loader: emulate immediate SSE4a instructions on macOS (#635)

* loader: emulate immediate SSE4a instructions on macOS

* tests: simplify macOS SSE4a test integration

Keep the Darwin target, CTest registration, and aggregate dependency in
one conditional block without a separate target list. Preserve the
standalone target when BUILD_TESTING is disabled.

Lowercase the test filename and diagnostic prefix, and drop unrelated
formatting changes from the instruction emulator.

Credit Brandon Strong's related SSE4a work in
https://github.com/KytyPS5/KytyPS5/pull/467.

Co-authored-by: Brandon Strong <brandostrong@gmail.com>

---------

Co-authored-by: nmzik <Nmzik@mail.ru>
Co-authored-by: Brandon Strong <brandostrong@gmail.com>

.github/workflows/build.yml

@@ -6,7 +6,12 @@ on:
     branches: [ main, master ]
   pull_request:
 
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
 env:
+  SCCACHE_GHA_ENABLED: "true"
   KYTY_BUILD_ORIGIN: ${{ github.repository != 'KytyPS5/KytyPS5' && 'Fork' || github.event_name == 'push' && 'Official' || 'Source' }}
   KYTY_BUILD_REPOSITORY: ${{ github.repository }}
 
@@ -78,6 +83,9 @@ jobs:
           arch: win64_msvc2022_64
           cache: true
 
+      - name: Setup sccache
+        uses: mozilla-actions/sccache-action@v0.0.11
+
       - name: Verify toolchain
         shell: cmd
         run: |
@@ -100,6 +108,8 @@ jobs:
             -DKYTY_RELEASE_TAG=${{ needs.metadata.outputs.release_tag }} ^
             -DCMAKE_C_COMPILER=clang-cl ^
             -DCMAKE_CXX_COMPILER=clang-cl ^
+            -DCMAKE_C_COMPILER_LAUNCHER=sccache ^
+            -DCMAKE_CXX_COMPILER_LAUNCHER=sccache ^
             -DCMAKE_PREFIX_PATH="%Qt6_DIR%"
 
       - name: Build
@@ -156,6 +166,9 @@ jobs:
           arch: clang_64
           cache: true
 
+      - name: Setup sccache
+        uses: mozilla-actions/sccache-action@v0.0.11
+
       - name: Verify toolchain
         shell: bash
         run: |
@@ -180,6 +193,8 @@ jobs:
             -DCMAKE_OSX_ARCHITECTURES=x86_64 \
             -DCMAKE_C_COMPILER=clang \
             -DCMAKE_CXX_COMPILER=clang++ \
+            -DCMAKE_C_COMPILER_LAUNCHER=sccache \
+            -DCMAKE_CXX_COMPILER_LAUNCHER=sccache \
             -DCMAKE_PREFIX_PATH="$QT_ROOT_DIR"
 
       - name: Build
@@ -304,6 +319,7 @@ jobs:
             libxss-dev \
             lld \
             ninja-build \
+            pkg-config \
             wayland-protocols
 
       - name: Install Qt 6.10.3
@@ -315,6 +331,9 @@ jobs:
           arch: linux_gcc_64
           cache: true
 
+      - name: Setup sccache
+        uses: mozilla-actions/sccache-action@v0.0.11
+
       - name: Verify toolchain
         shell: bash
         run: |
@@ -339,6 +358,8 @@ jobs:
             -DKYTY_RELEASE_TAG="${{ needs.metadata.outputs.release_tag }}" \
             -DCMAKE_C_COMPILER=clang \
             -DCMAKE_CXX_COMPILER=clang++ \
+            -DCMAKE_C_COMPILER_LAUNCHER=sccache \
+            -DCMAKE_CXX_COMPILER_LAUNCHER=sccache \
             -DCMAKE_PREFIX_PATH="$Qt6_DIR" 2>&1 | tee _Build/configure.log
           exit "${PIPESTATUS[0]}"
 

.gitmodules

@@ -22,7 +22,7 @@
 	url = https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator.git
 [submodule "3rdparty/ffmpeg-core"]
 	path = 3rdparty/ffmpeg-core
-	url = https://github.com/shadps4-emu/ext-ffmpeg-core.git
+	url = https://github.com/KytyPS5/ext-ffmpeg-core.git
 	shallow = true
 [submodule "3rdparty/fmt"]
 	path = 3rdparty/fmt

CMakeLists.txt

@@ -126,7 +126,7 @@ if (KYTY_LINKER STREQUAL LD AND NOT LINUX)
 	set(KYTY_LD_OPTIONS "-Wl,--image-base=0x100000000000")
 endif()
 
-project(Kyty${KYTY_PROJECT_NAME}${CMAKE_BUILD_TYPE}${KYTY_COMPILER} VERSION 0.2.2)
+project(Kyty${KYTY_PROJECT_NAME}${CMAKE_BUILD_TYPE}${KYTY_COMPILER} VERSION 0.3.0)
 
 include_directories(
 	${KYTY_THIRD_PARTY_DIR}/gtest/include
@@ -317,24 +317,10 @@ endif()
 
 set(kyty_emulator_link_libraries common Vulkan::Headers spirv-tools-opt spirv-tools SDL2-static imgui xxhash FFmpeg::ffmpeg fmt::fmt nlohmann_json::nlohmann_json LibAtrac9 kyty_git_version Zydis::Zydis xbyak::xbyak)
 
-# Linux system libraries required by the static FFmpeg archive.
+# Linux system libraries.
 if(LINUX)
 	find_package(Threads REQUIRED)
 	list(APPEND kyty_emulator_link_libraries m ${CMAKE_DL_LIBS} Threads::Threads)
-
-	# Optional FFmpeg dependencies.
-	find_package(ZLIB)
-	if(ZLIB_FOUND)
-		list(APPEND kyty_emulator_link_libraries ZLIB::ZLIB)
-	endif()
-	find_library(KYTY_BZ2_LIBRARY bz2)
-	if(KYTY_BZ2_LIBRARY)
-		list(APPEND kyty_emulator_link_libraries ${KYTY_BZ2_LIBRARY})
-	endif()
-	find_library(KYTY_LZMA_LIBRARY lzma)
-	if(KYTY_LZMA_LIBRARY)
-		list(APPEND kyty_emulator_link_libraries ${KYTY_LZMA_LIBRARY})
-	endif()
 endif()
 
 set(inc_headers
@@ -342,6 +328,7 @@ set(inc_headers
 	${KYTY_THIRD_PARTY_DIR}/SDL2/include
 	${KYTY_THIRD_PARTY_DIR}/VulkanMemoryAllocator/include
 	${KYTY_THIRD_PARTY_DIR}/SPIRV-Tools/include
+	${KYTY_THIRD_PARTY_DIR}/SPIRV-Headers/include
 	${KYTY_THIRD_PARTY_DIR}/fmt/include
 	${KYTY_THIRD_PARTY_DIR}/xxhash
 	${KYTY_THIRD_PARTY_DIR}/cpuinfo/include
@@ -431,16 +418,15 @@ add_executable(shader_vertex_metadata_tests EXCLUDE_FROM_ALL
 )
 target_include_directories(shader_vertex_metadata_tests PRIVATE ${inc_headers})
 
-add_executable(shader_stage_runtime_tests EXCLUDE_FROM_ALL
-	"${KYTY_TESTS_DIR}/ShaderStageRuntimeTests.cpp"
+add_executable(resource_materialization_tests EXCLUDE_FROM_ALL
+	"${KYTY_TESTS_DIR}/ResourceMaterializationTests.cpp"
 	"${KYTY_SOURCE_DIR}/graphics/guest_gpu/gpu_format.cpp"
 	"${KYTY_SOURCE_DIR}/graphics/host_gpu/hostMemory.cpp"
-	"${KYTY_SOURCE_DIR}/graphics/shader/shaderStageRuntime.cpp"
 	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/passes/ResourceMaterialization.cpp"
 	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/passes/SrtWalker.cpp"
 )
-target_link_libraries(shader_stage_runtime_tests fmt::fmt)
-target_include_directories(shader_stage_runtime_tests PRIVATE ${inc_headers})
+target_link_libraries(resource_materialization_tests fmt::fmt)
+target_include_directories(resource_materialization_tests PRIVATE ${inc_headers})
 
 add_executable(resource_tracking_tests EXCLUDE_FROM_ALL
 	"${KYTY_TESTS_DIR}/ResourceTrackingTests.cpp"
@@ -542,7 +528,7 @@ if(BUILD_TESTING)
 	add_test(NAME page_manager COMMAND $<TARGET_FILE:page_manager_tests>)
 	add_test(NAME bit_array COMMAND $<TARGET_FILE:bit_array_tests>)
 	add_test(NAME shader_vertex_metadata COMMAND $<TARGET_FILE:shader_vertex_metadata_tests>)
-	add_test(NAME shader_stage_runtime COMMAND $<TARGET_FILE:shader_stage_runtime_tests>)
+	add_test(NAME resource_materialization COMMAND $<TARGET_FILE:resource_materialization_tests>)
 	add_test(NAME resource_tracking COMMAND $<TARGET_FILE:resource_tracking_tests>)
 	add_test(NAME event_queue_lifetime COMMAND $<TARGET_FILE:event_queue_lifetime_tests>)
 	add_test(NAME kernel_file_system COMMAND $<TARGET_FILE:kernel_file_system_tests>)
@@ -551,6 +537,10 @@ if(BUILD_TESTING)
 	add_test(NAME shader_recompiler_compute COMMAND $<TARGET_FILE:shader_recompiler_compute_tests>)
 	add_test(NAME shader_recompiler_alignbyte
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --alignbyte-only)
+	add_test(NAME graphics_pipeline_rasterization
+		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --polygon-mode-only)
+	add_test(NAME graphics_draw_offsets
+		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --draw-offset-only)
 	add_test(NAME virtual_memory_allocation
 		COMMAND $<TARGET_FILE:virtual_memory_allocation_tests>)
 	add_test(NAME guest_red_zone_patcher
@@ -568,12 +558,16 @@ if(BUILD_TESTING)
 
 	add_test(NAME texture_cache_layered_image
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --layered-image-only)
+	add_test(NAME host_image_allocation
+		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --host-image-allocation-only)
 	add_test(NAME texture_cache_image_views
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --image-view-cache-only)
 	add_test(NAME texture_cache_storage_sampled
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --storage-sampled-only)
 	add_test(NAME texture_cache_depth_readback
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --depth-readback-only)
+	add_test(NAME compute_meta_clear_classification
+		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --compute-meta-clear-only)
 	add_test(NAME buffer_cache_dirty_gc
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --buffer-cache-gc-only)
 
@@ -596,7 +590,7 @@ if(BUILD_TESTING)
 		page_manager_tests
 		bit_array_tests
 		shader_vertex_metadata_tests
-		shader_stage_runtime_tests
+		resource_materialization_tests
 		resource_tracking_tests
 		event_queue_lifetime_tests
 		kernel_file_system_tests
@@ -607,6 +601,21 @@ if(BUILD_TESTING)
 	)
 endif()
 
+# Test Darwin signal-context restoration for SSE4a instructions.
+if(APPLE AND (CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" OR
+		(NOT CMAKE_OSX_ARCHITECTURES AND CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|AMD64)$")))
+	add_executable(macos_sse4a_tests EXCLUDE_FROM_ALL
+		"${KYTY_TESTS_DIR}/macosSse4aTests.cpp"
+		"${KYTY_SOURCE_DIR}/loader/x64InstructionEmulator.cpp"
+	)
+	target_link_libraries(macos_sse4a_tests common fmt::fmt Zydis::Zydis)
+	target_include_directories(macos_sse4a_tests PRIVATE ${inc_headers})
+	if(BUILD_TESTING)
+		add_test(NAME macos_sse4a COMMAND $<TARGET_FILE:macos_sse4a_tests>)
+		set_tests_properties(macos_sse4a PROPERTIES TIMEOUT 15)
+		add_dependencies(kyty_tests macos_sse4a_tests)
+	endif()
+endif()
 
 add_executable(kyty_emulator "${KYTY_SOURCE_DIR}/main.cpp" ${kyty_emulator_src})
 configure_macos_guest_address_space(kyty_emulator)

README.md

@@ -90,8 +90,14 @@ starting a large change.
 
 Set up the clang-format hook after cloning:
 
-```powershell
-python -m pip install pre-commit
+Install `pre-commit` using the method appropriate for your platform:
+
+- **Arch Linux / CachyOS:** `sudo pacman -S pre-commit`
+- **Other Linux / macOS / Windows:** `python -m pip install pre-commit`
+
+Then install the Git hook:
+
+```bash
 python -m pre_commit install --install-hooks
 ```
 
@@ -165,7 +171,7 @@ build has no working sound and no gamepad hotplug:
 
