / block) * block;
 				};
 				const auto pitch       = align(mip_pitch, layout.texel_block);
-				region.bufferRowLength =
-				    pitch > align(mip_width, layout.texel_block) ? pitch : 0;
+				region.bufferRowLength = pitch > align(mip_width, layout.texel_block) ? pitch : 0;
 			}
 			regions.push_back(region);
 		}
@@ -480,8 +482,7 @@ static bool SetGpuTileSize(uint64_t offset, uint64_t length, uint64_t capacity,
 	return true;
 }
 
-bool TextureBuildGpuTileInfos(uint64_t size,
-                              const std::vector<vk::BufferImageCopy>& regions,
+bool TextureBuildGpuTileInfos(uint64_t size, const std::vector<vk::BufferImageCopy>& regions,
                               const TextureUploadLayout& layout, uint32_t fmt, uint32_t depth,
                               uint64_t levels, std::vector<GpuTileInfo>& out_infos) {
 	if (size == 0 || levels == 0 || levels > 16 || depth == 0 ||
@@ -522,9 +523,8 @@ bool TextureBuildGpuTileInfos(uint64_t size,
 			for (uint32_t z = 0; z < mip_depth; z += block.block_depth) {
 				const uint32_t copy_depth = std::min(block.block_depth, mip_depth - z);
 				const auto&    region     = regions[region_base + z];
-				const auto     pitch = region.bufferRowLength != 0
-				                           ? region.bufferRowLength
-				                           : region.imageExtent.width;
+				const auto     pitch =
+				    region.bufferRowLength != 0 ? region.bufferRowLength : region.imageExtent.width;
 				const auto  logical_height = region.bufferImageHeight != 0
 				                                 ? region.bufferImageHeight
 				                                 : region.imageExtent.height;
@@ -544,16 +544,13 @@ bool TextureBuildGpuTileInfos(uint64_t size,
 					return false;
 				}
 				info.linear_slice_stride = linear_stride;
-				info.width = std::max(
-				    (region.imageExtent.width + element.wide - 1u) / element.wide, 1u);
-				info.height = std::max(
-				    (logical_height + element.tall - 1u) / element.tall, 1u);
+				info.width =
+				    std::max((region.imageExtent.width + element.wide - 1u) / element.wide, 1u);
+				info.height = std::max((logical_height + element.tall - 1u) / element.tall, 1u);
 				info.depth  = copy_depth;
-				info.surface_z = block.block_depth == 1
-				                     ? static_cast<uint32_t>(region.imageOffset.z)
-				                     : 0;
-				info.pitch =
-				    std::max((pitch + element.wide - 1u) / element.wide, 1u);
+				info.surface_z =
+				    block.block_depth == 1 ? static_cast<uint32_t>(region.imageOffset.z) : 0;
+				info.pitch        = std::max((pitch + element.wide - 1u) / element.wide, 1u);
 				info.tail_x       = tail ? volume.tail_x[level] : 0;
 				info.tail_y       = tail ? volume.tail_y[level] : 0;
 				info.tail         = tail;
@@ -581,9 +578,8 @@ bool TextureBuildGpuTileInfos(uint64_t size,
 			const auto level_depth = GetTextureLevelDepth(depth, level, layout.volume_texture);
 			for (uint32_t z = 0; z < level_depth; z++) {
 				const auto& region = regions[region_index++];
-				const auto pitch = region.bufferRowLength != 0
-				                       ? region.bufferRowLength
-				                       : region.imageExtent.width;
+				const auto  pitch =
+				    region.bufferRowLength != 0 ? region.bufferRowLength : region.imageExtent.width;
 				const auto  logical_height = region.bufferImageHeight != 0
 				                                 ? region.bufferImageHeight
 				                                 : region.imageExtent.height;
@@ -597,16 +593,14 @@ bool TextureBuildGpuTileInfos(uint64_t size,
 				                    info.tiled_size)) {
 					return false;
 				}
-				info.width = std::max(
-				    (region.imageExtent.width + element.wide - 1u) / element.wide, 1u);
-				info.height = std::max(
-				    (logical_height + element.tall - 1u) / element.tall, 1u);
+				info.width =
+				    std::max((region.imageExtent.width + element.wide - 1u) / element.wide, 1u);
+				info.height    = std::max((logical_height + element.tall - 1u) / element.tall, 1u);
 				info.surface_z = base_family == TileBlockFamily::RenderTarget64KB ||
 				                         base_family == TileBlockFamily::Depth64KB
 				                     ? region.imageSubresource.baseArrayLayer
 				                     : 0;
-				info.pitch =
-				    std::max((pitch + element.wide - 1u) / element.wide, 1u);
+				info.pitch     = std::max((pitch + element.wide - 1u) / element.wide, 1u);
 				info.tail      = tail;
 				info.tail_x    = tail ? level_size.x : 0;
 				info.tail_y    = tail ? level_size.y : 0;

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

@@ -40,12 +40,11 @@ TextureUploadLayout TextureCalcUploadLayout(uint32_t fmt, uint64_t width, uint64
                                                uint64_t tile, uint64_t upload_size,
                                                bool allow_depth_tile, bool volume_texture,
                                                const char* owner);
-std::vector<vk::BufferImageCopy>
-TextureBuildImageCopies(const TextureUploadLayout& layout, uint32_t width, uint32_t height,
-                        uint32_t depth, uint64_t levels, bool array_texture,
-                        bool volume_texture);
-bool TextureBuildGpuTileInfos(uint64_t size,
-                              const std::vector<vk::BufferImageCopy>& regions,
+std::vector<vk::BufferImageCopy> TextureBuildImageCopies(const TextureUploadLayout& layout,
+                                                         uint32_t width, uint32_t height,
+                                                         uint32_t depth, uint64_t levels,
+                                                         bool array_texture, bool volume_texture);
+bool TextureBuildGpuTileInfos(uint64_t size, const std::vector<vk::BufferImageCopy>& regions,
                               const TextureUploadLayout& layout, uint32_t fmt, uint32_t depth,
                               uint64_t levels, std::vector<GpuTileInfo>& infos);
 

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

@@ -14,9 +14,9 @@
 #include "gpu_tiler_shaders/gpu_tiler_standard64_spv.h"
 #include "gpu_tiler_shaders/gpu_tiler_swap_bgra16_spv.h"
 #include "graphics/host_gpu/graphicContext.h"
+#include "graphics/host_gpu/renderer/cache/streamBuffer.h"
 #include "graphics/host_gpu/renderer/commandScheduler.h"
 #include "graphics/host_gpu/renderer/image/image.h"
-#include "graphics/host_gpu/renderer/cache/streamBuffer.h"
 
 #include <algorithm>
 #include <array>

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

@@ -26,11 +26,15 @@ bool IsSampledImage(BindingKind kind) {
 		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;
 	}

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

@@ -73,6 +73,11 @@ static Prospero::ImageType TextureBaseType(Prospero::ImageType type) {
 	}
 }
 
+static bool IsMultisampledTexture(Prospero::ImageType type) {
+	return type == Prospero::ImageType::kColor2DMsaa ||
+	       type == Prospero::ImageType::kColor2DMsaaArray;
+}
+
 static BufferView NativeStorageBuffer(RenderContext& context, CommandBuffer& command_buffer,
                                       const ShaderBufferResource&                 descriptor,
                                       const ShaderRecompiler::IR::BufferResource& resource,
@@ -159,6 +164,8 @@ static bool IsSupportedSampledColorResource(const ShaderRecompiler::IR::ImageRes
 		case ShaderRecompiler::Decoder::ImageDimension::Dim1DArray:
 		case ShaderRecompiler::Decoder::ImageDimension::Dim2D:
 		case ShaderRecompiler::Decoder::ImageDimension::Dim2DArray:
+		case ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaa:
+		case ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaaArray:
 			supported_dimension = true;
 			break;
 		default: break;
@@ -195,6 +202,22 @@ TargetTextureViewInfo ResolveTargetTextureView(const ShaderRecompiler::IR::Image
 			           ? TargetTextureViewInfo {vk::ImageViewType::e2DArray, base_layer,
 			                                    image_layers - base_layer}
 			           : TargetTextureViewInfo {};
+		case Prospero::ImageType::kColor2DMsaa:
+			return resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaa &&
+			               base_layer == 0 && image_layers == 1
+			           ? TargetTextureViewInfo {vk::ImageViewType::e2D, 0, 1}
+			           : TargetTextureViewInfo {};
+		case Prospero::ImageType::kColor2DMsaaArray:
+			if (resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaa &&
+			    base_layer == 0 && image_layers == 1) {
+				return {vk::ImageViewType::e2D, 0, 1};
+			}
+			return resource.dimension ==
+			                   ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaaArray &&
+			               base_layer < image_layers
+			           ? TargetTextureViewInfo {vk::ImageViewType::e2DArray, base_layer,
+			                                    image_layers - base_layer}
+			           : TargetTextureViewInfo {};
 		default: return {};
 	}
 }
@@ -211,34 +234,57 @@ bool IsSupportedSampledVideoOutView(const ShaderRecompiler::IR::ImageResource& r
 bool IsSupportedDepthTargetDescriptor(const ShaderTextureResource& descriptor, const Image& image) {
 	const auto width        = static_cast<uint32_t>(descriptor.Width5()) + 1u;
 	const auto height       = static_cast<uint32_t>(descriptor.Height5()) + 1u;
-	const auto pitch  = TileGetTexturePitch(descriptor.Format(), width, 1, descriptor.TileMode());
 	const auto type         = static_cast<Prospero::ImageType>(descriptor.Type());
-	const bool supported_single_layer =
+	const bool multisampled = IsMultisampledTexture(type);
+	const auto samples      = multisampled ? 1u << descriptor.LastLevel() : 1u;
+	const auto pitch =
+	    multisampled ? TileGetDepthPitch(width, image.info.bytes_per_block, descriptor.LastLevel())
+	                 : TileGetTexturePitch(descriptor.Format(), width, 1, descriptor.TileMode());
+	const bool supported_2d    = type == Prospero::ImageType::kColor2D &&
 	                             image.info.resources.layers == 1 && descriptor.Depth() == 0 &&
-	    descriptor.BaseArray5() == 0 &&
-	    (type == Prospero::ImageType::kColor2D || type == Prospero::ImageType::kColor2DArray);
+	                             descriptor.BaseArray5() == 0;
+	const bool supported_array = type == Prospero::ImageType::kColor2DArray &&
+	                             descriptor.BaseArray5() <= descriptor.Depth() &&
+	                             descriptor.Depth() < image.info.resources.layers;
 	const bool supported_cube =
 	    type == Prospero::ImageType::kCube && width == height && image.info.resources.layers >= 6 &&
 	    image.info.resources.layers % 6u == 0 &&
 	    static_cast<uint32_t>(descriptor.Depth()) + 1u == image.info.resources.layers &&
 	    descriptor.BaseArray5() == 0;
+	const bool supported_msaa_2d    = type == Prospero::ImageType::kColor2DMsaa &&
+	                                  image.info.resources.layers == 1 && descriptor.Depth() == 0 &&
+	                                  descriptor.BaseArray5() == 0;
+	const bool supported_msaa_array = type == Prospero::ImageType::kColor2DMsaaArray &&
+	                                  descriptor.BaseArray5() <= descriptor.Depth() &&
+	                                  descriptor.Depth() < image.info.resources.layers;
+	const bool levels_ok =
+	    multisampled
+	        ? descriptor.BaseLevel() == 0 && descriptor.LastLevel() >= 1 &&
+	              descriptor.LastLevel() <= 3 && descriptor.MaxMip() == descriptor.LastLevel() &&
+	              image.info.resources.levels == 1 && image.info.samples == samples
+	        : descriptor.BaseLevel() == 0 && descriptor.LastLevel() == 0 &&
+	              descriptor.MaxMip() == 0 && image.info.samples == 1;
 	return image.info.IsDepth() && width == image.info.extent.width &&
-	       height == image.info.extent.height && (supported_single_layer || supported_cube) &&
-	       descriptor.BaseLevel() == 0 && descriptor.LastLevel() == 0 && descriptor.MaxMip() == 0 &&
-	       descriptor.MinLod() == 0 && descriptor.BaseArray5() == 0 &&
+	       height == image.info.extent.height &&
+	       (supported_2d || supported_array || supported_cube || supported_msaa_2d ||
+	        supported_msaa_array) &&
+	       levels_ok && descriptor.MinLod() == 0 &&
 	       descriptor.TileMode() == Prospero::GpuEnumValue(Prospero::TileMode::kDepth) &&
-	       descriptor.BCSwizzle() == 0 && !descriptor.MsaaDepth() && pitch >= width &&
-	       pitch == image.info.pitch;
+	       descriptor.BCSwizzle() == 0 && (!descriptor.MsaaDepth() || multisampled) &&
+	       pitch >= width && pitch == image.info.pitch;
 }
 
 bool IsSupportedDepthTextureEncoding(const ShaderTextureResource& descriptor, const Image& image) {
 	constexpr uint32_t field1_reserved_mask = 0x200fff00u;
 	constexpr uint32_t field2_reserved_mask = 0xf0003000u;
-	constexpr uint32_t field3_common        = 0x01800000u;
-	constexpr uint32_t field5_expected      = 0x00700000u;
-	const uint32_t     field3_expected =
-	    (descriptor.Type() << 28u) | field3_common | descriptor.DstSelXYZW();
+	const uint32_t     field3_expected = descriptor.DstSelXYZW() |
+	                                     (static_cast<uint32_t>(descriptor.BaseLevel()) << 12u) |
+	                                     (static_cast<uint32_t>(descriptor.LastLevel()) << 16u) |
+	                                     (static_cast<uint32_t>(descriptor.TileMode()) << 20u) |
+	                                     (static_cast<uint32_t>(descriptor.Type()) << 28u);
 	const uint32_t     field4_expected = descriptor.Depth() | (descriptor.BaseArray5() << 16u);
+	const uint32_t     field5_expected =
+	    0x00700000u | (static_cast<uint32_t>(descriptor.MaxMip()) << 4u);
 	const bool common = (descriptor.fields[1] & field1_reserved_mask) == 0 &&
 	                    (descriptor.fields[2] & field2_reserved_mask) == 0 &&
 	                    descriptor.fields[3] == field3_expected &&
@@ -251,8 +297,9 @@ bool IsSupportedDepthTextureEncoding(const ShaderTextureResource& descriptor, co
 		return true;
 	}
 	constexpr uint32_t htile_control = 0x00280000u;
+	const uint32_t expected_control  = htile_control | (descriptor.MsaaDepth() ? (1u << 10u) : 0u);
 	const auto     metadata_addr     = descriptor.MetaAddr() << 8u;
-	return (descriptor.fields[6] & 0x00ffffffu) == htile_control && metadata_addr != 0 &&
+	return (descriptor.fields[6] & 0x00ffffffu) == expected_control && metadata_addr != 0 &&
 	       descriptor.TileMode() == Prospero::GpuEnumValue(Prospero::TileMode::kDepth) &&
 	       image.info.tile_mode == Prospero::GpuEnumValue(Prospero::TileMode::kDepth) &&
 	       image.info.metadata.kind == ImageMetadataKind::Htile &&
@@ -377,10 +424,14 @@ void ValidateStorageTexture(const ShaderRecompiler::IR::ImageResource& resource,
 	const bool encoding_ok   = IsSupportedStorageTextureEncoding(descriptor);
 	const bool uint_resource =
 	    resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint;
+	const bool raw_sint_storage =
+	    format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32SInt) && uint_resource &&
+	    resource.written && !resource.read && !resource.atomic;
 	const bool format_ok =
-	    Prospero::IsSupportedTextureFormat(format) &&
+	    raw_sint_storage ||
+	    (Prospero::IsSupportedTextureFormat(format) &&
 	     uint_resource == Prospero::IsUintTextureFormat(format) &&
-	    (!resource.atomic || format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32UInt));
+	     (!resource.atomic || format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32UInt)));
 	if (resource_ok && descriptor_ok && encoding_ok && format_ok && size != 0) {
 		return;
 	}
@@ -518,6 +569,7 @@ static ImageViewInfo TextureViewInfo(const ShaderRecompiler::IR::ImageResource&
 			view.layer_count = 1;
 			break;
 		case ShaderRecompiler::Decoder::ImageDimension::Dim2DArray:
+		case ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaaArray:
 			view.type       = vk::ImageViewType::e2DArray;
 			view.base_layer = descriptor.BaseArray5();
 			if (view.base_layer >= image_layers) {
@@ -526,6 +578,7 @@ static ImageViewInfo TextureViewInfo(const ShaderRecompiler::IR::ImageResource&
 			view.layer_count = image_layers - view.base_layer;
 			break;
 		case ShaderRecompiler::Decoder::ImageDimension::Dim2D:
+		case ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaa:
 			view.type       = vk::ImageViewType::e2D;
 			view.base_layer = descriptor.BaseArray5();
 			if (view.base_layer >= image_layers) {
@@ -562,19 +615,26 @@ RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageResource&   reso
 	const auto base_level   = descriptor.BaseLevel();
 	const auto last_level   = descriptor.LastLevel();
 	const auto type         = TextureType(descriptor);
-	const bool multisampled =
-	    type == Prospero::ImageType::kColor2DMsaa || type == Prospero::ImageType::kColor2DMsaaArray;
+	const bool multisampled = IsMultisampledTexture(type);
 	const auto levels       = multisampled ? 1u : static_cast<uint32_t>(descriptor.MaxMip()) + 1u;
 	const auto tile         = descriptor.TileMode();
-	const bool msaa_tile  = tile == Prospero::GpuEnumValue(Prospero::TileMode::kRenderTarget);
+	const bool depth_tile = tile == Prospero::GpuEnumValue(Prospero::TileMode::kDepth);
+	const bool msaa_tile =
+	    depth_tile || tile == Prospero::GpuEnumValue(Prospero::TileMode::kRenderTarget);
 	const bool msaa_array = type == Prospero::ImageType::kColor2DMsaaArray;
 	if ((!multisampled && (base_level > last_level || last_level >= levels)) ||
 	    (multisampled &&
 	     (base_level != 0 || last_level == 0 || last_level > 3 ||
-	      descriptor.MaxMip() != last_level || !msaa_tile || descriptor.MsaaDepth() ||
+	      descriptor.MaxMip() != last_level || !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\n", base_level, last_level,
-		     levels);
+		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 "
+		     "dwords=%08x,%08x,%08x,%08x,%08x,%08x,%08x,%08x\n",
+		     base_level, last_level, levels, descriptor.MaxMip(), descriptor.Type(), 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]);
 	}
 	const auto samples = multisampled ? 1u << last_level : 1u;
 	const auto view_levels =
@@ -601,7 +661,8 @@ RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageResource&   reso
 	TileSizeAlign size {};
 	if (multisampled) {
 		const auto bytes = Prospero::NumBytesPerElement(format);
-		pitch            = TileGetRenderTargetPitch(width, bytes, last_level);
+		pitch = depth_tile ? TileGetDepthPitch(width, bytes, last_level)
+		                   : TileGetRenderTargetPitch(width, bytes, last_level);
 		if (pitch == 0 || !TileGetRenderTargetSize(width, height, pitch, bytes, size, last_level) ||
 		    size.size > UINT32_MAX / image_layers) {
 			EXIT("unsupported multisample texture layout\n");
@@ -616,12 +677,13 @@ RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageResource&   reso
 	                     (address & (static_cast<uint64_t>(size.align) - 1u)) != 0);
 	if (storage) {
 		ValidateStorageTexture(resource, descriptor, size.size);
-		m_context.GetBufferCache().ValidateGpuAccess(address, size.size, resource.read,
-		                                             resource.written);
 	}
 
 	const auto              pixel_format = TextureGetFormat(format);
-	const auto              storage_view_format = SrgbStorageViewFormat(pixel_format);
+	const auto              storage_view_format =
+	    storage && format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32SInt)
+	        ? vk::Format::eR32Uint
+	        : SrgbStorageViewFormat(pixel_format);
 	const auto              view_format = storage && storage_view_format != vk::Format::eUndefined
 	                                          ? storage_view_format
 	                                          : pixel_format;

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

@@ -116,8 +116,8 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 		if (attachment_samples == 0) {
 			attachment_samples = depth.samples;
 		} else if (attachment_samples != depth.samples) {
-			EXIT("mixed color/depth sample counts are unsupported: %u and %u\n",
-			     attachment_samples, depth.samples);
+			EXIT("mixed color/depth sample counts are unsupported: %u and %u\n", attachment_samples,
+			     depth.samples);
 		}
 	}
 	EXIT_IF(attachment_samples == 0 ||
@@ -203,9 +203,8 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 	LogPipelineTrace("CreatePipelineInternal begin", vs_id.hash0, vs_id.crc32, ps_id.hash0,
 	                 ps_id.crc32);
 	CreatePipelineInternal(m_graphics, m_descriptor_cache, *cached, rendering, vs_input_info,
-	                       vs_spirv, ps_input_info,
-	                       ps_spirv, static_params, vs_id.hash0, vs_id.crc32, ps_id.hash0,
-	                       ps_id.crc32, ps_active);
+	                       vs_spirv, ps_input_info, ps_spirv, static_params, vs_id.hash0,
+	                       vs_id.crc32, ps_id.hash0, ps_id.crc32, ps_active);
 	LogPipelineTrace("CreatePipelineInternal done", vs_id.hash0, vs_id.crc32, ps_id.hash0,
 	                 ps_id.crc32);
 

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

@@ -118,11 +118,12 @@ public:
 		ShaderId cs_shader_id;
 	};
 
-	GraphicsPipeline& CreateGraphicsPipeline(
-	    RenderColorInfo* colors, uint32_t color_count, RenderDepthInfo& depth,
+	GraphicsPipeline&
+	CreateGraphicsPipeline(RenderColorInfo* colors, uint32_t color_count, RenderDepthInfo& depth,
 	                       ShaderVertexInputInfo& vs_input_info, RenderCommandBuffer& command,
-	    ShaderPixelInputInfo* ps_input_info, vk::PrimitiveTopology topology, bool ps_active,
-	    std::span<const uint32_t> vs_spirv, std::span<const uint32_t> ps_spirv);
+	                       ShaderPixelInputInfo* ps_input_info, vk::PrimitiveTopology topology,
+	                       bool ps_active, std::span<const uint32_t> vs_spirv,
+	                       std::span<const uint32_t> ps_spirv);
 	ComputePipeline& CreateComputePipeline(ShaderComputeInputInfo&      input_info,
 	                                       const HW::ComputeShaderInfo& cs_regs,
 	                                       std::span<const uint32_t>    cs_spirv);
@@ -211,16 +212,13 @@ private:
 
 void LogPipelineTrace(const char* phase, uint32_t vs_hash0, uint32_t vs_crc32, uint32_t ps_hash0,
                       uint32_t ps_crc32);
-void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descriptor_cache,
-                            PipelineCache::GraphicsPipeline& pipeline,
-                            const PipelineRenderingState&   rendering,
-                            const ShaderVertexInputInfo&    vs_input_info,
-                            std::span<const uint32_t>       vs_shader,
-                            const ShaderPixelInputInfo*     ps_input_info,
-                            std::span<const uint32_t>       ps_shader,
-                            const PipelineStaticParameters& static_params, uint32_t vs_hash0,
-                            uint32_t vs_crc32, uint32_t ps_hash0, uint32_t ps_crc32,
-                            bool ps_active);
+void CreatePipelineInternal(
+    GraphicContext& graphics, DescriptorCache& descriptor_cache,
+    PipelineCache::GraphicsPipeline& pipeline, const PipelineRenderingState& rendering,
+    const ShaderVertexInputInfo& vs_input_info, std::span<const uint32_t> vs_shader,
+    const ShaderPixelInputInfo* ps_input_info, std::span<const uint32_t> ps_shader,
+    const PipelineStaticParameters& static_params, uint32_t vs_hash0, uint32_t vs_crc32,
+    uint32_t ps_hash0, uint32_t ps_crc32, bool ps_active);
 void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descriptor_cache,
                             PipelineCache::ComputePipeline& pipeline,
                             const ShaderComputeInputInfo&   input_info,

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

@@ -8,10 +8,10 @@
 #include "graphics/host_gpu/renderer/debug.h"
 #include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
+#include "graphics/host_gpu/renderer/pipeline/shaderSubgroup.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/renderer/renderTarget.h"
-#include "graphics/host_gpu/renderer/pipeline/shaderSubgroup.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 #include "graphics/shader/recompiler/ir/ShaderIR.h"
 #include "graphics/shader/shader.h"
@@ -386,8 +386,7 @@ static vk::BlendOp GetBlendOp(uint32_t op) {
 }
 
 static void CreateLayout(DescriptorCache&                   descriptor_cache,
-                         std::span<vk::DescriptorSetLayout> set_layouts,
-                         uint32_t& set_layouts_num,
+                         std::span<vk::DescriptorSetLayout> set_layouts, uint32_t& set_layouts_num,
                          std::span<vk::PushConstantRange>     push_constant_info,
                          uint32_t&                            push_constant_info_num,
                          const ShaderRecompiler::IR::Program& program,
@@ -412,8 +411,7 @@ static void CreateLayout(DescriptorCache& descriptor_cache,
 	}
 }
 
-static void ConfigureSubgroupSize(const GraphicContext&                                  graphics,
-                                  vk::ShaderStageFlagBits                                vk_stage,
+static void ConfigureSubgroupSize(const GraphicContext& graphics, vk::ShaderStageFlagBits vk_stage,
                                   const ShaderRecompiler::IR::Program&                   program,
                                   vk::PipelineShaderStageRequiredSubgroupSizeCreateInfo& required,
                                   vk::PipelineShaderStageCreateInfo&                     stage) {
@@ -456,16 +454,13 @@ static void ConfigureSubgroupSize(const GraphicContext&
 }
 