 ```bash
 sudo apt-get install --no-install-recommends \
-  clang lld ninja-build cmake git glslang-tools \
+  clang lld ninja-build cmake git glslang-tools pkg-config \
   libgl1-mesa-dev libx11-dev libxcursor-dev libxext-dev libxfixes-dev \
   libxi-dev libxrandr-dev libxss-dev libxkbcommon-dev \
   libasound2-dev libpulse-dev libudev-dev libdbus-1-dev libwayland-dev wayland-protocols
@@ -186,13 +192,37 @@ cmake --install _Build/linux --prefix _Build/linux/install
 ```
 
 The install step copies the Qt libraries and plugins next to the binaries, so
-`_Build/linux/install` runs without a matching system Qt.
+`_Build/linux/install` runs without a matching system Qt. FFmpeg is linked statically
+from the pinned [KytyPS5 FFmpeg core](https://github.com/KytyPS5/ext-ffmpeg-core)
+release, including VP9 and WebM support. System FFmpeg packages are not required.
 
 As on Windows, the MSVC compiler is not used; Clang is required. `cl.exe` is rejected at configure
 time.
 
 The CMake source root is the repository root.
 
+### Building on NixOS
+
+A development shell provides Clang, CMake, Ninja, Qt 6, the Vulkan headers, and the SDL2 backend
+libraries. Enter it and configure exactly as on other Linux distributions; the shell exports
+`CMAKE_PREFIX_PATH` and `QT_PLUGIN_PATH`, so the `-DCMAKE_PREFIX_PATH="$Qt6_DIR"` argument is not
+needed:
+
+```bash
+nix-shell # or: nix develop
+git submodule update --init --recursive
+
+cmake -S . -B _Build/linux -G Ninja -DCMAKE_BUILD_TYPE=Release \
+  -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++
+
+cmake --build _Build/linux --target launcher --parallel
+cmake --install _Build/linux --prefix _Build/linux/install
+```
+
+The configure step downloads the FFmpeg prebuilts and the `xbyak`/`zydis` sources, so it needs
+network access; a fully sandboxed `nix build` would require vendoring those inputs. A Vulkan 1.3
+driver must be available at runtime (on NixOS, `hardware.graphics.enable = true`).
+
 ### Building on macOS
 
 macOS builds target x86-64 and run under Rosetta 2 on Apple Silicon, so the PS5's x86-64 game

flake.lock

@@ -0,0 +1,27 @@
+{
+  "nodes": {
+    "nixpkgs": {
+      "locked": {
+        "lastModified": 1789426278,
+        "narHash": "sha256-GnotcvKtbTMnoVx6G4E7ZdUX6487Dky74dqq3pU7iRk=",
+        "owner": "NixOS",
+        "repo": "nixpkgs",
+        "rev": "efe6f071ede9d21c37462d2d6682d5e670099684",
+        "type": "github"
+      },
+      "original": {
+        "owner": "NixOS",
+        "ref": "nixos-unstable",
+        "repo": "nixpkgs",
+        "type": "github"
+      }
+    },
+    "root": {
+      "inputs": {
+        "nixpkgs": "nixpkgs"
+      }
+    }
+  },
+  "root": "root",
+  "version": 7
+}

flake.nix

@@ -0,0 +1,20 @@
+{
+  description = "KytyPS5 development shell";
+
+  inputs.nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
+
+  outputs = { self, nixpkgs }:
+    let
+      systems = [ "x86_64-linux" ];
+      forAllSystems = nixpkgs.lib.genAttrs systems;
+    in
+    {
+      devShells = forAllSystems (system:
+        let
+          pkgs = import nixpkgs { inherit system; };
+        in
+        {
+          default = import ./nix/devshell.nix { inherit pkgs; };
+        });
+    };
+}

nix/devshell.nix

@@ -0,0 +1,84 @@
+{ pkgs }:
+
+let
+  # Build-time dependencies. Every -dev package the bundled SDL2 probes for must
+  # be listed here, otherwise SDL2 quietly drops the audio/Wayland/udev backends.
+  buildDeps = with pkgs; [
+    qt6.qtbase # Concurrent, Network, Widgets (launcher)
+    vulkan-headers
+    vulkan-loader
+    vulkan-tools
+    vulkan-validation-layers
+    libglvnd
+    # X11 / input
+    libX11
+    libXcursor
+    libXext
+    libXfixes
+    libXi
+    libXrandr
+    libXScrnSaver
+    libxkbcommon
+    # Wayland
+    wayland
+    wayland-protocols
+    libdecor
+    # audio + device/dbus
+    alsa-lib
+    libpulseaudio
+    systemd # libudev
+    dbus
+  ];
+
+  # Libraries the built binaries need at runtime.
+  runtimeLibs = with pkgs; [
+    stdenv.cc.cc.lib # libstdc++ / libgcc_s
+    libgcc.lib
+    qt6.qtbase # libQt6Concurrent/Network/Widgets
+    vulkan-loader
+    libglvnd
+    alsa-lib
+    libpulseaudio
+    libX11
+    libXcursor
+    libXext
+    libXfixes
+    libXi
+    libXrandr
+    libXScrnSaver
+    libxkbcommon
+    wayland
+    libdecor
+    systemd
+    dbus
+  ];
+in
+pkgs.mkShell {
+  name = "kytyps5";
+
+  nativeBuildInputs = with pkgs; [
+    clang
+    lld
+    cmake
+    ninja
+    pkg-config
+    git
+    glslang # provides glslangValidator for the bundled shaders
+  ];
+
+  buildInputs = buildDeps;
+
+  shellHook = ''
+    export CMAKE_PREFIX_PATH="${pkgs.qt6.qtbase}:''${CMAKE_PREFIX_PATH:-}"
+    export QT_PLUGIN_PATH="${pkgs.qt6.qtbase}/lib/qt-6/plugins:${pkgs.qt6.qtbase}/lib/qt6/plugins:''${QT_PLUGIN_PATH:-}"
+    export LD_LIBRARY_PATH="${pkgs.lib.makeLibraryPath runtimeLibs}:''${LD_LIBRARY_PATH:-}"
+
+    if [ -z "''${KYTY_SHELL_QUIET:-}" ]; then
+      echo "KytyPS5 dev shell"
+      echo "  cmake -S . -B _Build/linux -G Ninja -DCMAKE_BUILD_TYPE=Release \\"
+      echo "    -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++"
+      echo "  cmake --build _Build/linux --target launcher --parallel"
+      echo "  cmake --install _Build/linux --prefix _Build/linux/install"
+    fi
+  '';
+}

shell.nix

@@ -0,0 +1,4 @@
+let
+  pkgs = import <nixpkgs> { };
+in
+import ./nix/devshell.nix { inherit pkgs; }

src/common/alignment.h

@@ -0,0 +1,25 @@
+// SPDX-FileCopyrightText: 2014 Jannik Vogel <email@jannikvogel.de>
+// SPDX-License-Identifier: CC0-1.0
+
+#pragma once
+
+#include <cstddef>
+#include <type_traits>
+
+namespace Common {
+
+template <typename T>
+[[nodiscard]] constexpr T AlignUp(T value, std::size_t size) {
+	static_assert(std::is_unsigned_v<T>, "T must be an unsigned value.");
+	auto mod{static_cast<T>(value % size)};
+	value -= mod;
+	return static_cast<T>(mod == T{0} ? value : value + size);
+}
+
+template <typename T>
+[[nodiscard]] constexpr T AlignDown(T value, std::size_t size) {
+	static_assert(std::is_unsigned_v<T>, "T must be an unsigned value.");
+	return static_cast<T>(value - value % size);
+}
+
+} // namespace Common

src/common/assert.cpp

@@ -1,6 +1,5 @@
 #include "common/assert.h"
 
-#include "common/debug.h"
 #include "common/logging/log.h"
 #include "common/subsystems.h"
 #include "kytyGitVersion.h"
@@ -12,25 +11,10 @@
 
 namespace Common {
 
-#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS && KYTY_BUILD == KYTY_BUILD_DEBUG &&                    \
-    KYTY_COMPILER == KYTY_COMPILER_CLANG
-constexpr int PRINT_STACK_FROM = 4;
-#else
-constexpr int PRINT_STACK_FROM = 2;
-#endif
-
 static std::string BuildFatalReport(const char* title, std::string_view text, const char* file,
                                     int line) {
-	DebugStack stack;
-	DebugStack::Trace(&stack);
-
-	std::string report = "--- Build ---\n" KYTY_BUILD_LABEL "\n--- Stack Trace ---\n";
-	for (int i = PRINT_STACK_FROM; i < stack.depth; i++) {
-		report += fmt::format("[{}] {:016x}\n", i - PRINT_STACK_FROM,
-		                      static_cast<uint64_t>(stack.GetAddr(i)));
-	}
-	report += fmt::format("{}\n{} in {}:{}\n", title, text, file, line);
-	return report;
+	return fmt::format("--- Build ---\n{}\n{}\n{} in {}:{}\n", KYTY_BUILD_LABEL, title, text, file,
+	                   line);
 }
 
 static int DbgReport(const char* title, std::string_view text, const char* file, int line) {
@@ -59,6 +43,7 @@ int DbgExitHandler(const char* file, int line, fmt::text_style style, std::strin
 }
 
 void DbgExit(int status) {
+	Subsystems::EmergencyShutdownActive();
 	std::fflush(nullptr);
 	std::_Exit(status);
 }

src/common/assert.h

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

src/common/debug.cpp

@@ -1,20 +1,7 @@
 #include "common/debug.h"
 
-#include "common/platform/sysDbg.h"
-
 namespace Common {
 
-void DebugStack::Print(int from, [[maybe_unused]] bool with_name) const {
-	for (int i = from; i < depth; i++) {
-		printf("[%d] %016" PRIx64 "\n", i - from, static_cast<uint64_t>(GetAddr(i)));
-	}
-}
-
-void DebugStack::Trace(DebugStack* stack) {
-	stack->depth = DEBUG_MAX_STACK_DEPTH;
-	SysStackWalk(stack->stack.data(), &stack->depth);
-}
-
 std::string Debug::GetCompiler() {
 #if KYTY_COMPILER == KYTY_COMPILER_CLANG
 	return "clang";

src/common/debug.h

@@ -2,42 +2,17 @@
 #define KYTY_COMMON_DEBUG_H_
 
 #include "common/common.h"
-#include "common/stringUtils.h"
 
-#include <array>
-#include <cstring>
+#include <string>
 
 namespace Common {
 
-#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS && KYTY_BUILD == KYTY_BUILD_DEBUG &&                    \
-    KYTY_COMPILER == KYTY_COMPILER_CLANG
-constexpr int DEBUG_MAX_STACK_DEPTH = 20;
-#else
-constexpr int DEBUG_MAX_STACK_DEPTH = 15;
-#endif
-
 namespace Debug {
 std::string GetCompiler();
 std::string GetLinker();
 std::string GetBitness();
 } // namespace Debug
 
-struct DebugStack {
-	int                                      depth {0};
-	std::array<void*, DEBUG_MAX_STACK_DEPTH> stack {};
-
-	[[nodiscard]] uintptr_t GetAddr(int i) const { return reinterpret_cast<uintptr_t>(stack[i]); }
-
-	void Print(int from, bool with_name = true) const;
-
-	void CopyTo(DebugStack* s) const {
-		std::memcpy(s->stack.data(), stack.data(), sizeof(void*) * depth);
-		s->depth = depth;
-	}
-
-	static void Trace(DebugStack* stack);
-};
-
 } // namespace Common
 
 #endif /* KYTY_COMMON_DEBUG_H_ */

src/common/emulatorConfig.cpp

@@ -43,14 +43,30 @@ int32_t GetUserId() {
 	return g_config->user_id;
 }
 
+const std::string& GetAudioInputDevice() {
+	return g_config->audio_input_device;
+}
+
 PresentMode GetPresentMode() {
 	return g_config->present_mode;
 }
 
+int32_t GetGpuIndex() {
+	return g_config->gpu_index;
+}
+
 bool FullscreenEnabled() {
 	return g_config->fullscreen_enabled;
 }
 
+bool VrEnabled() {
+	return g_config->vr_enabled;
+}
+
+bool AmdCpuEnabled() {
+	return g_config->amd_cpu_enabled;
+}
+
 uint32_t GetVblankFrequency() {
 	return std::clamp(g_config->vblank_frequency, 30u, 360u);
 }
@@ -71,7 +87,7 @@ ShaderOptimizationType GetShaderOptimizationType() {
 	return g_config->shader_optimization_type;
 }
 
-ShaderLogDirection GetShaderLogDirection() {
+LogDirection GetShaderLogDirection() {
 	return g_config->shader_log_direction;
 }
 
@@ -91,7 +107,7 @@ bool GraphicsDebugDumpEnabled() {
 	return g_config->graphics_debug_dump_enabled;
 }
 
-OutputDirection GetPrintfDirection() {
+LogDirection GetPrintfDirection() {
 	return g_config->printf_direction;
 }
 
@@ -99,8 +115,8 @@ std::filesystem::path GetPrintfOutputFile() {
 	return g_config->printf_output_file;
 }
 
-ProfilerDirection GetProfilerDirection() {
-	return g_config->profiler_direction;
+bool ProfilerEnabled() {
+	return g_config->profiler_enabled;
 }
 
 bool SpirvDebugPrintfEnabled() {

src/common/emulatorConfig.h

@@ -21,11 +21,7 @@ struct Lifecycle {
 
 enum class ShaderOptimizationType { None, Size, Performance };
 
-enum class ShaderLogDirection { Silent, Console, File };
-
-enum class ProfilerDirection { None, Network };
-
-enum class OutputDirection { Silent, Console, File };
+enum class LogDirection { Silent, Console, File };
 
 enum class PresentMode { Fifo, Mailbox, Immediate };
 
@@ -47,21 +43,25 @@ struct ConfigOptions {
 	uint32_t               screen_height               = 720;
 	std::string            user_name                   = "Kyty";
 	int32_t                user_id                     = DEFAULT_USER_ID;
-	PresentMode            present_mode                = PresentMode::Fifo;
+	std::string            audio_input_device;
+	PresentMode            present_mode                = PresentMode::Mailbox;
+	int32_t                gpu_index                   = -1;
 	bool                   fullscreen_enabled          = false;
+	bool                   vr_enabled                  = false;
+	bool                   amd_cpu_enabled             = false;
 	uint32_t               vblank_frequency            = 60;
 	uint32_t               console_language            = DEFAULT_CONSOLE_LANGUAGE;
 	bool                   vulkan_validation_enabled   = false;
 	bool                   shader_validation_enabled   = false;
 	ShaderOptimizationType shader_optimization_type    = ShaderOptimizationType::None;
-	ShaderLogDirection     shader_log_direction        = ShaderLogDirection::Silent;
+	LogDirection           shader_log_direction        = LogDirection::Silent;
 	std::filesystem::path  shader_log_folder           = "_Shaders";
 	bool                   command_buffer_dump_enabled = false;
 	std::filesystem::path  command_buffer_dump_folder  = "_Buffers";
 	bool                   graphics_debug_dump_enabled = false;
-	OutputDirection        printf_direction            = OutputDirection::Silent;
+	LogDirection           printf_direction            = LogDirection::Silent;
 	std::filesystem::path  printf_output_file          = "_kyty.txt";
-	ProfilerDirection      profiler_direction          = ProfilerDirection::None;
+	bool                   profiler_enabled            = false;
 	bool                   spirv_debug_printf_enabled  = false;
 	bool                   gpu_assisted_validation_enabled = false;
 	bool                   renderdoc_enabled           = false;
@@ -79,15 +79,19 @@ uint32_t GetScreenWidth();
 uint32_t GetScreenHeight();
 const std::string& GetUserName();
 int32_t  GetUserId();
+const std::string& GetAudioInputDevice();
 PresentMode GetPresentMode();
+int32_t GetGpuIndex();
 bool     FullscreenEnabled();
+bool     VrEnabled();
+bool     AmdCpuEnabled();
 uint32_t GetVblankFrequency();
 uint32_t GetConsoleLanguage();
 bool     VulkanValidationEnabled();
 
 bool                   ShaderValidationEnabled();
 ShaderOptimizationType GetShaderOptimizationType();
-ShaderLogDirection     GetShaderLogDirection();
+LogDirection           GetShaderLogDirection();
 std::filesystem::path  GetShaderLogFolder();
 
 bool                  CommandBufferDumpEnabled();
@@ -95,10 +99,10 @@ std::filesystem::path GetCommandBufferDumpFolder();
 
 bool GraphicsDebugDumpEnabled();
 
-OutputDirection       GetPrintfDirection();
+LogDirection          GetPrintfDirection();
 std::filesystem::path GetPrintfOutputFile();
 
-ProfilerDirection GetProfilerDirection();
+bool ProfilerEnabled();
 
 bool SpirvDebugPrintfEnabled();
 

src/common/file.cpp

@@ -69,7 +69,7 @@ File::File(): m_p(std::make_unique<FilePrivate>()) {
 }
 
 File::~File() {
-	EXIT_IF(m_p->f != nullptr);
+	Close();
 }
 
 File::File(const std::filesystem::path& name): m_p(std::make_unique<FilePrivate>()) {

src/common/hostException.cpp

@@ -5,14 +5,17 @@
 
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
 #include <windows.h> // IWYU pragma: keep
-#elif defined(__APPLE__)
+#else
+#include <algorithm>
 #include <csignal>
+#include <cstdlib>
+#include <initializer_list>
+#include <unistd.h>
+#if defined(__APPLE__)
 #include <sys/ucontext.h>
 #else
-#include <csignal>
-#include <initializer_list>
 #include <ucontext.h> // IWYU pragma: keep
-#include <unistd.h>
+#endif
 #endif
 
 // IWYU pragma: no_include <errhandlingapi.h>
@@ -32,6 +35,51 @@ static_assert(decltype(g_handler)::is_always_lock_free);
 static_assert(decltype(g_install_state)::is_always_lock_free);
 #endif
 
+#if KYTY_PLATFORM == KYTY_PLATFORM_LINUX
+
+// macOS uses the same POSIX platform setting and needs a signal stack too.
+class ThreadSignalStack {
+public:
+	ThreadSignalStack() {
+		const auto page_size = static_cast<size_t>(::getpagesize());
+		const auto stack_size =
+		    (std::max<size_t>(64 * 1024, MINSIGSTKSZ) + page_size - 1) & ~(page_size - 1);
+		if (::posix_memalign(&m_memory, page_size, stack_size) != 0) {
+			return;
+		}
+
+		stack_t stack {};
+		stack.ss_sp   = m_memory;
+		stack.ss_size = stack_size;
+		if (::sigaltstack(&stack, &m_previous) != 0) {
+			std::free(m_memory);
+			m_memory = nullptr;
+		}
+	}
+
+	~ThreadSignalStack() {
+		if (m_memory != nullptr && ::sigaltstack(&m_previous, nullptr) == 0) {
+			std::free(m_memory);
+		}
+	}
+
+	[[nodiscard]] bool IsInitialized() const { return m_memory != nullptr; }
+
+	KYTY_CLASS_NO_COPY(ThreadSignalStack)
+
+private:
+	void*   m_memory = nullptr;
+	stack_t m_previous {};
+};
+
+bool InitializeThreadSignalStack() {
+	// Keep fault handling off guest stacks, which GPU tracking can make read-only.
+	thread_local ThreadSignalStack signal_stack;
+	return signal_stack.IsInitialized();
+}
+
+#endif
+
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
 
 static LONG WINAPI ExceptionFilter(PEXCEPTION_POINTERS exception) noexcept {
@@ -253,7 +301,7 @@ bool InstallHandler(Handler handler) {
 #elif defined(__APPLE__)
 	struct sigaction sa {};
 	sa.sa_sigaction = SignalHandler;
-	sa.sa_flags     = SA_SIGINFO;
+	sa.sa_flags     = SA_SIGINFO | SA_ONSTACK;
 	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
@@ -274,8 +322,7 @@ bool InstallHandler(Handler handler) {
 	struct sigaction action {};
 	action.sa_sigaction = SignalHandler;
 	sigemptyset(&action.sa_mask);
-	// Fault resolution needs the normal thread stack.
-	action.sa_flags = SA_SIGINFO | SA_RESTART;
+	action.sa_flags = SA_SIGINFO | SA_RESTART | SA_ONSTACK;
 
 	for (const int signal_number: {SIGSEGV, SIGBUS, SIGILL}) {
 		if (::sigaction(signal_number, &action, nullptr) != 0) {

src/common/hostException.h

@@ -39,6 +39,10 @@ using Handler = bool (*)(const ExceptionInfo&);
 
 bool InstallHandler(Handler handler);
 
+#if KYTY_PLATFORM == KYTY_PLATFORM_LINUX
+bool InitializeThreadSignalStack();
+#endif
+
 } // namespace Common::HostException
 
 #endif /* KYTY_COMMON_HOST_EXCEPTION_H_ */

src/common/logging/log.cpp

@@ -129,6 +129,14 @@ static void WriteImpl(std::string_view text, fmt::text_style style = {}) {
 	}
 }
 
+void WriteToConsoleAndLog(std::string_view text) {
+	WriteImpl(text);
+	if (g_initialized && g_direction != Direction::Console) {
+		WriteStdout(text);
+	}
+	Flush();
+}
+
 void WriteFatal(std::string_view text) {
 	if (g_direction == Direction::Silent || !g_initialized) {
 		WriteStdout(text);
@@ -152,9 +160,9 @@ void WriteFatal(fmt::text_style style, std::string_view text) {
 void Initialize() {
 	g_initialized = true;
 	switch (Config::GetPrintfDirection()) {
-		case Config::OutputDirection::Silent: g_direction = Direction::Silent; break;
-		case Config::OutputDirection::Console: g_direction = Direction::Console; break;
-		case Config::OutputDirection::File: g_direction = Direction::File; break;
+		case Config::LogDirection::Silent: g_direction = Direction::Silent; break;
+		case Config::LogDirection::Console: g_direction = Direction::Console; break;
+		case Config::LogDirection::File: g_direction = Direction::File; break;
 	}
 	g_output_file =
 	    (g_direction == Direction::File ? Config::GetPrintfOutputFile() : std::filesystem::path {});

src/common/logging/log.h

@@ -25,6 +25,7 @@ Direction GetDirection();
 bool      IsSilent();
 void      Write(std::string_view text);
 void      Write(fmt::text_style style, std::string_view text);
+void      WriteToConsoleAndLog(std::string_view text);
 void      WriteFatal(std::string_view text);
 void      WriteFatal(fmt::text_style style, std::string_view text);
 void      Flush();

src/common/platform/sysDbg.h

@@ -30,8 +30,6 @@ struct sys_dbg_stack_info_t {
 #endif
 };
 
-void SysStackWalk(void** stack, int* depth);
-void SysStackUsage(sys_dbg_stack_info_t& s);          // NOLINT(google-runtime-references)
-void SysStackUsagePrint(sys_dbg_stack_info_t& stack); // NOLINT(google-runtime-references)
+void SysStackUsage(sys_dbg_stack_info_t& s); // NOLINT(google-runtime-references)
 
 #endif /* KYTY_COMMON_PLATFORM_SYSDBG_H_ */

src/common/platform/sysLinuxDbg.cpp

@@ -8,7 +8,6 @@
 
 #include <cstdlib>
 #include <cstring>
-#include <execinfo.h>
 #include <pthread.h>
 #include <sys/param.h>
 #include <sys/types.h>
@@ -17,56 +16,6 @@
 #include <libgen.h> // POSIX basename() lives here on macOS, not in <cstring>
 #endif
 
-// Avoid unwinding a guest-owned stack.
-static bool OnOwnStack() {
-	const char* probe = reinterpret_cast<const char*>(&probe);
-
-	pthread_attr_t attr {};
-#if defined(__APPLE__)
-	const auto* top  = static_cast<const char*>(pthread_get_stackaddr_np(pthread_self()));
-	const auto  size = pthread_get_stacksize_np(pthread_self());
-	(void)attr;
-	return top != nullptr && size != 0 && probe < top && probe >= top - size;
-#else
-	if (pthread_getattr_np(pthread_self(), &attr) != 0) {
-		return false;
-	}
-	void*      base = nullptr;
-	size_t     size = 0;
-	const bool ok = pthread_attr_getstack(&attr, &base, &size) == 0 && base != nullptr && size != 0;
-	pthread_attr_destroy(&attr);
-	if (!ok) {
-		return false;
-	}
-	const auto* low = static_cast<const char*>(base);
-	return probe >= low && probe < low + size;
-#endif
-}
-
-void SysStackWalk(void** stack, int* depth) {
-	if (stack == nullptr || depth == nullptr || *depth <= 0) {
-		if (depth != nullptr) {
-			*depth = 0;
-		}
-		return;
-	}
-
-	if (!OnOwnStack()) {
-		*depth = 0;
-		return;
-	}
-
-	const int n = ::backtrace(stack, *depth);
-	*depth      = (n < 0 ? 0 : n);
-}
-
-void SysStackUsagePrint(sys_dbg_stack_info_t& stack) {
-	printf("stack: (0x%" PRIx64 ", %" PRIu64 ")\n", static_cast<uint64_t>(stack.commited_addr),
-	       static_cast<uint64_t>(stack.commited_size));
-	printf("code: (0x%" PRIx64 ", %" PRIu64 ")\n", static_cast<uint64_t>(stack.code_addr),
-	       static_cast<uint64_t>(stack.code_size));
-}
-
 void SysStackUsage(sys_dbg_stack_info_t& s) {
 	pid_t pid = getpid();
 

src/common/platform/sysLinuxVirtual.cpp

@@ -31,8 +31,7 @@
 namespace Common {
 
 static pthread_mutex_t              g_virtual_mutex {};
-static std::map<uintptr_t, size_t>* g_allocs   = nullptr;
-static std::map<uintptr_t, int>*    g_protects = nullptr;
+static std::map<uintptr_t, size_t>* g_allocs = nullptr;
 
 void SysVirtualInit() {
 	pthread_mutexattr_t attr {};
@@ -46,8 +45,7 @@ void SysVirtualInit() {
 	pthread_mutex_init(&g_virtual_mutex, &attr);
 	pthread_mutexattr_destroy(&attr);
 
-	g_allocs   = new std::map<uintptr_t, size_t>;
-	g_protects = new std::map<uintptr_t, int>;
+	g_allocs = new std::map<uintptr_t, size_t>;
 }
 
 static int get_protection_flag(VirtualMemory::Mode mode) {
@@ -68,21 +66,6 @@ static int get_protection_flag(VirtualMemory::Mode mode) {
 	return protect;
 }
 
-static VirtualMemory::Mode get_protection_flag(int mode) {
-	switch (mode) {
-		case PROT_NONE: return VirtualMemory::Mode::NoAccess;
-		case PROT_READ: return VirtualMemory::Mode::Read;
-		case PROT_WRITE: return VirtualMemory::Mode::Write;
-		case PROT_READ | PROT_WRITE: return VirtualMemory::Mode::ReadWrite; // NOLINT
-		case PROT_EXEC: return VirtualMemory::Mode::Execute;
-		case PROT_EXEC | PROT_WRITE: return VirtualMemory::Mode::ExecuteWrite; // NOLINT
-		case PROT_EXEC | PROT_READ: return VirtualMemory::Mode::ExecuteRead;   // NOLINT
-		case PROT_EXEC | PROT_WRITE | PROT_READ:
-			return VirtualMemory::Mode::ExecuteReadWrite; // NOLINT
-		default: return VirtualMemory::Mode::NoAccess;
-	}
-}
-
 // Keep automatic mappings inside the guest and GPU-addressable low window.
 #ifdef KYTY_FIXED_NOREPLACE
 static constexpr uintptr_t LOW_ARENA_LIMIT = 0x000000FC00000000ULL; // libc mspace window ceiling
@@ -161,11 +144,6 @@ uint64_t SysVirtualAlloc(uint64_t address, uint64_t size, VirtualMemory::Mode mo
 	if (ptr != MAP_FAILED) {
 		pthread_mutex_lock(&g_virtual_mutex);
 		record_alloc(ret_addr, size);
-		uintptr_t page_start = ret_addr >> 12u;
-		uintptr_t page_end   = (ret_addr + size - 1) >> 12u;
-		for (uintptr_t page = page_start; page <= page_end; page++) {
-			(*g_protects)[page] = protect;
-		}
 		pthread_mutex_unlock(&g_virtual_mutex);
 	}
 
@@ -251,11 +229,6 @@ uint64_t SysVirtualAllocAligned(uint64_t address, uint64_t size, VirtualMemory::
 
 	pthread_mutex_lock(&g_virtual_mutex);
 	record_alloc(ret_addr, size);
-	uintptr_t page_start = ret_addr >> 12u;
-	uintptr_t page_end   = (ret_addr + size - 1) >> 12u;
-	for (uintptr_t page = page_start; page <= page_end; page++) {
-		(*g_protects)[page] = protect;
-	}
 	pthread_mutex_unlock(&g_virtual_mutex);
 
 	return ret_addr;
@@ -337,11 +310,6 @@ bool SysVirtualAllocFixed(uint64_t address, uint64_t size, VirtualMemory::Mode m
 	if (ptr != MAP_FAILED) {
 		pthread_mutex_lock(&g_virtual_mutex);
 		record_alloc(ret_addr, size);
-		uintptr_t page_start = ret_addr >> 12u;
-		uintptr_t page_end   = (ret_addr + size - 1) >> 12u;
-		for (uintptr_t page = page_start; page <= page_end; page++) {
-			(*g_protects)[page] = protect;
-		}
 		pthread_mutex_unlock(&g_virtual_mutex);
 
 		return true;
@@ -516,18 +484,7 @@ bool SysVirtualFree(uint64_t address) {
 		return false;
 	}
 
-	if (munmap(reinterpret_cast<void*>(addr), size) == 0) {
-		uintptr_t page_start = addr >> 12u;
-		uintptr_t page_end   = (addr + size - 1) >> 12u;
-		pthread_mutex_lock(&g_virtual_mutex);
-		for (uintptr_t page = page_start; page <= page_end; page++) {
-			g_protects->erase(page);
-		}
-		pthread_mutex_unlock(&g_virtual_mutex);
-		return true;
-	}
-
-	return false;
+	return munmap(reinterpret_cast<void*>(addr), size) == 0;
 }
 
 bool SysVirtualFreeRange(uint64_t address, uint64_t size) {
@@ -582,40 +539,16 @@ bool SysVirtualFreeRange(uint64_t address, uint64_t size) {
 	if (end < alloc_end) {
 		(*g_allocs)[end] = alloc_end - end;
 	}
-	for (uintptr_t page = addr >> 12u; page <= (end - 1u) >> 12u; page++) {
-		g_protects->erase(page);
-	}
 	pthread_mutex_unlock(&g_virtual_mutex);
 	return true;
 }
 
-bool SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode,
-                       VirtualMemory::Mode* old_mode) {
-	auto addr = static_cast<uintptr_t>(address);
-
-	pthread_mutex_lock(&g_virtual_mutex);
-	if (old_mode != nullptr) {
-		if (auto s = g_protects->find(addr >> 12u); s != g_protects->end()) {
-			*old_mode = get_protection_flag(s->second);
-		} else {
-			*old_mode = VirtualMemory::Mode::NoAccess;
-		}
-	}
-	pthread_mutex_unlock(&g_virtual_mutex);
-
-	uintptr_t page_start = addr >> 12u;
-	uintptr_t page_end   = (addr + size - 1) >> 12u;
-	if (mprotect(reinterpret_cast<void*>(page_start << 12u), (page_end - page_start + 1) << 12u,
-	             get_protection_flag(mode)) == 0) {
-		pthread_mutex_lock(&g_virtual_mutex);
-		for (uintptr_t page = page_start; page <= page_end; page++) {
-			(*g_protects)[page] = get_protection_flag(mode);
-		}
-		pthread_mutex_unlock(&g_virtual_mutex);
-		return true;
-	}
-
-	return false;
+bool SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
+	const auto addr       = static_cast<uintptr_t>(address);
+	const auto page_start = addr >> 12u;
+	const auto page_end   = (addr + size - 1) >> 12u;
+	return mprotect(reinterpret_cast<void*>(page_start << 12u), (page_end - page_start + 1) << 12u,
+	                get_protection_flag(mode)) == 0;
 }
 
 bool SysVirtualFlushInstructionCache(uint64_t /*address*/, uint64_t /*size*/) {

src/common/platform/sysVirtual.h

@@ -19,8 +19,7 @@ bool     SysVirtualReserveFixed(uint64_t address, uint64_t size);
 bool     SysVirtualDecommit(uint64_t address, uint64_t size);
 bool     SysVirtualFree(uint64_t address);
 bool     SysVirtualFreeRange(uint64_t address, uint64_t size);
-bool     SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode,
-                           VirtualMemory::Mode* old_mode = nullptr);
+bool     SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode);
 bool     SysVirtualFlushInstructionCache(uint64_t address, uint64_t size);
 
 } // namespace Common

src/common/platform/sysWindowsDbg.cpp

@@ -15,18 +15,6 @@
 #include "common/assert.h"
 #include "common/platform/sysDbg.h"
 
-void SysStackWalk(void** stack, int* depth) {
-	*depth = static_cast<int>(CaptureStackBackTrace(0, static_cast<DWORD>(*depth), stack, nullptr));
-}
-
-void SysStackUsagePrint(sys_dbg_stack_info_t& stack) {
-	printf("stack: (0x%" PRIx64 ", %" PRIu64 ") + (0x%" PRIx64 ", %" PRIu64 ") + (0x%" PRIx64
-	       ", %" PRIu64 ")\n",
-	       static_cast<uint64_t>(stack.reserved_addr), static_cast<uint64_t>(stack.reserved_size),
-	       static_cast<uint64_t>(stack.guard_addr), static_cast<uint64_t>(stack.guard_size),
-	       static_cast<uint64_t>(stack.commited_addr), static_cast<uint64_t>(stack.commited_size));
-}
-
 void SysStackUsage(sys_dbg_stack_info_t& s) {
 	MEMORY_BASIC_INFORMATION mbi {};
 	[[maybe_unused]] size_t  ss = VirtualQuery(&mbi, &mbi, sizeof(mbi));

src/common/platform/sysWindowsVirtual.cpp

@@ -45,18 +45,6 @@ static DWORD GetProtectionFlag(VirtualMemory::Mode mode) {
 	return protect;
 }
 
-static VirtualMemory::Mode GetProtectionFlag(DWORD mode) {
-	switch (mode) {
-		case PAGE_NOACCESS: return VirtualMemory::Mode::NoAccess;
-		case PAGE_READONLY: return VirtualMemory::Mode::Read;
-		case PAGE_READWRITE: return VirtualMemory::Mode::ReadWrite;
-		case PAGE_EXECUTE: return VirtualMemory::Mode::Execute;
-		case PAGE_EXECUTE_READ: return VirtualMemory::Mode::ExecuteRead;
-		case PAGE_EXECUTE_READWRITE: return VirtualMemory::Mode::ExecuteReadWrite;
-		default: return VirtualMemory::Mode::NoAccess;
-	}
-}
-
 void SysVirtualInit() {}
 
 uint64_t SysVirtualAlloc(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
@@ -328,17 +316,13 @@ bool SysVirtualFreeRange(uint64_t address, uint64_t size) {
 	return current - address == size && SysVirtualFree(address);
 }
 
-bool SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode,
-                       VirtualMemory::Mode* old_mode) {
+bool SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
 	DWORD old_protect = 0;
 	if (VirtualProtect(reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)), size,
 	                   GetProtectionFlag(mode), &old_protect) == 0) {
 		printf("VirtualProtect() failed: 0x%08" PRIx32 "\n", static_cast<uint32_t>(GetLastError()));
 		return false;
 	}
-	if (old_mode != nullptr) {
-		*old_mode = GetProtectionFlag(old_protect);
-	}
 	return true;
 }
 

src/common/profiler.cpp

@@ -55,19 +55,13 @@ void SetThreadName(const char* name) {
 }
 
 void Initialize() {
-	switch (Config::GetProfilerDirection()) {
-		case Config::ProfilerDirection::Network:
-			if (!tracy::ProfilerAvailable()) {
-				tracy::StartupProfiler();
-				TracySetProgramName("KytyPS5");
-				::printf("Tracy profiler enabled: client %d.%d.%d, protocol %u, "
-				         "broadcast %u, connect to 127.0.0.1:8086\n",
-				         tracy::Version::Major, tracy::Version::Minor, tracy::Version::Patch,
-				         tracy::ProtocolVersion, tracy::BroadcastVersion);
-			}
-			break;
-		case Config::ProfilerDirection::None:
-		default: break;
+	if (Config::ProfilerEnabled() && !tracy::ProfilerAvailable()) {
+		tracy::StartupProfiler();
+		TracySetProgramName("KytyPS5");
+		::printf("Tracy profiler enabled: client %d.%d.%d, protocol %u, "
+		         "broadcast %u, connect to 127.0.0.1:8086\n",
+		         tracy::Version::Major, tracy::Version::Minor, tracy::Version::Patch,
+		         tracy::ProtocolVersion, tracy::BroadcastVersion);
 	}
 }
 

src/common/stringUtils.h

@@ -316,10 +316,12 @@ inline std::string SafeCsv(std::string_view text) {
 	return std::string(text);
 }
 
-inline std::string Utf16ToUtf8(const char16_t* utf16) {
-	std::u16string                                                    input(utf16);
+inline std::string Utf16ToUtf8(std::u16string_view utf16) {
+	if (utf16.empty()) {
+		return {};
+	}
 	std::wstring_convert<std::codecvt_utf8_utf16<char16_t>, char16_t> convert;
-	return convert.to_bytes(input);
+	return convert.to_bytes(utf16.data(), utf16.data() + utf16.size());
 }
 
 inline ByteBuffer HexToBin(std::string_view text) {

src/common/threads.cpp

@@ -1,7 +1,6 @@
 #include "common/threads.h"
 
 #include "common/assert.h"
-#include "common/debug.h"
 
 #include <algorithm>
 #include <atomic>
@@ -9,7 +8,6 @@
 #include <chrono>             // IWYU pragma: keep
 #include <condition_variable> // IWYU pragma: keep
 #include <mutex>
-#include <vector>
 
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS && KYTY_COMPILER == KYTY_COMPILER_CLANG
 #define KYTY_WIN_CS
@@ -198,37 +196,7 @@ struct CondVarPrivate {
 #endif
 };
 
-static std::recursive_mutex                         g_cond_waiters_mutex;
-static std::vector<std::pair<int, CondVarPrivate*>> g_cond_waiters;
-static wait_poll_func_t                             g_cond_wait_poll_callback = nullptr;
-
-static void WakeCondVar(CondVarPrivate* cond_var) {
-#ifdef KYTY_WIN_CS
-	static auto func = ResolveWakeAllConditionVariable();
-	EXIT_NOT_IMPLEMENTED(func == nullptr);
-	func(&cond_var->m_cv);
-#else
-	cond_var->m_cv.notify_all();
-#endif
-}
-
-static void RegisterCondWaiter(CondVarPrivate* cond_var) {
-	std::lock_guard lock(g_cond_waiters_mutex);
-	g_cond_waiters.emplace_back(Thread::GetThreadIdUnique(), cond_var);
-}
-
-static void UnregisterCondWaiter(CondVarPrivate* cond_var) {
-	const auto      thread_id = Thread::GetThreadIdUnique();
-	std::lock_guard lock(g_cond_waiters_mutex);
-
-	const auto it = std::find_if(g_cond_waiters.begin(), g_cond_waiters.end(),
-	                             [thread_id, cond_var](const auto& waiter) {
-		                             return waiter.first == thread_id && waiter.second == cond_var;
-	                             });
-	if (it != g_cond_waiters.end()) {
-		g_cond_waiters.erase(it);
-	}
-}
+static wait_poll_func_t g_cond_wait_poll_callback = nullptr;
 
 struct ThreadPrivate {
 	ThreadPrivate(thread_func_t f, void* a): func(f), arg(a), m_thread(&Run, this) {}
@@ -366,7 +334,6 @@ CondVar::~CondVar() {
 }
 
 void CondVar::Wait(Mutex* mutex) {
-	RegisterCondWaiter(m_cond_var.get());
 #ifndef KYTY_WIN_CS
 	std::unique_lock<std::recursive_mutex> cpp_lock(mutex->m_mutex->m_mutex, std::adopt_lock_t());
 #endif
@@ -407,7 +374,6 @@ void CondVar::Wait(Mutex* mutex) {
 	}
 	cpp_lock.release();
 #endif
-	UnregisterCondWaiter(m_cond_var.get());
 }
 
 void CondVar::SetWaitPollCallback(wait_poll_func_t callback) {
@@ -416,7 +382,6 @@ void CondVar::SetWaitPollCallback(wait_poll_func_t callback) {
 
 bool CondVar::WaitFor(Mutex* mutex, uint32_t micros) {
 	bool ok = false;
-	RegisterCondWaiter(m_cond_var.get());
 #ifndef KYTY_WIN_CS
 	std::unique_lock<std::recursive_mutex> cpp_lock(mutex->m_mutex->m_mutex, std::adopt_lock_t());
 #endif
@@ -431,7 +396,6 @@ bool CondVar::WaitFor(Mutex* mutex, uint32_t micros) {
 	      std::cv_status::no_timeout);
 	cpp_lock.release();
 #endif
-	UnregisterCondWaiter(m_cond_var.get());
 	return ok;
 }
 
@@ -446,16 +410,13 @@ void CondVar::Signal() {
 }
 
 void CondVar::SignalAll() {
-	WakeCondVar(m_cond_var.get());
-}
-
-void CondVar::SignalThread(int thread_id) {
-	std::lock_guard lock(g_cond_waiters_mutex);
-	for (const auto& waiter: g_cond_waiters) {
-		if (waiter.first == thread_id) {
-			WakeCondVar(waiter.second);
-		}
-	}
+#ifdef KYTY_WIN_CS
+	static auto func = ResolveWakeAllConditionVariable();
+	EXIT_NOT_IMPLEMENTED(func == nullptr);
+	func(&m_cond_var->m_cv);
+#else
+	m_cond_var->m_cv.notify_all();
+#endif
 }
 
 int Thread::GetThreadIdUnique() {

src/common/threads.h

@@ -77,7 +77,6 @@ public:
 	void Signal();
 	void SignalAll();
 
-	static void SignalThread(int thread_id);
 	static void SetWaitPollCallback(wait_poll_func_t callback);
 
 	KYTY_CLASS_NO_COPY(CondVar);

src/common/virtualMemory.cpp

@@ -50,8 +50,8 @@ bool FreeRange(uint64_t address, uint64_t size) {
 	return SysVirtualFreeRange(address, size);
 }
 
-bool Protect(uint64_t address, uint64_t size, Mode mode, Mode* old_mode) {
-	return SysVirtualProtect(address, size, mode, old_mode);
+bool Protect(uint64_t address, uint64_t size, Mode mode) {
+	return SysVirtualProtect(address, size, mode);
 }
 
 bool FlushInstructionCache(uint64_t address, uint64_t size) {

src/common/virtualMemory.h

@@ -35,7 +35,7 @@ bool     ReserveFixed(uint64_t address, uint64_t size);
 bool     Decommit(uint64_t address, uint64_t size);
 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     Protect(uint64_t address, uint64_t size, Mode mode);
 bool     FlushInstructionCache(uint64_t address, uint64_t size);
 
 } // namespace VirtualMemory

src/emulator.cpp

@@ -196,6 +196,10 @@ void Run(const RunOptions& options) {
 	ClearDebugTextureFolder();
 
 	PrintSystemInfo();
+	std::string title_id;
+	if (Loader::SystemContentParamSfoGetString("TITLE_ID", &title_id) && !title_id.empty()) {
+		Log::WriteToConsoleAndLog(fmt::format("Title ID: {}\n", title_id));
+	}
 
 	int ok = atexit(KytyClose);
 	EXIT_NOT_IMPLEMENTED(ok != 0);

src/graphics/guest_gpu/command_processor/commandProcessor.h

@@ -32,11 +32,12 @@ private:
 
 	struct BufferCursor {
 		std::span<const uint32_t> commands;
-		uint32_t                  offset_dw           = 0;
-		uint32_t                  deferred_advance_dw = 0;
+		uint32_t                  offset_dw = 0;
 	};
 
 	std::vector<BufferCursor> m_buffer_stack;
+	std::span<const uint32_t> m_next_buffer;
+	bool                      m_chain         = false;
 	bool                      m_suspended     = false;
 	bool                      m_made_progress = false;
 };
@@ -73,13 +74,9 @@ public:
 	void SetDrawIndirectArgsBaseAddress(uint64_t draw_indirect_args_base_addr);
 	void SetDispatchIndirectArgsBaseAddress(uint64_t dispatch_indirect_args_base_addr);
 	void SetNumInstances(uint32_t num_instances);
-	void DrawIndex(uint32_t index_count, const void* index_addr, uint32_t flags, uint32_t type,
-	               uint32_t instance_count = 0, const void* object_ids = nullptr,
-	               uint32_t render_target_slice_offset = 0, int32_t vertex_offset_add = 0,
-	               uint32_t first_instance = 0);
-	void DrawIndexOffset(uint32_t index_offset, uint32_t index_count, uint32_t flags);
-	void DrawIndexAuto(uint32_t index_count, uint32_t flags,
-	                   uint32_t render_target_slice_offset = 0);
+	void DrawIndex(DrawIndexArgs args);
+	void DrawIndexOffset(uint32_t index_offset, uint32_t index_count);
+	void DrawIndexAuto(DrawAutoArgs args);
 	void DrawIndirect(uint32_t data_offset, uint32_t draw_initiator, bool indexed);
 	void DrawIndirectMulti(uint32_t data_offset, uint32_t max_count_or_count,
 	                       const volatile uint32_t* count_addr, uint32_t stride_in_bytes,
@@ -133,7 +130,7 @@ public:
 	[[nodiscard]] bool ShouldSkipPredicatedPackets() const { return m_predicate_skip; }
 
 	Pm4ProcessResult Process(Pm4Execution& execution, std::span<const uint32_t> commands);
-	void             ProcessIndirectBuffer(std::span<const uint32_t> commands);
+	void             ProcessIndirectBuffer(std::span<const uint32_t> commands, bool chain);
 
 	void SetFlip(const FlipInfo& flip) { m_flip = flip; }
 
@@ -147,16 +144,11 @@ private:
 	                      uint32_t cache_action, uint32_t event_index, uint32_t event_write_source,
 	                      void* dst_gpu_addr, T value, uint32_t interrupt_selector,
 	                      uint32_t interrupt_context_id);
-	void ProcessPm4(Pm4Execution& execution, size_t stop_depth);
+	void ProcessPm4(Pm4Execution& execution);
 	void SuspendPm4();
-	void SubmitNonIndexedDraw(uint32_t vertex_count, uint32_t flags,
-	                          uint32_t render_target_slice_offset, uint32_t first_vertex,
-	                          uint32_t first_instance);
-
 	CommandScheduler&   GetScheduler() const { return m_renderer.GetCommandScheduler(); }
 	CommandBuffer&      CurrentBuffer() { return GetScheduler().Current(); }
 	void                CheckBuffer() const { GetScheduler().CheckActive(); }
-	GpuResourceManager& GetGpuResources() const { return m_renderer.GetGpuResources(); }
 
 	RenderContext&   m_renderer;
 	HW::Context      m_ctx;

src/graphics/guest_gpu/command_processor/pm4Dispatch.cpp

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

src/graphics/guest_gpu/command_processor/pm4Dispatch.h

@@ -44,11 +44,7 @@ uint32_t HwCtxSetBorderColorTableAddr(CommandProcessor&, uint32_t, uint32_t, con
 uint32_t HwCtxSetStencilClear(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetDepthClear(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetScreenScissor(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwCtxSetDepthRenderTarget(CommandProcessor&, uint32_t, uint32_t, const uint32_t*,
-                                   uint32_t);
 uint32_t HwCtxSetStencilInfo(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwCtxSetDepthHtileSurface(CommandProcessor&, uint32_t, uint32_t, const uint32_t*,
-                                   uint32_t);
 uint32_t HwCtxSetHardwareScreenOffset(CommandProcessor&, uint32_t, uint32_t, const uint32_t*,
                                       uint32_t);
 uint32_t HwCtxSetWindowOffset(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
@@ -89,7 +85,6 @@ uint32_t HwCtxSetDrawPayloadControl(CommandProcessor&, uint32_t, uint32_t, const
 uint32_t HwCtxSetPrimitiveIdReset(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetObjprimIdControl(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetShaderStages(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwCtxSetRenderTarget(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetColorInfo(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetBlendControl(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwCtxSetViewportScissor(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
@@ -100,12 +95,10 @@ uint32_t HwShSetPsUserSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t
 uint32_t HwShSetGsUserSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwShSetHsUserSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwShSetCsUserSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwShSetPsUserAccumSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwShSetGsUserAccumSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwShSetHsUserAccumSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwShSetCsUserAccumSgpr(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
+uint32_t HwShIgnoreUserAccumulator(CommandProcessor&, uint32_t, uint32_t, const uint32_t*,
+                                   uint32_t);
 uint32_t HwShSetCsRegisters(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
-uint32_t HwShIgnoreRegisters(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
+uint32_t HwShSetRegisters(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwUcSetPrimitiveType(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwUcSetIndexType(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 uint32_t HwUcSetObjectId(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);

src/graphics/guest_gpu/gpu_defs.h

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

src/graphics/guest_gpu/gpu_format.cpp

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

src/graphics/guest_gpu/gpu_format.h

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

src/graphics/guest_gpu/graphicsRun.cpp

@@ -13,7 +13,6 @@
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/renderer/sync.h"
-#include "graphics/presentation/renderDoc.h"
 #include "graphics/presentation/videoOut.h"
 #include "graphics/presentation/window.h"
 #include "graphics/shader/shader.h"
@@ -40,6 +39,21 @@ static thread_local bool              g_gpu_mutex_owned   = false;
 static thread_local bool              g_gpu_thread        = false;
 static thread_local GuestGpu*         g_gpu_state         = nullptr;
 
+struct DrawIndirectArgs {
+	uint32_t vertex_count_per_instance;
+	uint32_t instance_count;
+	uint32_t start_vertex_location;
+	uint32_t start_instance_location;
+};
+
+struct DrawIndexedIndirectArgs {
+	uint32_t index_count_per_instance;
+	uint32_t instance_count;
+	uint32_t start_index_location;
+	uint32_t base_vertex_location;
+	uint32_t start_instance_location;
+};
+
 class GpuMutexLock final {
 public:
 	explicit GpuMutexLock(Common::Mutex& mutex): m_mutex(mutex) {
@@ -63,7 +77,7 @@ private:
 
 static bool GraphicsRunDebugDumpEnabled() {
 	return Config::GraphicsDebugDumpEnabled() &&
-	       Config::GetPrintfDirection() != Config::OutputDirection::Silent;
+	       Config::GetPrintfDirection() != Config::LogDirection::Silent;
 }
 
 GuestGpu::GuestGpu(RenderContext& renderer): m_renderer(renderer) {
@@ -372,8 +386,12 @@ void CommandProcessor::WriteReferenceClock(uint64_t dst_address, uint32_t num_by
 	}
 	const auto value = Sync::ReadReferenceClock();
 	std::memcpy(reinterpret_cast<void*>(dst_address), &value, num_bytes);
-	LOGF("\t copy_data reference clock: dst=0x%016" PRIx64 " value=0x%016" PRIx64 " size=%u\n",
-	     dst_address, value, num_bytes);
+	static std::atomic<uint32_t> clock_log_count {0};
+	if (clock_log_count.fetch_add(1) < 64) {
+		LOGF("\t copy_data reference clock: dst=0x%016" PRIx64 " value=0x%016" PRIx64
+		     " size=%u\n",
+		     dst_address, value, num_bytes);
+	}
 }
 
 void CommandProcessor::DmaData(uint8_t engine, uint8_t dst_sel, uint8_t dst_cache_policy,
@@ -413,7 +431,7 @@ void CommandProcessor::DmaData(uint8_t engine, uint8_t dst_sel, uint8_t dst_cach
 	if (!decode_gds(dst_sel, dst_gds)) {
 		EXIT("unsupported dmaData destination selector 0x%02" PRIx8 "\n", dst_sel);
 	}
-	auto& buffer_cache = GetGpuResources().GetBufferCache();
+	auto& buffer_cache = m_renderer.GetBufferCache();
 	if (src_sel == 2) {
 		buffer_cache.FillBuffer(
 		    dst_address_or_offset, num_bytes,
@@ -546,10 +564,6 @@ void GuestGpu::ThreadRun(void* data) {
 
 bool GuestGpu::Process(Submission& submission) {
 	const bool first_slice = !submission.started;
-	if (first_slice && RenderDocCaptureRequested()) {
-		Common::LockGuard render_lock(m_renderer.GetMutex());
-		RenderDocStartCapture();
-	}
 	auto& cp = GetProcessor(submission.queue_id);
 
 	if (first_slice && submission.reset_processor) {
@@ -593,11 +607,11 @@ bool GuestGpu::Process(Submission& submission) {
 			}
 			if (progressed) {
 				if (complete) {
-					m_renderer.GetGpuResources().RunGarbageCollector();
+					m_renderer.RunGarbageCollector();
 				}
 				cp.BufferFlush();
 			} else if (complete) {
-				m_renderer.GetGpuResources().RunGarbageCollector();
+				m_renderer.RunGarbageCollector();
 			}
 			break;
 		}
@@ -619,16 +633,16 @@ bool GuestGpu::Process(Submission& submission) {
 			           Pm4ProcessResult::Complete;
 			if (submission.command_execution.MadeProgress()) {
 				if (complete) {
-					m_renderer.GetGpuResources().RunGarbageCollector();
+					m_renderer.RunGarbageCollector();
 				}
 				cp.BufferFlush();
 			} else if (complete) {
-				m_renderer.GetGpuResources().RunGarbageCollector();
+				m_renderer.RunGarbageCollector();
 			}
 			break;
 		}
 		case SubmissionType::FlipPreparation:
-			m_renderer.GetGpuResources().RunGarbageCollector();
+			m_renderer.RunGarbageCollector();
 			cp.PrepareCpuFlip(submission.flip_request_id);
 			break;
 	}
@@ -662,20 +676,16 @@ Pm4ProcessResult CommandProcessor::Process(Pm4Execution&             execution,
 		Pm4Execution*     previous_execution;
 	} execution_scope(*this, execution);
 
-	ProcessPm4(execution, 0);
+	ProcessPm4(execution);
 	return execution.m_buffer_stack.empty() ? Pm4ProcessResult::Complete
 	                                        : Pm4ProcessResult::Blocked;
 }
 
-void CommandProcessor::ProcessIndirectBuffer(std::span<const uint32_t> commands) {
+void CommandProcessor::ProcessIndirectBuffer(std::span<const uint32_t> commands, bool chain) {
 	EXIT_IF(g_current_execution == nullptr);
-	if (commands.empty()) {
-		return;
-	}
-	auto&      execution  = *g_current_execution;
-	const auto stop_depth = execution.m_buffer_stack.size();
-	execution.m_buffer_stack.push_back({commands});
-	ProcessPm4(execution, stop_depth);
+	EXIT_IF(!g_current_execution->m_next_buffer.empty());
+	g_current_execution->m_next_buffer = commands;
+	g_current_execution->m_chain       = chain;
 }
 
 void CommandProcessor::SuspendPm4() {
@@ -683,21 +693,13 @@ void CommandProcessor::SuspendPm4() {
 	g_current_execution->m_suspended = true;
 }
 
-void CommandProcessor::ProcessPm4(Pm4Execution& execution, size_t stop_depth) {
-	while (execution.m_buffer_stack.size() > stop_depth) {
+void CommandProcessor::ProcessPm4(Pm4Execution& execution) {
+	while (!execution.m_buffer_stack.empty()) {
 		if (g_gpu_state != nullptr) {
 			g_gpu_state->ProcessCommands();
 		}
-		const auto buffer_index = execution.m_buffer_stack.size() - 1;
-		auto&      cursor       = execution.m_buffer_stack[buffer_index];
+		auto& cursor = execution.m_buffer_stack.back();
 		EXIT_IF(cursor.offset_dw > cursor.commands.size());
-		if (cursor.deferred_advance_dw != 0) {
-			EXIT_IF(cursor.deferred_advance_dw > cursor.commands.size() - cursor.offset_dw);
-			cursor.offset_dw += cursor.deferred_advance_dw;
-			cursor.deferred_advance_dw = 0;
-			execution.m_made_progress  = true;
-			continue;
-		}
 		if (cursor.offset_dw == cursor.commands.size()) {
 			execution.m_buffer_stack.pop_back();
 			continue;
@@ -773,14 +775,19 @@ void CommandProcessor::ProcessPm4(Pm4Execution& execution, size_t stop_depth) {
 		    handler(*this, packet_header & ~1u, packet + 1, remaining_dw, total_dw) + 1;
 		EXIT_IF(packet_dw > remaining_dw);
 		if (execution.m_suspended) {
-			if (execution.m_buffer_stack.size() > buffer_index + 1) {
-				execution.m_buffer_stack[buffer_index].deferred_advance_dw = packet_dw;
-			}
 			return;
 		}
-		EXIT_IF(execution.m_buffer_stack.size() != buffer_index + 1);
-		execution.m_buffer_stack[buffer_index].offset_dw += packet_dw;
+		cursor.offset_dw += packet_dw;
 		execution.m_made_progress = true;
+		if (!execution.m_next_buffer.empty()) {
+			// Chains and taken branches reuse the fetcher; only calls retain a return cursor.
+			if (execution.m_chain) {
+				cursor = {execution.m_next_buffer};
+			} else {
+				execution.m_buffer_stack.push_back({execution.m_next_buffer});
+			}
+			execution.m_next_buffer = {};
+		}
 	}
 }
 
@@ -815,68 +822,72 @@ void CommandProcessor::SetNumInstances(uint32_t num_instances) {
 
 void CommandProcessor::SetPredication(uint32_t condition, uint32_t op, uint32_t wait_op,
                                       const volatile void* address, uint32_t count_in_dwords) {
-	if (wait_op != 0) {
-		BufferFlushAndWait();
-	}
-
 	(void)count_in_dwords;
+	uint64_t value = 0;
 
 	switch (op) {
-		case 0x00: {
+		case 0x00:
 			m_predicate_skip = false;
-		} break;
-		case 0x03: {
+			return;
+		case 0x01: {
 			EXIT_NOT_IMPLEMENTED(address == nullptr);
-
-			auto value = *reinterpret_cast<const volatile uint64_t*>(address);
-
-			switch (condition) {
-				case 0x00: m_predicate_skip = (value != 0); break;
-				case 0x01: m_predicate_skip = (value == 0); break;
-				default: EXIT("unknown predication condition: 0x%08" PRIx32 "\n", condition);
-			}
-			static std::atomic<uint32_t> log_count {0};
-			if (log_count.fetch_add(1) < 128) {
-				LOGF("\t bool predication: addr=0x%016" PRIx64 ", value=0x%016" PRIx64
-				     ", condition=%" PRIu32 ", skip=%u, wait_op=%" PRIu32 "\n",
-				     reinterpret_cast<uint64_t>(address), value, condition,
-				     m_predicate_skip ? 1u : 0u, wait_op);
+			// One begin/end pair per DB; bit 63 marks each counter ready.
+			constexpr uint64_t ready_bit = 1ull << 63u;
+			const auto* results = reinterpret_cast<const volatile uint64_t*>(address);
+			for (uint32_t db = 0; db < 16u; db++) {
+				const auto begin = results[db * 2u];
+				const auto end   = results[db * 2u + 1u];
+				if ((begin & end & ready_bit) == 0) {
+					if (wait_op == 0) {
+						SuspendPm4();
+					} else {
+						m_predicate_skip = false;
+					}
+					return;
+				}
+				value += end - begin;
 			}
 		} break;
+		case 0x03:
+			if (wait_op != 0) {
+				BufferFlushAndWait();
+			}
+			EXIT_NOT_IMPLEMENTED(address == nullptr);
+			value = *reinterpret_cast<const volatile uint64_t*>(address);
+			break;
 		default: EXIT("unknown predication op: 0x%08" PRIx32 "\n", op);
 	}
+	switch (condition) {
+		case 0x00: m_predicate_skip = (value != 0); break;
+		case 0x01: m_predicate_skip = (value == 0); break;
+		default: EXIT("unknown predication condition: 0x%08" PRIx32 "\n", condition);
+	}
+	if (op == 0x03) {
+		static std::atomic<uint32_t> log_count {0};
+		if (log_count.fetch_add(1) < 128) {
+			LOGF("\t bool predication: addr=0x%016" PRIx64 ", value=0x%016" PRIx64
+			     ", condition=%" PRIu32 ", skip=%u, wait_op=%" PRIu32 "\n",
+			     reinterpret_cast<uint64_t>(address), value, condition,
+			     m_predicate_skip ? 1u : 0u, wait_op);
+		}
+	}
 }
 
-void CommandProcessor::DrawIndex(uint32_t index_count, const void* index_addr, uint32_t flags,
-                                 uint32_t type, uint32_t instance_count, const void* object_ids,
-                                 uint32_t render_target_slice_offset, int32_t vertex_offset_add,
-                                 uint32_t first_instance) {
+void CommandProcessor::DrawIndex(DrawIndexArgs args) {
 	CheckBuffer();
 
-	if (instance_count == 0) {
-		instance_count = m_num_instances;
+	args.index_type_and_size = m_index_type_and_size;
+	if (args.instance_count == 0) {
+		args.instance_count = m_num_instances;
 	}
-	if (object_ids != nullptr) {
-		LOGF("\t draw indexed multi-instanced objectIds = 0x%016" PRIx64 "\n",
-		     reinterpret_cast<uint64_t>(object_ids));
+	if (args.base_vertex != 0 || args.first_instance != 0) {
+		LOGF("\t draw indexed offsets: base_vertex = %" PRId32 ", first_instance = %" PRIu32 "\n",
+		     args.base_vertex, args.first_instance);
 	}
-	if (render_target_slice_offset != 0) {
-		LOGF("\t draw render target slice offset = %" PRIu32 "\n", render_target_slice_offset);
-	}
-	if (vertex_offset_add != 0 || first_instance != 0) {
-		LOGF("\t draw indexed offsets: vertex_offset_add = %" PRId32 ", first_instance = %" PRIu32
-		     "\n",
-		     vertex_offset_add, first_instance);
-	}
-	m_renderer.GetRenderExecutor().DrawIndex(
-	    m_submit_id, CurrentBuffer(), m_index_type_and_size, index_count, index_addr, flags, type,
-	    instance_count, render_target_slice_offset, vertex_offset_add, first_instance);
+	m_renderer.GetRenderExecutor().DrawIndex(m_submit_id, CurrentBuffer(), args);
 }
 
-void CommandProcessor::DrawIndexOffset(uint32_t index_offset, uint32_t index_count,
-                                       uint32_t flags) {
-	CheckBuffer();
-
+void CommandProcessor::DrawIndexOffset(uint32_t index_offset, uint32_t index_count) {
 	uint64_t index_size = 0;
 	switch (m_index_type_and_size) {
 		case 0: index_size = 2; break;
@@ -888,25 +899,10 @@ void CommandProcessor::DrawIndexOffset(uint32_t index_offset, uint32_t index_cou
 	auto* index_addr = reinterpret_cast<const void*>(
 	    m_index_base_addr + static_cast<uint64_t>(index_offset) * index_size);
 
-	m_renderer.GetRenderExecutor().DrawIndex(m_submit_id, CurrentBuffer(), m_index_type_and_size,
-	                                         index_count, index_addr, flags, 1, m_num_instances);
+	DrawIndex({.index_count = index_count, .index_addr = index_addr});
 }
 
 void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiator, bool indexed) {
-	struct DrawIndirectArgs {
-		uint32_t vertex_count_per_instance;
-		uint32_t instance_count;
-		uint32_t start_vertex_location;
-		uint32_t start_instance_location;
-	};
-	struct DrawIndexedIndirectArgs {
-		uint32_t index_count_per_instance;
-		uint32_t instance_count;
-		uint32_t start_index_location;
-		uint32_t base_vertex_location;
-		uint32_t start_instance_location;
-	};
-
 	EXIT_NOT_IMPLEMENTED((draw_initiator & ~0x20u) != 2u);
 	EXIT_NOT_IMPLEMENTED(m_draw_indirect_args_base_addr == 0);
 
@@ -928,8 +924,11 @@ void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiato
 			}
 		}
 		m_num_instances = args.instance_count;
-		SubmitNonIndexedDraw(args.vertex_count_per_instance, 0, 0, args.start_vertex_location,
-		                     args.start_instance_location);
+		DrawIndexAuto({.vertex_count   = args.vertex_count_per_instance,
+		               .instance_count = args.instance_count,
+		               .first_vertex   = args.start_vertex_location,
+		               .first_instance = args.start_instance_location,
+		               .offset_source  = DrawOffsetSource::IndirectArgs});
 		return;
 	}
 
@@ -970,28 +969,18 @@ void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiato
 	}
 
 	m_num_instances = args.instance_count;
-	DrawIndex(index_count, index_addr, 0, 1, args.instance_count, nullptr, 0,
-	          static_cast<int32_t>(args.base_vertex_location), args.start_instance_location);
+	DrawIndex({.index_count    = index_count,
+	           .index_addr     = index_addr,
+	           .instance_count = args.instance_count,
+	           .base_vertex    = static_cast<int32_t>(args.base_vertex_location),
+	           .first_instance = args.start_instance_location,
+	           .offset_source  = DrawOffsetSource::IndirectArgs});
 }
 
 void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_count_or_count,
                                          const volatile uint32_t* count_addr,
                                          uint32_t stride_in_bytes, uint32_t draw_initiator,
                                          bool indexed) {
-	struct DrawIndirectArgs {
-		uint32_t vertex_count_per_instance;
-		uint32_t instance_count;
-		uint32_t start_vertex_location;
-		uint32_t start_instance_location;
-	};
-	struct DrawIndexedIndirectArgs {
-		uint32_t index_count_per_instance;
-		uint32_t instance_count;
-		uint32_t start_index_location;
-		uint32_t base_vertex_location;
-		uint32_t start_instance_location;
-	};
-
 	EXIT_NOT_IMPLEMENTED((draw_initiator & ~0x20u) != 2u);
 	EXIT_NOT_IMPLEMENTED(m_draw_indirect_args_base_addr == 0);
 
@@ -1028,8 +1017,11 @@ void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_coun
 				}
 			}
 			m_num_instances = args->instance_count;
-			SubmitNonIndexedDraw(args->vertex_count_per_instance, 0, 0, args->start_vertex_location,
-			                     args->start_instance_location);
+			DrawIndexAuto({.vertex_count   = args->vertex_count_per_instance,
+			               .instance_count = args->instance_count,
+			               .first_vertex   = args->start_vertex_location,
+			               .first_instance = args->start_instance_location,
+			               .offset_source  = DrawOffsetSource::IndirectArgs});
 			continue;
 		}
 
@@ -1071,8 +1063,12 @@ void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_coun
 		}
 
 		m_num_instances = args->instance_count;
-		DrawIndex(index_count, index_addr, 0, 1, args->instance_count, nullptr, 0,
-		          static_cast<int32_t>(args->base_vertex_location), args->start_instance_location);
+		DrawIndex({.index_count    = index_count,
+		           .index_addr     = index_addr,
+		           .instance_count = args->instance_count,
+		           .base_vertex    = static_cast<int32_t>(args->base_vertex_location),
+		           .first_instance = args->start_instance_location,
+		           .offset_source  = DrawOffsetSource::IndirectArgs});
 	}
 }
 
@@ -1110,17 +1106,6 @@ void CommandProcessor::DispatchDirect(uint32_t thread_group_x, uint32_t thread_g
 		// local_x        = std::max(cs.num_thread_x, 1u);
 		// local_y        = std::max(cs.num_thread_y, 1u);
 		// local_z        = std::max(cs.num_thread_z, 1u);
-		if (cs.wave_size == 64u) {
-			static std::atomic_bool logged_wave64_shader {false};
-			if (!logged_wave64_shader.exchange(true, std::memory_order_relaxed)) {
-				LOGF("warning: executing wave64 compute shader cs=0x%016" PRIx64 "\n",
-				     cs.data_addr);
-				std::printf("warning: executing wave64 compute shader cs=0x%016" PRIx64 "\n",
-				            cs.data_addr);
-				std::fflush(stdout);
-			}
-		}
-
 		m_renderer.GetRenderExecutor().DispatchDirect(m_submit_id, CurrentBuffer(), thread_group_x,
 		                                              thread_group_y, thread_group_z, mode);
 	}
@@ -1163,19 +1148,13 @@ void CommandProcessor::DispatchIndirect(uint32_t data_offset, uint32_t mode) {
 	DispatchDirect(args->thread_group_x, args->thread_group_y, args->thread_group_z, mode);
 }
 
-void CommandProcessor::DrawIndexAuto(uint32_t index_count, uint32_t flags,
-                                     uint32_t render_target_slice_offset) {
-	SubmitNonIndexedDraw(index_count, flags, render_target_slice_offset, 0, 0);
-}
-
-void CommandProcessor::SubmitNonIndexedDraw(uint32_t vertex_count, uint32_t flags,
-                                            uint32_t render_target_slice_offset,
-                                            uint32_t first_vertex, uint32_t first_instance) {
+void CommandProcessor::DrawIndexAuto(DrawAutoArgs args) {
 	CheckBuffer();
 
-	m_renderer.GetRenderExecutor().DrawAuto(m_submit_id, CurrentBuffer(), vertex_count, flags,
-	                                        render_target_slice_offset, m_num_instances,
-	                                        first_vertex, first_instance);
+	if (args.instance_count == 0) {
+		args.instance_count = m_num_instances;
+	}
+	m_renderer.GetRenderExecutor().DrawAuto(m_submit_id, CurrentBuffer(), args);
 }
 
 void CommandProcessor::WaitFlipDone(uint32_t video_out_handle, uint32_t display_buffer_index) {
@@ -1318,6 +1297,7 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 								break;
 							case 0x14:
 							case 0x28:
+							case 0x2f:
 								if (event_index == 0x00) {
 									write64(false);
 									return;
@@ -1325,7 +1305,6 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 								break;
 							case 0x2b:
 							case 0x2d:
-							case 0x2f:
 							case 0x30:
 								if (event_index == 0x00 && !with_interrupt) {
 									write64(false);

src/graphics/guest_gpu/hardwareContext.h

@@ -11,15 +11,6 @@ struct ColorBase {
 	uint64_t addr = 0;
 };
 
-struct ColorPitch {
-	uint32_t pitch_div8_minus1       = 0;
-	uint32_t fmask_pitch_div8_minus1 = 0;
-};
-
-struct ColorSlice {
-	uint32_t slice_div64_minus1 = 0;
-};
-
 struct ColorView {
 	uint32_t base_array_slice_index = 0;
 	uint32_t last_array_slice_index = 0;
@@ -36,20 +27,15 @@ struct ColorInfo {
 	bool                    blend_clamp                    = false;
 	bool                    blend_bypass                   = false;
 	bool                    round_mode                     = false;
-	uint32_t                cmask_is_linear                = 0;
-	uint32_t                cmask_addr_type                = 0;
-	bool                    alt_tile_mode                  = false;
 	Prospero::ChannelLayout format                         = Prospero::ChannelLayout::kInvalid;
 	Prospero::ChannelType   channel_type                   = Prospero::ChannelType::kUNorm;
 	Prospero::ChannelOrder  channel_order                  = Prospero::ChannelOrder::kStandard;
 };
 
 struct ColorAttrib {
-	bool               force_dest_alpha_to_one = false;
-	Prospero::TileMode tile_mode               = Prospero::TileMode::kLinear;
-	Prospero::TileMode fmask_tile_mode         = Prospero::TileMode::kLinear;
-	uint32_t           num_samples             = 0;
-	uint32_t           num_fragments           = 0;
+	bool     force_dest_alpha_to_one = false;
+	uint32_t num_samples             = 0;
+	uint32_t num_fragments           = 0;
 };
 
 struct ColorAttrib2 {
@@ -59,41 +45,33 @@ struct ColorAttrib2 {
 };
 
 struct ColorAttrib3 {
-	uint32_t           depth              = 0;
-	Prospero::TileMode tile_mode          = Prospero::TileMode::kLinear;
-	uint32_t           dimension          = 0;
-	bool               cmask_pipe_aligned = false;
-	bool               dcc_pipe_aligned   = false;
+	uint32_t           depth                        = 0;
+	Prospero::TileMode tile_mode                    = Prospero::TileMode::kRenderTarget;
+	uint32_t           dimension                    = 0;
+	bool               metadata_pipe_aligned        = false;
+	bool               write_vrs_rate_hint_to_cmask = false;
 };
 
 struct ColorDccControl {
-	uint32_t max_uncompressed_block_size    = 0;
-	uint32_t max_compressed_block_size      = 0;
-	uint32_t min_compressed_block_size      = 0;
-	uint32_t color_transform                = 0;
-	bool     dcc_clear_key_enable           = false;
-	bool     overwrite_combiner_disable     = false;
-	bool     independent_64b_blocks         = false;
-	bool     independent_128b_blocks        = false;
-	bool     data_write_on_dcc_clear_to_reg = false;
+	enum class IndependentBlockSize : uint8_t { Disabled, Bytes64, Bytes128 };
+
+	uint32_t             max_uncompressed_block_size    = 2;
+	uint32_t             max_compressed_block_size      = 2;
+	uint32_t             color_transform                = 0;
+	bool                 dcc_clear_key_enable           = false;
+	bool                 overwrite_combiner_disable     = false;
+	IndependentBlockSize independent_block_size         = IndependentBlockSize::Disabled;
+	bool                 data_write_on_dcc_clear_to_reg = false;
 };
 
 struct ColorCmask {
 	uint64_t addr = 0;
 };
 
-struct ColorCmaskSlice {
-	uint32_t slice_minus1 = 0;
-};
-
 struct ColorFmask {
 	uint64_t addr = 0;
 };
 
-struct ColorFmaskSlice {
-	uint32_t slice_minus1 = 0;
-};
-
 struct ColorClearWord0 {
 	uint32_t word0 = 0;
 };
@@ -106,15 +84,8 @@ struct ColorDccAddr {
 	uint64_t addr = 0;
 };
 
-struct ColorSize {
-	uint32_t width  = 0;
-	uint32_t height = 0;
-};
-
 struct RenderTarget {
 	ColorBase       base;
-	ColorPitch      pitch;
-	ColorSlice      slice;
 	ColorView       view;
 	ColorInfo       info;
 	ColorAttrib     attrib;
@@ -122,13 +93,10 @@ struct RenderTarget {
 	ColorAttrib3    attrib3;
 	ColorDccControl dcc;
 	ColorCmask      cmask;
-	ColorCmaskSlice cmask_slice;
 	ColorFmask      fmask;
-	ColorFmaskSlice fmask_slice;
 	ColorClearWord0 clear_word0;
 	ColorClearWord1 clear_word1;
 	ColorDccAddr    dcc_addr;
-	ColorSize       size;
 };
 
 struct DepthZInfo {
@@ -136,7 +104,7 @@ struct DepthZInfo {
 	uint32_t                                    num_samples = 0;
 	Prospero::TextureCompatiblePlaneCompression texture_compatibility =
 	    Prospero::TextureCompatiblePlaneCompression::kDisable;
-	Prospero::ZCompareBase z_compare_base     = Prospero::ZCompareBase::kZMin;
+	Prospero::ZCompareBase z_compare_base     = Prospero::ZCompareBase::kZMax;
 	bool                   htile_acceleration = false;
 	bool                   expclear_enabled   = false;
 	bool                   partially_resident = false;
@@ -171,7 +139,7 @@ struct DepthStencilInfo {
 	Prospero::TextureCompatibleStencil texture_compatibility =
 	    Prospero::TextureCompatibleStencil::kDisable;
 	bool expclear_enabled       = false;
-	bool htile_stencil_disabled = false;
+	bool htile_stencil_disabled = true;
 	bool partially_resident     = false;
 
 	[[nodiscard]] static DepthStencilInfo Decode(uint32_t value) {
@@ -194,16 +162,6 @@ struct DepthStencilInfo {
 	}
 };
 
-struct DepthRenderTargetDepthInfo {
-	uint32_t addr5_swizzle_mask = 0;
-	uint32_t array_mode         = 0;
-	uint32_t pipe_config        = 0;
-	uint32_t bank_width         = 0;
-	uint32_t bank_height        = 0;
-	uint32_t macro_tile_aspect  = 0;
-	uint32_t num_banks          = 0;
-};
-
 struct DepthDepthView {
 	uint32_t slice_start           = 0;
 	uint32_t slice_max             = 0;
@@ -218,36 +176,18 @@ struct DepthDepthSizeXY {
 	bool     valid = false;
 };
 
-struct DepthRenderTargetHTileSurface {
-	uint32_t linear                  = 0;
-	uint32_t full_cache              = 0;
-	uint32_t htile_uses_preload_win  = 0;
-	uint32_t preload                 = 0;
-	uint32_t prefetch_width          = 0;
-	uint32_t prefetch_height         = 0;
-	uint32_t dst_outside_zero_to_one = 0;
-};
-
 struct DepthRenderTarget {
-	DepthZInfo                    z_info;
-	DepthStencilInfo              stencil_info;
-	DepthDepthView                depth_view;
-	DepthDepthSizeXY              size;
-	DepthRenderTargetDepthInfo    depth_info;
-	DepthRenderTargetHTileSurface htile_surface;
+	DepthZInfo       z_info;
+	DepthStencilInfo stencil_info;
+	DepthDepthView   depth_view;
+	DepthDepthSizeXY size;
 
 	uint64_t z_read_base_addr        = 0;
 	uint64_t stencil_read_base_addr  = 0;
 	uint64_t z_write_base_addr       = 0;
 	uint64_t stencil_write_base_addr = 0;
-	uint32_t pitch_div8_minus1       = 0;
-	uint32_t height_div8_minus1      = 0;
-	bool     pitch_height_valid      = false;
-	uint32_t slice_div64_minus1      = 0;
 	uint64_t htile_data_base_addr    = 0;
-	uint32_t width                   = 0;
-	uint32_t height                  = 0;
-	bool     width_height_valid      = false;
+	uint8_t  shading_rate_encoding   = 0;
 };
 
 struct RenderControl {
@@ -256,10 +196,19 @@ struct RenderControl {
 	bool    resummarize_enable       = false;
 	bool    stencil_compress_disable = false;
 	bool    depth_compress_disable   = false;
+	bool    copy_depth_to_color      = false;
+	bool    copy_stencil_to_color    = false;
 	bool    copy_centroid            = false;
 	uint8_t copy_sample              = 0;
 };
 
+struct DepthRenderOverride {
+	bool force_z_valid       = false;
+	bool force_z_dirty       = false;
+	bool force_stencil_valid = false;
+	bool force_stencil_dirty = false;
+};
+
 struct GdsOaCounter {
 	uint32_t counter = 0;
 	uint32_t address = 0;
@@ -299,16 +248,14 @@ struct ClipControl {
 };
 
 struct DepthControl {
-	bool    stencil_enable                     = false;
-	bool    z_enable                           = false;
-	bool    z_write_enable                     = false;
-	bool    depth_bounds_enable                = false;
-	uint8_t zfunc                              = 0;
-	bool    backface_enable                    = false;
-	uint8_t stencilfunc                        = 0;
-	uint8_t stencilfunc_bf                     = 0;
-	bool    color_writes_on_depth_fail_enable  = false;
-	bool    color_writes_on_depth_pass_disable = false;
+	bool    stencil_enable      = false;
+	bool    z_enable            = false;
+	bool    z_write_enable      = false;
+	bool    depth_bounds_enable = false;
+	uint8_t zfunc               = 0;
+	bool    backface_enable     = false;
+	uint8_t stencilfunc         = 0;
+	uint8_t stencilfunc_bf      = 0;
 };
 
 struct StencilControl {
@@ -422,16 +369,16 @@ struct AaConfig {
 struct Viewport {
 	float zmin                                  = 0.0f;
 	float zmax                                  = 0.0f;
-	float xscale                                = 0.0f;
+	float xscale                                = 1.0f;
 	float xoffset                               = 0.0f;
-	float yscale                                = 0.0f;
+	float yscale                                = 1.0f;
 	float yoffset                               = 0.0f;
-	float zscale                                = 0.0f;
+	float zscale                                = 1.0f;
 	float zoffset                               = 0.0f;
 	int   viewport_scissor_left                 = 0;
 	int   viewport_scissor_top                  = 0;
-	int   viewport_scissor_right                = 0;
-	int   viewport_scissor_bottom               = 0;
+	int   viewport_scissor_right                = 16384;
+	int   viewport_scissor_bottom               = 16384;
 	bool  viewport_scissor_window_offset_enable = false;
 };
 
@@ -440,26 +387,26 @@ struct ScreenViewport {
 	uint32_t transform_control                    = 1087;
 	int      screen_scissor_left                  = 0;
 	int      screen_scissor_top                   = 0;
-	int      screen_scissor_right                 = 0;
-	int      screen_scissor_bottom                = 0;
+	int      screen_scissor_right                 = 16384;
+	int      screen_scissor_bottom                = 16384;
 	int      window_scissor_left                  = 0;
 	int      window_scissor_top                   = 0;
-	int      window_scissor_right                 = 0;
-	int      window_scissor_bottom                = 0;
+	int      window_scissor_right                 = 16384;
+	int      window_scissor_bottom                = 16384;
 	bool     window_scissor_window_offset_enable  = false;
 	int      generic_scissor_left                 = 0;
 	int      generic_scissor_top                  = 0;
-	int      generic_scissor_right                = 0;
-	int      generic_scissor_bottom               = 0;
+	int      generic_scissor_right                = 16384;
+	int      generic_scissor_bottom               = 16384;
 	bool     generic_scissor_window_offset_enable = false;
 	int      window_offset_x                      = 0;
 	int      window_offset_y                      = 0;
 	uint32_t hw_offset_x                          = 0;
 	uint32_t hw_offset_y                          = 0;
-	float    guard_band_horz_clip                 = 0.0f;
-	float    guard_band_vert_clip                 = 0.0f;
-	float    guard_band_horz_discard              = 0.0f;
-	float    guard_band_vert_discard              = 0.0f;
+	float    guard_band_horz_clip                 = 1.0f;
+	float    guard_band_vert_clip                 = 1.0f;
+	float    guard_band_horz_discard              = 1.0f;
+	float    guard_band_vert_discard              = 1.0f;
 	uint16_t clip_rect_rule                       = 0xffffu;
 	int      clip_rect_left[4]                    = {};
 	int      clip_rect_top[4]                     = {};
@@ -468,27 +415,6 @@ struct ScreenViewport {
 	bool     clip_rect_window_offset_enable[4]    = {};
 };
 
-struct VsShaderResource1 {
-	uint8_t vgprs                    = 0;
-	uint8_t sgprs                    = 0;
-	uint8_t priority                 = 0;
-	uint8_t float_mode               = 0;
-	bool    dx10_clamp               = false;
-	bool    ieee_mode                = false;
-	uint8_t vgpr_component_count     = 0;
-	bool    cu_group_enable          = false;
-	bool    require_forward_progress = false;
-	bool    fp16_overflow            = false;
-};
-
-struct VsShaderResource2 {
-	bool    scratch_en        = false;
-	uint8_t user_sgpr         = 0;
-	bool    offchip_lds       = false;
-	bool    streamout_enabled = false;
-	uint8_t shared_vgprs      = 0;
-};
-
 struct HsShaderResource1 {
 	uint8_t vgprs                     = 0;
 	uint8_t priority                  = 0;
@@ -511,7 +437,6 @@ struct HsShaderResource2 {
 
 struct PsShaderResource1 {
 	uint8_t vgprs                    = 0;
-	uint8_t sgprs                    = 0;
 	uint8_t priority                 = 0;
 	uint8_t float_mode               = 0;
 	bool    dx10_clamp               = false;
@@ -535,7 +460,6 @@ struct PsStageRegisters {
 	uint64_t          data_addr = 0;
 	PsShaderResource1 rsrc1;
 	PsShaderResource2 rsrc2;
-	uint64_t          chksum = 0;
 };
 
 struct CsStageRegisters {
@@ -545,12 +469,13 @@ struct CsStageRegisters {
 	uint32_t num_thread_y              = 0;
 	uint32_t num_thread_z              = 0;
 	uint8_t  vgprs                     = 0;
-	uint8_t  sgprs                     = 0;
+	uint8_t  priority                  = 0;
 	uint8_t  float_mode                = 0;
 	bool     dx10_clamp                = false;
+	bool     debug_mode                = false;
 	bool     ieee_mode                 = false;
+	bool     require_forward_progress  = false;
 	bool     fp16_overflow             = false;
-	bool     bulky                     = false;
 	bool     threadgroup_configuration = false;
 	uint8_t  wave_size                 = 64;
 	bool     scratch_en                = false;
@@ -561,6 +486,7 @@ struct CsStageRegisters {
 	bool     tg_size_en                = false;
 	uint8_t  tidig_comp_cnt            = 0;
 	uint16_t lds_size                  = 0;
+	uint8_t  shared_vgprs              = 0;
 };
 
 struct EsStageRegisters {
@@ -568,21 +494,18 @@ struct EsStageRegisters {
 };
 
 struct LsStageRegisters {
-	uint64_t          data_addr = 0;
-	VsShaderResource1 rsrc1;
-	VsShaderResource2 rsrc2;
+	uint64_t data_addr = 0;
 };
 
 struct HsStageRegisters {
-	uint64_t          data_addr = 0;
+	uint64_t          data_addr      = 0;
+	uint64_t          user_data_addr = 0;
 	HsShaderResource1 rsrc1;
 	HsShaderResource2 rsrc2;
-	uint64_t          chksum = 0;
 };
 
 struct GsShaderResource1 {
 	uint8_t vgprs                     = 0;
-	uint8_t sgprs                     = 0;
 	uint8_t priority                  = 0;
 	uint8_t float_mode                = 0;
 	bool    dx10_clamp                = false;
@@ -605,10 +528,10 @@ struct GsShaderResource2 {
 };
 
 struct GsStageRegisters {
-	uint64_t          data_addr = 0;
+	uint64_t          data_addr      = 0;
+	uint64_t          user_data_addr = 0;
 	GsShaderResource1 rsrc1;
 	GsShaderResource2 rsrc2;
-	uint64_t          chksum = 0;
 };
 
 struct ShaderRegisters {
@@ -664,8 +587,7 @@ struct ShaderRegisters {
 enum class UserSgprType { Unknown, Region, Vsharp };
 
 struct UserSgprInfo {
-	// Changed from 16 to 64
-	static constexpr int SGPRS_MAX = 64;
+	static constexpr int SGPRS_MAX = 32;
 
 	uint32_t     value[SGPRS_MAX] = {0};
 	UserSgprType type[SGPRS_MAX]  = {};
@@ -712,12 +634,6 @@ public:
 	void Reset() { *this = Context(); }
 
 	void SetColorBase(uint32_t slot, const ColorBase& base) { m_render_targets[slot].base = base; }
-	void SetColorPitch(uint32_t slot, const ColorPitch& pitch) {
-		m_render_targets[slot].pitch = pitch;
-	}
-	void SetColorSlice(uint32_t slot, const ColorSlice& slice) {
-		m_render_targets[slot].slice = slice;
-	}
 	void SetColorView(uint32_t slot, const ColorView& view) { m_render_targets[slot].view = view; }
 	void SetColorInfo(uint32_t slot, const ColorInfo& info) { m_render_targets[slot].info = info; }
 	void SetColorAttrib(uint32_t slot, const ColorAttrib& attrib) {
@@ -735,15 +651,9 @@ public:
 	void SetColorCmask(uint32_t slot, const ColorCmask& cmask) {
 		m_render_targets[slot].cmask = cmask;
 	}
-	void SetColorCmaskSlice(uint32_t slot, const ColorCmaskSlice& cmask_slice) {
-		m_render_targets[slot].cmask_slice = cmask_slice;
-	}
 	void SetColorFmask(uint32_t slot, const ColorFmask& fmask) {
 		m_render_targets[slot].fmask = fmask;
 	}
-	void SetColorFmaskSlice(uint32_t slot, const ColorFmaskSlice& fmask_slice) {
-		m_render_targets[slot].fmask_slice = fmask_slice;
-	}
 	void SetColorClearWord0(uint32_t slot, const ColorClearWord0& clear_word0) {
 		m_render_targets[slot].clear_word0 = clear_word0;
 	}
@@ -753,7 +663,6 @@ public:
 	void SetColorDccAddr(uint32_t slot, const ColorDccAddr& dcc_addr) {
 		m_render_targets[slot].dcc_addr = dcc_addr;
 	}
-	void SetColorSize(uint32_t slot, const ColorSize& size) { m_render_targets[slot].size = size; }
 	[[nodiscard]] const RenderTarget& GetRenderTarget(uint32_t slot) const {
 		return m_render_targets[slot];
 	}
@@ -908,6 +817,12 @@ public:
 	void SetClipControl(const ClipControl& control) { m_clip_control = control; }
 	[[nodiscard]] const RenderControl& GetRenderControl() const { return m_render_control; }
 	void SetRenderControl(const RenderControl& control) { m_render_control = control; }
+	[[nodiscard]] const DepthRenderOverride& GetDepthRenderOverride() const {
+		return m_depth_render_override;
+	}
+	void SetDepthRenderOverride(const DepthRenderOverride& control) {
+		m_depth_render_override = control;
+	}
 	[[nodiscard]] const DepthControl& GetDepthControl() const { return m_depth_control; }
 	void SetDepthControl(const DepthControl& control) { m_depth_control = control; }
 	[[nodiscard]] const ModeControl& GetModeControl() const { return m_mode_control; }
@@ -1003,15 +918,16 @@ private:
 
 	uint32_t m_shader_stages = 0;
 
-	DepthRenderTarget m_depth_render_target;
-	RenderControl     m_render_control;
-	DepthControl      m_depth_control;
-	StencilControl    m_stencil_control;
-	StencilMask       m_stencil_mask;
-	float             m_depth_clear_value   = 0.0f;
-	float             m_depth_bounds_min    = 0.0f;
-	float             m_depth_bounds_max    = 1.0f;
-	uint8_t           m_stencil_clear_value = 0;
+	DepthRenderTarget   m_depth_render_target;
+	RenderControl       m_render_control;
+	DepthRenderOverride m_depth_render_override;
+	DepthControl        m_depth_control;
+	StencilControl      m_stencil_control;
+	StencilMask         m_stencil_mask;
+	float               m_depth_clear_value   = 0.0f;
+	float               m_depth_bounds_min    = 0.0f;
+	float               m_depth_bounds_max    = 1.0f;
+	uint8_t             m_stencil_clear_value = 0;
 
 	ModeControl m_mode_control;
 	PolyOffset  m_poly_offset;
@@ -1020,6 +936,12 @@ private:
 	ShaderRegisters m_sh_regs;
 };
 
+struct FsrView {
+	uint32_t control_points[2][4] {};
+	uint32_t alphas[2][2] {};
+	uint32_t window[2] {};
+};
+
 class UserConfig {
 public:
 	UserConfig()  = default;
@@ -1049,8 +971,17 @@ public:
 	void SetGdsOaCntl(uint32_t value) { m_gds_oa.cntl = value; }
 	void SetGdsOaCounter(uint32_t value) { m_gds_oa.counters[m_gds_oa.GetIndex()].counter = value; }
 	void SetGdsOaAddress(uint32_t value) { m_gds_oa.counters[m_gds_oa.GetIndex()].address = value; }
+	void SetFsrControlPoint(uint32_t axis, uint32_t index, uint32_t value) {
+		m_fsr_view.control_points[axis][index] = value;
+	}
+	void SetFsrAlpha(uint32_t axis, uint32_t index, uint32_t value) {
+		m_fsr_view.alphas[axis][index] = value;
+	}
+	void SetFsrWindow(uint32_t index, uint32_t value) { m_fsr_view.window[index] = value; }
+	[[nodiscard]] const FsrView& GetFsrView() const { return m_fsr_view; }
 
 private:
+	FsrView m_fsr_view;
 	Prospero::PrimitiveType m_prim_type               = Prospero::PrimitiveType::kNone;
 	uint32_t                m_index_offset            = 0;
 	uint32_t                m_object_id               = 0;
@@ -1072,21 +1003,22 @@ public:
 
 	void SetEsShaderBase(uint64_t addr) { m_vs.es_regs.data_addr = addr; }
 	void SetLsShaderBase(uint64_t addr) { m_vs.ls_regs.data_addr = addr; }
-	void SetLsShaderResource1(const VsShaderResource1& rsrc1) { m_vs.ls_regs.rsrc1 = rsrc1; }
-	void SetLsShaderResource2(const VsShaderResource2& rsrc2) { m_vs.ls_regs.rsrc2 = rsrc2; }
 	void SetHsShaderBase(uint64_t addr) { m_vs.hs_regs.data_addr = addr; }
+	void SetHsUserDataAddress(uint32_t word, uint32_t value) {
+		SetUserDataAddressWord(m_vs.hs_regs.user_data_addr, word, value);
+	}
 	void SetHsShaderResource1(const HsShaderResource1& rsrc1) { m_vs.hs_regs.rsrc1 = rsrc1; }
 	void SetHsShaderResource2(const HsShaderResource2& rsrc2) { m_vs.hs_regs.rsrc2 = rsrc2; }
-	void SetHsShaderChksum(uint32_t value) { m_vs.hs_regs.chksum = value; }
 	void SetGsShaderBase(uint64_t addr) { m_vs.gs_regs.data_addr = addr; }
+	void SetGsUserDataAddress(uint32_t word, uint32_t value) {
+		SetUserDataAddressWord(m_vs.gs_regs.user_data_addr, word, value);
+	}
 	void SetGsShaderResource1(const GsShaderResource1& rsrc1) { m_vs.gs_regs.rsrc1 = rsrc1; }
 	void SetGsShaderResource2(const GsShaderResource2& rsrc2) { m_vs.gs_regs.rsrc2 = rsrc2; }
-	void SetGsShaderChksum(uint32_t value) { m_vs.gs_regs.chksum = value; }
 
 	void SetPsShaderBase(uint64_t addr) { m_ps.ps_regs.data_addr = addr; }
 	void SetPsShaderResource1(const PsShaderResource1& rsrc1) { m_ps.ps_regs.rsrc1 = rsrc1; }
 	void SetPsShaderResource2(const PsShaderResource2& rsrc2) { m_ps.ps_regs.rsrc2 = rsrc2; }
-	void SetPsShaderChksum(uint32_t value) { m_ps.ps_regs.chksum = value; }
 
 	void SetCsShader(const CsStageRegisters& cs_regs) { m_cs.cs_regs = cs_regs; }
 	void SetCsWaveSize(uint8_t wave_size) { m_cs.cs_regs.wave_size = wave_size; }
@@ -1121,6 +1053,12 @@ public:
 	[[nodiscard]] const ComputeShaderInfo& GetCs() const { return m_cs; }
 
 private:
+	static void SetUserDataAddressWord(uint64_t& address, uint32_t word, uint32_t value) {
+		const auto shift = word * 32u;
+		address = (address & ~(uint64_t {0xffffffffu} << shift)) |
+		          (static_cast<uint64_t>(value) << shift);
+	}
+
 	VertexShaderInfo  m_vs;
 	PixelShaderInfo   m_ps;
 	ComputeShaderInfo m_cs;

src/graphics/guest_gpu/pm4.h

@@ -98,6 +98,10 @@ constexpr uint32_t DB_RENDER_CONTROL_DEPTH_CLEAR_ENABLE_SHIFT       = 0;
 constexpr uint32_t DB_RENDER_CONTROL_DEPTH_CLEAR_ENABLE_MASK        = 0x1;
 constexpr uint32_t DB_RENDER_CONTROL_STENCIL_CLEAR_ENABLE_SHIFT     = 1;
 constexpr uint32_t DB_RENDER_CONTROL_STENCIL_CLEAR_ENABLE_MASK      = 0x1;
+constexpr uint32_t DB_RENDER_CONTROL_COPY_DEPTH_TO_COLOR_SHIFT      = 2;
+constexpr uint32_t DB_RENDER_CONTROL_COPY_DEPTH_TO_COLOR_MASK       = 0x1;
+constexpr uint32_t DB_RENDER_CONTROL_COPY_STENCIL_TO_COLOR_SHIFT    = 3;
+constexpr uint32_t DB_RENDER_CONTROL_COPY_STENCIL_TO_COLOR_MASK     = 0x1;
 constexpr uint32_t DB_RENDER_CONTROL_RESUMMARIZE_ENABLE_SHIFT       = 4;
 constexpr uint32_t DB_RENDER_CONTROL_RESUMMARIZE_ENABLE_MASK        = 0x1;
 constexpr uint32_t DB_RENDER_CONTROL_STENCIL_COMPRESS_DISABLE_SHIFT = 5;
@@ -163,22 +167,6 @@ constexpr uint32_t PA_SC_SCREEN_SCISSOR_BR_BR_Y_MASK  = 0xFFFF;
 
 constexpr uint32_t DB_DFSM_CONTROL = 0xE;
 
-constexpr uint32_t DB_DEPTH_INFO                          = 0xF;
-constexpr uint32_t DB_DEPTH_INFO_ADDR5_SWIZZLE_MASK_SHIFT = 0;
-constexpr uint32_t DB_DEPTH_INFO_ADDR5_SWIZZLE_MASK_MASK  = 0xF;
-constexpr uint32_t DB_DEPTH_INFO_ARRAY_MODE_SHIFT         = 4;
-constexpr uint32_t DB_DEPTH_INFO_ARRAY_MODE_MASK          = 0xF;
-constexpr uint32_t DB_DEPTH_INFO_PIPE_CONFIG_SHIFT        = 8;
-constexpr uint32_t DB_DEPTH_INFO_PIPE_CONFIG_MASK         = 0x1F;
-constexpr uint32_t DB_DEPTH_INFO_BANK_WIDTH_SHIFT         = 13;
-constexpr uint32_t DB_DEPTH_INFO_BANK_WIDTH_MASK          = 0x3;
-constexpr uint32_t DB_DEPTH_INFO_BANK_HEIGHT_SHIFT        = 15;
-constexpr uint32_t DB_DEPTH_INFO_BANK_HEIGHT_MASK         = 0x3;
-constexpr uint32_t DB_DEPTH_INFO_MACRO_TILE_ASPECT_SHIFT  = 17;
-constexpr uint32_t DB_DEPTH_INFO_MACRO_TILE_ASPECT_MASK   = 0x3;
-constexpr uint32_t DB_DEPTH_INFO_NUM_BANKS_SHIFT          = 19;
-constexpr uint32_t DB_DEPTH_INFO_NUM_BANKS_MASK           = 0x3;
-
 constexpr uint32_t DB_Z_INFO                               = 0x10;
 constexpr uint32_t DB_Z_INFO_FORMAT_SHIFT                  = 0;
 constexpr uint32_t DB_Z_INFO_FORMAT_MASK                   = 0x3;
@@ -222,16 +210,6 @@ constexpr uint32_t DB_STENCIL_READ_BASE  = 0x13;
 constexpr uint32_t DB_Z_WRITE_BASE       = 0x14;
 constexpr uint32_t DB_STENCIL_WRITE_BASE = 0x15;
 
-constexpr uint32_t DB_DEPTH_SIZE                       = 0x16;
-constexpr uint32_t DB_DEPTH_SIZE_PITCH_TILE_MAX_SHIFT  = 0;
-constexpr uint32_t DB_DEPTH_SIZE_PITCH_TILE_MAX_MASK   = 0x7FF;
-constexpr uint32_t DB_DEPTH_SIZE_HEIGHT_TILE_MAX_SHIFT = 11;
-constexpr uint32_t DB_DEPTH_SIZE_HEIGHT_TILE_MAX_MASK  = 0x7FF;
-
-constexpr uint32_t DB_DEPTH_SLICE                      = 0x17;
-constexpr uint32_t DB_DEPTH_SLICE_SLICE_TILE_MAX_SHIFT = 0;
-constexpr uint32_t DB_DEPTH_SLICE_SLICE_TILE_MAX_MASK  = 0x3FFFFF;
-
 constexpr uint32_t DB_Z_READ_BASE_HI                  = 0x1A;
 constexpr uint32_t DB_STENCIL_READ_BASE_HI            = 0x1B;
 constexpr uint32_t DB_Z_WRITE_BASE_HI                 = 0x1C;
@@ -424,11 +402,6 @@ constexpr uint32_t DB_DEPTH_CONTROL_STENCILFUNC_SHIFT                        = 8
 constexpr uint32_t DB_DEPTH_CONTROL_STENCILFUNC_MASK                         = 0x7;
 constexpr uint32_t DB_DEPTH_CONTROL_STENCILFUNC_BF_SHIFT                     = 20;
 constexpr uint32_t DB_DEPTH_CONTROL_STENCILFUNC_BF_MASK                      = 0x7;
-constexpr uint32_t DB_DEPTH_CONTROL_ENABLE_COLOR_WRITES_ON_DEPTH_FAIL_SHIFT  = 30;
-constexpr uint32_t DB_DEPTH_CONTROL_ENABLE_COLOR_WRITES_ON_DEPTH_FAIL_MASK   = 0x1;
-constexpr uint32_t DB_DEPTH_CONTROL_DISABLE_COLOR_WRITES_ON_DEPTH_PASS_SHIFT = 31;
-constexpr uint32_t DB_DEPTH_CONTROL_DISABLE_COLOR_WRITES_ON_DEPTH_PASS_MASK  = 0x1;
-
 constexpr uint32_t DB_EQAA                                  = 0x201;
 constexpr uint32_t DB_EQAA_MAX_ANCHOR_SAMPLES_SHIFT         = 0;
 constexpr uint32_t DB_EQAA_MAX_ANCHOR_SAMPLES_MASK          = 0x7;
@@ -560,21 +533,9 @@ constexpr uint32_t VGT_DRAW_PAYLOAD_CNTL  = 0x2A6;
 constexpr uint32_t VGT_ESGS_RING_ITEMSIZE = 0x2AB;
 constexpr uint32_t VGT_REUSE_OFF          = 0x2AD;
 
-constexpr uint32_t DB_HTILE_SURFACE                               = 0x2AF;
-constexpr uint32_t DB_HTILE_SURFACE_LINEAR_SHIFT                  = 0;
-constexpr uint32_t DB_HTILE_SURFACE_LINEAR_MASK                   = 0x1;
-constexpr uint32_t DB_HTILE_SURFACE_FULL_CACHE_SHIFT              = 1;
-constexpr uint32_t DB_HTILE_SURFACE_FULL_CACHE_MASK               = 0x1;
-constexpr uint32_t DB_HTILE_SURFACE_HTILE_USES_PRELOAD_WIN_SHIFT  = 2;
-constexpr uint32_t DB_HTILE_SURFACE_HTILE_USES_PRELOAD_WIN_MASK   = 0x1;
-constexpr uint32_t DB_HTILE_SURFACE_PRELOAD_SHIFT                 = 3;
-constexpr uint32_t DB_HTILE_SURFACE_PRELOAD_MASK                  = 0x1;
-constexpr uint32_t DB_HTILE_SURFACE_PREFETCH_WIDTH_SHIFT          = 4;
-constexpr uint32_t DB_HTILE_SURFACE_PREFETCH_WIDTH_MASK           = 0x3F;
-constexpr uint32_t DB_HTILE_SURFACE_PREFETCH_HEIGHT_SHIFT         = 10;
-constexpr uint32_t DB_HTILE_SURFACE_PREFETCH_HEIGHT_MASK          = 0x3F;
-constexpr uint32_t DB_HTILE_SURFACE_DST_OUTSIDE_ZERO_TO_ONE_SHIFT = 16;
-constexpr uint32_t DB_HTILE_SURFACE_DST_OUTSIDE_ZERO_TO_ONE_MASK  = 0x1;
+constexpr uint32_t DB_SHADING_RATE_ENCODING                             = 0x2AF;
+constexpr uint32_t DB_SHADING_RATE_ENCODING_SHADING_RATE_ENCODING_SHIFT = 19;
+constexpr uint32_t DB_SHADING_RATE_ENCODING_SHADING_RATE_ENCODING_MASK  = 0x3;
 
 constexpr uint32_t DB_SRESULTS_COMPARE_STATE0                          = 0x2B0;
 constexpr uint32_t DB_SRESULTS_COMPARE_STATE1                          = 0x2B1;
@@ -621,7 +582,6 @@ constexpr uint32_t PA_SC_SHADER_CONTROL                  = 0x310;
 constexpr uint32_t PA_SC_BINNER_CNTL_0                   = 0x311;
 constexpr uint32_t PA_SC_BINNER_CNTL_1                   = 0x312;
 constexpr uint32_t PA_SC_CONSERVATIVE_RASTERIZATION_CNTL = 0x313;
-constexpr uint32_t PA_SC_NGG_MODE_CNTL                   = 0x314;
 constexpr uint32_t CB_COLOR0_BASE                        = 0x318;
 
 constexpr uint32_t CB_COLOR0_VIEW                   = 0x31B;
@@ -649,24 +609,13 @@ constexpr uint32_t CB_COLOR0_INFO_BLEND_BYPASS_SHIFT              = 16;
 constexpr uint32_t CB_COLOR0_INFO_BLEND_BYPASS_MASK               = 0x1;
 constexpr uint32_t CB_COLOR0_INFO_ROUND_MODE_SHIFT                = 18;
 constexpr uint32_t CB_COLOR0_INFO_ROUND_MODE_MASK                 = 0x1;
-constexpr uint32_t CB_COLOR0_INFO_CMASK_IS_LINEAR_SHIFT           = 19;
-constexpr uint32_t CB_COLOR0_INFO_CMASK_IS_LINEAR_MASK            = 0x1;
 constexpr uint32_t CB_COLOR0_INFO_FMASK_COMPRESSION_DISABLE_SHIFT = 26;
 constexpr uint32_t CB_COLOR0_INFO_FMASK_COMPRESSION_DISABLE_MASK  = 0x1;
 constexpr uint32_t CB_COLOR0_INFO_FMASK_COMPRESS_1FRAG_ONLY_SHIFT = 27;
 constexpr uint32_t CB_COLOR0_INFO_FMASK_COMPRESS_1FRAG_ONLY_MASK  = 0x1;
 constexpr uint32_t CB_COLOR0_INFO_DCC_ENABLE_SHIFT                = 28;
 constexpr uint32_t CB_COLOR0_INFO_DCC_ENABLE_MASK                 = 0x1;
-constexpr uint32_t CB_COLOR0_INFO_CMASK_ADDR_TYPE_SHIFT           = 29;
-constexpr uint32_t CB_COLOR0_INFO_CMASK_ADDR_TYPE_MASK            = 0x3;
-constexpr uint32_t CB_COLOR0_INFO_ALT_TILE_MODE_SHIFT             = 31;
-constexpr uint32_t CB_COLOR0_INFO_ALT_TILE_MODE_MASK              = 0x1;
-
 constexpr uint32_t CB_COLOR0_ATTRIB                             = 0x31D;
-constexpr uint32_t CB_COLOR0_ATTRIB_TILE_MODE_INDEX_SHIFT       = 0;
-constexpr uint32_t CB_COLOR0_ATTRIB_TILE_MODE_INDEX_MASK        = 0x1F;
-constexpr uint32_t CB_COLOR0_ATTRIB_FMASK_TILE_MODE_INDEX_SHIFT = 5;
-constexpr uint32_t CB_COLOR0_ATTRIB_FMASK_TILE_MODE_INDEX_MASK  = 0x1F;
 constexpr uint32_t CB_COLOR0_ATTRIB_NUM_SAMPLES_SHIFT           = 12;
 constexpr uint32_t CB_COLOR0_ATTRIB_NUM_SAMPLES_MASK            = 0x7;
 constexpr uint32_t CB_COLOR0_ATTRIB_NUM_FRAGMENTS_SHIFT         = 15;
@@ -681,8 +630,6 @@ constexpr uint32_t CB_COLOR0_DCC_CONTROL_KEY_CLEAR_ENABLE_SHIFT
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_KEY_CLEAR_ENABLE_MASK                  = 0x1;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_MAX_UNCOMPRESSED_BLOCK_SIZE_SHIFT      = 2;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_MAX_UNCOMPRESSED_BLOCK_SIZE_MASK       = 0x3;
-constexpr uint32_t CB_COLOR0_DCC_CONTROL_MIN_COMPRESSED_BLOCK_SIZE_SHIFT        = 4;
-constexpr uint32_t CB_COLOR0_DCC_CONTROL_MIN_COMPRESSED_BLOCK_SIZE_MASK         = 0x1;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_MAX_COMPRESSED_BLOCK_SIZE_SHIFT        = 5;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_MAX_COMPRESSED_BLOCK_SIZE_MASK         = 0x3;
 constexpr uint32_t CB_COLOR0_DCC_CONTROL_COLOR_TRANSFORM_SHIFT                  = 7;
@@ -739,31 +686,25 @@ constexpr uint32_t CB_COLOR0_ATTRIB3_CMASK_PIPE_ALIGNED_SHIFT = 26;
 constexpr uint32_t CB_COLOR0_ATTRIB3_CMASK_PIPE_ALIGNED_MASK  = 0x1;
 constexpr uint32_t CB_COLOR0_ATTRIB3_DCC_PIPE_ALIGNED_SHIFT   = 30;
 constexpr uint32_t CB_COLOR0_ATTRIB3_DCC_PIPE_ALIGNED_MASK    = 0x1;
+constexpr uint32_t CB_COLOR0_ATTRIB3_WRITE_VRS_RATE_HINT_TO_CMASK_SHIFT = 31;
+constexpr uint32_t CB_COLOR0_ATTRIB3_WRITE_VRS_RATE_HINT_TO_CMASK_MASK  = 0x1;
 
 constexpr uint32_t CB_COLOR7_ATTRIB3 = 0x3BF;
 
-/* Fake codes. Don't exist on real HW */
-
-constexpr uint32_t CX_NOP = 0x800003FF;
+constexpr uint32_t CX_NOP = 0xDB;
 
 constexpr uint32_t CX_NUM = 0x3FF + 1;
 
 /* Shader registers */
 
-constexpr uint32_t SPI_SHADER_TBA_LO_PS     = 0x0;
-constexpr uint32_t SPI_SHADER_PGM_RSRC4_PS  = 0x1;
-constexpr uint32_t SPI_SHADER_TMA_LO_PS     = 0x2;
-constexpr uint32_t SPI_SHADER_TMA_HI_PS     = 0x3;
-constexpr uint32_t SPI_SHADER_PGM_CHKSUM_PS = 0x6;
-constexpr uint32_t SPI_SHADER_PGM_RSRC3_PS  = 0x7;
-constexpr uint32_t SPI_SHADER_PGM_LO_PS     = 0x8;
-constexpr uint32_t SPI_SHADER_PGM_HI_PS     = 0x9;
+constexpr uint32_t SPI_SHADER_PACE_ID_PS            = 0x6;
+constexpr uint32_t SPI_GRAPHICS_SHADER_CONTROL_PS    = 0x7;
+constexpr uint32_t SPI_SHADER_PGM_LO_PS              = 0x8;
+constexpr uint32_t SPI_SHADER_PGM_HI_PS              = 0x9;
 
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS                        = 0xA;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_VGPRS_SHIFT            = 0;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_VGPRS_MASK             = 0x3F;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_SGPRS_SHIFT            = 6;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_SGPRS_MASK             = 0xF;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_PRIORITY_SHIFT         = 10;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_PRIORITY_MASK          = 0x3;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_FLOAT_MODE_SHIFT       = 12;
@@ -800,14 +741,11 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_SHARED_VGPR_CNT_MASK         = 0xF;
 constexpr uint32_t SPI_SHADER_USER_DATA_PS_0  = 0xC;
 constexpr uint32_t SPI_SHADER_USER_DATA_PS_15 = 0x1B;
 constexpr uint32_t SPI_SHADER_USER_DATA_PS_31 = 0x2B;
-constexpr uint32_t SPI_SHADER_REQ_CTRL_PS     = 0x30;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_PS_0 = 0x32;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_PS_3 = 0x35;
 
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_VGPRS_SHIFT           = 0;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_VGPRS_MASK            = 0x3F;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_SGPRS_SHIFT           = 6;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_SGPRS_MASK            = 0xF;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_PRIORITY_SHIFT        = 10;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_PRIORITY_MASK         = 0x3;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_VS_FLOAT_MODE_SHIFT      = 12;
@@ -838,19 +776,20 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC2_VS_USER_SGPR_MSB_MASK    = 0x1;
 constexpr uint32_t SPI_SHADER_PGM_RSRC2_VS_SHARED_VGPR_CNT_SHIFT = 28;
 constexpr uint32_t SPI_SHADER_PGM_RSRC2_VS_SHARED_VGPR_CNT_MASK  = 0xF;
 
-constexpr uint32_t SPI_SHADER_PGM_CHKSUM_GS        = 0x80;
+constexpr uint32_t SPI_SHADER_PACE_ID_GS           = 0x80;
+// AGC compiler records use this numeric tag as opaque metadata. It is not a shader checksum
+// register and intentionally has no checksum-state consumer in HardwareContext.
+constexpr uint32_t SPI_SHADER_PGM_CHKSUM_GS        = SPI_SHADER_PACE_ID_GS;
 constexpr uint32_t SPI_SHADER_PGM_RSRC4_GS         = 0x81;
 constexpr uint32_t SPI_SHADER_USER_DATA_ADDR_LO_GS = 0x82;
 constexpr uint32_t SPI_SHADER_USER_DATA_ADDR_HI_GS = 0x83;
-constexpr uint32_t SPI_SHADER_PGM_RSRC3_GS         = 0x87;
+constexpr uint32_t SPI_GRAPHICS_SHADER_CONTROL_GS  = 0x87;
 constexpr uint32_t SPI_SHADER_PGM_LO_GS            = 0x88;
 constexpr uint32_t SPI_SHADER_PGM_HI_GS            = 0x89;
 
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS                        = 0x8A;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_VGPRS_SHIFT            = 0;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_VGPRS_MASK             = 0x3F;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_SGPRS_SHIFT            = 6;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_SGPRS_MASK             = 0xF;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_PRIORITY_SHIFT         = 10;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_PRIORITY_MASK          = 0x3;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_GS_FLOAT_MODE_SHIFT       = 12;
@@ -891,22 +830,16 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC2_GS_SHARED_VGPR_CNT_MASK   = 0xF;
 constexpr uint32_t SPI_SHADER_USER_DATA_GS_0                      = 0x8C;
 constexpr uint32_t SPI_SHADER_USER_DATA_GS_15                     = 0x9B;
 constexpr uint32_t SPI_SHADER_USER_DATA_GS_31                     = 0xAB;
-constexpr uint32_t SPI_SHADER_REQ_CTRL_ESGS                       = 0xB0;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_ESGS_0                   = 0xB2;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_ESGS_3                   = 0xB5;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_ES_GS                     = 0xBC;
 constexpr uint32_t SPI_SHADER_PGM_LO_ES                           = 0xC8;
 constexpr uint32_t SPI_SHADER_PGM_HI_ES                           = 0xC9;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_ES                        = 0xCA;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_ES                        = 0xCB;
-constexpr uint32_t SPI_SHADER_USER_DATA_ES_0                      = 0xCC;
-constexpr uint32_t SPI_SHADER_USER_DATA_ES_15                     = 0xDB;
-constexpr uint32_t SPI_SHADER_USER_DATA_ES_31                     = 0xEB;
+// Private AGC registers used by shader binaries submitted through SET_SH_REG_INDIRECT.
 constexpr uint32_t SPI_SHADER_PGM_CHKSUM_HS                       = 0x100;
 constexpr uint32_t SPI_SHADER_PGM_RSRC4_HS                        = 0x101;
 constexpr uint32_t SPI_SHADER_USER_DATA_ADDR_LO_HS                = 0x102;
 constexpr uint32_t SPI_SHADER_USER_DATA_ADDR_HI_HS                = 0x103;
-constexpr uint32_t SPI_SHADER_PGM_RSRC3_HS                        = 0x107;
+constexpr uint32_t SPI_GRAPHICS_SHADER_CONTROL_HS                 = 0x107;
 constexpr uint32_t SPI_SHADER_PGM_LO_HS                           = 0x108;
 constexpr uint32_t SPI_SHADER_PGM_HI_HS                           = 0x109;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS                        = 0x10A;
@@ -944,13 +877,10 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_SHARED_VGPR_CNT_MASK   = 0xF;
 constexpr uint32_t SPI_SHADER_USER_DATA_HS_0                      = 0x10C;
 constexpr uint32_t SPI_SHADER_USER_DATA_HS_15                     = 0x11B;
 constexpr uint32_t SPI_SHADER_USER_DATA_HS_31                     = 0x12B;
-constexpr uint32_t SPI_SHADER_REQ_CTRL_LSHS                       = 0x130;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_LSHS_0                   = 0x132;
 constexpr uint32_t SPI_SHADER_USER_ACCUM_LSHS_3                   = 0x135;
 constexpr uint32_t SPI_SHADER_PGM_LO_LS                           = 0x148;
 constexpr uint32_t SPI_SHADER_PGM_HI_LS                           = 0x149;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_LS                        = 0x14A;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_LS                        = 0x14B;
 constexpr uint32_t COMPUTE_START_X                                = 0x204;
 constexpr uint32_t COMPUTE_START_Y                                = 0x205;
 constexpr uint32_t COMPUTE_START_Z                                = 0x206;
@@ -959,28 +889,26 @@ constexpr uint32_t COMPUTE_NUM_THREAD_Y                           = 0x208;
 constexpr uint32_t COMPUTE_NUM_THREAD_Z                           = 0x209;
 constexpr uint32_t COMPUTE_PGM_LO                                 = 0x20C;
 constexpr uint32_t COMPUTE_PGM_HI                                 = 0x20D;
-constexpr uint32_t COMPUTE_TBA_LO                                 = 0x20E;
-constexpr uint32_t COMPUTE_TBA_HI                                 = 0x20F;
-constexpr uint32_t COMPUTE_TMA_LO                                 = 0x210;
-constexpr uint32_t COMPUTE_TMA_HI                                 = 0x211;
 
 constexpr uint32_t COMPUTE_PGM_RSRC1                  = 0x212;
 constexpr uint32_t COMPUTE_PGM_RSRC1_VGPRS_SHIFT      = 0;
 constexpr uint32_t COMPUTE_PGM_RSRC1_VGPRS_MASK       = 0x3F;
-constexpr uint32_t COMPUTE_PGM_RSRC1_SGPRS_SHIFT      = 6;
-constexpr uint32_t COMPUTE_PGM_RSRC1_SGPRS_MASK       = 0xF;
+constexpr uint32_t COMPUTE_PGM_RSRC1_PRIORITY_SHIFT   = 10;
+constexpr uint32_t COMPUTE_PGM_RSRC1_PRIORITY_MASK    = 0x3;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FLOAT_MODE_SHIFT = 12;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FLOAT_MODE_MASK  = 0xFF;
 constexpr uint32_t COMPUTE_PGM_RSRC1_DX10_CLAMP_SHIFT = 21;
 constexpr uint32_t COMPUTE_PGM_RSRC1_DX10_CLAMP_MASK  = 0x1;
+constexpr uint32_t COMPUTE_PGM_RSRC1_DEBUG_MODE_SHIFT = 22;
+constexpr uint32_t COMPUTE_PGM_RSRC1_DEBUG_MODE_MASK  = 0x1;
 constexpr uint32_t COMPUTE_PGM_RSRC1_IEEE_MODE_SHIFT  = 23;
 constexpr uint32_t COMPUTE_PGM_RSRC1_IEEE_MODE_MASK   = 0x1;
-constexpr uint32_t COMPUTE_PGM_RSRC1_BULKY_SHIFT      = 24;
-constexpr uint32_t COMPUTE_PGM_RSRC1_BULKY_MASK       = 0x1;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FP16_OVFL_SHIFT  = 26;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FP16_OVFL_MASK   = 0x1;
 constexpr uint32_t COMPUTE_PGM_RSRC1_WGP_MODE_SHIFT   = 29;
 constexpr uint32_t COMPUTE_PGM_RSRC1_WGP_MODE_MASK    = 0x1;
+constexpr uint32_t COMPUTE_PGM_RSRC1_FWD_PROGRESS_SHIFT = 31;
+constexpr uint32_t COMPUTE_PGM_RSRC1_FWD_PROGRESS_MASK  = 0x1;
 
 constexpr uint32_t COMPUTE_DISPATCH_INITIATOR_CS_W32_EN_SHIFT = 15;
 constexpr uint32_t COMPUTE_DISPATCH_INITIATOR_CS_W32_EN_MASK  = 0x1;
@@ -1011,21 +939,16 @@ constexpr uint32_t COMPUTE_PGM_RSRC2_LDS_SIZE_SHIFT       = 15;
 constexpr uint32_t COMPUTE_PGM_RSRC2_LDS_SIZE_MASK        = 0x1FF;
 
 constexpr uint32_t COMPUTE_RESOURCE_LIMITS    = 0x215;
-constexpr uint32_t COMPUTE_DESTINATION_EN_SE0 = 0x216;
-constexpr uint32_t COMPUTE_DESTINATION_EN_SE1 = 0x217;
 constexpr uint32_t COMPUTE_TMPRING_SIZE       = 0x218;
-constexpr uint32_t COMPUTE_DESTINATION_EN_SE2 = 0x219;
-constexpr uint32_t COMPUTE_DESTINATION_EN_SE3 = 0x21A;
 constexpr uint32_t COMPUTE_USER_ACCUM_0       = 0x224;
 constexpr uint32_t COMPUTE_PGM_RSRC3          = 0x228;
-constexpr uint32_t COMPUTE_SHADER_CHKSUM      = 0x22A;
+constexpr uint32_t COMPUTE_PGM_RSRC3_SHARED_VGPRS_SHIFT = 0;
+constexpr uint32_t COMPUTE_PGM_RSRC3_SHARED_VGPRS_MASK  = 0xF;
+constexpr uint32_t COMPUTE_PACE_ID             = 0x22A;
 constexpr uint32_t COMPUTE_USER_DATA_0        = 0x240;
 constexpr uint32_t COMPUTE_USER_DATA_15       = 0x24F;
-constexpr uint32_t COMPUTE_DISPATCH_TUNNEL    = 0x27D;
 
-/* Fake codes. Don't exist on real HW */
-
-constexpr uint32_t SH_NOP = 0x800002FF;
+constexpr uint32_t SH_NOP = 0x280;
 
 constexpr uint32_t SH_NUM = 0x2FF + 1;
 
@@ -1042,6 +965,7 @@ constexpr uint32_t GE_INDX_OFFSET            = 0x24A;
 constexpr uint32_t GE_MULTI_PRIM_IB_RESET_EN = 0x24B;
 constexpr uint32_t VGT_HS_OFFCHIP_PARAM      = 0x24F;
 constexpr uint32_t VGT_TF_MEMORY_BASE        = 0x250;
+constexpr uint32_t IA_MULTI_VGT_PARAM        = 0x258;
 
 constexpr uint32_t GE_CNTL                     = 0x25B;
 constexpr uint32_t GE_CNTL_PRIM_GRP_SIZE_SHIFT = 0;
@@ -1053,7 +977,7 @@ constexpr uint32_t GE_USER_VGPR1  = 0x25C;
 constexpr uint32_t GE_USER_VGPR2  = 0x25D;
 constexpr uint32_t GE_USER_VGPR3  = 0x25E;
 constexpr uint32_t GE_STEREO_CNTL = 0x25F;
-constexpr uint32_t GE_PC_ALLOC    = 0x260;
+constexpr uint32_t UC_PARAMETER_OVERSUBSCRIPTION = 0x260;
 
 constexpr uint32_t GE_USER_VGPR_EN                     = 0x262;
 constexpr uint32_t GE_USER_VGPR_EN_EN_USER_VGPR1_SHIFT = 0;
@@ -1063,28 +987,21 @@ constexpr uint32_t GE_USER_VGPR_EN_EN_USER_VGPR2_MASK  = 0x1;
 constexpr uint32_t GE_USER_VGPR_EN_EN_USER_VGPR3_SHIFT = 2;
 constexpr uint32_t GE_USER_VGPR_EN_EN_USER_VGPR3_MASK  = 0x1;
 
-constexpr uint32_t IA_MULTI_VGT_PARAM       = 0x258;
+constexpr uint32_t FSR_WINDOW_LEFT           = 0x2BC;
+constexpr uint32_t FSR_CONTROL_POINTS_LEFT_X = 0x2C4;
+constexpr uint32_t FSR_CONTROL_POINTS_LEFT_Y = 0x2CC;
+constexpr uint32_t FSR_ALPHA_LEFT_X          = 0x2E4;
+constexpr uint32_t FSR_ALPHA_LEFT_Y          = 0x2E8;
+
 constexpr uint32_t TA_CS_BC_BASE_ADDR       = 0x380;
 constexpr uint32_t TA_CS_BC_BASE_ADDR_HI    = 0x381;
-constexpr uint32_t TEXTURE_GRADIENT_FACTORS = 0x382;
+constexpr uint32_t TEXTURE_GRADIENT_FACTORS = 0x383;
+constexpr uint32_t TEXTURE_GRADIENT_CONTROL = 0x384;
 constexpr uint32_t GDS_OA_CNTL              = 0x41D;
 constexpr uint32_t GDS_OA_COUNTER           = 0x41E;
 constexpr uint32_t GDS_OA_ADDRESS           = 0x41F;
 
-/* Fake codes. Don't exist on real HW */
-
-constexpr uint32_t FSR_EXTEND_SUBPIXEL_ROUNDING = 0x80003FF4;
-constexpr uint32_t FSR_ALPHA_VALUE0             = 0x80003FF5;
-constexpr uint32_t FSR_ALPHA_VALUE1             = 0x80003FF6;
-constexpr uint32_t FSR_CONTROL_POINT0           = 0x80003FF7;
-constexpr uint32_t FSR_CONTROL_POINT1           = 0x80003FF8;
-constexpr uint32_t FSR_CONTROL_POINT2           = 0x80003FF9;
-constexpr uint32_t FSR_CONTROL_POINT3           = 0x80003FFA;
-constexpr uint32_t FSR_WINDOW0                  = 0x80003FFB;
-constexpr uint32_t FSR_WINDOW1                  = 0x80003FFC;
-constexpr uint32_t TEXTURE_GRADIENT_CONTROL     = 0x80003FFD;
-constexpr uint32_t MEMORY_MAPPING_MASK          = 0x80003FFE;
-constexpr uint32_t UC_NOP                       = 0x80003FFF;
+constexpr uint32_t UC_NOP = 0xA2;
 
 constexpr uint32_t UC_NUM = 0x3FFF + 1;
 

src/graphics/guest_gpu/tile.cpp

@@ -1,5 +1,6 @@
 #include "graphics/guest_gpu/tile.h"
 
+#include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/profiler.h"
@@ -15,10 +16,6 @@
 
 namespace Libs::Graphics {
 
-static uint32_t AlignUp(uint32_t value, uint32_t alignment) {
-	return (value + alignment - 1u) & ~(alignment - 1u);
-}
-
 static uint32_t ShiftCeil(uint32_t value, uint32_t shift) {
 	return static_cast<uint32_t>((static_cast<uint64_t>(value) + (1ull << shift) - 1ull) >> shift);
 }
@@ -33,7 +30,7 @@ static uint32_t CalcLinearAlignedLevelPitch(uint32_t base_width, uint32_t base_h
 	const uint32_t level_width  = ShiftCeil(base_width, level);
 	const uint32_t level_height = ShiftCeil(base_height, level);
 	const uint32_t padded_width =
-	    AlignUp(std::max(level_width, 1u), CalcLinearBlockWidth(bytes_per_element));
+	    Common::AlignUp(std::max(level_width, 1u), CalcLinearBlockWidth(bytes_per_element));
 	const uint64_t size =
 	    static_cast<uint64_t>(padded_width) * std::max(level_height, 1u) * bytes_per_element;
 	EXIT_NOT_IMPLEMENTED(size > 0xffffffffull);
@@ -61,7 +58,7 @@ static uint32_t SetLinearMipChainLayout(uint32_t levels, const uint32_t* mip_pit
 		offset += mip_size[level];
 	}
 
-	return AlignUp(offset, 256u);
+	return Common::AlignUp(offset, 256u);
 }
 
 struct Log2BlockDimensions {
@@ -366,8 +363,8 @@ bool TileGetTiledTextureLayout(const TileSurfaceDescription& description, TileSu
 		    ((description.width >> level) + texture.texel_width - 1u) / texture.texel_width, 1u);
 		mip.height = std::max(
 		    ((description.height >> level) + texture.texel_height - 1u) / texture.texel_height, 1u);
-		mip.padded_width  = AlignUp(std::max(ShiftCeil(width0, level), 1u), block.block_width);
-		mip.padded_height = AlignUp(std::max(ShiftCeil(height0, level), 1u), block.block_height);
+		mip.padded_width  = Common::AlignUp(std::max(ShiftCeil(width0, level), 1u), block.block_width);
+		mip.padded_height = Common::AlignUp(std::max(ShiftCeil(height0, level), 1u), block.block_height);
 		mip.size = static_cast<uint64_t>(block.block_depth) * mip.padded_width * mip.padded_height *
 		           block.bytes_per_element;
 		result.block_slice_size += mip.size;
@@ -1121,8 +1118,8 @@ static bool TileGetHtileSize(uint32_t width, uint32_t height, TileSizeAlign& hti
 	}
 	// Prospero HTile stores one DWORD per depth tile. Its 32 KiB allocation blocks cover
 	// 1024x512 pixels, independently of the attachment's fragment count.
-	const uint64_t size = static_cast<uint64_t>(AlignUp(width, 1024u) / 1024u) *
-	                      (AlignUp(height, 512u) / 512u) * 32768u;
+	const uint64_t size = static_cast<uint64_t>(Common::AlignUp(width, 1024u) / 1024u) *
+	                      (Common::AlignUp(height, 512u) / 512u) * 32768u;
 	if (size == 0 || size > UINT32_MAX) {
 		return false;
 	}
@@ -1154,13 +1151,13 @@ bool TileGetDepthSize(uint32_t width, uint32_t height, uint32_t pitch,
 		                                           &stencil_block_height));
 		const uint32_t fragments = 1u << num_fragments_log2;
 		const uint64_t depth_bytes_total =
-		    valid_blocks ? static_cast<uint64_t>(AlignUp(width, depth_block_width)) *
-		                       AlignUp(height, depth_block_height) * depth_bytes * fragments
+		    valid_blocks ? static_cast<uint64_t>(Common::AlignUp(width, depth_block_width)) *
+		                       Common::AlignUp(height, depth_block_height) * depth_bytes * fragments
 		                 : 0;
 		const uint64_t stencil_bytes_total =
 		    stencil_format == Prospero::StencilFormat::k8UInt && valid_blocks
-		        ? static_cast<uint64_t>(AlignUp(width, stencil_block_width)) *
-		              AlignUp(height, stencil_block_height) * fragments
+		        ? static_cast<uint64_t>(Common::AlignUp(width, stencil_block_width)) *
+		              Common::AlignUp(height, stencil_block_height) * fragments
 		        : 0;
 		TileSizeAlign calculated_htile {};
 		const bool    htile_valid = !htile || TileGetHtileSize(width, height, calculated_htile);
@@ -1187,8 +1184,7 @@ uint32_t TileGetRenderTargetPitch(uint32_t width, uint32_t bytes_per_element,
 	                                                         &block_width, &block_height)) {
 		return 0;
 	}
-	const uint64_t pitch = (static_cast<uint64_t>(width) + block_width - 1u) &
-	                       ~static_cast<uint64_t>(block_width - 1u);
+	const uint64_t pitch = Common::AlignUp(static_cast<uint64_t>(width), block_width);
 	return pitch <= UINT32_MAX ? static_cast<uint32_t>(pitch) : 0;
 }
 
@@ -1209,8 +1205,7 @@ bool TileGetRenderTargetSize(uint32_t width, uint32_t height, uint32_t pitch,
 	    pitch != TileGetRenderTargetPitch(width, bytes_per_element, num_fragments_log2)) {
 		return false;
 	}
-	const uint64_t padded_height = (static_cast<uint64_t>(height) + block_height - 1u) &
-	                               ~static_cast<uint64_t>(block_height - 1u);
+	const uint64_t padded_height = Common::AlignUp(static_cast<uint64_t>(height), block_height);
 	const uint64_t size = static_cast<uint64_t>(pitch) * padded_height * bytes_per_element *
 	                      (1u << num_fragments_log2);
 	if (size == 0 || size > UINT32_MAX) {
@@ -1221,6 +1216,31 @@ bool TileGetRenderTargetSize(uint32_t width, uint32_t height, uint32_t pitch,
 	return true;
 }
 
+bool TileGetDccSize(uint32_t width, uint32_t height, uint32_t slices,
+                    uint32_t bytes_per_element, uint32_t levels, Prospero::TileMode tile,
+                    TileSizeAlign& total_size, uint32_t num_fragments_log2) {
+	total_size = {};
+	if (width == 0 || height == 0 || slices == 0 || levels != 1 || num_fragments_log2 != 0 ||
+	    !std::has_single_bit(bytes_per_element) || bytes_per_element > 16 ||
+	    (tile != Prospero::TileMode::kRenderTarget && tile != Prospero::TileMode::kDepth)) {
+		return false;
+	}
+	// Gen5 color metadata uses 4 KiB blocks, with one byte per 256 bytes of color data.
+	// Metadata stays thin for volume surfaces, so every depth slice has its own block raster.
+	const uint32_t coverage_bits = 20u - std::countr_zero(bytes_per_element);
+	const uint32_t block_width   = 1u << ((coverage_bits + 1u) / 2u);
+	const uint32_t block_height  = 1u << (coverage_bits / 2u);
+	const uint64_t blocks_x = (static_cast<uint64_t>(width) + block_width - 1u) / block_width;
+	const uint64_t blocks_y = (static_cast<uint64_t>(height) + block_height - 1u) / block_height;
+	const uint64_t blocks   = blocks_x * blocks_y;
+	if (blocks > UINT32_MAX / 4096u / slices) {
+		return false;
+	}
+	total_size.size  = static_cast<uint32_t>(blocks * slices * 4096u);
+	total_size.align = 4096;
+	return true;
+}
+
 bool TileGetRenderTargetMipLayout(uint32_t width, uint32_t height, uint32_t pitch,
                                   uint32_t bytes_per_element, uint32_t levels,
                                   TileSizeAlign& total_size, TileSizeOffset* level_sizes,
@@ -1347,7 +1367,7 @@ uint32_t TileGetTexturePitch(Prospero::BufferFormat format, uint32_t width,
 			TileTextureElementLayout element {};
 			if (TileGetTextureElementLayout(format, element) && element.texel_width == 1 &&
 			    element.texel_height == 1) {
-				pitch = AlignUp(pitch, CalcLinearBlockWidth(element.bytes));
+				pitch = Common::AlignUp(pitch, CalcLinearBlockWidth(element.bytes));
 			}
 			break;
 		}
@@ -1358,7 +1378,7 @@ uint32_t TileGetTexturePitch(Prospero::BufferFormat format, uint32_t width,
 		case Prospero::TileMode::kRenderTarget: {
 			TileTextureBlockLayout layout {};
 			if (TileGetTextureBlockLayout(format, tile, false, layout)) {
-				pitch = AlignUp(pitch, layout.block.block_width * layout.texel_width);
+				pitch = Common::AlignUp(pitch, layout.block.block_width * layout.texel_width);
 			}
 			break;
 		}

src/graphics/guest_gpu/tile.h

@@ -86,6 +86,7 @@ struct TileMipLayout {
 	uint32_t padded_height = 0;
 	uint32_t tail_x        = 0;
 	uint32_t tail_y        = 0;
+	bool     operator==(const TileMipLayout&) const = default;
 };
 
 struct TileSurfaceLayout {
@@ -122,6 +123,9 @@ uint32_t TileGetDepthPitch(uint32_t width, uint32_t bytes_per_element,
 bool     TileGetRenderTargetSize(uint32_t width, uint32_t height, uint32_t pitch,
                                  uint32_t bytes_per_element, TileSizeAlign& total_size,
                                  uint32_t num_fragments_log2 = 0);
+bool     TileGetDccSize(uint32_t width, uint32_t height, uint32_t slices,
+                        uint32_t bytes_per_element, uint32_t levels, Prospero::TileMode tile,
+                        TileSizeAlign& total_size, uint32_t num_fragments_log2 = 0);
 bool     TileGetRenderTargetMipLayout(uint32_t width, uint32_t height, uint32_t pitch,
                                       uint32_t bytes_per_element, uint32_t levels,
                                       TileSizeAlign& total_size, TileSizeOffset* level_sizes,

src/graphics/host_gpu/graphicContext.h

@@ -14,9 +14,7 @@
 
 namespace Libs::Graphics {
 
-struct VulkanBuffer;
 struct VulkanImage;
-struct VulkanMemory;
 
 inline constexpr uint32_t VULKAN_TARGET_API_VERSION = VK_API_VERSION_1_3;
 
@@ -29,10 +27,13 @@ struct GraphicContext {
 	vk::Device                         device                                = nullptr;
 	VmaAllocator                       allocator                             = nullptr;
 	bool                               memory_budget_ext_enabled             = false;
-	bool                               rt_extensions_enabled                 = false;
 	bool                               compute_subgroup_size_control_enabled = false;
-	bool                               compute_wave64_supported              = false;
 	bool                               sample_rate_shading_enabled           = false;
+	bool                               attachment_feedback_loop_enabled      = false;
+	bool                               provoking_vertex_last_enabled         = false;
+	bool                               supports_block_texel_view              = false;
+	bool                                      mesh_shader_enabled                   = false;
+	vk::PhysicalDeviceMeshShaderPropertiesEXT mesh_shader_properties                = {};
 	uint32_t                           subgroup_size                         = 0;
 	uint32_t                           min_subgroup_size                     = 0;
 	uint32_t                           max_subgroup_size                     = 0;
@@ -80,6 +81,10 @@ struct GraphicContext {
 		return it->second.first;
 	}
 
+	[[nodiscard]] bool SupportsComputeWave64() const noexcept {
+		return subgroup_size == 64u || compute_subgroup_size_control_enabled;
+	}
+
 	[[nodiscard]] vk::DeviceSize StorageMinAlignment() const {
 		const auto alignment = physical_device_properties.limits.minStorageBufferOffsetAlignment;
 		return alignment != 0 ? alignment : 1;
@@ -91,15 +96,8 @@ struct GraphicContext {
 	[[nodiscard]] bool CanReportMemoryUsage() const noexcept { return memory_budget_ext_enabled; }
 	[[nodiscard]] uint64_t GetDeviceMemoryUsage() const;
 	[[nodiscard]] uint64_t GetTotalMemoryBudget() const;
-	void                   CreateBuffer(uint64_t size, VulkanBuffer& buffer);
 	[[nodiscard]] bool     CreateImage(const vk::ImageCreateInfo& info, VulkanImage& image);
 	void                   DeleteImage(VulkanImage& image);
-	void                   MapMemory(VulkanMemory& memory, void*& data);
-	void                   UnmapMemory(VulkanMemory& memory);
-	void                   AppendHardwareRayTracingDeviceExtensions(
-	    const std::vector<vk::ExtensionProperties>& available_extensions,
-	    std::vector<const char*>&                   device_extensions);
-	void LoadHardwareRayTracingFunctions() const;
 
 	uint32_t screen_width  = 0;
 	uint32_t screen_height = 0;
@@ -114,18 +112,6 @@ private:
 	    m_image_format_properties;
 };
 
-struct VulkanMemory {
-	vk::MemoryRequirements  requirements       = {};
-	vk::MemoryPropertyFlags property           = {};
-	vk::MemoryPropertyFlags preferred_property = {};
-	vk::DeviceMemory        memory             = nullptr;
-	VmaAllocation           allocation         = nullptr;
-	VmaAllocationInfo       allocation_info    = {};
-	vk::DeviceSize          offset             = 0;
-	uint32_t                type               = 0;
-	uint64_t                unique_id          = 0;
-};
-
 struct VulkanImageState {
 	vk::PipelineStageFlags2 pl_stage    = vk::PipelineStageFlagBits2::eAllCommands;
 	vk::AccessFlags2        access_mask = vk::AccessFlagBits2::eNone;
@@ -139,7 +125,6 @@ struct VulkanImage {
 	vk::Format                    format      = vk::Format::eUndefined;
 	vk::ImageType                 image_type  = vk::ImageType::e2D;
 	vk::Extent3D                  extent      = {1, 1, 1};
-	uint32_t                      guest_pitch = 0;
 	uint32_t                      layers      = 1;
 	uint32_t                      mip_levels  = 1;
 	uint32_t                      samples     = 1;
@@ -148,15 +133,10 @@ struct VulkanImage {
 	vk::Image                     image       = nullptr;
 	VulkanImageState              state;
 	std::vector<VulkanImageState> subresource_states;
-	Graphics::VulkanMemory        memory;
+	VmaAllocation                allocation = nullptr;
 };
 
-struct VulkanBuffer {
-	vk::Buffer           buffer = nullptr;
-	VulkanMemory         memory;
-	vk::BufferUsageFlags usage       = {};
-	uint64_t             buffer_size = 0;
-};
+
 
 } // namespace Libs::Graphics
 

src/graphics/host_gpu/memoryTracker.cpp

@@ -1,5 +1,6 @@
 #include "graphics/host_gpu/memoryTracker.h"
 
+#include "common/alignment.h"
 #include "common/assert.h"
 
 namespace Libs::Graphics {
@@ -8,9 +9,6 @@ static_assert(std::atomic<void*>::is_always_lock_free);
 
 MemoryTracker::MemoryTracker(PageManager& page_manager): m_page_manager(page_manager) {
 	m_regions = std::make_unique<std::atomic<RegionManager*>[]>(REGION_COUNT);
-	for (size_t i = 0; i < REGION_COUNT; i++) {
-		m_regions[i].store(nullptr, std::memory_order_relaxed);
-	}
 }
 
 MemoryTracker::~MemoryTracker() = default;
@@ -18,15 +16,12 @@ MemoryTracker::~MemoryTracker() = default;
 #if KYTY_BUILD == KYTY_BUILD_DEBUG
 void MemoryTracker::ValidateGpuDirtyPages(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
                                           const char* operation) const noexcept {
-	if (vaddr == 0 || size == 0 || size > UINT64_MAX - vaddr ||
-	    (vaddr & (TRACKER_PAGE_SIZE - 1)) != 0 || (size & (TRACKER_PAGE_SIZE - 1)) != 0) {
+	if (!GuestRange {vaddr, size}.Valid() || (vaddr & (TRACKER_PAGE_SIZE - 1)) != 0 ||
+	    (size & (TRACKER_PAGE_SIZE - 1)) != 0) {
 		EXIT("MemoryTracker: invalid dirty-page validation range\n");
 	}
 	for (auto page = vaddr; page < vaddr + size; page += TRACKER_PAGE_SIZE) {
-		bool found = false;
-		dirty.ForEachIntersection(page, TRACKER_PAGE_SIZE,
-		                          [&found](RangeSet::Range) { found = true; });
-		if (!found) {
+		if (!dirty.Intersects(page, TRACKER_PAGE_SIZE)) {
 			EXIT("MemoryTracker: GPU-dirty tracker page has no dirty bytes, operation=%s "
 			     "addr=0x%016" PRIx64 "\n",
 			     operation, page);
@@ -37,15 +32,10 @@ void MemoryTracker::ValidateGpuDirtyPages(const RangeSet& dirty, uint64_t vaddr,
 void MemoryTracker::ValidateGpuDirtyOwnership(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
                                               const char* operation) {
 	ValidateRange(vaddr, size);
-	if (vaddr + size > UINT64_MAX - (TRACKER_PAGE_SIZE - 1)) {
-		EXIT("MemoryTracker: dirty ownership range alignment overflow\n");
-	}
-	const auto begin = vaddr & ~(TRACKER_PAGE_SIZE - 1);
-	const auto end   = (vaddr + size + TRACKER_PAGE_SIZE - 1) & ~(TRACKER_PAGE_SIZE - 1);
+	const auto begin = Common::AlignDown(vaddr, TRACKER_PAGE_SIZE);
+	const auto end   = Common::AlignUp(vaddr + size, TRACKER_PAGE_SIZE);
 	for (auto page = begin; page < end; page += TRACKER_PAGE_SIZE) {
-		bool has_dirty_bytes = false;
-		dirty.ForEachIntersection(page, TRACKER_PAGE_SIZE,
-		                          [&has_dirty_bytes](RangeSet::Range) { has_dirty_bytes = true; });
+		const bool has_dirty_bytes = dirty.Intersects(page, TRACKER_PAGE_SIZE);
 		if (IsRegionGpuModified(page, TRACKER_PAGE_SIZE) != has_dirty_bytes) {
 			EXIT("MemoryTracker: tracker and byte ownership disagree, operation=%s "
 			     "addr=0x%016" PRIx64 "\n",
@@ -56,8 +46,7 @@ void MemoryTracker::ValidateGpuDirtyOwnership(const RangeSet& dirty, uint64_t va
 #endif
 
 void MemoryTracker::ValidateRange(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("invalid memory tracker range\n");
 	}
 }
@@ -117,7 +106,8 @@ void MemoryTracker::UnmarkRegionAsGpuModified(uint64_t vaddr, uint64_t size) {
 	});
 }
 
-void MemoryTracker::UntrackMemoryImpl(uint64_t vaddr, uint64_t size) {
+void MemoryTracker::UntrackMemory(uint64_t vaddr, uint64_t size) {
+	CheckNotInUploadCallback();
 	std::vector<RegionManager*> managers;
 	managers.reserve((vaddr % TRACKER_REGION_SIZE + size + TRACKER_REGION_SIZE - 1) /
 	                 TRACKER_REGION_SIZE);
@@ -138,12 +128,6 @@ void MemoryTracker::UntrackMemoryImpl(uint64_t vaddr, uint64_t size) {
 	Iterate<false>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		manager->ChangeState<DirtySource::Cpu, true>(manager->GetCpuAddr() + offset, bytes);
 	});
-	locks.clear();
-}
-
-void MemoryTracker::UntrackMemory(uint64_t vaddr, uint64_t size) {
-	CheckNotInUploadCallback();
-	UntrackMemoryImpl(vaddr, size);
 }
 
 } // namespace Libs::Graphics

src/graphics/host_gpu/memoryTracker.h

@@ -34,7 +34,6 @@ public:
 	void InvalidateRegion(uint64_t vaddr, uint64_t size, Flush&& on_flush) noexcept {
 		static_assert(std::is_invocable_v<Flush&>);
 		CheckNotInUploadCallback();
-		ValidateRange(vaddr, size);
 
 		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 			const bool should_flush = [&] {
@@ -63,43 +62,18 @@ public:
 	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) {
-		static_assert(std::is_nothrow_invocable_v<Preflight&, uint64_t, uint64_t>);
-		static_assert(std::is_nothrow_invocable_v<Func&, uint64_t, uint64_t>);
-		CheckNotInUploadCallback();
-		std::vector<RegionManager*> managers;
-		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t, uint64_t) {
-			managers.push_back(manager);
-		});
-		std::vector<std::unique_lock<TrackingSpinLock>> locks;
-		locks.reserve(managers.size());
-		for (auto* manager: managers) {
-			locks.emplace_back(manager->lock);
-		}
-		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-			const auto address = manager->GetCpuAddr() + offset;
-			manager->template ForEachModifiedRange<DirtySource::Gpu, false>(address, bytes,
-			                                                                preflight);
-		});
-		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-			manager->template ForEachModifiedRange<DirtySource::Gpu, false>(
-			    manager->GetCpuAddr() + offset, bytes, func);
-		});
-		if constexpr (clear) {
-			Iterate<false>(vaddr, size,
-			               [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-				               const auto address = manager->GetCpuAddr() + offset;
-				               manager->template ForEachModifiedRange<DirtySource::Gpu, true>(
-				                   address, bytes, [](uint64_t, uint64_t) noexcept {});
-			               });
-		}
-	}
-
 	template <bool clear, typename Func>
 	void ForEachDownloadRange(uint64_t vaddr, uint64_t size, Func&& func) {
-		ForEachDownloadRange<clear>(
-		    vaddr, size, [](uint64_t, uint64_t) noexcept {}, std::forward<Func>(func));
+		static_assert(std::is_nothrow_invocable_v<Func&, uint64_t, uint64_t>);
+		CheckNotInUploadCallback();
+		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
+			std::scoped_lock lock(manager->lock);
+			const auto       address = manager->GetCpuAddr() + offset;
+			manager->template ForEachModifiedRange<DirtySource::Gpu, false>(address, bytes, func);
+			if constexpr (clear) {
+				manager->template ChangeState<DirtySource::Gpu, false>(address, bytes);
+			}
+		});
 	}
 
 	template <typename RangeFunc, typename UploadFunc>
@@ -171,7 +145,6 @@ private:
 	}
 
 	static void    ValidateRange(uint64_t vaddr, uint64_t size);
-	void           UntrackMemoryImpl(uint64_t vaddr, uint64_t 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 "common/platform/sysDbg.h"
+#include "common/alignment.h"
+#include "common/virtualMemory.h"
 #include "graphics/host_gpu/regionDefinitions.h"
 #include "kernel/memory.h"
 
@@ -21,10 +22,7 @@
 #include <windows.h>
 #undef min
 #undef max
-#elif defined(__APPLE__)
-#include <unistd.h>
 #else
-#include <execinfo.h>
 #include <unistd.h>
 #endif
 
@@ -36,38 +34,12 @@ constexpr uint64_t REGION_SIZE  = TRACKER_REGION_SIZE;
 constexpr uint64_t ADDRESS_SIZE = TRACKER_ADDRESS_SIZE;
 constexpr uint64_t REGION_COUNT = ADDRESS_SIZE / REGION_SIZE;
 
-#if KYTY_PLATFORM != KYTY_PLATFORM_WINDOWS
-// The tracker reuses Win32 memory-protection tags as internal page-state values.
-// Mirror their canonical numeric values so the shared state-machine logic is identical.
-constexpr uint32_t PAGE_NOACCESS  = 0x01;
-constexpr uint32_t PAGE_READONLY  = 0x02;
-constexpr uint32_t PAGE_READWRITE = 0x04;
-#endif
 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;
-
 [[noreturn]] void FailFast(const char* reason = nullptr) noexcept {
 	std::fputs("PageManager fail-fast: ", stderr);
 	std::fputs(reason != nullptr ? reason : "invalid page state", stderr);
 	std::fputc('\n', stderr);
-#if !defined(__APPLE__)
-	void* frames[16] {};
-	int   frame_count = static_cast<int>(std::size(frames));
-	SysStackWalk(frames, &frame_count);
-#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
-	const auto image_base = reinterpret_cast<uintptr_t>(GetModuleHandleW(nullptr));
-	for (int i = 0; i < frame_count; i++) {
-		const auto address = reinterpret_cast<uintptr_t>(frames[i]);
-		std::fprintf(stderr, "  frame[%d]=0x%016" PRIxPTR " image_rva=0x%016" PRIxPTR "\n", i,
-		             address, address >= image_base ? address - image_base : 0);
-	}
-#else
-	::backtrace_symbols_fd(frames, frame_count, STDERR_FILENO);
-#endif
-#endif
 	std::fflush(stderr);
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
 	TerminateProcess(GetCurrentProcess(), static_cast<UINT>(EXCEPTION_NONCONTINUABLE_EXCEPTION));
@@ -86,15 +58,6 @@ constexpr uint32_t READ_WRITE_PROTECTION = PAGE_READWRITE;
 	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);
-	}
-}
-
 class SpinGuard final {
 public:
 	explicit SpinGuard(std::atomic_flag& lock): m_lock(lock) {
@@ -110,20 +73,11 @@ private:
 };
 
 void ValidateRange(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= ADDRESS_SIZE || size > ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		Fatal("invalid range vaddr=0x%016" PRIx64 ", size=0x%016" PRIx64, vaddr, size);
 	}
 }
 
-uint64_t PageStart(uint64_t vaddr) {
-	return vaddr & ~(PAGE_SIZE - 1);
-}
-
-uint64_t PageEnd(uint64_t vaddr, uint64_t size) {
-	ValidateRange(vaddr, size);
-	return PageStart(vaddr + size - 1) + PAGE_SIZE;
-}
-
 } // namespace
 
 struct PageManager::Impl {
@@ -131,14 +85,14 @@ struct PageManager::Impl {
 		uint8_t write_watchers  : 7 = 0;
 		uint8_t access_watchers : 1 = 0;
 
-		[[nodiscard]] uint32_t Perms() const noexcept {
+		[[nodiscard]] Common::VirtualMemory::Mode Perms() const noexcept {
 			if (access_watchers != 0) {
-				return NO_ACCESS_PROTECTION;
+				return Common::VirtualMemory::Mode::NoAccess;
 			}
 			if (write_watchers != 0) {
-				return READ_ONLY_PROTECTION;
+				return Common::VirtualMemory::Mode::Read;
 			}
-			return READ_WRITE_PROTECTION;
+			return Common::VirtualMemory::Mode::ReadWrite;
 		}
 
 		template <int delta, bool is_read>
@@ -202,9 +156,6 @@ struct PageManager::Impl {
 		}
 #endif
 		regions = std::make_unique<std::atomic<Region*>[]>(REGION_COUNT);
-		for (uint64_t i = 0; i < REGION_COUNT; i++) {
-			regions[i].store(nullptr, std::memory_order_relaxed);
-		}
 	}
 
 	~Impl() {
@@ -239,11 +190,10 @@ struct PageManager::Impl {
 		return ptr;
 	}
 
-	void Protect(uint64_t vaddr, uint64_t size, uint32_t protection) noexcept {
-		if (!Libs::LibKernel::Memory::ProtectGuestHostMemory(vaddr, size,
-		                                                     ToMemoryMode(protection))) {
-			Fatal("address-space protection failed at 0x%016" PRIx64 ", new=0x%08" PRIx32, vaddr,
-			      protection);
+	void Protect(uint64_t vaddr, uint64_t size, Common::VirtualMemory::Mode mode) noexcept {
+		if (!Libs::LibKernel::Memory::ProtectGuestHostMemory(vaddr, size, mode)) {
+			Fatal("address-space protection failed at 0x%016" PRIx64 ", mode=0x%08" PRIx32, vaddr,
+			      static_cast<uint32_t>(mode));
 		}
 	}
 
@@ -300,11 +250,12 @@ struct PageManager::Impl {
 
 	template <bool track, bool is_read>
 	void UpdatePageWatchers(uint64_t vaddr, uint64_t size) {
-		const auto begin = PageStart(vaddr);
-		const auto end   = PageEnd(vaddr, size);
+		ValidateRange(vaddr, size);
+		const auto begin = Common::AlignDown(vaddr, PAGE_SIZE);
+		const auto end   = Common::AlignUp(vaddr + size, PAGE_SIZE);
 		for (auto chunk_begin = begin; chunk_begin < end;) {
-			const auto chunk_end   = std::min(end, (chunk_begin / REGION_SIZE + 1) * REGION_SIZE);
-			const auto region_base = chunk_begin / REGION_SIZE * REGION_SIZE;
+			const auto chunk_end = std::min(end, Common::AlignUp(chunk_begin + 1, REGION_SIZE));
+			const auto region_base = Common::AlignDown(chunk_begin, REGION_SIZE);
 			auto*      region = track ? GetOrCreateRegion(chunk_begin) : FindRegion(chunk_begin);
 			if (region == nullptr) {
 				Fatal("untracking unknown page 0x%016" PRIx64, chunk_begin);
@@ -371,8 +322,4 @@ template void PageManager::UpdatePageWatchersForRegion<true, false>(uint64_t, Re
 template void PageManager::UpdatePageWatchersForRegion<false, true>(uint64_t, RegionBits&);
 template void PageManager::UpdatePageWatchersForRegion<false, false>(uint64_t, RegionBits&);
 
-void PageManager::OnGpuMap(uint64_t, uint64_t) {}
-
-void PageManager::OnGpuUnmap(uint64_t, uint64_t) {}
-
 } // namespace Libs::Graphics

src/graphics/host_gpu/pageManager.h

@@ -24,8 +24,6 @@ public:
 	void UpdatePageWatchers(uint64_t vaddr, uint64_t size);
 	template <bool track, bool is_read = false>
 	void UpdatePageWatchersForRegion(uint64_t base_addr, RegionBits& mask);
-	void OnGpuMap(uint64_t vaddr, uint64_t size);
-	void OnGpuUnmap(uint64_t vaddr, uint64_t size);
 
 private:
 	struct Impl;

src/graphics/host_gpu/rangeSet.h

@@ -6,17 +6,11 @@
 #include <algorithm>
 #include <cstdint>
 #include <map>
-#include <vector>
 
 namespace Libs::Graphics {
 
 class RangeSet final {
 public:
-	struct Range {
-		uint64_t address = 0;
-		uint64_t size    = 0;
-	};
-
 	void Add(uint64_t address, uint64_t size) {
 		const auto end = End(address, size);
 		auto       it  = m_ranges.lower_bound(address);
@@ -62,12 +56,6 @@ public:
 		}
 	}
 
-	[[nodiscard]] std::vector<Range> Intersections(uint64_t address, uint64_t size) const {
-		std::vector<Range> result;
-		ForEachIntersection(address, size, [&result](Range range) { result.push_back(range); });
-		return result;
-	}
-
 	[[nodiscard]] bool Intersects(uint64_t address, uint64_t size) const {
 		const auto end = End(address, size);
 		auto       it  = m_ranges.lower_bound(address);
@@ -88,7 +76,7 @@ public:
 	}
 
 	template <typename Func>
-	void ForEachIntersection(uint64_t address, uint64_t size, Func&& func) const {
+	void ForEachInRange(uint64_t address, uint64_t size, Func&& func) const {
 		const auto end = End(address, size);
 		auto       it  = m_ranges.upper_bound(address);
 		if (it != m_ranges.begin()) {
@@ -98,7 +86,7 @@ public:
 			const auto begin = std::max(address, it->first);
 			const auto last  = std::min(end, it->second);
 			if (begin < last) {
-				func(Range {begin, last - begin});
+				func(begin, last);
 			}
 		}
 	}

src/graphics/host_gpu/regionDefinitions.h

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

src/graphics/host_gpu/regionManager.h

@@ -111,56 +111,43 @@ public:
 			bits.UnsetRange(start, end);
 		}
 		if constexpr (source == DirtySource::Cpu) {
-			UpdateCpuProtection<!enable>();
+			UpdateProtection<!enable, false>();
 		} else {
-			UpdateGpuProtection<enable>();
+			UpdateProtection<enable, true>();
 		}
 	}
 
 	template <DirtySource source, bool clear, typename Func>
 	void ForEachModifiedRange(uint64_t vaddr, uint64_t size, Func&& func) {
 		const auto [start, end] = GetPageRange(vaddr, size);
-		RegionBits mask(GetBits<source>(), start, end);
+		auto&      bits         = GetBits<source>();
+		RegionBits mask(bits, start, end);
 		if constexpr (clear) {
-			GetBits<source>().UnsetRange(start, end);
+			bits.UnsetRange(start, end);
+			if constexpr (source == DirtySource::Cpu) {
+				UpdateProtection<true, false>();
+			} else {
+				UpdateProtection<false, true>();
+			}
 		}
-		if constexpr (source == DirtySource::Cpu && clear) {
-			UpdateCpuProtection<true>();
-			ForEachRange(mask, std::forward<Func>(func));
-			return;
+		for (const auto [first, last]: mask) {
+			func(m_cpu_addr + first * TRACKER_PAGE_SIZE, (last - first) * TRACKER_PAGE_SIZE);
 		}
-		if constexpr (source == DirtySource::Gpu && clear) {
-			UpdateGpuProtection<false>();
-		}
-		ForEachRange(mask, std::forward<Func>(func));
 	}
 
 	TrackingSpinLock lock;
 
 private:
-	template <bool track>
-	void UpdateCpuProtection() {
-		auto mask  = m_cpu_dirty ^ m_writable;
-		m_writable = m_cpu_dirty;
+	template <bool track, bool is_read>
+	void UpdateProtection() {
+		const auto protection = is_read ? ~m_gpu_dirty : m_cpu_dirty;
+		auto&      previous   = is_read ? m_readable : m_writable;
+		auto       mask       = protection ^ previous;
 		if (mask.None()) {
 			return;
 		}
-		m_page_manager.UpdatePageWatchersForRegion<track>(m_cpu_addr, mask);
-	}
-
-	template <bool track>
-	void UpdateGpuProtection() {
-		auto readable = ~m_gpu_dirty;
-		auto mask     = readable ^ m_readable;
-		m_readable    = readable;
-		if (mask.None()) {
-			return;
-		}
-		if constexpr (track) {
-			m_page_manager.UpdatePageWatchersForRegion<true, true>(m_cpu_addr, mask);
-		} else {
-			m_page_manager.UpdatePageWatchersForRegion<false, true>(m_cpu_addr, mask);
-		}
+		previous = protection;
+		m_page_manager.UpdatePageWatchersForRegion<track, is_read>(m_cpu_addr, mask);
 	}
 
 	template <DirtySource source>
@@ -191,13 +178,6 @@ private:
 		        static_cast<size_t>((offset + size + TRACKER_PAGE_SIZE - 1) / TRACKER_PAGE_SIZE)};
 	}
 
-	template <typename Func>
-	void ForEachRange(const RegionBits& bits, Func&& func) const {
-		for (const auto [start, end]: bits) {
-			func(m_cpu_addr + start * TRACKER_PAGE_SIZE, (end - start) * TRACKER_PAGE_SIZE);
-		}
-	}
-
 	PageManager& m_page_manager;
 	uint64_t     m_cpu_addr = 0;
 	RegionBits   m_cpu_dirty;

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

@@ -1,5 +1,6 @@
 #include "graphics/host_gpu/renderer/cache/bufferCache.h"
 
+#include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/profiler.h"
@@ -13,7 +14,6 @@
 #include "kernel/memory.h"
 
 #include <algorithm>
-#include <array>
 #include <cinttypes>
 #include <cstring>
 #include <memory>
@@ -43,13 +43,6 @@ void BufferCache::WriteDataBuffer(Buffer& buffer, uint64_t address, const void*
 	}
 }
 
-struct BufferCache::DownloadCopy {
-	Buffer*  buffer        = nullptr;
-	uint64_t source_offset = 0;
-	uint64_t address       = 0;
-	uint64_t size          = 0;
-};
-
 void BufferCache::Register(BufferId id) {
 	ChangeRegister<true>(id);
 }
@@ -104,8 +97,7 @@ void BufferCache::TouchBuffer(const Buffer& buffer) {
 }
 
 void BufferCache::DeleteBuffer(BufferId id) {