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
-void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descriptor_cache,
-                            PipelineCache::GraphicsPipeline& pipeline,
-                            const PipelineRenderingState&   rendering,
-                            const ShaderVertexInputInfo&    vs_input_info,
-                            std::span<const uint32_t>       vs_shader,
-                            const ShaderPixelInputInfo*     ps_input_info,
-                            std::span<const uint32_t>       ps_shader,
-                            const PipelineStaticParameters& static_params, uint32_t vs_hash0,
-                            uint32_t vs_crc32, uint32_t ps_hash0, uint32_t ps_crc32,
-                            bool ps_active) {
+void CreatePipelineInternal(
+    GraphicContext& graphics, DescriptorCache& descriptor_cache,
+    PipelineCache::GraphicsPipeline& pipeline, const PipelineRenderingState& rendering,
+    const ShaderVertexInputInfo& vs_input_info, std::span<const uint32_t> vs_shader,
+    const ShaderPixelInputInfo* ps_input_info, std::span<const uint32_t> ps_shader,
+    const PipelineStaticParameters& static_params, uint32_t vs_hash0, uint32_t vs_crc32,
+    uint32_t ps_hash0, uint32_t ps_crc32, bool ps_active) {
 	EXIT_IF(ps_active && ps_input_info == nullptr);
 
 	vk::ShaderModule vert_shader_module = nullptr;
@@ -511,8 +506,7 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	vert_shader_stage_info.pName               = "main";
 	vert_shader_stage_info.pSpecializationInfo = nullptr;
 	EXIT_IF(!vs_input_info.stage);
-	ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eVertex,
-	                      *vs_input_info.stage.program,
+	ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eVertex, *vs_input_info.stage.program,
 	                      vert_subgroup_size, vert_shader_stage_info);
 
 	vk::PipelineShaderStageCreateInfo                     frag_shader_stage_info {};
@@ -527,8 +521,8 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	if (ps_active) {
 		EXIT_IF(!ps_input_info->stage);
 		ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eFragment,
-		                      *ps_input_info->stage.program,
-		                      frag_subgroup_size, frag_shader_stage_info);
+		                      *ps_input_info->stage.program, frag_subgroup_size,
+		                      frag_shader_stage_info);
 	}
 
 	vk::PipelineShaderStageCreateInfo shader_stages[]    = {vert_shader_stage_info,
@@ -838,15 +832,13 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 
 	EXIT_IF(!vs_input_info.stage);
 	CreateLayout(descriptor_cache, set_layouts, set_layouts_num, push_constant_info,
-	             push_constant_info_num,
-	             *vs_input_info.stage.program, vk::ShaderStageFlagBits::eVertex,
-	             DescriptorCache::Stage::Vertex);
+	             push_constant_info_num, *vs_input_info.stage.program,
+	             vk::ShaderStageFlagBits::eVertex, DescriptorCache::Stage::Vertex);
 	if (ps_active) {
 		EXIT_IF(!ps_input_info->stage);
 		CreateLayout(descriptor_cache, set_layouts, set_layouts_num, push_constant_info,
-		             push_constant_info_num,
-		             *ps_input_info->stage.program, vk::ShaderStageFlagBits::eFragment,
-		             DescriptorCache::Stage::Pixel);
+		             push_constant_info_num, *ps_input_info->stage.program,
+		             vk::ShaderStageFlagBits::eFragment, DescriptorCache::Stage::Pixel);
 	}
 
 	vk::PipelineLayoutCreateInfo pipeline_layout_info {};
@@ -1012,8 +1004,7 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	comp_shader_stage_info.pName               = "main";
 	comp_shader_stage_info.pSpecializationInfo = nullptr;
 	EXIT_IF(!input_info.stage);
-	ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eCompute,
-	                      *input_info.stage.program,
+	ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eCompute, *input_info.stage.program,
 	                      comp_subgroup_size, comp_shader_stage_info);
 
 	vk::DescriptorSetLayout set_layouts[1]  = {};
@@ -1024,9 +1015,8 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 
 	EXIT_IF(!input_info.stage);
 	CreateLayout(descriptor_cache, set_layouts, set_layouts_num, push_constant_info,
-	             push_constant_info_num,
-	             *input_info.stage.program, vk::ShaderStageFlagBits::eCompute,
-	             DescriptorCache::Stage::Compute);
+	             push_constant_info_num, *input_info.stage.program,
+	             vk::ShaderStageFlagBits::eCompute, DescriptorCache::Stage::Compute);
 
 	vk::PipelineLayoutCreateInfo pipeline_layout_info {};
 	pipeline_layout_info.sType                  = vk::StructureType::ePipelineLayoutCreateInfo;

src/graphics/host_gpu/renderer/renderCompute.cpp

@@ -10,14 +10,14 @@
 #include "graphics/guest_gpu/graphicsRun.h"
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/host_gpu/graphicContext.h"
+#include "graphics/host_gpu/renderer/image/imageInfo.h"
 #include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/pipeline/descriptors.h"
-#include "graphics/host_gpu/renderer/image/imageInfo.h"
 #include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
-#include "graphics/host_gpu/renderer/render.h"
-#include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/renderer/pipeline/shaderResourceBarrier.h"
 #include "graphics/host_gpu/renderer/pipeline/shaderSubgroup.h"
+#include "graphics/host_gpu/renderer/render.h"
+#include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 #include "graphics/shader/recompiler/ir/ResourceMaterialization.h"
 #include "graphics/shader/recompiler/ir/ShaderIR.h"

src/graphics/host_gpu/renderer/renderContext.cpp

@@ -14,8 +14,7 @@ namespace Libs::Graphics {
 RenderContext::RenderContext(GraphicContext& graphics)
     : m_graphics(graphics), m_render_executor(*this), m_command_scheduler(*this, graphics),
       m_descriptor_cache(graphics), m_pipeline_cache(graphics, m_descriptor_cache),
-      m_sampler_cache(graphics),
-      m_gpu_resources(graphics, m_command_scheduler) {
+      m_sampler_cache(graphics), m_gpu_resources(graphics, m_command_scheduler) {
 	EXIT_NOT_IMPLEMENTED(!Common::Thread::IsMainThread());
 }
 
@@ -99,8 +98,7 @@ void RenderContext::TriggerEopEvent(uint32_t context_id) {
 		    registration.eq, static_cast<uintptr_t>(registration.id),
 		    LibKernel::EventQueue::KERNEL_EVFILT_GRAPHICS,
 		    reinterpret_cast<void*>(static_cast<uintptr_t>(context_id)));
-		if (result == LibKernel::KERNEL_ERROR_EBADF ||
-		    result == LibKernel::KERNEL_ERROR_ENOENT) {
+		if (result == LibKernel::KERNEL_ERROR_EBADF || result == LibKernel::KERNEL_ERROR_ENOENT) {
 			DeleteEopEq(registration.eq, registration.id);
 			continue;
 		}

src/graphics/host_gpu/renderer/renderContext.h

@@ -6,12 +6,12 @@
 #include "common/common.h"
 #include "common/threads.h"
 #include "graphics/host_gpu/renderer/cache/bufferCache.h"
-#include "graphics/host_gpu/renderer/commandScheduler.h"
-#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/cache/gpuResourceManager.h"
-#include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
 #include "graphics/host_gpu/renderer/cache/samplerCache.h"
 #include "graphics/host_gpu/renderer/cache/textureCache.h"
+#include "graphics/host_gpu/renderer/commandScheduler.h"
+#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
+#include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
 #include "kernel/eventQueue.h"
 
 #include <memory>

src/graphics/host_gpu/renderer/sync.cpp

@@ -252,8 +252,8 @@ uint64_t PrepareVideoOutFlip(CommandBuffer& buffer, int handle, int index, int f
 	for (;;) {
 		uint64_t   request_id = 0;
 		auto&      video_out  = buffer.GetContext().GetVideoOut();
-		const auto result = video_out.SubmitFlipFromGpu(
-		    buffer, handle, index, flip_mode, flip_arg, request_id);
+		const auto result =
+		    video_out.SubmitFlipFromGpu(buffer, handle, index, flip_mode, flip_arg, request_id);
 		if (result == OK) {
 			EXIT_IF(request_id == 0);
 			return request_id;

src/graphics/host_gpu/vma.cpp

@@ -122,8 +122,8 @@ uint64_t GraphicContext::GetDeviceMemoryUsage() const {
 	    physical_device_properties.deviceType == vk::PhysicalDeviceType::eDiscreteGpu;
 	uint64_t usage = 0;
 	for (uint32_t heap = 0; heap < physical_device_memory_properties.memoryHeapCount; heap++) {
-		const bool device_local = static_cast<bool>(
-		    physical_device_memory_properties.memoryHeaps[heap].flags &
+		const bool device_local =
+		    static_cast<bool>(physical_device_memory_properties.memoryHeaps[heap].flags &
 		                      vk::MemoryHeapFlagBits::eDeviceLocal);
 		if (!discrete || device_local) {
 			usage += budgets[heap].usage;
@@ -160,8 +160,7 @@ uint64_t GraphicContext::GetTotalMemoryBudget() const {
 	}
 	constexpr uint64_t system_reserve = 8ull * 1024 * 1024 * 1024;
 	const auto         available      = budget > usage ? budget - usage : uint64_t {0};
-	return std::max(local,
-	                available > system_reserve ? available - system_reserve : uint64_t {0});
+	return std::max(local, available > system_reserve ? available - system_reserve : uint64_t {0});
 }
 
 void GraphicContext::CreateBuffer(uint64_t size, VulkanBuffer& buffer) {

src/graphics/host_gpu/vulkanCommon.cpp

@@ -37,6 +37,7 @@ constexpr FormatMapping kFormatMappings[] = {
     {Prospero::BufferFormat::k16_16Float, vk::Format::eR16G16Sfloat},
     {Prospero::BufferFormat::k11_11_10Float, vk::Format::eB10G11R11UfloatPack32},
     {Prospero::BufferFormat::k10_10_10_2UNorm, vk::Format::eA2B10G10R10UnormPack32},
+    {Prospero::BufferFormat::k10_10_10_2UInt, vk::Format::eA2B10G10R10UintPack32},
     {Prospero::BufferFormat::k8_8_8_8UNorm, vk::Format::eR8G8B8A8Unorm},
     {Prospero::BufferFormat::k8_8_8_8SNorm, vk::Format::eR8G8B8A8Snorm},
     {Prospero::BufferFormat::k8_8_8_8UInt, vk::Format::eR8G8B8A8Uint},
@@ -55,6 +56,10 @@ constexpr FormatMapping kFormatMappings[] = {
     {Prospero::BufferFormat::k32_32_32_32UInt, vk::Format::eR32G32B32A32Uint},
     {Prospero::BufferFormat::k32_32_32_32SInt, vk::Format::eR32G32B32A32Sint},
     {Prospero::BufferFormat::k32_32_32_32Float, vk::Format::eR32G32B32A32Sfloat},
+    // Narrow-channel sRGB formats are optional in Vulkan. Keep a same-width fallback until
+    // sampler-aware sRGB emulation is available.
+    {Prospero::BufferFormat::k8Srgb, vk::Format::eR8Unorm},
+    {Prospero::BufferFormat::k8_8Srgb, vk::Format::eR8G8Unorm},
     {Prospero::BufferFormat::k8_8_8_8Srgb, vk::Format::eR8G8B8A8Srgb},
     {Prospero::BufferFormat::k9_9_9_5Float, vk::Format::eE5B9G9R9UfloatPack32},
     {Prospero::BufferFormat::k5_6_5UNorm, vk::Format::eB5G6R5UnormPack16},

src/graphics/presentation/videoOut.cpp

@@ -359,9 +359,9 @@ static void TriggerVideoOutEvents(VideoOutConfig& video_out, VideoOutEventKind k
 		if (!registration || registration->generation != video_out.generation) {
 			continue;
 		}
-		const auto result = EventQueue::KernelTriggerEvent(
-		    registration->handle, VideoOutEventId(kind), EventQueue::KERNEL_EVFILT_VIDEO_OUT,
-		    trigger_data);
+		const auto result =
+		    EventQueue::KernelTriggerEvent(registration->handle, VideoOutEventId(kind),
+		                                   EventQueue::KERNEL_EVFILT_VIDEO_OUT, trigger_data);
 		EXIT_NOT_IMPLEMENTED(result != OK && result != LibKernel::KERNEL_ERROR_EBADF &&
 		                     result != LibKernel::KERNEL_ERROR_ENOENT);
 	}
@@ -372,9 +372,8 @@ static void DeleteVideoOutEvents(const VideoOutEventQueues& queues, VideoOutEven
 		if (!registration) {
 			continue;
 		}
-		const auto result =
-		    EventQueue::KernelDeleteEvent(registration->handle, VideoOutEventId(kind),
-		                                  EventQueue::KERNEL_EVFILT_VIDEO_OUT);
+		const auto result = EventQueue::KernelDeleteEvent(
+		    registration->handle, VideoOutEventId(kind), EventQueue::KERNEL_EVFILT_VIDEO_OUT);
 		EXIT_NOT_IMPLEMENTED(result != OK && result != LibKernel::KERNEL_ERROR_EBADF &&
 		                     result != LibKernel::KERNEL_ERROR_ENOENT);
 	}
@@ -425,17 +424,15 @@ static int RegisterVideoOutEvent(int handle, EventQueue::KernelEqueue eq, VideoO
 	bool                         add_queue = false;
 	{
 		Common::LockGuard event_lock(event_state->mutex);
-		const auto existing = std::find_if(queues.begin(), queues.end(), [&](const auto& candidate) {
+		const auto        existing =
+		    std::find_if(queues.begin(), queues.end(), [&](const auto& candidate) {
 			    return candidate->handle == eq && candidate->generation == generation;
 		    });
 		if (existing != queues.end()) {
 			registration = *existing;
 		} else {
-			registration = std::make_shared<VideoOutEventRegistration>(
-			    VideoOutEventRegistration {.handle     = eq,
-			                               .state      = event_state,
-			                               .generation = generation,
-			                               .kind       = kind});
+			registration = std::make_shared<VideoOutEventRegistration>(VideoOutEventRegistration {
+			    .handle = eq, .state = event_state, .generation = generation, .kind = kind});
 			queues.push_back(registration);
 			add_queue = true;
 		}
@@ -814,8 +811,8 @@ void VideoOutDriver::Impl::PresentThread(std::stop_token token) {
 				m_presenter.Present(*frame, true);
 			}
 			const auto frame_end = Common::Timer::QueryPerformanceCounter();
-			total_wait += static_cast<int64_t>(period) -
-			              static_cast<int64_t>(frame_end - frame_begin);
+			total_wait +=
+			    static_cast<int64_t>(period) - static_cast<int64_t>(frame_end - frame_begin);
 			continue;
 		}
 
@@ -841,8 +838,7 @@ void VideoOutDriver::Impl::PresentThread(std::stop_token token) {
 		VblankEnd();
 
 		const auto frame_end = Common::Timer::QueryPerformanceCounter();
-		total_wait += static_cast<int64_t>(period) -
-		              static_cast<int64_t>(frame_end - frame_begin);
+		total_wait += static_cast<int64_t>(period) - static_cast<int64_t>(frame_end - frame_begin);
 	}
 }
 
@@ -1000,8 +996,8 @@ void FlipQueue::Prepare(uint64_t request_id, Graphics::CommandBuffer& buffer) {
 	}
 	Graphics::Presenter::Frame* frame = nullptr;
 	if (special) {
-		frame = &m_presenter.PrepareBlankFrame(width, height,
-		                                       index == VIDEO_OUT_BUFFER_INDEX_BLACK, &buffer);
+		frame = &m_presenter.PrepareBlankFrame(width, height, index == VIDEO_OUT_BUFFER_INDEX_BLACK,
+		                                       &buffer);
 	} else {
 		frame = &m_presenter.PrepareFrame(buffer, source_info);
 	}

src/graphics/presentation/window/swapchain.cpp

@@ -206,21 +206,18 @@ private:
 	vk::Format                                     m_format     = vk::Format::eUndefined;
 };
 
-void Presenter::Frame::Configure(GraphicContext& graphics, vk::Extent2D extent,
-                                 vk::Format format) {
+void Presenter::Frame::Configure(GraphicContext& graphics, vk::Extent2D extent, vk::Format format) {
 	if (extent.width == 0 || extent.height == 0 || format == vk::Format::eUndefined) {
 		EXIT("unsupported prepared frame, extent=%ux%u format=%d\n", extent.width, extent.height,
 		     static_cast<int>(format));
 	}
 	const auto features = graphics.GetFormatProperties(format).optimalTilingFeatures;
-	const auto required = vk::FormatFeatureFlagBits::eBlitSrc |
-	                      vk::FormatFeatureFlagBits::eSampledImageFilterLinear |
-	                      vk::FormatFeatureFlagBits::eTransferSrc |
-	                      vk::FormatFeatureFlagBits::eTransferDst;
+	const auto required =
+	    vk::FormatFeatureFlagBits::eBlitSrc | vk::FormatFeatureFlagBits::eSampledImageFilterLinear |
+	    vk::FormatFeatureFlagBits::eTransferSrc | vk::FormatFeatureFlagBits::eTransferDst;
 	if ((features & required) != required) {
 		EXIT("prepared presentation format lacks optimal blit support: format=%d features=0x%x\n",
-		     static_cast<int>(format),
-		     static_cast<vk::FormatFeatureFlags::MaskType>(features));
+		     static_cast<int>(format), static_cast<vk::FormatFeatureFlags::MaskType>(features));
 	}
 
 	auto&      dst        = image;
@@ -267,8 +264,7 @@ void Presenter::Frame::Transit(vk::CommandBuffer command, vk::ImageLayout layout
 	                            ? vk::PipelineStageFlagBits2::eTransfer
 	                            : vk::PipelineStageFlagBits2::eAllCommands;
 	constexpr auto writes = vk::AccessFlagBits2::eTransferWrite |
-	                        vk::AccessFlagBits2::eShaderWrite |
-	                        vk::AccessFlagBits2::eMemoryWrite;
+	                        vk::AccessFlagBits2::eShaderWrite | vk::AccessFlagBits2::eMemoryWrite;
 	if (image.state.layout == layout && image.state.access_mask == access &&
 	    !static_cast<bool>(image.state.access_mask & writes)) {
 		return;
@@ -299,35 +295,27 @@ void Presenter::Frame::Transit(vk::CommandBuffer command, vk::ImageLayout layout
 void Presenter::Frame::CopyFrom(CommandBuffer& command_buffer, Image& source) {
 	command_buffer.EndRendering();
 	auto command = command_buffer.Handle();
-	source.Transit(vk::ImageLayout::eTransferSrcOptimal,
-	               vk::AccessFlagBits2::eTransferRead, {}, command);
-	Transit(command, vk::ImageLayout::eTransferDstOptimal,
-	        vk::AccessFlagBits2::eTransferWrite);
+	source.Transit(vk::ImageLayout::eTransferSrcOptimal, vk::AccessFlagBits2::eTransferRead, {},
+	               command);
+	Transit(command, vk::ImageLayout::eTransferDstOptimal, vk::AccessFlagBits2::eTransferWrite);
 	vk::ImageCopy copy {};
-	copy.srcSubresource = {vk::ImageAspectFlagBits::eColor, 0, 0,
-	                       source.backing.layers};
+	copy.srcSubresource = {vk::ImageAspectFlagBits::eColor, 0, 0, source.backing.layers};
 	copy.dstSubresource = {vk::ImageAspectFlagBits::eColor, 0, 0, image.layers};
 	copy.extent         = {std::min(source.backing.extent.width, image.extent.width),
 	                       std::min(source.backing.extent.height, image.extent.height), 1};
 	EXIT_IF(copy.srcSubresource.layerCount != copy.dstSubresource.layerCount);
-	command.copyImage(source.backing.image, vk::ImageLayout::eTransferSrcOptimal,
-	                  image.image, vk::ImageLayout::eTransferDstOptimal, copy);
-	Transit(command, vk::ImageLayout::eTransferSrcOptimal,
-	        vk::AccessFlagBits2::eTransferRead);
+	command.copyImage(source.backing.image, vk::ImageLayout::eTransferSrcOptimal, image.image,
+	                  vk::ImageLayout::eTransferDstOptimal, copy);
+	Transit(command, vk::ImageLayout::eTransferSrcOptimal, vk::AccessFlagBits2::eTransferRead);
 }
 
-void Presenter::Frame::Clear(CommandBuffer& command_buffer,
-                             const vk::ClearColorValue& color) {
+void Presenter::Frame::Clear(CommandBuffer& command_buffer, const vk::ClearColorValue& color) {
 	command_buffer.EndRendering();
 	auto command = command_buffer.Handle();
-	Transit(command, vk::ImageLayout::eTransferDstOptimal,
-	        vk::AccessFlagBits2::eTransferWrite);
-	const vk::ImageSubresourceRange range {
-	    vk::ImageAspectFlagBits::eColor, 0, 1, 0, 1};
-	command.clearColorImage(image.image, vk::ImageLayout::eTransferDstOptimal, &color, 1,
-	                        &range);
-	Transit(command, vk::ImageLayout::eTransferSrcOptimal,
-	        vk::AccessFlagBits2::eTransferRead);
+	Transit(command, vk::ImageLayout::eTransferDstOptimal, vk::AccessFlagBits2::eTransferWrite);
+	const vk::ImageSubresourceRange range {vk::ImageAspectFlagBits::eColor, 0, 1, 0, 1};
+	command.clearColorImage(image.image, vk::ImageLayout::eTransferDstOptimal, &color, 1, &range);
+	Transit(command, vk::ImageLayout::eTransferSrcOptimal, vk::AccessFlagBits2::eTransferRead);
 }
 
 class Swapchain final {
@@ -441,16 +429,12 @@ void Swapchain::Create() {
 	        ? vk::CompositeAlphaFlagBitsKHR::eOpaque
 	        : vk::CompositeAlphaFlagBitsKHR::eInherit;
 
-	vk::SurfaceFormatKHR format {vk::Format::eR8G8B8A8Unorm,
-	                             vk::ColorSpaceKHR::eSrgbNonlinear};
-	if (surface.formats.size() != 1 ||
-	    surface.formats.front().format != vk::Format::eUndefined) {
+	vk::SurfaceFormatKHR format {vk::Format::eR8G8B8A8Unorm, vk::ColorSpaceKHR::eSrgbNonlinear};
+	if (surface.formats.size() != 1 || surface.formats.front().format != vk::Format::eUndefined) {
 		const auto it = std::find_if(surface.formats.begin(), surface.formats.end(),
 		                             [](const vk::SurfaceFormatKHR& candidate) {
-			                             return candidate.format ==
-			                                        vk::Format::eB8G8R8A8Unorm ||
-			                                    candidate.format ==
-			                                        vk::Format::eR8G8B8A8Unorm;
+			                             return candidate.format == vk::Format::eB8G8R8A8Unorm ||
+			                                    candidate.format == vk::Format::eR8G8B8A8Unorm;
 		                             });
 		if (it == surface.formats.end()) {
 			EXIT("no supported UNORM swapchain format\n");
@@ -458,8 +442,7 @@ void Swapchain::Create() {
 		format = *it;
 	}
 	m_format                      = format.format;
-	const auto swapchain_features =
-	    graphics.GetFormatProperties(m_format).optimalTilingFeatures;
+	const auto swapchain_features = graphics.GetFormatProperties(m_format).optimalTilingFeatures;
 	if (!static_cast<bool>(swapchain_features & vk::FormatFeatureFlagBits::eBlitDst)) {
 		EXIT("swapchain format cannot be a blit destination: format=%d\n",
 		     static_cast<int>(m_format));
@@ -503,8 +486,7 @@ void Swapchain::Create() {
 		view.subresourceRange.baseMipLevel   = 0;
 		view.subresourceRange.layerCount     = 1;
 		view.subresourceRange.levelCount     = 1;
-		RequireVulkanSuccess(
-		    graphics.device.createImageView(&view, nullptr, &m_image_views[i]),
+		RequireVulkanSuccess(graphics.device.createImageView(&view, nullptr, &m_image_views[i]),
 		                     "vkCreateImageView");
 		EXIT_IF(m_image_views[i] == nullptr);
 	}
@@ -683,10 +665,9 @@ void Swapchain::RecordPresentCommands(CommandBuffer& command, VulkanImage& sourc
 	to_present.subresourceRange.levelCount     = 1;
 	to_present.subresourceRange.baseArrayLayer = 0;
 	to_present.subresourceRange.layerCount     = 1;
-	vk_command.pipelineBarrier(vk::PipelineStageFlagBits::eAllCommands,
-	                           vk::PipelineStageFlagBits::eAllCommands,
-	                           vk::DependencyFlagBits::eByRegion, 0,
-	                           nullptr, 0, nullptr, 1, &to_present);
+	vk_command.pipelineBarrier(
+	    vk::PipelineStageFlagBits::eAllCommands, vk::PipelineStageFlagBits::eAllCommands,
+	    vk::DependencyFlagBits::eByRegion, 0, nullptr, 0, nullptr, 1, &to_present);
 	command.End();
 }
 