-	auto* buffer = m_slot_buffers.try_get(id);
-	if (buffer == nullptr || buffer->is_deleted) {
+	if (IsBufferInvalid(id)) {
 		return;
 	}
 	Unregister(id);
@@ -116,94 +108,84 @@ void BufferCache::DeleteBuffer(BufferId id) {
 	}
 }
 
-std::pair<uint64_t, uint64_t> BufferCache::DownloadEnvelope(const DownloadCopy& copy) {
-	if (copy.buffer == nullptr || copy.size == 0 || copy.source_offset > copy.buffer->Size() ||
-	    copy.size > copy.buffer->Size() - copy.source_offset) {
-		EXIT("BufferCache: invalid download copy\n");
+bool BufferCache::DownloadBufferMemory(Buffer& buffer, uint64_t vaddr, uint64_t size) {
+	std::vector<vk::BufferCopy> copies;
+	uint64_t                    total_size     = 0;
+	const auto                  buffer_address = buffer.CpuAddress();
+	m_memory_tracker.ForEachDownloadRange<false>(
+	    vaddr, size, [&](uint64_t address, uint64_t bytes) noexcept {
+		    m_memory_tracker.ValidateGpuDirtyPages(m_gpu_modified_ranges, address, bytes,
+		                                           "buffer download");
+		    m_gpu_modified_ranges.ForEachInRange(address, bytes, [&](uint64_t start, uint64_t end) {
+			    copies.emplace_back(start - buffer_address, total_size, end - start);
+			    // Keep packed ranges on separate cache lines, as in shadPS4.
+			    total_size += Common::AlignUp(end - start, 64);
+		    });
+		    m_gpu_modified_ranges.Subtract(address, bytes);
+	    });
+	if (copies.empty()) {
+		return false;
 	}
-	const auto begin = copy.source_offset & ~uint64_t {3};
-	if (copy.source_offset > UINT64_MAX - copy.size ||
-	    copy.source_offset + copy.size > UINT64_MAX - 3) {
-		EXIT("BufferCache: download copy alignment overflow\n");
-	}
-	const auto end = (copy.source_offset + copy.size + 3) & ~uint64_t {3};
-	if (end > copy.buffer->Size()) {
-		EXIT("BufferCache: aligned download copy exceeds its owner\n");
-	}
-	return {begin, end - begin};
-}
 
-void BufferCache::DownloadBufferMemory(std::span<const DownloadCopy> copies) {
-	std::vector<DownloadCopy> batch;
-	batch.reserve(copies.size());
-	uint64_t                  packed_size = 0;
-	auto&                     download    = m_download_buffer;
-	const auto flush = [&] {
-		const auto [mapped, base_offset] = download.Map(packed_size, DOWNLOAD_ALIGNMENT);
-		EXIT_IF(mapped == nullptr);
-		uint64_t cursor = 0;
-		for (const auto& copy: batch) {
-			const auto [source_begin, envelope_size] = DownloadEnvelope(copy);
-			download.CopyFrom(m_scheduler.Current(), *copy.buffer, source_begin, base_offset + cursor,
-			                  envelope_size, vk::AccessFlagBits::eMemoryWrite, vk::AccessFlags {},
-			                  vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite,
-			                  vk::AccessFlagBits::eHostRead);
-			cursor += AlignDownload(envelope_size);
-		}
-		download.Commit();
-		const auto completion_tick = m_scheduler.CurrentTick();
-		m_scheduler.Finish();
-		m_scheduler.WaitPriorityOperations(completion_tick);
-		cursor = 0;
-		for (const auto& copy: batch) {
-			const auto [source_begin, envelope_size] = DownloadEnvelope(copy);
-			const auto offset = cursor + copy.source_offset - source_begin;
-			download.Invalidate(base_offset + offset, copy.size);
-			Libs::LibKernel::Memory::WriteBacking(copy.address, mapped + offset, copy.size);
-			cursor += AlignDownload(envelope_size);
-		}
-		batch.clear();
-		packed_size = 0;
-	};
-	for (auto copy: copies) {
-		while (copy.size != 0) {
-			const auto available = download.Size() - packed_size;
-			const auto prefix    = copy.source_offset & 3u;
-			const auto bytes     = std::min(copy.size, available - prefix);
-			DownloadCopy part {copy.buffer, copy.source_offset, copy.address, bytes};
-			const auto [source_begin, envelope_size] = DownloadEnvelope(part);
-			(void)source_begin;
-			packed_size += AlignDownload(envelope_size);
-			batch.push_back(part);
-			copy.source_offset += bytes;
-			copy.address += bytes;
-			copy.size -= bytes;
-			if (packed_size == download.Size()) {
-				flush();
-			}
-		}
+	const auto [mapped, offset] = m_download_buffer.Map(total_size, 64);
+	if (mapped == nullptr) {
+		EXIT("BufferCache: download exceeds 32 MiB staging buffer capacity\n");
 	}
-	if (!batch.empty()) {
-		flush();
-	}
-	for (const auto& copy: copies) {
-		m_gpu_modified_ranges.Subtract(copy.address, copy.size);
+	m_download_buffer.Commit();
+	for (auto& copy: copies) {
+		copy.dstOffset += offset;
 	}
+
+	auto& command = m_scheduler.Current();
+	command.EndRendering();
+	const auto              native = command.Handle();
+	vk::BufferMemoryBarrier before {};
+	before.srcAccessMask       = vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite;
+	before.dstAccessMask       = vk::AccessFlagBits::eTransferRead;
+	before.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+	before.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+	before.buffer              = buffer.Handle();
+	before.offset              = 0;
+	before.size                = buffer.Size();
+	native.pipelineBarrier(vk::PipelineStageFlagBits::eAllCommands,
+	                       vk::PipelineStageFlagBits::eTransfer, {}, 0, nullptr, 1, &before, 0,
+	                       nullptr);
+	native.copyBuffer(buffer.Handle(), m_download_buffer.Handle(),
+	                  static_cast<uint32_t>(copies.size()), copies.data());
+
+	auto after          = before;
+	after.srcAccessMask = vk::AccessFlagBits::eTransferWrite;
+	after.dstAccessMask = vk::AccessFlagBits::eHostRead;
+	after.buffer        = m_download_buffer.Handle();
+	after.offset        = offset;
+	after.size          = total_size;
+	native.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer,
+	                       vk::PipelineStageFlagBits::eAllCommands |
+	                           vk::PipelineStageFlagBits::eHost,
+	                       {}, 0, nullptr, 1, &after, 0, nullptr);
+	m_scheduler.DeferPriorityOperation([this, mapped, offset, total_size, buffer_address,
+	                                    copies = std::move(copies)] {
+		m_download_buffer.Invalidate(offset, total_size);
+		for (const auto& copy: copies) {
+			Libs::LibKernel::Memory::WriteBacking(buffer_address + copy.srcOffset,
+			                                      mapped + (copy.dstOffset - offset), copy.size);
+		}
+	});
+	return true;
 }
 
 BufferCache::BufferCache(GraphicContext& graphics, CommandScheduler& scheduler,
                          PageManager& page_manager, TextureCache& texture_cache)
-	: m_graphics(graphics), m_scheduler(scheduler),
-	  m_fault_manager(graphics, scheduler, *this, CACHING_PAGEBITS, CACHING_NUMPAGES),
-	  m_gds_buffer(graphics, scheduler, MemoryUsage::Stream, 0, AllFlags, GdsBufferSize),
-	  m_bda_pagetable_buffer(graphics, scheduler, MemoryUsage::DeviceLocal, 0, AllFlags,
-	                         BDA_PAGETABLE_SIZE),
-	  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),
-	  m_device_buffer(graphics, scheduler, MemoryUsage::DeviceLocal, 128 * MiB),
-	  m_texture_cache(texture_cache) {
+    : m_graphics(graphics), m_scheduler(scheduler), m_fault_manager(graphics, scheduler, *this),
+      m_gds_buffer(graphics, scheduler, MemoryUsage::Stream, 0, AllFlags, GdsBufferSize),
+      m_bda_pagetable_buffer(graphics, scheduler, MemoryUsage::DeviceLocal, 0, AllFlags,
+                             BDA_PAGETABLE_SIZE),
+      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),
+      m_device_buffer(graphics, scheduler, MemoryUsage::DeviceLocal, 128 * MiB),
+      m_texture_cache(texture_cache) {
 	std::memset(m_gds_buffer.Mapped().data(), 0, static_cast<size_t>(m_gds_buffer.Size()));
 	m_gds_buffer.Flush(0, m_gds_buffer.Size());
 	SetVulkanObjectNameF(m_graphics.device, m_bda_pagetable_buffer.Handle(),
@@ -241,8 +223,7 @@ BufferCache::~BufferCache() {
 }
 
 void BufferCache::InvalidateMemory(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: invalid memory-invalidation range\n");
 	}
 	m_memory_tracker.InvalidateRegion(vaddr, size,
@@ -255,66 +236,36 @@ void BufferCache::ReadMemory(uint64_t vaddr, uint64_t size, bool is_write) {
 		     "addr=0x%016" PRIx64 " size=0x%016" PRIx64 "\n",
 		     vaddr, size);
 	}
-	m_scheduler.Context().GetGpu().SendCommandSync(
-	    [this, vaddr, size, is_write] { ReadMemoryOnGpu(vaddr, size, is_write); });
-}
-
-void BufferCache::ReadMemoryOnGpu(uint64_t vaddr, uint64_t size, bool is_write) {
-	// CPU invalidation reaches this point only for a GPU-owned tracker page. Resolve the exact
-	// Buffer owner on the GPU thread so the cache index remains single-thread-owned.
-	if (is_write && !IsRegionRegistered(vaddr, size)) {
-		return;
-	}
-	std::vector<DownloadCopy> copies;
-	m_memory_tracker.ForEachDownloadRange<false>(
-	    vaddr, size,
-	    [&](uint64_t address, uint64_t bytes) noexcept {
-		    m_memory_tracker.ValidateGpuDirtyPages(m_gpu_modified_ranges, address, bytes,
-		                                           "memory invalidation");
-	    },
-		[&](uint64_t address, uint64_t bytes) noexcept {
-		    for (const auto range: m_gpu_modified_ranges.Intersections(address, bytes)) {
-			    for (uint64_t copied = 0; copied < range.size;) {
-				    const auto copy_address = range.address + copied;
-				    const auto* owner = m_page_table.Find(copy_address >> PageTable::kPageBits);
-				    if (owner == nullptr || !*owner) {
-					    EXIT("BufferCache: invalidation readback has no buffer owner\n");
-				    }
-				    auto& buffer = m_slot_buffers[*owner];
-				    if (!buffer.IsInBounds(copy_address, 1)) {
-					    EXIT("BufferCache: invalidation readback is outside its buffer owner\n");
-				    }
-				    const auto copy_size = std::min(
-				        range.size - copied, buffer.CpuAddress() + buffer.Size() - copy_address);
-				    copies.push_back(
-				        {&buffer, buffer.Offset(copy_address), copy_address, copy_size});
-				    copied += copy_size;
-			    }
-		    }
-	    });
-	if (copies.empty()) {
-		if (!is_write) {
+	m_scheduler.Context().GetGpu().SendCommandSync([this, vaddr, size, is_write] {
+		if (is_write && !IsRegionRegistered(vaddr, size)) {
 			return;
 		}
-		// A preceding read fault can consume the last GPU-owned copy after this write invalidation
-		// has already chosen to flush. Complete the CPU ownership handoff even though this callback
-		// no longer has bytes to download.
-		m_memory_tracker.MarkRegionAsCpuModified(vaddr, size);
-		return;
-	}
-	DownloadBufferMemory(copies);
-	// The enumeration above covered whole dirty pages and every exact interval on them.
-	m_memory_tracker.UnmarkRegionAsGpuModified(vaddr, size);
-	if (is_write) {
-		m_memory_tracker.MarkRegionAsCpuModified(vaddr, size);
-	}
+		auto& buffer = m_slot_buffers[FindBuffer(vaddr, size)];
+
+		// Widen nearby CPU reads so they share one GPU drain.
+		constexpr uint64_t WindowSize   = 512 * 1024;
+		const auto         buffer_begin = buffer.CpuAddress();
+		const auto         buffer_end   = buffer_begin + buffer.Size();
+		const auto window_begin = std::max(Common::AlignDown(vaddr, WindowSize), buffer_begin);
+		const auto window_end = std::min(std::max(window_begin + WindowSize, vaddr + size), buffer_end);
+
+		if (DownloadBufferMemory(buffer, window_begin, window_end - window_begin)) {
+			const auto tick = m_scheduler.CurrentTick();
+			m_scheduler.Wait(tick);
+			m_scheduler.WaitPriorityOperations(tick);
+			m_memory_tracker.UnmarkRegionAsGpuModified(window_begin, window_end - window_begin);
+		}
+		if (is_write) {
+			m_memory_tracker.MarkRegionAsCpuModified(vaddr, size);
+		}
+	});
 }
 
 BufferId BufferCache::FindBuffer(uint64_t vaddr, uint64_t size) {
 	if (vaddr == 0) {
 		return NULL_BUFFER_ID;
 	}
-	if (size == 0 || vaddr >= TRACKER_ADDRESS_SIZE || size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: invalid buffer discovery request\n");
 	}
 	const auto* owner = m_page_table.Find(vaddr >> PageTable::kPageBits);
@@ -327,38 +278,83 @@ BufferId BufferCache::FindBuffer(uint64_t vaddr, uint64_t size) {
 	return CreateBuffer(vaddr, size);
 }
 
-BufferId BufferCache::CreateBuffer(uint64_t vaddr, uint64_t size) {
-	auto& command = m_scheduler.Current();
-	EXIT_IF(command.IsInvalid());
-	auto       begin = vaddr & ~(CACHING_PAGESIZE - 1);
-	auto       end   = (vaddr + size + CACHING_PAGESIZE - 1) & ~(CACHING_PAGESIZE - 1);
-	auto       first = m_buffers.lower_bound(begin);
-	if (first != m_buffers.begin()) {
-		const auto previous = std::prev(first);
-		if (const auto& buffer = m_slot_buffers[previous->second];
-		    buffer.CpuAddress() + buffer.Size() > begin) {
-			first = previous;
+BufferCache::OverlapResult BufferCache::ResolveOverlaps(uint64_t vaddr, uint64_t size) {
+	static constexpr int      StreamLeapThreshold = 16;
+	static constexpr uint64_t StreamLeapSize      = CACHING_PAGESIZE * 128;
+
+	auto       begin      = vaddr;
+	auto       end        = vaddr + size;
+	const auto find_first = [&](uint64_t address) {
+		auto first = m_buffers.lower_bound(address);
+		if (first != m_buffers.begin()) {
+			const auto  previous = std::prev(first);
+			const auto& buffer   = m_slot_buffers[previous->second];
+			if (buffer.CpuAddress() + buffer.Size() > address) {
+				first = previous;
+			}
+		}
+		return first;
+	};
+	auto first           = find_first(begin);
+	auto last            = first;
+	int  stream_score    = 0;
+	bool has_stream_leap = false;
+	for (; last != m_buffers.end() && last->first < end; ++last) {
+		const auto& buffer        = m_slot_buffers[last->second];
+		const auto  buffer_begin  = buffer.CpuAddress();
+		const auto  buffer_end    = buffer_begin + buffer.Size();
+		const bool  expands_left  = buffer_begin < begin;
+		const bool  expands_right = buffer_end > end;
+		begin                     = std::min(begin, buffer_begin);
+		end                       = std::max(end, buffer_end);
+		if (!has_stream_leap && (stream_score += buffer.StreamScore()) > StreamLeapThreshold) {
+			has_stream_leap = true;
+			// Reserve space in the incoming stream's direction of growth.
+			// The old buffer extending left of the request predicts growth to the right, and vice versa.
+			if (expands_left) {
+				end += std::min(StreamLeapSize, PageTable::kAddressSpaceSize - end);
+			}
+			if (expands_right) {
+				const auto minimum = CACHING_PAGESIZE * 2;
+				if (begin > minimum) {
+					begin -= std::min(StreamLeapSize, begin - minimum);
+				}
+				first = find_first(begin);
+				begin = std::min(begin, first->first);
+			}
 		}
 	}
-	auto last = first;
-	for (; last != m_buffers.end() && last->first < end; ++last) {
-		const auto& buffer = m_slot_buffers[last->second];
-		begin              = std::min(begin, buffer.CpuAddress());
-		end                = std::max(end, buffer.CpuAddress() + buffer.Size());
+	return {first, last, begin, end, has_stream_leap};
+}
+
+void BufferCache::JoinOverlap(BufferId new_id, BufferId overlap_id, bool accumulate_stream_score) {
+	auto& new_buffer = m_slot_buffers[new_id];
+	auto& overlap    = m_slot_buffers[overlap_id];
+	if (accumulate_stream_score) {
+		new_buffer.IncreaseStreamScore(overlap.StreamScore() + 1);
 	}
+	new_buffer.CopyFrom(m_scheduler.Current(), overlap, 0,
+	                    overlap.CpuAddress() - new_buffer.CpuAddress(), overlap.Size());
+	DeleteBuffer(overlap_id);
+}
+
+BufferId BufferCache::CreateBuffer(uint64_t vaddr, uint64_t size) {
+	EXIT_IF(m_scheduler.Current().IsInvalid());
+	const auto end = Common::AlignUp(vaddr + size, CACHING_PAGESIZE);
+	vaddr = Common::AlignDown(vaddr, CACHING_PAGESIZE);
+	size               = end - vaddr;
+	const auto overlap = ResolveOverlaps(vaddr, size);
 
 	const auto id = m_slot_buffers.insert(
-	    m_graphics, m_scheduler, MemoryUsage::DeviceLocal, begin,
-	    AllFlags | vk::BufferUsageFlagBits::eShaderDeviceAddress, end - begin);
-	auto&      buffer = m_slot_buffers[id];
+	    m_graphics, m_scheduler, MemoryUsage::DeviceLocal, overlap.begin,
+	    AllFlags | vk::BufferUsageFlagBits::eShaderDeviceAddress, overlap.end - overlap.begin);
+	const auto& buffer = m_slot_buffers[id];
 	SetVulkanObjectNameF(m_graphics.device, buffer.Handle(),
-	                     "Kyty.GameBuffer[guest=0x{:016x} size=0x{:x}]", begin, end - begin);
-	for (auto overlap = first; overlap != last;) {
-		const auto current = overlap++;
-		const auto old_id  = current->second;
-		const auto& old    = m_slot_buffers[old_id];
-		buffer.CopyFrom(command, old, 0, old.CpuAddress() - begin, old.Size());
-		DeleteBuffer(old_id);
+	                     "Kyty.GameBuffer[guest=0x{:016x} size=0x{:x}]", overlap.begin,
+	                     overlap.end - overlap.begin);
+	for (auto it = overlap.first; it != overlap.last;) {
+		const auto old_id = (it++)->second;
+		JoinOverlap(id, old_id, !overlap.has_stream_leap);
 	}
 	Register(id);
 	return id;
@@ -381,7 +377,6 @@ bool BufferCache::SynchronizeBuffer(Buffer& buffer, uint64_t vaddr, uint64_t siz
 		command.EndRendering();
 		const auto native = command.Handle();
 		vk::BufferMemoryBarrier before {};
-		before.sType         = vk::StructureType::eBufferMemoryBarrier;
 		before.srcAccessMask = vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite |
 		                       vk::AccessFlagBits::eTransferRead |
 		                       vk::AccessFlagBits::eTransferWrite;
@@ -443,8 +438,7 @@ std::pair<Buffer*, uint64_t> BufferCache::ObtainBuffer(uint64_t vaddr, uint64_t
                                                        bool is_written, bool is_texel_buffer,
                                                        BufferId id) {
 	auto& command = m_scheduler.Current();
-	if (command.IsInvalid() || vaddr == 0 || size == 0 ||
-	    vaddr >= TRACKER_ADDRESS_SIZE || size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (command.IsInvalid() || !GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: buffer request requires a recording command buffer\n");
 	}
 
@@ -460,55 +454,34 @@ std::pair<Buffer*, uint64_t> BufferCache::ObtainBuffer(uint64_t vaddr, uint64_t
 		}
 	}
 
-	auto* buffer = m_slot_buffers.try_get(id);
-	if (buffer == nullptr || buffer->is_deleted || !buffer->IsInBounds(vaddr, size)) {
-		id     = FindBuffer(vaddr, size);
-		buffer = &m_slot_buffers[id];
+	if (IsBufferInvalid(id) || !m_slot_buffers[id].IsInBounds(vaddr, size)) {
+		id = FindBuffer(vaddr, size);
 	}
-	TouchBuffer(*buffer);
-	(void)SynchronizeBuffer(*buffer, vaddr, size, is_written, is_texel_buffer);
+	auto& buffer = m_slot_buffers[id];
+	TouchBuffer(buffer);
+	(void)SynchronizeBuffer(buffer, vaddr, size, is_written, is_texel_buffer);
 	if (is_written) {
 		m_gpu_modified_ranges.Add(vaddr, size);
 	}
-	return {buffer, buffer->Offset(vaddr)};
+	return {&buffer, buffer.Offset(vaddr)};
 }
 
 std::pair<Buffer*, uint64_t> BufferCache::ObtainBufferForImage(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: invalid image source\n");
 	}
-	auto find_owner = [&]() -> Buffer* {
-		const auto* owner = m_page_table.Find(vaddr >> PageTable::kPageBits);
-		if (owner == nullptr || !*owner) {
-			return nullptr;
-		}
+	const auto* owner = m_page_table.Find(vaddr >> PageTable::kPageBits);
+	if (owner != nullptr && *owner) {
 		auto& buffer = m_slot_buffers[*owner];
-		return buffer.IsInBounds(vaddr, size) ? &buffer : nullptr;
-	};
-
-	{
-		const bool cpu_modified            = m_memory_tracker.IsRegionCpuModified(vaddr, size);
-		const bool gpu_modified            = m_memory_tracker.IsRegionGpuModified(vaddr, size);
-		const bool has_dirty_buffer_source = m_gpu_modified_ranges.Intersects(vaddr, size);
-		m_memory_tracker.ValidateGpuDirtyOwnership(m_gpu_modified_ranges, vaddr, size,
-		                                           "image source");
-
-		auto* owner = find_owner();
-		if (has_dirty_buffer_source && owner == nullptr) {
-			if (!IsRegionRegistered(vaddr, size)) {
-				EXIT("BufferCache: GPU-dirty image source has no native buffer\n");
-			}
-			owner = &m_slot_buffers[FindBuffer(vaddr, size)];
-		}
-		if (owner != nullptr && !cpu_modified && (!gpu_modified || has_dirty_buffer_source)) {
-			TouchBuffer(*owner);
-			return {owner, owner->Offset(vaddr)};
-		}
-		if (has_dirty_buffer_source && owner == nullptr) {
-			EXIT("BufferCache: GPU-dirty image source could not resolve its native owner\n");
+		if (buffer.IsInBounds(vaddr, size)) {
+			TouchBuffer(buffer);
+			(void)SynchronizeBuffer(buffer, vaddr, size, false, false);
+			return {&buffer, buffer.Offset(vaddr)};
 		}
 	}
+	if (IsRegionGpuModified(vaddr, size)) {
+		return ObtainBuffer(vaddr, size, false, false);
+	}
 
 	auto [staging, stage_offset] = m_staging_buffer.Map(size, 16);
 	if (staging == nullptr || (!Libs::LibKernel::Memory::TryReadBacking(vaddr, staging, size) &&
@@ -516,52 +489,7 @@ std::pair<Buffer*, uint64_t> BufferCache::ObtainBufferForImage(uint64_t vaddr, u
 		EXIT("BufferCache: failed to read mapped guest image backing\n");
 	}
 	m_staging_buffer.Commit();
-
-	const bool has_dirty_buffer_source = m_gpu_modified_ranges.Intersects(vaddr, size);
-	auto*      owner                   = find_owner();
-	if (has_dirty_buffer_source && owner == nullptr) {
-		EXIT("BufferCache: GPU-dirty image source lost its native owner\n");
-	}
-	if (owner == nullptr ||
-	    (m_memory_tracker.IsRegionGpuModified(vaddr, size) && !has_dirty_buffer_source)) {
-		return {&m_staging_buffer, stage_offset};
-	}
-
-	TouchBuffer(*owner);
-	std::vector<std::pair<uint64_t, uint64_t>> uploads;
-	m_memory_tracker.ForEachUploadRange(
-	    vaddr, size, false,
-	    [&](uint64_t address, uint64_t upload_size) noexcept {
-		    uploads.emplace_back(address, upload_size);
-	    },
-	    [&]() noexcept {
-		    for (const auto& [address, upload_size]: uploads) {
-			    owner->CopyFrom(m_scheduler.Current(), m_staging_buffer,
-			                    stage_offset + address - vaddr, owner->Offset(address), upload_size,
-			                    vk::AccessFlagBits::eHostWrite);
-		    }
-	    });
-	return {owner, owner->Offset(vaddr)};
-}
-
-void BufferCache::WriteHostMemory(uint64_t vaddr, std::span<const uint8_t> data) {
-	if (vaddr == 0 || data.empty() || data.size() > UINT64_MAX - vaddr) {
-		EXIT("BufferCache: invalid host DMA write\n");
-	}
-	Libs::LibKernel::Memory::WriteBacking(vaddr, data.data(), data.size());
-
-	const auto end = vaddr + data.size();
-	for (const auto& [address, id]: m_buffers) {
-		auto&      buffer     = m_slot_buffers[id];
-		const auto buffer_end = address + buffer.Size();
-		const auto begin      = std::max(vaddr, address);
-		const auto range_end  = std::min(end, buffer_end);
-		if (begin >= range_end) {
-			continue;
-		}
-		WriteDataBuffer(buffer, begin, data.data() + begin - vaddr, range_end - begin);
-		TouchBuffer(buffer);
-	}
+	return {&m_staging_buffer, stage_offset};
 }
 
 void BufferCache::FillBuffer(uint64_t vaddr, uint64_t size, uint32_t value, bool is_gds) {
@@ -579,29 +507,15 @@ void BufferCache::FillBuffer(uint64_t vaddr, uint64_t size, uint32_t value, bool
 		EXIT("BufferCache: invalid fill memory address\n");
 	}
 	(void)m_texture_cache.ClearMeta(vaddr);
-	{
-		const auto region = m_texture_cache.QueryRegion(vaddr, size);
-		if (!HasGpuDirtyBytes(vaddr, size) && !region.gpu_image_bytes) {
-			if (region.image_bytes) {
-				m_texture_cache.InvalidateMemory(vaddr, size);
-			}
-			std::array<uint32_t, 4096> values;
-			values.fill(value);
-			const std::span<const uint8_t> bytes {reinterpret_cast<const uint8_t*>(values.data()),
-			                                      sizeof(values)};
-			for (uint64_t offset = 0; offset < size;) {
-				const auto chunk = std::min<uint64_t>(size - offset, bytes.size());
-				WriteHostMemory(vaddr + offset, bytes.first(chunk));
-				offset += chunk;
-			}
-			return;
-		}
+	if (!IsRegionGpuModified(vaddr, size)) {
+		// Access the guest mapping so write faults invalidate cached buffers and images.
+		auto* destination = reinterpret_cast<uint32_t*>(vaddr);
+		std::fill(destination, destination + size / sizeof(uint32_t), value);
+		return;
 	}
 
 	m_texture_cache.InvalidateMemoryFromGPU(vaddr, size);
-	const auto id          = FindBuffer(vaddr, size);
-	auto [dst, dst_offset] = ObtainBuffer(vaddr, size, true, true, id);
-	EXIT_IF(dst == nullptr);
+	auto [dst, dst_offset] = ObtainBuffer(vaddr, size, true, true);
 	dst->Fill(dst_offset, size, value);
 }
 
@@ -618,29 +532,11 @@ void BufferCache::CopyBuffer(uint64_t dst_vaddr, uint64_t src_vaddr, uint64_t si
 		     " size=0x%016" PRIx64 " src_gds=%d dst_gds=%d\n",
 		     src_vaddr, dst_vaddr, size, static_cast<int>(src_gds), static_cast<int>(dst_gds));
 	}
-	if (src_memory || dst_memory) {
-		const auto src_region =
-		    src_memory ? m_texture_cache.QueryRegion(src_vaddr, size) : TextureCache::RegionInfo {};
-		const auto dst_region =
-		    dst_memory ? m_texture_cache.QueryRegion(dst_vaddr, size) : TextureCache::RegionInfo {};
-		if (src_memory && dst_memory && !HasGpuDirtyBytes(src_vaddr, size) &&
-		    !HasGpuDirtyBytes(dst_vaddr, size) && !src_region.gpu_image_bytes &&
-		    !dst_region.gpu_image_bytes) {
-			if (dst_region.image_bytes) {
-				m_texture_cache.InvalidateMemory(dst_vaddr, size);
-			}
-			std::array<uint8_t, 64 * 1024> bytes;
-			for (uint64_t offset = 0; offset < size;) {
-				const auto chunk = std::min<uint64_t>(size - offset, bytes.size());
-				if (!Libs::LibKernel::Memory::TryReadBacking(src_vaddr + offset, bytes.data(),
-				                                             chunk)) {
-					EXIT("BufferCache: host DMA source has no direct backing\n");
-				}
-				WriteHostMemory(dst_vaddr + offset, std::span {bytes}.first(chunk));
-				offset += chunk;
-			}
-			return;
-		}
+	if (src_memory && dst_memory && !IsRegionGpuModified(dst_vaddr, size) &&
+	    !IsRegionGpuModified(src_vaddr, size) && !m_texture_cache.FindImageFromRange(src_vaddr, size)) {
+		std::memcpy(reinterpret_cast<void*>(dst_vaddr), reinterpret_cast<const void*>(src_vaddr),
+		            size);
+		return;
 	}
 
 	auto& command = m_scheduler.Current();
@@ -653,16 +549,11 @@ void BufferCache::CopyBuffer(uint64_t dst_vaddr, uint64_t src_vaddr, uint64_t si
 	                                    : std::pair {&m_gds_buffer, src_vaddr};
 	auto [dst, dst_offset] = dst_memory ? ObtainBuffer(dst_vaddr, size, true, true, dst_id)
 	                                    : std::pair {&m_gds_buffer, dst_vaddr};
-	EXIT_IF(src == nullptr || dst == nullptr);
-	if (src == dst && src_offset < dst_offset + size && dst_offset < src_offset + size) {
-		EXIT("BufferCache: resolved Vulkan copy ranges overlap\n");
-	}
 	dst->CopyFrom(command, *src, src_offset, dst_offset, size);
 }
 
 bool BufferCache::IsRegionRegistered(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
+	if (!GuestRange {vaddr, size}.Valid()) {
 		EXIT("BufferCache: invalid registered-region query\n");
 	}
 	// Cached buffers are ordered and non-overlapping. The last buffer beginning before the query
@@ -701,8 +592,7 @@ void BufferCache::RunGarbageCollector() {
 	const size_t   limit      = aggressive ? 64 : 32;
 
 	std::vector<BufferId> dirty_buffers;
-	std::vector<DownloadCopy> copies;
-	size_t                    retire_count = 0;
+	size_t                retire_count = 0;
 	m_lru_cache.ForEachItemBelow(tick - age, [&](BufferId id) {
 		auto& buffer = m_slot_buffers[id];
 		EXIT_IF(buffer.is_deleted);
@@ -713,19 +603,7 @@ void BufferCache::RunGarbageCollector() {
 			return false;
 		}
 		if (dirty) {
-			m_memory_tracker.ForEachDownloadRange<false>(
-			    buffer.CpuAddress(), buffer.Size(),
-			    [&](uint64_t dirty_address, uint64_t dirty_size) noexcept {
-				    m_memory_tracker.ValidateGpuDirtyPages(m_gpu_modified_ranges, dirty_address,
-				                                           dirty_size, "garbage collection");
-			    },
-			    [&](uint64_t dirty_address, uint64_t dirty_size) noexcept {
-				    m_gpu_modified_ranges.ForEachIntersection(
-				        dirty_address, dirty_size, [&](RangeSet::Range range) {
-					    copies.push_back({&buffer, range.address - buffer.CpuAddress(),
-					                      range.address, range.size});
-				        });
-				});
+			EXIT_IF(!DownloadBufferMemory(buffer, buffer.CpuAddress(), buffer.Size()));
 			dirty_buffers.push_back(id);
 		} else {
 			m_memory_tracker.UntrackMemory(buffer.CpuAddress(), buffer.Size());
@@ -737,8 +615,10 @@ void BufferCache::RunGarbageCollector() {
 		return;
 	}
 
-	EXIT_IF(copies.empty());
-	DownloadBufferMemory(copies);
+	// Publish all queued downloads before releasing their tracked pages and owners.
+	const auto completion_tick = m_scheduler.CurrentTick();
+	m_scheduler.Wait(completion_tick);
+	m_scheduler.WaitPriorityOperations(completion_tick);
 	for (const auto id: dirty_buffers) {
 		auto& buffer = m_slot_buffers[id];
 		m_memory_tracker.UnmarkRegionAsGpuModified(buffer.CpuAddress(), buffer.Size());

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

@@ -74,16 +74,26 @@ public:
 private:
 	friend struct BufferCacheTestAccess;
 
-	struct DownloadCopy;
+	bool IsBufferInvalid(BufferId id) const {
+		const auto* buffer = m_slot_buffers.try_get(id);
+		return buffer == nullptr || buffer->is_deleted;
+	}
+
+	using BufferMap = std::map<uint64_t, BufferId>;
+	struct OverlapResult {
+		BufferMap::iterator first;
+		BufferMap::iterator last;
+		uint64_t            begin;
+		uint64_t            end;
+		bool                has_stream_leap;
+	};
+
 	using PageTable = MultiLevelPageTable<BufferId, CACHING_PAGEBITS, 40, 16>;
 	static_assert(CACHING_PAGESIZE == (uint64_t {1} << PageTable::kPageBits));
-	static constexpr uint64_t               DOWNLOAD_ALIGNMENT = 64;
-	[[nodiscard]] static constexpr uint64_t AlignDownload(uint64_t size) noexcept {
-		return (size + DOWNLOAD_ALIGNMENT - 1) & ~(DOWNLOAD_ALIGNMENT - 1);
-	}
-	[[nodiscard]] static std::pair<uint64_t, uint64_t> DownloadEnvelope(const DownloadCopy& copy);
 	void WriteDataBuffer(Buffer& buffer, uint64_t address, const void* source, uint64_t size);
 	void TouchBuffer(const Buffer& buffer);
+	[[nodiscard]] OverlapResult ResolveOverlaps(uint64_t vaddr, uint64_t size);
+	void JoinOverlap(BufferId new_id, BufferId overlap_id, bool accumulate_stream_score);
 	[[nodiscard]] BufferId CreateBuffer(uint64_t vaddr, uint64_t size);
 	void                   Register(BufferId id);
 	void Unregister(BufferId id);
@@ -95,9 +105,8 @@ private:
 	[[nodiscard]] vk::Buffer UploadCopies(Buffer& buffer, std::span<vk::BufferCopy> copies,
 	                                      uint64_t total_size);
 	[[nodiscard]] bool SynchronizeBufferFromImage(Buffer& buffer, uint64_t vaddr, uint64_t size);
-	void DownloadBufferMemory(std::span<const DownloadCopy> copies);
-	void WriteHostMemory(uint64_t vaddr, std::span<const uint8_t> data);
-	void ReadMemoryOnGpu(uint64_t vaddr, uint64_t size, bool is_write);
+	// Queues backing publication; callers wait before clearing dirty pages or reusing their data.
+	[[nodiscard]] bool DownloadBufferMemory(Buffer& buffer, uint64_t vaddr, uint64_t size);
 
 	GraphicContext&                                   m_graphics;
 	CommandScheduler&                                 m_scheduler;
@@ -106,7 +115,7 @@ private:
 	Buffer                                            m_bda_pagetable_buffer;
 	Common::SlotVector<Buffer>                        m_slot_buffers;
 	Common::LeastRecentlyUsedCache<BufferId, uint64_t> m_lru_cache;
-	std::map<uint64_t, BufferId>                      m_buffers;
+	BufferMap                                         m_buffers;
 	PageTable                                         m_page_table;
 	RangeSet                                          m_gpu_modified_ranges;
 	MemoryTracker                                     m_memory_tracker;

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

@@ -23,13 +23,10 @@ constexpr size_t PageFaultAreaSize = MaxPageFaults * sizeof(uint64_t);
 } // namespace
 
 FaultManager::FaultManager(GraphicContext& graphics, CommandScheduler& scheduler,
-                           BufferCache& buffer_cache, uint32_t caching_pagebits,
-                           uint64_t caching_num_pages)
+                           BufferCache& buffer_cache)
     : m_graphics(graphics), m_scheduler(scheduler), m_buffer_cache(buffer_cache),
-      m_caching_pagesize(uint64_t {1} << caching_pagebits),
-      m_caching_num_pages(caching_num_pages), m_fault_buffer_size(caching_num_pages / 8),
       m_fault_buffer(graphics, scheduler, MemoryUsage::DeviceLocal, 0, AllFlags,
-                     m_fault_buffer_size),
+                     BufferCache::CACHING_NUMPAGES / 8),
       m_download_buffer(graphics, scheduler, MemoryUsage::Download, 0, AllFlags,
                         MaxPendingFaults * PageFaultAreaSize) {
 	SetVulkanObjectNameF(m_graphics.device, m_fault_buffer.Handle(), "Fault Buffer");
@@ -39,7 +36,6 @@ FaultManager::FaultManager(GraphicContext& graphics, CommandScheduler& scheduler
 	    {1, vk::DescriptorType::eStorageBuffer, 1, vk::ShaderStageFlagBits::eCompute, nullptr},
 	};
 	vk::DescriptorSetLayoutCreateInfo layout_info {};
-	layout_info.sType        = vk::StructureType::eDescriptorSetLayoutCreateInfo;
 	layout_info.flags        = vk::DescriptorSetLayoutCreateFlagBits::ePushDescriptorKHR;
 	layout_info.bindingCount = std::size(bindings);
 	layout_info.pBindings    = bindings;
@@ -48,16 +44,9 @@ FaultManager::FaultManager(GraphicContext& graphics, CommandScheduler& scheduler
 	                                                &m_fault_process_desc_layout),
 	    "create fault-buffer descriptor layout");
 
-	vk::ShaderModuleCreateInfo module_info {};
-	module_info.sType    = vk::StructureType::eShaderModuleCreateInfo;
-	module_info.codeSize = std::size(FAULT_BUFFER_PROCESS_SPV) * sizeof(uint32_t);
-	module_info.pCode    = FAULT_BUFFER_PROCESS_SPV;
-	vk::ShaderModule module = nullptr;
-	RequireVulkanSuccess(m_graphics.device.createShaderModule(&module_info, nullptr, &module),
-	                     "create fault-buffer shader module");
+	const auto module = CompileSPV(FAULT_BUFFER_PROCESS_SPV, m_graphics.device);
 
 	vk::PipelineLayoutCreateInfo pipeline_layout_info {};
-	pipeline_layout_info.sType          = vk::StructureType::ePipelineLayoutCreateInfo;
 	pipeline_layout_info.setLayoutCount = 1;
 	pipeline_layout_info.pSetLayouts    = &m_fault_process_desc_layout;
 	RequireVulkanSuccess(
@@ -66,12 +55,10 @@ FaultManager::FaultManager(GraphicContext& graphics, CommandScheduler& scheduler
 	    "create fault-buffer pipeline layout");
 
 	vk::PipelineShaderStageCreateInfo stage {};
-	stage.sType  = vk::StructureType::ePipelineShaderStageCreateInfo;
 	stage.stage  = vk::ShaderStageFlagBits::eCompute;
 	stage.module = module;
 	stage.pName  = "main";
 	vk::ComputePipelineCreateInfo pipeline_info {};
-	pipeline_info.sType  = vk::StructureType::eComputePipelineCreateInfo;
 	pipeline_info.stage  = stage;
 	pipeline_info.layout = m_fault_process_pipeline_layout;
 	const auto result = m_graphics.device.createComputePipelines(
@@ -99,14 +86,13 @@ void FaultManager::ProcessFaultBuffer() {
 	m_download_buffer.Flush(offset, PageFaultAreaSize);
 
 	vk::BufferMemoryBarrier2 pre_barrier {};
-	pre_barrier.sType         = vk::StructureType::eBufferMemoryBarrier2;
 	pre_barrier.srcStageMask  = vk::PipelineStageFlagBits2::eAllCommands;
 	pre_barrier.srcAccessMask = vk::AccessFlagBits2::eShaderWrite;
 	pre_barrier.dstStageMask  = vk::PipelineStageFlagBits2::eComputeShader;
 	pre_barrier.dstAccessMask = vk::AccessFlagBits2::eShaderRead;
 	pre_barrier.buffer        = m_fault_buffer.Handle();
 	pre_barrier.offset        = 0;
-	pre_barrier.size          = m_fault_buffer_size;
+	pre_barrier.size           = m_fault_buffer.Size();
 	auto post_barrier         = pre_barrier;
 	post_barrier.srcStageMask  = vk::PipelineStageFlagBits2::eComputeShader;
 	post_barrier.srcAccessMask = vk::AccessFlagBits2::eShaderWrite;
@@ -114,12 +100,11 @@ void FaultManager::ProcessFaultBuffer() {
 	post_barrier.dstAccessMask = vk::AccessFlagBits2::eShaderWrite;
 
 	const vk::DescriptorBufferInfo infos[] {
-	    {m_fault_buffer.Handle(), 0, m_fault_buffer_size},
+	    {m_fault_buffer.Handle(), 0, m_fault_buffer.Size()},
 	    {m_download_buffer.Handle(), offset, PageFaultAreaSize},
 	};
 	std::array<vk::WriteDescriptorSet, 2> writes {};
 	for (uint32_t index = 0; index < writes.size(); ++index) {
-		writes[index].sType           = vk::StructureType::eWriteDescriptorSet;
 		writes[index].dstBinding      = index;
 		writes[index].descriptorCount = 1;
 		writes[index].descriptorType  = vk::DescriptorType::eStorageBuffer;
@@ -129,7 +114,6 @@ void FaultManager::ProcessFaultBuffer() {
 	m_scheduler.EndRendering();
 	auto command = m_scheduler.Current().Handle();
 	vk::DependencyInfo dependency {};
-	dependency.sType                    = vk::StructureType::eDependencyInfo;
 	dependency.dependencyFlags          = vk::DependencyFlagBits::eByRegion;
 	dependency.bufferMemoryBarrierCount = 1;
 	dependency.pBufferMemoryBarriers    = &pre_barrier;
@@ -137,7 +121,7 @@ void FaultManager::ProcessFaultBuffer() {
 	command.bindPipeline(vk::PipelineBindPoint::eCompute, m_fault_process_pipeline);
 	command.pushDescriptorSetKHR(vk::PipelineBindPoint::eCompute,
 	                             m_fault_process_pipeline_layout, 0, writes);
-	const auto num_threads    = m_caching_num_pages / 32;
+	const auto num_threads    = BufferCache::CACHING_NUMPAGES / 32;
 	const auto num_workgroups = (num_threads + 63) / 64;
 	command.dispatch(static_cast<uint32_t>(num_workgroups), 1, 1);
 	dependency.pBufferMemoryBarriers = &post_barrier;
@@ -146,14 +130,14 @@ void FaultManager::ProcessFaultBuffer() {
 	const auto area = m_current_area;
 	m_scheduler.DeferOperation([this, mapped, offset, area] {
 		m_download_buffer.Invalidate(offset, PageFaultAreaSize);
-		m_fault_ranges.Clear();
+		RangeSet    fault_ranges;
 		const auto* faults = std::bit_cast<const uint64_t*>(mapped);
 		const auto  count  = static_cast<uint32_t>(faults[0]);
 		for (uint32_t index = 1; index <= count; ++index) {
-			m_fault_ranges.Add(faults[index], m_caching_pagesize);
+			fault_ranges.Add(faults[index], BufferCache::CACHING_PAGESIZE);
 			LOGF("Accessed non-GPU cached memory at 0x%016" PRIx64 "\n", faults[index]);
 		}
-		m_fault_ranges.ForEach([this](uint64_t start, uint64_t end) {
+		fault_ranges.ForEach([this](uint64_t start, uint64_t end) {
 			EXIT_IF(end - start > std::numeric_limits<uint32_t>::max());
 			(void)m_buffer_cache.FindBuffer(start, end - start);
 		});

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

@@ -2,7 +2,6 @@
 #define EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_CACHE_FAULTMANAGER_H_
 
 #include "common/abi.h"
-#include "graphics/host_gpu/rangeSet.h"
 #include "graphics/host_gpu/renderer/cache/streamBuffer.h"
 
 #include <array>
@@ -16,8 +15,7 @@ class FaultManager {
 	static constexpr size_t MaxPendingFaults = 8;
 
 public:
-	FaultManager(GraphicContext& graphics, CommandScheduler& scheduler, BufferCache& buffer_cache,
-	             uint32_t caching_pagebits, uint64_t caching_num_pages);
+	FaultManager(GraphicContext& graphics, CommandScheduler& scheduler, BufferCache& buffer_cache);
 	~FaultManager();
 	KYTY_CLASS_NO_COPY(FaultManager);
 
@@ -28,10 +26,6 @@ private:
 	GraphicContext&                            m_graphics;
 	CommandScheduler&                          m_scheduler;
 	BufferCache&                               m_buffer_cache;
-	RangeSet                                   m_fault_ranges;
-	uint64_t                                   m_caching_pagesize;
-	uint64_t                                   m_caching_num_pages;
-	uint64_t                                   m_fault_buffer_size;
 	Buffer                                     m_fault_buffer;
 	Buffer                                     m_download_buffer;
 	std::array<uint64_t, MaxPendingFaults>      m_fault_areas {};

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

@@ -1,97 +0,0 @@
-#include "graphics/host_gpu/renderer/cache/gpuResourceManager.h"
-
-#include "common/assert.h"
-#include "graphics/guest_gpu/graphicsRun.h"
-#include "graphics/host_gpu/renderer/commandScheduler.h"
-namespace Libs::Graphics {
-
-GpuResourceManager::GpuResourceManager(GraphicContext& graphics, CommandScheduler& scheduler)
-    : m_scheduler(scheduler), m_buffer_cache(graphics, scheduler, m_page_manager, m_texture_cache),
-      m_texture_cache(graphics, scheduler, m_page_manager, m_buffer_cache) {}
-
-GpuResourceManager::~GpuResourceManager() = default;
-
-bool GpuResourceManager::HandleFault(PageFaultAccess access, uint64_t fault_vaddr) noexcept {
-	constexpr uint64_t fault_size = 8;
-	if (!IsMapped(fault_vaddr, fault_size)) {
-		return false;
-	}
-	if (access == PageFaultAccess::Write) {
-		m_buffer_cache.InvalidateMemory(fault_vaddr, fault_size);
-		m_texture_cache.InvalidateMemory(fault_vaddr, fault_size);
-	} else {
-		m_buffer_cache.ReadMemory(fault_vaddr, fault_size);
-	}
-	return true;
-}
-
-bool GpuResourceManager::InvalidateMemory(uint64_t vaddr, uint64_t size) {
-	if (!IsMapped(vaddr, size)) {
-		return false;
-	}
-	m_buffer_cache.InvalidateMemory(vaddr, size);
-	m_texture_cache.InvalidateMemory(vaddr, size);
-	return true;
-}
-
-bool GpuResourceManager::IsMapped(uint64_t vaddr, uint64_t size) const noexcept {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
-		return false;
-	}
-	std::shared_lock lock(m_mapped_ranges_mutex);
-	return m_mapped_ranges.Contains(vaddr, size);
-}
-
-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);
-}
-
-void GpuResourceManager::UnmapMemory(uint64_t vaddr, uint64_t size) {
-	if (CommandScheduler::InDeferredOperation()) {
-		EXIT("unsupported memory unmap from an asynchronous GPU completion, "
-		     "addr=0x%016" PRIx64 " size=0x%016" PRIx64 "\n",
-		     vaddr, size);
-	}
-	const auto unmap = [this, vaddr, size] {
-		if (m_scheduler.Active()) {
-			const auto tick = m_scheduler.CurrentTick();
-			m_scheduler.Finish();
-			m_scheduler.WaitPriorityOperations(tick);
-		}
-		m_buffer_cache.InvalidateMemory(vaddr, size);
-		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);
-	};
-	if (m_gpu == nullptr) {
-		unmap();
-		return;
-	}
-	m_gpu->SendCommandSync(unmap);
-}
-
-void GpuResourceManager::PrepareBda() {
-	std::shared_lock lock(m_mapped_ranges_mutex);
-	m_mapped_ranges.ForEach([this](uint64_t start, uint64_t end) {
-		m_buffer_cache.SynchronizeBuffersInRange(start, end - start);
-	});
-	m_fault_process_pending = true;
-}
-
-void GpuResourceManager::RunGarbageCollector() {
-	if (m_fault_process_pending) {
-		m_fault_process_pending = false;
-		m_buffer_cache.ProcessFaultBuffer();
-	}
-	m_texture_cache.ProcessDownloadImages();
-	m_texture_cache.RunGarbageCollector();
-	m_buffer_cache.RunGarbageCollector();
-}
-
-} // namespace Libs::Graphics

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

@@ -1,49 +0,0 @@
-#ifndef EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_GPURESOURCEMANAGER_H_
-#define EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_GPURESOURCEMANAGER_H_
-
-#include "common/abi.h"
-#include "common/common.h"
-#include "graphics/host_gpu/pageManager.h"
-#include "graphics/host_gpu/renderer/cache/bufferCache.h"
-#include "graphics/host_gpu/renderer/cache/textureCache.h"
-
-#include <cstdint>
-#include <shared_mutex>
-
-namespace Libs::Graphics {
-
-class CommandScheduler;
-class GuestGpu;
-
-class GpuResourceManager {
-public:
-	GpuResourceManager(GraphicContext& graphics, CommandScheduler& scheduler);
-	~GpuResourceManager();
-	KYTY_CLASS_NO_COPY(GpuResourceManager);
-
-	[[nodiscard]] BufferCache&  GetBufferCache() { return m_buffer_cache; }
-	[[nodiscard]] TextureCache& GetTextureCache() { return m_texture_cache; }
-	void                        SetGpu(GuestGpu* gpu) noexcept { m_gpu = gpu; }
-
-	[[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);
-	void               UnmapMemory(uint64_t vaddr, uint64_t size);
-	void               PrepareBda();
-	void               RunGarbageCollector();
-
-private:
-	PageManager               m_page_manager;
-	CommandScheduler&         m_scheduler;
-	BufferCache               m_buffer_cache;
-	TextureCache              m_texture_cache;
-	mutable std::shared_mutex m_mapped_ranges_mutex;
-	RangeSet                  m_mapped_ranges;
-	GuestGpu*                 m_gpu = nullptr;
-	bool                      m_fault_process_pending = false;
-};
-
-} // namespace Libs::Graphics
-
-#endif // EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_GPURESOURCEMANAGER_H_

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

@@ -256,17 +256,6 @@ private:
 	size_type                              m_inline_size = 0;
 };
 
-// Removes only the requested owner. Other owners in the same page entry remain intact.
-template <typename Container, typename Value>
-[[nodiscard]] bool EraseExact(Container& owners, const Value& owner) {
-	const auto it = std::find(owners.begin(), owners.end(), owner);
-	if (it == owners.end()) {
-		return false;
-	}
-	owners.erase(it);
-	return true;
-}
-
 } // namespace Libs::Graphics
 
 #endif // EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_MULTILEVELPAGETABLE_H_

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

@@ -116,9 +116,6 @@ vk::Sampler SamplerCache::GetSampler(const ShaderSamplerResource& r) {
 		default: EXIT("unknown border color: %d", static_cast<int>(r.BorderColorType()));
 	}
 
-	sampler_info.sType     = vk::StructureType::eSamplerCreateInfo;
-	sampler_info.pNext     = nullptr;
-	sampler_info.flags     = {};
 	sampler_info.magFilter = to_vk_filter(mag_filter);
 	sampler_info.minFilter = to_vk_filter(min_filter);
 	sampler_info.mipmapMode =

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

@@ -1,13 +1,12 @@
 #include "graphics/host_gpu/renderer/cache/streamBuffer.h"
 
+#include "common/alignment.h"
 #include "common/assert.h"
 #include "common/profiler.h"
 #include "graphics/host_gpu/graphicContext.h"
 #include "graphics/host_gpu/renderer/commandScheduler.h"
-#include "graphics/host_gpu/vma.h"
 
 #include <cstring>
-#include <limits>
 #include <numeric>
 #include <vk_mem_alloc.h>
 
@@ -46,16 +45,11 @@ constexpr size_t WATCHES_RESERVE_CHUNK   = 0x1000;
 		result = value;
 		return true;
 	}
-	const auto remainder = value % alignment;
-	if (remainder == 0) {
-		result = value;
-		return true;
-	}
-	const auto increment = alignment - remainder;
-	if (value > std::numeric_limits<uint64_t>::max() - increment) {
+	const auto aligned = Common::AlignUp(value, alignment);
+	if (aligned < value) {
 		return false;
 	}
-	result = value + increment;
+	result = aligned;
 	return true;
 }
 
@@ -64,14 +58,13 @@ constexpr size_t WATCHES_RESERVE_CHUNK   = 0x1000;
 Buffer::Buffer(GraphicContext& graphics, CommandScheduler& scheduler, MemoryUsage usage,
                uint64_t cpu_address, vk::BufferUsageFlags flags, uint64_t size)
     : m_graphics(&graphics), m_scheduler(&scheduler), m_usage(usage), m_cpu_address(cpu_address),
-      m_size(size), m_buffer(std::make_unique<VulkanBuffer>()) {
+      m_size(size) {
 	KYTY_PROFILER_FUNCTION();
 	EXIT_IF(graphics.allocator == nullptr || size == 0);
 
 	vk::BufferCreateInfo buffer_info {};
 	buffer_info.size        = size;
 	buffer_info.usage       = flags;
-	buffer_info.sharingMode = vk::SharingMode::eExclusive;
 
 	const bool with_bda = bool(flags & vk::BufferUsageFlagBits::eShaderDeviceAddress);
 	const VmaAllocationCreateFlags bda_flag =
@@ -88,93 +81,70 @@ Buffer::Buffer(GraphicContext& graphics, CommandScheduler& scheduler, MemoryUsag
 	VkBuffer          native_buffer = VK_NULL_HANDLE;
 	const auto        result        = static_cast<vk::Result>(vmaCreateBuffer(
 	    graphics.allocator, static_cast<const VkBufferCreateInfo*>(buffer_info), &allocation_info,
-	    &native_buffer, &m_buffer->memory.allocation, &allocation_result));
+	    &native_buffer, &m_allocation, &allocation_result));
 	if (result != vk::Result::eSuccess) {
 		graphics.LogMemoryBudget();
 	}
 	EXIT_NOT_IMPLEMENTED(result != vk::Result::eSuccess);
 
-	m_buffer->buffer                 = native_buffer;
-	m_buffer->usage                  = flags;
-	m_buffer->buffer_size            = size;
-	m_buffer->memory.allocation_info = allocation_result;
-	m_buffer->memory.memory          = allocation_result.deviceMemory;
-	m_buffer->memory.offset          = allocation_result.offset;
-	m_buffer->memory.type            = allocation_result.memoryType;
-	m_buffer->memory.unique_id       = VulkanNextMemoryUniqueId();
-	graphics.device.getBufferMemoryRequirements(m_buffer->buffer, &m_buffer->memory.requirements);
-	if (static_cast<bool>(flags & vk::BufferUsageFlagBits::eShaderDeviceAddress)) {
+	m_buffer = native_buffer;
+	if (with_bda) {
 		vk::BufferDeviceAddressInfo address_info {};
-		address_info.sType  = vk::StructureType::eBufferDeviceAddressInfo;
-		address_info.buffer = m_buffer->buffer;
+		address_info.buffer = m_buffer;
 		m_device_address    = graphics.device.getBufferAddress(address_info);
 		EXIT_IF(m_device_address == 0);
 	}
 
 	VkMemoryPropertyFlags properties = 0;
-	vmaGetAllocationMemoryProperties(graphics.allocator, m_buffer->memory.allocation, &properties);
-	m_buffer->memory.property = vk::MemoryPropertyFlags(properties);
-	m_is_coherent             = (properties & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) != 0;
+	vmaGetAllocationMemoryProperties(graphics.allocator, m_allocation, &properties);
+	m_coherent = (properties & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) != 0;
 	if (allocation_result.pMappedData != nullptr) {
 		m_mapped = {static_cast<uint8_t*>(allocation_result.pMappedData),
 		            static_cast<size_t>(size)};
 	}
-	VulkanTrackAllocation(m_buffer->memory);
 }
 
 Buffer::~Buffer() {
-	if (m_buffer->buffer != nullptr) {
-		VulkanUntrackAllocation(m_buffer->memory);
-		vmaDestroyBuffer(m_graphics->allocator, m_buffer->buffer, m_buffer->memory.allocation);
+	if (m_buffer != nullptr) {
+		vmaDestroyBuffer(m_graphics->allocator, m_buffer, m_allocation);
 	}
 }
 
-vk::Buffer Buffer::Handle() const noexcept {
-	return m_buffer->buffer;
-}
-
 vk::DeviceAddress Buffer::BufferDeviceAddress() const noexcept {
 	EXIT_IF(m_device_address == 0);
 	return m_device_address;
 }
 
 bool Buffer::IsInBounds(uint64_t address, uint64_t size) const noexcept {
-	return address >= m_cpu_address && size <= m_size && address - m_cpu_address <= m_size - size;
-}
-
-void Buffer::Write(uint64_t offset, const void* source, uint64_t size) {
-	EXIT_IF(source == nullptr || m_mapped.empty() || offset > m_size || size > m_size - offset);
-	std::memcpy(m_mapped.data() + offset, source, static_cast<size_t>(size));
-	Flush(offset, size);
+	return address >= m_cpu_address && size <= Size() && address - m_cpu_address <= Size() - 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) {
+	EXIT_IF(m_mapped.empty() || offset > Size() || size > Size() - offset);
+	if (!IsCoherent() && size != 0) {
 		const auto result =
-		    vmaFlushAllocation(m_graphics->allocator, m_buffer->memory.allocation, offset, size);
+		    vmaFlushAllocation(m_graphics->allocator, m_allocation, offset, size);
 		EXIT_NOT_IMPLEMENTED(static_cast<vk::Result>(result) != vk::Result::eSuccess);
 	}
 }
 
 void Buffer::Invalidate(uint64_t offset, uint64_t size) {
-	EXIT_IF(m_usage != MemoryUsage::Download || offset > m_size || size > m_size - offset);
-	if (!m_is_coherent && size != 0) {
+	EXIT_IF(m_usage != MemoryUsage::Download || offset > Size() || size > Size() - offset);
+	if (!IsCoherent() && size != 0) {
 		const auto result =
-		    vmaInvalidateAllocation(m_graphics->allocator, m_buffer->memory.allocation, offset, size);
+		    vmaInvalidateAllocation(m_graphics->allocator, m_allocation, offset, size);
 		EXIT_NOT_IMPLEMENTED(static_cast<vk::Result>(result) != vk::Result::eSuccess);
 	}
 }
 
 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) {
+	if (Handle() == nullptr || size == 0 || offset > Size() || size > Size() - offset) {
 		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);
+		     static_cast<const void*>(Handle()), offset, size, Size());
 	}
 	vk::BufferMemoryBarrier barrier {};
-	barrier.sType               = vk::StructureType::eBufferMemoryBarrier;
 	barrier.srcAccessMask       = source;
 	barrier.dstAccessMask       = destination;
 	barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
@@ -189,8 +159,8 @@ void Buffer::CopyFrom(CommandBuffer& command, const Buffer& source, uint64_t sou
                       uint64_t destination_offset, uint64_t size, vk::AccessFlags source_before,
                       vk::AccessFlags destination_before, vk::AccessFlags source_after,
                       vk::AccessFlags destination_after) {
-	if (size == 0 || source_offset > source.m_size || size > source.m_size - source_offset ||
-	    destination_offset > m_size || size > m_size - destination_offset) {
+	if (size == 0 || source_offset > source.Size() || size > source.Size() - source_offset ||
+	    destination_offset > Size() || size > Size() - destination_offset) {
 		EXIT("Buffer: invalid copy range\n");
 	}
 	if (source.Handle() == Handle() && source_offset < destination_offset + size &&
@@ -247,10 +217,8 @@ void Buffer::Fill(uint64_t offset, uint64_t size, uint32_t value) {
 
 StreamBuffer::StreamBuffer(GraphicContext& graphics, CommandScheduler& scheduler, MemoryUsage usage,
                            uint64_t size)
-    : Buffer(graphics, scheduler, usage, 0, AllFlags, size) {
-	ReserveWatches(m_current_watches, WATCHES_INITIAL_RESERVE);
-	ReserveWatches(m_previous_watches, WATCHES_INITIAL_RESERVE);
-}
+    : Buffer(graphics, scheduler, usage, 0, AllFlags, size),
+      m_current_watches(WATCHES_INITIAL_RESERVE), m_previous_watches(WATCHES_INITIAL_RESERVE) {}
 
 bool StreamBuffer::NormalizeReservation(bool coherent, uint64_t atom, uint64_t& size,
                                         uint64_t& alignment) {
@@ -315,10 +283,8 @@ std::pair<uint8_t*, uint64_t> StreamBuffer::Map(uint64_t size, uint64_t alignmen
 }
 
 void StreamBuffer::Commit() {
-	if (!IsCoherent() && Usage() != MemoryUsage::Download && m_mapped_size != 0) {
-		const auto result = vmaFlushAllocation(
-		    Graphics().allocator, NativeBuffer().memory.allocation, m_offset, m_mapped_size);
-		EXIT_NOT_IMPLEMENTED(static_cast<vk::Result>(result) != vk::Result::eSuccess);
+	if (Usage() != MemoryUsage::Download && m_mapped_size != 0) {
+		Flush(m_offset, m_mapped_size);
 	}
 
 	m_offset += m_mapped_size;
@@ -328,7 +294,7 @@ void StreamBuffer::Commit() {
 		return;
 	}
 	if (m_current_watch_cursor + 1 >= m_current_watches.size()) {
-		ReserveWatches(m_current_watches, WATCHES_RESERVE_CHUNK);
+		m_current_watches.resize(m_current_watches.size() + WATCHES_RESERVE_CHUNK);
 	}
 	auto& watch       = m_current_watches[m_current_watch_cursor++];
 	watch.upper_bound = m_offset;
@@ -344,10 +310,6 @@ uint64_t StreamBuffer::Copy(const void* source, uint64_t size, uint64_t alignmen
 	return offset;
 }
 
-void StreamBuffer::ReserveWatches(std::vector<Watch>& watches, size_t grow_size) {
-	watches.resize(watches.size() + grow_size);
-}
-
 bool StreamBuffer::WaitPendingOperations(const std::vector<Watch>& watches,
                                          std::optional<size_t>     invalidation_mark,
                                          uint64_t requested_upper_bound, bool allow_wait,

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

@@ -6,19 +6,19 @@
 #include "graphics/host_gpu/vulkanCommon.h"
 
 #include <cstdint>
-#include <memory>
 #include <optional>
 #include <span>
 #include <utility>
 #include <vector>
 
+VK_DEFINE_HANDLE(VmaAllocation)
+
 namespace Libs::Graphics {
 
 class CommandBuffer;
 class CommandScheduler;
 struct StreamBufferTestAccess;
 struct GraphicContext;
-struct VulkanBuffer;
 
 enum class MemoryUsage : uint8_t {
 	DeviceLocal,
@@ -42,10 +42,10 @@ public:
 	~Buffer();
 	KYTY_CLASS_NO_COPY(Buffer);
 
-	[[nodiscard]] vk::Buffer         Handle() const noexcept;
+	[[nodiscard]] vk::Buffer         Handle() const noexcept { return m_buffer; }
 	[[nodiscard]] uint64_t           Size() const noexcept { return m_size; }
 	[[nodiscard]] std::span<uint8_t> Mapped() const noexcept { return m_mapped; }
-	[[nodiscard]] bool               IsCoherent() const noexcept { return m_is_coherent; }
+	[[nodiscard]] bool               IsCoherent() const noexcept { return m_coherent; }
 	[[nodiscard]] MemoryUsage        Usage() const noexcept { return m_usage; }
 	[[nodiscard]] uint64_t           CpuAddress() const noexcept { return m_cpu_address; }
 	[[nodiscard]] vk::DeviceAddress BufferDeviceAddress() const noexcept;
@@ -53,7 +53,8 @@ public:
 		return address - m_cpu_address;
 	}
 	[[nodiscard]] bool IsInBounds(uint64_t address, uint64_t size) const noexcept;
-	void               Write(uint64_t offset, const void* source, uint64_t size);
+	void               IncreaseStreamScore(int score) noexcept { stream_score += score; }
+	[[nodiscard]] int  StreamScore() const noexcept { return stream_score; }
 	void               Flush(uint64_t offset, uint64_t size);
 	void               Invalidate(uint64_t offset, uint64_t size);
 	void CopyFrom(CommandBuffer& command, const Buffer& source, uint64_t source_offset,
@@ -68,13 +69,13 @@ public:
 	void Fill(uint64_t offset, uint64_t size, uint32_t value);
 
 	// BufferCache state lives directly on the resource.
-	bool   is_deleted = false;
-	size_t lru_id     = 0;
+	bool   is_deleted   = false;
+	int    stream_score = 0;
+	size_t lru_id       = 0;
 
 protected:
 	[[nodiscard]] GraphicContext&   Graphics() const noexcept { return *m_graphics; }
 	[[nodiscard]] CommandScheduler& Scheduler() const noexcept { return *m_scheduler; }
-	[[nodiscard]] VulkanBuffer&     NativeBuffer() noexcept { return *m_buffer; }
 
 private:
 	[[nodiscard]] vk::BufferMemoryBarrier Barrier(uint64_t offset, uint64_t size,
@@ -85,11 +86,12 @@ private:
 	CommandScheduler*             m_scheduler   = nullptr;
 	MemoryUsage                   m_usage       = MemoryUsage::DeviceLocal;
 	uint64_t                      m_cpu_address = 0;
-	uint64_t                      m_size        = 0;
 	vk::DeviceAddress             m_device_address = 0;
-	std::unique_ptr<VulkanBuffer> m_buffer;
+	vk::Buffer                    m_buffer     = nullptr;
+	VmaAllocation                 m_allocation = nullptr;
+	uint64_t                      m_size;
+	bool                          m_coherent = false;
 	std::span<uint8_t>            m_mapped;
-	bool                          m_is_coherent = false;
 };
 
 class StreamBuffer final: public Buffer {
@@ -110,7 +112,6 @@ private:
 		uint64_t upper_bound = 0;
 	};
 
-	void                      ReserveWatches(std::vector<Watch>& watches, size_t grow_size);
 	[[nodiscard]] static bool NormalizeReservation(bool coherent, uint64_t atom, uint64_t& size,
 	                                               uint64_t& alignment);
 	[[nodiscard]] bool        WaitPendingOperations(const std::vector<Watch>& watches,

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

@@ -13,6 +13,7 @@
 #include "graphics/host_gpu/renderer/image/tiler.h"
 
 #include <map>
+#include <type_traits>
 #include <unordered_map>
 #include <unordered_set>
 #include <vector>
@@ -37,12 +38,6 @@ public:
 		BindingType   type = BindingType::Texture;
 	};
 
-	struct RegionInfo {
-		bool image_pages     = false;
-		bool image_bytes     = false;
-		bool gpu_image_bytes = false;
-	};
-
 	TextureCache(GraphicContext& graphics, CommandScheduler& scheduler, PageManager& page_manager,
 	             BufferCache& buffer_cache);
 	~TextureCache();
@@ -66,15 +61,11 @@ public:
 	                                        uint32_t packed_clear);
 	void               InvalidateMemory(uint64_t address, uint64_t size);
 	void               InvalidateMemoryFromGPU(uint64_t address, uint64_t size);
-	[[nodiscard]] RegionInfo QueryRegion(uint64_t address, uint64_t size);
+	[[nodiscard]] bool IsRegionGpuModified(uint64_t address, uint64_t size);
 
 	[[nodiscard]] bool IsMeta(uint64_t address);
-	[[nodiscard]] bool IsMetaCleared(uint64_t address, uint32_t slice,
-	                                 uint32_t* fill_value = nullptr);
+	[[nodiscard]] bool IsMetaCleared(uint64_t address, uint32_t slice);
 	[[nodiscard]] bool ClearMeta(uint64_t address);
-	// Returns true when registered DCC absorbed the fill and the caller may skip the dispatch.
-	// False may still record PendingDcc state, but the guest dispatch must execute.
-	[[nodiscard]] bool TryConsumeDccFill(uint64_t address, uint64_t size, uint32_t fill_value);
 	[[nodiscard]] bool TouchMeta(uint64_t address, uint32_t slice, bool is_clear);
 
 	void UnmapMemory(uint64_t address, uint64_t size);
@@ -83,21 +74,14 @@ public:
 
 private:
 	enum class TransferDirection { Upload, Download };
-	struct ColorTransferPlan;
-	struct DownloadPlan;
+	struct TextureTransfer;
+	struct ImageDownload;
 
 	struct MetaDataInfo {
-		// A guest metadata-fill dispatch may initialize DCC before its render target is bound.
-		// PendingDcc retains that exact fill until FindRenderTarget classifies the address,
-		// without exposing an unconfirmed buffer address to the normal metadata heuristics.
-		// Keep all surface metadata in one entry so CMask/FMask can be
-		// registered beside HTile and DCC without introducing parallel tracking paths.
-		enum class Type : uint8_t { PendingDcc, CMask, FMask, HTile, Dcc };
+		enum class Type : uint8_t { CMask, FMask, HTile };
 
-		Type     type       = Type::PendingDcc;
-		uint32_t clear_mask = 0;
-		uint32_t fill_value = 0xffffffffu;
-		uint64_t fill_size  = 0;
+		Type     type;
+		uint32_t clear_mask = UINT32_MAX;
 	};
 
 	struct OverlapResult {
@@ -109,6 +93,23 @@ private:
 	using ImageIds       = InlinePageOwnerList<ImageId, 16>;
 	using ImagePageTable = MultiLevelPageTable<ImageIds, 20, 40, 10>;
 
+	// Callers have validated the nonempty 40-bit range with TryGetPageRange.
+	template <typename Func>
+	static void ForEachPage(uint64_t address, size_t size, Func&& func) {
+		using FuncReturn = typename std::invoke_result<Func, uint64_t>::type;
+		static constexpr bool RETURNS_BOOL = std::is_same_v<FuncReturn, bool>;
+		const uint64_t page_end = (address + size - 1) >> ImagePageTable::kPageBits;
+		for (uint64_t page = address >> ImagePageTable::kPageBits; page <= page_end; ++page) {
+			if constexpr (RETURNS_BOOL) {
+				if (func(page)) {
+					break;
+				}
+			} else {
+				func(page);
+			}
+		}
+	}
+
 	[[nodiscard]] ImageId     InsertImage(const ImageInfo& info);
 	[[nodiscard]] ImageId     GetNullImage(const ImageDesc& desc);
 	void                      RegisterImage(ImageId id);
@@ -137,16 +138,21 @@ private:
 	[[nodiscard]] ImageId       ResolveDepthOverlap(const ImageInfo& requested, BindingType binding,
 	                                                ImageId cached);
 	[[nodiscard]] ImageId       ExpandImage(const ImageInfo& info, ImageId source);
-	void                        RefreshImage(ImageId id, const ImageDesc& desc);
-	void                        InitializeImage(ImageId id, const ImageDesc& desc);
-	[[nodiscard]] ColorTransferPlan BuildColorTransfer(const Image& image, BindingType binding,
-	                                                   TransferDirection direction) const;
-	[[nodiscard]] DownloadPlan      BuildDownload(const Image& image) const;
-	void UploadImage(Image& image, const ImageDesc& desc, Buffer& source, uint64_t source_offset);
-	void DownloadImageData(Image& image, Buffer& destination, uint64_t destination_offset,
-	                       uint64_t destination_size, DownloadPlan plan);
+	void                        RefreshImage(ImageId id);
+	void                        MaterializeDccClear(ImageId id, const ImageDesc& desc,
+	                                                uint32_t metadata_base_layer);
+	void                        InitializeImage(ImageId id);
+	[[nodiscard]] TextureTransfer
+	BuildTextureTransfer(const Image& image, BindingType binding, TransferDirection direction) const;
+	[[nodiscard]] ImageDownload BuildDownload(const Image& image) const;
+	void UploadImage(Image& image, Buffer& source, uint64_t source_offset);
+	void DownloadImage(Image& image, Buffer& destination, uint64_t destination_offset,
+	                       uint64_t destination_size, ImageDownload transfer);
 	void DownloadDepth(Image& image, Buffer& destination, uint64_t destination_offset);
 	void CommitGpuWrite(Image& image);
+	// Caller holds m_lock. Volume layer ranges select depth slices.
+	void ClearImage(CommandBuffer& command, ImageId id, vk::Format format,
+	                const vk::ImageSubresourceRange& range, const vk::ClearValue& clear);
 	void PrepareImageCopy(Image& image);
 	void RefreshCopySource(ImageId id);
 	[[nodiscard]] bool CopyD16(Image& destination, Image& source);
@@ -156,7 +162,7 @@ private:
 	void ValidateImageDesc(const ImageDesc& desc) const;
 
 	void               InvalidateCpuAliases(uint64_t address, uint64_t size);
-	[[nodiscard]] bool TryDownloadImage(ImageId id);
+	[[nodiscard]] bool DownloadImageMemory(ImageId id);
 
 	GraphicContext&                                   m_graphics;
 	CommandScheduler&                                 m_scheduler;

src/graphics/host_gpu/renderer/colorRenderTarget.cpp

@@ -4,6 +4,7 @@
 #include "common/logging/log.h"
 #include "common/profiler.h"
 #include "graphics/guest_gpu/gpu_defs.h"
+#include "graphics/guest_gpu/gpu_format.h"
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/guest_gpu/tile.h"
 #include "graphics/host_gpu/graphicContext.h"
@@ -21,47 +22,40 @@ namespace Libs::Graphics {
 
 static std::atomic<uint32_t> g_render_color_log_count = 0;
 
-static void ResolveDccClearInfo(RenderColorInfo& info, vk::Format format, bool has_dcc,
-                                uint32_t packed_clear) {
-	// Register-backed DCC clears use the target's packed clear value. Decode one-word guest
-	// formats here; unsupported encodings remain tracked without unsafe materialization.
-	info.metadata_clear_supported =
-	    has_dcc && DecodePackedColorClear(format, packed_clear, info.color_clear_value);
-	switch (format) {
-		case vk::Format::eR8G8B8A8Unorm:
-		case vk::Format::eR8G8B8A8Srgb:
-		case vk::Format::eB8G8R8A8Unorm:
-		case vk::Format::eB8G8R8A8Srgb:
-		case vk::Format::eA2B10G10R10UnormPack32:
-		case vk::Format::eA2R10G10B10UnormPack32:
-			info.metadata_fixed_clear_supported = has_dcc;
-			break;
-		default: info.metadata_fixed_clear_supported = false; break;
+static bool DccAlphaOnMsb(const HW::ColorInfo& info) {
+	switch (info.format) {
+		case Prospero::ChannelLayout::k10_10_10_2:
+		case Prospero::ChannelLayout::k10_10_10_2Float:
+		case Prospero::ChannelLayout::k5_5_5_1: return true;
+		case Prospero::ChannelLayout::k2_10_10_10:
+		case Prospero::ChannelLayout::k1_5_5_5: return false;
+		default: break;
 	}
-	if (!info.metadata_clear_supported) {
-		info.color_clear_value = {};
+	const auto components =
+	    Prospero::ResolveRenderTargetFormat(info.format, info.channel_type).components;
+	if (components == 1) {
+		return info.channel_order != Prospero::ChannelOrder::kStandard;
 	}
+	return components == 3 || info.channel_order == Prospero::ChannelOrder::kStandard ||
+	       info.channel_order == Prospero::ChannelOrder::kAlt;
 }
 
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
-void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer& buffer,
-                                              RenderColorInfo& r,
+void RenderExecutor::ResolveRenderColorTarget(CommandBuffer& buffer, RenderColorInfo& r,
                                               uint32_t         render_target_slice_offset,
-                                              uint32_t render_target_slot, bool ignore_target_mask,
+                                              uint32_t rt_slot, bool ignore_target_mask,
                                               bool exact_format) {
 	KYTY_PROFILER_FUNCTION();
 	const auto& hw = buffer.GetRegisters();
 
-	const auto  rt_slot = (render_target_slot == UINT32_MAX ? render_target_first_bound_slot(buffer)
-	                                                        : render_target_slot);
 	const auto& rt      = hw.GetRenderTarget(rt_slot);
 	auto        mask    = render_target_mask_slot(hw.GetRenderTargetMask(), rt_slot);
 	if (ignore_target_mask && rt.base.addr != 0 && mask == 0) {
 		mask = 0x0f;
 	}
 
-	r.target_slot    = rt_slot;
-	r.export_mapping = {};
+	r             = {};
+	r.target_slot = rt_slot;
 
 	if (rt.base.addr == 0 || mask == 0) {
 		if (graphics_debug_dump_enabled()) {
@@ -76,23 +70,6 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 			}
 		}
 
-		// No color output
-		r.type                           = RenderColorType::NoColorOutput;
-		r.desc                           = {};
-		r.base_addr                      = 0;
-		r.image_id                       = {};
-		r.image_view                     = nullptr;