@@ -752,8 +733,7 @@ Presenter::Frame& Presenter::PrepareFrame(CommandBuffer& buffer, const ImageInfo
 		default: break;
 	}
 	frame->Configure(m_impl->window.graphic_ctx,
-	                 {image.backing.extent.width, image.backing.extent.height},
-	                 frame_format);
+	                 {image.backing.extent.width, image.backing.extent.height}, frame_format);
 	frame->CopyFrom(buffer, image);
 	return *frame;
 }
@@ -772,8 +752,7 @@ Presenter::Frame& Presenter::PrepareBlankFrame(uint32_t width, uint32_t height,
 		frame->Clear(*producer, clear);
 	} else {
 		if (frame->present_commands == nullptr) {
-			frame->present_commands =
-			    std::make_unique<CommandBuffer>(m_impl->present_scheduler);
+			frame->present_commands = std::make_unique<CommandBuffer>(m_impl->present_scheduler);
 		}
 		auto& command = *frame->present_commands;
 		command.WaitForFenceAndReset();
@@ -830,8 +809,7 @@ void Presenter::Present(Frame& frame, bool reuse) {
 			continue;
 		}
 		if (frame.present_commands == nullptr) {
-			frame.present_commands =
-			    std::make_unique<CommandBuffer>(m_impl->present_scheduler);
+			frame.present_commands = std::make_unique<CommandBuffer>(m_impl->present_scheduler);
 		}
 		{
 			Common::LockGuard render_lock(m_impl->renderer.GetMutex());

src/graphics/presentation/window/vulkanWindow.cpp

@@ -32,11 +32,11 @@
 #include "graphics/host_gpu/vma.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 #include "graphics/presentation/presenter.h"
-#include "kernel/memory.h"
 #include "graphics/presentation/renderDoc.h"
 #include "graphics/presentation/videoOut.h"
 #include "graphics/presentation/window.h"
 #include "graphics/presentation/window/windowInternal.h"
+#include "kernel/memory.h"
 #include "libs/controller.h"
 #include "loader/systemContent.h"
 
@@ -909,10 +909,9 @@ void WindowContext::CreateVulkan() {
 	}
 	surface = native_surface;
 
-	std::vector<const char*> device_extensions = {VK_KHR_SWAPCHAIN_EXTENSION_NAME,
-	                                              VK_EXT_DEPTH_CLIP_CONTROL_EXTENSION_NAME,
-	                                              VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME,
-	                                              "VK_KHR_maintenance1"};
+	std::vector<const char*> device_extensions = {
+	    VK_KHR_SWAPCHAIN_EXTENSION_NAME, VK_EXT_DEPTH_CLIP_CONTROL_EXTENSION_NAME,
+	    VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME, "VK_KHR_maintenance1"};
 
 #if defined(__APPLE__)
 	// MoltenVK lacks VK_EXT_depth_clip_enable and VK_EXT_color_write_enable; the renderer
@@ -932,8 +931,8 @@ void WindowContext::CreateVulkan() {
 
 	uint32_t queue_family = static_cast<uint32_t>(-1);
 
-	VulkanFindPhysicalDevice(graphic_ctx.instance, surface, device_extensions,
-	                         surface_capabilities, graphic_ctx.physical_device, queue_family);
+	VulkanFindPhysicalDevice(graphic_ctx.instance, surface, device_extensions, surface_capabilities,
+	                         graphic_ctx.physical_device, queue_family);
 
 	if (graphic_ctx.physical_device == nullptr) {
 		EXIT("Could not find suitable device");
@@ -949,9 +948,8 @@ void WindowContext::CreateVulkan() {
 		auto available_extensions = EnumerateVulkan<vk::ExtensionProperties>(
 		    "vkEnumerateDeviceExtensionProperties",
 		    [&](uint32_t* count, vk::ExtensionProperties* values) {
-			    return graphic_ctx.physical_device.enumerateDeviceExtensionProperties(nullptr,
-			                                                                           count,
-			                                                                           values);
+			    return graphic_ctx.physical_device.enumerateDeviceExtensionProperties(
+			        nullptr, count, values);
 		    });
 
 		if (HasExtension(available_extensions, VK_EXT_MEMORY_BUDGET_EXTENSION_NAME)) {

src/graphics/presentation/window/window.cpp

@@ -1,7 +1,5 @@
 #include "graphics/presentation/window.h"
 
-#include <cstdlib>
-
 #include "SDL.h"
 #include "SDL_error.h"
 #include "SDL_events.h"
@@ -40,6 +38,7 @@
 
 #include <algorithm>
 #include <cstdio>
+#include <cstdlib>
 #include <cstring>
 #include <memory>
 #include <string>
@@ -251,9 +250,7 @@ static void GameEventKeyboard(WindowLoopState& game, const EventKeyboard& key) {
 	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_SPACE: SetPause(game, !game.paused.load(std::memory_order_acquire)); break;
 			case SDLK_F1:
 				if (!key.repeat) {
 					RenderDocRequestCapture();
@@ -390,7 +387,9 @@ void WindowContext::Resize(uint32_t new_width, uint32_t new_height) {
 void WindowContext::ProcessWindowEvent(const SDL_WindowEvent& event) {
 	const auto& window_event = event;
 	switch (window_event.event) {
-		case SDL_WINDOWEVENT_SHOWN: LOGF("Window %" PRIu32 " shown\n", window_event.windowID); break;
+		case SDL_WINDOWEVENT_SHOWN:
+			LOGF("Window %" PRIu32 " shown\n", window_event.windowID);
+			break;
 
 		case SDL_WINDOWEVENT_HIDDEN:
 			LOGF("Window %" PRIu32 " hidden\n", window_event.windowID);
@@ -401,13 +400,13 @@ void WindowContext::ProcessWindowEvent(const SDL_WindowEvent& event) {
 			break;
 
 		case SDL_WINDOWEVENT_MOVED:
-			LOGF("Window %" PRIu32 " moved to %" PRId32 ",%" PRId32 "\n",
-			     window_event.windowID, window_event.data1, window_event.data2);
+			LOGF("Window %" PRIu32 " moved to %" PRId32 ",%" PRId32 "\n", window_event.windowID,
+			     window_event.data1, window_event.data2);
 			break;
 
 		case SDL_WINDOWEVENT_RESIZED:
-			LOGF("Window %" PRIu32 " resized to %" PRId32 "x%" PRId32 "\n",
-			     window_event.windowID, window_event.data1, window_event.data2);
+			LOGF("Window %" PRIu32 " resized to %" PRId32 "x%" PRId32 "\n", window_event.windowID,
+			     window_event.data1, window_event.data2);
 
 			LOGF("m: %d\n", static_cast<int>(SDL_ThreadID()));
 			Resize(window_event.data1, window_event.data2);
@@ -807,8 +806,7 @@ static void WindowCreate(WindowContext& context) {
 		window_flags |= static_cast<uint32_t>(SDL_WINDOW_BORDERLESS);
 	}
 #endif
-	context.window =
-	    SDL_CreateWindow(KYTY_SDL_WINDOW_CAPTION, KYTY_SDL_WINDOWPOS_CENTERED,
+	context.window = SDL_CreateWindow(KYTY_SDL_WINDOW_CAPTION, KYTY_SDL_WINDOWPOS_CENTERED,
 	                                  KYTY_SDL_WINDOWPOS_CENTERED, width, height, window_flags);
 
 	context.window_hidden = true;
@@ -950,11 +948,11 @@ void WindowContext::UpdateTitle() {
 		fps_frames  = 0;
 	}
 
-	auto fps = fmt::format("{}{}{}{}{}{}[{}] [{}], frame: {}, fps: {:f}", (has_title ? title : ""),
+	auto fps =
+	    fmt::format("{}{}{}{}{}{}[{}] [{}], frame: {}, fps: {:f}", (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);
+	                (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.

src/graphics/presentation/window/windowInternal.h

@@ -38,8 +38,7 @@ struct WindowContext {
 	~WindowContext();
 	KYTY_CLASS_NO_COPY(WindowContext);
 
-	[[nodiscard]] static vk::PhysicalDeviceVulkan13Features
-	RequiredVulkan13Features() noexcept;
+	[[nodiscard]] static vk::PhysicalDeviceVulkan13Features RequiredVulkan13Features() noexcept;
 	void                                                    CreateVulkan();
 	void                                                    RecreateSurface();
 	void                                                    RefreshSurfaceCapabilities();

src/graphics/shader/recompiler/ShaderRecompiler.cpp

@@ -2,15 +2,16 @@
 
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "graphics/shader/recompiler/cfg/ShaderCFG.h"
+#include "graphics/shader/recompiler/decompiler/ShaderDecoder.h"
+#include "graphics/shader/recompiler/emitter/SpirvEmitter.h"
 #include "graphics/shader/recompiler/ir/BindingLayout.h"
+#include "graphics/shader/recompiler/ir/ReadLaneElimination.h"
 #include "graphics/shader/recompiler/ir/ResourceMaterialization.h"
 #include "graphics/shader/recompiler/ir/ResourceTracking.h"
 #include "graphics/shader/recompiler/ir/ScalarProvenance.h"
-#include "graphics/shader/recompiler/cfg/ShaderCFG.h"
-#include "graphics/shader/recompiler/decompiler/ShaderDecoder.h"
 #include "graphics/shader/recompiler/ir/ShaderIR.h"
 #include "graphics/shader/recompiler/ir/ShaderInfoCollection.h"
-#include "graphics/shader/recompiler/emitter/SpirvEmitter.h"
 #include "graphics/shader/recompiler/ir/SrtPatcher.h"
 #include "graphics/shader/recompiler/ir/SrtWalker.h"
 
@@ -838,6 +839,11 @@ bool TryRecompile(std::span<const uint32_t> code, const CompileOptions& options,
 	if (!IR::AllocateBindings(ir, layout_options, error)) {
 		return false;
 	}
+	const auto read_lane_stats = IR::EliminateReadLane(ir);
+	if (read_lane_stats.rewritten_reads != 0) {
+		LOGF("%s read-lane elimination: reads=%" PRIu32 " shadow_writes=%" PRIu32 "\n",
+		     GetDumpLabel(options), read_lane_stats.rewritten_reads, read_lane_stats.shadow_writes);
+	}
 	std::string ir_dump;
 	if (options.dump_ir) {
 		ir_dump = MakeIrDump(cfg, ir);

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

@@ -871,19 +871,19 @@ std::vector<uint32_t> DominatedBlocks(const Graph& graph, uint32_t header,
 	return blocks;
 }
 
-uint32_t AppendSyntheticMergeBlock(Graph& graph, uint32_t old_merge) {
-	const auto* merge = graph.FindBlock(old_merge);
+uint32_t AppendSyntheticBranchBlock(Graph& graph, uint32_t target) {
+	const auto* target_block = graph.FindBlock(target);
 
 	BasicBlock block;
 	block.id                    = static_cast<uint32_t>(graph.blocks.size());
-	block.start_pc              = merge != nullptr ? merge->start_pc : 0u;
+	block.start_pc              = target_block != nullptr ? target_block->start_pc : 0u;
 	block.end_pc                = block.start_pc;
-	block.inst_begin            = merge != nullptr ? merge->inst_begin : 0u;
+	block.inst_begin            = target_block != nullptr ? target_block->inst_begin : 0u;
 	block.inst_end              = block.inst_begin;
-	block.successors            = {old_merge};
+	block.successors            = {target};
 	block.terminator.kind       = TerminatorKind::Branch;
 	block.terminator.condition  = BranchCondition::Always;
-	block.terminator.true_block = old_merge;
+	block.terminator.true_block = target;
 	graph.blocks.push_back(std::move(block));
 	return graph.blocks.back().id;
 }
@@ -897,15 +897,50 @@ bool IsSyntheticMergeForwarder(const Graph& graph, uint32_t block_id, uint32_t m
 	       block->terminator.true_block == merge;
 }
 
-bool IsInsideLoopConstruct(const Graph& graph, const NaturalLoop& loop, uint32_t block_id) {
-	return block_id != UINT32_MAX && block_id != loop.merge && block_id != loop.continue_block &&
-	       graph.Dominates(loop.header, block_id) &&
-	       (loop.merge == UINT32_MAX || !graph.Dominates(loop.merge, block_id));
+const NaturalLoop* FindInnermostContainingLoop(const Graph& graph, uint32_t block_id) {
+	const NaturalLoop* innermost = nullptr;
+	for (const auto& loop: graph.natural_loops) {
+		if (Contains(loop.body_blocks, block_id) &&
+		    (innermost == nullptr || loop.body_blocks.size() < innermost->body_blocks.size())) {
+			innermost = &loop;
+		}
+	}
+	return innermost;
 }
 
-bool SelectionMergeLeavesContainingLoop(const Graph& graph, uint32_t header, uint32_t merge) {
+bool IsInsideLoopConstruct(const Graph& graph, const NaturalLoop& loop, uint32_t block_id) {
+	return block_id != UINT32_MAX && block_id != loop.merge && block_id != loop.continue_block &&
+	       graph.Dominates(loop.header, block_id) && !graph.Dominates(loop.merge, block_id);
+}
+
+bool IsInnermostLoopControlConditional(const Graph& graph, const BasicBlock& block) {
+	if (block.terminator.kind != TerminatorKind::ConditionalBranch) {
+		return false;
+	}
+	const auto* loop = FindInnermostContainingLoop(graph, block.id);
+	if (loop == nullptr || loop->merge == UINT32_MAX || loop->continue_block == UINT32_MAX) {
+		return false;
+	}
+	const auto true_target  = block.terminator.true_block;
+	const auto false_target = block.terminator.false_block;
+	if (block.id == loop->continue_block) {
+		const auto is_repeat_target = [&](uint32_t target) {
+			return target == loop->header || target == loop->merge;
+		};
+		return is_repeat_target(true_target) && is_repeat_target(false_target);
+	}
+	const auto is_control_target = [&](uint32_t target) {
+		return target == loop->merge || target == loop->continue_block;
+	};
+	return (is_control_target(true_target) &&
+	        (is_control_target(false_target) ||
+	         IsInsideLoopConstruct(graph, *loop, false_target))) ||
+	       (is_control_target(false_target) && IsInsideLoopConstruct(graph, *loop, true_target));
+}
+
+bool MergeLeavesContainingLoop(const Graph& graph, uint32_t header, uint32_t merge) {
 	for (const auto& loop: graph.natural_loops) {
-		if (IsInsideLoopConstruct(graph, loop, header) &&
+		if (loop.header != header && IsInsideLoopConstruct(graph, loop, header) &&
 		    !IsInsideLoopConstruct(graph, loop, merge)) {
 			return true;
 		}
@@ -913,6 +948,80 @@ bool SelectionMergeLeavesContainingLoop(const Graph& graph, uint32_t header, uin
 	return false;
 }
 
+bool CanonicalizeNaturalLoops(Graph& graph, std::string* error) {
+	const auto rewrite_budget = graph.blocks.size() * 2u + 16u;
+	for (size_t rewrite = 0; rewrite < rewrite_budget; rewrite++) {
+		bool changed = false;
+		for (const auto& loop: graph.natural_loops) {
+			std::vector<uint32_t> latches;
+			for (const auto& edge: graph.back_edges) {
+				if (edge.to == loop.header) {
+					AddUnique(latches, edge.from);
+				}
+			}
+			if (latches.size() <= 1u) {
+				continue;
+			}
+
+			const auto continue_block = AppendSyntheticBranchBlock(graph, loop.header);
+			for (auto latch: latches) {
+				auto* block = graph.FindBlock(latch);
+				if (block != nullptr) {
+					ReplaceValue(block->successors, loop.header, continue_block);
+					ReplaceTerminatorTarget(block->terminator, loop.header, continue_block);
+				}
+			}
+			RebuildPredecessors(graph);
+			RecomputeAnalyses(graph);
+			changed = true;
+			break;
+		}
+		if (changed) {
+			continue;
+		}
+
+		for (const auto& loop: graph.natural_loops) {
+			const auto* header                 = graph.FindBlock(loop.header);
+			const auto  is_loop_control_target = [&](uint32_t target) {
+				return target == loop.merge || target == loop.continue_block;
+			};
+			if (header == nullptr || header->terminator.kind != TerminatorKind::ConditionalBranch ||
+			    is_loop_control_target(header->terminator.true_block) ||
+			    is_loop_control_target(header->terminator.false_block) ||
+			    !Contains(loop.body_blocks, header->terminator.true_block) ||
+			    !Contains(loop.body_blocks, header->terminator.false_block)) {
+				continue;
+			}
+
+			const auto old_header   = loop.header;
+			const auto predecessors = header->predecessors;
+			const auto new_header   = AppendSyntheticBranchBlock(graph, old_header);
+			for (auto pred: predecessors) {
+				auto* block = graph.FindBlock(pred);
+				if (block != nullptr) {
+					ReplaceValue(block->successors, old_header, new_header);
+					ReplaceTerminatorTarget(block->terminator, old_header, new_header);
+				}
+			}
+			if (graph.entry_block == old_header) {
+				graph.entry_block = new_header;
+			}
+			MoveBlockBefore(graph, new_header, old_header);
+			RebuildPredecessors(graph);
+			RecomputeAnalyses(graph);
+			changed = true;
+			break;
+		}
+		if (!changed) {
+			return true;
+		}
+	}
+
+	SetFailure(graph, FailureKind::StructuredControlFlow, graph.entry_block,
+	           "CFG loop canonicalization exceeded rewrite budget", error);
+	return false;
+}
+
 bool SplitSharedMergeBlock(Graph& graph, uint32_t merge,
                            const std::vector<uint32_t>& construct_blocks,
                            bool                         force_split = false) {
@@ -948,7 +1057,7 @@ bool SplitSharedMergeBlock(Graph& graph, uint32_t merge,
 		return false;
 	}
 
-	const auto synthetic_merge = AppendSyntheticMergeBlock(graph, merge);
+	const auto synthetic_merge = AppendSyntheticBranchBlock(graph, merge);
 	auto*      synthetic_block = graph.FindBlock(synthetic_merge);
 	if (synthetic_block != nullptr) {
 		synthetic_block->predecessors = predecessors_to_split;
@@ -980,14 +1089,111 @@ bool SplitSharedMergeBlock(Graph& graph, uint32_t merge,
 bool SplitOneLoopMerge(Graph& graph) {
 	const auto& loops = graph.natural_loops;
 	for (const auto& loop: loops) {
-		if (SplitSharedMergeBlock(graph, loop.merge, loop.body_blocks)) {
+		const auto construct_blocks = DominatedBlocks(graph, loop.header, loop.merge);
+		const auto force_split      = MergeLeavesContainingLoop(graph, loop.header, loop.merge);
+		if (SplitSharedMergeBlock(graph, loop.merge, construct_blocks, force_split)) {
 			return true;
 		}
 	}
 	return false;
 }
 
-bool SplitOneSelectionMerge(Graph& graph) {
+std::vector<uint32_t> SelectionRegion(const Graph& graph, const BasicBlock& header,
+                                      uint32_t merge) {
+	std::vector<uint32_t> region;
+	std::vector<uint32_t> pending = {header.terminator.true_block,
+	                                 header.terminator.false_block};
+	while (!pending.empty()) {
+		const auto block_id = pending.back();
+		pending.pop_back();
+		if (block_id == merge || Contains(region, block_id)) {
+			continue;
+		}
+		const auto* block = graph.FindBlock(block_id);
+		if (block == nullptr) {
+			continue;
+		}
+		AddUnique(region, block_id);
+		pending.insert(pending.end(), block->successors.begin(), block->successors.end());
+	}
+	SortUnique(region);
+	return region;
+}
+
+bool DuplicateSelectionRegion(Graph& graph, uint32_t header_id, uint32_t merge,
+                              const std::vector<uint32_t>& region, uint32_t block_budget) {
+	std::vector<uint32_t> cloned_blocks;
+	for (auto block_id: region) {
+		if (!graph.Dominates(header_id, block_id)) {
+			cloned_blocks.push_back(block_id);
+		}
+	}
+	if (cloned_blocks.empty() || graph.FindBlock(header_id) == nullptr || header_id >= merge ||
+	    graph.blocks.size() + cloned_blocks.size() + 1u > block_budget) {
+		return false;
+	}
+
+	const auto first_clone = static_cast<uint32_t>(graph.blocks.size());
+	std::map<uint32_t, uint32_t> clones;
+	for (uint32_t i = 0; i < cloned_blocks.size(); i++) {
+		clones.emplace(cloned_blocks[i], first_clone + i);
+	}
+
+	for (auto block_id: cloned_blocks) {
+		BasicBlock clone = *graph.FindBlock(block_id);
+		clone.id         = clones.at(block_id);
+		clone.predecessors.clear();
+		clone.dominators.clear();
+		clone.post_dominators.clear();
+		graph.blocks.push_back(std::move(clone));
+	}
+
+	const auto remap_block = [&](BasicBlock& block) {
+		const auto remap_target = [&](uint32_t& target) {
+			if (const auto it = clones.find(target); it != clones.end()) {
+				target = it->second;
+			}
+		};
+		for (auto& successor: block.successors) {
+			remap_target(successor);
+		}
+		remap_target(block.terminator.true_block);
+		remap_target(block.terminator.false_block);
+		remap_target(block.terminator.merge_block);
+		remap_target(block.terminator.continue_block);
+		for (auto& target: block.terminator.indirect_targets) {
+			remap_target(target);
+		}
+	};
+	for (auto block_id: region) {
+		const auto owned_id = clones.contains(block_id) ? clones.at(block_id) : block_id;
+		remap_block(*graph.FindBlock(owned_id));
+	}
+
+	const auto private_merge = AppendSyntheticBranchBlock(graph, merge);
+	auto&      header        = *graph.FindBlock(header_id);
+	remap_block(header);
+
+	for (auto block_id: region) {
+		const auto owned_id = clones.contains(block_id) ? clones.at(block_id) : block_id;
+		auto*      block    = graph.FindBlock(owned_id);
+		if (block != nullptr) {
+			ReplaceValue(block->successors, merge, private_merge);
+			ReplaceTerminatorTarget(block->terminator, merge, private_merge);
+		}
+	}
+	ReplaceValue(header.successors, merge, private_merge);
+	ReplaceTerminatorTarget(header.terminator, merge, private_merge);
+
+	for (uint32_t i = 0; i <= cloned_blocks.size(); i++) {
+		MoveBlockBefore(graph, first_clone + i, merge + i);
+	}
+	RebuildPredecessors(graph);
+	RecomputeAnalyses(graph);
+	return true;
+}
+
+bool SplitOneSelectionMerge(Graph& graph, uint32_t block_budget) {
 	std::vector<uint32_t> loop_headers;
 	loop_headers.reserve(graph.natural_loops.size());
 	for (const auto& loop: graph.natural_loops) {
@@ -1001,11 +1207,26 @@ bool SplitOneSelectionMerge(Graph& graph) {
 		    Contains(loop_headers, block_id)) {
 			continue;
 		}
+		if (IsInnermostLoopControlConditional(graph, *block)) {
+			continue;
+		}
 
 		const auto merge = graph.FindNearestCommonPostDominator(block->terminator.true_block,
 		                                                        block->terminator.false_block);
+		if (merge == UINT32_MAX || graph.FindBlock(merge) == nullptr) {
+			continue;
+		}
+		const auto region = SelectionRegion(graph, *block, merge);
+		if (std::any_of(region.begin(), region.end(),
+		                [&](uint32_t member) { return !graph.Dominates(block_id, member); })) {
+			if (graph.natural_loops.empty() &&
+			    DuplicateSelectionRegion(graph, block_id, merge, region, block_budget)) {
+				return true;
+			}
+			continue;
+		}
 		const auto construct_blocks = DominatedBlocks(graph, block_id, merge);
-		const auto force_split      = SelectionMergeLeavesContainingLoop(graph, block_id, merge);
+		const auto force_split      = MergeLeavesContainingLoop(graph, block_id, merge);
 		if (SplitSharedMergeBlock(graph, merge, construct_blocks, force_split)) {
 			return true;
 		}
@@ -1015,10 +1236,12 @@ bool SplitOneSelectionMerge(Graph& graph) {
 
 bool SplitSharedMergeBlocks(Graph& graph, std::string* error) {
 	const auto original_block_count = static_cast<uint32_t>(graph.blocks.size());
-	const auto split_budget =
-	    std::max<uint32_t>(16u, std::min<uint32_t>(128u, original_block_count));
+	const auto split_budget = std::max<uint32_t>(
+	    16u, std::min<uint32_t>(128u, original_block_count * 4u));
+	const auto block_budget = std::max<uint32_t>(
+	    32u, std::min<uint32_t>(512u, original_block_count * 8u));
 	for (uint32_t splits = 0; splits < split_budget; splits++) {
-		if (!SplitOneLoopMerge(graph) && !SplitOneSelectionMerge(graph)) {
+		if (!SplitOneLoopMerge(graph) && !SplitOneSelectionMerge(graph, block_budget)) {
 			return true;
 		}
 		RebuildPredecessors(graph);
@@ -1353,6 +1576,9 @@ bool Structurize(Graph& graph, std::string* error) {
 		return false;
 	}
 
+	if (!CanonicalizeNaturalLoops(graph, error)) {
+		return false;
+	}
 	if (!SplitSharedMergeBlocks(graph, error)) {
 		return false;
 	}
@@ -1395,6 +1621,9 @@ bool Structurize(Graph& graph, std::string* error) {
 		    block.terminator.loop_header) {
 			continue;
 		}
+		if (IsInnermostLoopControlConditional(graph, block)) {
+			continue;
+		}
 
 		const auto merge = graph.FindNearestCommonPostDominator(block.terminator.true_block,
 		                                                        block.terminator.false_block);

src/graphics/shader/recompiler/decompiler/ImageOps.cpp

@@ -35,9 +35,9 @@ constexpr ImageDimension DecodeImageDimension(uint32_t dim) {
 		case 2u: return ImageDimension::Dim3D;
 		case 3u: return ImageDimension::Dim2DArray;
 		case 4u: return ImageDimension::Dim1DArray;
-		case 5u:
-		case 7u: return ImageDimension::Dim2DArray;
-		case 6u: return ImageDimension::Dim2D;
+		case 5u: return ImageDimension::Dim2DArray;
+		case 6u: return ImageDimension::Dim2DMsaa;
+		case 7u: return ImageDimension::Dim2DMsaaArray;
 		default: return ImageDimension::Unknown;
 	}
 }
@@ -46,8 +46,10 @@ constexpr uint32_t ImageCoordComponents(ImageDimension dimension) {
 	switch (dimension) {
 		case ImageDimension::Dim1D: return 1u;
 		case ImageDimension::Dim1DArray: return 2u;
+		case ImageDimension::Dim2DMsaa:
 		case ImageDimension::Dim3D:
 		case ImageDimension::Dim2DArray: return 3u;
+		case ImageDimension::Dim2DMsaaArray: return 4u;
 		default: return 2u;
 	}
 }

src/graphics/shader/recompiler/decompiler/ShaderDecoder.cpp

@@ -194,6 +194,8 @@ const char* ImageDimensionToString(ImageDimension dimension) {
 		case ImageDimension::Dim2D: return "2d";
 		case ImageDimension::Dim3D: return "3d";
 		case ImageDimension::Dim2DArray: return "2d_array";
+		case ImageDimension::Dim2DMsaa: return "2d_msaa";
+		case ImageDimension::Dim2DMsaaArray: return "2d_msaa_array";
 		default: return "unknown";
 	}
 }

src/graphics/shader/recompiler/decompiler/ShaderDecoder.h

@@ -575,6 +575,8 @@ enum class ImageDimension : uint32_t {
 	Dim2D,
 	Dim3D,
 	Dim2DArray,
+	Dim2DMsaa,
+	Dim2DMsaaArray,
 };
 
 constexpr uint32_t MaxInstructionRawWords       = 5u;

src/graphics/shader/recompiler/emitter/SpirvEmitter.cpp

@@ -30,10 +30,16 @@ bool ImageBinding(const IR::ImageResource& image, IR::DescriptorBindingKind& kin
 				kind = integer ? Kind::SampledUint1DArray : Kind::Sampled1DArray;
 				return true;
 			case Dim::Dim2D: kind = integer ? Kind::SampledUint2D : Kind::Sampled2D; return true;
+			case Dim::Dim2DMsaa:
+				kind = integer ? Kind::SampledUint2DMsaa : Kind::Sampled2DMsaa;
+				return true;
 			case Dim::Dim3D: kind = integer ? Kind::SampledUint3D : Kind::Sampled3D; return true;
 			case Dim::Dim2DArray:
 				kind = integer ? Kind::SampledUint2DArray : Kind::Sampled2DArray;
 				return true;
+			case Dim::Dim2DMsaaArray:
+				kind = integer ? Kind::SampledUint2DMsaaArray : Kind::Sampled2DMsaaArray;
+				return true;
 			case Dim::Unknown: return false;
 		}
 	}
@@ -51,6 +57,8 @@ bool ImageBinding(const IR::ImageResource& image, IR::DescriptorBindingKind& kin
 		case Dim::Dim2DArray:
 			kind = uint_image ? Kind::StorageUint2DArray : Kind::Storage2DArray;
 			return true;
+		case Dim::Dim2DMsaa:
+		case Dim::Dim2DMsaaArray: return false;
 		case Dim::Unknown: return false;
 	}
 	return false;

src/graphics/shader/recompiler/emitter/spirvEmitterAnalysis.cpp

@@ -276,8 +276,7 @@ void CopyProgramInputsAndOutputs(EmitterState& state, const IR::Program& program
 		if (HasOutput(state.outputs, output.kind, output.index)) {
 			continue;
 		}
-		state.outputs.push_back(
-		    {output.kind, output.index, output.location, 0, output.debug_name});
+		state.outputs.push_back({output.kind, output.index, output.location, 0, output.debug_name});
 	}
 }
 
@@ -576,6 +575,8 @@ ImageViewKind ImageViewKindFromDimension(Decoder::ImageDimension dimension) {
 		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;
 	}
 }
@@ -601,7 +602,9 @@ uint32_t ImageViewCoordinateComponents(ImageViewKind view) {
 		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;
 	}
 }
@@ -611,7 +614,9 @@ uint32_t ImageViewSpatialComponents(ImageViewKind view) {
 		case ImageViewKind::Dim1D:
 		case ImageViewKind::Dim1DArray: return 1u;
 		case ImageViewKind::Dim2D:
-		case ImageViewKind::Dim2DArray: return 2u;
+		case ImageViewKind::Dim2DArray:
+		case ImageViewKind::Dim2DMsaa:
+		case ImageViewKind::Dim2DMsaaArray: return 2u;
 		case ImageViewKind::Dim3D: return 3u;
 		default: return 0u;
 	}
@@ -663,8 +668,7 @@ uint32_t LoadSampledImageDescriptor(EmitterState& state, const IR::MemoryInfo& m
 
 uint32_t LoadSamplerDescriptor(EmitterState& state, uint32_t sampler, uint32_t use_pc) {
 	(void)use_pc;
-	const auto binding =
-	    ResourceForDescriptor(state, IR::DescriptorBindingKind::Samplers, sampler);
+	const auto binding = ResourceForDescriptor(state, IR::DescriptorBindingKind::Samplers, sampler);
 	const auto pointer = DescriptorElementPointer(
 	    state, state.ptr_uniform_sampler, state.sampler_variable, binding.array_index,
 	    IR::DescriptorBindingKind::Samplers, sampler, "sampler descriptor array was not emitted");

src/graphics/shader/recompiler/emitter/spirvEmitterExport.cpp

@@ -36,8 +36,8 @@ uint32_t EmitExportVec4F32(EmitterState& state, const IR::Instruction& inst) {
 			}
 		}
 		const auto vec = state.builder.AllocateId();
-		state.builder.AddFunction({OpCompositeConstruct, state.vec4_float_type, vec,
-		                            components[0], components[1], components[2], components[3]});
+		state.builder.AddFunction({OpCompositeConstruct, state.vec4_float_type, vec, components[0],
+		                           components[1], components[2], components[3]});
 		return vec;
 	}
 
@@ -50,7 +50,59 @@ uint32_t EmitExportVec4F32(EmitterState& state, const IR::Instruction& inst) {
 	return vec;
 }
 
-uint32_t ApplyMrtExportMapping(EmitterState& state, const IR::Instruction& inst, uint32_t value) {
+uint32_t EmitExportComponentU32(EmitterState& state, const IR::Instruction& inst,
+                                uint32_t component) {
+	const bool enabled = ((inst.export_info.en >> component) & 1u) != 0;
+	if (!enabled || component >= inst.src_count || component >= 4u) {
+		return ConstantU32(state, component == 3u ? 1u : 0u);
+	}
+	return EmitValueLoad(state, inst.src[component]);
+}
+
+uint32_t EmitExportVec4U32(EmitterState& state, const IR::Instruction& inst) {
+	uint32_t components[4] = {
+	    ConstantU32(state, 0u),
+	    ConstantU32(state, 0u),
+	    ConstantU32(state, 0u),
+	    ConstantU32(state, 1u),
+	};
+
+	if (inst.export_info.compr) {
+		for (uint32_t pair_index = 0; pair_index < 2u && pair_index < inst.src_count;
+		     pair_index++) {
+			const auto raw = EmitValueLoad(state, inst.src[pair_index]);
+			for (uint32_t lane = 0; lane < 2u; lane++) {
+				const auto component = pair_index * 2u + lane;
+				if (((inst.export_info.en >> component) & 1u) == 0) {
+					continue;
+				}
+				components[component] = state.builder.AllocateId();
+				state.builder.AddFunction(
+				    {OpBitFieldUExtract, state.uint_type, components[component], raw,
+				     ConstantU32(state, lane * 16u), ConstantU32(state, 16u)});
+			}
+		}
+	} else {
+		for (uint32_t component = 0; component < 4u; component++) {
+			components[component] = EmitExportComponentU32(state, inst, component);
+		}
+	}
+
+	const auto vec = state.builder.AllocateId();
+	state.builder.AddFunction({OpCompositeConstruct, state.vec4_uint_type, vec, components[0],
+	                           components[1], components[2], components[3]});
+	return vec;
+}
+
+static bool MrtUsesUintOutput(const EmitterState& state, const IR::Instruction& inst) {
+	return inst.export_info.kind == IR::ExportTargetKind::Mrt &&
+	       state.pixel_input_info != nullptr &&
+	       inst.export_info.index < std::size(state.pixel_input_info->target_output_mode) &&
+	       state.pixel_input_info->target_output_mode[inst.export_info.index] == 7u;
+}
+
+uint32_t ApplyMrtExportMapping(EmitterState& state, const IR::Instruction& inst, uint32_t value,
+                               uint32_t vector_type) {
 	if (inst.export_info.kind != IR::ExportTargetKind::Mrt || state.pixel_input_info == nullptr ||
 	    inst.export_info.index >= state.pixel_input_info->target_export_mapping.size()) {
 		return value;
@@ -62,8 +114,8 @@ uint32_t ApplyMrtExportMapping(EmitterState& state, const IR::Instruction& inst,
 	}
 
 	const auto mapped = state.builder.AllocateId();
-	state.builder.AddFunction({OpVectorShuffle, state.vec4_float_type, mapped, value, value,
-	                            mapping.Map(0), mapping.Map(1), mapping.Map(2), mapping.Map(3)});
+	state.builder.AddFunction({OpVectorShuffle, vector_type, mapped, value, value, mapping.Map(0),
+	                           mapping.Map(1), mapping.Map(2), mapping.Map(3)});
 	return mapped;
 }
 
@@ -114,11 +166,15 @@ void EmitExport(EmitterState& state, const IR::Instruction& inst) {
 		return;
 	}
 
-	const auto value = ApplyMrtExportMapping(state, inst, EmitExportVec4F32(state, inst));
+	const auto uint_output = MrtUsesUintOutput(state, inst);
+	const auto vector_type = uint_output ? state.vec4_uint_type : state.vec4_float_type;
+	const auto value       = ApplyMrtExportMapping(
+	    state, inst, uint_output ? EmitExportVec4U32(state, inst) : EmitExportVec4F32(state, inst),
+	    vector_type);
 	if (inst.export_info.kind == IR::ExportTargetKind::Position) {
 		const auto pointer = state.builder.AllocateId();
-		state.builder.AddFunction({OpAccessChain, state.ptr_output_vec4_float, pointer, variable,
-		                            ConstantU32(state, 0)});
+		state.builder.AddFunction(
+		    {OpAccessChain, state.ptr_output_vec4_float, pointer, variable, ConstantU32(state, 0)});
 		state.builder.AddFunction({OpStore, pointer, value});
 		return;
 	}

src/graphics/shader/recompiler/emitter/spirvEmitterFlowOps.cpp

@@ -150,8 +150,8 @@ void EmitWqmB64(EmitterState& state, const IR::Instruction& inst) {
 		EmitPerInvocationMask(state, inst.dst, active);
 	} else {
 		const auto result = state.builder.AllocateId();
-		state.builder.AddFunction({OpSelect, state.uint_type, result, active,
-		                            ConstantU32(state, 1), ConstantU32(state, 0)});
+		state.builder.AddFunction({OpSelect, state.uint_type, result, active, ConstantU32(state, 1),
+		                           ConstantU32(state, 0)});
 		EmitStoreU32(state, inst.dst, result);
 		EmitStoreU32(state, OffsetRegisterOperand(inst.dst, 1), ConstantU32(state, 0));
 	}
@@ -205,8 +205,7 @@ void EmitSaveexecB32(EmitterState& state, const IR::Instruction& inst) {
 
 	const auto cond = state.builder.AllocateId();
 	const auto scc  = state.builder.AllocateId();
-	state.builder.AddFunction(
-	    {OpINotEqual, state.bool_type, cond, new_low, ConstantU32(state, 0)});
+	state.builder.AddFunction({OpINotEqual, state.bool_type, cond, new_low, ConstantU32(state, 0)});
 	state.builder.AddFunction(
 	    {OpSelect, state.uint_type, scc, cond, ConstantU32(state, 1), ConstantU32(state, 0)});
 	EmitStoreU32(state, SccOperand(), scc);
@@ -279,8 +278,9 @@ void EmitReadFirstLaneU32(EmitterState& state, const IR::Instruction& inst) {
 
 uint32_t EmitLaneIndex(EmitterState& state, const IR::Operand& operand) {
 	const auto lane = state.builder.AllocateId();
+	const auto mask = state.wave_size == 32u ? 31u : 63u;
 	state.builder.AddFunction({OpBitwiseAnd, state.uint_type, lane, EmitValueLoad(state, operand),
-	                            ConstantU32(state, 63)});
+	                           ConstantU32(state, mask)});
 	return lane;
 }
 
@@ -336,10 +336,8 @@ void EmitPermlaneB32(EmitterState& state, const IR::Instruction& inst, bool x16)
 		state.builder.AddFunction(
 		    {OpBitwiseXor, state.uint_type, row_value, row, ConstantU32(state, 16)});
 	}
-	state.builder.AddFunction(
-	    {OpBitwiseAnd, state.uint_type, lane, subid, ConstantU32(state, 15)});
-	state.builder.AddFunction(
-	    {OpBitwiseAnd, state.uint_type, lane8, lane, ConstantU32(state, 7)});
+	state.builder.AddFunction({OpBitwiseAnd, state.uint_type, lane, subid, ConstantU32(state, 15)});
+	state.builder.AddFunction({OpBitwiseAnd, state.uint_type, lane8, lane, ConstantU32(state, 7)});
 	state.builder.AddFunction(
 	    {OpShiftLeftLogical, state.uint_type, shift, lane8, ConstantU32(state, 2)});
 	state.builder.AddFunction(

src/graphics/shader/recompiler/emitter/spirvEmitterImageHelpers.cpp

@@ -2,6 +2,66 @@
 #include "graphics/shader/recompiler/emitter/spirvEmitterInternal.h"
 
 namespace Libs::Graphics::ShaderRecompiler::Spirv::Emitter {
+namespace {
+
+uint32_t EmitCubeAxisF32(EmitterState& state, uint32_t value) {
+	const auto normalized = state.builder.AllocateId();
+	state.builder.AddFunction(
+	    {OpFSub, state.float_type, normalized, value, ConstantF32(state, 0x3f800000u)});
+	return normalized;
+}
+
+uint32_t EmitCubeLayerF32(EmitterState& state, uint32_t face_id) {
+	// Sampled RDNA2 cubemaps encode face_id as slice * 8 + face. The native
+	// 2D-array view stores six contiguous faces per slice, so remove the two
+	// reserved face IDs from every preceding slice.
+	const auto guest_layer = state.builder.AllocateId();
+	const auto slice       = state.builder.AllocateId();
+	const auto padding     = state.builder.AllocateId();
+	const auto host_layer  = state.builder.AllocateId();
+	const auto result      = state.builder.AllocateId();
+	state.builder.AddFunction({OpConvertFToU, state.uint_type, guest_layer, face_id});
+	state.builder.AddFunction(
+	    {OpShiftRightLogical, state.uint_type, slice, guest_layer, ConstantU32(state, 3)});
+	state.builder.AddFunction(
+	    {OpShiftLeftLogical, state.uint_type, padding, slice, ConstantU32(state, 1)});
+	state.builder.AddFunction({OpISub, state.uint_type, host_layer, guest_layer, padding});
+	state.builder.AddFunction({OpConvertUToF, state.float_type, result, host_layer});
+	return result;
+}
+
+uint32_t EmitImageCoordF32Impl(EmitterState& state, const IR::Instruction& inst,
+                               const IR::Operand& address, uint32_t first_component,
+                               uint32_t components) {
+	auto x = EmitImageAddressFloatLoad(state, inst, address, first_component);
+	if (components == 1u) {
+		return x;
+	}
+	auto y = inst.memory.image_address_components > first_component + 1u
+	             ? EmitImageAddressFloatLoad(state, inst, address, first_component + 1u)
+	             : EmitZeroF32(state);
+	if (inst.memory.image_cube) {
+		// RDNA2 sampled cubemap S/T coordinates are biased by +1 relative to
+		// normalized 2D-array coordinates.
+		x = EmitCubeAxisF32(state, x);
+		y = EmitCubeAxisF32(state, y);
+	}
+	const auto coord = state.builder.AllocateId();
+	if (components == 3u) {
+		auto z = inst.memory.image_address_components > first_component + 2u
+		             ? EmitImageAddressFloatLoad(state, inst, address, first_component + 2u)
+		             : EmitZeroF32(state);
+		if (inst.memory.image_cube) {
+			z = EmitCubeLayerF32(state, z);
+		}
+		state.builder.AddFunction({OpCompositeConstruct, state.vec3_float_type, coord, x, y, z});
+	} else {
+		state.builder.AddFunction({OpCompositeConstruct, state.vec2_float_type, coord, x, y});
+	}
+	return coord;
+}
+
+} // namespace
 
 bool HasImageSampleFlag(const IR::Instruction& inst, uint32_t flag) {
 	return (inst.memory.image_sample_flags & flag) != 0;
@@ -36,24 +96,8 @@ ImageSampleLayout MakeImageSampleLayout(const IR::Instruction& inst, ImageViewKi
 
 uint32_t EmitImageCoordF32(EmitterState& state, const IR::Instruction& inst,
                            const ImageSampleLayout& layout, ImageViewKind view) {
-	const auto x          = EmitImageAddressFloatLoad(state, inst, inst.src[0], layout.coord);
-	const auto components = ImageViewCoordinateComponents(view);
-	if (components == 1u) {
-		return x;
-	}
-	const auto y = inst.memory.image_address_components > layout.coord + 1u
-	                       ? EmitImageAddressFloatLoad(state, inst, inst.src[0], layout.coord + 1u)
-	                       : EmitZeroF32(state);
-	const auto coord = state.builder.AllocateId();
-	if (components == 3u) {
-		const auto z = inst.memory.image_address_components > layout.coord + 2u
-		                   ? EmitImageAddressFloatLoad(state, inst, inst.src[0], layout.coord + 2u)
-		                   : EmitZeroF32(state);
-		state.builder.AddFunction({OpCompositeConstruct, state.vec3_float_type, coord, x, y, z});
-	} else {
-		state.builder.AddFunction({OpCompositeConstruct, state.vec2_float_type, coord, x, y});
-	}
-	return coord;
+	return EmitImageCoordF32Impl(state, inst, inst.src[0], layout.coord,
+	                             ImageViewCoordinateComponents(view));
 }
 
 uint32_t EmitImageLodF32(EmitterState& state, const IR::Instruction& inst,
@@ -95,9 +139,9 @@ uint32_t EmitImageGradientF32(EmitterState& state, const IR::Instruction& inst,
 	                   : EmitZeroF32(state);
 	const auto grad = state.builder.AllocateId();
 	if (components == 3u) {
-		const auto z = inst.memory.image_address_components > first_component + 2u
-		                   ? EmitImageAddressFloatLoad(state, inst, inst.src[0],
-		                                               first_component + 2u)
+		const auto z =
+		    inst.memory.image_address_components > first_component + 2u
+		        ? EmitImageAddressFloatLoad(state, inst, inst.src[0], first_component + 2u)
 		        : EmitZeroF32(state);
 		state.builder.AddFunction({OpCompositeConstruct, state.vec3_float_type, grad, x, y, z});
 	} else {
@@ -120,8 +164,7 @@ uint32_t EmitImagePackedOffsetI32(EmitterState& state, const IR::Instruction& in
 			state.builder.AddFunction(
 			    {OpCompositeConstruct, state.vec3_int_type, ret, zero, zero, zero});
 		} else {
-			state.builder.AddFunction(
-			    {OpCompositeConstruct, state.vec2_int_type, ret, zero, zero});
+			state.builder.AddFunction({OpCompositeConstruct, state.vec2_int_type, ret, zero, zero});
 		}
 		return ret;
 	}
@@ -209,24 +252,8 @@ uint32_t EmitImageMipLodU32(EmitterState& state, const IR::Instruction& inst,
 
 uint32_t EmitImageQueryCoordF32(EmitterState& state, const IR::Instruction& inst,
                                 ImageViewKind view) {
-	const auto x          = EmitImageAddressFloatLoad(state, inst, inst.src[0], 0);
-	const auto components = ImageViewCoordinateComponents(view);
-	if (components == 1u) {
-		return x;
-	}
-	const auto y = inst.memory.image_address_components > 1u
-	                       ? EmitImageAddressFloatLoad(state, inst, inst.src[0], 1)
-	                       : EmitZeroF32(state);
-	const auto coord = state.builder.AllocateId();
-	if (components == 3u) {
-		const auto z = inst.memory.image_address_components > 2u
-		                   ? EmitImageAddressFloatLoad(state, inst, inst.src[0], 2)
-		                   : EmitZeroF32(state);
-		state.builder.AddFunction({OpCompositeConstruct, state.vec3_float_type, coord, x, y, z});
-	} else {
-		state.builder.AddFunction({OpCompositeConstruct, state.vec2_float_type, coord, x, y});
-	}
-	return coord;
+	// OpImageQueryLod takes only the spatial coordinates, even for arrayed images.
+	return EmitImageCoordF32Impl(state, inst, inst.src[0], 0, ImageViewSpatialComponents(view));
 }
 
 uint32_t DmaskComponentIndex(uint32_t dmask, uint32_t component) {

src/graphics/shader/recompiler/emitter/spirvEmitterImageOps.cpp

@@ -133,10 +133,18 @@ void EmitImageLoad(EmitterState& state, const IR::Instruction& inst) {
 	const bool integer = inst.memory.kind == IR::ResourceKind::ImageUint;
 
 	const auto color = state.builder.AllocateId();
-	state.builder.AddFunction({OpImageFetch, integer ? state.vec4_uint_type : state.vec4_float_type,
-	                           color, image, EmitImageLoadCoordU32(state, inst, view),
-	                           ImageOperandsLodMask,
-	                           EmitImageMipLodU32(state, inst, inst.src[0], view)});
+	const auto coord = EmitImageLoadCoordU32(state, inst, view);
+	if (ImageSpirvMultisampled(view) != 0) {
+		const auto sample = EmitImageAddressValueLoad(state, inst, inst.src[0],
+		                                              ImageViewCoordinateComponents(view));
+		state.builder.AddFunction({OpImageFetch,
+		                           integer ? state.vec4_uint_type : state.vec4_float_type, color,
+		                           image, coord, ImageOperandsSampleMask, sample});
+	} else {
+		state.builder.AddFunction(
+		    {OpImageFetch, integer ? state.vec4_uint_type : state.vec4_float_type, color, image,
+		     coord, ImageOperandsLodMask, EmitImageMipLodU32(state, inst, inst.src[0], view)});
+	}
 
 	const auto dmask     = inst.memory.dmask != 0 ? inst.memory.dmask : 1u;
 	uint32_t   dst_index = 0;
@@ -158,8 +166,8 @@ void EmitImageLoad(EmitterState& state, const IR::Instruction& inst) {
 void EmitImageStore(EmitterState& state, const IR::Instruction& inst) {
 	const auto uint_image = inst.memory.kind == IR::ResourceKind::StorageImageUint;
 	const auto view       = StorageImageViewKind(state, inst.memory, uint_image, inst.pc);
-	const auto binding    = ResourceForDescriptor(state, StorageBindingKind(uint_image, view),
-	                                              inst.memory.resource);
+	const auto binding =
+	    ResourceForDescriptor(state, StorageBindingKind(uint_image, view), inst.memory.resource);
 	const auto image = LoadStorageImageDescriptorAtIndex(state, inst.memory.resource,
 	                                                     binding.array_index, uint_image, view);
 

src/graphics/shader/recompiler/emitter/spirvEmitterInternal.h

@@ -99,6 +99,7 @@ enum : uint32_t {
 	ImageOperandsGradMask         = 0x00000004u,
 	ImageOperandsOffsetMask       = 0x00000010u,
 	ImageOperandsConstOffsetsMask = 0x00000020u,
+	ImageOperandsSampleMask       = 0x00000040u,
 };
 
 enum : uint32_t {
@@ -150,7 +151,6 @@ enum : uint32_t {
 	OpImageGather                  = 96,
 	OpImageDrefGather              = 97,
 	OpImageWrite                   = 99,
-	OpImage                        = 100,
 	OpImageQuerySizeLod            = 103,
 	OpImageQueryLod                = 105,
 	OpImageQueryLevels             = 106,
@@ -382,7 +382,7 @@ struct EmitterState {
 	uint32_t                                         ptr_workgroup_array       = 0;
 	uint32_t                                         ptr_workgroup_uint        = 0;
 	uint32_t                                         lds_variable              = 0;
-	std::array<SampledImageDescriptors, 10>          sampled_images;
+	std::array<SampledImageDescriptors, 14>          sampled_images;
 	std::array<StorageImageDescriptors, 10>          storage_images;
 	uint32_t                                         sampler_type                          = 0;
 	uint32_t                                         sampler_array_type                    = 0;
@@ -453,17 +453,20 @@ enum class ImageViewKind {
 	Dim2D,
 	Dim2DArray,
 	Dim3D,
+	Dim2DMsaa,
+	Dim2DMsaaArray,
 	Count,
 };
 