-		r.format                         = vk::Format::eUndefined;
-		r.extent                         = {};
-		r.base_mip_level                 = 0;
-		r.base_array_layer               = 0;
-		r.buffer_size                    = 0;
-		r.samples                        = 1;
-		r.export_mapping                 = {};
-		r.color_clear_enable             = false;
-		r.metadata_clear_supported       = false;
-		r.metadata_fixed_clear_supported = false;
-		r.color_clear_value              = {};
 		return;
 	}
 	const auto samples = render_sample_count(rt.attrib.num_fragments);
@@ -100,50 +77,11 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 		EXIT("unsupported render-target sample configuration: samples=%u fragments=%u\n",
 		     rt.attrib.num_samples, rt.attrib.num_fragments);
 	}
-	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:
-		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,
-			     render_target_slice_offset);
-	}
-	r.base_array_layer    = view.base_layer;
 	const uint32_t levels = rt.attrib2.num_mip_levels + 1u;
 	if (levels == 0 || levels > 16 || rt.view.current_mip_level >= levels) {
 		EXIT("unsupported render-target mip range: current=%u levels=%u\n",
 		     rt.view.current_mip_level, levels);
 	}
-	if (graphics_debug_dump_enabled()) {
-		static std::atomic_uint log_count = 0;
-		const auto              log_id    = log_count.fetch_add(1, std::memory_order_relaxed);
-		if (log_id < 128) {
-			LOGF("RenderColorTarget: inspect slot=%" PRIu32 " base=0x%010" PRIx64
-			     " mask=0x%01" PRIx32 " attrib2_width=%" PRIu32 " attrib2_height=%" PRIu32
-			     " attrib3_tile=0x%08" PRIx32 " attrib3_dim=0x%08" PRIx32 " fmt=0x%08" PRIx32
-			     " nfmt=0x%08" PRIx32 " order=0x%08" PRIx32 "\n",
-			     rt_slot, rt.base.addr, mask, rt.attrib2.width, rt.attrib2.height,
-			     static_cast<uint32_t>(rt.attrib3.tile_mode), rt.attrib3.dimension,
-			     static_cast<uint32_t>(rt.info.format), static_cast<uint32_t>(rt.info.channel_type),
-			     static_cast<uint32_t>(rt.info.channel_order));
-		}
-	}
-
-	// Color-control state selects the color-buffer operation and logical blend operation.
-	// The normal copy operation is a regular color write, not an attachment clear.
-	// Nonlinear clear values are still stored as normalized components.
-	// Fast color clears are metadata driven and must be handled explicitly when
-	// that metadata path is implemented; render-pass load must preserve contents.
-	r.color_clear_enable = false;
-	r.color_clear_value  = {};
-
-	uint32_t   width  = 0;
-	uint32_t   height = 0;
-	uint32_t   pitch  = 0;
-	uint64_t   size   = 0;
-	bool       tile   = false;
 	static constexpr std::array image_types {Prospero::ImageType::kColor1D,
 	                                         Prospero::ImageType::kColor2D,
 	                                         Prospero::ImageType::kColor3D};
@@ -165,7 +103,48 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 	if (volume && samples != 1) {
 		EXIT("multisampled 3D render targets are unsupported\n");
 	}
-	const uint32_t depth        = volume ? rt.attrib3.depth + 1u : 1u;
+	const uint32_t depth = volume ? rt.attrib3.depth + 1u : 1u;
+	// For volumes, CB_COLOR_VIEW bounds exported slices; ATTRIB3 defines storage depth.
+	// The host attachment contains only the selected slices that exist in this mip.
+	const uint32_t last_layer = volume
+	                                ? std::min(rt.view.last_array_slice_index,
+	                                           std::max(depth >> rt.view.current_mip_level, 1u) - 1u)
+	                                : rt.view.last_array_slice_index;
+	const auto view = ResolveTargetViewInfo(
+	    rt.view.base_array_slice_index, last_layer, render_target_slice_offset);
+	switch (view.type) {
+		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,
+			     render_target_slice_offset);
+	}
+	if (graphics_debug_dump_enabled()) {
+		static std::atomic_uint log_count = 0;
+		const auto              log_id    = log_count.fetch_add(1, std::memory_order_relaxed);
+		if (log_id < 128) {
+			LOGF("RenderColorTarget: inspect slot=%" PRIu32 " base=0x%010" PRIx64
+			     " mask=0x%01" PRIx32 " attrib2_width=%" PRIu32 " attrib2_height=%" PRIu32
+			     " attrib3_tile=0x%08" PRIx32 " attrib3_dim=0x%08" PRIx32 " fmt=0x%08" PRIx32
+			     " nfmt=0x%08" PRIx32 " order=0x%08" PRIx32 "\n",
+			     rt_slot, rt.base.addr, mask, rt.attrib2.width, rt.attrib2.height,
+			     static_cast<uint32_t>(rt.attrib3.tile_mode), rt.attrib3.dimension,
+			     static_cast<uint32_t>(rt.info.format), static_cast<uint32_t>(rt.info.channel_type),
+			     static_cast<uint32_t>(rt.info.channel_order));
+		}
+	}
+
+	// Color-control state selects the color-buffer operation and logical blend operation.
+	// The normal copy operation is a regular color write, not an attachment clear.
+	// Nonlinear clear values are still stored as normalized components.
+	// Fast color clears are metadata driven and must be handled explicitly when
+	// that metadata path is implemented; render-pass load must preserve contents.
+	uint32_t   width  = 0;
+	uint32_t   height = 0;
+	uint32_t   pitch  = 0;
+	uint64_t   size   = 0;
+	bool       tile   = false;
 	const bool     standard4    = rt.attrib3.tile_mode == Prospero::TileMode::kStandard4KB;
 	const bool     standard64   = rt.attrib3.tile_mode == Prospero::TileMode::kStandard64KB;
 	const bool     depth_tile   = rt.attrib3.tile_mode == Prospero::TileMode::kDepth;
@@ -204,13 +183,11 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 	if (texture_tile &&
 	    (!TileGetTextureBlockLayout(transfer_format, rt.attrib3.tile_mode, volume,
 	                                texture_tile_layout) ||
-	     rt.pitch.pitch_div8_minus1 != 0 ||
 	     (rt.base.addr & (texture_tile_layout.block.block_size - 1u)) != 0 ||
 	     rt.info.fmask_compression_enable || rt.info.fmask_data_compression_disable ||
 	     rt.info.fmask_one_frag_mode || rt.info.cmask_fast_clear_enable ||
-	     rt.info.dcc_compression_enable || rt.info.cmask_is_linear != 0 ||
-	     rt.info.cmask_addr_type != 0 || rt.info.alt_tile_mode || rt.cmask.addr != 0 ||
-	     rt.fmask.addr != 0 || rt.dcc_addr.addr != 0 || rt.dcc.data_write_on_dcc_clear_to_reg)) {
+	     rt.info.dcc_compression_enable || rt.cmask.addr != 0 || rt.fmask.addr != 0 ||
+	     rt.dcc_addr.addr != 0 || rt.dcc.data_write_on_dcc_clear_to_reg)) {
 		EXIT("unsupported texture-tiled render target: addr=0x%016" PRIx64 " tile=%u"
 		     " dimension=%u depth=%u levels=%u layer=%u/%u samples=%u fragments=%u bpe=%u"
 		     " cmask=0x%016" PRIx64 " fmask=0x%016" PRIx64 " dcc=0x%016" PRIx64 "\n",
@@ -226,21 +203,14 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 		     " layer=%u/%u\n",
 		     rt.attrib3.dimension, rt.attrib3.depth, view.base_layer, view.image_layers);
 	}
-	if (rt.pitch.pitch_div8_minus1 != 0) {
-		pitch = (rt.pitch.pitch_div8_minus1 + 1u) << 3u;
-	} else if (tile) {
-		if (volume) {
-			pitch = TileGetTexturePitch(transfer_format, width, rt.attrib3.tile_mode);
-		} else if (texture_tile) {
-			pitch = TileGetTexturePitch(transfer_format, width, rt.attrib3.tile_mode);
-		} else {
-			pitch = TileGetRenderTargetPitch(width, bytes_per_element, rt.attrib.num_fragments);
-		}
-		if (pitch == 0) {
-			EXIT("unsupported render-target pitch: width=%u bytes=%u\n", width, bytes_per_element);
-		}
+	// PPSA28068: the linear EULA target and its sampled view must share the padded pitch.
+	if (!tile || volume || texture_tile) {
+		pitch = TileGetTexturePitch(transfer_format, width, rt.attrib3.tile_mode);
 	} else {
-		pitch = width;
+		pitch = TileGetRenderTargetPitch(width, bytes_per_element, rt.attrib.num_fragments);
+	}
+	if (pitch == 0) {
+		EXIT("unsupported render-target pitch: width=%u bytes=%u\n", width, bytes_per_element);
 	}
 
 	TileSizeOffset    mip_sizes[16] {};
@@ -294,12 +264,6 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 		mip_sizes[0]  = {static_cast<uint32_t>(size), 0, 0, 0, 0, 0};
 		mip_padded[0] = {pitch, height};
 	}
-	if (rt.slice.slice_div64_minus1 != 0 &&
-	    (static_cast<uint64_t>(rt.slice.slice_div64_minus1) + 1u) * 64u != size) {
-		EXIT("render-target slice span mismatch: encoded=0x%016" PRIx64 " derived=0x%016" PRIx64
-		     "\n",
-		     (static_cast<uint64_t>(rt.slice.slice_div64_minus1) + 1u) * 64u, size);
-	}
 	if (size == 0 || (!volume && size > UINT64_MAX / view.image_layers)) {
 		EXIT("render-target memory footprint is invalid\n");
 	}
@@ -309,18 +273,12 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 	if (backing_size == 0) {
 		EXIT("render-target backing is empty\n");
 	}
-	if (backing_size > TRACKER_ADDRESS_SIZE - rt.base.addr) {
+	if (!GuestRange {rt.base.addr, backing_size}.Valid()) {
 		EXIT("render-target backing range is invalid\n");
 	}
 
 	const vk::Extent2D view_extent = {std::max(width >> rt.view.current_mip_level, 1u),
 	                                  std::max(height >> rt.view.current_mip_level, 1u)};
-	const uint32_t     view_depth  = std::max(depth >> rt.view.current_mip_level, 1u);
-	if (volume &&
-	    (view.base_layer >= view_depth || view.layer_count > view_depth - view.base_layer)) {
-		EXIT("3D render-target view exceeds mip depth: base=%u count=%u depth=%u mip=%u\n",
-		     view.base_layer, view.layer_count, view_depth, rt.view.current_mip_level);
-	}
 
 	auto decision_log_id = g_render_color_log_count.fetch_add(1);
 	if (decision_log_id < 128) {
@@ -347,10 +305,14 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 	desc.info.tile_mode       = rt.attrib3.tile_mode;
 	const bool has_dcc        = rt.info.dcc_compression_enable && rt.dcc_addr.addr != 0;
 	if (has_dcc) {
-		// DCC uses a separate metadata allocation. Carry its address through ImageInfo so
-		// TextureCache can associate metadata fills observed before target registration.
-		desc.info.metadata.kind          = ImageMetadataKind::Dcc;
-		desc.info.metadata.range.address = rt.dcc_addr.addr;
+		TileSizeAlign metadata_size {};
+		(void)TileGetDccSize(width, height, volume ? depth : view.image_layers, bytes_per_element,
+		                     levels, rt.attrib3.tile_mode, metadata_size, rt.attrib.num_fragments);
+		desc.info.metadata.kind                     = ImageMetadataKind::Dcc;
+		desc.info.metadata.range                    = {rt.dcc_addr.addr, metadata_size.size};
+		desc.info.metadata.dcc_clear_word           = rt.clear_word0.word0;
+		desc.info.metadata.dcc_clear_register_valid = true;
+		desc.info.metadata.dcc_alpha_msb            = DccAlphaOnMsb(rt.info);
 	}
 	for (uint32_t level = 0; level < levels; level++) {
 		if (volume) {
@@ -392,18 +354,10 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer&
 	desc.view_info.usage       = vk::ImageUsageFlagBits::eColorAttachment;
 	auto& texture_cache        = m_context.GetTextureCache();
 	r.desc                     = std::move(desc);
+	r.guest_mip_level          = rt.view.current_mip_level;
+	r.guest_array_layer        = view.base_layer;
 	r.image_id                 = texture_cache.FindImage(r.desc, exact_format);
-	r.type                     = RenderColorType::RenderTexture;
-	r.base_addr                = rt.base.addr;
-	r.image_view               = nullptr;
-	r.format                   = r.desc.view_info.format;
-	r.extent                   = view_extent;
-	r.base_mip_level           = rt.view.current_mip_level;
-	r.buffer_size              = backing_size;
-	r.samples                  = samples;
 	r.export_mapping           = target_format.export_mapping;
-	r.color_clear_enable       = false;
-	ResolveDccClearInfo(r, target_format.format, has_dcc, rt.clear_word0.word0);
 	BindRenderTarget(r.image_id);
 }
 

src/graphics/host_gpu/renderer/colorRenderTarget.h

@@ -6,33 +6,24 @@
 #include "graphics/host_gpu/renderer/renderTarget.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 
+#include <algorithm>
 #include <cstdint>
 
 namespace Libs::Graphics {
 
-enum class RenderColorType {
-	NoColorOutput,
-	RenderTexture,
-};
-
 struct RenderColorInfo {
-	RenderColorType                 type = RenderColorType::NoColorOutput;
+	// Discovery keeps guest image information but can remap the view into a larger cache image.
 	TextureCache::ImageDesc         desc;
 	ImageId                         image_id;
-	vk::ImageView                   image_view       = nullptr;
-	vk::Format                      format           = vk::Format::eUndefined;
-	vk::Extent2D                    extent           = {};
-	uint32_t                        base_mip_level   = 0;
-	uint32_t                        base_array_layer = 0;
-	uint64_t                        base_addr        = 0;
-	uint64_t                        buffer_size      = 0;
 	uint32_t                        target_slot      = 0;
-	uint32_t                        samples          = 1;
+	uint32_t                        guest_mip_level   = 0;
+	uint32_t                        guest_array_layer = 0;
 	Prospero::ColorComponentMapping export_mapping;
-	bool                            color_clear_enable             = false;
-	bool                            metadata_clear_supported       = false;
-	bool                            metadata_fixed_clear_supported = false;
-	vk::ClearColorValue             color_clear_value {};
+
+	[[nodiscard]] vk::Extent2D Extent() const {
+		return {std::max(desc.info.extent.width >> guest_mip_level, 1u),
+		        std::max(desc.info.extent.height >> guest_mip_level, 1u)};
+	}
 };
 
 } // namespace Libs::Graphics

src/graphics/host_gpu/renderer/commandScheduler.cpp

@@ -19,11 +19,11 @@ void ReportVulkanFatal(const char* what, vk::Result result, uint64_t tick, uint3
                        uint32_t arg3, uint64_t arg4) {
 	LOGF("%s failed: %s (%d), tick=%" PRIu64 " debug_op=%u debug_submit=%" PRIu64
 	     " args=%u,%u,%u,%u,0x%016" PRIx64 "\n",
-	     what, VulkanToString(result).c_str(), static_cast<int>(result), tick, debug_op,
+	     what, vk::to_string(result).c_str(), static_cast<int>(result), tick, debug_op,
 	     debug_submit, arg0, arg1, arg2, arg3, arg4);
 	std::printf("%s failed: %s (%d), tick=%" PRIu64 " debug_op=%u debug_submit=%" PRIu64
 	            " args=%u,%u,%u,%u,0x%016" PRIx64 "\n",
-	            what, VulkanToString(result).c_str(), static_cast<int>(result), tick, debug_op,
+	            what, vk::to_string(result).c_str(), static_cast<int>(result), tick, debug_op,
 	            debug_submit, arg0, arg1, arg2, arg3, arg4);
 	std::fflush(stdout);
 }
@@ -34,7 +34,6 @@ CommandScheduler::CommandPool::CommandPool(GraphicContext& graphics, MasterSemap
     : m_graphics(graphics), m_master(master) {
 	EXIT_IF(graphics.queue_family == static_cast<uint32_t>(-1));
 	vk::CommandPoolCreateInfo create {};
-	create.sType            = vk::StructureType::eCommandPoolCreateInfo;
 	create.queueFamilyIndex = graphics.queue_family;
 	create.flags            = vk::CommandPoolCreateFlagBits::eTransient |
 	                          vk::CommandPoolCreateFlagBits::eResetCommandBuffer;
@@ -52,7 +51,6 @@ size_t CommandScheduler::CommandPool::Grow() {
 	m_buffers.resize(first + GrowStep);
 
 	vk::CommandBufferAllocateInfo allocate {};
-	allocate.sType              = vk::StructureType::eCommandBufferAllocateInfo;
 	allocate.commandPool        = m_pool;
 	allocate.level              = vk::CommandBufferLevel::ePrimary;
 	allocate.commandBufferCount = static_cast<uint32_t>(GrowStep);
@@ -87,10 +85,7 @@ vk::CommandBuffer CommandScheduler::CommandPool::Commit() {
 		m_ticks[*found] = m_master.CurrentTick();
 	}
 
-	m_hint = *found + 1;
-	if (m_hint == m_ticks.size()) {
-		m_hint = 0;
-	}
+	m_hint = (*found + 1) % m_ticks.size();
 	return m_buffers[*found];
 }
 
@@ -153,14 +148,10 @@ void CommandScheduler::Begin(HW::Context& registers, HW::UserConfig& user_config
 		std::lock_guard lock(m_operation_mutex);
 		EXIT_IF(m_operation_state != OperationState::Open);
 	}
-	m_registers   = &registers;
-	m_user_config = &user_config;
-	m_shaders     = &shaders;
+	m_command.Bind(registers, user_config, shaders);
 
 	if (m_command.IsInvalid()) {
 		BeginNext();
-	} else {
-		BindCurrent();
 	}
 }
 
@@ -217,13 +208,7 @@ void CommandScheduler::Wait(uint64_t tick) {
 }
 
 void CommandScheduler::PopPendingOperations() {
-	PopPendingOperations(true);
-}
-
-void CommandScheduler::PopPendingOperations(bool refresh_gpu_tick) {
-	if (refresh_gpu_tick) {
-		m_master.Refresh();
-	}
+	m_master.Refresh();
 	for (;;) {
 		PendingOperation operation;
 		{
@@ -351,11 +336,6 @@ CommandBuffer& CommandScheduler::Current() {
 	return m_command;
 }
 
-void CommandScheduler::BindCurrent() {
-	EXIT_IF(m_registers == nullptr || m_user_config == nullptr || m_shaders == nullptr);
-	m_command.Bind(*m_registers, *m_user_config, *m_shaders);
-}
-
 CommandBuffer& CommandScheduler::BeginCommand() {
 	EXIT_IF(!m_command.IsInvalid());
 	m_command.m_buffer = m_command_pool.Commit();
@@ -381,14 +361,12 @@ uint64_t CommandScheduler::Submit(SubmitInfo submit) {
 		submit.AddSignal(m_master.Handle(), tick);
 
 		vk::TimelineSemaphoreSubmitInfo timeline_info {};
-		timeline_info.sType                     = vk::StructureType::eTimelineSemaphoreSubmitInfo;
 		timeline_info.waitSemaphoreValueCount   = submit.num_wait_semaphores;
 		timeline_info.pWaitSemaphoreValues      = submit.wait_ticks.data();
 		timeline_info.signalSemaphoreValueCount = submit.num_signal_semaphores;
 		timeline_info.pSignalSemaphoreValues    = submit.signal_ticks.data();
 
 		vk::SubmitInfo submit_info {};
-		submit_info.sType                = vk::StructureType::eSubmitInfo;
 		submit_info.pNext                = &timeline_info;
 		submit_info.waitSemaphoreCount   = submit.num_wait_semaphores;
 		submit_info.pWaitSemaphores      = submit.wait_semaphores.data();
@@ -414,8 +392,7 @@ uint64_t CommandScheduler::Submit(SubmitInfo submit) {
 }
 
 void CommandScheduler::BeginNext() {
-	EXIT_IF(!m_command.IsInvalid());
-	BindCurrent();
+	CheckActive();
 	BeginCommand();
 }
 

src/graphics/host_gpu/renderer/commandScheduler.h

@@ -42,7 +42,7 @@ public:
 	void                      DeferPriorityOperation(Common::UniqueFunction<void>&& operation);
 	[[nodiscard]] static bool InDeferredOperation() noexcept;
 
-	[[nodiscard]] bool             Active() const noexcept { return m_registers != nullptr; }
+	[[nodiscard]] bool Active() const noexcept { return m_command.m_registers != nullptr; }
 	void                           CheckActive() const;
 	CommandBuffer&                 Current();
 	[[nodiscard]] uint64_t         CurrentTick() const noexcept { return m_master.CurrentTick(); }
@@ -80,9 +80,7 @@ private:
 		uint64_t                     tick = 0;
 	};
 
-	void BindCurrent();
 	void BeginNext();
-	void PopPendingOperations(bool refresh_gpu_tick);
 	void PriorityOperationsThread(std::stop_token stop);
 	void RunOperation(Common::UniqueFunction<void>&& operation);
 
@@ -99,9 +97,6 @@ private:
 	bool                         m_priority_active      = false;
 	uint64_t                     m_priority_active_tick = 0;
 	OperationState               m_operation_state      = OperationState::Open;
-	HW::Context*                 m_registers            = nullptr;
-	HW::UserConfig*              m_user_config          = nullptr;
-	HW::Shader*                  m_shaders              = nullptr;
 };
 
 } // namespace Libs::Graphics

src/graphics/host_gpu/renderer/context.cpp

@@ -9,7 +9,6 @@
 #include "graphics/host_gpu/renderer/image/imageView.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
-#include "graphics/host_gpu/vma.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 
 #include <algorithm>
@@ -34,10 +33,7 @@ void CommandBuffer::Begin() {
 	auto buffer = Handle();
 
 	vk::CommandBufferBeginInfo begin_info {};
-	begin_info.sType            = vk::StructureType::eCommandBufferBeginInfo;
-	begin_info.pNext            = nullptr;
-	begin_info.flags            = vk::CommandBufferUsageFlagBits::eOneTimeSubmit;
-	begin_info.pInheritanceInfo = nullptr;
+	begin_info.flags = vk::CommandBufferUsageFlagBits::eOneTimeSubmit;
 
 	auto result = buffer.begin(&begin_info);
 
@@ -65,17 +61,16 @@ void CommandBuffer::SetDebugInfo(uint32_t op, uint64_t submit_id, uint32_t arg0,
 }
 
 void CommandBuffer::BeginRendering(const RenderState& state) const {
-	EXIT_IF(state.width == 0 || state.height == 0 || state.num_layers == 0 ||
-	        state.num_color_attachments > RENDER_COLOR_ATTACHMENTS_MAX);
 	if (m_rendering && m_render_state == state) {
 		return;
 	}
+	EXIT_IF(state.width == 0 || state.height == 0 || state.num_layers == 0 ||
+	        state.num_color_attachments > RENDER_COLOR_ATTACHMENTS_MAX);
 	EndRendering();
 
 	std::array<vk::RenderingAttachmentInfo, RENDER_COLOR_ATTACHMENTS_MAX> colors {};
 	for (uint32_t i = 0; i < state.num_color_attachments; i++) {
 		const auto& attachment = state.color_attachments[i];
-		colors[i].sType        = vk::StructureType::eRenderingAttachmentInfo;
 		colors[i].imageView    = attachment.image_view;
 		colors[i].imageLayout  = attachment.image_layout;
 		colors[i].loadOp =
@@ -86,7 +81,6 @@ void CommandBuffer::BeginRendering(const RenderState& state) const {
 
 	const auto&                 depth_stencil = state.depth_stencil_attachment;
 	vk::RenderingAttachmentInfo depth {};
-	depth.sType       = vk::StructureType::eRenderingAttachmentInfo;
 	depth.imageView   = depth_stencil.image_view;
 	depth.imageLayout = depth_stencil.image_layout;
 	depth.loadOp =
@@ -95,7 +89,6 @@ void CommandBuffer::BeginRendering(const RenderState& state) const {
 	depth.clearValue.depthStencil.depth = std::bit_cast<float>(depth_stencil.clear_value[0]);
 
 	vk::RenderingAttachmentInfo stencil {};
-	stencil.sType       = vk::StructureType::eRenderingAttachmentInfo;
 	stencil.imageView   = depth_stencil.image_view;
 	stencil.imageLayout = depth_stencil.image_layout;
 	stencil.loadOp =
@@ -104,7 +97,6 @@ void CommandBuffer::BeginRendering(const RenderState& state) const {
 	stencil.clearValue.depthStencil.stencil = depth_stencil.clear_value[1];
 
 	vk::RenderingInfo rendering {};
-	rendering.sType                = vk::StructureType::eRenderingInfo;
 	rendering.renderArea.extent    = {state.width, state.height};
 	rendering.layerCount           = state.num_layers;
 	rendering.colorAttachmentCount = state.num_color_attachments;

src/graphics/host_gpu/renderer/debug.cpp

@@ -4,7 +4,6 @@
 #include "common/common.h"
 #include "common/emulatorConfig.h"
 #include "common/logging/log.h"
-#include "common/stringUtils.h"
 #include "graphics/guest_gpu/gpu_defs.h"
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/host_gpu/renderer/render.h"
@@ -17,8 +16,6 @@
 
 namespace Libs::Graphics {
 
-static std::atomic<uint32_t> g_scissor_default_log_count = 0;
-
 uint32_t render_target_mask_slot(uint32_t mask, uint32_t slot) {
 	return (mask >> (slot * 4u)) & 0x0fu;
 }
@@ -59,7 +56,7 @@ uint32_t render_target_first_bound_slot(const CommandBuffer& buffer) {
 
 bool graphics_debug_dump_enabled() {
 	return Config::GraphicsDebugDumpEnabled() &&
-	       Config::GetPrintfDirection() != Config::OutputDirection::Silent;
+	       Config::GetPrintfDirection() != Config::LogDirection::Silent;
 }
 
 void uc_print(const char* func, const HW::UserConfig& uc) {
@@ -91,117 +88,75 @@ void uc_check(const HW::UserConfig& uc) {
 	EXIT_NOT_IMPLEMENTED(user_en.vgpr3 != false);
 }
 
-void sh_print(const char* func, const HW::Shader& /*uc*/) {
-	LOGF("%s\n", func);
-}
+std::string rt_print(const char* func, const HW::RenderTarget& rt) {
+	std::string dst;
+	dst.reserve(4096);
 
-std::vector<std::string> rt_print(const char* func, const HW::RenderTarget& rt) {
-	std::vector<std::string> dst;
-	dst.reserve(53);
+	dst += fmt::format("{}\n", func);
 
-	dst.push_back(fmt::format("{}\n", func));
+	dst += fmt::format("\t base.addr                       = 0x{:016x}\n", rt.base.addr);
+	dst += fmt::format("\t view.base_array_slice_index     = 0x{:08x}\n",
+	                  rt.view.base_array_slice_index);
+	dst += fmt::format("\t view.last_array_slice_index     = 0x{:08x}\n",
+	                  rt.view.last_array_slice_index);
+	dst += fmt::format("\t view.current_mip_level          = 0x{:08x}\n", rt.view.current_mip_level);
+	dst += fmt::format("\t info.fmask_compression_enable   = {}\n",
+	                  rt.info.fmask_compression_enable ? "true" : "false");
 
-	dst.push_back(fmt::format("\t base.addr                       = 0x{:016x}\n", rt.base.addr));
-	dst.push_back(
-	    fmt::format("\t pitch.pitch_div8_minus1         = 0x{:08x}\n", rt.pitch.pitch_div8_minus1));
-	dst.push_back(fmt::format("\t pitch.fmask_pitch_div8_minus1   = 0x{:08x}\n",
-	                          rt.pitch.fmask_pitch_div8_minus1));
-	dst.push_back(fmt::format("\t slice.slice_div64_minus1        = 0x{:08x}\n",
-	                          rt.slice.slice_div64_minus1));
-	dst.push_back(fmt::format("\t view.base_array_slice_index     = 0x{:08x}\n",
-	                          rt.view.base_array_slice_index));
-	dst.push_back(fmt::format("\t view.last_array_slice_index     = 0x{:08x}\n",
-	                          rt.view.last_array_slice_index));
-	dst.push_back(
-	    fmt::format("\t view.current_mip_level          = 0x{:08x}\n", rt.view.current_mip_level));
-	dst.push_back(fmt::format("\t info.fmask_compression_enable   = {}\n",
-	                          rt.info.fmask_compression_enable ? "true" : "false"));
+	dst += fmt::format("\t info.fmask_data_compression_disable = {}\n",
+	                  rt.info.fmask_data_compression_disable ? "true" : "false");
+	dst += fmt::format("\t info.fmask_one_frag_mode        = {}\n",
+	                  rt.info.fmask_one_frag_mode ? "true" : "false");
 
-	// dst.push_back(fmt::format("\t info.fmask_compression_mode     = 0x{:08x}\n", //
-	// rt.info.fmask_compression_mode));
-	dst.push_back(fmt::format("\t info.fmask_data_compression_disable = {}\n",
-	                          rt.info.fmask_data_compression_disable ? "true" : "false"));
-	dst.push_back(fmt::format("\t info.fmask_one_frag_mode        = {}\n",
-	                          rt.info.fmask_one_frag_mode ? "true" : "false"));
-
-	dst.push_back(fmt::format("\t info.cmask_fast_clear_enable    = {}\n",
-	                          rt.info.cmask_fast_clear_enable ? "true" : "false"));
-	dst.push_back(fmt::format("\t info.dcc_compression_enable     = {}\n",
-	                          rt.info.dcc_compression_enable ? "true" : "false"));
-	dst.push_back(
-	    fmt::format("\t info.cmask_is_linear            = 0x{:08x}\n", rt.info.cmask_is_linear));
-	dst.push_back(
-	    fmt::format("\t info.cmask_addr_type            = 0x{:08x}\n", rt.info.cmask_addr_type));
-	dst.push_back(fmt::format("\t info.alt_tile_mode              = {}\n",
-	                          rt.info.alt_tile_mode ? "true" : "false"));
-	dst.push_back(fmt::format("\t info.format                     = 0x{:08x}\n",
-	                          static_cast<uint32_t>(rt.info.format)));
-	dst.push_back(fmt::format("\t info.channel_type               = 0x{:08x}\n",
-	                          static_cast<uint32_t>(rt.info.channel_type)));
-	dst.push_back(fmt::format("\t info.channel_order              = 0x{:08x}\n",
-	                          static_cast<uint32_t>(rt.info.channel_order)));
-	dst.push_back(fmt::format("\t info.blend_bypa                 = {}\n",
-	                          rt.info.blend_bypass ? "true" : "false"));
-	dst.push_back(fmt::format("\t info.blend_clamp                = {}\n",
-	                          rt.info.blend_clamp ? "true" : "false"));
-	dst.push_back(fmt::format("\t info.round_mode                 = {}\n",
-	                          rt.info.round_mode ? "true" : "false"));
-	dst.push_back(fmt::format("\t attrib.force_dest_alpha_to_one  = {}\n",
-	                          rt.attrib.force_dest_alpha_to_one ? "true" : "false"));
-	dst.push_back(fmt::format("\t attrib.tile_mode                = 0x{:08x}\n",
-	                          static_cast<uint32_t>(rt.attrib.tile_mode)));
-	dst.push_back(fmt::format("\t attrib.fmask_tile_mode          = 0x{:08x}\n",
-	                          static_cast<uint32_t>(rt.attrib.fmask_tile_mode)));
-	dst.push_back(
-	    fmt::format("\t attrib.num_samples              = 0x{:08x}\n", rt.attrib.num_samples));
-	dst.push_back(
-	    fmt::format("\t attrib.num_fragments            = 0x{:08x}\n", rt.attrib.num_fragments));
-	dst.push_back(fmt::format("\t attrib2.width                   = 0x{:08x}\n", rt.attrib2.width));
-	dst.push_back(
-	    fmt::format("\t attrib2.height                  = 0x{:08x}\n", rt.attrib2.height));
-	dst.push_back(
-	    fmt::format("\t attrib2.num_mip_levels          = 0x{:08x}\n", rt.attrib2.num_mip_levels));
-	dst.push_back(fmt::format("\t attrib3.depth                   = 0x{:08x}\n", rt.attrib3.depth));
-	dst.push_back(fmt::format("\t attrib3.tile_mode               = 0x{:08x}\n",
-	                          static_cast<uint32_t>(rt.attrib3.tile_mode)));
-	dst.push_back(
-	    fmt::format("\t attrib3.dimension               = 0x{:08x}\n", rt.attrib3.dimension));
-	dst.push_back(fmt::format("\t attrib3.cmask_pipe_aligned      = {}\n",
-	                          rt.attrib3.cmask_pipe_aligned ? "true" : "false"));
-	dst.push_back(fmt::format("\t attrib3.dcc_pipe_aligned        = {}\n",
-	                          rt.attrib3.dcc_pipe_aligned ? "true" : "false"));
-	dst.push_back(fmt::format("\t dcc.max_uncompressed_block_size = 0x{:08x}\n",
-	                          rt.dcc.max_uncompressed_block_size));
-	dst.push_back(fmt::format("\t dcc.max_compressed_block_size   = 0x{:08x}\n",
-	                          rt.dcc.max_compressed_block_size));
-	dst.push_back(fmt::format("\t dcc.min_compressed_block_size   = 0x{:08x}\n",
-	                          rt.dcc.min_compressed_block_size));
-	dst.push_back(
-	    fmt::format("\t dcc.color_transform             = 0x{:08x}\n", rt.dcc.color_transform));
-	dst.push_back(fmt::format("\t dcc.overwrite_combiner_disable  = {}\n",
-	                          rt.dcc.overwrite_combiner_disable ? "true" : "false"));
-	dst.push_back(fmt::format("\t dcc.independent_64b_blocks      = {}\n",
-	                          rt.dcc.independent_64b_blocks ? "true" : "false"));
-	dst.push_back(fmt::format("\t dcc.independent_128b_blocks     = {}\n",
-	                          rt.dcc.independent_128b_blocks ? "true" : "false"));
-	dst.push_back(fmt::format("\t data_write_on_dcc_clear_to_reg  = {}\n",
-	                          rt.dcc.data_write_on_dcc_clear_to_reg ? "true" : "false"));
-	dst.push_back(fmt::format("\t dcc.dcc_clear_key_enable        = {}\n",
-	                          rt.dcc.dcc_clear_key_enable ? "true" : "false"));
-	dst.push_back(fmt::format("\t cmask.addr                      = 0x{:016x}\n", rt.cmask.addr));
-	dst.push_back(fmt::format("\t cmask_slice.slice_minus1        = 0x{:08x}\n",
-	                          rt.cmask_slice.slice_minus1));
-	dst.push_back(fmt::format("\t fmask.addr                      = 0x{:016x}\n", rt.fmask.addr));
-	dst.push_back(fmt::format("\t fmask_slice.slice_minus1        = 0x{:08x}\n",
-	                          rt.fmask_slice.slice_minus1));
-	dst.push_back(
-	    fmt::format("\t clear_word0.word0               = 0x{:08x}\n", rt.clear_word0.word0));
-	dst.push_back(
-	    fmt::format("\t clear_word1.word1               = 0x{:08x}\n", rt.clear_word1.word1));
-	dst.push_back(
-	    fmt::format("\t dcc_addr.addr                   = 0x{:016x}\n", rt.dcc_addr.addr));
-	dst.push_back(fmt::format("\t size.width                      = 0x{:08x}\n", rt.size.width));
-	dst.push_back(fmt::format("\t size.height                     = 0x{:08x}\n", rt.size.height));
+	dst += fmt::format("\t info.cmask_fast_clear_enable    = {}\n",
+	                  rt.info.cmask_fast_clear_enable ? "true" : "false");
+	dst += fmt::format("\t info.dcc_compression_enable     = {}\n",
+	                  rt.info.dcc_compression_enable ? "true" : "false");
+	dst += fmt::format("\t info.format                     = 0x{:08x}\n",
+	                  static_cast<uint32_t>(rt.info.format));
+	dst += fmt::format("\t info.channel_type               = 0x{:08x}\n",
+	                  static_cast<uint32_t>(rt.info.channel_type));
+	dst += fmt::format("\t info.channel_order              = 0x{:08x}\n",
+	                  static_cast<uint32_t>(rt.info.channel_order));
+	dst += fmt::format("\t info.blend_bypa                 = {}\n",
+	                  rt.info.blend_bypass ? "true" : "false");
+	dst += fmt::format("\t info.blend_clamp                = {}\n",
+	                  rt.info.blend_clamp ? "true" : "false");
+	dst += fmt::format("\t info.round_mode                 = {}\n",
+	                  rt.info.round_mode ? "true" : "false");
+	dst += fmt::format("\t attrib.force_dest_alpha_to_one  = {}\n",
+	                  rt.attrib.force_dest_alpha_to_one ? "true" : "false");
+	dst += fmt::format("\t attrib.num_samples              = 0x{:08x}\n", rt.attrib.num_samples);
+	dst += fmt::format("\t attrib.num_fragments            = 0x{:08x}\n", rt.attrib.num_fragments);
+	dst += fmt::format("\t attrib2.width                   = 0x{:08x}\n", rt.attrib2.width);
+	dst += fmt::format("\t attrib2.height                  = 0x{:08x}\n", rt.attrib2.height);
+	dst += fmt::format("\t attrib2.num_mip_levels          = 0x{:08x}\n", rt.attrib2.num_mip_levels);
+	dst += fmt::format("\t attrib3.depth                   = 0x{:08x}\n", rt.attrib3.depth);
+	dst += fmt::format("\t attrib3.tile_mode               = 0x{:08x}\n",
+	                  static_cast<uint32_t>(rt.attrib3.tile_mode));
+	dst += fmt::format("\t attrib3.dimension               = 0x{:08x}\n", rt.attrib3.dimension);
+	dst += fmt::format("\t attrib3.metadata_pipe_aligned   = {}\n",
+	                  rt.attrib3.metadata_pipe_aligned ? "true" : "false");
+	dst += fmt::format("\t attrib3.write_vrs_rate_hint_to_cmask = {}\n",
+	                  rt.attrib3.write_vrs_rate_hint_to_cmask ? "true" : "false");
+	dst += fmt::format("\t dcc.max_uncompressed_block_size = 0x{:08x}\n",
+	                  rt.dcc.max_uncompressed_block_size);
+	dst += fmt::format("\t dcc.max_compressed_block_size   = 0x{:08x}\n",
+	                  rt.dcc.max_compressed_block_size);
+	dst += fmt::format("\t dcc.color_transform             = 0x{:08x}\n", rt.dcc.color_transform);
+	dst += fmt::format("\t dcc.overwrite_combiner_disable  = {}\n",
+	                  rt.dcc.overwrite_combiner_disable ? "true" : "false");
+	dst += fmt::format("\t dcc.independent_block_size      = 0x{:02x}\n",
+	                  static_cast<uint8_t>(rt.dcc.independent_block_size));
+	dst += fmt::format("\t data_write_on_dcc_clear_to_reg  = {}\n",
+	                  rt.dcc.data_write_on_dcc_clear_to_reg ? "true" : "false");
+	dst += fmt::format("\t dcc.dcc_clear_key_enable        = {}\n",
+	                  rt.dcc.dcc_clear_key_enable ? "true" : "false");
+	dst += fmt::format("\t cmask.addr                      = 0x{:016x}\n", rt.cmask.addr);
+	dst += fmt::format("\t fmask.addr                      = 0x{:016x}\n", rt.fmask.addr);
+	dst += fmt::format("\t clear_word0.word0               = 0x{:08x}\n", rt.clear_word0.word0);
+	dst += fmt::format("\t clear_word1.word1               = 0x{:08x}\n", rt.clear_word1.word1);
+	dst += fmt::format("\t dcc_addr.addr                   = 0x{:016x}\n", rt.dcc_addr.addr);
 
 	return dst;
 }
@@ -231,13 +186,8 @@ static bool RenderIsColorDimension(uint32_t dimension) {
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
 static void RtCheck(const HW::RenderTarget& rt) {
 	if (rt.base.addr != 0) {
-		// bool render_to_texture = (rt.attrib.tile_mode == 0x0d);
 		//  EXIT_NOT_IMPLEMENTED(rt.base_addr == 0);
 
-		EXIT_NOT_IMPLEMENTED(rt.pitch.pitch_div8_minus1 != 0);
-		EXIT_NOT_IMPLEMENTED(rt.pitch.fmask_pitch_div8_minus1 != 0);
-		EXIT_NOT_IMPLEMENTED(rt.slice.slice_div64_minus1 != 0);
-
 		EXIT_NOT_IMPLEMENTED(rt.view.base_array_slice_index > rt.view.last_array_slice_index);
 		if (rt.view.base_array_slice_index != 0x00000000 ||
 		    rt.view.last_array_slice_index != 0x00000000) {
@@ -257,11 +207,6 @@ static void RtCheck(const HW::RenderTarget& rt) {
 				logged = true;
 			}
 		}
-		const bool native_2x_one_frag =
-		    rt.info.fmask_compression_enable && !rt.info.fmask_data_compression_disable &&
-		    rt.info.cmask_fast_clear_enable && rt.fmask.addr != 0 && rt.cmask.addr != 0 &&
-		    rt.attrib.num_samples == 1 && rt.attrib.num_fragments == 1;
-		EXIT_NOT_IMPLEMENTED(rt.info.fmask_one_frag_mode && !native_2x_one_frag);
 		if (rt.info.fmask_compression_enable) {
 			EXIT_NOT_IMPLEMENTED(rt.attrib.num_samples == 0 && rt.attrib.num_fragments == 0);
 			static bool logged = false;
@@ -286,22 +231,6 @@ static void RtCheck(const HW::RenderTarget& rt) {
 				logged = true;
 			}
 		}
-		if (rt.info.cmask_is_linear != 0x00000000 || rt.info.cmask_addr_type != 0x00000000) {
-			static bool logged = false;
-			if (!logged) {
-				LOGF("RenderTarget: temporary: ignoring PS5 CMASK metadata fields "
-				     "is_linear=%u addr_type=%u\n",
-				     rt.info.cmask_is_linear, rt.info.cmask_addr_type);
-				logged = true;
-			}
-		}
-		if (rt.info.alt_tile_mode) {
-			static bool logged = false;
-			if (!logged) {
-				LOGF("RenderTarget: temporary: ignoring alternate tile mode flag\n");
-				logged = true;
-			}
-		}
 		if (rt.info.blend_bypass) {
 			static bool logged = false;
 			if (!logged) {
@@ -328,8 +257,6 @@ static void RtCheck(const HW::RenderTarget& rt) {
 				logged = true;
 			}
 		}
-		// EXIT_NOT_IMPLEMENTED(rt.tile_mode != 0x0000000a);
-		// EXIT_NOT_IMPLEMENTED(rt.fmask_tile_mode != 0x0000000a);
 		if (rt.attrib.num_samples != 0x00000000 || rt.attrib.num_fragments != 0x00000000) {
 			static bool logged = false;
 			if (!logged) {
@@ -364,34 +291,22 @@ static void RtCheck(const HW::RenderTarget& rt) {
 		if (!RenderIsColorDimension(rt.attrib3.dimension)) {
 			EXIT("unknown PS5 render-target dimension: 0x%08" PRIx32 "\n", rt.attrib3.dimension);
 		}
-		if (!rt.attrib3.cmask_pipe_aligned) {
+		if (!rt.attrib3.metadata_pipe_aligned) {
 			static bool logged = false;
 			if (!logged) {
-				LOGF("RenderTarget: temporary: accepting unaligned PS5 CMASK pipe flag\n");
-				logged = true;
-			}
-		}
-		if (!rt.attrib3.dcc_pipe_aligned) {
-			static bool logged = false;
-			if (!logged) {
-				LOGF("RenderTarget: temporary: accepting unaligned PS5 DCC pipe flag\n");
+				LOGF("RenderTarget: temporary: accepting unaligned PS5 metadata pipe flag\n");
 				logged = true;
 			}
 		}
+		EXIT_NOT_IMPLEMENTED(rt.attrib3.write_vrs_rate_hint_to_cmask);
 
 		// EXIT_NOT_IMPLEMENTED(rt.dcc_max_uncompressed_block_size != 0x00000002);
 		// EXIT_NOT_IMPLEMENTED(rt.dcc.max_compressed_block_size != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(rt.dcc.min_compressed_block_size != 0x00000000);
 		// EXIT_NOT_IMPLEMENTED(rt.dcc.color_transform != 0x00000000);
 		EXIT_NOT_IMPLEMENTED(rt.dcc.overwrite_combiner_disable != false);
-		// EXIT_NOT_IMPLEMENTED(rt.dcc.force_independent_blocks != false);
-		// EXIT_NOT_IMPLEMENTED(rt.dcc.independent_128b_blocks != false);
 		// EXIT_NOT_IMPLEMENTED(rt.dcc.data_write_on_dcc_clear_to_reg != false);
 		EXIT_NOT_IMPLEMENTED(rt.dcc.dcc_clear_key_enable != false);
-		if (rt.cmask.addr != 0x0000000000000000 || rt.cmask_slice.slice_minus1 != 0x00000000 ||
-		    rt.fmask.addr != 0x0000000000000000 ||
-		    (rt.fmask_slice.slice_minus1 != 0x00000000 &&
-		     rt.fmask_slice.slice_minus1 != rt.slice.slice_div64_minus1) ||
+		if (rt.cmask.addr != 0x0000000000000000 || rt.fmask.addr != 0x0000000000000000 ||
 		    rt.dcc_addr.addr != 0x0000000000000000) {
 			static bool logged = false;
 			if (!logged) {
@@ -401,9 +316,6 @@ static void RtCheck(const HW::RenderTarget& rt) {
 				logged = true;
 			}
 		}
-
-		EXIT_NOT_IMPLEMENTED(rt.size.width != 0);
-		EXIT_NOT_IMPLEMENTED(rt.size.height != 0);
 	}
 }
 
@@ -423,37 +335,20 @@ static void ZPrint(const char* func, const HW::DepthRenderTarget& z) {
 	     "\t stencil_info.htile_stencil_disabled   = %s\n"
 	     "\t stencil_info.expclear_enabled         = %s\n"
 	     "\t stencil_info.partially_resident       = %s\n"
-	     "\t depth_info.addr5_swizzle_mask         = 0x%08" PRIx32 "\n"
-	     "\t depth_info.array_mode                 = 0x%08" PRIx32 "\n"
-	     "\t depth_info.pipe_config                = 0x%08" PRIx32 "\n"
-	     "\t depth_info.bank_width                 = 0x%08" PRIx32 "\n"
-	     "\t depth_info.bank_height                = 0x%08" PRIx32 "\n"
-	     "\t depth_info.macro_tile_aspect          = 0x%08" PRIx32 "\n"
-	     "\t depth_info.num_banks                  = 0x%08" PRIx32 "\n"
 	     "\t depth_view.slice_start                = 0x%08" PRIx32 "\n"
 	     "\t depth_view.slice_max                  = 0x%08" PRIx32 "\n"
 	     "\t depth_view.current_mip_level          = 0x%02" PRIx8 "\n"
 	     "\t depth_view.depth_write_disable        = %s\n"
 	     "\t depth_view.stencil_write_disable      = %s\n"
-	     "\t htile_surface.linear                  = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.full_cache              = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.htile_uses_preload_win  = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.preload                 = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.prefetch_width          = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.prefetch_height         = 0x%08" PRIx32 "\n"
-	     "\t htile_surface.dst_outside_zero_to_one = 0x%08" PRIx32 "\n"
 	     "\t z_read_base_addr                      = 0x%016" PRIx64 "\n"
 	     "\t stencil_read_base_addr                = 0x%016" PRIx64 "\n"
 	     "\t z_write_base_addr                     = 0x%016" PRIx64 "\n"
 	     "\t stencil_write_base_addr               = 0x%016" PRIx64 "\n"
-	     "\t pitch_div8_minus1                     = 0x%08" PRIx32 "\n"
-	     "\t height_div8_minus1                    = 0x%08" PRIx32 "\n"
-	     "\t slice_div64_minus1                    = 0x%08" PRIx32 "\n"
 	     "\t htile_data_base_addr                  = 0x%016" PRIx64 "\n"
-	     "\t width                                 = 0x%08" PRIx32 "\n"
-	     "\t height                                = 0x%08" PRIx32 "\n"
+	     "\t shading_rate_encoding                 = 0x%02" PRIx8 "\n"
 	     "\t size.x_max                            = 0x%04" PRIx16 "\n"
-	     "\t size.y_max                            = 0x%04" PRIx16 "\n",
+	     "\t size.y_max                            = 0x%04" PRIx16 "\n"
+	     "\t size.valid                            = %s\n",
 	     static_cast<uint32_t>(z.z_info.format), z.z_info.num_samples,
 	     static_cast<uint32_t>(z.z_info.texture_compatibility),
 	     z.z_info.htile_acceleration ? "true" : "false",
@@ -464,120 +359,13 @@ static void ZPrint(const char* func, const HW::DepthRenderTarget& z) {
 	     static_cast<uint32_t>(z.stencil_info.texture_compatibility),
 	     z.stencil_info.htile_stencil_disabled ? "true" : "false",
 	     z.stencil_info.expclear_enabled ? "true" : "false",
-	     z.stencil_info.partially_resident ? "true" : "false", z.depth_info.addr5_swizzle_mask,
-	     z.depth_info.array_mode, z.depth_info.pipe_config, z.depth_info.bank_width,
-	     z.depth_info.bank_height, z.depth_info.macro_tile_aspect, z.depth_info.num_banks,
-	     z.depth_view.slice_start, z.depth_view.slice_max, z.depth_view.current_mip_level,
+	     z.stencil_info.partially_resident ? "true" : "false", z.depth_view.slice_start,
+	     z.depth_view.slice_max, z.depth_view.current_mip_level,
 	     z.depth_view.depth_write_disable ? "true" : "false",
-	     z.depth_view.stencil_write_disable ? "true" : "false", z.htile_surface.linear,
-	     z.htile_surface.full_cache, z.htile_surface.htile_uses_preload_win,
-	     z.htile_surface.preload, z.htile_surface.prefetch_width, z.htile_surface.prefetch_height,
-	     z.htile_surface.dst_outside_zero_to_one, z.z_read_base_addr, z.stencil_read_base_addr,
-	     z.z_write_base_addr, z.stencil_write_base_addr, z.pitch_div8_minus1, z.height_div8_minus1,
-	     z.slice_div64_minus1, z.htile_data_base_addr, z.width, z.height, z.size.x_max,
-	     z.size.y_max);
-}
-
-// NOLINTNEXTLINE(readability-function-cognitive-complexity)
-static void ZCheck(const HW::DepthRenderTarget& z, const HW::DepthControl& dc,
-                   const HW::RenderControl& rc) {
-	const bool depth_active =
-	    dc.z_enable || dc.z_write_enable || dc.depth_bounds_enable || rc.depth_clear_enable;
-	const bool stencil_active = dc.stencil_enable || rc.stencil_clear_enable;
-	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 == Prospero::DepthFormat::kInvalid) {
-		EXIT_NOT_IMPLEMENTED(z.z_info.format != Prospero::DepthFormat::kInvalid);
-		EXIT_NOT_IMPLEMENTED(z.z_info.num_samples != 0);
-		EXIT_NOT_IMPLEMENTED(z.z_info.htile_acceleration != false);
-		EXIT_NOT_IMPLEMENTED(z.z_info.expclear_enabled != false);
-		EXIT_NOT_IMPLEMENTED(z.z_info.partially_resident != false);
-		EXIT_NOT_IMPLEMENTED(z.z_info.max_mip_level != 0);
-	} else {
-		EXIT_NOT_IMPLEMENTED(z.z_info.format != Prospero::DepthFormat::kZ16 &&
-		                     z.z_info.format != Prospero::DepthFormat::kZ32F);
-		if (z.z_info.num_samples != 0x00000000) {
-			static bool logged = false;
-			if (!logged) {
-				LOGF("DepthTarget: using native num_samples=0x%08" PRIx32 "\n",
-				     z.z_info.num_samples);
-				logged = true;
-			}
-		}
-		EXIT_NOT_IMPLEMENTED(z.z_info.expclear_enabled != false);
-		EXIT_NOT_IMPLEMENTED(z.z_info.partially_resident != false);
-		EXIT_NOT_IMPLEMENTED(z.z_info.max_mip_level != 0);
-	}
-
-	if (z.stencil_info.format == Prospero::StencilFormat::kInvalid) {
-		EXIT_NOT_IMPLEMENTED(z.stencil_info.expclear_enabled != false);
-		EXIT_NOT_IMPLEMENTED(z.stencil_info.partially_resident != false);
-	} else {
-		EXIT_NOT_IMPLEMENTED(z.stencil_info.format != Prospero::StencilFormat::k8UInt);
-		EXIT_NOT_IMPLEMENTED(z.stencil_info.expclear_enabled != false);
-		EXIT_NOT_IMPLEMENTED(z.stencil_info.partially_resident != false);
-	}
-
-	if (z.z_info.format != Prospero::DepthFormat::kInvalid ||
-	    z.stencil_info.format != Prospero::StencilFormat::kInvalid) {
-
-		EXIT_NOT_IMPLEMENTED(z.depth_info.addr5_swizzle_mask != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.array_mode != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.pipe_config != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.bank_width != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.bank_height != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.macro_tile_aspect != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.depth_info.num_banks != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.linear != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.full_cache != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.htile_uses_preload_win != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.preload != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.prefetch_width != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.prefetch_height != 0x00000000);
-		EXIT_NOT_IMPLEMENTED(z.htile_surface.dst_outside_zero_to_one != 0x00000000);
-
-		if (z.depth_view.current_mip_level != 0x00000000) {
-			static std::atomic<uint32_t> log_count {0};
-			if (log_count.fetch_add(1, std::memory_order_relaxed) < 16) {
-				LOGF("DepthTarget: temporary: ignoring PS5 current mip level=0x%08" PRIx32 "\n",
-				     z.depth_view.current_mip_level);
-			}
-		}
-		if (z.depth_view.depth_write_disable || z.depth_view.stencil_write_disable) {
-			static std::atomic<uint32_t> log_count {0};
-			if (log_count.fetch_add(1, std::memory_order_relaxed) < 16) {
-				LOGF("DepthTarget: honoring write disable depth=%s, stencil=%s\n",
-				     z.depth_view.depth_write_disable ? "true" : "false",
-				     z.depth_view.stencil_write_disable ? "true" : "false");
-			}
-		}
-		EXIT_NOT_IMPLEMENTED(!z.depth_view.depth_write_disable &&
-		                     z.z_read_base_addr != z.z_write_base_addr);
-		EXIT_NOT_IMPLEMENTED(!z.depth_view.stencil_write_disable &&
-		                     z.stencil_read_base_addr != z.stencil_write_base_addr);
-		EXIT_NOT_IMPLEMENTED(!z.depth_view.depth_write_disable && z.z_write_base_addr == 0);
-		// EXIT_NOT_IMPLEMENTED(z.pitch_div8_minus1 != 0x000000ff);
-		// EXIT_NOT_IMPLEMENTED(z.height_div8_minus1 != 0x0000008f);
-		// EXIT_NOT_IMPLEMENTED(z.slice_div64_minus1 != 0x00008fff);
-		// EXIT_NOT_IMPLEMENTED(z.htile_data_base_addr == 0);
-		// EXIT_NOT_IMPLEMENTED(z.width != 0x00000780);
-		// EXIT_NOT_IMPLEMENTED(z.height != 0x00000438);
-
-		EXIT_NOT_IMPLEMENTED(z.width != 0);
-		EXIT_NOT_IMPLEMENTED(z.height != 0);
-		if (z.size.x_max == 0 || z.size.y_max == 0) {
-			static std::atomic<uint32_t> log_count {0};
-			if (log_count.fetch_add(1) < 16) {
-				LOGF("DepthTarget: temporary: inferring PS5 default depth extent x_max=0x%04" PRIx16
-				     ", y_max=0x%04" PRIx16 "\n",
-				     z.size.x_max, z.size.y_max);
-			}
-		}
-	}
+	     z.depth_view.stencil_write_disable ? "true" : "false", z.z_read_base_addr,
+	     z.stencil_read_base_addr, z.z_write_base_addr, z.stencil_write_base_addr,
+	     z.htile_data_base_addr, z.shading_rate_encoding, z.size.x_max, z.size.y_max,
+	     z.size.valid ? "true" : "false");
 }
 
 static void ClipPrint(const char* func, const HW::ClipControl& c) {
@@ -608,7 +396,7 @@ static void ClipCheck(const HW::ClipControl& c) {
 	// dx_linear_attr_clip_enable preserves linear (noperspective) attributes at clip-generated
 	// vertices, which Vulkan provides as part of clipping and interpolation.
 	EXIT_NOT_IMPLEMENTED(c.user_clip_planes != 0 || c.user_clip_plane_mode != 0 ||
-	                     c.vertex_kill_any || c.user_clip_plane_negate_y || c.clip_disable ||
+	                     c.vertex_kill_any || c.user_clip_plane_negate_y ||
 	                     c.user_clip_plane_cull_only || c.cull_on_clipping_error_disable ||
 	                     c.force_viewport_index_from_vs_enable);
 }
@@ -664,16 +452,6 @@ static void McCheck(const HW::ModeControl& c) {
 	// EXIT_NOT_IMPLEMENTED(c.cull_front != false);
 	// EXIT_NOT_IMPLEMENTED(c.cull_back != false);
 	// EXIT_NOT_IMPLEMENTED(c.face != false);
-	if (c.poly_mode != 0) {
-		static bool logged = false;
-		if (!logged) {
-			LOGF("\t temporary: PA_SU_SC_MODE_CNTL.POLY_MODE is not fully implemented; continuing "
-			     "with filled polygons\n");
-			logged = true;
-		}
-	}
-	EXIT_NOT_IMPLEMENTED(c.polymode_front_ptype != 0 && c.polymode_front_ptype != 2);
-	EXIT_NOT_IMPLEMENTED(c.polymode_back_ptype != 0 && c.polymode_back_ptype != 2);
 	if (c.vtx_window_offset_enable) {
 		static bool logged = false;
 		if (!logged) {
@@ -683,7 +461,6 @@ static void McCheck(const HW::ModeControl& c) {
 			logged = true;
 		}
 	}
-	EXIT_NOT_IMPLEMENTED(c.provoking_vtx_last != false);
 	EXIT_NOT_IMPLEMENTED(c.persp_corr_dis != false);
 }
 
@@ -710,16 +487,7 @@ static void BcPrint(const char* func, const HW::BlendControl& c, const HW::Blend
 	     color.red, color.green, color.blue, color.alpha, cc.mode, cc.op);
 }
 
-static void BcCheck(const HW::BlendControl& /*c*/, const HW::BlendColor& color,
-                    const HW::ColorControl& cc) {
-	// EXIT_NOT_IMPLEMENTED(c.color_srcblend != 0);
-	// EXIT_NOT_IMPLEMENTED(c.color_comb_fcn != 0);
-	// EXIT_NOT_IMPLEMENTED(c.color_destblend != 0);
-	// EXIT_NOT_IMPLEMENTED(c.alpha_srcblend != 0);
-	// EXIT_NOT_IMPLEMENTED(c.alpha_comb_fcn != 0);
-	// EXIT_NOT_IMPLEMENTED(c.alpha_destblend != 0);
-	// EXIT_NOT_IMPLEMENTED(c.separate_alpha_blend != false);
-	// EXIT_NOT_IMPLEMENTED(c.enable != false);
+static void BcCheck(const HW::BlendColor& color, const HW::ColorControl& cc) {
 	if (color.red != 0.0f || color.green != 0.0f || color.blue != 0.0f || color.alpha != 0.0f) {
 		static bool logged = false;
 		if (!logged) {
@@ -759,8 +527,6 @@ static void DPrint(const char* func, const HW::DepthControl& c, const HW::Stenci
 	     "\t backface_enable      = %s\n"
 	     "\t stencilfunc          = %" PRIu8 "\n"
 	     "\t stencilfunc_bf       = %" PRIu8 "\n"
-	     "\t color_writes_on_depth_fail_enable  = %s\n"
-	     "\t color_writes_on_depth_pass_disable = %s\n"
 	     "\t stencil_fail         = %" PRIu8 "\n"
 	     "\t stencil_zpass        = %" PRIu8 "\n"
 	     "\t stencil_zfail        = %" PRIu8 "\n"
@@ -777,37 +543,11 @@ static void DPrint(const char* func, const HW::DepthControl& c, const HW::Stenci
 	     "\t stencil_opval_bf     = %" PRIu8 "\n",
 	     c.stencil_enable ? "true" : "false", c.z_enable ? "true" : "false",
 	     c.z_write_enable ? "true" : "false", c.depth_bounds_enable ? "true" : "false", c.zfunc,
-	     c.backface_enable ? "true" : "false", c.stencilfunc, c.stencilfunc_bf,
-	     c.color_writes_on_depth_fail_enable ? "true" : "false",
-	     c.color_writes_on_depth_pass_disable ? "true" : "false", s.stencil_fail, s.stencil_zpass,
-	     s.stencil_zfail, s.stencil_fail_bf, s.stencil_zpass_bf, s.stencil_zfail_bf,
-	     sm.stencil_testval, sm.stencil_mask, sm.stencil_writemask, sm.stencil_opval,
-	     sm.stencil_testval_bf, sm.stencil_mask_bf, sm.stencil_writemask_bf, sm.stencil_opval_bf);
-}
-
-static void DCheck(const HW::DepthControl& c, const HW::StencilControl& s,
-                   const HW::StencilMask& /*sm*/) {
-	// EXIT_NOT_IMPLEMENTED(c.stencil_enable != false);
-	// EXIT_NOT_IMPLEMENTED(c.z_enable != false);
-	// EXIT_NOT_IMPLEMENTED(c.z_write_enable != false);
-	// EXIT_NOT_IMPLEMENTED(c.zfunc != 0);
-	// Back-face stencil state is handled separately when enabled.
-	// EXIT_NOT_IMPLEMENTED(c.stencilfunc != 0);
-	// EXIT_NOT_IMPLEMENTED(c.stencilfunc_bf != 0);
-	EXIT_NOT_IMPLEMENTED(c.color_writes_on_depth_fail_enable != false);
-	EXIT_NOT_IMPLEMENTED(c.color_writes_on_depth_pass_disable != false);
-	// EXIT_NOT_IMPLEMENTED(s.stencil_fail != 0);
-	// EXIT_NOT_IMPLEMENTED(s.stencil_zpass != 0);
-	// EXIT_NOT_IMPLEMENTED(s.stencil_zfail != 0);
-	// Back-face stencil ops may legitimately differ from the front-face ops.
-	// EXIT_NOT_IMPLEMENTED(sm.stencil_testval != 0);
-	// EXIT_NOT_IMPLEMENTED(sm.stencil_mask != 0);
-	// EXIT_NOT_IMPLEMENTED(sm.stencil_writemask != 0);
-	// EXIT_NOT_IMPLEMENTED(sm.stencil_opval != 0);
-	// EXIT_NOT_IMPLEMENTED(sm.stencil_testval_bf != 0);
-	// EXIT_NOT_IMPLEMENTED(sm.stencil_mask_bf != 0);
-	// EXIT_NOT_IMPLEMENTED(sm.stencil_writemask_bf != 0);
-	// EXIT_NOT_IMPLEMENTED(sm.stencil_opval_bf != 0);
+	     c.backface_enable ? "true" : "false", c.stencilfunc, c.stencilfunc_bf, s.stencil_fail,
+	     s.stencil_zpass, s.stencil_zfail, s.stencil_fail_bf, s.stencil_zpass_bf,
+	     s.stencil_zfail_bf, sm.stencil_testval, sm.stencil_mask, sm.stencil_writemask,
+	     sm.stencil_opval, sm.stencil_testval_bf, sm.stencil_mask_bf, sm.stencil_writemask_bf,
+	     sm.stencil_opval_bf);
 }
 
 static void EqaaPrint(const char* func, const HW::EqaaControl& c) {
@@ -968,15 +708,6 @@ static void VpCheck(const HW::ScreenViewport& vp, const HW::ScanModeControl& smc
 	// EXIT_NOT_IMPLEMENTED(smc.vport_scissor_enable);
 	EXIT_NOT_IMPLEMENTED(smc.line_stipple_enable);
 
-	if (vp.viewports[0].zmin > 0.000000 || vp.viewports[0].zmax != 1.000000) {
-		static bool logged = false;
-		if (!logged) {
-			LOGF("\t warning: non-default viewport depth clamp zmin = %f, zmax = %f; using "
-			     "viewport scale/offset for Vulkan depth range\n",
-			     vp.viewports[0].zmin, vp.viewports[0].zmax);
-			logged = true;
-		}
-	}
 	// EXIT_NOT_IMPLEMENTED(vp.viewports[0].xscale != 960.000000);
 	// EXIT_NOT_IMPLEMENTED(vp.viewports[0].xoffset != 960.000000);
 	// EXIT_NOT_IMPLEMENTED(vp.viewports[0].yscale != -540.000000);
@@ -996,7 +727,7 @@ static void VpCheck(const HW::ScreenViewport& vp, const HW::ScanModeControl& smc
 	// EXIT_NOT_IMPLEMENTED(fabsf(vp.guard_band_horz_clip - 33.133327f) > 0.001f);
 	// EXIT_NOT_IMPLEMENTED(fabsf(vp.guard_band_vert_clip - 59.629623f) > 0.001f);
 
-	if (vp.guard_band_horz_discard != 0.0f || vp.guard_band_vert_discard != 0.0f) {
+	if (vp.guard_band_horz_discard != 1.0f || vp.guard_band_vert_discard != 1.0f) {
 		static std::atomic<uint32_t> log_count {0};
 		if (log_count.fetch_add(1) < 16) {
 			LOGF("\t warning: unsupported PS5 guard band discard = %f, %f, continuing\n",
@@ -1016,10 +747,6 @@ static bool ScissorRectValid(const ScissorRect& r) {
 	return r.right > r.left && r.bottom > r.top;
 }
 
-static bool ScissorRectSet(const ScissorRect& r) {
-	return r.left != 0 || r.top != 0 || r.right != 0 || r.bottom != 0;
-}
-
 static ScissorRect ScissorRectOffset(ScissorRect r, int x, int y) {
 	r.left += x;
 	r.right += x;
@@ -1029,8 +756,8 @@ static ScissorRect ScissorRectOffset(ScissorRect r, int x, int y) {
 }
 
 static ScissorRect ScissorRectIntersect(const ScissorRect& a, const ScissorRect& b) {
-	return {a.left > b.left ? a.left : b.left, a.top > b.top ? a.top : b.top,
-	        a.right < b.right ? a.right : b.right, a.bottom < b.bottom ? a.bottom : b.bottom};
+	return {std::max(a.left, b.left), std::max(a.top, b.top),
+	        std::min(a.right, b.right), std::min(a.bottom, b.bottom)};
 }
 
 static ScissorRect ScissorRectClamp(ScissorRect r, uint32_t width, uint32_t height) {
@@ -1077,49 +804,26 @@ static bool ScissorClipRuleToIntersectionMask(uint16_t rule, uint8_t* mask) {
 }
 
 ScissorRect calc_final_scissor(const HW::ScreenViewport& vp, const HW::ScanModeControl& smc,
-                               vk::Extent2D extent) {
-	ScissorRect screen {vp.screen_scissor_left, vp.screen_scissor_top, vp.screen_scissor_right,
-	                    vp.screen_scissor_bottom};
-	ScissorRect final = screen;
-
-	if (!ScissorRectSet(screen)) {
-		final = {0, 0, static_cast<int>(extent.width), static_cast<int>(extent.height)};
-
-		auto log_id = g_scissor_default_log_count.fetch_add(1);
-		if (log_id < 32) {
-			LOGF("temporary: default unset screen scissor to framebuffer extent %ux%u\n",
-			     extent.width, extent.height);
+                               vk::Extent2D extent, uint32_t viewport_index) {
+	EXIT_IF(viewport_index >= std::size(vp.viewports));
+	ScissorRect final {vp.screen_scissor_left, vp.screen_scissor_top, vp.screen_scissor_right,
+	                   vp.screen_scissor_bottom};
+	const auto intersect = [&](ScissorRect rect, bool window_offset) {
+		if (window_offset) {
+			rect = ScissorRectOffset(rect, vp.window_offset_x, vp.window_offset_y);
 		}
-	}
+		final = ScissorRectIntersect(final, rect);
+	};
+	intersect({vp.window_scissor_left, vp.window_scissor_top, vp.window_scissor_right,
+	           vp.window_scissor_bottom}, vp.window_scissor_window_offset_enable);
+	intersect({vp.generic_scissor_left, vp.generic_scissor_top, vp.generic_scissor_right,
+	           vp.generic_scissor_bottom}, vp.generic_scissor_window_offset_enable);
 
-	ScissorRect window {vp.window_scissor_left, vp.window_scissor_top, vp.window_scissor_right,
-	                    vp.window_scissor_bottom};
-	if (ScissorRectSet(window)) {
-		if (vp.window_scissor_window_offset_enable) {
-			window = ScissorRectOffset(window, vp.window_offset_x, vp.window_offset_y);
-		}
-		final = ScissorRectIntersect(final, window);
-	}
-
-	ScissorRect generic {vp.generic_scissor_left, vp.generic_scissor_top, vp.generic_scissor_right,
-	                     vp.generic_scissor_bottom};
-	if (ScissorRectSet(generic)) {
-		if (vp.generic_scissor_window_offset_enable) {
-			generic = ScissorRectOffset(generic, vp.window_offset_x, vp.window_offset_y);
-		}
-		final = ScissorRectIntersect(final, generic);
-	}
-
-	const auto& viewport = vp.viewports[0];
-	ScissorRect viewport_scissor {viewport.viewport_scissor_left, viewport.viewport_scissor_top,
-	                              viewport.viewport_scissor_right,
-	                              viewport.viewport_scissor_bottom};
-	if (smc.vport_scissor_enable && ScissorRectSet(viewport_scissor)) {
-		if (viewport.viewport_scissor_window_offset_enable) {
-			viewport_scissor =
-			    ScissorRectOffset(viewport_scissor, vp.window_offset_x, vp.window_offset_y);
-		}
-		final = ScissorRectIntersect(final, viewport_scissor);
+	const auto& viewport = vp.viewports[viewport_index];
+	if (smc.vport_scissor_enable) {
+		intersect({viewport.viewport_scissor_left, viewport.viewport_scissor_top,
+		           viewport.viewport_scissor_right, viewport.viewport_scissor_bottom},
+		          viewport.viewport_scissor_window_offset_enable);
 	}
 
 	if (vp.clip_rect_rule == 0) {
@@ -1132,12 +836,8 @@ ScissorRect calc_final_scissor(const HW::ScreenViewport& vp, const HW::ScanModeC
 					continue;
 				}
 
-				ScissorRect clip {vp.clip_rect_left[i], vp.clip_rect_top[i], vp.clip_rect_right[i],
-				                  vp.clip_rect_bottom[i]};
-				if (vp.clip_rect_window_offset_enable[i]) {
-					clip = ScissorRectOffset(clip, vp.window_offset_x, vp.window_offset_y);
-				}
-				final = ScissorRectIntersect(final, clip);
+				intersect({vp.clip_rect_left[i], vp.clip_rect_top[i], vp.clip_rect_right[i],
+				           vp.clip_rect_bottom[i]}, vp.clip_rect_window_offset_enable[i]);
 			}
 		} else {
 			static std::atomic<uint32_t> log_count {0};
@@ -1155,15 +855,10 @@ void hw_check(const CommandBuffer& buffer) {
 	const auto& hw      = buffer.GetRegisters();
 	const auto  rt_slot = render_target_first_bound_slot(buffer);
 	const auto& rt      = hw.GetRenderTarget(rt_slot);
-	const auto& bc      = hw.GetBlendControl(rt_slot);
 	const auto& bclr    = hw.GetBlendColor();
 	const auto& vp      = hw.GetScreenViewport();
-	const auto& z       = hw.GetDepthRenderTarget();
 	const auto& c       = hw.GetClipControl();
 	const auto& rc      = hw.GetRenderControl();
-	const auto& d       = hw.GetDepthControl();
-	const auto& s       = hw.GetStencilControl();
-	const auto& sm      = hw.GetStencilMask();
 	const auto& mc      = hw.GetModeControl();
 	const auto& eqaa    = hw.GetEqaaControl();
 	const auto& cc      = hw.GetColorControl();
@@ -1184,25 +879,22 @@ void hw_check(const CommandBuffer& buffer) {
 	RtCheck(rt);
 	log_phase("vp");
 	VpCheck(vp, smc);
-	log_phase("z");
-	ZCheck(z, d, rc);
 	log_phase("clip");
 	ClipCheck(c);
 	log_phase("rc");
 	RcCheck(rc);
 	log_phase("depth");
-	DCheck(d, s, sm);
 	log_phase("mode");
 	McCheck(mc);
 	log_phase("blend");
-	BcCheck(bc, bclr, cc);
+	BcCheck(bclr, cc);
 	log_phase("eqaa");
 	EqaaCheck(eqaa, ac);
 	log_phase("aa");
 	AaCheck(aa, ac);
 	log_phase("done");
 
-	if (RenderTargetMaskHasBoundMrt(buffer)) {
+	if (graphics_debug_dump_enabled() && RenderTargetMaskHasBoundMrt(buffer)) {
 		LOGF("MRT render target mask: 0x%08" PRIx32 "\n", hw.GetRenderTargetMask());
 		for (uint32_t i = 0; i < 8; i++) {
 			const auto& mrt = hw.GetRenderTarget(i);
@@ -1254,7 +946,7 @@ void hw_print(const CommandBuffer& buffer) {
 		     hw.GetRenderTargetMask(), hw.GetDepthClearValue(), hw.GetStencilClearValue(),
 		     hw.GetLineWidth(), hw.GetPrimitiveResetIndex());
 
-		LOGF("%s", Common::Concat(rt_print("RenderTraget:", rt), "").c_str());
+		LOGF("%s", rt_print("RenderTraget:", rt).c_str());
 
 		ZPrint("DepthRenderTraget:", z);
 		VpPrint("ScreenViewport:", vp, smc);

src/graphics/host_gpu/renderer/debug.h

@@ -5,7 +5,6 @@
 
 #include <cstdint>
 #include <string>
-#include <vector>
 
 namespace Libs::Graphics {
 
@@ -13,7 +12,6 @@ class CommandBuffer;
 
 namespace HW {
 class Context;
-class Shader;
 class UserConfig;
 struct RenderTarget;
 struct ScanModeControl;
@@ -32,14 +30,13 @@ uint32_t                 render_target_first_bound_slot(const CommandBuffer& buf
 bool                     graphics_debug_dump_enabled();
 void                     uc_print(const char* func, const HW::UserConfig& uc);
 void                     uc_check(const HW::UserConfig& uc);
-void                     sh_print(const char* func, const HW::Shader& uc);
-std::vector<std::string> rt_print(const char* func, const HW::RenderTarget& rt);
+std::string              rt_print(const char* func, const HW::RenderTarget& rt);
 bool                     RenderIsColorTileModeLinear(Prospero::TileMode tile_mode);
 void                     hw_print(const CommandBuffer& buffer);
 void                     hw_check(const CommandBuffer& buffer);
 void                     LogDrawPhase(const char* draw_name, const char* phase);
 ScissorRect calc_final_scissor(const HW::ScreenViewport& vp, const HW::ScanModeControl& smc,
-                               vk::Extent2D extent);
+                               vk::Extent2D extent, uint32_t viewport_index);
 
 } // namespace Libs::Graphics
 

src/graphics/host_gpu/renderer/depthRenderTarget.cpp

@@ -10,6 +10,7 @@
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/guest_gpu/tile.h"
 #include "graphics/host_gpu/graphicContext.h"
+#include "graphics/host_gpu/renderer/commandScheduler.h"
 #include "graphics/host_gpu/renderer/debug.h"
 #include "graphics/host_gpu/renderer/image/imageView.h"
 #include "graphics/host_gpu/renderer/image/textureCommon.h"
@@ -18,6 +19,7 @@
 #include "graphics/host_gpu/vulkanCommon.h"
 
 #include <algorithm>
+#include <array>
 #include <atomic>
 #include <cmath>
 #include <cstdarg>
@@ -44,8 +46,12 @@ static vk::StencilOp ConvertStencilOp(uint8_t value, uint8_t write_mask, uint8_t
 	switch (static_cast<Prospero::StencilOp>(value)) {
 		case Prospero::StencilOp::kKeep: return vk::StencilOp::eKeep;
 		case Prospero::StencilOp::kZero: return vk::StencilOp::eZero;
-		case Prospero::StencilOp::kReplaceTest:
-		case Prospero::StencilOp::kReplaceOp: return vk::StencilOp::eReplace;
+		case Prospero::StencilOp::kReplaceTest: return vk::StencilOp::eReplace;
+		case Prospero::StencilOp::kReplaceOp:
+			if ((op_value & write_mask) == 0) {
+				return vk::StencilOp::eZero;
+			}
+			return vk::StencilOp::eReplace;
 		case Prospero::StencilOp::kAddClamp: return vk::StencilOp::eIncrementAndClamp;
 		case Prospero::StencilOp::kSubClamp: return vk::StencilOp::eDecrementAndClamp;
 		case Prospero::StencilOp::kInvert: return vk::StencilOp::eInvert;
@@ -65,9 +71,38 @@ static vk::StencilOp ConvertStencilOp(uint8_t value, uint8_t write_mask, uint8_t
 	}
 }
 
-static bool UsesStencilOpValue(uint8_t fail, uint8_t pass, uint8_t depth_fail) {
-	constexpr auto replace_op = static_cast<uint8_t>(Prospero::StencilOp::kReplaceOp);
-	return fail == replace_op || pass == replace_op || depth_fail == replace_op;
+static vk::StencilOpState ConvertStencilState(
+    uint8_t compare, const std::array<uint8_t, 3>& operations, uint8_t op_value,
+    const vk::StencilOpState& state) {
+	const auto test_value = state.reference;
+	auto reference       = test_value;
+	auto required_bits   = state.compareMask;
+	if (compare == static_cast<uint8_t>(vk::CompareOp::eAlways) ||
+	    compare == static_cast<uint8_t>(vk::CompareOp::eNever)) {
+		required_bits = 0;
+	}
+	std::array<vk::StencilOp, 3> converted {};
+	for (size_t i = 0; i < operations.size(); i++) {
+		converted[i] = ConvertStencilOp(operations[i], state.writeMask, op_value);
+		if (converted[i] != vk::StencilOp::eReplace) {
+			continue;
+		}
+		auto replacement = test_value;
+		if (static_cast<Prospero::StencilOp>(operations[i]) == Prospero::StencilOp::kReplaceOp) {
+			replacement = op_value;
+		}
+		if (((reference ^ replacement) & required_bits & state.writeMask) != 0) {
+			DepthFatal("unsupported stencil replacement: compare=%u, compare mask=0x%02" PRIx32
+			           ", write mask=0x%02" PRIx32 ", operation value=0x%02" PRIx8
+			           ", test value=0x%02" PRIx32,
+			           compare, state.compareMask, state.writeMask, op_value, test_value);
+		}
+		// Vulkan shares one reference between comparison and every replacement on this face.
+		reference = (reference & ~state.writeMask) | (replacement & state.writeMask);
+		required_bits |= state.writeMask;
+	}
+	return {converted[0], converted[1], converted[2], static_cast<vk::CompareOp>(compare),
+	        state.compareMask, state.writeMask, reference};
 }
 
 [[nodiscard]] static vk::Format ResolveHostDepthAttachmentFormat(const CommandBuffer&     buffer,
@@ -96,61 +131,20 @@ static bool UsesStencilOpValue(uint8_t fail, uint8_t pass, uint8_t depth_fail) {
 	return vk::Format::eUndefined;
 }
 
-// NOLINTNEXTLINE(readability-function-cognitive-complexity)
-void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer& buffer,
-                                              RenderDepthInfo& r) {
-	KYTY_PROFILER_FUNCTION();
-	(void)submit_id;
-	const auto& hw          = buffer.GetRegisters();
-	const auto& z           = hw.GetDepthRenderTarget();
-	const auto& rc          = hw.GetRenderControl();
-	const auto& dc          = hw.GetDepthControl();
-	const auto& sc          = hw.GetStencilControl();
-	const auto& sm          = hw.GetStencilMask();
-	const bool  has_stencil = z.stencil_info.format != Prospero::StencilFormat::kInvalid;
-	const bool  depth_active =
-	    dc.z_enable || dc.z_write_enable || dc.depth_bounds_enable || rc.depth_clear_enable;
-	const bool stencil_active = has_stencil && (dc.stencil_enable || rc.stencil_clear_enable);
-	if (!depth_active && !stencil_active) {
-		return;
-	}
+static TextureCache::ImageDesc MakeDepthTargetDesc(const CommandBuffer& buffer,
+                                                  const HW::DepthRenderTarget& z,
+                                                  bool write_buffer = false) {
+	const bool has_stencil = z.stencil_info.format != Prospero::StencilFormat::kInvalid;
+	const auto depth_address = write_buffer ? z.z_write_base_addr : z.z_read_base_addr;
+	const auto stencil_address =
+	    write_buffer ? z.stencil_write_base_addr : z.stencil_read_base_addr;
 	if (!z.z_info.HasValidTextureCompatibility() ||
 	    !z.stencil_info.HasValidTextureCompatibility()) {
 		DepthFatal("invalid PS5 depth texture-compatibility encoding");
 	}
-	const bool attachment_unbound =
-	    z.z_info.format == Prospero::DepthFormat::kInvalid &&
-	    z.stencil_info.format == Prospero::StencilFormat::kInvalid && z.z_info.num_samples == 0 &&
-	    z.z_info.texture_compatibility == Prospero::TextureCompatiblePlaneCompression::kDisable &&
-	    !z.z_info.expclear_enabled && !z.z_info.partially_resident && z.z_info.max_mip_level == 0 &&
-	    z.stencil_info.texture_compatibility == Prospero::TextureCompatibleStencil::kDisable &&
-	    !z.stencil_info.expclear_enabled && !z.stencil_info.partially_resident &&
-	    z.depth_view.slice_start == 0 && z.depth_view.slice_max == 0 &&
-	    z.depth_view.current_mip_level == 0 && !z.depth_view.depth_write_disable &&
-	    !z.depth_view.stencil_write_disable && z.depth_info.addr5_swizzle_mask == 0 &&
-	    z.depth_info.array_mode == 0 && z.depth_info.pipe_config == 0 &&
-	    z.depth_info.bank_width == 0 && z.depth_info.bank_height == 0 &&
-	    z.depth_info.macro_tile_aspect == 0 && z.depth_info.num_banks == 0 &&
-	    z.htile_surface.linear == 0 && z.htile_surface.full_cache == 0 &&
-	    z.htile_surface.htile_uses_preload_win == 0 && z.htile_surface.preload == 0 &&
-	    z.htile_surface.prefetch_width == 0 && z.htile_surface.prefetch_height == 0 &&
-	    z.htile_surface.dst_outside_zero_to_one == 0 && z.z_read_base_addr == 0 &&
-	    z.z_write_base_addr == 0 && z.stencil_read_base_addr == 0 &&
-	    z.stencil_write_base_addr == 0 && z.htile_data_base_addr == 0 &&
-	    // DB_DEPTH_SIZE_XY is independent state and may remain programmed after the attachment
-	    // formats and addresses are unbound. A zero encoding is the valid 1x1 value, so its
-	    // presence alone must not manufacture a depth attachment.
-	    !z.z_info.htile_acceleration && !z.width_height_valid && !z.pitch_height_valid &&
-	    z.size.x_max == 0 && z.size.y_max == 0 && z.pitch_div8_minus1 == 0 &&
-	    z.height_div8_minus1 == 0 && z.slice_div64_minus1 == 0 && z.width == 0 && z.height == 0;
-	if (attachment_unbound) {
-		static std::atomic_bool logged = false;
-		if (!logged.exchange(true, std::memory_order_relaxed)) {
-			LOGF("DepthTarget: ignoring enabled depth/stencil state without a bound attachment\n");
-		}
-		return;
-	}
 	const bool has_htile = z.z_info.htile_acceleration;
+	const bool unsupported_shading_rate_encoding =
+	    z.shading_rate_encoding > 1 || (z.shading_rate_encoding != 0 && !has_htile);
 	const auto samples   = render_sample_count(z.z_info.num_samples);
 	if (samples == 0) {
 		DepthFatal("unsupported depth fragment count: %u", z.z_info.num_samples);
@@ -165,34 +159,19 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 			DepthFatal("invalid depth view: base=%u last=%u", z.depth_view.slice_start,
 			           z.depth_view.slice_max);
 	}
-	if (rc.copy_centroid || rc.copy_sample != 0 || z.z_info.expclear_enabled ||
-	    z.stencil_info.expclear_enabled || z.z_info.partially_resident ||
+	// EXPCLEAR permits an HTile acceleration state; the host attachment is already expanded.
+	if (z.z_info.partially_resident ||
 	    z.stencil_info.partially_resident || z.z_info.max_mip_level != 0 ||
-	    z.depth_view.current_mip_level != 0 || z.depth_info.addr5_swizzle_mask != 0 ||
-	    z.depth_info.array_mode != 0 || z.depth_info.pipe_config != 0 ||
-	    z.depth_info.bank_width != 0 || z.depth_info.bank_height != 0 ||
-	    z.depth_info.macro_tile_aspect != 0 || z.depth_info.num_banks != 0 ||
-	    z.htile_surface.linear != 0 || z.htile_surface.full_cache != 0 ||
-	    z.htile_surface.htile_uses_preload_win != 0 || z.htile_surface.preload != 0 ||
-	    z.htile_surface.prefetch_width != 0 || z.htile_surface.prefetch_height != 0 ||
-	    z.htile_surface.dst_outside_zero_to_one != 0 || z.z_read_base_addr == 0 ||
-	    (!z.depth_view.depth_write_disable && z.z_write_base_addr != z.z_read_base_addr) ||
-	    (z.z_read_base_addr & 0xffffu) != 0 ||
-	    dc.zfunc > static_cast<uint8_t>(vk::CompareOp::eAlways)) {
+	    z.depth_view.current_mip_level != 0 || unsupported_shading_rate_encoding ||
+	    depth_address == 0 || (depth_address & 0xffffu) != 0) {
 		DepthFatal("unsupported depth register state");
 	}
 	if (has_stencil) {
 		if (z.stencil_info.format != Prospero::StencilFormat::k8UInt || !htile_stencil_compat ||
-		    z.stencil_read_base_addr == 0 ||
-		    (!z.depth_view.stencil_write_disable &&
-		     z.stencil_write_base_addr != z.stencil_read_base_addr) ||
-		    (z.stencil_read_base_addr & 0xffffu) != 0) {
+		    stencil_address == 0 || (stencil_address & 0xffffu) != 0) {
 			DepthFatal("unsupported stencil attachment state");
 		}
-	} else if (z.stencil_read_base_addr != 0 || z.stencil_write_base_addr != 0 ||
-	           !htile_stencil_compat ||
-	           z.stencil_info.texture_compatibility !=
-	               Prospero::TextureCompatibleStencil::kDisable) {
+	} else if (z.stencil_read_base_addr != 0 || z.stencil_write_base_addr != 0) {
 		DepthFatal("stencil state without an active stencil attachment");
 	}
 	if (has_htile) {
@@ -203,37 +182,27 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 			DepthFatal("HTile clear tracking supports at most 32 slices");
 		}
 	}
-	const bool size_xy_valid = z.size.valid;
-	const bool wh_valid      = z.width_height_valid && z.width != 0 && z.height != 0;
-	if (!size_xy_valid && !wh_valid) {
+	if (!z.size.valid) {
 		DepthFatal("missing depth extent");
 	}
-	const uint32_t width  = size_xy_valid ? static_cast<uint32_t>(z.size.x_max) + 1u : z.width;
-	const uint32_t height = size_xy_valid ? static_cast<uint32_t>(z.size.y_max) + 1u : z.height;
-	if (width > 16384 || height > 16384 ||
-	    (size_xy_valid && wh_valid && (width != z.width || height != z.height)) ||
-	    (!z.pitch_height_valid &&
-	     (z.pitch_div8_minus1 != 0 || z.height_div8_minus1 != 0 || z.slice_div64_minus1 != 0))) {
-		DepthFatal("inconsistent depth extent or encoded layout");
+	const uint32_t width  = static_cast<uint32_t>(z.size.x_max) + 1u;
+	const uint32_t height = static_cast<uint32_t>(z.size.y_max) + 1u;
+	if (width > 16384 || height > 16384) {
+		DepthFatal("invalid depth extent");
 	}
 	const auto* policy = FindDepthFormatPolicy(z.z_info.format);
 	if (policy == nullptr) {
 		DepthFatal("unsupported depth/stencil format pair");
 	}
 	const auto ideal_format = DepthAttachmentFormat(*policy, has_stencil);
-	r.format = ResolveHostDepthAttachmentFormat(buffer, *policy, has_stencil, samples);
-	if (r.format == vk::Format::eUndefined) {
+	const auto format = ResolveHostDepthAttachmentFormat(buffer, *policy, has_stencil, samples);
+	if (format == vk::Format::eUndefined) {
 		DepthFatal("no host depth/stencil format supports required usage for %s",
-		           VulkanToString(ideal_format).c_str());
+		           vk::to_string(ideal_format).c_str());
 	}
 	const auto     guest_format = policy->guest_format;
 	const uint32_t bytes        = policy->bytes_per_element;
 	const auto     pitch        = TileGetDepthPitch(width, bytes, z.z_info.num_samples);
-	if (z.pitch_height_valid && ((static_cast<uint64_t>(z.pitch_div8_minus1) + 1u) * 8u != pitch ||
-	                             (static_cast<uint64_t>(z.height_div8_minus1) + 1u) * 8u !=
-	                                 ((static_cast<uint64_t>(height) + 7u) & ~7ull))) {
-		DepthFatal("encoded depth pitch or height mismatch");
-	}
 	TileSizeAlign depth_size {};
 	TileSizeAlign stencil_size {};
 	TileSizeAlign htile_size {};
@@ -244,14 +213,6 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 	    (has_htile != (htile_size.align == 32768 && htile_size.size != 0))) {
 		DepthFatal("unsupported depth/stencil/HTile footprint");
 	}
-	if (z.pitch_height_valid &&
-	    (static_cast<uint64_t>(z.slice_div64_minus1) + 1u) * 64u != depth_size.size) {
-		DepthFatal("depth footprint mismatch: extent=%ux%u pitch=%u expected=0x%016" PRIx64
-		           " align=0x%016" PRIx64 " encoded_valid=%u encoded=0x%016" PRIx64,
-		           width, height, pitch, depth_size.size, depth_size.align,
-		           z.pitch_height_valid ? 1u : 0u,
-		           (static_cast<uint64_t>(z.slice_div64_minus1) + 1u) * 64u);
-	}
 	if (depth_size.size > UINT64_MAX / view.image_layers ||
 	    stencil_size.size > UINT64_MAX / view.image_layers ||
 	    htile_size.size > UINT64_MAX / view.image_layers) {
@@ -260,27 +221,98 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 	const auto depth_backing_size   = depth_size.size * view.image_layers;
 	const auto stencil_backing_size = stencil_size.size * view.image_layers;
 	const auto htile_backing_size   = htile_size.size * view.image_layers;
-	if (depth_backing_size > TRACKER_ADDRESS_SIZE - z.z_read_base_addr ||
-	    (has_stencil && stencil_backing_size > TRACKER_ADDRESS_SIZE - z.stencil_read_base_addr) ||
-	    (has_htile && htile_backing_size > TRACKER_ADDRESS_SIZE - z.htile_data_base_addr)) {
+	if (!GuestRange {depth_address, depth_backing_size}.Valid() ||
+	    (has_stencil && !GuestRange {stencil_address, stencil_backing_size}.Valid()) ||
+	    (has_htile && !GuestRange {z.htile_data_base_addr, htile_backing_size}.Valid())) {
 		DepthFatal("layered depth backing range is invalid");
 	}
-	r.htile                   = has_htile;
-	r.width                   = width;
-	r.height                  = height;
-	r.samples                 = samples;
-	r.depth_buffer_size       = depth_backing_size;
-	r.depth_buffer_vaddr      = z.z_read_base_addr;
-	r.stencil_buffer_size     = has_stencil ? stencil_backing_size : 0;
-	r.stencil_buffer_vaddr    = has_stencil ? z.stencil_read_base_addr : 0;
-	r.htile_buffer_size       = has_htile ? htile_backing_size : 0;
-	r.htile_buffer_vaddr      = has_htile ? z.htile_data_base_addr : 0;
+	TextureCache::ImageDesc desc {};
+	desc.type                 = TextureCache::BindingType::DepthTarget;
+	desc.info.data            = {depth_address, depth_backing_size};
+	desc.info.stencil =
+	    has_stencil ? GuestRange {stencil_address, stencil_backing_size} : GuestRange {};
+	desc.info.pixel_format    = format;
+	desc.info.guest_format    = guest_format;
+	desc.info.type            = Prospero::ImageType::kColor2D;
+	desc.info.extent          = {width, height, 1};
+	desc.info.resources       = {1, view.image_layers};
+	desc.info.pitch           = pitch;
+	desc.info.bytes_per_block = bytes;
+	desc.info.samples         = samples;
+	desc.info.tile_mode       = Prospero::TileMode::kDepth;
+	desc.info.mip_layout[0]   = {0, depth_backing_size, pitch, height};
+	desc.info.metadata.range =
+	    has_htile ? GuestRange {z.htile_data_base_addr, htile_backing_size} : GuestRange {};
+	desc.info.metadata.kind   = has_htile ? ImageMetadataKind::Htile : ImageMetadataKind::None;
+	desc.info.metadata.stencil_compressed =
+	    has_stencil && has_htile && !z.stencil_info.htile_stencil_disabled;
+	desc.view_info.format = format;
+	desc.view_info.type =
+	    view.layer_count == 1 ? vk::ImageViewType::e2D : vk::ImageViewType::e2DArray;
+	desc.view_info.aspect      = ImageViewOps::DepthAspectMask(format);
+	desc.view_info.base_level  = 0;
+	desc.view_info.level_count = 1;
+	desc.view_info.base_layer  = view.base_layer;
+	desc.view_info.layer_count = view.layer_count;
+	desc.view_info.usage       = vk::ImageUsageFlagBits::eDepthStencilAttachment;
+	return desc;
+}
+
+// NOLINTNEXTLINE(readability-function-cognitive-complexity)
+void RenderExecutor::ResolveRenderDepthTarget(CommandBuffer& buffer, RenderDepthInfo& r) {
+	KYTY_PROFILER_FUNCTION();
+	const auto& hw          = buffer.GetRegisters();
+	const auto& z           = hw.GetDepthRenderTarget();
+	const auto& rc          = hw.GetRenderControl();
+	const auto& dc          = hw.GetDepthControl();
+	const auto& sc          = hw.GetStencilControl();
+	const auto& sm          = hw.GetStencilMask();
+	const bool  has_stencil = z.stencil_info.format != Prospero::StencilFormat::kInvalid;
+	const bool depth_active = dc.z_enable || dc.depth_bounds_enable ||
+	                          rc.depth_clear_enable || rc.copy_depth_to_color;
+	const bool stencil_active =
+	    has_stencil && (dc.stencil_enable || rc.stencil_clear_enable || rc.copy_stencil_to_color);
+	if (!depth_active && !stencil_active) {
+		return;
+	}
+	const bool attachment_unbound =
+	    z.z_info.format == Prospero::DepthFormat::kInvalid &&
+	    z.stencil_info.format == Prospero::StencilFormat::kInvalid && z.z_info.num_samples == 0 &&
+	    z.z_info.texture_compatibility == Prospero::TextureCompatiblePlaneCompression::kDisable &&
+	    !z.z_info.expclear_enabled && !z.z_info.partially_resident && z.z_info.max_mip_level == 0 &&
+	    z.stencil_info.texture_compatibility == Prospero::TextureCompatibleStencil::kDisable &&
+	    !z.stencil_info.expclear_enabled && !z.stencil_info.partially_resident &&
+	    z.depth_view.slice_start == 0 && z.depth_view.slice_max == 0 &&
+	    z.depth_view.current_mip_level == 0 && !z.depth_view.depth_write_disable &&
+	    !z.depth_view.stencil_write_disable && z.z_read_base_addr == 0 && z.z_write_base_addr == 0 &&
+	    z.stencil_read_base_addr == 0 && z.stencil_write_base_addr == 0 &&
+	    z.htile_data_base_addr == 0 &&
+	    // DB_DEPTH_SIZE_XY is independent state and may remain programmed after the attachment
+	    // formats and addresses are unbound. A zero encoding is the valid 1x1 value, so its
+	    // presence alone must not manufacture a depth attachment.
+	    !z.z_info.htile_acceleration && z.shading_rate_encoding == 0 && z.size.x_max == 0 &&
+	    z.size.y_max == 0;
+	if (attachment_unbound) {
+		static std::atomic_bool logged = false;
+		if (!logged.exchange(true, std::memory_order_relaxed)) {
+			LOGF("DepthTarget: ignoring enabled depth/stencil state without a bound attachment\n");
+		}
+		return;
+	}
+	if (rc.copy_depth_to_color || rc.copy_stencil_to_color || rc.copy_centroid ||
+	    rc.copy_sample != 0 || dc.zfunc > static_cast<uint8_t>(vk::CompareOp::eAlways) ||
+	    (!z.depth_view.depth_write_disable && z.z_write_base_addr != z.z_read_base_addr) ||
+	    (has_stencil && !z.depth_view.stencil_write_disable &&
+	     z.stencil_write_base_addr != z.stencil_read_base_addr)) {
+		DepthFatal("unsupported depth register state");
+	}
+	r.desc = MakeDepthTargetDesc(buffer, z);
 	r.depth_clear_enable      = rc.depth_clear_enable;
-	r.depth_meta_clear_enable = false;
 	r.depth_load_clear_enable = r.depth_clear_enable;
 	r.depth_clear_value       = hw.GetDepthClearValue();
 	r.depth_test_enable       = dc.z_enable;
-	r.depth_write_enable      = dc.z_write_enable && !z.depth_view.depth_write_disable;
+	r.depth_write_enable      = r.depth_test_enable && dc.z_write_enable &&
+	                            !z.depth_view.depth_write_disable && !r.depth_clear_enable;
 	r.depth_compare_op        = static_cast<vk::CompareOp>(dc.zfunc);
 
 	r.depth_bounds_test_enable = dc.depth_bounds_enable;
@@ -298,75 +330,91 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer&
 		const uint8_t back_write_mask  = stencil_ops_disabled ? 0 : sm.stencil_writemask_bf;
 		if (dc.stencilfunc > static_cast<uint8_t>(vk::CompareOp::eAlways) ||
 		    (dc.backface_enable &&
-		     dc.stencilfunc_bf > static_cast<uint8_t>(vk::CompareOp::eAlways)) ||
-		    (front_write_mask != 0 &&
-		     UsesStencilOpValue(sc.stencil_fail, sc.stencil_zpass, sc.stencil_zfail) &&
-		     sm.stencil_opval != sm.stencil_testval) ||
-		    (dc.backface_enable && back_write_mask != 0 &&
-		     UsesStencilOpValue(sc.stencil_fail_bf, sc.stencil_zpass_bf, sc.stencil_zfail_bf) &&
-		     sm.stencil_opval_bf != sm.stencil_testval_bf)) {
-			DepthFatal("unsupported stencil compare or replacement state");
+		     dc.stencilfunc_bf > static_cast<uint8_t>(vk::CompareOp::eAlways))) {
+			DepthFatal("unsupported stencil compare state");
 		}
-		r.stencil_static_front = {
-		    ConvertStencilOp(sc.stencil_fail, front_write_mask, sm.stencil_opval),
-		    ConvertStencilOp(sc.stencil_zpass, front_write_mask, sm.stencil_opval),
-		    ConvertStencilOp(sc.stencil_zfail, front_write_mask, sm.stencil_opval),
-		    static_cast<vk::CompareOp>(dc.stencilfunc)};
-		r.stencil_dynamic_front = {sm.stencil_mask, front_write_mask, sm.stencil_testval};
+		r.stencil_front = ConvertStencilState(
+		    dc.stencilfunc, {sc.stencil_fail, sc.stencil_zpass, sc.stencil_zfail},
+		    sm.stencil_opval, {.compareMask = sm.stencil_mask, .writeMask = front_write_mask,
+		                       .reference = sm.stencil_testval});
 		if (dc.backface_enable) {
-			r.stencil_static_back = {
-			    ConvertStencilOp(sc.stencil_fail_bf, back_write_mask, sm.stencil_opval_bf),
-			    ConvertStencilOp(sc.stencil_zpass_bf, back_write_mask, sm.stencil_opval_bf),
-			    ConvertStencilOp(sc.stencil_zfail_bf, back_write_mask, sm.stencil_opval_bf),
-			    static_cast<vk::CompareOp>(dc.stencilfunc_bf)};
-			r.stencil_dynamic_back = {sm.stencil_mask_bf, back_write_mask, sm.stencil_testval_bf};
+			r.stencil_back = ConvertStencilState(
+			    dc.stencilfunc_bf, {sc.stencil_fail_bf, sc.stencil_zpass_bf, sc.stencil_zfail_bf},
+			    sm.stencil_opval_bf, {.compareMask = sm.stencil_mask_bf, .writeMask = back_write_mask,
+			                          .reference = sm.stencil_testval_bf});
 		} else {
-			r.stencil_static_back  = r.stencil_static_front;
-			r.stencil_dynamic_back = r.stencil_dynamic_front;
+			r.stencil_back = r.stencil_front;
 		}
 	}
-	r.vaddr_num = has_stencil ? 2 : 1;
-	r.vaddr[0]  = r.depth_buffer_vaddr;
-	r.size[0]   = r.depth_buffer_size;
-	if (has_stencil) {
-		r.vaddr[1] = r.stencil_buffer_vaddr;
-		r.size[1]  = r.stencil_buffer_size;
-	}
-	TextureCache::ImageDesc desc {};
-	desc.type                 = TextureCache::BindingType::DepthTarget;
-	desc.info.data            = {r.depth_buffer_vaddr, r.depth_buffer_size};
-	desc.info.stencil         = {r.stencil_buffer_vaddr, r.stencil_buffer_size};
-	desc.info.pixel_format    = r.format;
-	desc.info.guest_format    = guest_format;
-	desc.info.type            = Prospero::ImageType::kColor2D;
-	desc.info.extent          = {width, height, 1};
-	desc.info.resources       = {1, view.image_layers};
-	desc.info.pitch           = pitch;
-	desc.info.bytes_per_block = bytes;
-	desc.info.samples         = samples;
-	desc.info.tile_mode       = Prospero::TileMode::kDepth;
-	desc.info.mip_layout[0]   = {0, r.depth_buffer_size, pitch, height};
-	desc.info.metadata.range  = {r.htile_buffer_vaddr, r.htile_buffer_size};
-	desc.info.metadata.kind   = has_htile ? ImageMetadataKind::Htile : ImageMetadataKind::None;
-	desc.info.metadata.stencil_compressed =
-	    has_stencil && has_htile && !z.stencil_info.htile_stencil_disabled;
-	desc.view_info.format = r.format;
-	desc.view_info.type =
-	    view.layer_count == 1 ? vk::ImageViewType::e2D : vk::ImageViewType::e2DArray;
-	desc.view_info.aspect      = ImageViewOps::DepthAspectMask(r.format);
-	desc.view_info.base_level  = 0;
-	desc.view_info.level_count = 1;
-	desc.view_info.base_layer  = view.base_layer;
-	desc.view_info.layer_count = view.layer_count;
-	desc.view_info.usage       = vk::ImageUsageFlagBits::eDepthStencilAttachment;
-	r.desc                     = std::move(desc);
-	auto& cache                = m_context.GetTextureCache();
-	r.image_id                 = cache.FindImage(r.desc);
-	r.image_view               = nullptr;
+	auto& cache = m_context.GetTextureCache();
+	r.image_id = cache.FindImage(r.desc);
 	BindRenderTarget(r.image_id);
 }
 
+bool RenderExecutor::DepthStencilCopy(CommandBuffer& buffer) {
+	const auto& hw       = buffer.GetRegisters();
+	const auto& z        = hw.GetDepthRenderTarget();
+	const auto& override = hw.GetDepthRenderOverride();
+	if (hw.GetColorControl().mode != 0) {
+		return false;
+	}
+	const bool depth_copy = override.force_z_dirty && override.force_z_valid &&
+	                        z.z_info.format != Prospero::DepthFormat::kInvalid &&
+	                        z.z_read_base_addr != 0 && z.z_write_base_addr != 0 &&
+	                        z.z_read_base_addr != z.z_write_base_addr;
+	const bool stencil_copy = override.force_stencil_dirty && override.force_stencil_valid &&
+	                          z.stencil_info.format != Prospero::StencilFormat::kInvalid &&
+	                          z.stencil_read_base_addr != 0 && z.stencil_write_base_addr != 0 &&
+	                          z.stencil_read_base_addr != z.stencil_write_base_addr;
+	if (!depth_copy && !stencil_copy) {
+		return false;
+	}
+
+	auto  read_desc  = MakeDepthTargetDesc(buffer, z);
+	auto  write_desc = MakeDepthTargetDesc(buffer, z, true);
+	auto& cache      = m_context.GetTextureCache();
+	const auto read_id = cache.FindImage(read_desc);
+	BindRenderTarget(read_id);
+	const auto write_id = cache.FindImage(write_desc);
+	BindRenderTarget(write_id);
+	cache.UpdateImage(read_id);
+	cache.UpdateImage(write_id);
+	cache.MarkGpuWritten(write_id);
+	auto& source      = cache.GetImage(read_id);
+	auto& destination = cache.GetImage(write_id);
+	EXIT_IF(read_id == write_id || source.backing.format != destination.backing.format);
+
+	auto& scheduler = m_context.GetCommandScheduler();
+	scheduler.EndRendering();
+	const auto command = scheduler.Current().Handle();
+	const ImageSubresourceRange range {read_desc.view_info.base_level, 1,
+	                                  read_desc.view_info.base_layer,
+	                                  read_desc.view_info.layer_count};
+	source.Transit(vk::ImageLayout::eTransferSrcOptimal, vk::AccessFlagBits2::eTransferRead,
+	               range, command);
+	destination.Transit(vk::ImageLayout::eTransferDstOptimal, vk::AccessFlagBits2::eTransferWrite,
+	                    range, command);
+	std::array<vk::ImageCopy, 2> regions {};
+	uint32_t                   count = 0;
+	for (const auto aspect: {vk::ImageAspectFlagBits::eDepth, vk::ImageAspectFlagBits::eStencil}) {
+		if ((aspect == vk::ImageAspectFlagBits::eDepth && !depth_copy) ||
+		    (aspect == vk::ImageAspectFlagBits::eStencil && !stencil_copy)) {
+			continue;
+		}
+		auto& region = regions[count++];
+		region.srcSubresource = {aspect, range.base_level, range.base_layer,
+		                         range.layer_count};
+		region.dstSubresource = region.srcSubresource;
+		region.extent = write_desc.info.extent;
+	}
+	command.copyImage(source.backing.image, vk::ImageLayout::eTransferSrcOptimal,
+	                  destination.backing.image, vk::ImageLayout::eTransferDstOptimal,
+	                  count, regions.data());
+	return true;
+}
+
 vk::ImageAspectFlags RenderDepthInfo::AttachmentWriteAspects() const {
+	const auto format = desc.view_info.format;
 	if (format == vk::Format::eUndefined) {
 		return {};
 	}
@@ -374,21 +422,20 @@ vk::ImageAspectFlags RenderDepthInfo::AttachmentWriteAspects() const {
 	const auto           available = ImageViewOps::DepthAspectMask(format);
 	vk::ImageAspectFlags writes {};
 	if ((available & vk::ImageAspectFlagBits::eDepth) &&
-	    (depth_load_clear_enable || (depth_test_enable && depth_write_enable))) {
+	    (depth_load_clear_enable || depth_write_enable)) {
 		writes |= vk::ImageAspectFlagBits::eDepth;
 	}
 	if (!(available & vk::ImageAspectFlagBits::eStencil)) {
 		return writes;
 	}
 
-	const auto face_writes = [&](const PipelineStencilStaticState&  state,
-	                             const PipelineStencilDynamicState& dynamic) {
-		if (dynamic.writeMask == 0) {
+	const auto face_writes = [&](const vk::StencilOpState& state) {
+		if (state.writeMask == 0) {
 			return false;
 		}
 		bool can_pass = state.compareOp != vk::CompareOp::eNever;
 		bool can_fail = state.compareOp != vk::CompareOp::eAlways;
-		if (dynamic.compareMask == 0) {
+		if (state.compareMask == 0) {
 			switch (state.compareOp) {
 				case vk::CompareOp::eEqual:
 				case vk::CompareOp::eLessOrEqual:
@@ -414,8 +461,7 @@ vk::ImageAspectFlags RenderDepthInfo::AttachmentWriteAspects() const {
 		       (can_pass && depth_fail && state.depthFailOp != vk::StencilOp::eKeep);
 	};
 	if (stencil_clear_enable ||
-	    (stencil_test_enable && (face_writes(stencil_static_front, stencil_dynamic_front) ||
-	                             face_writes(stencil_static_back, stencil_dynamic_back)))) {
+	    (stencil_test_enable && (face_writes(stencil_front) || face_writes(stencil_back)))) {
 		writes |= vk::ImageAspectFlagBits::eStencil;
 	}
 	return writes;

src/graphics/host_gpu/renderer/depthRenderTarget.h

@@ -11,32 +11,21 @@
 
 namespace Libs::Graphics {
 
+// Without a stencil plane, Hi-Stencil fields are inactive. An active plane is compatible when
+// Hi-Stencil is disabled or HTile backing is present.
 inline constexpr bool depth_htile_stencil_acceleration_compatible(bool has_stencil, bool has_htile,
                                                                   bool htile_stencil_disabled) {
-	return htile_stencil_disabled || (has_stencil && has_htile);
+	return !has_stencil || htile_stencil_disabled || has_htile;
 }
 
 struct RenderDepthInfo {
+	// Discovery keeps guest image information but can remap the view into a larger cache image.
 	TextureCache::ImageDesc     desc;
-	vk::Format                  format                   = vk::Format::eUndefined;
-	uint32_t                    width                    = 0;
-	uint32_t                    height                   = 0;
-	uint32_t                    samples                  = 1;
-	bool                        htile                    = false;
-	uint64_t                    depth_buffer_size        = 0;
-	uint64_t                    depth_buffer_vaddr       = 0;
-	uint64_t                    depth_tile_swizzle       = 0;
-	uint64_t                    stencil_buffer_size      = 0;
-	uint64_t                    stencil_buffer_vaddr     = 0;
-	uint64_t                    stencil_tile_swizzle     = 0;
-	uint64_t                    htile_buffer_size        = 0;
-	uint64_t                    htile_buffer_vaddr       = 0;
-	uint64_t                    htile_tile_swizzle       = 0;
 	bool                        depth_clear_enable       = false;
 	bool                        depth_load_clear_enable  = false;
-	bool                        depth_meta_clear_enable  = false;
 	float                       depth_clear_value        = 0.0f;
 	bool                        depth_test_enable        = false;
+	// Effective draw writes; discovery applies test, target-write and clear controls.
 	bool                        depth_write_enable       = false;
 	vk::CompareOp               depth_compare_op         = vk::CompareOp::eNever;
 	bool                        depth_bounds_test_enable = false;
@@ -45,29 +34,20 @@ struct RenderDepthInfo {
 	bool                        stencil_clear_enable     = false;
 	uint8_t                     stencil_clear_value      = 0;
 	bool                        stencil_test_enable      = false;
-	PipelineStencilStaticState  stencil_static_front;
-	PipelineStencilStaticState  stencil_static_back;
-	PipelineStencilDynamicState stencil_dynamic_front;
-	PipelineStencilDynamicState stencil_dynamic_back;
+	vk::StencilOpState          stencil_front;
+	vk::StencilOpState          stencil_back;
 	ImageId                     image_id;
-	vk::ImageView               image_view = nullptr;
-	uint64_t                    vaddr[3]   = {};
-	uint64_t                    size[3]    = {};
-	int                         vaddr_num  = 0;
 
 	[[nodiscard]] vk::ImageAspectFlags AttachmentWriteAspects() const;
 };
 
-inline bool depth_attachment_read_only(const RenderDepthInfo& depth) {
-	return !depth.AttachmentWriteAspects();
-}
-
 inline vk::ImageLayout depth_attachment_layout(const RenderDepthInfo& depth) {
-	const auto available     = ImageViewOps::DepthAspectMask(depth.format);
+	const auto available     = ImageViewOps::DepthAspectMask(depth.desc.view_info.format);
 	const auto writes        = depth.AttachmentWriteAspects();
 	const bool has_depth     = static_cast<bool>(available & vk::ImageAspectFlagBits::eDepth);
 	const bool has_stencil   = static_cast<bool>(available & vk::ImageAspectFlagBits::eStencil);
-	const bool depth_write   = static_cast<bool>(writes & vk::ImageAspectFlagBits::eDepth);
+	// LoadOp clears are separate from guest depth writes.
+	const bool depth_write   = depth.depth_write_enable;
 	const bool stencil_write = static_cast<bool>(writes & vk::ImageAspectFlagBits::eStencil);
 	if (!has_stencil) {
 		return depth_write ? vk::ImageLayout::eDepthAttachmentOptimal

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

@@ -10,7 +10,6 @@
 
 #include <algorithm>
 #include <array>
-#include <iterator>
 
 namespace Libs::Graphics {
 
@@ -23,7 +22,6 @@ BlitHelper::BlitHelper(GraphicContext& graphics, CommandScheduler& scheduler)
 	texture_binding.stageFlags      = vk::ShaderStageFlagBits::eFragment;
 
 	vk::DescriptorSetLayoutCreateInfo descriptor_info {};
-	descriptor_info.sType        = vk::StructureType::eDescriptorSetLayoutCreateInfo;
 	descriptor_info.flags        = vk::DescriptorSetLayoutCreateFlagBits::ePushDescriptorKHR;
 	descriptor_info.bindingCount = 1;
 	descriptor_info.pBindings    = &texture_binding;
@@ -32,16 +30,14 @@ BlitHelper::BlitHelper(GraphicContext& graphics, CommandScheduler& scheduler)
 	                     "create BlitHelper descriptor layout");
 
 	vk::PipelineLayoutCreateInfo layout_info {};
-	layout_info.sType          = vk::StructureType::ePipelineLayoutCreateInfo;
 	layout_info.setLayoutCount = 1;
 	layout_info.pSetLayouts    = &m_descriptor_layout;
 	RequireVulkanSuccess(
 	    m_graphics.device.createPipelineLayout(&layout_info, nullptr, &m_pipeline_layout),
 	    "create BlitHelper pipeline layout");
 
-	m_vertex_shader = CreateShader(GPU_BLIT_FS_TRIANGLE_SPV, std::size(GPU_BLIT_FS_TRIANGLE_SPV));
-	m_fragment_shader =
-	    CreateShader(GPU_BLIT_COLOR_TO_MS_DEPTH_SPV, std::size(GPU_BLIT_COLOR_TO_MS_DEPTH_SPV));
+	m_vertex_shader   = CompileSPV(GPU_BLIT_FS_TRIANGLE_SPV, m_graphics.device);
+	m_fragment_shader = CompileSPV(GPU_BLIT_COLOR_TO_MS_DEPTH_SPV, m_graphics.device);
 }
 
 BlitHelper::~BlitHelper() {
@@ -62,18 +58,6 @@ BlitHelper::~BlitHelper() {
 	}
 }
 
-vk::ShaderModule BlitHelper::CreateShader(const uint32_t* code, size_t words) const {
-	EXIT_IF(code == nullptr || words == 0);
-	vk::ShaderModuleCreateInfo create {};
-	create.sType            = vk::StructureType::eShaderModuleCreateInfo;
-	create.codeSize         = words * sizeof(uint32_t);
-	create.pCode            = code;
-	vk::ShaderModule module = nullptr;
-	RequireVulkanSuccess(m_graphics.device.createShaderModule(&create, nullptr, &module),
-	                     "create BlitHelper shader module");
-	return module;
-}
-
 vk::Pipeline BlitHelper::GetPipeline(PipelineKey key) {
 	const auto cached = std::ranges::find(m_pipelines, key, &Pipeline::key);
 	if (cached != m_pipelines.end()) {
@@ -84,51 +68,37 @@ vk::Pipeline BlitHelper::GetPipeline(PipelineKey key) {
 	EXIT_IF(samples == vk::SampleCountFlagBits {} || key.format == vk::Format::eUndefined);
 
 	std::array<vk::PipelineShaderStageCreateInfo, 2> stages {};
-	stages[0].sType  = vk::StructureType::ePipelineShaderStageCreateInfo;
 	stages[0].stage  = vk::ShaderStageFlagBits::eVertex;
 	stages[0].module = m_vertex_shader;
 	stages[0].pName  = "main";
-	stages[1].sType  = vk::StructureType::ePipelineShaderStageCreateInfo;
 	stages[1].stage  = vk::ShaderStageFlagBits::eFragment;
 	stages[1].module = m_fragment_shader;
 	stages[1].pName  = "main";
 
 	vk::PipelineVertexInputStateCreateInfo vertex_input {};
-	vertex_input.sType = vk::StructureType::ePipelineVertexInputStateCreateInfo;
 	vk::PipelineInputAssemblyStateCreateInfo input_assembly {};
-	input_assembly.sType    = vk::StructureType::ePipelineInputAssemblyStateCreateInfo;
 	input_assembly.topology = vk::PrimitiveTopology::eTriangleList;
 	vk::PipelineViewportStateCreateInfo viewport {};
-	viewport.sType         = vk::StructureType::ePipelineViewportStateCreateInfo;
 	viewport.viewportCount = 1;
 	viewport.scissorCount  = 1;
 	vk::PipelineRasterizationStateCreateInfo rasterization {};
-	rasterization.sType       = vk::StructureType::ePipelineRasterizationStateCreateInfo;
-	rasterization.polygonMode = vk::PolygonMode::eFill;
-	rasterization.cullMode    = vk::CullModeFlagBits::eNone;
-	rasterization.lineWidth   = 1.0f;
+	rasterization.lineWidth = 1.0f;
 	vk::PipelineMultisampleStateCreateInfo multisample {};
-	multisample.sType                = vk::StructureType::ePipelineMultisampleStateCreateInfo;
 	multisample.rasterizationSamples = samples;
 	vk::PipelineDepthStencilStateCreateInfo depth {};
-	depth.sType            = vk::StructureType::ePipelineDepthStencilStateCreateInfo;
 	depth.depthTestEnable  = VK_TRUE;
 	depth.depthWriteEnable = VK_TRUE;
 	depth.depthCompareOp   = vk::CompareOp::eAlways;
 	vk::PipelineColorBlendStateCreateInfo color_blend {};
-	color_blend.sType = vk::StructureType::ePipelineColorBlendStateCreateInfo;
 	const std::array dynamic_states {vk::DynamicState::eViewport, vk::DynamicState::eScissor};
 	vk::PipelineDynamicStateCreateInfo dynamic {};
-	dynamic.sType             = vk::StructureType::ePipelineDynamicStateCreateInfo;
 	dynamic.dynamicStateCount = static_cast<uint32_t>(dynamic_states.size());
 	dynamic.pDynamicStates    = dynamic_states.data();
 
 	vk::PipelineRenderingCreateInfo rendering {};
-	rendering.sType                 = vk::StructureType::ePipelineRenderingCreateInfo;
 	rendering.depthAttachmentFormat = key.format;
 
 	vk::GraphicsPipelineCreateInfo create {};
-	create.sType               = vk::StructureType::eGraphicsPipelineCreateInfo;
 	create.pNext               = &rendering;
 	create.stageCount          = static_cast<uint32_t>(stages.size());
 	create.pStages             = stages.data();
@@ -186,7 +156,6 @@ void BlitHelper::ReinterpretColorAsMsDepth(Image& source, Image& destination) {
 	                    command);
 
 	vk::RenderingAttachmentInfo depth_attachment {};
-	depth_attachment.sType                   = vk::StructureType::eRenderingAttachmentInfo;
 	depth_attachment.imageView               = destination_view;
 	depth_attachment.imageLayout             = ColorToMsDepthLayout;
 	depth_attachment.loadOp                  = vk::AttachmentLoadOp::eClear;
@@ -194,7 +163,6 @@ void BlitHelper::ReinterpretColorAsMsDepth(Image& source, Image& destination) {
 	depth_attachment.clearValue.depthStencil = {0.0f, 0};
 
 	vk::RenderingInfo rendering {};
-	rendering.sType             = vk::StructureType::eRenderingInfo;
 	rendering.renderArea.extent = {destination_info.extent.width, destination_info.extent.height};
 	rendering.layerCount        = 1;
 	rendering.pDepthAttachment  = &depth_attachment;
@@ -204,7 +172,6 @@ void BlitHelper::ReinterpretColorAsMsDepth(Image& source, Image& destination) {
 	descriptor_image.imageView   = source_view;
 	descriptor_image.imageLayout = vk::ImageLayout::eShaderReadOnlyOptimal;
 	vk::WriteDescriptorSet descriptor_write {};
-	descriptor_write.sType           = vk::StructureType::eWriteDescriptorSet;
 	descriptor_write.dstBinding      = 0;
 	descriptor_write.descriptorCount = 1;
 	descriptor_write.descriptorType  = vk::DescriptorType::eSampledImage;

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

@@ -37,8 +37,7 @@ private:
 		vk::Pipeline handle = nullptr;
 	};
 
-	[[nodiscard]] vk::ShaderModule CreateShader(const uint32_t* code, size_t words) const;
-	[[nodiscard]] vk::Pipeline     GetPipeline(PipelineKey key);
+	[[nodiscard]] vk::Pipeline GetPipeline(PipelineKey key);
 
 	GraphicContext&         m_graphics;
 	CommandScheduler&       m_scheduler;

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

@@ -26,12 +26,13 @@ namespace {
 	}
 }
 
-[[nodiscard]] vk::ImageCreateFlags ImageCreateFlags(const ImageInfo& info) {
+[[nodiscard]] vk::ImageCreateFlags ImageCreateFlags(const GraphicContext& graphics,
+                                                   const ImageInfo& info) {
 	vk::ImageCreateFlags flags {};
 	if (DepthAspectTransferFormat(info.pixel_format) == vk::Format::eUndefined) {
 		flags |= vk::ImageCreateFlagBits::eMutableFormat;
 		flags |= vk::ImageCreateFlagBits::eExtendedUsage;
-		if (Prospero::BlockCompressedBytesPerBlock(info.guest_format) != 0) {
+		if (info.IsBlock() && graphics.supports_block_texel_view) {
 			flags |= vk::ImageCreateFlagBits::eBlockTexelViewCompatible;
 		}
 	}
@@ -47,6 +48,10 @@ namespace {
 }
 
 [[nodiscard]] vk::ImageUsageFlags ImageUsageFlags(GraphicContext& graphics, const ImageInfo& info) {
+	if (info.IsBlock()) {
+		return vk::ImageUsageFlagBits::eTransferSrc | vk::ImageUsageFlagBits::eTransferDst |
+		       vk::ImageUsageFlagBits::eSampled;
+	}
 	const auto properties = graphics.GetFormatProperties(info.pixel_format);
 	auto       usage = vk::ImageUsageFlagBits::eTransferSrc | vk::ImageUsageFlagBits::eTransferDst;
 	if (HasFormatFeature(properties, vk::FormatFeatureFlagBits::eSampledImage)) {
@@ -54,29 +59,22 @@ namespace {
 	}
 	if (DepthAspectTransferFormat(info.pixel_format) != vk::Format::eUndefined) {
 		usage |= vk::ImageUsageFlagBits::eDepthStencilAttachment;
+		if (graphics.attachment_feedback_loop_enabled && (usage & vk::ImageUsageFlagBits::eSampled)) {
+			usage |= vk::ImageUsageFlagBits::eAttachmentFeedbackLoopEXT;
+		}
 		return usage;
 	}
 	if (HasFormatFeature(properties, vk::FormatFeatureFlagBits::eColorAttachment)) {
 		usage |= vk::ImageUsageFlagBits::eColorAttachment;
 	}
-	if (info.samples == 1 &&
-	    HasFormatFeature(properties, vk::FormatFeatureFlagBits::eStorageImage)) {
+	if (info.samples == 1) {
 		usage |= vk::ImageUsageFlagBits::eStorage;
-	} else if (info.samples == 1) {
-		const auto compatible = SrgbStorageViewFormat(info.pixel_format);
-		if (compatible != vk::Format::eUndefined &&
-		    HasFormatFeature(graphics.GetFormatProperties(compatible),
-		                     vk::FormatFeatureFlagBits::eStorageImage)) {
-			usage |= vk::ImageUsageFlagBits::eStorage;
-		}
 	}
 	return usage;
 }
 
-void ValidateRange(GuestRange range, const char* name) {
-	if ((range.address == 0) != (range.size == 0) ||
-	    (range.address != 0 && (range.address >= TRACKER_ADDRESS_SIZE ||
-	                            range.size > TRACKER_ADDRESS_SIZE - range.address))) {
+void ValidateOptionalRange(GuestRange range, const char* name) {
+	if (!range.ValidOrEmpty()) {
 		EXIT("invalid %s image range: address=0x%016llx size=0x%016llx\n", name,
 		     static_cast<unsigned long long>(range.address),
 		     static_cast<unsigned long long>(range.size));
@@ -212,9 +210,7 @@ void Image::Transit(vk::ImageLayout destination_layout, vk::AccessFlags2 destina
 	if (barriers.empty()) {
 		return;
 	}
-	if (m_scheduler != nullptr) {
-		m_scheduler->EndRendering();
-	}
+	m_scheduler.EndRendering();
 	vk::DependencyInfo dependency {};
 	dependency.imageMemoryBarrierCount = static_cast<uint32_t>(barriers.size());
 	dependency.pImageMemoryBarriers    = barriers.data();
@@ -223,8 +219,8 @@ void Image::Transit(vk::ImageLayout destination_layout, vk::AccessFlags2 destina
 
 void Image::Upload(std::span<const vk::BufferImageCopy> copies, vk::Buffer buffer, uint64_t offset,
                    uint64_t size) {
-	EXIT_IF(m_scheduler == nullptr || copies.empty() || buffer == nullptr || size == 0);
-	m_scheduler->EndRendering();
+	EXIT_IF(copies.empty() || buffer == nullptr || size == 0);
+	m_scheduler.EndRendering();
 	vk::BufferMemoryBarrier2 buffer_barrier {};
 	buffer_barrier.srcStageMask        = vk::PipelineStageFlagBits2::eAllCommands;
 	buffer_barrier.srcAccessMask       = vk::AccessFlagBits2::eMemoryWrite;
@@ -244,7 +240,7 @@ void Image::Upload(std::span<const vk::BufferImageCopy> copies, vk::Buffer buffe
 	dependency.pBufferMemoryBarriers    = &buffer_barrier;
 	dependency.imageMemoryBarrierCount  = static_cast<uint32_t>(image_barriers.size());
 	dependency.pImageMemoryBarriers     = image_barriers.data();
-	auto command                        = m_scheduler->Current().Handle();
+	auto command                        = m_scheduler.Current().Handle();
 	command.pipelineBarrier2(dependency);
 	command.copyBufferToImage(buffer, backing.image, vk::ImageLayout::eTransferDstOptimal,
 	                          static_cast<uint32_t>(copies.size()), copies.data());
@@ -262,8 +258,8 @@ void Image::Upload(std::span<const vk::BufferImageCopy> copies, vk::Buffer buffe
 
 void Image::Download(std::span<const vk::BufferImageCopy> copies, vk::Buffer buffer,
                      uint64_t offset, uint64_t size) {
-	EXIT_IF(m_scheduler == nullptr || copies.empty() || buffer == nullptr || size == 0);
-	m_scheduler->EndRendering();
+	EXIT_IF(copies.empty() || buffer == nullptr || size == 0);
+	m_scheduler.EndRendering();
 	vk::BufferMemoryBarrier2 buffer_barrier {};
 	buffer_barrier.srcStageMask = vk::PipelineStageFlagBits2::eAllCommands;
 	buffer_barrier.srcAccessMask =
@@ -284,7 +280,7 @@ void Image::Download(std::span<const vk::BufferImageCopy> copies, vk::Buffer buf
 	dependency.pBufferMemoryBarriers    = &buffer_barrier;
 	dependency.imageMemoryBarrierCount  = static_cast<uint32_t>(image_barriers.size());
 	dependency.pImageMemoryBarriers     = image_barriers.data();
-	auto command                        = m_scheduler->Current().Handle();
+	auto command                        = m_scheduler.Current().Handle();
 	command.pipelineBarrier2(dependency);
 	command.copyImageToBuffer(backing.image, vk::ImageLayout::eTransferSrcOptimal, buffer,
 	                          static_cast<uint32_t>(copies.size()), copies.data());
@@ -321,8 +317,8 @@ std::pair<uint32_t, uint32_t> Image::SanitizeCopyLayers(const Image& source,
 }
 
 void Image::CopyImage(Image& source) {
-	EXIT_IF(m_scheduler == nullptr || source.backing.samples != backing.samples);
-	m_scheduler->EndRendering();
+	EXIT_IF(source.backing.samples != backing.samples);
+	m_scheduler.EndRendering();
 	const uint32_t levels     = std::min(source.backing.mip_levels, backing.mip_levels);
 	const uint32_t base_depth = backing.image_type == vk::ImageType::e3D
 	                                ? backing.extent.depth
@@ -361,7 +357,7 @@ void Image::CopyImage(Image& source) {
 	if (copies.empty()) {
 		return;
 	}
-	auto command = m_scheduler->Current().Handle();
+	auto command = m_scheduler.Current().Handle();
 	source.Transit(vk::ImageLayout::eTransferSrcOptimal, vk::AccessFlagBits2::eTransferRead, {},
 	               command);
 	Transit(vk::ImageLayout::eTransferDstOptimal, vk::AccessFlagBits2::eTransferWrite, {}, command);
@@ -374,8 +370,7 @@ void Image::CopyImage(Image& source) {
 
 void Image::Resolve(Image& source, const ImageSubresourceRange& source_range,
                     const ImageSubresourceRange& destination_range) {
-	EXIT_IF(m_scheduler == nullptr || backing.samples != 1 ||
-	        source.backing.image_type != vk::ImageType::e2D ||
+	EXIT_IF(backing.samples != 1 || source.backing.image_type != vk::ImageType::e2D ||
 	        backing.image_type != vk::ImageType::e2D || source_range.level_count != 1 ||
 	        destination_range.level_count != 1 ||
 	        source_range.base_level >= source.backing.mip_levels ||
@@ -403,8 +398,8 @@ void Image::Resolve(Image& source, const ImageSubresourceRange& source_range,
 	resolved_destination_range.layer_count = layers;
 	const vk::Extent3D resolve_extent {info.extent.width, info.extent.height, 1};
 
-	m_scheduler->EndRendering();
-	auto command = m_scheduler->Current().Handle();
+	m_scheduler.EndRendering();
+	auto command = m_scheduler.Current().Handle();
 	source.Transit(vk::ImageLayout::eTransferSrcOptimal, vk::AccessFlagBits2::eTransferRead,
 	               resolved_source_range, command);
 	Transit(vk::ImageLayout::eTransferDstOptimal, vk::AccessFlagBits2::eTransferWrite,
@@ -440,9 +435,8 @@ uint32_t Image::CopyRows(uint64_t row_size, uint32_t rows, uint64_t capacity) no
 }
 
 void Image::CopyImageWithBuffer(Image& source, Buffer& buffer) {
-	EXIT_IF(m_scheduler == nullptr || buffer.Handle() == nullptr || source.backing.samples != 1 ||
-	        backing.samples != 1);
-	m_scheduler->EndRendering();
+	EXIT_IF(buffer.Handle() == nullptr || source.backing.samples != 1 || backing.samples != 1);
+	m_scheduler.EndRendering();
 	const uint32_t levels = std::min(source.backing.mip_levels, backing.mip_levels);
 	const auto     source_aspect =
 	    FullAspectMask(source.backing.format) & ~vk::ImageAspectFlagBits::eStencil;
@@ -472,7 +466,7 @@ void Image::CopyImageWithBuffer(Image& source, Buffer& buffer) {
 	dependency.dependencyFlags          = vk::DependencyFlagBits::eByRegion;
 	dependency.bufferMemoryBarrierCount = 1;
 	dependency.pBufferMemoryBarriers    = &barrier;
-	auto command                        = m_scheduler->Current().Handle();
+	auto command                        = m_scheduler.Current().Handle();
 	source.Transit(vk::ImageLayout::eTransferSrcOptimal, vk::AccessFlagBits2::eTransferRead, {},
 	               command);
 	Transit(vk::ImageLayout::eTransferDstOptimal, vk::AccessFlagBits2::eTransferWrite, {}, command);
@@ -532,9 +526,9 @@ void Image::CopyImageWithBuffer(Image& source, Buffer& buffer) {
 }
 
 void Image::CopyMip(Image& source, uint32_t mip, uint32_t layer) {
-	EXIT_IF(m_scheduler == nullptr || source.backing.samples != backing.samples ||
-	        mip >= backing.mip_levels || layer >= backing.layers);
-	m_scheduler->EndRendering();
+	EXIT_IF(source.backing.samples != backing.samples || mip >= backing.mip_levels ||
+	        layer >= backing.layers);
+	m_scheduler.EndRendering();
 	const auto width  = std::max(backing.extent.width >> mip, 1u);
 	const auto height = std::max(backing.extent.height >> mip, 1u);
 	const auto depth  = std::max(backing.extent.depth >> mip, 1u);
@@ -554,7 +548,7 @@ void Image::CopyMip(Image& source, uint32_t mip, uint32_t layer) {
 		copy.dstSubresource = {aspect, mip, layer, destination_layers};
 		copy.extent         = {width, height, depth};
 	}
-	auto command = m_scheduler->Current().Handle();
+	auto command = m_scheduler.Current().Handle();
 	Transit(vk::ImageLayout::eTransferDstOptimal, vk::AccessFlagBits2::eTransferWrite, {}, command);
 	source.Transit(vk::ImageLayout::eTransferSrcOptimal, vk::AccessFlagBits2::eTransferRead, {},
 	               command);
@@ -567,13 +561,13 @@ void Image::CopyMip(Image& source, uint32_t mip, uint32_t layer) {
 namespace ImageOps {
 
 void Validate(const ImageInfo& info) {
-	ValidateRange(info.data, "data");
-	ValidateRange(info.stencil, "stencil");
+	ValidateOptionalRange(info.data, "data");
+	ValidateOptionalRange(info.stencil, "stencil");
 
 	if (info.pixel_format == vk::Format::eUndefined) {
 		const bool metadata_empty =
-		    info.metadata.range.address == 0 && info.metadata.range.size == 0 &&
-		    info.metadata.kind == ImageMetadataKind::None && info.metadata.control == 0 &&
+		    info.metadata.range.Empty() && info.metadata.kind == ImageMetadataKind::None &&
+		    info.metadata.control == 0 &&
 		    info.metadata.compression == VideoOutCompression::Uncompressed &&
 		    !info.metadata.stencil_compressed;
 		if (info.data.Empty() || info.HasStencil() || !metadata_empty || info.extent.width == 0 ||
@@ -620,16 +614,14 @@ void Validate(const ImageInfo& info) {
 	}
 	switch (info.metadata.kind) {
 		case ImageMetadataKind::None:
-			if (info.metadata.range.address != 0 || info.metadata.range.size != 0 ||
-			    info.metadata.control != 0 ||
+			if (!info.metadata.range.Empty() || info.metadata.control != 0 ||
 			    info.metadata.compression != VideoOutCompression::Uncompressed ||
 			    info.metadata.stencil_compressed) {
 				EXIT("metadata-free image has metadata state\n");
 			}
 			break;
 		case ImageMetadataKind::Htile:
-			ValidateRange(info.metadata.range, "HTILE");
-			if (info.metadata.range.Empty() ||
+			if (!info.metadata.range.Valid() ||
 			    info.metadata.compression != VideoOutCompression::Uncompressed) {
 				EXIT("invalid HTILE metadata\n");
 			}
@@ -660,7 +652,7 @@ Prospero::BufferFormat RenderTargetTransferFormat(uint32_t bytes_per_element) {
 } // namespace ImageOps
 
 Image::Image(GraphicContext& graphics, CommandScheduler& scheduler, const ImageInfo& image_info)
-    : info(image_info), m_graphics(&graphics), m_scheduler(&scheduler) {
+    : info(image_info), m_graphics(graphics), m_scheduler(scheduler) {
 	KYTY_PROFILER_FUNCTION();
 	ImageOps::Validate(info);
 	m_cpu_dirty =
@@ -669,29 +661,17 @@ Image::Image(GraphicContext& graphics, CommandScheduler& scheduler, const ImageI
 		return;
 	}
 
-	backing.format      = info.pixel_format;
-	backing.image_type  = HostImageType(info.type);
-	backing.extent      = info.extent;
-	backing.guest_pitch = info.pitch;
-	backing.layers      = info.IsVolume() ? 1u : info.resources.layers;
-	backing.mip_levels  = info.resources.levels;
-	backing.samples     = info.samples;
-	backing.flags       = ImageCreateFlags(info);
-	backing.usage       = ImageUsageFlags(graphics, info);
-
 	vk::ImageCreateInfo create {};
-	create.sType         = vk::StructureType::eImageCreateInfo;
-	create.flags         = backing.flags;
-	create.imageType     = backing.image_type;
-	create.extent        = backing.extent;
-	create.mipLevels     = backing.mip_levels;
-	create.arrayLayers   = backing.layers;
-	create.format        = backing.format;
+	create.flags         = ImageCreateFlags(graphics, info);
+	create.imageType     = HostImageType(info.type);
+	create.extent        = info.extent;
+	create.mipLevels     = info.resources.levels;
+	create.arrayLayers   = info.IsVolume() ? 1u : info.resources.layers;
+	create.format        = info.pixel_format;
 	create.tiling        = vk::ImageTiling::eOptimal;
-	create.initialLayout = backing.state.layout;
-	create.usage         = backing.usage;
-	create.sharingMode   = vk::SharingMode::eExclusive;
-	create.samples       = vulkan_sample_count(backing.samples);
+	create.initialLayout = vk::ImageLayout::eUndefined;
+	create.usage         = ImageUsageFlags(graphics, info);
+	create.samples       = vulkan_sample_count(info.samples);
 
 	vk::ImageFormatProperties properties {};
 	if (graphics.GetImageFormatProperties(create.format, create.imageType, create.tiling,
@@ -702,10 +682,9 @@ Image::Image(GraphicContext& graphics, CommandScheduler& scheduler, const ImageI
 		     "flags=0x%x samples=%u\n",
 		     static_cast<int>(create.format), static_cast<int>(create.imageType),
 		     static_cast<vk::ImageUsageFlags::MaskType>(create.usage),
-		     static_cast<vk::ImageCreateFlags::MaskType>(create.flags), backing.samples);
+		     static_cast<vk::ImageCreateFlags::MaskType>(create.flags), info.samples);
 	}
 
-	backing.memory.property = vk::MemoryPropertyFlagBits::eDeviceLocal;
 	if (!graphics.CreateImage(create, backing)) {
 		EXIT("failed to create image: extent=%ux%ux%u format=%d layers=%u levels=%u\n",
 		     create.extent.width, create.extent.height, create.extent.depth,
@@ -714,11 +693,12 @@ Image::Image(GraphicContext& graphics, CommandScheduler& scheduler, const ImageI
 }
 
 uint64_t Image::HashGuestEdges() const {
-	constexpr uint64_t                         page_mask = TRACKER_PAGE_SIZE - 1;
 	std::array<uint8_t, TRACKER_PAGE_SIZE * 2> bytes {};
 	const auto                                 range = info.data;
-	const uint64_t head_end     = std::min(range.End(), (range.address + page_mask) & ~page_mask);
-	const uint64_t tail_begin   = std::max(range.address, range.End() & ~page_mask);
+	const uint64_t head_end =
+	    std::min(range.End(), Common::AlignUp(range.address, TRACKER_PAGE_SIZE));
+	const uint64_t tail_begin =
+	    std::max(range.address, Common::AlignDown(range.End(), TRACKER_PAGE_SIZE));
 	const uint64_t head_size    = head_end - range.address;
 	const uint64_t tail_address = tail_begin < head_end ? head_end : tail_begin;
 	const uint64_t tail_size    = range.End() - tail_address;
@@ -733,16 +713,13 @@ uint64_t Image::HashGuestEdges() const {
 
 Image::~Image() {
 	KYTY_PROFILER_FUNCTION();
-	if (m_graphics == nullptr) {
-		return;
-	}
 	for (const auto& cached: views) {
 		if (cached.view != nullptr) {
-			m_graphics->device.destroyImageView(cached.view, nullptr);
+			m_graphics.device.destroyImageView(cached.view, nullptr);
 		}
 	}
 	if (backing.image != nullptr) {
-		m_graphics->DeleteImage(backing);
+		m_graphics.DeleteImage(backing);
 	}
 }
 

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

@@ -1,6 +1,7 @@
 #ifndef EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_IMAGE_H_
 #define EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_IMAGE_H_
 
+#include "common/alignment.h"
 #include "common/assert.h"
 #include "common/slotVector.h"
 #include "graphics/host_gpu/graphicContext.h"
@@ -35,11 +36,13 @@ struct ImageUsage {
 };
 
 struct ImageBinding {
-	bool is_bound      = false;
-	bool is_target     = false;
-	bool needs_rebind  = false;
-	bool force_general = false;
-	bool shader_write  = false;
+	vk::ImageLayout  attachment_layout = vk::ImageLayout::eUndefined;
+	vk::AccessFlags2 attachment_access;
+	bool             is_bound      = false;
+	bool             is_target     = false;
+	bool             needs_rebind  = false;
+	bool             force_general = false;
+	bool             shader_write  = false;
 };
 
 class Image final {
@@ -49,7 +52,6 @@ public:
 	KYTY_CLASS_NO_COPY(Image);
 
 	[[nodiscard]] vk::ImageView FindView(const ImageViewInfo& view_info);
-	void                        AssociateDepth(ImageId image_id) { depth_id = image_id; }
 	using Barriers = std::vector<vk::ImageMemoryBarrier2>;
 	[[nodiscard]] Barriers GetBarriers(vk::ImageLayout                      destination_layout,
 	                                   vk::AccessFlags2                     destination_access,
@@ -127,15 +129,12 @@ public:
 		return pages ? ImagePageRangesOverlap(info.data.address, info.data.size, address, size)
 		             : ImageRangeOverlaps(info.data.address, info.data.size, address, size);
 	}
-	[[nodiscard]] bool GpuOverlaps(uint64_t address, uint64_t size) const noexcept {
-		return IsGpuModified() && Overlaps(address, size);
-	}
 	[[nodiscard]] bool SafeToDownload() const noexcept {
 		return IsGpuModified() && !IsBufferModified() && !IsCpuDirty();
 	}
 	[[nodiscard]] bool IsTracked() const noexcept { return track_addr != 0 && track_addr_end != 0; }
 	[[nodiscard]] uint64_t AccountedSize() const noexcept {
-		return backing.image == nullptr ? 0 : (info.data.size + 1023) & ~uint64_t {1023};
+		return backing.image == nullptr ? 0 : Common::AlignUp(info.data.size, 1024);
 	}
 	[[nodiscard]] uint64_t HashGuestEdges() const;
 
@@ -161,8 +160,8 @@ private:
 	[[nodiscard]] static std::pair<uint32_t, uint32_t>
 	SanitizeCopyLayers(const Image& source, const Image& destination, uint32_t depth);
 
-	GraphicContext*   m_graphics         = nullptr;
-	CommandScheduler* m_scheduler        = nullptr;
+	GraphicContext&   m_graphics;
+	CommandScheduler& m_scheduler;
 	uint64_t          m_maybe_cpu_hash   = 0;
 	bool              m_cpu_dirty        = false;
 	bool              m_maybe_cpu_dirty  = false;

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

@@ -15,18 +15,6 @@
 
 namespace Libs::Graphics {
 
-struct GuestRange {
-	uint64_t address = 0;
-	uint64_t size    = 0;
-
-	[[nodiscard]] constexpr bool Empty() const noexcept { return address == 0 || size == 0; }
-	[[nodiscard]] constexpr bool Valid() const noexcept {
-		return !Empty() && address < TRACKER_ADDRESS_SIZE && size <= TRACKER_ADDRESS_SIZE - address;
-	}
-	[[nodiscard]] constexpr uint64_t End() const noexcept { return address + size; }
-	auto                             operator<=>(const GuestRange&) const = default;
-};
-
 enum class VideoOutCompression : uint8_t { Uncompressed, Dcc256_256_0, Dcc256_64_64, Unsupported };
 
 enum class ImageMetadataKind : uint8_t { None, Htile, Dcc };
@@ -35,8 +23,11 @@ struct ImageMetadataInfo {
 	GuestRange          range;
 	ImageMetadataKind   kind               = ImageMetadataKind::None;
 	uint32_t            control            = 0;
+	uint32_t            dcc_clear_word           = 0;
 	VideoOutCompression compression        = VideoOutCompression::Uncompressed;
 	bool                stencil_compressed = false;
+	bool                dcc_clear_register_valid = false;
+	bool                dcc_alpha_msb            = true;
 };
 
 struct ImageSubresources {
@@ -56,6 +47,7 @@ struct ImageSubresourceRange {
 struct ImageMipInfo {
 	uint64_t offset                                 = 0;
 	uint64_t size                                   = 0;
+	// Padded dimensions in storage elements (compressed blocks for BC formats).
 	uint32_t pitch                                  = 0;
 	uint32_t height                                 = 0;
 	auto     operator<=>(const ImageMipInfo&) const = default;
@@ -90,9 +82,6 @@ struct ImageInfo {
 	[[nodiscard]] constexpr bool IsVolume() const noexcept {
 		return type == Prospero::ImageType::kColor3D;
 	}
-	[[nodiscard]] constexpr bool IsLayered() const noexcept {
-		return !IsVolume() && resources.layers > 1;
-	}
 	[[nodiscard]] constexpr uint32_t TransferLayers() const noexcept {
 		return IsVolume() ? extent.depth : resources.layers;
 	}
@@ -105,9 +94,24 @@ struct ImageInfo {
 		       bytes_per_block == other.bytes_per_block;
 	}
 	[[nodiscard]] int32_t MipOf(const ImageInfo& container) const noexcept {
-		if (!IsCompatible(container) || tile_mode != container.tile_mode || resources.levels != 1 ||
-		    container.resources.layers == 0 ||
-		    container.resources.levels > container.mip_layout.size()) {
+		if (container.resources.levels > container.mip_layout.size()) {
+			return -1;
+		}
+		for (uint32_t level = 0; level < container.resources.levels; level++) {
+			if (SliceOf(container, static_cast<int32_t>(level)) >= 0) {
+				return static_cast<int32_t>(level);
+			}
+		}
+		return -1;
+	}
+	[[nodiscard]] int32_t SliceOf(const ImageInfo& container, int32_t mip) const noexcept {
+		if (!IsCompatible(container) || tile_mode != container.tile_mode || type != container.type ||
+		    IsVolume() || resources.levels != 1 || resources.layers == 0 ||
+		    container.resources.layers == 0 || mip < 0 ||
+		    static_cast<uint32_t>(mip) >= container.resources.levels ||
+		    container.resources.levels > container.mip_layout.size() ||
+		    !data.Valid() || !container.data.Valid() || data.address < container.data.address ||
+		    data.End() > container.data.End()) {
 			return -1;
 		}
 		if (HasStencil() != container.HasStencil() ||
@@ -116,75 +120,36 @@ struct ImageInfo {
 			return -1;
 		}
 
-		int32_t mip = -1;
-		for (uint32_t level = 0; level < container.resources.levels; level++) {
-			const auto& layout = container.mip_layout[level];
-			if (layout.size == 0 || layout.size % container.resources.layers != 0 ||
-			    container.data.address > UINT64_MAX - layout.offset) {
-				continue;
-			}
-			const auto mip_base   = container.data.address + layout.offset;
-			const auto slice_size = layout.size / container.resources.layers;
-			if (slice_size == 0 || mip_base > UINT64_MAX - layout.size) {
-				continue;
-			}
-			const auto mip_end = mip_base + layout.size;
-			if (data.address >= mip_base && data.address < mip_end &&
-			    (data.address - mip_base) % slice_size == 0) {
-				mip = static_cast<int32_t>(level);
-				break;
-			}
-		}
-		if (mip < 0) {
+		const auto  level  = static_cast<uint32_t>(mip);
+		const auto& layout = container.mip_layout[level];
+		if (extent.width != std::max(container.extent.width >> level, 1u) ||
+		    extent.height != std::max(container.extent.height >> level, 1u) ||
+		    extent.depth != container.extent.depth || mip_layout[0].pitch != layout.pitch ||
+		    mip_layout[0].height != layout.height || layout.size == 0 ||
+		    layout.size % container.resources.layers != 0 ||
+		    container.data.size % container.resources.layers != 0 ||
+		    data.size % resources.layers != 0) {
 			return -1;
 		}
 
-		const auto level = static_cast<uint32_t>(mip);
-		if (extent.width != std::max(container.extent.width >> level, 1u) ||
-		    extent.height != std::max(container.extent.height >> level, 1u)) {
+		// PS5 block slices contain the entire reversed mip chain.
+		const auto slice_stride = container.data.size / container.resources.layers;
+		const auto child_stride = data.size / resources.layers;
+		if (child_stride != layout.size / container.resources.layers ||
+		    layout.offset >= slice_stride || child_stride > slice_stride - layout.offset ||
+		    (resources.layers > 1 && child_stride != slice_stride)) {
 			return -1;
 		}
-		const auto mip_depth = std::max(container.extent.depth >> level, 1u);
-		if (container.type == Prospero::ImageType::kColor3D &&
-		    type == Prospero::ImageType::kColor2D) {
-			if (resources.layers != mip_depth) {
-				return -1;
-			}
-		} else if (type != container.type) {
+		const auto address_delta = data.address - container.data.address;
+		if (address_delta < layout.offset || (address_delta - layout.offset) % slice_stride != 0) {
 			return -1;
 		}
-		return mip;
-	}
-	[[nodiscard]] int32_t SliceOf(const ImageInfo& container, int32_t mip) const noexcept {
-		if (!IsCompatible(container) || type != container.type || mip < 0 ||
-		    static_cast<uint32_t>(mip) >= container.resources.levels ||
-		    container.resources.levels > container.mip_layout.size() ||
-		    container.resources.layers == 0 || data.size == 0) {
+		const auto layer = (address_delta - layout.offset) / slice_stride;
+		if (layer > INT32_MAX || layer >= container.resources.layers ||
+		    resources.layers > container.resources.layers - layer) {
 			return -1;
 		}
-		const auto level = static_cast<uint32_t>(mip);
-		if (extent.width != std::max(container.extent.width >> level, 1u) ||
-		    extent.height != std::max(container.extent.height >> level, 1u)) {
-			return -1;
-		}
-		const auto& layout = container.mip_layout[level];
-		if (layout.size == 0 || layout.size % container.resources.layers != 0 ||
-		    container.data.address > UINT64_MAX - layout.offset) {
-			return -1;
-		}
-		const auto slice_size = layout.size / container.resources.layers;
-		if (slice_size == 0 || data.size % slice_size != 0) {
-			return -1;
-		}
-		const auto mip_base = container.data.address + layout.offset;
-		if (data.address < mip_base) {
-			return -1;
-		}
-		const auto address_delta = data.address - mip_base;
-		if (address_delta % data.size != 0 || address_delta / data.size > INT32_MAX) {
-			return -1;
-		}
-		return static_cast<int32_t>(address_delta / data.size);
+		return static_cast<int32_t>(layer);
 	}
 };
 
@@ -194,6 +159,7 @@ struct ImageViewInfo {
 	vk::ImageAspectFlags aspect      = vk::ImageAspectFlagBits::eColor;
 	uint32_t             base_level  = 0;
 	uint32_t             level_count = 1;
+	uint32_t             min_lod     = 0; // U4.8 clamp relative to base_level.
 	uint32_t             base_layer  = 0;
 	uint32_t             layer_count = 1;
 	vk::ComponentMapping mapping     = {};
@@ -202,9 +168,10 @@ struct ImageViewInfo {
 	[[nodiscard]] bool operator==(const ImageViewInfo& rhs) const noexcept {
 		return format == rhs.format && type == rhs.type && aspect == rhs.aspect &&
 		       base_level == rhs.base_level && level_count == rhs.level_count &&
-		       base_layer == rhs.base_layer && layer_count == rhs.layer_count &&
-		       mapping.r == rhs.mapping.r && mapping.g == rhs.mapping.g &&
-		       mapping.b == rhs.mapping.b && mapping.a == rhs.mapping.a && usage == rhs.usage;
+		       min_lod == rhs.min_lod && base_layer == rhs.base_layer &&
+		       layer_count == rhs.layer_count && mapping.r == rhs.mapping.r &&
+		       mapping.g == rhs.mapping.g && mapping.b == rhs.mapping.b &&
+		       mapping.a == rhs.mapping.a && usage == rhs.usage;
 	}
 };
 
@@ -350,14 +317,6 @@ ClassifyVideoOutCompression(bool compressed, uint64_t metadata_address, uint32_t
 	}
 }
 
-[[nodiscard]] inline constexpr bool
-CanUseVideoOutNativeWithoutUpload(VideoOutCompression compression, bool render_target,
-                                  bool gpu_modified, bool guest_modified) noexcept {
-	return compression != VideoOutCompression::Uncompressed &&
-	       compression != VideoOutCompression::Unsupported && !guest_modified &&
-	       (render_target || gpu_modified);
-}
-
 struct VideoOutPixelFormatInfo {
 	vk::Format             format            = vk::Format::eUndefined;
 	Prospero::BufferFormat guest_format      = Prospero::BufferFormat::kInvalid;
@@ -408,32 +367,6 @@ inline constexpr std::array<VideoOutFormatPolicy, 6> VIDEO_OUT_FORMAT_POLICIES {
 	return false;
 }
 
-[[nodiscard]] inline constexpr bool
-IsSupportedDisplayRenderTargetTileMode(Prospero::TileMode tile_mode) noexcept {
-	return tile_mode == Prospero::TileMode::kRenderTarget;
-}
-
-[[nodiscard]] inline constexpr bool IsSupportedStandard64RenderTarget(const ImageInfo& info) {
-	if (info.tile_mode != Prospero::TileMode::kStandard64KB || info.data.address == 0 ||
-	    (info.data.address & 0xffffu) != 0 || info.extent.width == 0 || info.extent.height == 0 ||
-	    info.bytes_per_block != 4 || info.resources.levels != 1 || info.resources.layers != 1 ||
-	    info.samples != 1) {
-		return false;
-	}
-	const auto expected_pitch =
-	    (static_cast<uint64_t>(info.extent.width) + 127u) & ~uint64_t {127u};
-	const auto padded_height =
-	    (static_cast<uint64_t>(info.extent.height) + 127u) & ~uint64_t {127u};
-	return expected_pitch <= UINT32_MAX && info.pitch == expected_pitch &&
-	       expected_pitch <= UINT64_MAX / padded_height / info.bytes_per_block &&
-	       info.data.size == expected_pitch * padded_height * info.bytes_per_block;
-}
-
-[[nodiscard]] inline constexpr bool IsTiledRenderTarget(const ImageInfo& info) noexcept {
-	return info.tile_mode == Prospero::TileMode::kRenderTarget ||
-	       IsSupportedStandard64RenderTarget(info);
-}
-
 [[nodiscard]] inline bool DecodePackedColorClear(vk::Format format, uint32_t packed,
                                                  vk::ClearColorValue& clear) {
 	vk::ClearColorValue next {};
@@ -443,6 +376,10 @@ IsSupportedDisplayRenderTargetTileMode(Prospero::TileMode tile_mode) noexcept {
 		return encoded <= 0.04045f ? encoded / 12.92f : std::pow((encoded + 0.055f) / 1.055f, 2.4f);
 	};
 	switch (format) {
+		// A single-plane float target carries its clear as raw float bits, the same encoding the
+		// depth decoder below uses. Without this the clear is discarded and the target keeps stale
+		// contents.
+		case vk::Format::eR32Sfloat: next.float32[0] = std::bit_cast<float>(packed); break;
 		case vk::Format::eR32Uint: next.uint32[0] = packed; break;
 		case vk::Format::eR32Sint: next.int32[0] = static_cast<int32_t>(packed); break;
 		case vk::Format::eR8G8B8A8Srgb:

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

@@ -21,10 +21,6 @@ namespace {
 	}
 }
 
-[[nodiscard]] bool IsCompatibleViewFormat(vk::Format image_format, vk::Format view_format) {
-	return ImageViewOps::FormatsCompatible(image_format, view_format);
-}
-
 [[nodiscard]] bool IsStencilViewFormat(vk::Format format) {
 	switch (format) {
 		case vk::Format::eS8Uint:
@@ -326,8 +322,14 @@ vk::ImageView Image::FindView(const ImageViewInfo& view_info) {
 		normalized.aspect = vk::ImageAspectFlagBits::eStencil;
 	}
 	normalized.usage = is_storage ? vk::ImageUsageFlagBits::eStorage : vk::ImageUsageFlags {};
+	for (const auto& cached: views) {
+		if (cached.info == normalized) {
+			return cached.view;
+		}
+	}
+
 	const bool format_compatible = normalized.format != vk::Format::eUndefined &&
-	                               IsCompatibleViewFormat(image.format, normalized.format);
+	                               ImageViewOps::FormatsCompatible(image.format, normalized.format);
 	const bool slice_view =
 	    image.image_type == vk::ImageType::e3D && (normalized.type == vk::ImageViewType::e2D ||
 	                                               normalized.type == vk::ImageViewType::e2DArray);
@@ -355,20 +357,18 @@ vk::ImageView Image::FindView(const ImageViewInfo& view_info) {
 		     image.layers);
 	}
 
-	for (const auto& cached: views) {
-		if (cached.info == normalized) {
-			return cached.view;
-		}
-	}
-
 	vk::ImageViewUsageCreateInfo usage {};
-	usage.sType = vk::StructureType::eImageViewUsageCreateInfo;
 	usage.usage = image.usage;
 	if (!is_storage) {
 		usage.usage &= ~vk::ImageUsageFlagBits::eStorage;