-constexpr uint32_t ImageViewKindCount = static_cast<uint32_t>(ImageViewKind::Count);
+constexpr uint32_t SampledImageViewKindCount = static_cast<uint32_t>(ImageViewKind::Count);
+constexpr uint32_t StorageImageViewKindCount = static_cast<uint32_t>(ImageViewKind::Dim2DMsaa);
 
 constexpr uint32_t SampledImageIndex(bool integer, ImageViewKind view) {
-	return static_cast<uint32_t>(view) + (integer ? ImageViewKindCount : 0u);
+	return static_cast<uint32_t>(view) + (integer ? SampledImageViewKindCount : 0u);
 }
 
 constexpr uint32_t StorageImageIndex(bool integer, ImageViewKind view) {
-	return static_cast<uint32_t>(view) + (integer ? ImageViewKindCount : 0u);
+	return static_cast<uint32_t>(view) + (integer ? StorageImageViewKindCount : 0u);
 }
 
 constexpr IR::DescriptorBindingKind SampledBindingKind(bool integer, ImageViewKind view) {
@@ -474,6 +477,9 @@ constexpr IR::DescriptorBindingKind SampledBindingKind(bool integer, ImageViewKi
 			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;
 		}
 	}
@@ -483,6 +489,8 @@ constexpr IR::DescriptorBindingKind SampledBindingKind(bool integer, ImageViewKi
 		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;
@@ -516,6 +524,8 @@ constexpr uint32_t ImageSpirvDimension(ImageViewKind view) {
 		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;
 	}
@@ -523,7 +533,14 @@ constexpr uint32_t ImageSpirvDimension(ImageViewKind view) {
 }
 
 constexpr uint32_t ImageSpirvArrayed(ImageViewKind view) {
-	return view == ImageViewKind::Dim1DArray || view == ImageViewKind::Dim2DArray ? 1u : 0u;
+	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 AddCarryResult {

src/graphics/shader/recompiler/emitter/spirvEmitterModule.cpp

@@ -174,6 +174,12 @@ uint32_t VertexParameterInputPointerType(const EmitterState& state, VertexInputS
 	}
 }
 
+static bool MrtUsesUintOutput(const EmitterState& state, uint32_t index) {
+	return state.stage == ShaderType::Pixel && state.pixel_input_info != nullptr &&
+	       index < std::size(state.pixel_input_info->target_output_mode) &&
+	       state.pixel_input_info->target_output_mode[index] == 7u;
+}
+
 void AllocateInputVariables(EmitterState& state) {
 	for (auto& binding: state.inputs) {
 		binding.variable_id = state.builder.AllocateId();
@@ -323,23 +329,39 @@ void AddDescriptorAnnotationsAndNames(EmitterState& state) {
 		Decorate(state.address_memory_variable, "address_memory",
 		         IR::DescriptorBindingKind::AddressMemory);
 	}
-	constexpr const char* SampledNames[] = {
-	    "sampled_1d",      "sampled_1d_array",      "sampled_2d",      "sampled_2d_array",
-	    "sampled_3d",      "sampled_uint_1d",       "sampled_uint_1d_array",
-	    "sampled_uint_2d", "sampled_uint_2d_array", "sampled_uint_3d"};
+	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_images.size(); i++) {
-		const auto view = static_cast<ImageViewKind>(i % ImageViewKindCount);
+		const auto view = static_cast<ImageViewKind>(i % SampledImageViewKindCount);
 		Decorate(state.sampled_images[i].variable, SampledNames[i],
-		         SampledBindingKind(i >= ImageViewKindCount, view));
+		         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"};
+	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"};
 	for (uint32_t i = 0; i < state.storage_images.size(); i++) {
-		const auto view = static_cast<ImageViewKind>(i % ImageViewKindCount);
+		const auto view = static_cast<ImageViewKind>(i % StorageImageViewKindCount);
 		Decorate(state.storage_images[i].variable, StorageNames[i],
-		         StorageBindingKind(i >= ImageViewKindCount, view));
+		         StorageBindingKind(i >= StorageImageViewKindCount, view));
 	}
 	if (state.sampler_variable != 0) {
 		Decorate(state.sampler_variable, "samplers", IR::DescriptorBindingKind::Samplers);
@@ -409,6 +431,7 @@ void EmitHeaderAndTypes(EmitterState& state) {
 	state.ptr_output_sample_mask_array = state.builder.AllocateId();
 	state.ptr_output_float             = state.builder.AllocateId();
 	state.ptr_output_vec4_float        = state.builder.AllocateId();
+	const auto ptr_output_vec4_uint    = state.builder.AllocateId();
 	state.per_vertex_type              = state.builder.AllocateId();
 	state.ptr_output_per_vertex        = state.builder.AllocateId();
 	state.storage_runtime_array_type   = state.builder.AllocateId();
@@ -462,7 +485,7 @@ void EmitHeaderAndTypes(EmitterState& state) {
 		state.builder.AddCapability({CapabilityImageGatherExtended});
 	}
 	if (std::any_of(state.storage_images.begin(),
-	                state.storage_images.begin() + ImageViewKindCount,
+	                state.storage_images.begin() + StorageImageViewKindCount,
 	                [](const auto& image) { return image.variable != 0; })) {
 		state.builder.AddCapability({CapabilityStorageImageReadWithoutFormat});
 		state.builder.AddCapability({CapabilityStorageImageWriteWithoutFormat});
@@ -605,6 +628,8 @@ void EmitHeaderAndTypes(EmitterState& state) {
 	    {OpTypePointer, state.ptr_output_int, StorageClassOutput, state.int_type});
 	state.builder.AddType(
 	    {OpTypePointer, state.ptr_output_vec4_float, StorageClassOutput, state.vec4_float_type});
+	state.builder.AddType(
+	    {OpTypePointer, ptr_output_vec4_uint, StorageClassOutput, state.vec4_uint_type});
 	if (state.per_vertex_variable != 0) {
 		state.builder.AddType({OpTypeStruct, state.per_vertex_type, state.vec4_float_type});
 		state.builder.AddType({OpTypePointer, state.ptr_output_per_vertex, StorageClassOutput,
@@ -615,8 +640,12 @@ void EmitHeaderAndTypes(EmitterState& state) {
 	for (const auto& binding: state.outputs) {
 		if (binding.kind == IR::StageOutputKind::Parameter ||
 		    binding.kind == IR::StageOutputKind::Mrt) {
+			const auto pointer_type =
+			    binding.kind == IR::StageOutputKind::Mrt && MrtUsesUintOutput(state, binding.index)
+			        ? ptr_output_vec4_uint
+			        : state.ptr_output_vec4_float;
 			state.builder.AddType(
-			    {OpVariable, state.ptr_output_vec4_float, binding.variable_id, StorageClassOutput});
+			    {OpVariable, pointer_type, binding.variable_id, StorageClassOutput});
 		}
 	}
 	if (state.depth_variable != 0) {
@@ -700,11 +729,11 @@ void EmitHeaderAndTypes(EmitterState& state) {
 	}
 	for (uint32_t i = 0; i < state.sampled_images.size(); i++) {
 		auto&      image     = state.sampled_images[i];
-		const auto view      = static_cast<ImageViewKind>(i % ImageViewKindCount);
-		const bool integer   = i >= ImageViewKindCount;
+		const auto view      = static_cast<ImageViewKind>(i % SampledImageViewKindCount);
+		const bool integer   = i >= SampledImageViewKindCount;
 		const auto component = integer ? state.uint_type : state.float_type;
-		state.builder.AddType({OpTypeImage, image.image_type, component,
-		                       ImageSpirvDimension(view), 0, ImageSpirvArrayed(view), 0, 1,
+		state.builder.AddType({OpTypeImage, image.image_type, component, ImageSpirvDimension(view),
+		                       0, ImageSpirvArrayed(view), ImageSpirvMultisampled(view), 1,
 		                       ImageFormatUnknown});
 		state.builder.AddType({OpTypeSampledImage, image.sampled_image_type, image.image_type});
 		state.builder.AddType(
@@ -733,13 +762,12 @@ void EmitHeaderAndTypes(EmitterState& state) {
 	}
 	for (uint32_t i = 0; i < state.storage_images.size(); i++) {
 		auto&      image     = state.storage_images[i];
-		const auto view      = static_cast<ImageViewKind>(i % ImageViewKindCount);
-		const bool integer   = i >= ImageViewKindCount;
+		const auto view      = static_cast<ImageViewKind>(i % StorageImageViewKindCount);
+		const bool integer   = i >= StorageImageViewKindCount;
 		const auto component = integer ? state.uint_type : state.float_type;
 		const auto format    = integer ? ImageFormatR32ui : ImageFormatUnknown;
-		state.builder.AddType({OpTypeImage, image.image_type, component,
-		                       ImageSpirvDimension(view), 0, ImageSpirvArrayed(view), 0, 2,
-		                       format});
+		state.builder.AddType({OpTypeImage, image.image_type, component, ImageSpirvDimension(view),
+		                       0, ImageSpirvArrayed(view), 0, 2, format});
 		state.builder.AddType(
 		    {OpTypePointer, image.pointer_type, StorageClassUniformConstant, image.image_type});
 		if (image.variable != 0) {
@@ -786,15 +814,15 @@ void AllocateDescriptorVariables(EmitterState& state) {
 		state.flattened_srt_variable = state.builder.AllocateId();
 	}
 	for (uint32_t i = 0; i < state.sampled_images.size(); i++) {
-		const auto view = static_cast<ImageViewKind>(i % ImageViewKindCount);
-		if (DescriptorBinding(state, SampledBindingKind(i >= ImageViewKindCount, view)) !=
+		const auto view = static_cast<ImageViewKind>(i % SampledImageViewKindCount);
+		if (DescriptorBinding(state, SampledBindingKind(i >= SampledImageViewKindCount, view)) !=
 		    nullptr) {
 			state.sampled_images[i].variable = state.builder.AllocateId();
 		}
 	}
 	for (uint32_t i = 0; i < state.storage_images.size(); i++) {
-		const auto view = static_cast<ImageViewKind>(i % ImageViewKindCount);
-		if (DescriptorBinding(state, StorageBindingKind(i >= ImageViewKindCount, view)) !=
+		const auto view = static_cast<ImageViewKind>(i % StorageImageViewKindCount);
+		if (DescriptorBinding(state, StorageBindingKind(i >= StorageImageViewKindCount, view)) !=
 		    nullptr) {
 			state.storage_images[i].variable = state.builder.AllocateId();
 		}

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

@@ -13,16 +13,30 @@ namespace {
 constexpr uint32_t MaxPushConstantBytes = 128;
 
 constexpr std::array ImageBindingKinds = {
-    DescriptorBindingKind::Sampled1D,          DescriptorBindingKind::Sampled1DArray,
-    DescriptorBindingKind::Sampled2D,          DescriptorBindingKind::Sampled2DArray,
-    DescriptorBindingKind::Sampled3D,          DescriptorBindingKind::SampledUint1D,
-    DescriptorBindingKind::SampledUint1DArray, DescriptorBindingKind::SampledUint2D,
-    DescriptorBindingKind::SampledUint2DArray, DescriptorBindingKind::SampledUint3D,
-    DescriptorBindingKind::Storage1D,          DescriptorBindingKind::Storage1DArray,
-    DescriptorBindingKind::Storage2D,          DescriptorBindingKind::Storage2DArray,
-    DescriptorBindingKind::Storage3D,          DescriptorBindingKind::StorageUint1D,
-    DescriptorBindingKind::StorageUint1DArray, DescriptorBindingKind::StorageUint2D,
-    DescriptorBindingKind::StorageUint2DArray, DescriptorBindingKind::StorageUint3D,
+    DescriptorBindingKind::Sampled1D,
+    DescriptorBindingKind::Sampled1DArray,
+    DescriptorBindingKind::Sampled2D,
+    DescriptorBindingKind::Sampled2DArray,
+    DescriptorBindingKind::Sampled2DMsaa,
+    DescriptorBindingKind::Sampled2DMsaaArray,
+    DescriptorBindingKind::Sampled3D,
+    DescriptorBindingKind::SampledUint1D,
+    DescriptorBindingKind::SampledUint1DArray,
+    DescriptorBindingKind::SampledUint2D,
+    DescriptorBindingKind::SampledUint2DArray,
+    DescriptorBindingKind::SampledUint2DMsaa,
+    DescriptorBindingKind::SampledUint2DMsaaArray,
+    DescriptorBindingKind::SampledUint3D,
+    DescriptorBindingKind::Storage1D,
+    DescriptorBindingKind::Storage1DArray,
+    DescriptorBindingKind::Storage2D,
+    DescriptorBindingKind::Storage2DArray,
+    DescriptorBindingKind::Storage3D,
+    DescriptorBindingKind::StorageUint1D,
+    DescriptorBindingKind::StorageUint1DArray,
+    DescriptorBindingKind::StorageUint2D,
+    DescriptorBindingKind::StorageUint2DArray,
+    DescriptorBindingKind::StorageUint3D,
 };
 
 bool ImageBinding(const ImageResource& image, DescriptorBindingKind& result) {
@@ -36,6 +50,8 @@ bool ImageBinding(const ImageResource& image, DescriptorBindingKind& result) {
 				case Dimension::Dim1DArray: result = Kind::Sampled1DArray; return true;
 				case Dimension::Dim2D: result = Kind::Sampled2D; return true;
 				case Dimension::Dim2DArray: result = Kind::Sampled2DArray; return true;
+				case Dimension::Dim2DMsaa: result = Kind::Sampled2DMsaa; return true;
+				case Dimension::Dim2DMsaaArray: result = Kind::Sampled2DMsaaArray; return true;
 				case Dimension::Dim3D: result = Kind::Sampled3D; return true;
 				default: return false;
 			}
@@ -45,6 +61,8 @@ bool ImageBinding(const ImageResource& image, DescriptorBindingKind& result) {
 				case Dimension::Dim1DArray: result = Kind::SampledUint1DArray; return true;
 				case Dimension::Dim2D: result = Kind::SampledUint2D; return true;
 				case Dimension::Dim2DArray: result = Kind::SampledUint2DArray; return true;
+				case Dimension::Dim2DMsaa: result = Kind::SampledUint2DMsaa; return true;
+				case Dimension::Dim2DMsaaArray: result = Kind::SampledUint2DMsaaArray; return true;
 				case Dimension::Dim3D: result = Kind::SampledUint3D; return true;
 				default: return false;
 			}
@@ -165,8 +183,7 @@ bool CollectUserData(const Program& program, std::vector<uint32_t>& result) {
 				return false;
 			}
 			for (uint32_t i = 0; i < inst.src_count; i++) {
-				if (!CollectValue(program.provenance, inst.scalar_sources[i], visited,
-				                  registers)) {
+				if (!CollectValue(program.provenance, inst.scalar_sources[i], visited, registers)) {
 					return false;
 				}
 			}

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

@@ -0,0 +1,323 @@
+#include "graphics/shader/recompiler/ir/ReadLaneElimination.h"
+
+#include "graphics/shader/recompiler/ir/SrtWalker.h"
+
+#include <algorithm>
+#include <iterator>
+#include <map>
+#include <set>
+#include <utility>
+
+namespace Libs::Graphics::ShaderRecompiler::IR {
+
+namespace {
+
+constexpr uint32_t FirstTemporaryScalarRegister = 128;
+
+struct LaneKey {
+	uint32_t reg  = 0;
+	uint32_t lane = 0;
+
+	auto operator<=>(const LaneKey&) const = default;
+};
+
+using LaneSet = std::set<LaneKey>;
+
+bool PairDwordOpcode(Opcode op) {
+	switch (op) {
+		case Opcode::MoveU64:
+		case Opcode::WqmB64:
+		case Opcode::SaveexecB64:
+		case Opcode::BitwiseAndU64:
+		case Opcode::BitwiseAndNotU64:
+		case Opcode::BitwiseOrU64:
+		case Opcode::BitwiseOrNotU64:
+		case Opcode::BitwiseXorU64:
+		case Opcode::BitwiseNandU64:
+		case Opcode::BitwiseNorU64:
+		case Opcode::BitwiseXnorU64:
+		case Opcode::BitwiseNotU64:
+		case Opcode::BitFieldMaskU64:
+		case Opcode::BitFieldExtractU64:
+		case Opcode::BitReplicateB64B32:
+		case Opcode::ShiftLeftLogicalU64:
+		case Opcode::ShiftRightLogicalU64:
+		case Opcode::SelectU64: return true;
+		default: return false;
+	}
+}
+
+bool ResolveLane(const Program& program, const Instruction& inst, uint32_t source_index,
+                 uint32_t& lane) {
+	if (source_index >= inst.src_count || (program.wave_size != 32 && program.wave_size != 64)) {
+		return false;
+	}
+	const auto& selector = inst.src[source_index];
+	if (selector.kind == OperandKind::ImmediateU32) {
+		lane = selector.imm % program.wave_size;
+		return true;
+	}
+	uint32_t folded = 0;
+	if (!FoldScalarConstant(program.provenance, inst.scalar_sources[source_index], folded)) {
+		return false;
+	}
+	lane = folded % program.wave_size;
+	return true;
+}
+
+bool UniformWriteSource(const Instruction& inst) {
+	if (inst.src_count == 0) {
+		return false;
+	}
+	const auto& source = inst.src[0];
+	if (source.kind == OperandKind::ImmediateU32 || source.kind == OperandKind::PcRelativeU32) {
+		return true;
+	}
+	return source.kind == OperandKind::Register &&
+	       (source.reg.file == RegisterFile::Scalar || source.reg.file == RegisterFile::Scc ||
+	        source.reg.file == RegisterFile::M0);
+}
+
+bool WriteLaneKey(const Program& program, const Instruction& inst, LaneKey& key) {
+	if (inst.op != Opcode::WriteLaneU32 || inst.dst.kind != OperandKind::Register ||
+	    inst.dst.reg.file != RegisterFile::Vector || !UniformWriteSource(inst)) {
+		return false;
+	}
+	uint32_t lane = 0;
+	if (!ResolveLane(program, inst, 1, lane)) {
+		return false;
+	}
+	key = {inst.dst.reg.index, lane};
+	return true;
+}
+
+bool ReadLaneKey(const Program& program, const Instruction& inst, LaneKey& key) {
+	if (inst.op != Opcode::ReadLaneU32 || inst.src_count < 2 ||
+	    inst.src[0].kind != OperandKind::Register || inst.src[0].reg.file != RegisterFile::Vector) {
+		return false;
+	}
+	uint32_t lane = 0;
+	if (!ResolveLane(program, inst, 1, lane)) {
+		return false;
+	}
+	key = {inst.src[0].reg.index, lane};
+	return true;
+}
+
+void InvalidateRegister(LaneSet& valid, uint32_t reg) {
+	const auto first = valid.lower_bound({reg, 0});
+	const auto last  = valid.lower_bound({reg + 1u, 0});
+	valid.erase(first, last);
+}
+
+void ApplyInstruction(const Program& program, const Instruction& inst, LaneSet& valid) {
+	if (inst.op == Opcode::WriteLaneU32 && inst.dst.kind == OperandKind::Register &&
+	    inst.dst.reg.file == RegisterFile::Vector) {
+		LaneKey key;
+		if (WriteLaneKey(program, inst, key)) {
+			valid.insert(key);
+			return;
+		}
+		uint32_t lane = 0;
+		if (ResolveLane(program, inst, 1, lane)) {
+			valid.erase({inst.dst.reg.index, lane});
+		} else {
+			InvalidateRegister(valid, inst.dst.reg.index);
+		}
+		return;
+	}
+
+	if (inst.op == Opcode::MoveRelDestU32 && inst.dst.kind == OperandKind::Register &&
+	    inst.dst.reg.file == RegisterFile::Vector) {
+		valid.clear();
+		return;
+	}
+
+	if (inst.dst.kind == OperandKind::Register && inst.dst.reg.file == RegisterFile::Vector) {
+		uint32_t dwords = std::max(inst.memory.data_dwords, 1u);
+		if (PairDwordOpcode(inst.op) || inst.op == Opcode::UMadU64U32) {
+			dwords = std::max(dwords, 2u);
+		}
+		for (uint32_t i = 0; i < dwords && inst.dst.reg.index <= UINT32_MAX - i; i++) {
+			InvalidateRegister(valid, inst.dst.reg.index + i);
+		}
+	}
+	if (inst.dst2.kind == OperandKind::Register && inst.dst2.reg.file == RegisterFile::Vector) {
+		InvalidateRegister(valid, inst.dst2.reg.index);
+	}
+}
+
+LaneSet TransferBlock(const Program& program, const BasicBlock& block, LaneSet state) {
+	for (const auto& inst: block.instructions) {
+		ApplyInstruction(program, inst, state);
+	}
+	return state;
+}
+
+LaneSet Intersect(const LaneSet& left, const LaneSet& right) {
+	LaneSet result;
+	std::set_intersection(left.begin(), left.end(), right.begin(), right.end(),
+	                      std::inserter(result, result.end()));
+	return result;
+}
+
+uint32_t NextTemporaryScalarRegister(const Program& program) {
+	uint32_t   next     = FirstTemporaryScalarRegister;
+	const auto consider = [&next](const Operand& operand) {
+		if (operand.kind == OperandKind::Register && operand.reg.file == RegisterFile::Scalar &&
+		    operand.reg.index >= next && operand.reg.index != UINT32_MAX) {
+			next = operand.reg.index + 1u;
+		}
+	};
+	for (const auto& block: program.blocks) {
+		for (const auto& inst: block.instructions) {
+			consider(inst.dst);
+			consider(inst.dst2);
+			for (uint32_t i = 0; i < inst.src_count; i++) {
+				consider(inst.src[i]);
+			}
+		}
+	}
+	return next;
+}
+
+Operand ScalarRegisterOperand(uint32_t reg) {
+	Operand operand;
+	operand.kind      = OperandKind::Register;
+	operand.reg.file  = RegisterFile::Scalar;
+	operand.reg.index = reg;
+	return operand;
+}
+
+Instruction ShadowWrite(const Instruction& write, uint32_t temporary) {
+	Instruction shadow;
+	shadow.pc        = write.pc;
+	shadow.op        = Opcode::MoveU32;
+	shadow.dst       = ScalarRegisterOperand(temporary);
+	shadow.src[0]    = write.src[0];
+	shadow.src_count = 1;
+	return shadow;
+}
+
+Instruction ShadowRead(const Instruction& read, uint32_t temporary) {
+	Instruction rewritten;
+	rewritten.pc        = read.pc;
+	rewritten.op        = Opcode::MoveU32;
+	rewritten.dst       = read.dst;
+	rewritten.src[0]    = ScalarRegisterOperand(temporary);
+	rewritten.src_count = 1;
+	return rewritten;
+}
+
+} // namespace
+
+ReadLaneEliminationStats EliminateReadLane(Program& program) {
+	ReadLaneEliminationStats stats;
+	if (program.blocks.empty() || (program.wave_size != 32 && program.wave_size != 64)) {
+		return stats;
+	}
+
+	LaneSet universe;
+	for (const auto& block: program.blocks) {
+		for (const auto& inst: block.instructions) {
+			LaneKey key;
+			if (WriteLaneKey(program, inst, key)) {
+				universe.insert(key);
+			}
+		}
+	}
+	if (universe.empty()) {
+		return stats;
+	}
+
+	const size_t         block_count = program.blocks.size();
+	std::vector<LaneSet> entry(block_count, universe);
+	std::vector<LaneSet> exit(block_count, universe);
+	entry[0].clear();
+	for (size_t block = 0; block < block_count; block++) {
+		exit[block] = TransferBlock(program, program.blocks[block], entry[block]);
+	}
+
+	bool changed = true;
+	while (changed) {
+		changed = false;
+		for (size_t block_index = 0; block_index < block_count; block_index++) {
+			LaneSet     next_entry;
+			const auto& block = program.blocks[block_index];
+			if (block_index != 0 && !block.predecessors.empty()) {
+				next_entry = universe;
+				for (const auto predecessor: block.predecessors) {
+					if (predecessor >= block_count) {
+						next_entry.clear();
+						break;
+					}
+					next_entry = Intersect(next_entry, exit[predecessor]);
+				}
+			}
+			auto next_exit = TransferBlock(program, block, next_entry);
+			if (next_entry != entry[block_index] || next_exit != exit[block_index]) {
+				entry[block_index] = std::move(next_entry);
+				exit[block_index]  = std::move(next_exit);
+				changed            = true;
+			}
+		}
+	}
+
+	LaneSet forwarded;
+	for (size_t block_index = 0; block_index < block_count; block_index++) {
+		auto state = entry[block_index];
+		for (const auto& inst: program.blocks[block_index].instructions) {
+			LaneKey key;
+			if (ReadLaneKey(program, inst, key) && state.contains(key)) {
+				forwarded.insert(key);
+			}
+			ApplyInstruction(program, inst, state);
+		}
+	}
+	if (forwarded.empty()) {
+		return stats;
+	}
+
+	std::map<LaneKey, uint32_t> temporaries;
+	auto                        next_temporary = NextTemporaryScalarRegister(program);
+	for (const auto& key: forwarded) {
+		if (next_temporary == UINT32_MAX) {
+			return {};
+		}
+		temporaries.emplace(key, next_temporary++);
+	}
+
+	for (size_t block_index = 0; block_index < block_count; block_index++) {
+		const auto original  = std::move(program.blocks[block_index].instructions);
+		auto&      rewritten = program.blocks[block_index].instructions;
+		rewritten.clear();
+		rewritten.reserve(original.size() + temporaries.size());
+		auto state = entry[block_index];
+		for (const auto& inst: original) {
+			LaneKey read_key;
+			if (ReadLaneKey(program, inst, read_key) && state.contains(read_key)) {
+				const auto temporary = temporaries.find(read_key);
+				if (temporary != temporaries.end()) {
+					rewritten.push_back(ShadowRead(inst, temporary->second));
+					stats.rewritten_reads++;
+					ApplyInstruction(program, inst, state);
+					continue;
+				}
+			}
+
+			rewritten.push_back(inst);
+			LaneKey write_key;
+			if (WriteLaneKey(program, inst, write_key)) {
+				const auto temporary = temporaries.find(write_key);
+				if (temporary != temporaries.end()) {
+					rewritten.push_back(ShadowWrite(inst, temporary->second));
+					stats.shadow_writes++;
+				}
+			}
+			ApplyInstruction(program, inst, state);
+		}
+	}
+	return stats;
+}
+
+} // namespace Libs::Graphics::ShaderRecompiler::IR