 	}
+	vk::ImageViewMinLodCreateInfoEXT min_lod {};
+	if (normalized.min_lod != 0) {
+		min_lod.minLod = static_cast<float>(normalized.base_level) +
+		                 static_cast<float>(normalized.min_lod) / 256.0f;
+		usage.pNext    = &min_lod;
+	}
 	vk::ImageViewCreateInfo create {};
-	create.sType                           = vk::StructureType::eImageViewCreateInfo;
 	create.pNext                           = &usage;
 	create.image                           = image.image;
 	create.viewType                        = normalized.type;
@@ -381,7 +381,7 @@ vk::ImageView Image::FindView(const ImageViewInfo& view_info) {
 	create.subresourceRange.layerCount     = normalized.layer_count;
 
 	vk::ImageView view   = nullptr;
-	const auto    result = m_graphics->device.createImageView(&create, nullptr, &view);
+	const auto    result = m_graphics.device.createImageView(&create, nullptr, &view);
 	if (result != vk::Result::eSuccess || view == nullptr) {
 		EXIT("failed to create image view: result=%d image_format=%d view_format=%d type=%d "
 		     "aspect=0x%x mip=%u+%u layer=%u+%u usage=0x%x\n",

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

@@ -92,29 +92,18 @@ SelectSampledColorView(vk::Format image_format, vk::Format view_format, uint32_t
 	}
 }
 
-[[nodiscard]] inline uint32_t
-SelectSampledDepthView(vk::Format image_format, vk::Format view_format, uint32_t swizzle) noexcept {
-	if (IsSupportedSampledDepthView(image_format, view_format, swizzle)) {
-		return swizzle;
-	}
-	EXIT("unsupported sampled depth image view: image_format=%d view_format=%d swizzle=0x%03x\n",
-	     static_cast<int>(image_format), static_cast<int>(view_format), swizzle);
-}
-
 [[nodiscard]] inline bool
 IsSupportedSampledDepthResource(const ShaderRecompiler::IR::ImageResource& resource) noexcept {
-	if (resource.kind != ShaderRecompiler::IR::ResourceKind::Image &&
-	    resource.kind != ShaderRecompiler::IR::ResourceKind::ImageUint) {
+	if (resource.resource_class != ShaderRecompiler::IR::ImageResourceClass::Sampled) {
 		return false;
 	}
-	if (resource.kind == ShaderRecompiler::IR::ResourceKind::ImageUint && resource.depth_compare) {
+	if (resource.numeric_class == Prospero::TextureNumericClass::Unsupported) {
 		return false;
 	}
-	return (resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2D ||
-	        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 &&
+	if (resource.numeric_class != Prospero::TextureNumericClass::Float && resource.depth_compare) {
+		return false;
+	}
+	return resource.mip_mode == ShaderRecompiler::IR::ImageMipMode::None && resource.read &&
 	       !resource.written && !resource.atomic;
 }
 
@@ -128,8 +117,9 @@ inline void ValidateStorageColorView(vk::Format image_format, vk::Format view_fo
 
 [[nodiscard]] inline bool
 IsSupportedStorageImageResource(const ShaderRecompiler::IR::ImageResource& resource) noexcept {
-	return (resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImage ||
-	        resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint) &&
+	return resource.resource_class == ShaderRecompiler::IR::ImageResourceClass::Storage &&
+	       (resource.numeric_class == Prospero::TextureNumericClass::Float ||
+	        resource.numeric_class == Prospero::TextureNumericClass::Uint) &&
 	       (resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim1D ||
 	        resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim1DArray ||
 	        resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2D ||
@@ -141,19 +131,19 @@ IsSupportedStorageImageResource(const ShaderRecompiler::IR::ImageResource& resou
 	         resource.mip_count != 0u)) &&
 	       resource.written &&
 	       (!resource.atomic ||
-	        (resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint &&
-	         resource.read)) &&
+	        (resource.numeric_class == Prospero::TextureNumericClass::Uint && resource.read)) &&
 	       !resource.depth_compare;
 }
 
 inline void
 ValidateStorageImageResource(const ShaderRecompiler::IR::ImageResource& resource) noexcept {
 	if (!IsSupportedStorageImageResource(resource)) {
-		EXIT("unsupported storage color image resource: kind=%u dimension=%u mip=%u "
+		EXIT("unsupported storage color image resource: class=%u numeric=%u dimension=%u mip=%u "
 		     "read=%d written=%d atomic=%d depth_compare=%d\n",
-		     static_cast<uint32_t>(resource.kind), static_cast<uint32_t>(resource.dimension),
-		     static_cast<uint32_t>(resource.mip_mode), resource.read, resource.written,
-		     resource.atomic, resource.depth_compare);
+		     static_cast<uint32_t>(resource.resource_class),
+		     static_cast<uint32_t>(resource.numeric_class),
+		     static_cast<uint32_t>(resource.dimension), static_cast<uint32_t>(resource.mip_mode),
+		     resource.read, resource.written, resource.atomic, resource.depth_compare);
 	}
 }
 

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

@@ -1,15 +1,16 @@
 #include "graphics/host_gpu/renderer/image/textureCommon.h"
 
+#include "common/alignment.h"
 #include "common/assert.h"
 #include "graphics/guest_gpu/gpu_defs.h"
 #include "graphics/guest_gpu/gpu_format.h"
 #include "graphics/host_gpu/renderer/image/tiler.h"
 #include "graphics/host_gpu/vulkanCommon.h"
+#include "graphics/shader/shader.h"
 
 #include <algorithm>
-#include <bit>
+#include <array>
 #include <cinttypes>
-#include <cstring>
 
 namespace Libs::Graphics {
 namespace {
@@ -27,13 +28,13 @@ constexpr Prospero::ColorComponentMapping kRenderTargetColorMappings[4][4] = {
      Prospero::ColorMappingArgb},
 };
 
-struct RenderTargetHostFormatInfo {
+struct HostFormatInfo {
 	vk::Format                      format = vk::Format::eUndefined;
 	Prospero::ColorComponentMapping host_to_storage;
 };
 
-RenderTargetHostFormatInfo ResolveRenderTargetHostFormat(Prospero::BufferFormat guest_format,
-                                                         Prospero::ChannelOrder order) {
+HostFormatInfo ResolveHostFormat(Prospero::BufferFormat guest_format,
+                                 Prospero::ChannelOrder order) {
 	if (order == Prospero::ChannelOrder::kAlt) {
 		switch (guest_format) {
 			case Prospero::BufferFormat::k8_8_8_8UNorm:
@@ -47,17 +48,77 @@ RenderTargetHostFormatInfo ResolveRenderTargetHostFormat(Prospero::BufferFormat
 			default: break;
 		}
 	}
+	const auto format = VulkanFormat(guest_format);
 	switch (guest_format) {
-		case Prospero::BufferFormat::k5_5_5_1UNorm:
-			return {vk::Format::eA1R5G5B5UnormPack16, Prospero::ColorMappingBgra};
+		case Prospero::BufferFormat::k5_5_5_1UNorm: return {format, Prospero::ColorMappingBgra};
 		case Prospero::BufferFormat::k1_5_5_5UNorm:
-			return {vk::Format::eR5G5B5A1UnormPack16, Prospero::ColorMappingAbgr};
-		case Prospero::BufferFormat::k4_4_4_4UNorm:
-			return {vk::Format::eR4G4B4A4UnormPack16, Prospero::ColorMappingAbgr};
-		default: return {VulkanFormat(guest_format), {}};
+		case Prospero::BufferFormat::k4_4_4_4UNorm: return {format, Prospero::ColorMappingAbgr};
+		default: return {format, {}};
 	}
 }
 
+uint32_t GetTextureLevelDepth(uint32_t depth, uint32_t level, bool volume_texture) {
+	return volume_texture ? std::max(depth >> level, 1u) : depth;
+}
+
+size_t GetTextureRegionCount(uint32_t depth, uint32_t levels, bool volume_texture) {
+	size_t count = 0;
+	for (uint32_t level = 0; level < levels; level++) {
+		count += GetTextureLevelDepth(depth, level, volume_texture);
+	}
+	return count;
+}
+
+uint64_t CalcTextureSliceStride(const TextureUploadMipLayout* mips, uint32_t levels,
+                                uint64_t total_size, uint32_t depth) {
+	uint64_t stride = 0;
+	for (uint32_t mip_level = 0; mip_level < levels; ++mip_level) {
+		stride = std::max(stride, mips[mip_level].offset + mips[mip_level].size);
+	}
+
+	if (depth > 1 && total_size != 0 && total_size % depth == 0) {
+		stride = std::max(stride, total_size / depth);
+	}
+
+	return stride;
+}
+
+uint64_t CalcLinearUploadLevelSize(const TileTextureElementLayout& element, uint32_t pitch,
+                                   uint32_t height) {
+	const uint32_t element_columns =
+	    std::max((pitch + element.texel_width - 1u) / element.texel_width, 1u);
+	const uint32_t element_rows =
+	    std::max((height + element.texel_height - 1u) / element.texel_height, 1u);
+	return static_cast<uint64_t>(element_columns) * element_rows * element.bytes;
+}
+
+uint64_t SetLinearUploadLevels(TextureUploadMipLayout*         mips,
+                               const TileTextureElementLayout& element, uint32_t height,
+                               uint32_t levels, uint32_t base_pitch) {
+	uint64_t offset     = 0;
+	auto     mip_pitch  = base_pitch;
+	auto     mip_height = height;
+
+	for (uint32_t mip_level = 0; mip_level < levels; ++mip_level) {
+		const auto size = CalcLinearUploadLevelSize(element, mip_pitch, mip_height);
+		mips[mip_level] = {offset, size};
+
+		offset += size;
+		if (mip_pitch > 1) {
+			mip_pitch /= 2;
+		}
+		if (mip_height > 1) {
+			mip_height /= 2;
+		}
+	}
+
+	return offset;
+}
+
+bool FitsBufferRange(uint64_t offset, uint64_t size, uint64_t capacity) {
+	return offset <= capacity && size <= capacity - offset;
+}
+
 } // namespace
 
 RenderTargetFormatInfo TextureGetRenderTargetFormat(Prospero::ChannelLayout layout,
@@ -65,135 +126,61 @@ RenderTargetFormatInfo TextureGetRenderTargetFormat(Prospero::ChannelLayout layo
                                                     Prospero::ChannelOrder  order) {
 	const auto encoding = Prospero::ResolveRenderTargetFormat(layout, type);
 	if (encoding.IsValid() && encoding.SupportsOrder(order)) {
-		const auto host  = ResolveRenderTargetHostFormat(encoding.buffer_format, order);
-		const auto bytes = Prospero::RenderTargetBytesPerElement(encoding.buffer_format);
-		if (host.format != vk::Format::eUndefined && bytes != 0) {
+		const auto host_format = ResolveHostFormat(encoding.buffer_format, order);
+		const auto bytes       = Prospero::RenderTargetBytesPerElement(encoding.buffer_format);
+		if (host_format.format != vk::Format::eUndefined && bytes != 0) {
 			const auto order_mapping =
 			    kRenderTargetColorMappings[static_cast<size_t>(order)][encoding.components - 1u];
-			return {host.format, bytes, host.host_to_storage.Then(order_mapping)};
+			return {host_format.format, bytes, host_format.host_to_storage.Then(order_mapping)};
 		}
 	}
 	EXIT("unsupported render-target format combination: layout=%u type=%u order=%u\n",
 	     static_cast<uint32_t>(layout), static_cast<uint32_t>(type), static_cast<uint32_t>(order));
 }
 
-namespace {
-
-static uint32_t GetTextureLevelDepth(uint32_t depth, uint32_t level, bool volume_texture) {
-	return volume_texture ? std::max(depth >> level, 1u) : depth;
-}
-
-static size_t GetTextureRegionCount(uint32_t depth, uint64_t levels, bool volume_texture) {
-	size_t count = 0;
-	for (uint32_t level = 0; level < levels; level++) {
-		count += GetTextureLevelDepth(depth, level, volume_texture);
+vk::ComponentMapping TextureGetComponentMapping(uint32_t                        swizzle,
+                                                Prospero::ColorComponentMapping host_to_storage) {
+	constexpr std::array host_components {vk::ComponentSwizzle::eR, vk::ComponentSwizzle::eG,
+	                                      vk::ComponentSwizzle::eB, vk::ComponentSwizzle::eA};
+	std::array<vk::ComponentSwizzle, 4> storage_components {};
+	for (uint32_t host_component = 0; host_component < host_components.size(); ++host_component) {
+		storage_components[host_to_storage.Map(host_component)] = host_components[host_component];
 	}
-	return count;
-}
-
-vk::ComponentSwizzle TextureGetComponentSwizzle(uint8_t s) {
-	switch (static_cast<Prospero::CompSwizzle>(s)) {
-		case Prospero::CompSwizzle::kZero: return vk::ComponentSwizzle::eZero;
-		case Prospero::CompSwizzle::kOne: return vk::ComponentSwizzle::eOne;
-		case Prospero::CompSwizzle::kRed: return vk::ComponentSwizzle::eR;
-		case Prospero::CompSwizzle::kGreen: return vk::ComponentSwizzle::eG;
-		case Prospero::CompSwizzle::kBlue: return vk::ComponentSwizzle::eB;
-		case Prospero::CompSwizzle::kAlpha: return vk::ComponentSwizzle::eA;
-		default: EXIT("unknown swizzle: %d\n", static_cast<int>(s));
-	}
-	return vk::ComponentSwizzle::eIdentity;
-}
-
-static uint32_t TextureGetDstSel(uint32_t swizzle, uint32_t channel) {
-	return (swizzle >> (channel * 3u)) & 0x7u;
-}
-
-} // namespace
-
-vk::ComponentMapping TextureGetComponentMapping(uint32_t swizzle) {
-	vk::ComponentMapping components {};
-	components.r = TextureGetComponentSwizzle(static_cast<uint8_t>(TextureGetDstSel(swizzle, 0)));
-	components.g = TextureGetComponentSwizzle(static_cast<uint8_t>(TextureGetDstSel(swizzle, 1)));
-	components.b = TextureGetComponentSwizzle(static_cast<uint8_t>(TextureGetDstSel(swizzle, 2)));
-	components.a = TextureGetComponentSwizzle(static_cast<uint8_t>(TextureGetDstSel(swizzle, 3)));
-	return components;
+	const auto resolve = [&](uint32_t component) {
+		const auto selector = static_cast<Prospero::CompSwizzle>(GetDstSel(swizzle, component));
+		switch (selector) {
+			case Prospero::CompSwizzle::kZero: return vk::ComponentSwizzle::eZero;
+			case Prospero::CompSwizzle::kOne: return vk::ComponentSwizzle::eOne;
+			case Prospero::CompSwizzle::kRed: return storage_components[0];
+			case Prospero::CompSwizzle::kGreen: return storage_components[1];
+			case Prospero::CompSwizzle::kBlue: return storage_components[2];
+			case Prospero::CompSwizzle::kAlpha: return storage_components[3];
+			default: EXIT("unknown swizzle: %u\n", static_cast<uint32_t>(selector));
+		}
+	};
+	return {resolve(0), resolve(1), resolve(2), resolve(3)};
 }
 
 SurfaceFormatInfo TextureGetSurfaceFormatInfo(Prospero::BufferFormat format) {
-	const auto backing_format    = Prospero::RemapTextureFormat(format);
-	const auto vk_format         = VulkanFormat(backing_format);
+	const auto backing_format = Prospero::RemapTextureFormat(format);
+	const auto host_format = ResolveHostFormat(backing_format, Prospero::ChannelOrder::kStandard);
 	const auto conversion_format =
 	    backing_format != format ? format : Prospero::BufferFormat::kInvalid;
-	if (vk_format != vk::Format::eUndefined) {
-		return SurfaceFormatInfo(vk_format, conversion_format);
+	if (host_format.format != vk::Format::eUndefined) {
+		return {host_format.format, conversion_format, host_format.host_to_storage};
 	}
 	EXIT("unknown format: fmt = %u\n", static_cast<uint32_t>(format));
-	return SurfaceFormatInfo(vk::Format::eUndefined, Prospero::BufferFormat::kInvalid);
 }
 
-namespace {
-
-uint64_t CalcTextureSliceStride(const TextureUploadMipLayout* mips, uint64_t levels,
-                                uint64_t total_size, uint32_t depth) {
-	uint64_t stride = 0;
-	for (uint32_t i = 0; i < levels; i++) {
-		stride = std::max(stride, mips[i].offset + mips[i].size);
-	}
-
-	if (depth > 1 && total_size != 0 && total_size % depth == 0) {
-		const auto guest_stride = total_size / depth;
-		if (guest_stride >= stride) {
-			stride = guest_stride;
-		}
-	}
-
-	return stride;
-}
-
-uint64_t CalcLinearUploadLevelSize(const TileTextureElementLayout& element, uint32_t pitch,
-                                   uint32_t height) {
-	const uint32_t elements_w =
-	    std::max((pitch + element.texel_width - 1u) / element.texel_width, 1u);
-	const uint32_t elements_h =
-	    std::max((height + element.texel_height - 1u) / element.texel_height, 1u);
-	return static_cast<uint64_t>(elements_w) * elements_h * element.bytes;
-}
-
-uint64_t SetLinearUploadLevels(TextureUploadMipLayout*         mips,
-                               const TileTextureElementLayout& element, uint64_t height,
-                               uint64_t levels, uint32_t base_pitch) {
-	uint64_t offset = 0;
-	auto     pitch  = base_pitch;
-	auto     h      = static_cast<uint32_t>(height);
-
-	for (uint32_t i = 0; i < levels; i++) {
-		const auto size = CalcLinearUploadLevelSize(element, pitch, h);
-		mips[i].size    = size;
-		mips[i].offset  = offset;
-
-		offset += size;
-		if (pitch > 1) {
-			pitch /= 2;
-		}
-		if (h > 1) {
-			h /= 2;
-		}
-	}
-
-	return offset;
-}
-
-} // namespace
-
 TextureUploadLayout TextureCalcUploadLayout(Prospero::BufferFormat format, uint32_t width,
                                             uint32_t height, uint32_t levels, uint32_t depth,
-                                            Prospero::TileMode tile, uint64_t upload_size,
+                                            Prospero::TileMode tile_mode, uint64_t upload_size,
                                             bool allow_depth_tile, bool volume_texture,
                                             const char* owner) {
 	TextureUploadLayout layout {};
 	layout.surface.description = {
 	    format,
-	    tile,
+	    tile_mode,
 	    volume_texture ? TileSurfaceDimension::Dim3D : TileSurfaceDimension::Dim2D,
 	    width,
 	    height,
@@ -202,7 +189,6 @@ TextureUploadLayout TextureCalcUploadLayout(Prospero::BufferFormat format, uint3
 	    volume_texture ? 1u : depth,
 	};
 	const auto&              description = layout.surface.description;
-	const auto               tile_mode   = description.tile_mode;
 	TileTextureElementLayout element {};
 
 	if (format == Prospero::BufferFormat::kInvalid) {
@@ -212,34 +198,19 @@ TextureUploadLayout TextureCalcUploadLayout(Prospero::BufferFormat format, uint3
 		     levels);
 	}
 
-	switch (tile_mode) {
-		case Prospero::TileMode::kLinear:
-			if (!TileGetTextureElementLayout(format, element)) {
-				EXIT("%s: unsupported linear texture format: fmt=%u\n", owner,
-				     static_cast<uint32_t>(format));
-			}
-			break;
-		case Prospero::TileMode::kDepth:
-			if (!allow_depth_tile || !TileGetTiledTextureLayout(description, layout.surface)) {
-				EXIT("%s: unsupported typed tiled upload: fmt=%u tile=%u "
-				     "size=%" PRIu64 " extent=%ux%u levels=%u\n",
-				     owner, static_cast<uint32_t>(format),
-				     static_cast<uint32_t>(description.tile_mode), upload_size, width, height,
-				     levels);
-			}
-			break;
-		default:
-			if (!TileGetTiledTextureLayout(description, layout.surface)) {
-				EXIT("%s: unsupported typed tiled upload: fmt=%u tile=%u "
-				     "size=%" PRIu64 " extent=%ux%u levels=%u\n",
-				     owner, static_cast<uint32_t>(format),
-				     static_cast<uint32_t>(description.tile_mode), upload_size, width, height,
-				     levels);
-			}
-			break;
-	}
-
-	if (tile_mode != Prospero::TileMode::kLinear) {
+	if (tile_mode == Prospero::TileMode::kLinear) {
+		if (!TileGetTextureElementLayout(format, element)) {
+			EXIT("%s: unsupported linear texture format: fmt=%u\n", owner,
+			     static_cast<uint32_t>(format));
+		}
+	} else {
+		if ((tile_mode == Prospero::TileMode::kDepth && !allow_depth_tile) ||
+		    !TileGetTiledTextureLayout(description, layout.surface)) {
+			EXIT("%s: unsupported typed tiled upload: fmt=%u tile=%u "
+			     "size=%" PRIu64 " extent=%ux%u levels=%u\n",
+			     owner, static_cast<uint32_t>(format), static_cast<uint32_t>(tile_mode),
+			     upload_size, width, height, levels);
+		}
 		element = {layout.surface.texture.block.bytes_per_element,
 		           layout.surface.texture.texel_width, layout.surface.texture.texel_height};
 	}
@@ -272,40 +243,33 @@ TextureUploadLayout TextureCalcUploadLayout(Prospero::BufferFormat format, uint3
 std::vector<vk::BufferImageCopy> TextureBuildImageCopies(const TextureUploadLayout& layout) {
 	const auto& description    = layout.surface.description;
 	const bool  volume_texture = description.dimension == TileSurfaceDimension::Dim3D;
+	const bool  linear_texture = description.tile_mode == Prospero::TileMode::kLinear;
 	const auto  depth          = volume_texture ? description.depth : description.layers;
 	uint32_t    mip_width      = description.width;
 	uint32_t    mip_height     = description.height;
-	uint32_t    mip_pitch = volume_texture && description.tile_mode != Prospero::TileMode::kLinear
-	                            ? description.width
-	                            : layout.pitch;
+	uint32_t    mip_pitch = volume_texture && !linear_texture ? description.width : layout.pitch;
 
 	std::vector<vk::BufferImageCopy> regions;
 	regions.reserve(GetTextureRegionCount(depth, description.levels, volume_texture));
-	for (uint32_t i = 0; i < description.levels; i++) {
-		EXIT_NOT_IMPLEMENTED(layout.mips[i].size == 0);
-
-		const auto mip_depth = GetTextureLevelDepth(depth, i, volume_texture);
-
-		for (uint32_t z = 0; z < mip_depth; z++) {
-			const auto          slice_offset = z * layout.slice_stride;
-			vk::BufferImageCopy region {};
-			region.bufferOffset     = layout.mips[i].offset + slice_offset;
-			region.imageSubresource = {vk::ImageAspectFlagBits::eColor, i, volume_texture ? 0u : z,
-			                           1};
-			region.imageOffset.z    = volume_texture ? static_cast<int>(z) : 0;
-			region.imageExtent      = {mip_width, mip_height, 1};
-			const bool linear       = description.tile_mode == Prospero::TileMode::kLinear;
-			if (linear) {
-				region.bufferRowLength   = layout.mips[i].row_length;
-				region.bufferImageHeight = layout.mips[i].image_height;
-			} else {
-				const auto texel_width = layout.surface.texture.texel_width;
-				const auto align       = [](uint32_t value, uint32_t block) {
-					return ((value + block - 1u) / block) * block;
-				};
-				const auto pitch       = align(mip_pitch, texel_width);
-				region.bufferRowLength = pitch > align(mip_width, texel_width) ? pitch : 0;
-			}
+	for (uint32_t mip_level = 0; mip_level < description.levels; ++mip_level) {
+		EXIT_NOT_IMPLEMENTED(layout.mips[mip_level].size == 0);
+		const auto          mip_depth = GetTextureLevelDepth(depth, mip_level, volume_texture);
+		vk::BufferImageCopy region {};
+		region.imageSubresource = {vk::ImageAspectFlagBits::eColor, mip_level, 0, 1};
+		region.imageExtent      = {mip_width, mip_height, 1};
+		if (linear_texture) {
+			region.bufferRowLength   = layout.mips[mip_level].row_length;
+			region.bufferImageHeight = layout.mips[mip_level].image_height;
+		} else {
+			const auto texel_width   = layout.surface.texture.texel_width;
+			const auto aligned_pitch = Common::AlignUp(mip_pitch, texel_width);
+			region.bufferRowLength =
+			    aligned_pitch > Common::AlignUp(mip_width, texel_width) ? aligned_pitch : 0;
+		}
+		for (uint32_t slice_index = 0; slice_index < mip_depth; ++slice_index) {
+			region.bufferOffset = layout.mips[mip_level].offset + slice_index * layout.slice_stride;
+			region.imageSubresource.baseArrayLayer = volume_texture ? 0u : slice_index;
+			region.imageOffset.z = volume_texture ? static_cast<int>(slice_index) : 0;
 			regions.push_back(region);
 		}
 
@@ -323,17 +287,9 @@ std::vector<vk::BufferImageCopy> TextureBuildImageCopies(const TextureUploadLayo
 	return regions;
 }
 
-static bool SetGpuTileSize(uint64_t offset, uint64_t length, uint64_t capacity, uint64_t& size) {
-	if (offset > capacity || length > capacity - offset) {
-		return false;
-	}
-	size = length;
-	return true;
-}
-
 bool TextureBuildGpuTileInfos(uint64_t tiled_size, const std::vector<vk::BufferImageCopy>& regions,
                               const TextureUploadLayout& layout, uint32_t levels,
-                              std::vector<GpuTileInfo>& out_infos) {
+                              std::vector<GpuTileInfo>& out_tile_infos) {
 	const auto& description    = layout.surface.description;
 	const bool  volume_texture = description.dimension == TileSurfaceDimension::Dim3D;
 	const auto  depth          = volume_texture ? description.depth : description.layers;
@@ -351,111 +307,69 @@ bool TextureBuildGpuTileInfos(uint64_t tiled_size, const std::vector<vk::BufferI
 		return false;
 	}
 
-	std::vector<GpuTileInfo> infos;
-	infos.reserve(regions.size());
-	if (volume_texture) {
-		size_t region_base = 0;
-		for (uint32_t level = 0; level < levels; ++level) {
-			const uint32_t mip_depth     = GetTextureLevelDepth(depth, level, true);
-			const bool     tail          = level >= surface.first_tail_level;
-			const uint64_t linear_stride = layout.slice_stride;
-			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  logical_height = region.bufferImageHeight != 0
-				                                 ? region.bufferImageHeight
-				                                 : region.imageExtent.height;
-				GpuTileInfo info {};
-				info.family            = block.family;
-				info.bytes_per_element = block.bytes_per_element;
-				info.linear_offset     = region.bufferOffset;
-				info.tiled_offset =
-				    static_cast<uint64_t>(z / block.block_depth) * surface.block_slice_size +
-				    surface.mips[level].offset;
-				const uint64_t linear_span =
-				    static_cast<uint64_t>(copy_depth - 1u) * linear_stride +
-				    layout.mips[level].size;
-				if (!SetGpuTileSize(info.linear_offset, linear_span, UINT64_MAX,
-				                    info.linear_size) ||
-				    !SetGpuTileSize(info.tiled_offset, surface.mips[level].size, tiled_size,
-				                    info.tiled_size)) {
-					return false;
-				}
-				info.linear_slice_stride = linear_stride;
-				info.width  = std::max((region.imageExtent.width + texture.texel_width - 1u) /
-				                           texture.texel_width,
-				                       1u);
-				info.height = std::max(
-				    (logical_height + texture.texel_height - 1u) / texture.texel_height, 1u);
-				info.depth = copy_depth;
-				info.surface_z =
+	std::vector<GpuTileInfo> tile_infos;
+	tile_infos.reserve(regions.size());
+	const auto slices_per_copy = volume_texture ? block.block_depth : 1u;
+	size_t     region_base     = 0;
+	for (uint32_t mip_level = 0; mip_level < levels; ++mip_level) {
+		const auto& mip       = surface.mips[mip_level];
+		const auto  mip_depth = GetTextureLevelDepth(depth, mip_level, volume_texture);
+		for (uint32_t slice_index = 0; slice_index < mip_depth; slice_index += slices_per_copy) {
+			const auto& region = regions[region_base + slice_index];
+			GpuTileInfo tile_info {};
+			tile_info.family            = block.family;
+			tile_info.bytes_per_element = block.bytes_per_element;
+			tile_info.linear_offset     = region.bufferOffset;
+			tile_info.linear_size       = layout.mips[mip_level].size;
+			tile_info.tiled_size        = mip.size;
+			if (volume_texture) {
+				tile_info.depth               = std::min(slices_per_copy, mip_depth - slice_index);
+				tile_info.linear_slice_stride = layout.slice_stride;
+				tile_info.linear_size += (tile_info.depth - 1u) * layout.slice_stride;
+				tile_info.tiled_offset =
+				    mip.offset +
+				    static_cast<uint64_t>(slice_index / slices_per_copy) * surface.block_slice_size;
+				tile_info.surface_z =
 				    block.block_depth == 1 ? static_cast<uint32_t>(region.imageOffset.z) : 0;
-				info.pitch = std::max((pitch + texture.texel_width - 1u) / texture.texel_width, 1u);
-				info.tail_x       = tail ? surface.mips[level].tail_x : 0;
-				info.tail_y       = tail ? surface.mips[level].tail_y : 0;
-				info.tail         = tail;
-				info.tiled_width  = surface.mips[level].padded_width;
-				info.tiled_height = surface.mips[level].padded_height;
-				infos.push_back(info);
-			}
-			region_base += mip_depth;
-		}
-	} else {
-		size_t region_index = 0;
-		for (uint32_t level = 0; level < levels; level++) {
-			const auto& level_size  = layout.mips[level];
-			const auto& mip         = surface.mips[level];
-			const bool  tail        = level >= surface.first_tail_level;
-			const auto  level_depth = GetTextureLevelDepth(depth, level, false);
-			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  logical_height = region.bufferImageHeight != 0
-				                                 ? region.bufferImageHeight
-				                                 : region.imageExtent.height;
-				GpuTileInfo info {};
-				info.family              = block.family;
-				info.bytes_per_element   = block.bytes_per_element;
-				info.linear_offset       = region.bufferOffset;
+			} else {
 				const auto source_stride = layout.source_slice_stride != 0
 				                               ? layout.source_slice_stride
 				                               : surface.block_slice_size;
-				if (z > (UINT64_MAX - mip.offset) / source_stride) {
+				if (slice_index > (UINT64_MAX - mip.offset) / source_stride) {
 					return false;
 				}
-				info.tiled_offset = mip.offset + static_cast<uint64_t>(z) * source_stride;
-				if (!SetGpuTileSize(info.linear_offset, level_size.size, UINT64_MAX,
-				                    info.linear_size) ||
-				    !SetGpuTileSize(info.tiled_offset, mip.size, tiled_size, info.tiled_size)) {
-					return false;
-				}
-				info.width  = std::max((region.imageExtent.width + texture.texel_width - 1u) /
-				                           texture.texel_width,
-				                       1u);
-				info.height = std::max(
-				    (logical_height + texture.texel_height - 1u) / texture.texel_height, 1u);
-				info.surface_z = block.family == TileBlockFamily::RenderTarget64KB ||
-				                         block.family == TileBlockFamily::Depth64KB
-				                     ? region.imageSubresource.baseArrayLayer
-				                     : 0;
-				info.pitch = std::max((pitch + texture.texel_width - 1u) / texture.texel_width, 1u);
-				info.tail  = tail;
-				info.tail_x       = tail ? mip.tail_x : 0;
-				info.tail_y       = tail ? mip.tail_y : 0;
-				info.tiled_width  = mip.padded_width;
-				info.tiled_height = mip.padded_height;
-				infos.push_back(info);
+				tile_info.tiled_offset =
+				    mip.offset + static_cast<uint64_t>(slice_index) * source_stride;
+				tile_info.surface_z = block.family == TileBlockFamily::RenderTarget64KB ||
+				                              block.family == TileBlockFamily::Depth64KB
+				                          ? region.imageSubresource.baseArrayLayer
+				                          : 0;
 			}
+			if (!FitsBufferRange(tile_info.linear_offset, tile_info.linear_size, UINT64_MAX) ||
+			    !FitsBufferRange(tile_info.tiled_offset, tile_info.tiled_size, tiled_size)) {
+				return false;
+			}
+			const auto row_length =
+			    region.bufferRowLength != 0 ? region.bufferRowLength : region.imageExtent.width;
+			const auto image_height = region.bufferImageHeight != 0 ? region.bufferImageHeight
+			                                                        : region.imageExtent.height;
+			tile_info.width         = std::max(
+			    (region.imageExtent.width + texture.texel_width - 1u) / texture.texel_width, 1u);
+			tile_info.height =
+			    std::max((image_height + texture.texel_height - 1u) / texture.texel_height, 1u);
+			tile_info.pitch =
+			    std::max((row_length + texture.texel_width - 1u) / texture.texel_width, 1u);
+			tile_info.tail         = mip_level >= surface.first_tail_level;
+			tile_info.tail_x       = tile_info.tail ? mip.tail_x : 0;
+			tile_info.tail_y       = tile_info.tail ? mip.tail_y : 0;
+			tile_info.tiled_width  = mip.padded_width;
+			tile_info.tiled_height = mip.padded_height;
+			tile_infos.push_back(tile_info);
 		}
+		region_base += mip_depth;
 	}
 
-	if (infos.empty()) {
-		return false;
-	}
-	out_infos = std::move(infos);
+	out_tile_infos = std::move(tile_infos);
 	return true;
 }
 

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

@@ -1,9 +1,7 @@
 #ifndef EMULATOR_INCLUDE_EMULATOR_GRAPHICS_HOST_GPU_RENDERER_IMAGE_TEXTURECOMMON_H_
 #define EMULATOR_INCLUDE_EMULATOR_GRAPHICS_HOST_GPU_RENDERER_IMAGE_TEXTURECOMMON_H_
 
-#include "common/abi.h"
-#include "common/common.h"
-#include "graphics/guest_gpu/gpu_format.h"
+#include "graphics/guest_gpu/gpu_defs.h"
 #include "graphics/guest_gpu/tile.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 
@@ -20,12 +18,9 @@ struct RenderTargetFormatInfo {
 };
 
 struct SurfaceFormatInfo {
-	explicit constexpr SurfaceFormatInfo(vk::Format format,
-	                                     Prospero::BufferFormat conversion_format)
-	    : vk_format(format), conversion_format(conversion_format) {}
-
-	vk::Format             vk_format;
-	Prospero::BufferFormat conversion_format;
+	vk::Format                      vk_format;
+	Prospero::BufferFormat          conversion_format;
+	Prospero::ColorComponentMapping host_to_storage;
 };
 
 struct TextureUploadMipLayout {
@@ -43,20 +38,21 @@ struct TextureUploadLayout {
 	TextureUploadMipLayout mips[16] = {};
 };
 
-vk::ComponentMapping   TextureGetComponentMapping(uint32_t swizzle);
+vk::ComponentMapping   TextureGetComponentMapping(uint32_t                        swizzle,
+                                                  Prospero::ColorComponentMapping host_to_storage);
 SurfaceFormatInfo      TextureGetSurfaceFormatInfo(Prospero::BufferFormat format);
 RenderTargetFormatInfo TextureGetRenderTargetFormat(Prospero::ChannelLayout layout,
                                                     Prospero::ChannelType   type,
                                                     Prospero::ChannelOrder  order);
 TextureUploadLayout    TextureCalcUploadLayout(Prospero::BufferFormat format, uint32_t width,
                                                uint32_t height, uint32_t levels, uint32_t depth,
-                                               Prospero::TileMode tile, uint64_t upload_size,
+                                               Prospero::TileMode tile_mode, uint64_t upload_size,
                                                bool allow_depth_tile, bool volume_texture,
                                                const char* owner);
 std::vector<vk::BufferImageCopy> TextureBuildImageCopies(const TextureUploadLayout& layout);
 bool TextureBuildGpuTileInfos(uint64_t tiled_size, const std::vector<vk::BufferImageCopy>& regions,
                               const TextureUploadLayout& layout, uint32_t levels,
-                              std::vector<GpuTileInfo>& infos);
+                              std::vector<GpuTileInfo>& out_tile_infos);
 
 } // namespace Libs::Graphics
 

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

@@ -1,5 +1,6 @@
 #include "graphics/host_gpu/renderer/image/tiler.h"
 
+#include "common/alignment.h"
 #include "common/assert.h"
 #include "gpu_tiler_shaders/gpu_tiler_demote_d16_spv.h"
 #include "gpu_tiler_shaders/gpu_tiler_depth_spv.h"
@@ -40,7 +41,6 @@ TileManager::TileManager(GraphicContext& graphics, CommandScheduler& scheduler,
 	               nullptr};
 
 	vk::DescriptorSetLayoutCreateInfo descriptor_info {};
-	descriptor_info.sType        = vk::StructureType::eDescriptorSetLayoutCreateInfo;
 	descriptor_info.flags        = vk::DescriptorSetLayoutCreateFlagBits::ePushDescriptorKHR;
 	descriptor_info.bindingCount = static_cast<uint32_t>(bindings.size());
 	descriptor_info.pBindings    = bindings.data();
@@ -50,7 +50,6 @@ TileManager::TileManager(GraphicContext& graphics, CommandScheduler& scheduler,
 
 	const vk::PushConstantRange  push_range {vk::ShaderStageFlagBits::eCompute, 0, sizeof(Push)};
 	vk::PipelineLayoutCreateInfo layout_info {};
-	layout_info.sType                  = vk::StructureType::ePipelineLayoutCreateInfo;
 	layout_info.setLayoutCount         = 1;
 	layout_info.pSetLayouts            = &m_descriptor_layout;
 	layout_info.pushConstantRangeCount = 1;
@@ -92,11 +91,9 @@ TileManager::~TileManager() {
 TileManager::Scratch TileManager::AllocateScratch(uint64_t size) {
 	EXIT_IF(size == 0);
 	vk::BufferCreateInfo create {};
-	create.sType = vk::StructureType::eBufferCreateInfo;
 	create.size  = size;
 	create.usage = vk::BufferUsageFlagBits::eStorageBuffer | vk::BufferUsageFlagBits::eTransferSrc |
 	               vk::BufferUsageFlagBits::eTransferDst;
-	create.sharingMode = vk::SharingMode::eExclusive;
 
 	VmaAllocationCreateInfo allocate {};
 	allocate.usage       = VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE;
@@ -225,7 +222,7 @@ void TileManager::Prepare(bool tile, uint64_t tiled_capacity, uint64_t linear_ca
 
 	const uint64_t uniform_alignment =
 	    std::max<uint64_t>(limits.minUniformBufferOffsetAlignment, 1);
-	const uint64_t stride = (sizeof(Push) + uniform_alignment - 1) & ~(uniform_alignment - 1);
+	const uint64_t stride = Common::AlignUp<uint64_t>(sizeof(Push), uniform_alignment);
 	EXIT_NOT_IMPLEMENTED(dispatches.size() > UINT64_MAX / stride);
 	const uint64_t bytes  = dispatches.size() * stride;
 	auto [mapped, offset] = m_stream_buffer.Map(bytes, uniform_alignment);
@@ -263,21 +260,14 @@ vk::Pipeline TileManager::GetPipeline(uint32_t slot) {
 	const uint32_t                   values[] {1u << element_index, direction_index};
 	const vk::SpecializationMapEntry entries[] {{0, 0, 4}, {1, 4, 4}};
 	const vk::SpecializationInfo     specialization {2, entries, sizeof(values), values};
-	vk::ShaderModuleCreateInfo       module_info {};
-	module_info.sType       = vk::StructureType::eShaderModuleCreateInfo;
-	module_info.codeSize    = shaders[family_index].words * sizeof(uint32_t);
-	module_info.pCode       = shaders[family_index].code;
-	vk::ShaderModule module = nullptr;
-	RequireVulkanSuccess(m_graphics.device.createShaderModule(&module_info, nullptr, &module),
-	                     "create TileManager shader module");
+	const auto module =
+	    CompileSPV({shaders[family_index].code, shaders[family_index].words}, m_graphics.device);
 	vk::PipelineShaderStageCreateInfo stage {};
-	stage.sType               = vk::StructureType::ePipelineShaderStageCreateInfo;
 	stage.stage               = vk::ShaderStageFlagBits::eCompute;
 	stage.module              = module;
 	stage.pName               = "main";
 	stage.pSpecializationInfo = &specialization;
 	vk::ComputePipelineCreateInfo create {};
-	create.sType  = vk::StructureType::eComputePipelineCreateInfo;
 	create.stage  = stage;
 	create.layout = m_pipeline_layout;
 	const auto result =
@@ -287,19 +277,19 @@ vk::Pipeline TileManager::GetPipeline(uint32_t slot) {
 	return m_pipelines[slot];
 }
 
-void TileManager::Record(bool tile, vk::Buffer source, uint64_t source_offset,
+void TileManager::Record(vk::Buffer source, uint64_t source_offset,
                          uint64_t source_capacity, vk::Buffer target, uint64_t target_offset,
                          uint64_t target_capacity, std::span<Dispatch> dispatches,
                          bool clear_target) {
 	const auto&    limits = m_graphics.GetPhysicalDeviceProperties().limits;
 	const uint64_t descriptor_alignment =
 	    std::max<uint64_t>(limits.minStorageBufferOffsetAlignment, 4);
-	const uint64_t source_descriptor_offset = source_offset & ~(descriptor_alignment - 1);
-	const uint64_t target_descriptor_offset = target_offset & ~(descriptor_alignment - 1);
+	const uint64_t source_descriptor_offset = Common::AlignDown(source_offset, descriptor_alignment);
+	const uint64_t target_descriptor_offset = Common::AlignDown(target_offset, descriptor_alignment);
 	const uint64_t source_base              = source_offset - source_descriptor_offset;
 	const uint64_t target_base              = target_offset - target_descriptor_offset;
-	const uint64_t source_range             = (source_base + source_capacity + 3u) & ~uint64_t {3};
-	const uint64_t target_range             = (target_base + target_capacity + 3u) & ~uint64_t {3};
+	const uint64_t source_range             = Common::AlignUp(source_base + source_capacity, 4);
+	const uint64_t target_range             = Common::AlignUp(target_base + target_capacity, 4);
 	EXIT_NOT_IMPLEMENTED(source_range > limits.maxStorageBufferRange ||
 	                     target_range > limits.maxStorageBufferRange || target_offset % 4 != 0 ||
 	                     target_capacity % 4 != 0);
@@ -307,7 +297,6 @@ void TileManager::Record(bool tile, vk::Buffer source, uint64_t source_offset,
 	m_scheduler.EndRendering();
 	auto                    command = m_scheduler.Current().Handle();
 	vk::BufferMemoryBarrier barriers[3] {};
-	barriers[0].sType         = vk::StructureType::eBufferMemoryBarrier;
 	barriers[0].srcAccessMask = vk::AccessFlagBits::eMemoryWrite | vk::AccessFlagBits::eHostWrite;
 	barriers[0].dstAccessMask = vk::AccessFlagBits::eShaderRead;
 	barriers[0].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
@@ -316,7 +305,6 @@ void TileManager::Record(bool tile, vk::Buffer source, uint64_t source_offset,
 	barriers[0].offset              = source_offset;
 	barriers[0].size                = source_capacity;
 	barriers[1]                     = barriers[0];
-	barriers[1].srcAccessMask = vk::AccessFlagBits::eMemoryWrite | vk::AccessFlagBits::eHostWrite;
 	barriers[1].dstAccessMask =
 	    clear_target ? vk::AccessFlagBits::eTransferWrite
 	                 : vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite;
@@ -352,7 +340,6 @@ void TileManager::Record(bool tile, vk::Buffer source, uint64_t source_offset,
 		const vk::DescriptorBufferInfo        infos[] {source_info, target_info, params_info};
 		std::array<vk::WriteDescriptorSet, 3> writes {};
 		for (uint32_t index = 0; index < writes.size(); index++) {
-			writes[index].sType           = vk::StructureType::eWriteDescriptorSet;
 			writes[index].dstBinding      = index;
 			writes[index].descriptorCount = 1;
 			writes[index].descriptorType  = index == 2 ? vk::DescriptorType::eUniformBuffer
@@ -382,9 +369,9 @@ TileManager::Result TileManager::Detile(vk::Buffer tiled, uint64_t tiled_offset,
 	const uint64_t        source_base = tiled_offset & (descriptor_alignment - 1);
 	std::vector<Dispatch> dispatches;
 	Prepare(false, tiled_capacity, linear_capacity, infos, source_base, 0, dispatches);
-	auto scratch = AllocateScratch((linear_capacity + 3u) & ~uint64_t {3});
+	auto scratch = AllocateScratch(Common::AlignUp(linear_capacity, 4));
 	DeferDestroy(scratch);
-	Record(false, tiled, tiled_offset, tiled_capacity, scratch.buffer, 0, scratch.size, dispatches,
+	Record(tiled, tiled_offset, tiled_capacity, scratch.buffer, 0, scratch.size, dispatches,
 	       true);
 	return {scratch.buffer, 0, linear_capacity};
 }
@@ -399,7 +386,7 @@ void TileManager::Tile(vk::Buffer linear, uint64_t linear_offset, uint64_t linea
 	const uint64_t        target_base = tiled_offset & (descriptor_alignment - 1);
 	std::vector<Dispatch> dispatches;
 	Prepare(true, tiled_capacity, linear_capacity, infos, source_base, target_base, dispatches);
-	Record(true, linear, linear_offset, linear_capacity, tiled, tiled_offset, tiled_capacity,
+	Record(linear, linear_offset, linear_capacity, tiled, tiled_offset, tiled_capacity,
 	       dispatches, false);
 }
 
@@ -416,19 +403,19 @@ void TileManager::TileImage(Image& image, std::span<const vk::BufferImageCopy> r
 	// Reserve all stream parameters before creating a scheduler-lived scratch dependency:
 	// StreamBuffer::Map is allowed to submit the current tick when it wraps.
 	Prepare(true, tiled_capacity, linear_capacity, infos, 0, target_base, dispatches);
-	auto linear = AllocateScratch((linear_capacity + 3u) & ~uint64_t {3});
+	auto linear = AllocateScratch(Common::AlignUp(linear_capacity, 4));
 	DeferDestroy(linear);
 	image.Download(regions, linear.buffer, 0, linear.size);
 	Result source {linear.buffer, 0, linear.size};
 	if (transform == ColorTransform::SwapBgra16) {
 		source = SwapBgra16(source);
 	}
-	Record(true, source.buffer, source.offset, linear_capacity, tiled, tiled_offset, tiled_capacity,
+	Record(source.buffer, source.offset, linear_capacity, tiled, tiled_offset, tiled_capacity,
 	       dispatches, false);
 }
 
 TileManager::Result TileManager::GetScratchBuffer(uint64_t size) {
-	auto scratch = AllocateScratch((size + 3u) & ~uint64_t {3});
+	auto scratch = AllocateScratch(Common::AlignUp(size, 4));
 	DeferDestroy(scratch);
 	return {scratch.buffer, 0, scratch.size};
 }
@@ -436,11 +423,11 @@ TileManager::Result TileManager::GetScratchBuffer(uint64_t size) {
 TileManager::StorageBinding TileManager::BindStorage(Result buffer, uint64_t size) const {
 	const auto& limits            = m_graphics.GetPhysicalDeviceProperties().limits;
 	const auto  alignment         = std::max<uint64_t>(limits.minStorageBufferOffsetAlignment, 4);
-	const auto  descriptor_offset = buffer.offset - buffer.offset % alignment;
+	const auto  descriptor_offset = Common::AlignDown(buffer.offset, alignment);
 	const auto  base              = buffer.offset - descriptor_offset;
 	EXIT_IF(buffer.buffer == nullptr || size == 0 || buffer.size < size || base > UINT32_MAX ||
 	        size > UINT64_MAX - base || base + size > UINT64_MAX - 3);
-	const auto range = (base + size + 3) & ~uint64_t {3};
+	const auto range = Common::AlignUp(base + size, 4);
 	EXIT_IF(range > limits.maxStorageBufferRange || range > UINT32_MAX);
 	return {{buffer.buffer, descriptor_offset, range}, static_cast<uint32_t>(base)};
 }
@@ -481,21 +468,13 @@ void TileManager::ConvertD16(Result source, Result target, D16Direction directio
 			code  = GPU_TILER_DEMOTE_D16_SPV;
 			words = std::size(GPU_TILER_DEMOTE_D16_SPV);
 		}
-		vk::ShaderModuleCreateInfo module_info {};
-		module_info.sType       = vk::StructureType::eShaderModuleCreateInfo;
-		module_info.codeSize    = words * sizeof(uint32_t);
-		module_info.pCode       = code;
-		vk::ShaderModule module = nullptr;
-		RequireVulkanSuccess(m_graphics.device.createShaderModule(&module_info, nullptr, &module),
-		                     "create D16 conversion shader module");
+		const auto module = CompileSPV({code, words}, m_graphics.device);
 		vk::PipelineShaderStageCreateInfo stage {};
-		stage.sType               = vk::StructureType::ePipelineShaderStageCreateInfo;
 		stage.stage               = vk::ShaderStageFlagBits::eCompute;
 		stage.module              = module;
 		stage.pName               = "main";
 		stage.pSpecializationInfo = &specialization;
 		vk::ComputePipelineCreateInfo create {};
-		create.sType  = vk::StructureType::eComputePipelineCreateInfo;
 		create.stage  = stage;
 		create.layout = m_pipeline_layout;
 		const auto result =
@@ -525,14 +504,13 @@ void TileManager::ConvertD16(Result source, Result target, D16Direction directio
 	const auto target_required = required(layout.height, layout.layers, layout.target_row_stride,
 	                                      layout.target_slice_stride, target_active);
 	EXIT_IF(source_required > UINT64_MAX - 3 || target_required > UINT64_MAX - 3);
-	const auto source_barrier_size = (source_required + 3) & ~uint64_t {3};
-	const auto target_barrier_size = (target_required + 3) & ~uint64_t {3};
+	const auto source_barrier_size = Common::AlignUp(source_required, 4);
+	const auto target_barrier_size = Common::AlignUp(target_required, 4);
 	EXIT_IF(source.size < source_barrier_size || target.size < target_barrier_size);
 
 	m_scheduler.EndRendering();
 	auto                    command = m_scheduler.Current().Handle();
 	vk::BufferMemoryBarrier barriers[2] {};
-	barriers[0].sType               = vk::StructureType::eBufferMemoryBarrier;
 	barriers[0].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
 	barriers[0].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
 	barriers[0].buffer              = source.buffer;
@@ -542,7 +520,6 @@ void TileManager::ConvertD16(Result source, Result target, D16Direction directio
 	                            vk::AccessFlagBits::eTransferWrite |
 	                            vk::AccessFlagBits::eShaderWrite;
 	barriers[0].dstAccessMask = vk::AccessFlagBits::eShaderRead;
-	barriers[1].sType         = vk::StructureType::eBufferMemoryBarrier;
 	barriers[1].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
 	barriers[1].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
 	barriers[1].buffer              = target.buffer;
@@ -596,7 +573,6 @@ void TileManager::ConvertD16(Result source, Result target, D16Direction directio
 			};
 			std::array<vk::WriteDescriptorSet, 2> writes {};
 			for (uint32_t index = 0; index < writes.size(); index++) {
-				writes[index].sType           = vk::StructureType::eWriteDescriptorSet;
 				writes[index].dstBinding      = index;
 				writes[index].descriptorCount = 1;
 				writes[index].descriptorType  = vk::DescriptorType::eStorageBuffer;
@@ -626,20 +602,12 @@ void TileManager::ConvertD16(Result source, Result target, D16Direction directio
 
 void TileManager::SwapBgra16(Result input, Result output, uint32_t pixels) {
 	if (m_swap_bgra16 == nullptr) {
-		vk::ShaderModuleCreateInfo module_info {};
-		module_info.sType       = vk::StructureType::eShaderModuleCreateInfo;
-		module_info.codeSize    = std::size(GPU_TILER_SWAP_BGRA16_SPV) * sizeof(uint32_t);
-		module_info.pCode       = GPU_TILER_SWAP_BGRA16_SPV;
-		vk::ShaderModule module = nullptr;
-		RequireVulkanSuccess(m_graphics.device.createShaderModule(&module_info, nullptr, &module),
-		                     "create BGRA16 swap shader module");
+		const auto module = CompileSPV(GPU_TILER_SWAP_BGRA16_SPV, m_graphics.device);
 		vk::PipelineShaderStageCreateInfo stage {};
-		stage.sType  = vk::StructureType::ePipelineShaderStageCreateInfo;
 		stage.stage  = vk::ShaderStageFlagBits::eCompute;
 		stage.module = module;
 		stage.pName  = "main";
 		vk::ComputePipelineCreateInfo create {};
-		create.sType  = vk::StructureType::eComputePipelineCreateInfo;
 		create.stage  = stage;
 		create.layout = m_pipeline_layout;
 		const auto result =
@@ -658,7 +626,6 @@ void TileManager::SwapBgra16(Result input, Result output, uint32_t pixels) {
 	};
 	std::array<vk::WriteDescriptorSet, 2> writes {};
 	for (uint32_t index = 0; index < writes.size(); index++) {
-		writes[index].sType           = vk::StructureType::eWriteDescriptorSet;
 		writes[index].dstBinding      = index;
 		writes[index].descriptorCount = 1;
 		writes[index].descriptorType  = vk::DescriptorType::eStorageBuffer;
@@ -666,7 +633,6 @@ void TileManager::SwapBgra16(Result input, Result output, uint32_t pixels) {
 	}
 	vk::BufferMemoryBarrier barriers[2] {};
 	for (uint32_t index = 0; index < 2; index++) {
-		barriers[index].sType               = vk::StructureType::eBufferMemoryBarrier;
 		barriers[index].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
 		barriers[index].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
 		barriers[index].buffer              = infos[index].buffer;

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

@@ -125,7 +125,7 @@ private:
 	void Prepare(bool tile, uint64_t tiled_capacity, uint64_t linear_capacity,
 	             std::span<const GpuTileInfo> infos, uint64_t source_base, uint64_t target_base,
 	             std::vector<Dispatch>& dispatches);
-	void Record(bool tile, vk::Buffer source, uint64_t source_offset, uint64_t source_capacity,
+	void Record(vk::Buffer source, uint64_t source_offset, uint64_t source_capacity,
 	            vk::Buffer target, uint64_t target_offset, uint64_t target_capacity,
 	            std::span<Dispatch> dispatches, bool clear_target);
 	[[nodiscard]] vk::Pipeline GetPipeline(uint32_t slot);

src/graphics/host_gpu/renderer/masterSemaphore.cpp

@@ -7,12 +7,10 @@ namespace Libs::Graphics {
 
 MasterSemaphore::MasterSemaphore(GraphicContext& graphics): m_graphics(graphics) {
 	vk::SemaphoreTypeCreateInfo type_info {};
-	type_info.sType         = vk::StructureType::eSemaphoreTypeCreateInfo;
 	type_info.semaphoreType = vk::SemaphoreType::eTimeline;
 	type_info.initialValue  = 0;
 
 	vk::SemaphoreCreateInfo create_info {};
-	create_info.sType = vk::StructureType::eSemaphoreCreateInfo;
 	create_info.pNext = &type_info;
 
 	const auto result = m_graphics.device.createSemaphore(&create_info, nullptr, &m_semaphore);
@@ -47,7 +45,6 @@ void MasterSemaphore::Wait(uint64_t tick) {
 	}
 
 	vk::SemaphoreWaitInfo wait_info {};
-	wait_info.sType          = vk::StructureType::eSemaphoreWaitInfo;
 	wait_info.semaphoreCount = 1;
 	wait_info.pSemaphores    = &m_semaphore;
 	wait_info.pValues        = &tick;

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

@@ -20,7 +20,7 @@ constexpr std::array DescriptorPoolSizes = {
 
 DescriptorHeap::DescriptorHeap(GraphicContext& graphics, MasterSemaphore& master_semaphore)
     : m_graphics(graphics), m_master_semaphore(master_semaphore) {
-	CreatePool();
+	CreateDescriptorPool();
 }
 
 DescriptorHeap::~DescriptorHeap() {
@@ -49,7 +49,7 @@ vk::DescriptorSet DescriptorHeap::Commit(vk::DescriptorSetLayout layout) {
 		m_pending_pools.pop_front();
 		EXIT_IF(m_graphics.device.resetDescriptorPool(m_current_pool, {}) != vk::Result::eSuccess);
 	} else {
-		CreatePool();
+		CreateDescriptorPool();
 	}
 
 	m_sets.clear();
@@ -63,7 +63,6 @@ bool DescriptorHeap::Allocate(vk::DescriptorSetLayout layout, Batch& batch) {
 	layouts.fill(layout);
 
 	vk::DescriptorSetAllocateInfo allocate {};
-	allocate.sType          = vk::StructureType::eDescriptorSetAllocateInfo;
 	allocate.descriptorPool = m_current_pool;
 	allocate.pSetLayouts    = layouts.data();
 
@@ -83,10 +82,8 @@ bool DescriptorHeap::Allocate(vk::DescriptorSetLayout layout, Batch& batch) {
 	}
 }
 
-void DescriptorHeap::CreatePool() {
+void DescriptorHeap::CreateDescriptorPool() {
 	vk::DescriptorPoolCreateInfo create {};
-	create.sType         = vk::StructureType::eDescriptorPoolCreateInfo;
-	create.flags         = {};
 	create.maxSets       = DescriptorHeapCount;
 	create.poolSizeCount = static_cast<uint32_t>(DescriptorPoolSizes.size());
 	create.pPoolSizes    = DescriptorPoolSizes.data();

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

@@ -31,7 +31,7 @@ private:
 	};
 
 	[[nodiscard]] bool Allocate(vk::DescriptorSetLayout layout, Batch& batch);
-	void               CreatePool();
+	void               CreateDescriptorPool();
 
 	GraphicContext&                                     m_graphics;
 	MasterSemaphore&                                    m_master_semaphore;

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

@@ -16,15 +16,7 @@
 
 namespace Libs::Graphics {
 
-namespace ShaderRecompiler::IR {
-struct ResourceSnapshot;
-}
-
-struct BufferView {
-	vk::Buffer     buffer = nullptr;
-	vk::DeviceSize offset = 0;
-	vk::DeviceSize range  = VK_WHOLE_SIZE;
-};
+struct ShaderStageRuntime;
 
 struct TextureBinding {
 	ImageId                    image_id;
@@ -34,23 +26,25 @@ struct TextureBinding {
 	std::vector<vk::ImageView> mip_views;
 };
 
-struct NativeDescriptors {
-	std::vector<BufferView>     buffers;
-	std::vector<TextureBinding> images;
-	std::vector<vk::Sampler>    samplers;
-	BufferView                  gds;
-	BufferView                  flattened_srt;
-	BufferView                  user_data;
-};
-
 struct PreparedBindings {
-	const ShaderRecompiler::IR::Program*          program  = nullptr;
-	const ShaderRecompiler::IR::ResourceSnapshot* snapshot = nullptr;
-	NativeDescriptors                             resources;
-	std::vector<BufferId>                         buffer_ids;
-	std::vector<uint32_t>                         flattened_srt;
-	std::vector<uint32_t>                         user_data;
-	bool                                          committed = false;
+	struct BufferSource {
+		uint64_t address = 0;
+		uint64_t size    = 0;
+		BufferId id;
+	};
+
+	// The draw owns the immutable compiled-program/runtime-snapshot association through commit.
+	const ShaderStageRuntime* runtime = nullptr;
+	// Keep the resolved guest range through cache preparation; only the host buffer ID may
+	// become stale and need resolving again when bindings are rebound.
+	std::vector<BufferSource>             buffer_sources;
+	std::vector<vk::DescriptorBufferInfo> buffers;
+	std::vector<TextureBinding>           images;
+	std::vector<vk::Sampler>              samplers;
+	vk::DescriptorBufferInfo              gds {nullptr, 0, VK_WHOLE_SIZE};
+	vk::DescriptorBufferInfo              flattened_srt;
+	vk::DescriptorBufferInfo              shader_data_buffer;
+	std::vector<uint32_t>                 shader_data;
 };
 
 [[nodiscard]] vk::DescriptorType
@@ -70,23 +64,8 @@ template <typename T>
 	return result;
 }
 
-struct TargetTextureViewInfo {
-	vk::ImageViewType type        = static_cast<vk::ImageViewType>(VK_IMAGE_VIEW_TYPE_MAX_ENUM);
-	uint32_t          base_layer  = 0;
-	uint32_t          layer_count = 0;
-};
-
-[[nodiscard]] TargetTextureViewInfo
-ResolveTargetTextureView(const ShaderRecompiler::IR::ImageResource& resource,
-                         Prospero::ImageType type, uint32_t base_layer, uint32_t image_layers);
-
-[[nodiscard]] bool IsSupportedDepthTargetDescriptor(const ShaderTextureResource& descriptor,
-                                                    const Image& image, bool r128 = false);
 [[nodiscard]] bool IsSupportedDepthTextureEncoding(const ShaderTextureResource& descriptor,
-                                                   const Image& image, bool r128 = false);
-[[nodiscard]] bool
-IsSupportedSampledVideoOutView(const ShaderRecompiler::IR::ImageResource& resource,
-                               const ShaderTextureResource& descriptor, const Image& image);
+                                                   bool r128 = false);
 void ValidateStorageTexture(const ShaderRecompiler::IR::ImageResource& resource,
                             const ShaderTextureResource& descriptor, uint64_t size);
 

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

@@ -1,6 +1,7 @@
 #include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
 
 #include "common/assert.h"
+#include "common/emulatorConfig.h"
 #include "common/file.h"
 #include "common/logging/log.h"
 #include "common/profiler.h"
@@ -11,28 +12,54 @@
 #include "graphics/host_gpu/renderer/image/imageView.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
+#include "graphics/shader/recompiler/ShaderRecompiler.h"
 #include "graphics/shader/shaderCompiler.h"
+#include "kernel/memory.h"
 #include "kytyGitVersion.h"
 #include "loader/systemContent.h"
 
 #include <algorithm>
+#include <array>
 #include <atomic>
 #include <cctype>
 #include <cstdio>
 #include <cstring>
+#include <fmt/format.h>
 #include <limits>
 #include <span>
+#include <spirv-tools/libspirv.hpp>
 #include <string_view>
+#include <tuple>
 #include <utility>
 #include <vector>
-
-#include <fmt/format.h>
 #include <xxhash.h>
 
 namespace Libs::Graphics {
 
 namespace {
 
+vk::PolygonMode ResolvePolygonMode(const HW::ModeControl& mode, bool cull_front, bool cull_back) {
+	// CxPrimitiveSetup::PolygonMode disables both per-face modes when it is zero.
+	if (mode.poly_mode == 0) {
+		return vk::PolygonMode::eFill;
+	}
+	EXIT_NOT_IMPLEMENTED(mode.poly_mode != 1);
+	if (cull_front && cull_back) {
+		return vk::PolygonMode::eFill;
+	}
+	if (!cull_front && !cull_back && mode.polymode_front_ptype != mode.polymode_back_ptype) {
+		EXIT("Pipeline: different polygon modes for two visible faces are unsupported\n");
+	}
+	// Vulkan has one polygon mode. A culled face does not constrain that mode.
+	const auto polygon_mode = cull_front ? mode.polymode_back_ptype : mode.polymode_front_ptype;
+	switch (polygon_mode) {
+		case 0: return vk::PolygonMode::ePoint;
+		case 1: return vk::PolygonMode::eLine;
+		case 2: return vk::PolygonMode::eFill;
+		default: EXIT("Pipeline: invalid polygon mode %u\n", polygon_mode);
+	}
+}
+
 std::string DriverCacheSignature(const vk::PhysicalDeviceProperties& properties) {
 	constexpr char hex[] = "0123456789abcdef";
 	std::string    uuid(VK_UUID_SIZE * 2, '0');
@@ -62,105 +89,306 @@ template <typename... Args>
 void PipelineCacheLog(fmt::format_string<Args...> format, Args&&... args) {
 	auto message = fmt::format(format, std::forward<Args>(args)...);
 	message += '\n';
-	if (Log::GetDirection() != Log::Direction::Console) {
-		std::fwrite(message.data(), 1, message.size(), stdout);
-		std::fflush(stdout);
-	}
-	Log::Write(message);
-	Log::Flush();
+	Log::WriteToConsoleAndLog(message);
 }
 
-void NormalizeStaticParamsForDynamicState(PipelineStaticParameters& static_params) {
-	static_params.viewport_scale[0]  = 0.5f;
-	static_params.viewport_scale[1]  = 0.5f;
-	static_params.viewport_scale[2]  = 1.0f;
-	static_params.viewport_offset[0] = 0.5f;
-	static_params.viewport_offset[1] = 0.5f;
-	static_params.viewport_offset[2] = 0.0f;
+bool ReadShaderGuestMemory(void*, uint64_t address, uint32_t* value) {
+	return value != nullptr &&
+	       Libs::LibKernel::Memory::TryReadGpuCleanBacking(address, value, sizeof(*value));
+}
 
-	static_params.scissor_ltrb[0] = 0;
-	static_params.scissor_ltrb[1] = 0;
-	static_params.scissor_ltrb[2] = 1;
-	static_params.scissor_ltrb[3] = 1;
+void DumpShaderSpirv(const char* stage_name, uint64_t shader_hash,
+                     const std::vector<uint32_t>& spirv) {
+	if (!Config::GraphicsDebugDumpEnabled()) {
+		return;
+	}
+	static std::atomic_int id = 0;
+	const auto path = Config::GetShaderLogFolder() / fmt::format("{:04d}_new_shader_{}_{:016x}.spv",
+	                                                             id++, stage_name, shader_hash);
+	Common::File::CreateDirectories(path.parent_path());
+	Common::File file(path);
+	if (file.IsInvalid()) {
+		const auto path_text = Common::PathToString(path);
+		LOGF_COLOR(Log::Color::BrightRed, "Can't create file: %s\n", path_text.c_str());
+		return;
+	}
+	file.Write(spirv.data(), spirv.size() * sizeof(uint32_t));
+}
+
+void DumpShaderOriginal(const char* stage_name, uint64_t shader_hash,
+                        std::span<const uint32_t> code, const std::string& decoded_dump) {
+	if (!Config::GraphicsDebugDumpEnabled()) {
+		return;
+	}
+	EXIT_IF(code.empty());
+	static std::atomic_int id = 0;
+	const auto base = Config::GetShaderLogFolder() / "original" /
+	                  fmt::format("{:04d}_new_shader_{}_{:016x}", id++, stage_name, shader_hash);
+	Common::File::CreateDirectories(base.parent_path());
+	for (const auto& [suffix, data, size]: {
+	         std::tuple {".bin", static_cast<const void*>(code.data()), code.size_bytes()},
+	         std::tuple {".rdna2", static_cast<const void*>(decoded_dump.data()),
+	                     decoded_dump.size()},
+	     }) {
+		if (size == 0) {
+			continue;
+		}
+		auto path = base;
+		path += suffix;
+		Common::File file(path);
+		if (file.IsInvalid()) {
+			const auto path_text = Common::PathToString(path);
+			LOGF_COLOR(Log::Color::BrightRed, "Can't create file: %s\n", path_text.c_str());
+		} else {
+			file.Write(data, size);
+		}
+	}
+}
+
+bool ValidateShaderSpirv(const char* label, uint64_t shader_hash,
+                         const std::vector<uint32_t>& spirv) {
+	if (!Config::ShaderValidationEnabled()) {
+		return true;
+	}
+	spvtools::SpirvTools tools(SPV_ENV_VULKAN_1_3);
+	std::string          messages;
+	tools.SetMessageConsumer([&messages](spv_message_level_t, const char*,
+	                                     const spv_position_t& position, const char* message) {
+		messages += fmt::format("{}: {} ({}) {}\n", static_cast<int>(position.line),
+		                        static_cast<int>(position.column), static_cast<int>(position.index),
+		                        message);
+	});
+	if (tools.Validate(spirv)) {
+		return true;
+	}
+	spvtools::SpirvTools disassembler(SPV_ENV_VULKAN_1_2);
+	std::string          text;
+	disassembler.Disassemble(spirv, &text,
+	                         static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_NO_HEADER) |
+	                             static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_FRIENDLY_NAMES) |
+	                             static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_COMMENT) |
+	                             static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_INDENT) |
+	                             static_cast<uint32_t>(SPV_BINARY_TO_TEXT_OPTION_COLOR));
+	LOGF_COLOR(Log::Color::BrightRed, "%s SPIR-V validation failed hash=0x%016" PRIx64 ":\n%s",
+	           label, shader_hash, messages.c_str());
+	LOGF("%s\n", text.c_str());
+	return false;
 }
 
 } // namespace
 
 struct PipelineCache::ProgramCache {
-	struct SourceKey {
-		ShaderType stage = ShaderType::Unknown;
-		uint64_t   hash  = 0;
+	struct ProgramKey {
+		ShaderType            stage           = ShaderType::Unknown;
+		uint64_t              hash            = 0;
+		uint32_t              user_data_count = 0;
+		uint32_t              code_size       = 0;
+		std::vector<uint32_t> static_state;
 
-		bool operator==(const SourceKey&) const = default;
+		bool operator==(const ProgramKey&) const = default;
 	};
 
 	struct Permutation {
-		std::vector<uint32_t>                               static_key;
-		std::shared_ptr<const ShaderRecompiler::IR::Program> program;
-		ShaderProgram                                       handle;
+		ShaderRecompiler::IR::ResourceSpecialization specialization;
+		ShaderRecompiler::IR::CompiledShaderInfo     program;
+		ShaderProgram                                handle;
 	};
 
-	struct SourceKeyHash {
-		std::size_t operator()(const SourceKey& key) const {
+	struct SourceEntry {
+		explicit SourceEntry(ShaderRecompiler::IR::ResourcePlan plan)
+		    : resource_plan(std::move(plan)) {
+			permutations.reserve(8);
+		}
+
+		ShaderRecompiler::IR::ResourcePlan resource_plan;
+		std::vector<Permutation>           permutations;
+	};
+
+	struct ProgramKeyHash {
+		std::size_t operator()(const ProgramKey& key) const {
 			std::size_t hash = static_cast<std::size_t>(key.stage);
 			PipelineKeyHash::Mix(hash, static_cast<std::size_t>(key.hash));
 			if constexpr (sizeof(std::size_t) < sizeof(uint64_t)) {
 				PipelineKeyHash::Mix(hash, static_cast<std::size_t>(key.hash >> 32u));
 			}
+			PipelineKeyHash::Mix(hash, key.user_data_count);
+			PipelineKeyHash::Mix(hash, key.code_size);
+			PipelineKeyHash::Mix(hash, key.static_state.size());
+			// Bucket same-shape static variants by source. ProgramKey equality performs the one
+			// exact state comparison needed on a stable hit without hashing up to 429 words first.
 			return hash;
 		}
 	};
 
-	static uint64_t MixId(uint64_t hash, uint64_t value) {
-		return hash ^ (value + 0x9e3779b97f4a7c15ull + (hash << 6u) + (hash >> 2u));
+	static constexpr std::size_t MaxStaticKeyWords = 13 + ShaderVertexInputInfo::RES_MAX * 13;
+
+	Permutation CompilePermutation(const ShaderParams&                          params,
+	                               const ShaderRecompiler::CompileOptions&      options,
+	                               ShaderRecompiler::TranslateResult            translated,
+	                               ShaderRecompiler::IR::ResourceSpecialization specialization,
+	                               uint32_t push_data_start_dword) {
+		const char* stage_name = nullptr;
+		switch (options.stage) {
+			case ShaderType::Vertex: stage_name = "vs"; break;
+			case ShaderType::Mesh: stage_name = "ms"; break;
+			case ShaderType::Local: stage_name = "ls"; break;
+			case ShaderType::TessellationControl: stage_name = "hs"; break;
+			case ShaderType::TessellationEvaluation: stage_name = "ds"; break;
+			case ShaderType::Pixel: stage_name = "ps"; break;
+			case ShaderType::Compute: stage_name = "cs"; break;
+			default: EXIT("invalid pipeline shader stage\n");
+		}
+		auto result = ShaderRecompiler::CompileProgram(std::move(translated), options,
+		                                               specialization, push_data_start_dword);
+		DumpShaderOriginal(stage_name, options.shader_hash, params.code, result.decoded_dump);
+		if (!ValidateShaderSpirv(options.dump_label, options.shader_hash, result.spirv)) {
+			DumpShaderSpirv(stage_name, options.shader_hash, result.spirv);
+			EXIT("%s failed hash=0x%016" PRIx64 ": SPIR-V validation failed\n", options.dump_label,
+			     options.shader_hash);
+		}
+		DumpShaderSpirv(stage_name, options.shader_hash, result.spirv);
+
+		const auto module = CompileSPV(result.spirv, device);
+		EXIT_IF(module == nullptr);
+		if (options.dump_ir) {
+			LOGF("%s SPIR-V words=%" PRIu64 " wave_size=%u\n", options.dump_label,
+			     static_cast<uint64_t>(result.spirv.size()), options.wave_size);
+		}
+		return {
+		    .specialization = std::move(specialization),
+		    .program        = std::move(result.program).TakeCompiledInfo(),
+		    .handle         = {.id = ++next_shader_id, .module = module},
+		};
 	}
 
-	static constexpr std::size_t MaxStaticKeyWords =
-	    5 + ShaderVertexInputInfo::RES_MAX * 17;
-
 	template <typename InputInfo>
-	ShaderProgram Get(const ShaderParams& params, InputInfo& input_info) {
-		constexpr ShaderType stage = [] {
-			if constexpr (std::is_same_v<InputInfo, ShaderVertexInputInfo>) {
-				return ShaderType::Vertex;
-			} else if constexpr (std::is_same_v<InputInfo, ShaderPixelInputInfo>) {
-				return ShaderType::Pixel;
-			} else {
-				static_assert(std::is_same_v<InputInfo, ShaderComputeInputInfo>);
-				return ShaderType::Compute;
-			}
-		}();
+	ShaderProgram Get(const ShaderParams& params, InputInfo& input_info,
+	                  uint32_t& push_data_cursor) {
+		ShaderType stage;
+		if constexpr (std::is_same_v<InputInfo, ShaderVertexInputInfo>) {
+			stage = input_info.logical_stage;