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

@@ -0,0 +1,20 @@
+#ifndef EMULATOR_INCLUDE_EMULATOR_GRAPHICS_SHADER_RECOMPILER_READLANEELIMINATION_H_
+#define EMULATOR_INCLUDE_EMULATOR_GRAPHICS_SHADER_RECOMPILER_READLANEELIMINATION_H_
+
+#include "graphics/shader/recompiler/ir/ShaderIR.h"
+
+namespace Libs::Graphics::ShaderRecompiler::IR {
+
+struct ReadLaneEliminationStats {
+	uint32_t rewritten_reads = 0;
+	uint32_t shadow_writes   = 0;
+};
+
+// Replaces fixed-lane ReadLane operations that are reached by a matching WriteLane on every
+// control-flow path. A synthetic scalar register snapshots the value at WriteLane execution time,
+// so the rewrite remains valid when the source SGPR is subsequently overwritten.
+[[nodiscard]] ReadLaneEliminationStats EliminateReadLane(Program& program);
+
+} // namespace Libs::Graphics::ShaderRecompiler::IR
+
+#endif /* EMULATOR_INCLUDE_EMULATOR_GRAPHICS_SHADER_RECOMPILER_READLANEELIMINATION_H_ */

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

@@ -15,7 +15,8 @@ constexpr uint64_t AddressMask = 0x0000ffffffffffffull;
 Decoder::ImageDimension DescriptorDimension(const DescriptorValue&  descriptor,
                                             Decoder::ImageDimension requested) {
 	const bool is_array = requested == Decoder::ImageDimension::Dim1DArray ||
-	                      requested == Decoder::ImageDimension::Dim2DArray;
+	                      requested == Decoder::ImageDimension::Dim2DArray ||
+	                      requested == Decoder::ImageDimension::Dim2DMsaaArray;
 	switch (static_cast<Prospero::ImageType>((descriptor.dwords[3] >> 28u) & 0xfu)) {
 		case Prospero::ImageType::kColor1D: return Decoder::ImageDimension::Dim1D;
 		case Prospero::ImageType::kColor1DArray:
@@ -26,13 +27,17 @@ Decoder::ImageDimension DescriptorDimension(const DescriptorValue&       descrip
 		case Prospero::ImageType::kColor3D: return Decoder::ImageDimension::Dim3D;
 		case Prospero::ImageType::kCube: return Decoder::ImageDimension::Dim2DArray;
 		case Prospero::ImageType::kColor2DArray:
-		case Prospero::ImageType::kColor2DMsaaArray:
 			if (is_array) {
 				return Decoder::ImageDimension::Dim2DArray;
 			}
 			return Decoder::ImageDimension::Dim2D;
-		case Prospero::ImageType::kColor2D:
-		case Prospero::ImageType::kColor2DMsaa: return Decoder::ImageDimension::Dim2D;
+		case Prospero::ImageType::kColor2DMsaaArray:
+			if (is_array) {
+				return Decoder::ImageDimension::Dim2DMsaaArray;
+			}
+			return Decoder::ImageDimension::Dim2DMsaa;
+		case Prospero::ImageType::kColor2D: return Decoder::ImageDimension::Dim2D;
+		case Prospero::ImageType::kColor2DMsaa: return Decoder::ImageDimension::Dim2DMsaa;
 		default: return Decoder::ImageDimension::Unknown;
 	}
 }
@@ -43,7 +48,8 @@ bool NullImageDescriptor(const DescriptorValue& descriptor) {
 
 bool ValidImageDescriptor(const DescriptorValue& descriptor) {
 	const auto type   = static_cast<Prospero::ImageType>((descriptor.dwords[3] >> 28u) & 0xfu);
-	if (type < Prospero::ImageType::kColor1D) {
+	const auto format = static_cast<Prospero::BufferFormat>((descriptor.dwords[1] >> 20u) & 0x1ffu);
+	if (type < Prospero::ImageType::kColor1D || format == Prospero::BufferFormat::kInvalid) {
 		return false;
 	}
 	if (type == Prospero::ImageType::kColor2DMsaa ||
@@ -51,8 +57,7 @@ bool ValidImageDescriptor(const DescriptorValue& descriptor) {
 		const auto base_level = (descriptor.dwords[3] >> 12u) & 0xfu;
 		const auto fragments  = (descriptor.dwords[3] >> 16u) & 0xfu;
 		const auto max_mip    = (descriptor.dwords[5] >> 4u) & 0xfu;
-		return base_level == 0 && fragments >= 1 && fragments <= 3 &&
-		       max_mip == fragments;
+		return base_level == 0 && fragments >= 1 && fragments <= 3 && max_mip == fragments;
 	}
 	return true;
 }
@@ -61,6 +66,11 @@ uint32_t DescriptorImageSwizzle(const DescriptorValue& descriptor) {
 	return descriptor.dwords[3] & 0xfffu;
 }
 
+bool DescriptorIsCube(const DescriptorValue& descriptor) {
+	return static_cast<Prospero::ImageType>((descriptor.dwords[3] >> 28u) & 0xfu) ==
+	       Prospero::ImageType::kCube;
+}
+
 bool DecodeBufferDescriptor(const DescriptorValue& descriptor, ShaderBufferResource& result) {
 	if (descriptor.dword_count != std::size(result.fields)) {
 		return false;
@@ -171,12 +181,13 @@ bool ValidateResourceSpecialization(const Program& program, const ResourceSnapsh
 		const auto& image      = program.info.images[i];
 		const auto& descriptor = snapshot.images[i];
 		if (NullImageDescriptor(descriptor)) {
-			bool canonical_kind = image.kind == ResourceKind::Image ||
-			                      image.kind == ResourceKind::StorageImage;
+			bool canonical_kind =
+			    image.kind == ResourceKind::Image || image.kind == ResourceKind::StorageImage;
 			if (image.atomic) {
 				canonical_kind = image.kind == ResourceKind::StorageImageUint;
 			}
-			if (image.dimension != Decoder::ImageDimension::Dim2D || !canonical_kind) {
+			if (image.dimension != Decoder::ImageDimension::Dim2D || image.cube ||
+			    !canonical_kind) {
 				if (error != nullptr) {
 					*error = fmt::format(
 					    "image descriptor {} no longer matches canonical null specialization", i);
@@ -186,10 +197,15 @@ bool ValidateResourceSpecialization(const Program& program, const ResourceSnapsh
 			continue;
 		}
 		const auto dimension = DescriptorDimension(descriptor, image.dimension);
-		if (dimension == Decoder::ImageDimension::Unknown || dimension != image.dimension) {
+		if (dimension == Decoder::ImageDimension::Unknown || dimension != image.dimension ||
+		    DescriptorIsCube(descriptor) != image.cube) {
 			if (error != nullptr) {
-				*error =
-				    fmt::format("image descriptor {} no longer matches specialized dimension", i);
+				*error = fmt::format(
+				    "image descriptor {} no longer matches specialized dimension: "
+				    "{:08x},{:08x},{:08x},{:08x},{:08x},{:08x},{:08x},{:08x}",
+				    i, descriptor.dwords[0], descriptor.dwords[1], descriptor.dwords[2],
+				    descriptor.dwords[3], descriptor.dwords[4], descriptor.dwords[5],
+				    descriptor.dwords[6], descriptor.dwords[7]);
 			}
 			return false;
 		}
@@ -204,8 +220,12 @@ bool ValidateResourceSpecialization(const Program& program, const ResourceSnapsh
 				}
 				return false;
 			}
-			const auto uint_descriptor =
-			    Prospero::IsUintTextureFormat((descriptor.dwords[1] >> 20u) & 0x1ffu);
+			const auto format = (descriptor.dwords[1] >> 20u) & 0x1ffu;
+			const bool raw_sint_storage =
+			    storage && format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32SInt) &&
+			    !image.read && !image.atomic;
+			const bool uint_descriptor =
+			    Prospero::IsUintTextureFormat(format) || raw_sint_storage;
 			const auto uint_program = image.kind == ResourceKind::ImageUint ||
 			                          image.kind == ResourceKind::StorageImageUint;
 			if (uint_descriptor != uint_program && !(image.atomic && uint_program)) {
@@ -361,6 +381,7 @@ bool SpecializeResources(Program& program, const ResourceSnapshot& snapshot, std
 		auto&       image      = next.images[i];
 		if (NullImageDescriptor(descriptor)) {
 			image.dimension = Decoder::ImageDimension::Dim2D;
+			image.cube      = false;
 			switch (image.kind) {
 				case ResourceKind::ImageUint: image.kind = ResourceKind::Image; break;
 				case ResourceKind::StorageImageUint:
@@ -386,11 +407,19 @@ bool SpecializeResources(Program& program, const ResourceSnapshot& snapshot, std
 			return false;
 		}
 		image.dimension = descriptor_dimension;
+		image.cube      = DescriptorIsCube(descriptor);
 		if (image.kind == ResourceKind::StorageImage ||
 		    image.kind == ResourceKind::StorageImageUint) {
 			image.storage_swizzle = DescriptorImageSwizzle(descriptor);
 		}
-		if (Prospero::IsUintTextureFormat((descriptor.dwords[1] >> 20u) & 0x1ffu)) {
+		const auto format  = (descriptor.dwords[1] >> 20u) & 0x1ffu;
+		const bool storage = image.kind == ResourceKind::StorageImage ||
+		                     image.kind == ResourceKind::StorageImageUint;
+		const bool raw_sint_storage =
+		    storage && format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32SInt) &&
+		    !image.read && !image.atomic;
+		const bool uint_image = Prospero::IsUintTextureFormat(format) || raw_sint_storage;
+		if (uint_image) {
 			switch (image.kind) {
 				case ResourceKind::Image: image.kind = ResourceKind::ImageUint; break;
 				case ResourceKind::StorageImage: image.kind = ResourceKind::StorageImageUint; break;
@@ -402,6 +431,7 @@ bool SpecializeResources(Program& program, const ResourceSnapshot& snapshot, std
 		std::reference_wrapper<Instruction> inst;
 		ResourceKind                        kind;
 		Decoder::ImageDimension             dimension;
+		bool                                cube;
 	};
 	std::vector<ImagePatch> patches;
 	for (auto& block: program.blocks) {
@@ -420,13 +450,14 @@ bool SpecializeResources(Program& program, const ResourceSnapshot& snapshot, std
 				return false;
 			}
 			const auto& image = next.images[inst.memory.resource];
-			patches.push_back({std::ref(inst), image.kind, image.dimension});
+			patches.push_back({std::ref(inst), image.kind, image.dimension, image.cube});
 		}
 	}
 	program.info = std::move(next);
 	for (const auto& patch: patches) {
 		patch.inst.get().memory.kind            = patch.kind;
 		patch.inst.get().memory.image_dimension = patch.dimension;
+		patch.inst.get().memory.image_cube      = patch.cube;
 	}
 	return true;
 }

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

@@ -45,6 +45,8 @@ namespace {
 
 constexpr uint32_t ScalarRegisters = 128;
 constexpr uint32_t VectorRegisters = 256;
+// Clamp X/Y/Z are consecutive three-bit fields; the high bit of each selects a border mode.
+constexpr uint32_t SamplerBorderClampMask = (1u << 2u) | (1u << 5u) | (1u << 8u);
 
 struct ScalarState {
 	std::array<uint32_t, ScalarRegisters> regs          = {};
@@ -647,6 +649,24 @@ private:
 		return AddDescriptor(descriptor);
 	}
 
+	uint32_t AddSamplerDescriptor(const ScalarState& state, uint32_t base) {
+		if (base >= ScalarRegisters || 4u > ScalarRegisters - base) {
+			return ScalarProvenance::Unknown;
+		}
+		DescriptorValue descriptor;
+		descriptor.dword_count = 4;
+		for (uint32_t i = 0; i < 4; i++) {
+			descriptor.dwords[i] = state.regs[base + i];
+		}
+		if (auto d0 = descriptor.dwords[0];
+		    d0 < m_graph.values.size() && m_graph.values[d0].op == ScalarValueOp::Constant &&
+		    (m_graph.values[d0].imm & SamplerBorderClampMask) == 0) {
+			// Without a border clamp, the border color and table index in dword 3 are unused.
+			descriptor.dwords[3] = Constant(0);
+		}
+		return AddDescriptor(descriptor);
+	}
+
 	uint32_t AddFlatAddressDescriptor(const Instruction& inst, const ScalarState& state) {
 		const uint32_t first = FlatStore(inst.op) ? 1u : 0u;
 		if (inst.src_count < first + 2u) {
@@ -704,7 +724,7 @@ private:
 			inst.memory.resource_source = AddDescriptor(state, inst.memory.resource * 4u, 8);
 			if (inst.op == Opcode::ImageSample || inst.op == Opcode::ImageGather4 ||
 			    inst.op == Opcode::ImageGetLod) {
-				inst.memory.sampler_source = AddDescriptor(state, inst.memory.sampler * 4u, 4);
+				inst.memory.sampler_source = AddSamplerDescriptor(state, inst.memory.sampler * 4u);
 			}
 		}
 	}

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

@@ -432,6 +432,7 @@ struct MemoryInfo {
 	bool     typed           = false;
 	bool     formatted       = false;
 	bool     image_has_mip   = false;
+	bool     image_cube      = false;
 	bool     glc             = false;
 	bool     slc             = false;
 	bool     idxen           = false;
@@ -607,6 +608,7 @@ struct ImageResource {
 	bool                    written         = false;
 	bool                    atomic          = false;
 	bool                    depth_compare   = false;
+	bool                    cube            = false;
 
 	bool operator==(const ImageResource& other) const = default;
 };
@@ -677,11 +679,15 @@ enum class DescriptorBindingKind {
 	Sampled1DArray,
 	Sampled2D,
 	Sampled2DArray,
+	Sampled2DMsaa,
+	Sampled2DMsaaArray,
 	Sampled3D,
 	SampledUint1D,
 	SampledUint1DArray,
 	SampledUint2D,
 	SampledUint2DArray,
+	SampledUint2DMsaa,
+	SampledUint2DMsaaArray,
 	SampledUint3D,
 	Storage1D,
 	Storage1DArray,

src/graphics/shader/shader.cpp

@@ -13,8 +13,8 @@
 #include "graphics/guest_gpu/graphicsRun.h"
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
-#include "graphics/shader/recompiler/decompiler/ShaderDecoder.h"
 #include "graphics/shader/recompiler/ShaderRecompiler.h"
+#include "graphics/shader/recompiler/decompiler/ShaderDecoder.h"
 #include "graphics/shader/shaderVertexMetadata.h"
 #include "libs/errno.h"
 #include "spirv-tools/libspirv.h"
@@ -828,8 +828,7 @@ static void ShaderGetStaticInputInfoPS(
 	    vs_info.stage.program != nullptr && !vs_info.stage.program->bindings.descriptors.empty()
 	        ? 1
 	        : 0;
-	ps_info.push_constant_offset =
-	    vs_info.stage.program != nullptr
+	ps_info.push_constant_offset = vs_info.stage.program != nullptr
 	                                   ? vs_info.stage.program->bindings.push_constant_offset +
 	                                         vs_info.stage.program->bindings.push_constant_size
 	                                   : 0;
@@ -1294,8 +1293,7 @@ static void DumpShaderRecompilerSpirv(const char* type, uint64_t shader_hash,
 
 	static std::atomic_int id = 0;
 
-	const auto base_name =
-	    Config::GetShaderLogFolder() /
+	const auto base_name = Config::GetShaderLogFolder() /
 	                       fmt::format("{:04d}_new_shader_{}_{:016x}", id++, type, shader_hash);
 	Common::File::CreateDirectories(base_name.parent_path());
 
@@ -1345,8 +1343,7 @@ static void DumpShaderRecompilerOriginal(const char* type, uint64_t shader_hash,
 
 	static std::atomic_int id = 0;
 
-	const auto base_name =
-	    Config::GetShaderLogFolder() / "original" /
+	const auto base_name = Config::GetShaderLogFolder() / "original" /
 	                       fmt::format("{:04d}_new_shader_{}_{:016x}", id++, type, shader_hash);
 	Common::File::CreateDirectories(base_name.parent_path());
 

src/kernel/eventQueue.cpp

@@ -459,8 +459,7 @@ int KYTY_SYSV_ABI KernelAddUserEvent(KernelEqueue eq, int id) {
 int KYTY_SYSV_ABI KernelAddUserEventEdge(KernelEqueue eq, int id) {
 	PRINT_NAME();
 
-	LOGF("\t user event edge add: eq = 0x%016" PRIx64 ", id = %d\n", static_cast<uint64_t>(eq),
-	     id);
+	LOGF("\t user event edge add: eq = 0x%016" PRIx64 ", id = %d\n", static_cast<uint64_t>(eq), id);
 
 	KernelEqueueEvent event {};
 	event.event.ident         = static_cast<uintptr_t>(id);
@@ -507,8 +506,7 @@ int KYTY_SYSV_ABI KernelTriggerUserEventForAll(int id, void* udata) {
 int KYTY_SYSV_ABI KernelDeleteUserEvent(KernelEqueue eq, int id) {
 	PRINT_NAME();
 
-	LOGF("\t user event delete: eq = 0x%016" PRIx64 ", id = %d\n", static_cast<uint64_t>(eq),
-	     id);
+	LOGF("\t user event delete: eq = 0x%016" PRIx64 ", id = %d\n", static_cast<uint64_t>(eq), id);
 
 	return KernelDeleteEvent(eq, static_cast<uintptr_t>(id), KERNEL_EVFILT_USER);
 }
@@ -577,8 +575,7 @@ int KYTY_SYSV_ABI KernelAddAmprSystemEvent(KernelEqueue eq, int id, void* udata)
 int KYTY_SYSV_ABI KernelDeleteAmprEvent(KernelEqueue eq, int id) {
 	PRINT_NAME();
 
-	LOGF("\t AMPR event delete: eq = 0x%016" PRIx64 ", id = %d\n", static_cast<uint64_t>(eq),
-	     id);
+	LOGF("\t AMPR event delete: eq = 0x%016" PRIx64 ", id = %d\n", static_cast<uint64_t>(eq), id);
 
 	if (eq != KERNEL_EQUEUE_INVALID) {
 		(void)KernelDeleteEvent(eq, static_cast<uintptr_t>(id), KERNEL_EVFILT_USER);
@@ -590,8 +587,8 @@ int KYTY_SYSV_ABI KernelDeleteAmprEvent(KernelEqueue eq, int id) {
 int KYTY_SYSV_ABI KernelDeleteAmprSystemEvent(KernelEqueue eq, int id) {
 	PRINT_NAME();
 
-	LOGF("\t AMPR system event delete: eq = 0x%016" PRIx64 ", id = %d\n",
-	     static_cast<uint64_t>(eq), id);
+	LOGF("\t AMPR system event delete: eq = 0x%016" PRIx64 ", id = %d\n", static_cast<uint64_t>(eq),
+	     id);
 
 	return KernelDeleteAmprEvent(eq, id);
 }

src/kernel/macosGuestAddressSpace.cpp

@@ -0,0 +1,8 @@
+#if defined(__APPLE__) && defined(__x86_64__)
+
+// Make the process own the guest ranges before any runtime initialization.
+asm(".zerofill SYSTEM_MANAGED,SYSTEM_MANAGED,__kyty_system_managed,0x7fffbc000");
+asm(".zerofill SYSTEM_RESERVED,SYSTEM_RESERVED,__kyty_system_reserved,0x7c0004000");
+asm(".zerofill USER_AREA,USER_AREA,__kyty_user_area,0x8c00000000");
+
+#endif

src/kernel/memory.h

@@ -145,7 +145,7 @@ int KYTY_SYSV_ABI KernelIsStack(void* addr, void** start, void** end);
 int KYTY_SYSV_ABI KernelReserveVirtualRange(void** addr, size_t len, int flags, size_t alignment);
 bool              KernelHandleReservedRangeAccessViolation(uint64_t vaddr);
 int KYTY_SYSV_ABI KernelAvailableFlexibleMemorySize(size_t* size);
-int KYTY_SYSV_ABI KernelConfiguredFlexibleMemorySize(uint64_t* size);
+int KYTY_SYSV_ABI KernelConfiguredFlexibleMemorySize(size_t* size);
 int KYTY_SYSV_ABI KernelMprotect(const void* addr, size_t len, int prot);
 int KYTY_SYSV_ABI KernelMtypeprotect(const void* addr, size_t len, int type, int prot);
 int KYTY_SYSV_ABI KernelBatchMap(KernelBatchMapEntry* entries, int num_entries,
@@ -163,17 +163,30 @@ int KYTY_SYSV_ABI KernelMemoryPoolBatch(const KernelMemoryPoolBatchEntry* entrie
 int KYTY_SYSV_ABI KernelMemoryPoolGetBlockStats(KernelMemoryPoolBlockStats* output,
                                                 size_t                      output_size);
 
-void RegisterProgramMemory(uint64_t vaddr, uint64_t size, Common::VirtualMemory::Mode mode,
-                           const char* name);
-void UpdateProgramMemoryProtection(uint64_t vaddr, uint64_t size, Common::VirtualMemory::Mode mode);
-void UnregisterProgramMemory(uint64_t vaddr, uint64_t size);
+uint64_t AllocateProgramMemory(uint64_t search_addr, uint64_t size,
+                               Common::VirtualMemory::Mode mode, const char* name);
+void SetProgramMemoryProtection(uint64_t vaddr, uint64_t size, Common::VirtualMemory::Mode mode);
+uint64_t AllocateRuntimeMemory(uint64_t search_addr, uint64_t size,
+                               Common::VirtualMemory::Mode mode, const char* name,
+                               bool fixed = false);
+uint64_t AllocateGuestStackMemory(uint64_t search_addr, uint64_t size,
+                                  Common::VirtualMemory::Mode mode, const char* name);
+bool     ProtectGuestMemory(uint64_t vaddr, uint64_t size, Common::VirtualMemory::Mode mode,
+                            Common::VirtualMemory::Mode* old_mode = nullptr);
+// Transient PageManager watch state; does not change the guest mapping's semantic protection.
+bool ProtectGuestHostMemory(uint64_t vaddr, uint64_t size, Common::VirtualMemory::Mode mode);
+bool FreeGuestMemory(uint64_t vaddr, uint64_t size);
 
 #if defined(KYTY_VIRTUAL_MEMORY_ALLOCATION_TESTS)
 void     TestFailNextPhysicalMemoryUnmap();
 void     TestFailPhysicalMemoryUnmapAfter(uint32_t successful_unmaps);
-void TestFailHostReservationAfter(uint32_t successful_pages);
+void     TestFailGuestBackingStoreUnmapAfter(uint32_t successful_unmaps);
 void     TestFailNextFixedReserveRangeRegistration();
 bool     TestPlaceholderRangeIsFree(uint64_t vaddr, uint64_t size);
+bool     TestGuestAddressRangeIsOwned(uint64_t vaddr, uint64_t size);
+bool     TestGuestBackingOutsideAddressSpace();
+uint64_t TestGuestBackingSize();
+bool     TestGuestFreeRangeBounds();
 #endif
 
 } // namespace Libs::LibKernel::Memory

src/kernel/pthread.cpp

@@ -65,6 +65,10 @@
 
 namespace Libs {
 
+namespace LibcInternalExt {
+void RunThreadAtexitDestructors();
+} // namespace LibcInternalExt
+
 namespace LibKernel {
 
 LIB_NAME("libkernel", "libkernel");
@@ -80,10 +84,10 @@ constexpr int      DESTRUCTOR_ITERATIONS     = 4;
 constexpr size_t   PTHREAD_STACK_DEFAULT   = 0x100000;
 constexpr size_t   GUEST_PTHREAD_STACK_MIN = 0x4000;
 constexpr size_t   PTHREAD_STACK_PAGE      = 0x4000;
-constexpr size_t   PTHREAD_STACK_GRANULARITY = 0x10000;
 constexpr size_t   PTHREAD_STACK_INITIAL   = 0x200000;
 constexpr size_t   PTHREAD_STACK_EXTRA     = 0x100000;
 constexpr uint64_t PTHREAD_STACK_TOP       = 0x7efff8000ull;
+constexpr uint64_t PTHREAD_STACK_BOTTOM    = 0x0000040000ull;
 constexpr uint32_t SIGNAL_APC_POLL_MICROS  = 10000;
 
 static constexpr KernelClockid KERNEL_CLOCK_REALTIME          = 0;
@@ -697,16 +701,17 @@ static std::atomic<int32_t> g_pthread_thread_id      = 0;
 
 static Common::Mutex g_guest_stack_mutex;
 static uint64_t      g_guest_stack_last = 0;
+struct CachedGuestStack {
+	uint64_t address;
+	size_t   map_size;
+	size_t   guard_size;
+};
+static std::vector<CachedGuestStack> g_guest_stack_cache;
 
 static size_t RoundStackSize(size_t size) {
 	return ((size + PTHREAD_STACK_PAGE - 1) / PTHREAD_STACK_PAGE) * PTHREAD_STACK_PAGE;
 }
 
-static size_t RoundStackMappingSize(size_t size) {
-	return ((size + PTHREAD_STACK_GRANULARITY - 1) / PTHREAD_STACK_GRANULARITY) *
-	       PTHREAD_STACK_GRANULARITY;
-}
-
 static int CreateGuestStack(PthreadAttr attr) {
 	if (attr == nullptr) {
 		return KERNEL_ERROR_EINVAL;
@@ -722,35 +727,42 @@ static int CreateGuestStack(PthreadAttr attr) {
 
 	const auto stack_size = RoundStackSize(attr->stack_size);
 	const auto guard_size = RoundStackSize(attr->guard_size);
-	const auto map_size   = RoundStackMappingSize(stack_size + guard_size);
+	const auto map_size   = stack_size + guard_size;
 
 	uint64_t stack_addr = 0;
+	bool     cached     = false;
 	{
 		Common::LockGuard lock(g_guest_stack_mutex);
 
+		auto cached_stack =
+		    std::find_if(g_guest_stack_cache.begin(), g_guest_stack_cache.end(),
+		                 [map_size, guard_size](const auto& stack) {
+			                 return stack.map_size == map_size && stack.guard_size == guard_size;
+		                 });
+		if (cached_stack != g_guest_stack_cache.end()) {
+			stack_addr = cached_stack->address;
+			g_guest_stack_cache.erase(cached_stack);
+			cached = true;
+		} else {
 			if (g_guest_stack_last == 0) {
-			g_guest_stack_last = (PTHREAD_STACK_TOP - PTHREAD_STACK_INITIAL - PTHREAD_STACK_PAGE) &
-			                     ~(static_cast<uint64_t>(PTHREAD_STACK_GRANULARITY) - 1);
+				g_guest_stack_last = PTHREAD_STACK_TOP - PTHREAD_STACK_INITIAL - PTHREAD_STACK_PAGE;
+			}
+			if (map_size > g_guest_stack_last - PTHREAD_STACK_BOTTOM) {
+				return KERNEL_ERROR_EAGAIN;
 			}
-
 			stack_addr = g_guest_stack_last - map_size;
 			g_guest_stack_last -= map_size;
 		}
+	}
 
-	void* mapped_addr = reinterpret_cast<void*>(stack_addr);
-
-	constexpr int GUEST_PROT_READ_WRITE = 0x03;
-	constexpr int GUEST_MAP_PRIVATE     = 0x02;
-	constexpr int GUEST_MAP_FIXED       = 0x10;
-	constexpr int GUEST_MAP_STACK       = 0x400;
-	constexpr int GUEST_MAP_ANON        = 0x1000;
-
-	int result = Memory::KernelMapNamedFlexibleMemory(
-	    &mapped_addr, map_size, GUEST_PROT_READ_WRITE,
-	    GUEST_MAP_PRIVATE | GUEST_MAP_FIXED | GUEST_MAP_STACK | GUEST_MAP_ANON, "stack");
-	if (result != OK) {
+	int result = OK;
+	if (!cached) {
+		stack_addr = Memory::AllocateGuestStackMemory(
+		    stack_addr, map_size, Common::VirtualMemory::Mode::ReadWrite, "stack");
+		if (stack_addr == 0) {
 			return KERNEL_ERROR_EAGAIN;
 		}
+	}
 
 	if (guard_size != 0) {
 		result = Memory::KernelMprotect(reinterpret_cast<void*>(stack_addr), guard_size, 0);
@@ -761,7 +773,7 @@ static int CreateGuestStack(PthreadAttr attr) {
 	}
 
 	attr->stack_addr     = reinterpret_cast<void*>(stack_addr + guard_size);
-	attr->stack_size     = map_size - guard_size;
+	attr->stack_size     = stack_size;
 	attr->stack_user     = false;
 	attr->stack_map_addr = stack_addr;
 	attr->stack_map_size = map_size;
@@ -777,13 +789,90 @@ static void FreeGuestStack(PthreadAttr attr) {
 		return;
 	}
 
-	Memory::KernelMunmap(attr->stack_map_addr, attr->stack_map_size);
+	const auto guard_size = attr->stack_map_size - attr->stack_size;
+	{
+		Common::LockGuard lock(g_guest_stack_mutex);
+		g_guest_stack_cache.push_back({attr->stack_map_addr, attr->stack_map_size, guard_size});
+	}
 
 	attr->stack_addr     = nullptr;
 	attr->stack_map_addr = 0;
 	attr->stack_map_size = 0;
 }
 
+#if defined(KYTY_VIRTUAL_MEMORY_ALLOCATION_TESTS)
+bool TestGuestStackOwnerLifecycle(uint64_t* first_address, uint64_t* second_address,
+                                  uint64_t* map_size) {
+	if (first_address == nullptr || second_address == nullptr || map_size == nullptr) {
+		return false;
+	}
+
+	size_t flexible_before = 0;
+	if (Memory::KernelAvailableFlexibleMemorySize(&flexible_before) != OK) {
+		return false;
+	}
+
+	PthreadAttr attr = nullptr;
+	if (PthreadAttrInit(&attr) != OK) {
+		return false;
+	}
+	if (CreateGuestStack(attr) != OK) {
+		PthreadAttrDestroy(&attr);
+		return false;
+	}
+	*first_address = attr->stack_map_addr;
+	*map_size      = attr->stack_map_size;
+	const bool first_owned =
+	    Memory::TestGuestAddressRangeIsOwned(*first_address, static_cast<uint64_t>(*map_size));
+	uint64_t   backing_value = 0;
+	const bool first_private =
+	    !Memory::TryReadBacking(*first_address, &backing_value, sizeof(backing_value));
+	size_t     flexible_during_first = 0;
+	const bool first_capacity_unchanged =
+	    Memory::KernelAvailableFlexibleMemorySize(&flexible_during_first) == OK &&
+	    flexible_during_first == flexible_before;
+	FreeGuestStack(attr);
+
+	if (CreateGuestStack(attr) != OK) {
+		PthreadAttrDestroy(&attr);
+		return false;
+	}
+	*second_address = attr->stack_map_addr;
+	const bool second_owned =
+	    Memory::TestGuestAddressRangeIsOwned(*second_address, static_cast<uint64_t>(*map_size));
+	const bool second_private =
+	    !Memory::TryReadBacking(*second_address, &backing_value, sizeof(backing_value));
+	size_t     flexible_during_second = 0;
+	const bool second_capacity_unchanged =
+	    Memory::KernelAvailableFlexibleMemorySize(&flexible_during_second) == OK &&
+	    flexible_during_second == flexible_before;
+	FreeGuestStack(attr);
+
+	CachedGuestStack cached {};
+	bool             found = false;
+	{
+		Common::LockGuard lock(g_guest_stack_mutex);
+		const auto        entry = std::find_if(
+		    g_guest_stack_cache.begin(), g_guest_stack_cache.end(),
+		    [second_address](const auto& stack) { return stack.address == *second_address; });
+		if (entry != g_guest_stack_cache.end()) {
+			cached = *entry;
+			g_guest_stack_cache.erase(entry);
+			found = true;
+		}
+	}
+
+	const bool unmapped = found && Memory::KernelMunmap(cached.address, cached.map_size) == OK;
+	size_t     flexible_after = 0;
+	const bool final_capacity_unchanged =
+	    Memory::KernelAvailableFlexibleMemorySize(&flexible_after) == OK &&
+	    flexible_after == flexible_before;
+	return PthreadAttrDestroy(&attr) == OK && first_owned && first_private &&
+	       first_capacity_unchanged && second_owned && second_private &&
+	       second_capacity_unchanged && unmapped && final_capacity_unchanged;
+}
+#endif
+
 static KYTY_SYSV_ABI void* RunOnGuestStack(void* arg, pthread_entry_func_t func, void* stack_top) {
 #if defined(__x86_64__) || defined(_M_X64)
 	void*      ret       = nullptr;
@@ -3262,6 +3351,8 @@ int PthreadGetCurrentPriorityForKernel() {
 static void CleanupThread(void* arg) {
 	auto* thread = static_cast<Pthread>(arg);
 
+	LibcInternalExt::RunThreadAtexitDestructors();
+
 	auto thread_dtors = g_pthread_context->GetThreadDtors();
 
 	if (thread_dtors != nullptr) {
@@ -3795,8 +3886,12 @@ int KYTY_SYSV_ABI KernelGettimeofday(KernelTimeval* tp) {
 	tp->tv_sec  = static_cast<int64_t>(ticks / 1000000);
 	tp->tv_usec = static_cast<int64_t>(ticks % 1000000);
 #else
-	auto dt = Common::DateTime::FromSystemUTC();
-	sec_to_timeval(tp, dt.ToUnix());
+	struct timespec ts {};
+	result = ::clock_gettime(CLOCK_REALTIME, &ts);
+	if (result == 0) {
+		tp->tv_sec  = static_cast<int64_t>(ts.tv_sec);
+		tp->tv_usec = static_cast<int64_t>(ts.tv_nsec / 1000);
+	}
 #endif
 
 	if (result == 0) {

src/kernel/pthread.h

@@ -112,6 +112,10 @@ void                  PthreadQueuePendingSignal(Pthread thread, int signum);
 bool                  PthreadHasPendingSignal(Pthread thread, int signum);
 bool                  PthreadTakePendingSignal(Pthread thread, int signum);
 bool PthreadGetGuestStack(Pthread thread, uint64_t* stack_addr, uint64_t* stack_size);
+#if defined(KYTY_VIRTUAL_MEMORY_ALLOCATION_TESTS)
+bool TestGuestStackOwnerLifecycle(uint64_t* first_address, uint64_t* second_address,
+                                  uint64_t* map_size);
+#endif
 #if KYTY_PLATFORM != KYTY_PLATFORM_WINDOWS
 bool PthreadKillHost(Pthread thread, int host_signal);
 #endif

src/launcher/src/configurationItem.cpp

@@ -206,8 +206,8 @@ bool ConfigurationItem::operator<(const QTreeWidgetItem& other) const {
 			       GetStatusText(other_item->m_info->game_status);
 		case GameVersionColumn:
 		case FirmwareVersionColumn: {
-			const auto& version = column == GameVersionColumn ? m_info->gameVersion
-			                                                  : m_info->firmwareVer;
+			const auto& version =
+			    column == GameVersionColumn ? m_info->gameVersion : m_info->firmwareVer;
 			const auto& other_version = column == GameVersionColumn
 			                                ? other_item->m_info->gameVersion
 			                                : other_item->m_info->firmwareVer;

src/launcher/src/configurationListWidget.cpp

@@ -1,6 +1,5 @@
 #include "configurationListWidget.h"
 
-#include "patchesDialog.h"
 #include "common.h"
 #include "compatibilityDatabase.h"
 #include "configuration.h"
@@ -8,6 +7,7 @@
 #include "configurationItem.h"
 #include "gameListTreeWidget.h"
 #include "mainDialog.h"
+#include "patchesDialog.h"
 #include "trophyViewerDialog.h"
 
 #include <QAbstractItemModel>

src/launcher/src/mainDialog.cpp

@@ -272,10 +272,18 @@ static bool FindTerminal(QString* program, QStringList* prefix) {
 	};
 
 	static const TerminalSpec candidates[] = {
-	    {"x-terminal-emulator", "-e"}, {"gnome-terminal", "--"}, {"konsole", "-e"},
-	    {"xfce4-terminal", "-x"},      {"mate-terminal", "--"},  {"tilix", "-e"},
-	    {"alacritty", "-e"},           {"kitty", nullptr},       {"foot", nullptr},
-	    {"wezterm", "-e"},             {"urxvt", "-e"},          {"xterm", "-e"},
+	    {"x-terminal-emulator", "-e"},
+	    {"gnome-terminal", "--"},
+	    {"konsole", "-e"},
+	    {"xfce4-terminal", "-x"},
+	    {"mate-terminal", "--"},
+	    {"tilix", "-e"},
+	    {"alacritty", "-e"},
+	    {"kitty", nullptr},
+	    {"foot", nullptr},
+	    {"wezterm", "-e"},
+	    {"urxvt", "-e"},
+	    {"xterm", "-e"},
 	};
 
 	const auto try_candidate = [program, prefix](const QString& executable, const char* separator) {
@@ -379,9 +387,9 @@ void MainDialog::RunInterpreter(QProcess* process, const Configuration& info) {
 #if !defined(_WIN32)
 	// Report immediate launch failures.
 	if (!process->waitForStarted(5000)) {
-		QMessageBox::critical(this, tr("Error"),
-		                      tr("Failed to start:\n%1\n\n%2")
-		                          .arg(process->program(), process->errorString()));
+		QMessageBox::critical(
+		    this, tr("Error"),
+		    tr("Failed to start:\n%1\n\n%2").arg(process->program(), process->errorString()));
 		return;
 	}
 #endif

src/launcher/src/patchesDialog.cpp

@@ -59,10 +59,8 @@ void PatchesDialog::Load() {
 		return;
 	}
 
-	const auto patches = QJsonDocument::fromJson(file.readAll())
-	                         .object()
-	                         .value(QStringLiteral("patches"))
-	                         .toArray();
+	const auto patches =
+	    QJsonDocument::fromJson(file.readAll()).object().value(QStringLiteral("patches")).toArray();
 	for (const auto& value: patches) {
 		const auto patch = value.toObject();
 		auto* item = new QListWidgetItem(patch.value(QStringLiteral("name")).toString(), m_patches);

src/libs/agc.cpp

@@ -834,6 +834,104 @@ int KYTY_SYSV_ABI GraphicsUnknownFuseShaderHalves(Shader* fused_result, const Sh
 	return OK;
 }
 
+static void merge_shader_register_max_field(ShaderRegister* dst, const ShaderRegister* src,
+                                            uint32_t shift, uint32_t mask) {
+	const auto dst_field = (dst->value >> shift) & mask;
+	const auto src_field = (src->value >> shift) & mask;
+	const auto field     = std::max(dst_field, src_field);
+
+	dst->value &= ~(mask << shift);
+	dst->value |= field << shift;
+}
+
+int KYTY_SYSV_ABI GraphicsUnknownNApJjpKNBl4(Shader* fused_result, const Shader* front,
+                                             const Shader* back, void* scratch_mem) {
+	PRINT_NAME();
+
+	LOGF("\t fused_result = 0x%016" PRIx64 "\n"
+	     "\t front        = 0x%016" PRIx64 "\n"
+	     "\t back         = 0x%016" PRIx64 "\n"
+	     "\t scratch_mem  = 0x%016" PRIx64 "\n",
+	     reinterpret_cast<uint64_t>(fused_result), reinterpret_cast<uint64_t>(front),
+	     reinterpret_cast<uint64_t>(back), reinterpret_cast<uint64_t>(scratch_mem));
+
+	const auto front_type = static_cast<Prospero::ShaderBinaryType>(front->type);
+	const auto is_gs      = front_type == Prospero::ShaderBinaryType::kGsFront;
+	const auto is_hs      = front_type == Prospero::ShaderBinaryType::kHsFront;
+	if ((!is_gs && !is_hs) ||
+	    (is_gs && back->type != static_cast<uint8_t>(Prospero::ShaderBinaryType::kGsBack)) ||
+	    (is_hs && back->type != static_cast<uint8_t>(Prospero::ShaderBinaryType::kHsBack))) {
+		return GRAPHICS5_ERROR_INVALID_SHADER_HALVES;
+	}
+
+	*fused_result      = *back;
+	fused_result->type = static_cast<uint8_t>(is_gs ? Prospero::ShaderBinaryType::kGs
+	                                                : Prospero::ShaderBinaryType::kHs);
+
+	const auto back_stages  = back->specials->vgt_shader_stages_en.value;
+	const auto front_stages = front->specials->vgt_shader_stages_en.value;
+	const auto mismatch_bit = is_gs ? (1u << 22u) : (1u << 21u);
+	if (((front_stages ^ back_stages) & mismatch_bit) != 0) {
+		return GRAPHICS5_ERROR_INVALID_SHADER_HALVES;
+	}
+
+	if (scratch_mem != nullptr) {
+		auto* sh_registers = static_cast<ShaderRegister*>(scratch_mem);
+		memcpy(sh_registers, back->sh_registers,
+		       static_cast<size_t>(back->num_sh_registers) * sizeof(ShaderRegister));
+		fused_result->sh_registers = sh_registers;
+	}
+
+	auto*      fused_regs      = fused_result->sh_registers;
+	const auto fused_reg_count = static_cast<uint32_t>(fused_result->num_sh_registers);
+	const auto front_reg_count = static_cast<uint32_t>(front->num_sh_registers);
+	const auto checksum_offset =
+	    is_gs ? Pm4::SPI_SHADER_PGM_CHKSUM_GS : Pm4::SPI_SHADER_PGM_CHKSUM_HS;
+	const auto* front_checksum0 =
+	    find_shader_register(front->sh_registers, front_reg_count, checksum_offset, 0);
+	const auto* front_checksum1 =
+	    find_shader_register(front->sh_registers, front_reg_count, checksum_offset, 1);
+	auto* fused_checksum0  = find_shader_register(fused_regs, fused_reg_count, checksum_offset, 0);
+	auto* fused_checksum1  = find_shader_register(fused_regs, fused_reg_count, checksum_offset, 1);
+	fused_checksum0->value = front_checksum0->value;
+	fused_checksum1->value = front_checksum1->value;
+
+	const auto  rsrc1_offset = is_gs ? Pm4::SPI_SHADER_PGM_RSRC1_GS : Pm4::SPI_SHADER_PGM_RSRC1_HS;
+	const auto  rsrc2_offset = is_gs ? Pm4::SPI_SHADER_PGM_RSRC2_GS : Pm4::SPI_SHADER_PGM_RSRC2_HS;
+	const auto* front_rsrc1 =
+	    find_shader_register(front->sh_registers, front_reg_count, rsrc1_offset);
+	const auto* front_rsrc2 =
+	    find_shader_register(front->sh_registers, front_reg_count, rsrc2_offset);
+	auto* fused_rsrc1 = find_shader_register(fused_regs, fused_reg_count, rsrc1_offset);
+	auto* fused_rsrc2 = find_shader_register(fused_regs, fused_reg_count, rsrc2_offset);
+
+	merge_shader_register_max_field(fused_rsrc1, front_rsrc1, 0, 0x3fu);
+	merge_shader_register_max_field(fused_rsrc2, front_rsrc2, 28, 0x0fu);
+	if (is_gs) {
+		merge_shader_register_max_field(fused_rsrc1, front_rsrc1, 29, 0x03u);
+		merge_shader_register_max_field(fused_rsrc2, front_rsrc2, 16, 0x03u);
+		fused_rsrc2->value =
+		    (fused_rsrc2->value & 0xf7ffffc1u) | (front_rsrc2->value & 0x0800003eu);
+		fused_rsrc2->value =
+		    (fused_rsrc2->value & 0xfffbffffu) | (front_rsrc2->value & 0x00040000u);
+	} else {
+		merge_shader_register_max_field(fused_rsrc1, front_rsrc1, 28, 0x03u);
+		fused_rsrc2->value =
+		    (fused_rsrc2->value & 0xf7ffffc1u) | (front_rsrc2->value & 0x0800003eu);
+	}
+
+	const auto program_lo_offset = is_gs ? Pm4::SPI_SHADER_PGM_LO_ES : Pm4::SPI_SHADER_PGM_LO_LS;
+	auto*      program_lo = find_shader_register(fused_regs, fused_reg_count, program_lo_offset);
+	const auto address    = reinterpret_cast<uint64_t>(front->code);
+	program_lo->value     = static_cast<uint32_t>(address >> 8u);
+	(program_lo + 1)->value &= 0xffffff00u;
+	(program_lo + 1)->value |= static_cast<uint32_t>((address >> 40u) & 0xffu);
+
+	fused_result->user_data = front->user_data;
+
+	return OK;
+}
+
 static constexpr int GRAPHICS5_ERROR_INVALID_PACKET = static_cast<int>(0x8a6c000cu);
 
 enum class RegIndirectPacket : uint32_t {
@@ -1347,7 +1445,7 @@ int KYTY_SYSV_ABI GraphicsWriteDataPatchSetAddressOrOffset(uint32_t* cmd,
 		return OK;
 	}
 
-	return static_cast<int>(0x8a6c000cu);
+	return GRAPHICS5_ERROR_INVALID_PACKET;
 }
 
 int KYTY_SYSV_ABI GraphicsUnknownJumpPatchSetTarget(uint32_t* cmd, const volatile uint32_t* target,
@@ -2380,7 +2478,7 @@ int KYTY_SYSV_ABI GraphicsUnknownIkfdtRIqCE(uint32_t* cmd, uint64_t arg1,
 
 	auto op = (cmd[0] >> 8u) & 0xffu;
 	if (op != Pm4::IT_INDIRECT_BUFFER) {
-		return 0x8a6c000c;
+		return GRAPHICS5_ERROR_INVALID_PACKET;
 	}
 
 	auto vaddr        = reinterpret_cast<uint64_t>(target);
@@ -2800,6 +2898,10 @@ uint32_t KYTY_SYSV_ABI GraphicsAcbCondExecGetSize() {
 	return GraphicsDcbCondExecGetSize();
 }
 
+uint32_t KYTY_SYSV_ABI GraphicsAcbJumpGetSize() {
+	return 0x10u;
+}
+
 uint32_t* KYTY_SYSV_ABI GraphicsAcbWaitRegMem(CommandBuffer* buf, uint8_t size,
                                               uint8_t compare_function, uint8_t cache_policy,
                                               const volatile void* address, uint64_t reference,
@@ -3119,7 +3221,7 @@ int KYTY_SYSV_ABI GraphicsDmaDataPatchSetDstAddressOrOffset(uint32_t* cmd,
 		return OK;
 	}
 
-	return static_cast<int>(0x8a6c000cu);
+	return GRAPHICS5_ERROR_INVALID_PACKET;
 }
 
 int KYTY_SYSV_ABI GraphicsDmaDataPatchSetSrcAddressOrOffsetOrImmediate(
@@ -3133,7 +3235,7 @@ int KYTY_SYSV_ABI GraphicsDmaDataPatchSetSrcAddressOrOffsetOrImmediate(
 		return OK;
 	}
 
-	return static_cast<int>(0x8a6c000cu);
+	return GRAPHICS5_ERROR_INVALID_PACKET;
 }
 
 uint32_t KYTY_SYSV_ABI GraphicsGetPacketSize(uint32_t* packet) {
@@ -3197,6 +3299,15 @@ int KYTY_SYSV_ABI GraphicsSetRangePredication(uint32_t* start, const volatile ui
 	return OK;
 }
 
+int KYTY_SYSV_ABI GraphicsRewindPatchSetRewindState(uint32_t* cmd, uint8_t state) {
+	if (((cmd[0] >> 8u) & 0xffu) != Pm4::IT_REWIND) {
+		return GRAPHICS5_ERROR_INVALID_PACKET;
+	}
+
+	cmd[1] = (cmd[1] & 0x7fffffffu) | (static_cast<uint32_t>(state) << 31u);
+	return OK;
+}
+
 int KYTY_SYSV_ABI GraphicsCondExecPatchSetEnd(uint32_t* cmd, const volatile uint32_t* buffer) {
 	PRINT_NAME();
 
@@ -3205,23 +3316,23 @@ int KYTY_SYSV_ABI GraphicsCondExecPatchSetEnd(uint32_t* cmd, const volatile uint
 	     reinterpret_cast<uint64_t>(cmd), reinterpret_cast<uint64_t>(buffer));
 
 	if (cmd == nullptr || buffer == nullptr) {
-		return static_cast<int>(0x8a6c000cu);
+		return GRAPHICS5_ERROR_INVALID_PACKET;
 	}
 
 	auto op = (cmd[0] >> 8u) & 0xffu;
 	if (op != Pm4::IT_COND_EXEC) {
-		return static_cast<int>(0x8a6c000cu);
+		return GRAPHICS5_ERROR_INVALID_PACKET;
 	}
 
 	auto* packet_end = cmd + 5;
 	auto* range_end  = const_cast<uint32_t*>(reinterpret_cast<const volatile uint32_t*>(buffer));
 	if (range_end < packet_end) {
-		return static_cast<int>(0x8a6c000cu);
+		return GRAPHICS5_ERROR_INVALID_PACKET;
 	}
 
 	auto num_dwords = static_cast<uint64_t>(range_end - packet_end);
 	if (num_dwords > 0x3fffu) {
-		return static_cast<int>(0x8a6c000cu);
+		return GRAPHICS5_ERROR_INVALID_PACKET;
 	}
 
 	cmd[4] = (cmd[4] & ~0x3fffu) | static_cast<uint32_t>(num_dwords);
@@ -3237,12 +3348,12 @@ int KYTY_SYSV_ABI GraphicsCondExecPatchSetCommandAddress(uint32_t*
 	     reinterpret_cast<uint64_t>(cmd), reinterpret_cast<uint64_t>(command));
 
 	if (cmd == nullptr || command == nullptr) {
-		return static_cast<int>(0x8a6c000cu);
+		return GRAPHICS5_ERROR_INVALID_PACKET;
 	}
 
 	auto op = (cmd[0] >> 8u) & 0xffu;
 	if (op != Pm4::IT_COND_EXEC || (reinterpret_cast<uintptr_t>(command) & 0x3u) != 0) {
-		return static_cast<int>(0x8a6c000cu);
+		return GRAPHICS5_ERROR_INVALID_PACKET;
 	}
 
 	auto addr = reinterpret_cast<uint64_t>(command);

src/libs/agc.h

@@ -38,6 +38,8 @@ int KYTY_SYSV_ABI   GraphicsUnknownGetFusedShaderSize(SizeAlign* dst, const Shad
                                                       const Shader* back);
 int KYTY_SYSV_ABI   GraphicsUnknownFuseShaderHalves(Shader* fused_result, const Shader* front,
                                                     const Shader* back, void* scratch_mem);
+int KYTY_SYSV_ABI   GraphicsUnknownNApJjpKNBl4(Shader* fused_result, const Shader* front,
+                                               const Shader* back, void* scratch_mem);
 int KYTY_SYSV_ABI   GraphicsSetCxRegIndirectPatchSetAddress(uint32_t*                      cmd,
                                                             const volatile ShaderRegister* regs);
 int KYTY_SYSV_ABI   GraphicsSetShRegIndirectPatchSetAddress(uint32_t*                      cmd,
@@ -183,6 +185,7 @@ uint32_t KYTY_SYSV_ABI  GraphicsAcbAcquireMemGetSize();
 uint32_t* KYTY_SYSV_ABI GraphicsAcbCondExec(CommandBuffer* buf, const volatile uint32_t* address,
                                             uint32_t num_dwords);
 uint32_t KYTY_SYSV_ABI  GraphicsAcbCondExecGetSize();
+uint32_t KYTY_SYSV_ABI  GraphicsAcbJumpGetSize();
 uint32_t* KYTY_SYSV_ABI GraphicsAcbWaitRegMem(CommandBuffer* buf, uint8_t size,
                                               uint8_t compare_function, uint8_t cache_policy,
                                               const volatile void* address, uint64_t reference,
@@ -241,6 +244,7 @@ uint32_t KYTY_SYSV_ABI  GraphicsGetPacketSize(uint32_t* packet);
 int KYTY_SYSV_ABI       GraphicsSetPacketPredication(uint32_t* packet, uint32_t predication);
 int KYTY_SYSV_ABI       GraphicsSetRangePredication(uint32_t* start, const volatile uint32_t* end,
                                                     uint32_t predication);
+int KYTY_SYSV_ABI       GraphicsRewindPatchSetRewindState(uint32_t* cmd, uint8_t state);
 int KYTY_SYSV_ABI       GraphicsCondExecPatchSetEnd(uint32_t* cmd, const volatile uint32_t* buffer);
 int KYTY_SYSV_ABI       GraphicsCondExecPatchSetCommandAddress(uint32_t*                cmd,
                                                                const volatile uint32_t* command);

src/libs/audio.cpp

@@ -364,7 +364,12 @@ bool Audio::QueueSdlAudio(PortOut* port, const void* data, bool blocking) {
 	}
 
 	if (blocking) {
-		const auto min_queued_size = queue_size * 2u;
+		constexpr uint64_t target_latency_us = 40000;
+		const auto buffer_us = port->freq != 0 ? (1000000ULL * port->samples_num) / port->freq : 0;
+		const auto buffers =
+		    buffer_us != 0 ? static_cast<uint32_t>((target_latency_us + buffer_us - 1) / buffer_us)
+		                   : 2u;
+		const auto min_queued_size = queue_size * std::clamp(buffers, 2u, 16u);
 		const auto wait_start      = LibKernel::KernelGetProcessTime();
 		while (SDL_GetQueuedAudioSize(port->audio_device) > min_queued_size) {
 			if (LibKernel::KernelGetProcessTime() - wait_start > 200000) {
@@ -511,7 +516,16 @@ uint32_t Audio::AudioOutOutputs(OutputParam* params, uint32_t num, bool blocking
 		max_wait_time      = (wait_time > max_wait_time ? wait_time : max_wait_time);
 	}
 
-	if (blocking && max_wait_time != 0) {
+	bool all_ports_have_device = true;
+	for (uint32_t i = 0; i < num; i++) {
+		if (m_out_ports[params[i].handle.GetId()].audio_device == 0) {
+			all_ports_have_device = false;
+			break;
+		}
+	}
+
+	// Device-backed ports are paced by the SDL queue above.
+	if (blocking && max_wait_time != 0 && !all_ports_have_device) {
 		Common::Thread::SleepMicro(max_wait_time);
 	}
 
@@ -1416,6 +1430,12 @@ struct Ngs2CustomSubmixerRackOption {
 	uint32_t             max_inputs   = 0;
 };
 
+struct Ngs2CustomMasteringRackOption {
+	Ngs2CustomRackOption custom_rack_option;
+	uint32_t             max_channels = 0;
+	uint32_t             max_inputs   = 0;
+};
+
 struct Ngs2CustomSamplerRackOption {
 	Ngs2CustomRackOption custom_rack_option;
 	uint32_t             max_channel_works        = 0;
@@ -1433,6 +1453,7 @@ union Ngs2RackOptionUnion {
 	Ngs2SubmixerRackOption       submixer;
 	Ngs2ReverbRackOption         reverb;
 	Ngs2CustomSubmixerRackOption custom_submixer;
+	Ngs2CustomMasteringRackOption custom_mastering;
 	Ngs2CustomSamplerRackOption  custom_sampler;
 };
 
@@ -1588,6 +1609,7 @@ enum class Ngs2RackType {
 	Mastering,
 	Reverb,
 	CustomSubmixer,
+	CustomMastering,
 	CustomSampler,
 };
 
@@ -1665,6 +1687,20 @@ struct Ngs2VoiceCallbackParam {
 
 struct Ngs2VoiceState {
 	uint32_t state_flags;
+	int32_t  error_code;
+};
+
+struct Ngs2SubmixerVoiceState {
+	Ngs2VoiceState voice_state;
+	float          envelope_height;
+	float          peak_height;
+	float          compressor_height;
+};
+
+struct Ngs2CustomMasteringVoiceState {
+	Ngs2VoiceState voice_state;
+	uint32_t       reserved;
+	uint32_t       reserved2;
 };
 
 struct Ngs2SamplerVoiceState {
@@ -1683,6 +1719,10 @@ static Ngs2RackInternal* g_racks_list = nullptr;
 
 static_assert(sizeof(Ngs2SystemOption) == 144);
 static_assert(sizeof(Ngs2RackOption) == 176);
+static_assert(sizeof(Ngs2VoiceState) == 8);
+static_assert(sizeof(Ngs2SubmixerVoiceState) == 20);
+static_assert(sizeof(Ngs2CustomMasteringVoiceState) == 16);
+static_assert(sizeof(Ngs2SamplerVoiceState) == 56);
 
 static uint32_t Ngs2GetStateFlags(const Ngs2VoiceInternal* voice) {
 	switch (voice->state) {
@@ -1726,6 +1766,7 @@ static Ngs2Internal* Ngs2CreateSystemInternal(const Ngs2SystemOption* option, vo
 static bool Ngs2RackIsCustom(Ngs2RackType type) {
 	switch (type) {
 		case Ngs2RackType::CustomSubmixer:
+		case Ngs2RackType::CustomMastering:
 		case Ngs2RackType::CustomSampler: return true;
 		default: return false;
 	}
@@ -2037,6 +2078,12 @@ int KYTY_SYSV_ABI Ngs2RackCreate(uintptr_t system_handle, uint32_t rack_id,
 			    *reinterpret_cast<const Ngs2CustomSubmixerRackOption*>(option);
 			rack->type = Ngs2RackType::CustomSubmixer;
 			break;
+		case 0x4003:
+			EXIT_NOT_IMPLEMENTED(option->size != sizeof(Ngs2CustomMasteringRackOption));
+			rack->option.custom_mastering =
+			    *reinterpret_cast<const Ngs2CustomMasteringRackOption*>(option);
+			rack->type = Ngs2RackType::CustomMastering;
+			break;
 		case 0x4001:
 			EXIT_NOT_IMPLEMENTED(option->size != sizeof(Ngs2CustomSamplerRackOption));
 			rack->option.custom_sampler =
@@ -2550,6 +2597,9 @@ int KYTY_SYSV_ABI Ngs2VoiceControl(uintptr_t voice_handle, const Ngs2VoiceParamH
 			case 0x4002:
 				EXIT_NOT_IMPLEMENTED(voice->rack->type != Ngs2RackType::CustomSubmixer);
 				break;
+			case 0x4003:
+				EXIT_NOT_IMPLEMENTED(voice->rack->type != Ngs2RackType::CustomMastering);
+				break;
 			default: EXIT("unknown rack_id: 0x%" PRIx32 "\n", rack_id);
 		}
 
@@ -2587,6 +2637,25 @@ int KYTY_SYSV_ABI Ngs2VoiceGetState(uintptr_t voice_handle, Ngs2VoiceState* stat
 	Common::LockGuard lock(voice->rack->ngs->mutex);
 
 	switch (voice->rack->type) {
+		case Ngs2RackType::Submixer: {
+			EXIT_NOT_IMPLEMENTED(state_size != sizeof(Ngs2SubmixerVoiceState));
+			auto* submixer = reinterpret_cast<Ngs2SubmixerVoiceState*>(state);
+			*submixer     = {};
+			submixer->voice_state.state_flags = Ngs2GetStateFlags(voice);
+			LOGF("\t state_flags = %u\n", submixer->voice_state.state_flags);
+			break;
+		}
+		case Ngs2RackType::CustomMastering: {
+			const auto configured_size =
+			    voice->rack->option.custom_mastering.custom_rack_option.state_size;
+			EXIT_NOT_IMPLEMENTED(configured_size < sizeof(Ngs2CustomMasteringVoiceState));
+			EXIT_NOT_IMPLEMENTED(state_size != configured_size);
+			std::memset(state, 0, state_size);
+			auto* mastering = reinterpret_cast<Ngs2CustomMasteringVoiceState*>(state);
+			mastering->voice_state.state_flags = Ngs2GetStateFlags(voice);
+			LOGF("\t state_flags = %u\n", mastering->voice_state.state_flags);
+			break;
+		}
 		case Ngs2RackType::Sampler:
 		case Ngs2RackType::CustomSampler: {
 			if (state_size != sizeof(Ngs2SamplerVoiceState)) {

src/libs/libAudio2.cpp

@@ -336,18 +336,6 @@ static AudioOut2PortStateEntry* audioout2_find_port_locked(AudioOut2PortHandle p
 	return nullptr;
 }
 
-static uint32_t audioout2_context_grains(AudioOut2ContextHandle ctx) {
-	uint32_t samples_num = 512;
-
-	g_audioout2_context_mutex.Lock();
-	if (auto* state = audioout2_find_context_locked(ctx); state != nullptr) {
-		samples_num = (state->num_grains == 0 ? 512u : state->num_grains);
-	}
-	g_audioout2_context_mutex.Unlock();
-
-	return samples_num;
-}
-
 static void audioout2_queue_context_audio(AudioOut2ContextHandle ctx, bool blocking) {
 	std::vector<AudioInternal::OutputParam> params;
 	params.reserve(AudioInternal::OUT_PORTS_MAX);
@@ -455,6 +443,12 @@ int KYTY_SYSV_ABI AudioOut2ContextDestroy(AudioOut2ContextHandle ctx) {
 	PRINT_NAME();
 	LOGF("\t ctx = 0x%016" PRIx64 "\n", ctx);
 
+	g_audioout2_context_mutex.Lock();
+	if (auto* state = audioout2_find_context_locked(ctx); state != nullptr) {
+		*state = AudioOut2ContextState {};
+	}
+	g_audioout2_context_mutex.Unlock();
+
 	std::array<int, 256> audio_handles {};
 	size_t               audio_handles_num = 0;
 
@@ -473,12 +467,6 @@ int KYTY_SYSV_ABI AudioOut2ContextDestroy(AudioOut2ContextHandle ctx) {
 		audioout2_close_audio_handle(audio_handles[i]);
 	}
 
-	g_audioout2_context_mutex.Lock();
-	if (auto* state = audioout2_find_context_locked(ctx); state != nullptr) {
-		*state = AudioOut2ContextState {};
-	}
-	g_audioout2_context_mutex.Unlock();
-
 	return OK;
 }
 
@@ -560,28 +548,43 @@ int KYTY_SYSV_ABI AudioOut2PortCreate(AudioOut2ContextHandle ctx, const AudioOut
 	EXIT_NOT_IMPLEMENTED(port == nullptr);
 
 	const auto next_port    = g_audioout2_next_port.fetch_add(1, std::memory_order_relaxed);
+	const auto audio_format = audioout2_data_format_to_audio_format(params->data_format);
+	const auto audio_type   = audioout2_port_type_to_audio_out_type(params->port_type);
+
+	g_audioout2_context_mutex.Lock();
+	const auto* context_state = audioout2_find_context_locked(ctx);
+	if (context_state == nullptr) {
+		g_audioout2_context_mutex.Unlock();
+		return AUDIO_OUT2_ERROR_INVALID_PARAM;
+	}
+	const auto samples_num = context_state->num_grains == 0 ? 512u : context_state->num_grains;
 
 	g_audioout2_port_mutex.Lock();
-	auto* port_state = audioout2_find_port_locked(0);
-	if (port_state == nullptr) {
+	AudioOut2PortStateEntry* port_state = nullptr;
 	for (auto& candidate: g_audioout2_ports) {
 		if (!candidate.used) {
 			port_state = &candidate;
 			break;
 		}
 	}
+	if (port_state != nullptr) {
+		*port_state               = AudioOut2PortStateEntry {};
+		port_state->used          = true;
+		port_state->handle        = next_port;
+		port_state->context       = ctx;
+		port_state->port_type     = params->port_type;
+		port_state->data_format   = params->data_format;
+		port_state->sampling_freq = params->sampling_freq;
+		port_state->samples_num   = samples_num;
+		port_state->audio_format  = audio_format;
 	}
 	g_audioout2_port_mutex.Unlock();
+	g_audioout2_context_mutex.Unlock();
 
-	if (next_port > g_audioout2_ports.size() || port_state == nullptr) {
+	if (port_state == nullptr) {
 		return AUDIO_OUT2_ERROR_PORT_FULL;
 	}
 
-	*port = next_port;
-
-	const auto samples_num  = audioout2_context_grains(ctx);
-	const auto audio_format = audioout2_data_format_to_audio_format(params->data_format);
-	const auto audio_type   = audioout2_port_type_to_audio_out_type(params->port_type);
 	int audio_handle = 0;
 
 	if (audio_format != AudioInternal::Format::Unknown &&
@@ -591,17 +594,17 @@ int KYTY_SYSV_ABI AudioOut2PortCreate(AudioOut2ContextHandle ctx, const AudioOut
 	}
 
 	g_audioout2_port_mutex.Lock();
-	*port_state               = AudioOut2PortStateEntry {};
-	port_state->used          = true;
-	port_state->handle        = *port;
-	port_state->context       = ctx;
-	port_state->port_type     = params->port_type;
-	port_state->data_format   = params->data_format;
-	port_state->sampling_freq = params->sampling_freq;
-	port_state->samples_num   = samples_num;
-	port_state->audio_format  = audio_format;
+	const bool reserved = port_state->used && port_state->handle == next_port;
+	if (reserved) {
 		port_state->audio_handle = audio_handle;
+	}
 	g_audioout2_port_mutex.Unlock();
+	if (!reserved) {
+		audioout2_close_audio_handle(audio_handle);
+		return AUDIO_OUT2_ERROR_INVALID_PARAM;
+	}
+
+	*port = next_port;
 
 	if (next_port <= 16 || (next_port % 600) == 0) {
 		PRINT_NAME();

src/libs/libC.cpp

@@ -633,6 +633,15 @@ LIB_VERSION("LibcInternalExt", 1, "LibcInternal", 1, 1);
 static uint64_t g_mspace_atomic_id_mask = 0;
 static uint64_t g_mstate_table[64]      = {0};
 
+using thread_atexit_destructor_t = KYTY_SYSV_ABI void (*)(void*);
+
+struct ThreadAtexitDestructor {
+	thread_atexit_destructor_t destructor;
+	void*                      object;
+};
+
+static thread_local std::vector<ThreadAtexitDestructor> g_thread_atexit_destructors;
+
 struct Info {
 	uint64_t  size;
 	uint32_t  unknown1;
@@ -650,25 +659,29 @@ void KYTY_SYSV_ABI LibcHeapGetTraceInfo(Info* info) {
 	info->mstate_table          = g_mstate_table;
 }
 
-uint64_t KYTY_SYSV_ABI LibcInternalExtUnknownQBS714Jr3g(uint64_t arg0, uint64_t arg1, uint64_t arg2,
-                                                        uint64_t arg3, uint64_t arg4,
-                                                        uint64_t arg5) {
+int KYTY_SYSV_ABI LibcInternalExtCxaThreadAtexit(thread_atexit_destructor_t destructor, void* object,
+                                                 void* /*module_id*/) {
 	PRINT_NAME();
 
-	LOGF("\t arg0 = 0x%016" PRIx64 "\n"
-	     "\t arg1 = 0x%016" PRIx64 "\n"
-	     "\t arg2 = 0x%016" PRIx64 "\n"
-	     "\t arg3 = 0x%016" PRIx64 "\n"
-	     "\t arg4 = 0x%016" PRIx64 "\n"
-	     "\t arg5 = 0x%016" PRIx64 "\n",
-	     arg0, arg1, arg2, arg3, arg4, arg5);
+	g_thread_atexit_destructors.push_back({destructor, object});
 
 	return 0;
 }
 
+void RunThreadAtexitDestructors() {
+	while (!g_thread_atexit_destructors.empty()) {
+		auto destructor = g_thread_atexit_destructors.back();
+		g_thread_atexit_destructors.pop_back();
+
+		if (destructor.destructor != nullptr) {
+			destructor.destructor(destructor.object);
+		}
+	}
+}
+
 LIB_DEFINE(InitLibcInternalExt_1) {
 	LIB_FUNC("NWtTN10cJzE", LibcInternalExt::LibcHeapGetTraceInfo);
-	LIB_FUNC("qBS714-Jr3g", LibcInternalExt::LibcInternalExtUnknownQBS714Jr3g);
+	LIB_FUNC("qBS714-Jr3g", LibcInternalExt::LibcInternalExtCxaThreadAtexit);
 }
 
 } // namespace LibcInternalExt

src/libs/libGraphicsDriver.cpp

@@ -159,6 +159,7 @@ LIB_DEFINE(InitGraphicsDriver_1) {
 	LIB_FUNC("f3dg2CSgRKY", Gen5::GraphicsCreateShader);
 	LIB_FUNC("dolOmWH+huQ", Gen5::GraphicsUnknownGetFusedShaderSize);
 	LIB_FUNC("fd5Bp5tGTgo", Gen5::GraphicsUnknownFuseShaderHalves);
+	LIB_FUNC("nApJjpKNBl4", Gen5::GraphicsUnknownNApJjpKNBl4);
 	LIB_FUNC("vcmNN+AAXnY", Gen5::GraphicsSetCxRegIndirectPatchSetAddress);
 	LIB_FUNC("Qrj4c+61z4A", Gen5::GraphicsSetShRegIndirectPatchSetAddress);
 	LIB_FUNC("6lNcCp+fxi4", Gen5::GraphicsSetUcRegIndirectPatchSetAddress);
@@ -207,6 +208,7 @@ LIB_DEFINE(InitGraphicsDriver_1) {
 	LIB_FUNC("ewobAQeMo5k", Gen5::GraphicsAcbAcquireMemGetSize);
 	LIB_FUNC("qyM2bxYFPAk", Gen5::GraphicsAcbCondExec);
 	LIB_FUNC("ozKzBP4aki4", Gen5::GraphicsAcbCondExecGetSize);
+	LIB_FUNC("b-oySn+G2tE", Gen5::GraphicsAcbJumpGetSize);
 	LIB_FUNC("htn36gPnBk4", Gen5::GraphicsAcbWaitRegMem);
 	LIB_FUNC("-RnpfpxIhec", Gen5::GraphicsAcbDmaData);
 	LIB_FUNC("qzMN2XKGA4k", Gen5::GraphicsAcbCopyData);
@@ -282,6 +284,7 @@ LIB_DEFINE(InitGraphicsDriver_1) {
 	LIB_FUNC("YWTKOju587o", Gen5::GraphicsCondExecPatchSetCommandAddress);
 	LIB_FUNC("k-JpyR2dYAM", Gen5::GraphicsCondExecPatchSetEnd);
 	LIB_FUNC("3ZWa3AoyWZQ", Gen5::GraphicsCondExecPatchSetCommandAddress);
+	LIB_FUNC("ziVA3whp3p4", Gen5::GraphicsRewindPatchSetRewindState);
 	LIB_FUNC("YUeqkyT7mEQ", Gen5::GraphicsDcbSetFlip);
 }
 

src/libs/libKernel.cpp

@@ -40,7 +40,10 @@
 #define NOMINMAX
 #endif
 #include <windows.h>
-#elif !defined(__APPLE__)
+#elif defined(__APPLE__)
+#include <csignal>
+#include <sys/ucontext.h>
+#else
 #include <csignal>
 #include <ucontext.h>
 #endif
@@ -837,7 +840,7 @@ static void ApplySignalUcontext(CONTEXT* dst_ctx, const SignalUcontext& src_ctx)
 }
 #endif
 
-#if KYTY_PLATFORM != KYTY_PLATFORM_WINDOWS && !defined(__APPLE__) && defined(__x86_64__)
+#if KYTY_PLATFORM != KYTY_PLATFORM_WINDOWS && defined(__x86_64__)
 
 static SignalUcontext CreateSignalUcontextFromHost(const ucontext_t* host_ctx) {
 	SignalUcontext ctx = {};
@@ -845,6 +848,34 @@ static SignalUcontext CreateSignalUcontextFromHost(const ucontext_t* host_ctx) {
 		return ctx;
 	}
 
+#if defined(__APPLE__)
+	const auto& ss = host_ctx->uc_mcontext->__ss;
+
+	ctx.uc_mcontext.mc_rdi    = ss.__rdi;
+	ctx.uc_mcontext.mc_rsi    = ss.__rsi;
+	ctx.uc_mcontext.mc_rdx    = ss.__rdx;
+	ctx.uc_mcontext.mc_rcx    = ss.__rcx;
+	ctx.uc_mcontext.mc_r8     = ss.__r8;
+	ctx.uc_mcontext.mc_r9     = ss.__r9;
+	ctx.uc_mcontext.mc_rax    = ss.__rax;
+	ctx.uc_mcontext.mc_rbx    = ss.__rbx;
+	ctx.uc_mcontext.mc_rbp    = ss.__rbp;
+	ctx.uc_mcontext.mc_r10    = ss.__r10;
+	ctx.uc_mcontext.mc_r11    = ss.__r11;
+	ctx.uc_mcontext.mc_r12    = ss.__r12;
+	ctx.uc_mcontext.mc_r13    = ss.__r13;
+	ctx.uc_mcontext.mc_r14    = ss.__r14;
+	ctx.uc_mcontext.mc_r15    = ss.__r15;
+	ctx.uc_mcontext.mc_rip    = ss.__rip;
+	ctx.uc_mcontext.mc_rsp    = ss.__rsp;
+	ctx.uc_mcontext.mc_rflags = ss.__rflags;
+	ctx.uc_mcontext.mc_cs     = ss.__cs & 0xffffu;
+	ctx.uc_mcontext.mc_gs     = static_cast<uint16_t>(ss.__gs & 0xffffu);
+	ctx.uc_mcontext.mc_fs     = static_cast<uint16_t>(ss.__fs & 0xffffu);
+	ctx.uc_mcontext.mc_len    = sizeof(SignalMcontext);
+
+	return ctx;
+#else
 	const auto* gregs = host_ctx->uc_mcontext.gregs;
 
 	ctx.uc_mcontext.mc_rdi    = static_cast<uint64_t>(gregs[REG_RDI]);
@@ -874,6 +905,7 @@ static SignalUcontext CreateSignalUcontextFromHost(const ucontext_t* host_ctx) {
 	ctx.uc_mcontext.mc_len = sizeof(SignalMcontext);
 
 	return ctx;
+#endif
 }
 
 static void ApplySignalUcontextToHost(ucontext_t* dst_ctx, const SignalUcontext& src_ctx) {
@@ -881,6 +913,29 @@ static void ApplySignalUcontextToHost(ucontext_t* dst_ctx, const SignalUcontext&
 		return;
 	}
 
+#if defined(__APPLE__)
+	auto& ss = dst_ctx->uc_mcontext->__ss;
+
+	ss.__rdi    = src_ctx.uc_mcontext.mc_rdi;
+	ss.__rsi    = src_ctx.uc_mcontext.mc_rsi;
+	ss.__rdx    = src_ctx.uc_mcontext.mc_rdx;
+	ss.__rcx    = src_ctx.uc_mcontext.mc_rcx;
+	ss.__r8     = src_ctx.uc_mcontext.mc_r8;
+	ss.__r9     = src_ctx.uc_mcontext.mc_r9;
+	ss.__rax    = src_ctx.uc_mcontext.mc_rax;
+	ss.__rbx    = src_ctx.uc_mcontext.mc_rbx;
+	ss.__rbp    = src_ctx.uc_mcontext.mc_rbp;
+	ss.__r10    = src_ctx.uc_mcontext.mc_r10;
+	ss.__r11    = src_ctx.uc_mcontext.mc_r11;
+	ss.__r12    = src_ctx.uc_mcontext.mc_r12;
+	ss.__r13    = src_ctx.uc_mcontext.mc_r13;
+	ss.__r14    = src_ctx.uc_mcontext.mc_r14;
+	ss.__r15    = src_ctx.uc_mcontext.mc_r15;
+	ss.__rip    = src_ctx.uc_mcontext.mc_rip;
+	ss.__rsp    = src_ctx.uc_mcontext.mc_rsp;
+	ss.__rflags = src_ctx.uc_mcontext.mc_rflags;
+	// Segment selectors are left untouched; XNU validates them on sigreturn.
+#else
 	auto* gregs = dst_ctx->uc_mcontext.gregs;
 
 	gregs[REG_RDI] = static_cast<greg_t>(src_ctx.uc_mcontext.mc_rdi);
@@ -903,10 +958,17 @@ static void ApplySignalUcontextToHost(ucontext_t* dst_ctx, const SignalUcontext&
 	gregs[REG_EFL] = static_cast<greg_t>(src_ctx.uc_mcontext.mc_rflags);
 
 	// The kernel validates packed segment selectors on sigreturn.
+#endif
 }
 
 static int SignalDispatchHostSignal() {
+#if defined(__APPLE__)
+	// macOS has no realtime signals; SIGUSR1 is otherwise unused on the host side (the
+	// guest's SIGUSR1 is an emulated signal number, not a host registration).
+	static const int host_signal = SIGUSR1;
+#else
 	static const int host_signal = SIGRTMIN + 3;
+#endif
 	return host_signal;
 }
 
@@ -1174,7 +1236,7 @@ static int KYTY_SYSV_ABI KernelRaiseException(Pthread thread, int signum) {
 		}
 		CloseHandle(target_thread);
 		return OK;
-#elif !defined(__APPLE__) && defined(__x86_64__)
+#elif defined(__x86_64__)
 		// Deliver on the target thread.
 		if (thread == PthreadSelfOrNull()) {
 			SignalDispatchScope scope;