+		} else if constexpr (std::is_same_v<InputInfo, ShaderPixelInputInfo>) {
+			stage = ShaderType::Pixel;
+		} else {
+			static_assert(std::is_same_v<InputInfo, ShaderComputeInputInfo>);
+			stage = ShaderType::Compute;
+		}
 
-		BuildStageStaticKey(input_info, key_scratch);
-		auto& permutations = programs[{stage, params.hash}];
-		for (const auto& permutation: permutations) {
-			if (permutation.static_key == key_scratch &&
-			    MaterializeProgram(permutation.program, params, input_info)) {
-				return permutation.handle;
+		lookup_key.stage           = stage;
+		lookup_key.hash            = params.hash;
+		lookup_key.user_data_count = static_cast<uint32_t>(params.user_data.size());
+		lookup_key.code_size       = static_cast<uint32_t>(params.code.size());
+		BuildStageStaticKey(input_info, lookup_key.static_state);
+		auto                                         entry = programs.find(lookup_key);
+		ShaderRecompiler::IR::ResourceSnapshot       resources;
+		ShaderRecompiler::IR::ResourceSpecialization specialization;
+		const ShaderRecompiler::IR::SrtRuntime       runtime {
+		    .user_data                  = params.user_data,
+		    .shader_base                = params.Base(),
+		    .read_specialization_memory = ReadShaderGuestMemory,
+		};
+		if (entry != programs.end()) {
+			EXIT_IF(!ShaderRecompiler::IR::MaterializeResources(
+			    entry->second.resource_plan, runtime, resources, specialization));
+			if (const auto permutation = std::ranges::find_if(
+			        entry->second.permutations, [&](const Permutation& candidate) {
+				        const auto& layout = candidate.program.bindings;
+				        return layout.push_data_start_dword ==
+				                   ShaderRecompiler::IR::PushData::StartFor(
+				                       push_data_cursor, layout.ShaderDataDwords()) &&
+				               candidate.specialization == specialization;
+			        });
+			    permutation != entry->second.permutations.end()) {
+				input_info.stage = {.program   = &permutation->program,
+				                    .resources = std::move(resources)};
+				permutation->program.bindings.AdvancePushData(push_data_cursor);
+				return permutation->handle;
 			}
 		}
 
-		const auto module = CompileProgram(device, params, input_info);
-		EXIT_IF(module == nullptr || !input_info.stage);
-		uint64_t id = MixId(MixId(params.hash, static_cast<uint64_t>(stage)), permutations.size());
-		if (id == 0) {
-			id = 1;
+		ShaderStageInputInfo stage_input {};
+		if constexpr (std::is_same_v<InputInfo, ShaderVertexInputInfo>) {
+			stage_input.vertex = &input_info;
+		} else if constexpr (std::is_same_v<InputInfo, ShaderPixelInputInfo>) {
+			stage_input.pixel = &input_info;
+		} else {
+			stage_input.compute = &input_info;
 		}
-		const ShaderProgram handle {.id = id, .module = module};
-		permutations.push_back({key_scratch, input_info.stage.program, handle});
+		const char* label = nullptr;
+		switch (stage) {
+			case ShaderType::Vertex: label = "ShaderRecompiler VS"; break;
+			case ShaderType::Mesh: label = "ShaderRecompiler MS"; break;
+			case ShaderType::Local: label = "ShaderRecompiler LS"; break;
+			case ShaderType::TessellationControl: label = "ShaderRecompiler HS"; break;
+			case ShaderType::TessellationEvaluation: label = "ShaderRecompiler DS"; break;
+			case ShaderType::Pixel: label = "ShaderRecompiler PS"; break;
+			case ShaderType::Compute: label = "ShaderRecompiler CS"; break;
+			default: EXIT("invalid pipeline shader stage\n");
+		}
+		ShaderRecompiler::CompileOptions options;
+		options.stage       = stage;
+		options.shader_hash = params.hash;
+		options.user_data   = params.user_data;
+		options.back_code      = params.back_code;
+		options.dump_ir     = Config::GetShaderLogDirection() != Config::LogDirection::Silent;
+		options.early_dump  = options.dump_ir;
+		options.dump_label  = label;
+		options.input_info  = stage_input;
 
-		std::printf("Num compiled %u shaders\n", ++num_compiled);
-		return handle;
+		if constexpr (std::is_same_v<InputInfo, ShaderVertexInputInfo>) {
+			options.user_data_base = 8;
+			if (stage == ShaderType::Mesh || stage == ShaderType::TessellationControl) {
+				options.user_data_base = 0;
+				options.wave_size = stage == ShaderType::Mesh ? input_info.mesh.wave_size : 64u;
+			}
+		} else {
+			options.wave_size = input_info.wave_size;
+		}
+		auto translated = ShaderRecompiler::TranslateProgram(params.code, options);
+		if (entry == programs.end()) {
+			auto resource_plan = ShaderRecompiler::IR::ExtractResourcePlan(translated.program);
+			EXIT_IF(!ShaderRecompiler::IR::MaterializeResources(resource_plan, runtime, resources,
+			                                                    specialization));
+			entry = programs.try_emplace(lookup_key, std::move(resource_plan)).first;
+		}
+		entry->second.permutations.push_back(CompilePermutation(
+		    params, options, std::move(translated), std::move(specialization), push_data_cursor));
+		const auto& permutation = entry->second.permutations.back();
+		input_info.stage = {.program = &permutation.program, .resources = std::move(resources)};
+		permutation.program.bindings.AdvancePushData(push_data_cursor);
+
+		std::array<size_t, static_cast<size_t>(ShaderType::TessellationEvaluation) + 1> counts {};
+		for (const auto& [key, source]: programs) {
+			counts[static_cast<size_t>(key.stage)] += source.permutations.size();
+		}
+		// Guest geometry shaders are compiled through the host mesh stage.
+		std::printf("Shaders: VS %zu | PS %zu | CS %zu | GS %zu | LS %zu | HS %zu | TES %zu\n",
+		            counts[static_cast<size_t>(ShaderType::Vertex)],
+		            counts[static_cast<size_t>(ShaderType::Pixel)],
+		            counts[static_cast<size_t>(ShaderType::Compute)],
+		            counts[static_cast<size_t>(ShaderType::Mesh)],
+		            counts[static_cast<size_t>(ShaderType::Local)],
+		            counts[static_cast<size_t>(ShaderType::TessellationControl)],
+		            counts[static_cast<size_t>(ShaderType::TessellationEvaluation)]);
+		return permutation.handle;
 	}
 
 	explicit ProgramCache(vk::Device device): device(device) {
-		key_scratch.reserve(MaxStaticKeyWords);
+		lookup_key.static_state.reserve(MaxStaticKeyWords);
+	}
+	~ProgramCache() {
+		for (const auto& [key, entry]: programs) {
+			(void)key;
+			for (const auto& permutation: entry.permutations) {
+				device.destroyShaderModule(permutation.handle.module, nullptr);
+			}
+		}
 	}
 
-	std::unordered_map<SourceKey, std::vector<Permutation>, SourceKeyHash> programs;
-	std::vector<uint32_t> key_scratch;
-	vk::Device device;
-	uint32_t   num_compiled = 0;
+	std::unordered_map<ProgramKey, SourceEntry, ProgramKeyHash> programs;
+	ProgramKey                                                  lookup_key;
+	vk::Device                                                  device;
+	uint64_t                                                    next_shader_id = 0;
 };
 
 PipelineCache::PipelineCache(GraphicContext& graphics)
@@ -181,12 +409,6 @@ PipelineCache::~PipelineCache() {
 	};
 	destroy(m_graphics_pipelines);
 	destroy(m_compute_pipelines);
-	for (const auto& [key, permutations]: m_program_cache->programs) {
-		(void)key;
-		for (const auto& permutation: permutations) {
-			m_graphics.device.destroyShaderModule(permutation.handle.module, nullptr);
-		}
-	}
 	if (m_driver_cache != nullptr) {
 		m_graphics.device.destroyPipelineCache(m_driver_cache, nullptr);
 	}
@@ -212,7 +434,7 @@ void PipelineCache::InitializeDriverCache() {
 		return;
 	}
 
-	m_driver_cache_path = std::filesystem::path("_PipelineCache") / (title_id + ".bin");
+	m_driver_cache_path     = std::filesystem::path("_PipelineCache") / (title_id + ".bin");
 	const auto path         = Common::PathToString(m_driver_cache_path);
 	const bool cache_exists = Common::File::IsFileExisting(m_driver_cache_path);
 	if (cache_exists) {
@@ -224,7 +446,7 @@ void PipelineCache::InitializeDriverCache() {
 	if (cache_exists) {
 		Common::File file(m_driver_cache_path, Common::File::Mode::Read);
 		const auto   file_size = file.IsInvalid() ? 0 : file.Size();
-		const auto signature = DriverCacheSignature(m_graphics.GetPhysicalDeviceProperties());
+		const auto   signature = DriverCacheSignature(m_graphics.GetPhysicalDeviceProperties());
 		if (file_size >= signature.size() + sizeof(uint64_t) &&
 		    file_size <= std::numeric_limits<uint32_t>::max()) {
 			std::string cached_signature(signature.size(), '\0');
@@ -236,7 +458,8 @@ void PipelineCache::InitializeDriverCache() {
 			file.Read(cached_signature.data(), static_cast<uint32_t>(cached_signature.size()),
 			          &signature_read);
 			file.Read(&payload_hash, sizeof(payload_hash), &hash_read);
-			file.Read(initial_data.data(), static_cast<uint32_t>(initial_data.size()), &payload_read);
+			file.Read(initial_data.data(), static_cast<uint32_t>(initial_data.size()),
+			          &payload_read);
 			file.Close();
 			if (signature_read != cached_signature.size() || hash_read != sizeof(payload_hash) ||
 			    payload_read != initial_data.size() || cached_signature != signature ||
@@ -248,26 +471,24 @@ void PipelineCache::InitializeDriverCache() {
 			}
 		} else {
 			file.Close();
-			PipelineCacheLog("Vulkan pipeline cache: invalidating {} (invalid file size)",
-			                 path);
+			PipelineCacheLog("Vulkan pipeline cache: invalidating {} (invalid file size)", path);
 		}
 	}
 
 	vk::PipelineCacheCreateInfo create {};
-	create.sType           = vk::StructureType::ePipelineCacheCreateInfo;
 	create.initialDataSize = initial_data.size();
 	create.pInitialData    = initial_data.empty() ? nullptr : initial_data.data();
 	auto result = m_graphics.device.createPipelineCache(&create, nullptr, &m_driver_cache);
 	if (result != vk::Result::eSuccess && !initial_data.empty()) {
 		PipelineCacheLog("Vulkan pipeline cache: driver rejected {} ({}); starting empty", path,
-		                 VulkanToString(result));
+		                 vk::to_string(result));
 		initial_data.clear();
 		create.initialDataSize = 0;
 		create.pInitialData    = nullptr;
 		result = m_graphics.device.createPipelineCache(&create, nullptr, &m_driver_cache);
 	}
 	if (result != vk::Result::eSuccess) {
-		PipelineCacheLog("Vulkan pipeline cache: disabled ({})", VulkanToString(result));
+		PipelineCacheLog("Vulkan pipeline cache: disabled ({})", vk::to_string(result));
 		m_driver_cache = nullptr;
 		return;
 	}
@@ -304,11 +525,11 @@ void PipelineCache::Save() {
 	if (result != vk::Result::eSuccess || size == 0 ||
 	    size > std::numeric_limits<uint32_t>::max()) {
 		PipelineCacheLog("Vulkan pipeline cache: save failed ({}, {} bytes)",
-		                 VulkanToString(result), size);
+		                 vk::to_string(result), size);
 		return;
 	}
 	payload.resize(size);
-	auto prefix = DriverCacheSignature(m_graphics.GetPhysicalDeviceProperties());
+	auto       prefix       = DriverCacheSignature(m_graphics.GetPhysicalDeviceProperties());
 	const auto payload_hash = XXH3_64bits(payload.data(), payload.size());
 	prefix.append(reinterpret_cast<const char*>(&payload_hash), sizeof(payload_hash));
 	if (!Common::File::CreateDirectories(m_driver_cache_path.parent_path())) {
@@ -338,43 +559,90 @@ void PipelineCache::Save() {
 	m_driver_cache = nullptr;
 }
 
-ShaderProgram PipelineCache::GetVertexProgram(const HW::VertexShaderInfo& regs,
-                                              const HW::ShaderRegisters&  sh,
-                                              ShaderVertexInputInfo&      input_info) {
-	const auto params = PrepareProgram(regs, sh, input_info);
+PipelineCache::GraphicsPrograms PipelineCache::GetGraphicsPrograms(
+    const HW::VertexShaderInfo& vertex_regs, const HW::PixelShaderInfo& pixel_regs,
+    const HW::ShaderRegisters& sh, const HW::Context& context, const HW::UserConfig& user_config,
+    std::span<const Prospero::ColorComponentMapping, 8> target_export_mapping, bool pixel_active,
+    std::array<ShaderVertexInputInfo, 3>& vertex_info, ShaderPixelInputInfo& pixel_info) {
+	const bool tess_active = user_config.GetPrimType() == Prospero::PrimitiveType::kPatch;
+	std::array<ShaderParams, 3> vertex_params;
+	if (tess_active) {
+		vertex_params = PrepareTessellationPrograms(vertex_regs, context, vertex_info);
+	} else {
+		vertex_params[0] = PrepareProgram(vertex_regs, context, user_config, vertex_info[0]);
+	}
+	const bool mesh_active = vertex_info[0].logical_stage == ShaderType::Mesh;
+	if (mesh_active) {
+		EXIT_NOT_IMPLEMENTED(!m_graphics.mesh_shader_enabled);
+		auto& mesh              = vertex_info[0].mesh;
+		mesh.host_subgroup_size = m_graphics.subgroup_size;
+		const auto& limits      = m_graphics.mesh_shader_properties;
+		const auto  logical_threads =
+		    mesh.threads_num[0] * mesh.threads_num[1] * mesh.threads_num[2];
+		const auto host_threads = ((logical_threads + mesh.wave_size - 1u) / mesh.wave_size) *
+		                          std::min(mesh.host_subgroup_size, mesh.wave_size);
+		if (host_threads > limits.maxMeshWorkGroupInvocations ||
+		    host_threads > limits.maxMeshWorkGroupSize[0] ||
+		    mesh.max_vertices > limits.maxMeshOutputVertices ||
+		    mesh.max_primitives > limits.maxMeshOutputPrimitives ||
+		    mesh.lds_size_dwords * sizeof(uint32_t) > limits.maxMeshSharedMemorySize) {
+			EXIT("mesh shader exceeds host limits: threads=%u vertices=%u primitives=%u LDS=%u\n",
+			     host_threads, mesh.max_vertices, mesh.max_primitives, mesh.lds_size_dwords);
+		}
+	}
+	ShaderParams pixel_params;
+	if (pixel_active) {
+		pixel_params = PrepareProgram(pixel_regs, sh, target_export_mapping, pixel_info);
+	}
+	if (context.GetClipControl().clip_disable) {
+		const auto& viewport = context.GetScreenViewport().viewports[0];
+		const auto& limits   = m_graphics.GetPhysicalDeviceProperties().limits;
+		auto&       clip     = vertex_info[tess_active ? 2u : 0u].clip_space;
+		clip.scale[0]        = viewport.xscale;
+		clip.scale[1]        = viewport.yscale;
+		clip.offset[0]       = viewport.xoffset;
+		clip.offset[1]       = viewport.yoffset;
+		clip.half_extent[0] =
+		    static_cast<float>(std::min(limits.maxViewportDimensions[0], 16384u)) * 0.5f;
+		clip.half_extent[1] =
+		    static_cast<float>(std::min(limits.maxViewportDimensions[1], 16384u)) * 0.5f;
+		clip.enabled = true;
+	}
 	Common::LockGuard lock(m_mutex);
-	return m_program_cache->Get(params, input_info);
-}
-
-ShaderProgram PipelineCache::GetPixelProgram(
-    const HW::PixelShaderInfo& regs, const HW::ShaderRegisters& sh,
-    const ShaderVertexInputInfo&                        vertex_info,
-    std::span<const Prospero::ColorComponentMapping, 8> target_export_mapping,
-    ShaderPixelInputInfo&                               input_info) {
-	const auto params = PrepareProgram(regs, sh, vertex_info, target_export_mapping, input_info);
-	Common::LockGuard lock(m_mutex);
-	return m_program_cache->Get(params, input_info);
+	uint32_t          push_data_cursor =
+	    mesh_active ? ShaderRecompiler::IR::PushData::MeshDrawDwordCount : 0;
+	GraphicsPrograms  result;
+	if (pixel_active) {
+		result.pixel = m_program_cache->Get(pixel_params, pixel_info, push_data_cursor);
+	}
+	for (uint32_t i = 0; i < (tess_active ? 3u : 1u); i++) {
+		result.vertex[i] = m_program_cache->Get(vertex_params[i], vertex_info[i], push_data_cursor);
+	}
+	return result;
 }
 
 ShaderProgram PipelineCache::GetComputeProgram(const HW::ComputeShaderInfo& regs,
                                                const HW::ShaderRegisters&   sh,
                                                ShaderComputeInputInfo&      input_info) {
-	input_info.needs_lds_barriers = !m_graphics.compute_wave64_supported;
-	const auto params             = PrepareProgram(regs, sh, input_info);
+	input_info.host_subgroup_size = m_graphics.SupportsComputeWave64() ? 64u : 32u;
+	const auto        params      = PrepareProgram(regs, sh, input_info);
 	Common::LockGuard lock(m_mutex);
-	return m_program_cache->Get(params, input_info);
+	uint32_t          push_data_cursor = 0;
+	return m_program_cache->Get(params, input_info, push_data_cursor);
 }
 
 bool PipelineStaticParameters::operator==(const PipelineStaticParameters& other) const noexcept {
 	return std::memcmp(this, &other, sizeof(*this)) == 0;
 }
 
-PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
+PipelineCache::Pipeline& PipelineCache::GetGraphicsPipeline(
     std::span<const RenderColorInfo> colors, const RenderDepthInfo& depth,
-    const ShaderVertexInputInfo& vs_input_info, CommandBuffer& command,
+    std::span<const ShaderVertexInputInfo> vertex_info, CommandBuffer& command,
     const ShaderPixelInputInfo* ps_input_info, vk::PrimitiveTopology topology,
-    bool primitive_restart_enable, const ShaderProgram& vertex_program,
-    const ShaderProgram& pixel_program) {
+    bool primitive_restart_enable, const GraphicsPrograms& programs) {
+	const auto& vs_input_info  = vertex_info.front();
+	const auto& vertex_program = programs.vertex[0];
+	const auto& pixel_program  = programs.pixel;
 	KYTY_PROFILER_BLOCK("PipelineCache::CreatePipeline(Gfx)", profiler::colors::DeepOrangeA200);
 
 	EXIT_IF(colors.size() > RENDER_COLOR_ATTACHMENTS_MAX);
@@ -386,51 +654,60 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 	Common::LockGuard lock(m_mutex);
 	auto&             ctx = command.GetRegisters();
 
-	const HW::BlendColor& bclr                                     = ctx.GetBlendColor();
-	uint32_t              color_mask[RENDER_COLOR_ATTACHMENTS_MAX] = {};
-	for (uint32_t i = 0; i < color_count; i++) {
-		color_mask[i] =
-		    (colors[i].image_id ? colors[i].export_mapping.ApplyMask(render_target_mask_slot(
-		                              ctx.GetRenderTargetMask(), colors[i].target_slot))
-		                        : 0);
-	}
 	const HW::ModeControl& mc = ctx.GetModeControl();
 
 	const auto vs_id = vertex_program.id;
 	const auto ps_id = ps_active ? pixel_program.id : 0;
 
-	PipelineStaticParameters static_params {};
-	GraphicsPipeline         p {};
-	p.ps_shader_id = ps_id;
-	p.vs_shader_id = vs_id;
-
-	static_params.color_count = color_count;
-	PipelineRenderingState rendering {};
-	rendering.color_count       = color_count;
+	GraphicsPipelineKey key {};
+	for (uint32_t i = 0; i < programs.vertex.size(); i++) {
+		key.vertex_shader_ids[i] = programs.vertex[i].id;
+	}
+	key.ps_shader_id            = ps_id;
+	auto& static_params         = key.static_params;
+	auto& rendering             = key.rendering;
+	rendering.color_count       = 0;
 	uint32_t attachment_samples = 0;
 	for (uint32_t i = 0; i < color_count; i++) {
-		EXIT_IF(!colors[i].image_id || colors[i].format == vk::Format::eUndefined);
-		rendering.color_formats[i] = colors[i].format;
+		const auto slot = colors[i].target_slot;
+		EXIT_IF(slot >= RENDER_COLOR_ATTACHMENTS_MAX);
+		rendering.color_count = std::max(rendering.color_count, slot + 1);
+		EXIT_IF(!colors[i].image_id || colors[i].desc.view_info.format == vk::Format::eUndefined);
+		static_params.color_mask[slot] = colors[i].export_mapping.ApplyMask(
+		    render_target_mask_slot(ctx.GetRenderTargetMask(), colors[i].target_slot));
+		rendering.color_formats[slot] = colors[i].desc.view_info.format;
 		if (attachment_samples == 0) {
-			attachment_samples = colors[i].samples;
-		} else if (attachment_samples != colors[i].samples) {
+			attachment_samples = colors[i].desc.info.samples;
+		} else if (attachment_samples != colors[i].desc.info.samples) {
 			EXIT("mixed color attachment sample counts are unsupported: %u and %u\n",
-			     attachment_samples, colors[i].samples);
+			     attachment_samples, colors[i].desc.info.samples);
 		}
+		const auto& rt                        = ctx.GetRenderTarget(colors[i].target_slot);
+		const auto& bc                        = ctx.GetBlendControl(colors[i].target_slot);
+		static_params.color_srcblend[slot]       = bc.color_srcblend;
+		static_params.color_comb_fcn[slot]       = bc.color_comb_fcn;
+		static_params.color_destblend[slot]      = bc.color_destblend;
+		static_params.alpha_srcblend[slot]       = bc.alpha_srcblend;
+		static_params.alpha_comb_fcn[slot]       = bc.alpha_comb_fcn;
+		static_params.alpha_destblend[slot]      = bc.alpha_destblend;
+		static_params.separate_alpha_blend[slot] = bc.separate_alpha_blend;
+		static_params.blend_enable[slot]         = bc.enable && !rt.info.blend_bypass;
 	}
 	const bool with_depth =
-	    depth.format != vk::Format::eUndefined && static_cast<bool>(depth.image_id);
+	    depth.desc.view_info.format != vk::Format::eUndefined && static_cast<bool>(depth.image_id);
 	if (with_depth) {
-		const auto aspects = ImageViewOps::DepthAspectMask(depth.format);
-		rendering.depth_format =
-		    aspects & vk::ImageAspectFlagBits::eDepth ? depth.format : vk::Format::eUndefined;
-		rendering.stencil_format =
-		    aspects & vk::ImageAspectFlagBits::eStencil ? depth.format : vk::Format::eUndefined;
+		const auto aspects       = ImageViewOps::DepthAspectMask(depth.desc.view_info.format);
+		rendering.depth_format   = aspects & vk::ImageAspectFlagBits::eDepth
+		                               ? depth.desc.view_info.format
+		                               : vk::Format::eUndefined;
+		rendering.stencil_format = aspects & vk::ImageAspectFlagBits::eStencil
+		                               ? depth.desc.view_info.format
+		                               : vk::Format::eUndefined;
 		if (attachment_samples == 0) {
-			attachment_samples = depth.samples;
-		} else if (attachment_samples != depth.samples) {
+			attachment_samples = depth.desc.info.samples;
+		} else if (attachment_samples != depth.desc.info.samples) {
 			EXIT("mixed color/depth sample counts are unsupported: %u and %u\n", attachment_samples,
-			     depth.samples);
+			     depth.desc.info.samples);
 		}
 	}
 	if (color_count == 0 && !with_depth) {
@@ -460,49 +737,44 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 	if (static_params.sample_shading_enable && !m_graphics.sample_rate_shading_enabled) {
 		EXIT("Pipeline: sample-rate shading is required but unsupported by the host\n");
 	}
-	static_params.with_depth         = with_depth;
-	static_params.depth_test_enable  = depth.depth_test_enable;
-	static_params.depth_write_enable = (depth.depth_write_enable && !depth.depth_clear_enable);
-	static_params.depth_compare_op   = depth.depth_compare_op;
 	static_params.depth_bounds_test_enable = depth.depth_bounds_test_enable;
 	static_params.depth_min_bounds         = depth.depth_min_bounds;
 	static_params.depth_max_bounds         = depth.depth_max_bounds;
-	static_params.stencil_test_enable      = depth.stencil_test_enable;
-	static_params.stencil_front            = depth.stencil_static_front;
-	static_params.stencil_back             = depth.stencil_static_back;
-	for (uint32_t i = 0; i < RENDER_COLOR_ATTACHMENTS_MAX; i++) {
-		static_params.color_mask[i] = color_mask[i];
-	}
-	const bool rect_list     = topology == vk::PrimitiveTopology::ePatchList;
+	const bool rect_list =
+	    command.GetUserConfig().GetPrimType() == Prospero::PrimitiveType::kRectList;
 	static_params.cull_back  = !rect_list && mc.cull_back;
 	static_params.cull_front = !rect_list && mc.cull_front;
 	static_params.face       = mc.face;
+	static_params.provoking_vtx_last = mc.provoking_vtx_last;
+	static_params.polygon_mode =
+	    ResolvePolygonMode(mc, static_params.cull_front, static_params.cull_back);
 
-	for (uint32_t i = 0; i < color_count; i++) {
-		const auto& rt                        = ctx.GetRenderTarget(colors[i].target_slot);
-		const auto& bc                        = ctx.GetBlendControl(colors[i].target_slot);
-		static_params.color_srcblend[i]       = bc.color_srcblend;
-		static_params.color_comb_fcn[i]       = bc.color_comb_fcn;
-		static_params.color_destblend[i]      = bc.color_destblend;
-		static_params.alpha_srcblend[i]       = bc.alpha_srcblend;
-		static_params.alpha_comb_fcn[i]       = bc.alpha_comb_fcn;
-		static_params.alpha_destblend[i]      = bc.alpha_destblend;
-		static_params.separate_alpha_blend[i] = bc.separate_alpha_blend;
-		static_params.blend_enable[i]         = bc.enable;
-		static_params.blend_bypass[i]         = rt.info.blend_bypass;
+	if (vs_input_info.stage.program->stage != ShaderType::Mesh) {
+		EXIT_IF(vs_input_info.buffers_num < 0 ||
+		        vs_input_info.buffers_num > ShaderVertexInputInfo::RES_MAX ||
+		        vs_input_info.resources_num < 0 ||
+		        vs_input_info.resources_num > ShaderVertexInputInfo::RES_MAX);
+		key.vertex_input.binding_count   = static_cast<uint8_t>(vs_input_info.buffers_num);
+		key.vertex_input.attribute_count = static_cast<uint8_t>(vs_input_info.resources_num);
+		uint32_t attributes_num          = 0;
+		for (int binding = 0; binding < vs_input_info.buffers_num; binding++) {
+			const auto& buffer = vs_input_info.buffers[binding];
+			EXIT_IF(buffer.attr_num < 0 || buffer.attr_num > ShaderVertexInputBuffer::ATTR_MAX);
+			attributes_num += static_cast<uint32_t>(buffer.attr_num);
+			EXIT_IF(attributes_num > static_cast<uint32_t>(vs_input_info.resources_num));
+			key.vertex_input.bindings[binding] = {.stride   = buffer.stride,
+			                                      .instance = buffer.fetch_index != 0};
+			for (int attribute = 0; attribute < buffer.attr_num; attribute++) {
+				const auto index = buffer.attr_indices[attribute];
+				EXIT_IF(index < 0 || index >= vs_input_info.resources_num);
+				key.vertex_input.attributes[index] = {
+				    .offset  = buffer.attr_offsets[attribute],
+				    .binding = static_cast<uint8_t>(binding),
+				};
+			}
+		}
+		EXIT_IF(attributes_num != static_cast<uint32_t>(vs_input_info.resources_num));
 	}
-	static_params.blend_color_red   = bclr.red;
-	static_params.blend_color_green = bclr.green;
-	static_params.blend_color_blue  = bclr.blue;
-	static_params.blend_color_alpha = bclr.alpha;
-
-	NormalizeStaticParamsForDynamicState(static_params);
-
-	GraphicsPipelineKey key {};
-	key.rendering     = rendering;
-	key.vs_shader_id  = p.vs_shader_id;
-	key.ps_shader_id  = p.ps_shader_id;
-	key.static_params = static_params;
 
 	if (auto iter = m_graphics_pipelines.find(key); iter != m_graphics_pipelines.end()) {
 		return *iter->second;
@@ -513,16 +785,15 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 		if (ps_active) {
 			ShaderDbgDumpInputInfo(*ps_input_info);
 		}
-		LOGF("PipelineTrace: shader modules VS=%" PRIu64 " module=%p PS=%" PRIu64
-		     " module=%p\n",
+		LOGF("PipelineTrace: shader modules VS=%" PRIu64 " module=%p PS=%" PRIu64 " module=%p\n",
 		     vs_id, static_cast<void*>(vertex_program.module), ps_id,
 		     static_cast<void*>(pixel_program.module));
 	}
 
-	auto cached = std::make_unique<GraphicsPipeline>(p);
+	auto cached = std::make_unique<Pipeline>();
 	LogPipelineTrace("CreatePipelineInternal begin", vs_id, ps_id);
-	CreatePipelineInternal(m_graphics, *cached, rendering, vs_input_info, vertex_program.module,
-	                       ps_input_info, pixel_program.module, static_params, m_driver_cache);
+	CreatePipelineInternal(m_graphics, *cached, rendering, key.vertex_input, vertex_info,
+	                       ps_input_info, programs, static_params, m_driver_cache);
 	LogPipelineTrace("CreatePipelineInternal done", vs_id, ps_id);
 
 	EXIT_NOT_IMPLEMENTED(cached->pipeline == nullptr);
@@ -534,22 +805,17 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 	return *iter->second;
 }
 
-PipelineCache::ComputePipeline&
-PipelineCache::CreateComputePipeline(ShaderComputeInputInfo& input_info,
-                                     const ShaderProgram&    compute_program) {
+PipelineCache::Pipeline&
+PipelineCache::GetComputePipeline(const ShaderComputeInputInfo& input_info,
+                                  const ShaderProgram&          compute_program) {
 	KYTY_PROFILER_BLOCK("PipelineCache::CreatePipeline(Compute)", profiler::colors::RedA100);
 
 	EXIT_IF(!compute_program);
 
 	Common::LockGuard lock(m_mutex);
 
-	ComputePipeline p {};
-	p.cs_shader_id = compute_program.id;
-
-	ComputePipelineKey key {};
-	key.cs_shader_id = p.cs_shader_id;
-
-	if (auto iter = m_compute_pipelines.find(key); iter != m_compute_pipelines.end()) {
+	if (auto iter = m_compute_pipelines.find(compute_program.id);
+	    iter != m_compute_pipelines.end()) {
 		return *iter->second;
 	}
 
@@ -557,13 +823,13 @@ PipelineCache::CreateComputePipeline(ShaderComputeInputInfo& input_info,
 		ShaderDbgDumpInputInfo(input_info);
 	}
 
-	auto cached = std::make_unique<ComputePipeline>(p);
+	auto cached = std::make_unique<Pipeline>();
 	CreatePipelineInternal(m_graphics, *cached, input_info, compute_program.module, m_driver_cache);
 
 	EXIT_NOT_IMPLEMENTED(cached->pipeline == nullptr);
 	EXIT_NOT_IMPLEMENTED(cached->pipeline_layout == nullptr);
 
-	auto [iter, inserted] = m_compute_pipelines.emplace(std::move(key), std::move(cached));
+	auto [iter, inserted] = m_compute_pipelines.emplace(compute_program.id, std::move(cached));
 	EXIT_IF(!inserted);
 
 	return *iter->second;

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

@@ -26,36 +26,28 @@ class CommandBuffer;
 namespace HW {
 class Context;
 class Shader;
+class UserConfig;
 struct ComputeShaderInfo;
 } // namespace HW
 
 #pragma pack(push, 1)
 
 struct PipelineStaticParameters {
-	float                      viewport_scale[3]        = {};
-	float                      viewport_offset[3]       = {};
 	bool                       negative_one_to_one      = false;
 	bool                       depth_clip_enable        = true;
-	int                        scissor_ltrb[4]          = {0};
 	vk::PrimitiveTopology      topology                 = vk::PrimitiveTopology::ePointList;
 	bool                       primitive_restart_enable = false;
 	uint32_t                   samples                  = 1;
 	bool                       sample_shading_enable    = false;
-	bool                       with_depth               = false;
-	bool                       depth_test_enable        = false;
-	bool                       depth_write_enable       = false;
-	vk::CompareOp              depth_compare_op         = vk::CompareOp::eNever;
 	bool                       depth_bounds_test_enable = false;
 	float                      depth_min_bounds         = 0.0f;
 	float                      depth_max_bounds         = 0.0f;
-	bool                       stencil_test_enable      = false;
-	PipelineStencilStaticState stencil_front;
-	PipelineStencilStaticState stencil_back;
-	uint32_t                   color_count                                        = 1;
 	uint32_t                   color_mask[RENDER_COLOR_ATTACHMENTS_MAX]           = {};
 	bool                       cull_front                                         = false;
 	bool                       cull_back                                          = false;
 	bool                       face                                               = false;
+	bool                       provoking_vtx_last                                 = false;
+	vk::PolygonMode            polygon_mode                                       = vk::PolygonMode::eFill;
 	uint8_t                    color_srcblend[RENDER_COLOR_ATTACHMENTS_MAX]       = {};
 	uint8_t                    color_comb_fcn[RENDER_COLOR_ATTACHMENTS_MAX]       = {};
 	uint8_t                    color_destblend[RENDER_COLOR_ATTACHMENTS_MAX]      = {};
@@ -64,11 +56,6 @@ struct PipelineStaticParameters {
 	uint8_t                    alpha_destblend[RENDER_COLOR_ATTACHMENTS_MAX]      = {};
 	bool                       separate_alpha_blend[RENDER_COLOR_ATTACHMENTS_MAX] = {};
 	bool                       blend_enable[RENDER_COLOR_ATTACHMENTS_MAX]         = {};
-	bool                       blend_bypass[RENDER_COLOR_ATTACHMENTS_MAX]         = {};
-	float                      blend_color_red                                    = 0.0f;
-	float                      blend_color_green                                  = 0.0f;
-	float                      blend_color_blue                                   = 0.0f;
-	float                      blend_color_alpha                                  = 0.0f;
 
 	bool operator==(const PipelineStaticParameters& other) const noexcept;
 };
@@ -78,14 +65,7 @@ struct PipelineStaticParameters {
 static_assert(std::is_trivially_copyable_v<PipelineStaticParameters>);
 static_assert(std::is_standard_layout_v<PipelineStaticParameters>);
 static_assert(alignof(PipelineStaticParameters) == 1);
-static_assert(sizeof(PipelineStaticParameters) ==
-              sizeof(float[3]) + sizeof(float[3]) + sizeof(bool) * 2 + sizeof(int[4]) +
-                  sizeof(vk::PrimitiveTopology) + sizeof(bool) + sizeof(uint32_t) +
-                  sizeof(bool) * 4 + sizeof(vk::CompareOp) + sizeof(bool) + sizeof(float) * 2 +
-                  sizeof(bool) + sizeof(PipelineStencilStaticState) * 2 + sizeof(uint32_t) +
-                  sizeof(uint32_t[RENDER_COLOR_ATTACHMENTS_MAX]) + sizeof(bool) * 3 +
-                  sizeof(uint8_t[RENDER_COLOR_ATTACHMENTS_MAX]) * 6 +
-                  sizeof(bool[RENDER_COLOR_ATTACHMENTS_MAX]) * 3 + sizeof(float) * 4);
+static_assert(sizeof(PipelineStaticParameters) == 125);
 
 struct PipelineRenderingState {
 	std::array<vk::Format, RENDER_COLOR_ATTACHMENTS_MAX> color_formats {};
@@ -96,13 +76,31 @@ struct PipelineRenderingState {
 	bool operator==(const PipelineRenderingState&) const = default;
 };
 
+struct PipelineVertexInputState {
+	struct Binding {
+		uint32_t stride                           = 0;
+		bool     instance                         = false;
+		bool     operator==(const Binding&) const = default;
+	};
+	struct Attribute {
+		uint32_t offset                             = 0;
+		uint8_t  binding                            = 0;
+		bool     operator==(const Attribute&) const = default;
+	};
+
+	std::array<Binding, ShaderVertexInputInfo::RES_MAX>   bindings {};
+	std::array<Attribute, ShaderVertexInputInfo::RES_MAX> attributes {};
+	uint8_t                                               binding_count   = 0;
+	uint8_t                                               attribute_count = 0;
+
+	bool operator==(const PipelineVertexInputState&) const = default;
+};
+
 struct ShaderProgram {
 	uint64_t         id     = 0;
 	vk::ShaderModule module = nullptr;
 
-	explicit operator bool() const {
-		return id != 0 && module != nullptr;
-	}
+	explicit operator bool() const { return id != 0 && module != nullptr; }
 };
 
 class PipelineCache {
@@ -119,56 +117,47 @@ public:
 		bool                    uses_push_descriptors = false;
 	};
 
-	struct GraphicsPipeline: Pipeline {
-		uint64_t vs_shader_id = 0;
-		uint64_t ps_shader_id = 0;
+	struct GraphicsPrograms {
+		std::array<ShaderProgram, 3> vertex;
+		ShaderProgram pixel;
+
+		[[nodiscard]] uint32_t VertexStageCount() const { return vertex[1] ? 3u : 1u; }
 	};
 
-	struct ComputePipeline: Pipeline {
-		uint64_t cs_shader_id = 0;
-	};
-
-	ShaderProgram GetVertexProgram(const HW::VertexShaderInfo& regs,
-	                               const HW::ShaderRegisters&  sh,
-	                               ShaderVertexInputInfo&      input_info);
-	ShaderProgram
-	GetPixelProgram(const HW::PixelShaderInfo& regs, const HW::ShaderRegisters& sh,
-	                const ShaderVertexInputInfo&                        vertex_info,
-	                std::span<const Prospero::ColorComponentMapping, 8> target_export_mapping,
-	                ShaderPixelInputInfo&                               input_info);
+	GraphicsPrograms
+	GetGraphicsPrograms(const HW::VertexShaderInfo& vertex_regs,
+	                    const HW::PixelShaderInfo& pixel_regs, const HW::ShaderRegisters& sh,
+	                    const HW::Context& context, const HW::UserConfig& user_config,
+	                    std::span<const Prospero::ColorComponentMapping, 8> target_export_mapping,
+	                    bool pixel_active, std::array<ShaderVertexInputInfo, 3>& vertex_info,
+	                    ShaderPixelInputInfo& pixel_info);
 	ShaderProgram GetComputeProgram(const HW::ComputeShaderInfo& regs,
-	                                const HW::ShaderRegisters& sh,
-	                                ShaderComputeInputInfo& input_info);
+	                                const HW::ShaderRegisters&   sh,
+	                                ShaderComputeInputInfo&      input_info);
 
-	GraphicsPipeline& CreateGraphicsPipeline(
-	    std::span<const RenderColorInfo> colors, const RenderDepthInfo& depth,
-	    const ShaderVertexInputInfo& vs_input_info, CommandBuffer& command,
-	    const ShaderPixelInputInfo* ps_input_info, vk::PrimitiveTopology topology,
-	    bool primitive_restart_enable, const ShaderProgram& vertex_program,
-	    const ShaderProgram& pixel_program);
-	ComputePipeline& CreateComputePipeline(ShaderComputeInputInfo& input_info,
-	                                       const ShaderProgram&    compute_program);
+	Pipeline& GetGraphicsPipeline(std::span<const RenderColorInfo>       colors,
+	                              const RenderDepthInfo&                 depth,
+	                              std::span<const ShaderVertexInputInfo> vertex_info,
+	                              CommandBuffer& command, const ShaderPixelInputInfo* ps_input_info,
+	                              vk::PrimitiveTopology topology, bool primitive_restart_enable,
+	                              const GraphicsPrograms& programs);
+	Pipeline& GetComputePipeline(const ShaderComputeInputInfo& input_info,
+	                             const ShaderProgram&          compute_program);
 
 private:
 	struct ProgramCache;
 
 	struct GraphicsPipelineKey {
 		PipelineRenderingState   rendering;
-		uint64_t                 vs_shader_id = 0;
+		std::array<uint64_t, 3>  vertex_shader_ids {};
 		uint64_t                 ps_shader_id = 0;
+		PipelineVertexInputState vertex_input;
 		PipelineStaticParameters static_params;
 
 		bool operator==(const GraphicsPipelineKey& other) const {
-			return rendering == other.rendering && vs_shader_id == other.vs_shader_id &&
-			       ps_shader_id == other.ps_shader_id && static_params == other.static_params;
-		}
-	};
-
-	struct ComputePipelineKey {
-		uint64_t cs_shader_id = 0;
-
-		bool operator==(const ComputePipelineKey& other) const {
-			return cs_shader_id == other.cs_shader_id;
+			return rendering == other.rendering && vertex_shader_ids == other.vertex_shader_ids &&
+			       ps_shader_id == other.ps_shader_id && vertex_input == other.vertex_input &&
+			       static_params == other.static_params;
 		}
 	};
 
@@ -199,45 +188,49 @@ private:
 		std::size_t operator()(const GraphicsPipelineKey& key) const {
 			std::size_t hash = 0;
 			PipelineKeyHash::MixRendering(hash, key.rendering);
-			PipelineKeyHash::Mix(hash, key.vs_shader_id);
+			for (const auto id: key.vertex_shader_ids) {
+				PipelineKeyHash::Mix(hash, id);
+			}
 			PipelineKeyHash::Mix(hash, key.ps_shader_id);
+			PipelineKeyHash::Mix(hash, key.vertex_input.binding_count);
+			for (uint32_t i = 0; i < key.vertex_input.binding_count; i++) {
+				PipelineKeyHash::Mix(hash, key.vertex_input.bindings[i].stride);
+				PipelineKeyHash::Mix(hash, key.vertex_input.bindings[i].instance);
+			}
+			PipelineKeyHash::Mix(hash, key.vertex_input.attribute_count);
+			for (uint32_t i = 0; i < key.vertex_input.attribute_count; i++) {
+				PipelineKeyHash::Mix(hash, key.vertex_input.attributes[i].offset);
+				PipelineKeyHash::Mix(hash, key.vertex_input.attributes[i].binding);
+			}
 			PipelineKeyHash::MixStaticParams(hash, key.static_params);
 			return hash;
 		}
 	};
 
-	struct ComputePipelineKeyHash {
-		std::size_t operator()(const ComputePipelineKey& key) const {
-			std::size_t hash = 0;
-			PipelineKeyHash::Mix(hash, key.cs_shader_id);
-			return hash;
-		}
-	};
-
-	GraphicContext&                m_graphics;
+	GraphicContext&               m_graphics;
 	std::unique_ptr<ProgramCache> m_program_cache;
 	vk::PipelineCache             m_driver_cache = nullptr;
 	std::filesystem::path         m_driver_cache_path;
-	std::unordered_map<GraphicsPipelineKey, std::unique_ptr<GraphicsPipeline>,
-	                   GraphicsPipelineKeyHash>
-	    m_graphics_pipelines;
-	std::unordered_map<ComputePipelineKey, std::unique_ptr<ComputePipeline>, ComputePipelineKeyHash>
-	              m_compute_pipelines;
+	std::unordered_map<GraphicsPipelineKey, std::unique_ptr<Pipeline>, GraphicsPipelineKeyHash>
+	                                                        m_graphics_pipelines;
+	std::unordered_map<uint64_t, std::unique_ptr<Pipeline>> m_compute_pipelines;
 	Common::Mutex m_mutex;
 
 	void InitializeDriverCache();
 };
 
 void LogPipelineTrace(const char* phase, uint64_t vertex_program_id, uint64_t pixel_program_id);
-void CreatePipelineInternal(
-    GraphicContext& graphics, PipelineCache::GraphicsPipeline& pipeline,
-    const PipelineRenderingState& rendering, const ShaderVertexInputInfo& vs_input_info,
-    vk::ShaderModule vertex_module, const ShaderPixelInputInfo* ps_input_info,
-	vk::ShaderModule pixel_module, const PipelineStaticParameters& static_params,
-	vk::PipelineCache driver_cache);
-void CreatePipelineInternal(GraphicContext& graphics, PipelineCache::ComputePipeline& pipeline,
+void CreatePipelineInternal(GraphicContext& graphics, PipelineCache::Pipeline& pipeline,
+                            const PipelineRenderingState&          rendering,
+                            const PipelineVertexInputState&        vertex_input,
+                            std::span<const ShaderVertexInputInfo> vertex_info,
+                            const ShaderPixelInputInfo*            ps_input_info,
+                            const PipelineCache::GraphicsPrograms& programs,
+                            const PipelineStaticParameters&        static_params,
+                            vk::PipelineCache                      driver_cache);
+void CreatePipelineInternal(GraphicContext& graphics, PipelineCache::Pipeline& pipeline,
                             const ShaderComputeInputInfo& input_info,
-	vk::ShaderModule compute_module, vk::PipelineCache driver_cache);
+                            vk::ShaderModule compute_module, vk::PipelineCache driver_cache);
 
 } // namespace Libs::Graphics
 

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

@@ -8,11 +8,34 @@
 
 namespace Libs::Graphics {
 
+vk::ShaderStageFlagBits NativeShaderStage(ShaderType stage) {
+	switch (stage) {
+		case ShaderType::Local:
+		case ShaderType::Vertex: return vk::ShaderStageFlagBits::eVertex;
+		case ShaderType::Mesh: return vk::ShaderStageFlagBits::eMeshEXT;
+		case ShaderType::TessellationControl: return vk::ShaderStageFlagBits::eTessellationControl;
+		case ShaderType::TessellationEvaluation:
+			return vk::ShaderStageFlagBits::eTessellationEvaluation;
+		case ShaderType::Pixel: return vk::ShaderStageFlagBits::eFragment;
+		case ShaderType::Compute: return vk::ShaderStageFlagBits::eCompute;
+		default: EXIT("unknown native shader stage\n");
+	}
+}
+
 vk::PipelineStageFlags ShaderPipelineStages(vk::ShaderStageFlags stages) {
 	vk::PipelineStageFlags result = {};
 	if (stages & vk::ShaderStageFlagBits::eVertex) {
 		result |= vk::PipelineStageFlagBits::eVertexShader;
 	}
+	if (stages & vk::ShaderStageFlagBits::eMeshEXT) {
+		result |= vk::PipelineStageFlagBits::eMeshShaderEXT;
+	}
+	if (stages & vk::ShaderStageFlagBits::eTessellationControl) {
+		result |= vk::PipelineStageFlagBits::eTessellationControlShader;
+	}
+	if (stages & vk::ShaderStageFlagBits::eTessellationEvaluation) {
+		result |= vk::PipelineStageFlagBits::eTessellationEvaluationShader;
+	}
 	if (stages & vk::ShaderStageFlagBits::eFragment) {
 		result |= vk::PipelineStageFlagBits::eFragmentShader;
 	}
@@ -25,7 +48,6 @@ vk::PipelineStageFlags ShaderPipelineStages(vk::ShaderStageFlags stages) {
 
 VulkanMemoryBarrier MakeShaderWriteDependency() {
 	VulkanMemoryBarrier barrier {};
-	barrier.sType         = vk::StructureType::eMemoryBarrier;
 	barrier.srcAccessMask = vk::AccessFlagBits::eShaderWrite;
 	barrier.dstAccessMask = vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite |
 	                        vk::AccessFlagBits::eVertexAttributeRead |
@@ -38,7 +60,6 @@ VulkanMemoryBarrier MakeShaderWriteDependency() {
 
 VulkanMemoryBarrier MakeShaderAccessDependency() {
 	VulkanMemoryBarrier barrier {};
-	barrier.sType         = vk::StructureType::eMemoryBarrier;
 	barrier.srcAccessMask = vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite;
 	barrier.dstAccessMask = vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite;
 	return barrier;
@@ -46,7 +67,6 @@ VulkanMemoryBarrier MakeShaderAccessDependency() {
 
 VulkanMemoryBarrier MakeShaderWriteHazardDependency() {
 	VulkanMemoryBarrier barrier {};
-	barrier.sType         = vk::StructureType::eMemoryBarrier;
 	barrier.srcAccessMask = vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite;
 	barrier.dstAccessMask = vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite;
 	return barrier;
@@ -56,7 +76,6 @@ vk::BufferMemoryBarrier MakeGdsDependency(vk::Buffer buffer) {
 	EXIT_IF(buffer == nullptr);
 
 	vk::BufferMemoryBarrier barrier {};
-	barrier.sType         = vk::StructureType::eBufferMemoryBarrier;
 	barrier.srcAccessMask = vk::AccessFlagBits::eHostWrite | vk::AccessFlagBits::eTransferWrite |
 	                        vk::AccessFlagBits::eShaderWrite;
 	barrier.dstAccessMask = vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite;
@@ -68,11 +87,12 @@ vk::BufferMemoryBarrier MakeGdsDependency(vk::Buffer buffer) {
 	return barrier;
 }
 
-std::vector<ShaderBufferWriteRange>
-CollectShaderBufferWrites(const ShaderRecompiler::IR::Program&          program,
-                          const ShaderRecompiler::IR::ResourceSnapshot& resources) {
+bool HasShaderBufferWrites(const ShaderStageRuntime& runtime) {
+	EXIT_IF(!runtime);
+	const auto& program   = *runtime.program;
+	const auto& resources = runtime.resources;
 	EXIT_IF(resources.buffers.size() != program.info.buffers.size());
-	std::vector<ShaderBufferWriteRange> writes;
+	bool has_writes = false;
 	for (uint32_t i = 0; i < program.info.buffers.size(); i++) {
 		if (!program.info.buffers[i].written) {
 			continue;
@@ -81,23 +101,11 @@ CollectShaderBufferWrites(const ShaderRecompiler::IR::Program&          program,
 		EXIT_IF(value.dword_count < 4);
 		ShaderBufferResource descriptor;
 		std::memcpy(descriptor.fields, value.dwords.data(), sizeof(descriptor.fields));
-		const auto address = descriptor.Base48();
-		const auto records = static_cast<uint64_t>(descriptor.NumRecords());
-		const auto stride  = static_cast<uint64_t>(descriptor.Stride());
-		if (stride != 0 && records > UINT64_MAX / stride) {
-			EXIT("shader resource barrier buffer footprint overflow\n");
-		}
-		const auto size = stride == 0 ? records : stride * records;
-		if (address != 0 && size != 0) {
-			writes.push_back({address, size});
-		}
+		// A zero stride means byte addressing. For either addressing mode a nonzero record
+		// count is exactly the condition for a nonempty descriptor range.
+		has_writes |= descriptor.Base48() != 0 && descriptor.NumRecords() != 0;
 	}
-	return writes;
-}
-
-bool HasShaderBufferWrites(const ShaderStageRuntime& runtime) {
-	EXIT_IF(!runtime);
-	return !CollectShaderBufferWrites(*runtime.program, *runtime.resources).empty();
+	return has_writes;
 }
 
 void ShaderAccessBarrier(vk::CommandBuffer vk_buffer, vk::PipelineStageFlags source_stages) {
@@ -118,13 +126,11 @@ void ShaderWriteHazardBarrier(vk::CommandBuffer      vk_buffer,
 void ShaderWriteBarrier(vk::CommandBuffer vk_buffer, vk::PipelineStageFlags source_stages) {
 	EXIT_IF(vk_buffer == nullptr || !source_stages);
 	const auto barrier = MakeShaderWriteDependency();
-	vk_buffer.pipelineBarrier(
-	    source_stages,
-	    vk::PipelineStageFlagBits::eComputeShader | vk::PipelineStageFlagBits::eVertexInput |
-	        vk::PipelineStageFlagBits::eVertexShader | vk::PipelineStageFlagBits::eFragmentShader |
-	        vk::PipelineStageFlagBits::eTransfer |
-	        vk::PipelineStageFlagBits::eColorAttachmentOutput,
-	    vk::DependencyFlags {}, 1, &barrier, 0, nullptr, 0, nullptr);
+	vk_buffer.pipelineBarrier(source_stages,
+	                          vk::PipelineStageFlagBits::eComputeShader |
+	                              vk::PipelineStageFlagBits::eAllGraphics |
+	                              vk::PipelineStageFlagBits::eTransfer,
+	                          vk::DependencyFlags {}, 1, &barrier, 0, nullptr, 0, nullptr);
 }
 
 } // namespace Libs::Graphics

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

@@ -4,27 +4,16 @@
 #include "graphics/host_gpu/graphicContext.h"
 #include "graphics/shader/recompiler/ir/passes/ResourceMaterialization.h"
 
-#include <vector>
-
 namespace Libs::Graphics {
 
 struct ShaderStageRuntime;
 
-struct ShaderBufferWriteRange {
-	uint64_t address = 0;
-	uint64_t size    = 0;
-
-	bool operator==(const ShaderBufferWriteRange& other) const = default;
-};
-
+vk::ShaderStageFlagBits NativeShaderStage(ShaderType stage);
 vk::PipelineStageFlags  ShaderPipelineStages(vk::ShaderStageFlags stages);
 VulkanMemoryBarrier     MakeShaderAccessDependency();
 VulkanMemoryBarrier     MakeShaderWriteHazardDependency();
 VulkanMemoryBarrier     MakeShaderWriteDependency();
 vk::BufferMemoryBarrier MakeGdsDependency(vk::Buffer buffer);
-std::vector<ShaderBufferWriteRange>
-CollectShaderBufferWrites(const ShaderRecompiler::IR::Program&          program,
-                          const ShaderRecompiler::IR::ResourceSnapshot& resources);
 bool HasShaderBufferWrites(const ShaderStageRuntime& runtime);
 void ShaderAccessBarrier(vk::CommandBuffer vk_buffer, vk::PipelineStageFlags source_stages);
 void ShaderWriteHazardBarrier(vk::CommandBuffer      vk_buffer,

src/graphics/host_gpu/renderer/render.h

@@ -48,6 +48,31 @@ enum class CommandBufferDebugOp : uint32_t {
 	Unknown,
 };
 
+enum class DrawOffsetSource : uint8_t {
+	DrawState,
+	IndirectArgs,
+};
+
+struct DrawIndexArgs {
+	uint32_t         index_count                = 0;
+	const void*      index_addr                 = nullptr;
+	uint32_t         instance_count             = 0;
+	uint32_t         index_type_and_size        = 0;
+	int32_t          base_vertex                = 0;
+	uint32_t         first_instance             = 0;
+	DrawOffsetSource offset_source              = DrawOffsetSource::DrawState;
+	uint32_t         render_target_slice_offset = 0;
+};
+
+struct DrawAutoArgs {
+	uint32_t         vertex_count               = 0;
+	uint32_t         instance_count             = 0;
+	uint32_t         first_vertex               = 0;
+	uint32_t         first_instance             = 0;
+	DrawOffsetSource offset_source              = DrawOffsetSource::DrawState;
+	uint32_t         render_target_slice_offset = 0;
+};
+
 struct SubmitInfo {
 	static constexpr uint32_t MaxSemaphores = 3;
 
@@ -129,13 +154,6 @@ public:
 	explicit RenderExecutor(RenderContext& context): m_context(context) {}
 	KYTY_CLASS_NO_COPY(RenderExecutor);
 
-	void DrawIndex(uint64_t submit_id, CommandBuffer& buffer, uint32_t index_type_and_size,
-	               uint32_t index_count, const void* index_addr, uint32_t flags, uint32_t type,
-	               uint32_t instance_count = 1, uint32_t render_target_slice_offset = 0,
-	               int32_t vertex_offset_add = 0, uint32_t first_instance = 0);
-	void DrawAuto(uint64_t submit_id, CommandBuffer& buffer, uint32_t index_count, uint32_t flags,
-	              uint32_t render_target_slice_offset = 0, uint32_t instance_count = 1,
-	              uint32_t first_vertex = 0, uint32_t first_instance = 0);
 	void DispatchDirect(uint64_t submit_id, CommandBuffer& buffer, uint32_t thread_group_x,
 	                    uint32_t thread_group_y, uint32_t thread_group_z, uint32_t mode);
 
@@ -148,41 +166,44 @@ public:
 	                    std::span<PreparedBindings* const> bindings);
 
 private:
+	void DrawIndex(uint64_t submit_id, CommandBuffer& buffer, const DrawIndexArgs& args);
+	void DrawAuto(uint64_t submit_id, CommandBuffer& buffer, const DrawAutoArgs& args);
+
 	struct GraphicsBindings {
-		PreparedBindings                vertex;
+		std::array<PreparedBindings, 3> vertex;
 		std::optional<PreparedBindings> pixel;
 	};
 
 	[[nodiscard]] TextureBinding ResolveTexture(const ShaderRecompiler::IR::ImageResource& resource,
 	                                            const ShaderRecompiler::IR::DescriptorValue& value);
-	[[nodiscard]] GraphicsBindings PrepareGraphicsBindings(const ShaderStageRuntime& vertex,
-	                                                       const ShaderStageRuntime& pixel,
-	                                                       bool                      pixel_active);
-	void ResolveRenderColorTarget(uint64_t submit_id, CommandBuffer& buffer,
-	                              RenderColorInfo& target, uint32_t render_target_slice_offset = 0,
-	                              uint32_t render_target_slot = UINT32_MAX,
+	void PrepareGraphicsBindings(std::span<PreparedBindings* const> stages,
+	                             std::span<RenderColorInfo> colors);
+	void ResolveRenderColorTarget(CommandBuffer& buffer, RenderColorInfo& target,
+	                              uint32_t render_target_slice_offset, uint32_t render_target_slot,
 	                              bool ignore_target_mask = false, bool exact_format = false);
-	void ResolveRenderDepthTarget(uint64_t submit_id, CommandBuffer& buffer,
-	                              RenderDepthInfo& target);
-	[[nodiscard]] bool PrepareDrawRenderState(uint64_t submit_id, CommandBuffer& buffer,
+	void ResolveRenderDepthTarget(CommandBuffer& buffer, RenderDepthInfo& target);
+	[[nodiscard]] bool DepthStencilCopy(CommandBuffer& buffer);
+	[[nodiscard]] bool PrepareDrawRenderState(CommandBuffer& buffer,
 	                                          const DrawCallInfo& draw,
 	                                          uint32_t            render_target_slice_offset,
-	                                          bool log_setup_phases, DrawRenderState& state);
+	                                          DrawRenderState& state);
 	void ExecutePreparedDraw(uint64_t submit_id, CommandBuffer& buffer, const DrawCallInfo& draw,
 	                         DrawRenderState& state, vk::PrimitiveTopology topology,
 	                         const DrawEmitInfo& emit, const DrawIndexBufferSource& index_source,
-	                         bool primitive_restart_enable, bool log_pipeline_phase,
-	                         bool set_bind_debug, bool set_auto_debug);
+	                         bool primitive_restart_enable);
 	[[nodiscard]] RenderState AcquireRenderTargets(CommandBuffer& buffer, RenderColorInfo* colors,
-	                                               uint32_t color_count, RenderDepthInfo& depth);
-	[[nodiscard]] bool        ResolveColorTargets(uint64_t submit_id, CommandBuffer& buffer,
+	                                               uint32_t color_count, RenderDepthInfo& depth,
+	                                               const std::optional<PreparedBindings>& pixel = std::nullopt);
+	[[nodiscard]] bool        ResolveColorTargets(CommandBuffer& buffer,
 	                                              uint32_t render_target_slice_offset);
 	void                      BindImage(ImageId id, bool storage);
 	void                      BindRenderTarget(ImageId id);
-	void                      TrackImageBinding(ImageId id);
 	void                      ResetBindings();
 	[[nodiscard]] bool        TryConsumeComputeMetaClear(const ShaderComputeInputInfo& input,
 	                                                     const CommandBuffer&          buffer);
+	[[nodiscard]] bool TryConsumeComputeImageClear(const ShaderComputeInputInfo& input,
+	                                              CommandBuffer& command, uint32_t group_x,
+	                                              uint32_t group_y, uint32_t group_z, uint32_t mode);
 
 	RenderContext&                        m_context;
 	std::vector<ImageId>                  m_bound_images;
@@ -190,13 +211,12 @@ private:
 	std::vector<vk::DescriptorImageInfo>  m_descriptor_images;
 	std::vector<vk::WriteDescriptorSet>   m_descriptor_writes;
 	std::vector<uint32_t>                 m_image_occurrences;
-	std::array<uint32_t, ShaderRecompiler::IR::NativePushConstantSize / sizeof(uint32_t)>
-	    m_push_constants {};
 
+	friend class CommandProcessor;
 	friend struct RenderExecutorTestAccess;
 };
 
-[[nodiscard]] bool ResolveComputeImageClear(const ShaderComputeInputInfo& input, uint32_t group_x,
+[[nodiscard]] bool ResolveComputeBufferFill(const ShaderComputeInputInfo& input, uint32_t group_x,
                                             uint32_t group_y, uint32_t group_z, uint32_t mode,
                                             ShaderBufferResource& descriptor,
                                             uint32_t& packed_clear, uint64_t& size);

src/graphics/host_gpu/renderer/renderCompute.cpp

@@ -25,28 +25,33 @@
 #include "libs/errno.h"
 
 #include <algorithm>
+#include <array>
 #include <atomic>
 #include <cmath>
 #include <cstring>
 #include <limits>
+#include <mutex>
 #include <span>
 #include <unordered_map>
 #include <vector>
 
 namespace Libs::Graphics {
-static uint64_t BufferDescriptorSize(const ShaderBufferResource& descriptor) {
-	const uint64_t records = descriptor.NumRecords();
-	const uint64_t stride  = descriptor.Stride();
-	if (stride != 0 && records > UINT64_MAX / stride) {
-		EXIT("compute buffer descriptor footprint overflow\n");
+static bool FillSourcesDisjoint(std::span<const ShaderRecompiler::IR::DescriptorValue> sources,
+                                 GuestRange destination, uint32_t output_buffer = UINT32_MAX) {
+	for (uint32_t i = 0; i < sources.size(); ++i) {
+		if (i == output_buffer) continue;
+		const auto source = DecodeNativeDescriptor<ShaderBufferResource>(sources[i]);
+		const auto bytes  = source.GetSize();
+		if (source.Base48() < destination.End() && destination.address < source.Base48() + bytes)
+			return false;
 	}
-	return stride == 0 ? records : records * stride;
+	return true;
 }
 
 bool RenderExecutor::TryConsumeComputeMetaClear(const ShaderComputeInputInfo& input,
                                                 const CommandBuffer&          buffer) {
 	const auto& program   = *input.stage.program;
-	const auto& resources = *input.stage.resources;
+	const auto& resources = input.stage.resources;
 	if (resources.buffers.size() != program.info.buffers.size()) {
 		EXIT("compute runtime buffer count does not match shader metadata\n");
 	}
@@ -54,7 +59,9 @@ bool RenderExecutor::TryConsumeComputeMetaClear(const ShaderComputeInputInfo& in
 	for (uint32_t i = 0; i < program.info.buffers.size(); i++) {
 		const auto& resource   = program.info.buffers[i];
 		const auto  descriptor = DecodeNativeDescriptor<ShaderBufferResource>(resources.buffers[i]);
-		if (!resource.written && cache.IsMeta(descriptor.Base48())) {
+		// A metadata resource that is also read is not proven to be a full overwrite. Execute it
+		// conservatively instead of replacing the dispatch with a coarse full-surface clear.
+		if ((!resource.written || resource.read) && cache.IsMeta(descriptor.Base48())) {
 			return false;
 		}
 	}
@@ -74,83 +81,108 @@ bool RenderExecutor::TryConsumeComputeMetaClear(const ShaderComputeInputInfo& in
 	return false;
 }
 
-bool ResolveComputeImageClear(const ShaderComputeInputInfo& input, uint32_t group_x,
+bool ResolveComputeBufferFill(const ShaderComputeInputInfo& input, uint32_t group_x,
                               uint32_t group_y, uint32_t group_z, uint32_t mode,
                               ShaderBufferResource& resolved_descriptor, uint32_t& resolved_clear,
                               uint64_t& resolved_size) {
-	const auto& program   = *input.stage.program;
-	const auto& resources = *input.stage.resources;
-	if (program.info.buffers.size() != 1 || resources.buffers.size() != 1 ||
-	    !program.info.images.empty() || !program.info.samplers.empty() || program.info.uses_dma ||
-	    !resources.images.empty() || !resources.samplers.empty()) {
+	const auto& resources = input.stage.resources;
+	const auto& fill      = resources.uniform_fill;
+	if (fill.kind != ShaderRecompiler::IR::UniformFillKind::Buffer) {
 		return false;
 	}
-	const auto& resource   = program.info.buffers.front();
-	const auto& raw        = resources.buffers.front();
-	const auto  descriptor = DecodeNativeDescriptor<ShaderBufferResource>(raw);
-	if (!resource.formatted || !resource.written || resource.read || resource.atomic ||
-	    resource.scalar || resource.max_byte_extent != 16 || descriptor.Stride() != 16 ||
-	    descriptor.Format() != Prospero::BufferFormat::k32_32_32_32UInt ||
+	const auto element_size = fill.words * sizeof(uint32_t);
+	const auto descriptor =
+	    DecodeNativeDescriptor<ShaderBufferResource>(resources.buffers[fill.resource]);
+	constexpr std::array formats {
+	    Prospero::BufferFormat::k32UInt, Prospero::BufferFormat::k32_32UInt,
+	    Prospero::BufferFormat::k32_32_32UInt, Prospero::BufferFormat::k32_32_32_32UInt};
+	if (descriptor.Stride() != element_size || descriptor.Format() != formats[fill.words - 1] ||
 	    descriptor.SwizzleEnabled() || descriptor.IndexStride() != 0 || descriptor.AddTid() ||
-	    resource.packed_stride != descriptor.PackedStride() || raw.dword_count != 4 ||
-	    program.user_data_base != 0 || resources.user_data.size() != 8) {
+	    descriptor.Base48() == 0) {
 		return false;
 	}
-	for (uint32_t i = 0; i < raw.dword_count; i++) {
-		if (raw.dwords[i] != resources.user_data[i]) {
-			return false;
-		}
-	}
-	const uint32_t clear = resources.user_data[4];
-	if (resources.user_data[5] != clear || resources.user_data[6] != clear ||
-	    resources.user_data[7] != clear) {
+	if (input.threads_num[0] == 0 || input.threads_num[0] != fill.group_stride[0] ||
+	    input.threads_num[1] != 1 || input.threads_num[2] != 1 || group_x == 0 || group_y != 1 ||
+	    group_z != 1 || mode != (input.dispatch_thread_dimensions ? 0x61u : 0x41u)) {
 		return false;
 	}
-	const bool full_dispatch =
-	    input.dispatch_thread_dimensions && input.threads_num[0] == 64 &&
-	    input.threads_num[1] == 1 && input.threads_num[2] == 1 && group_x != 0 && group_y == 1 &&
-	    group_z == 1 && input.dispatch_threads_num[0] == group_x &&
-	    input.dispatch_threads_num[1] == 1 && input.dispatch_threads_num[2] == 1 &&
-	    input.group_id[0] && !input.group_id[1] && !input.group_id[2] &&
-	    input.thread_ids_num == 1 && input.wave_size == 64 && !input.tg_size_en && mode == 0x61u &&
-	    group_x % input.threads_num[0] == 0 && descriptor.NumRecords() == group_x;
-	const auto size = BufferDescriptorSize(descriptor);
-	if (!full_dispatch || size == 0) {
+	const uint64_t invocations = input.dispatch_thread_dimensions
+	                                 ? group_x
+	                                 : static_cast<uint64_t>(group_x) * input.threads_num[0];
+	const auto     size        = descriptor.GetSize();
+	if (invocations != descriptor.NumRecords() || size == 0 || size > UINT32_MAX ||
+	    (input.dispatch_thread_dimensions &&
+	     (group_x % input.threads_num[0] != 0 || input.dispatch_threads_num[0] != group_x ||
+	      input.dispatch_threads_num[1] != 1 || input.dispatch_threads_num[2] != 1))) {
 		return false;
 	}
+	if (!FillSourcesDisjoint(resources.buffers, {descriptor.Base48(), size}, fill.resource))
+		return false;
 	resolved_descriptor = descriptor;
-	resolved_clear      = clear;
+	resolved_clear      = fill.value;
 	resolved_size       = size;
 	return true;
 }
 
-static bool TryConsumeComputeImageClear(const ShaderComputeInputInfo& input, CommandBuffer& command,
-                                        uint32_t group_x, uint32_t group_y, uint32_t group_z,
-                                        uint32_t mode) {
+bool RenderExecutor::TryConsumeComputeImageClear(const ShaderComputeInputInfo& input,
+                                                CommandBuffer& command, uint32_t group_x,
+                                                uint32_t group_y, uint32_t group_z, uint32_t mode) {
+	const auto& program   = *input.stage.program;
+	const auto& resources = input.stage.resources;
+	const auto& fill      = resources.uniform_fill;
+	auto&       cache     = command.GetContext().GetTextureCache();
+	if (fill.kind == ShaderRecompiler::IR::UniformFillKind::Image) {
+		if (mode != 0x41u || input.dispatch_thread_dimensions || fill.value > 255 ||
+		    input.threads_num[2] != 1)
+			return false;
+		const auto  descriptor = DecodeNativeDescriptor<ShaderTextureResource>(resources.images[0]);
+		const auto& resource   = program.info.images[0];
+		if (descriptor.IsNull() || descriptor.Format() != Prospero::BufferFormat::k8UInt ||
+		    descriptor.Type() != Prospero::ImageType::kColor2DArray || descriptor.MetaCompress() ||
+		    descriptor.WriteCompress() || descriptor.BaseLevel() > descriptor.LastLevel() ||
+		    descriptor.BaseLevel() > descriptor.MaxMip() ||
+		    descriptor.BaseArray5() > descriptor.Depth() || descriptor.DstSelX() != 4)
+			return false;
+		const std::array extents {
+		    std::max(1u, (descriptor.Width5() + 1u) >> descriptor.BaseLevel()),
+		    std::max(1u, (descriptor.Height5() + 1u) >> descriptor.BaseLevel()),
+		    descriptor.Depth() - descriptor.BaseArray5() + 1u};
+		const std::array groups {group_x, group_y, group_z};
+		for (uint32_t axis = 0; axis < 3; ++axis) {
+			const uint64_t threads = input.threads_num[axis];
+			// Guest image writes outside the descriptor dimensions are discarded. Only the
+			// final workgroup may extend beyond the selected image view.
+			if (threads == 0 || threads != fill.group_stride[axis] ||
+			    groups[axis] != (extents[axis] + threads - 1) / threads ||
+			    groups[axis] * threads > UINT32_MAX) return false;
+		}
+		const auto  binding     = ResolveTexture(resource, resources.images[0]);
+		const auto& destination = binding.desc.info.data;
+		if (!FillSourcesDisjoint(resources.buffers, destination)) return false;
+		std::scoped_lock lock {cache.m_lock};
+		const auto&      image = cache.GetImage(binding.image_id);
+		const auto&      view  = binding.desc.view_info;
+		if (image.backing.format != vk::Format::eD32SfloatS8Uint || image.info.samples != 1 ||
+		    image.info.stencil != destination || view.base_level >= image.backing.mip_levels ||
+		    view.base_layer >= image.backing.layers || view.layer_count != extents[2] ||
+		    view.layer_count > image.backing.layers - view.base_layer ||
+		    std::max(1u, image.info.extent.width >> view.base_level) != extents[0] ||
+		    std::max(1u, image.info.extent.height >> view.base_level) != extents[1]) return false;
+		const vk::ImageSubresourceRange range {vk::ImageAspectFlagBits::eStencil, view.base_level,
+		                                       1, view.base_layer, view.layer_count};
+		vk::ClearValue clear {};
+		clear.depthStencil = vk::ClearDepthStencilValue {0.0f, fill.value};
+		cache.ClearImage(command, binding.image_id, image.backing.format, range, clear);
+		return true;
+	}
 	ShaderBufferResource descriptor;
 	uint32_t             packed_clear = 0;
 	uint64_t             size         = 0;
-	if (!ResolveComputeImageClear(input, group_x, group_y, group_z, mode, descriptor, packed_clear,
+	if (!ResolveComputeBufferFill(input, group_x, group_y, group_z, mode, descriptor, packed_clear,
 	                              size)) {
 		return false;
 	}
-	auto& cache = command.GetContext().GetTextureCache();
 	if (!cache.ClearImageFromBuffer(command, descriptor.Base48(), size, packed_clear)) {
-		// Recognized metadata-fill shaders access DCC as an ordinary storage buffer and may run
-		// before the render target is bound. TryConsumeDccFill either consumes registered state
-		// or retains a PendingDcc fill while allowing the dispatch to run.
-		const bool registered_metadata =
-		    cache.TryConsumeDccFill(descriptor.Base48(), size, packed_clear);
-		static std::atomic<uint32_t> logged_metadata_clears {0};
-		if (logged_metadata_clears.fetch_add(1, std::memory_order_relaxed) < 32) {
-			LOGF("GraphicsRenderDispatchDirect: %s metadata clear shader=0x%016" PRIx64
-			     " addr=0x%016" PRIx64 " size=0x%016" PRIx64 " value=0x%08" PRIx32 "\n",
-			     registered_metadata ? "tracked" : "deferred", input.stage.program->shader_hash,
-			     descriptor.Base48(), size, packed_clear);
-		}
-		if (registered_metadata) {
-			return true;
-		}
 		return false;
 	}
 	static std::atomic<uint32_t> logged_clears {0};
@@ -174,6 +206,17 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 	                    thread_group_x, thread_group_y, thread_group_z, mode,
 	                    sh_ctx.GetCs().cs_regs.data_addr);
 
+	if (thread_group_x == 0 || thread_group_y == 0 || thread_group_z == 0) {
+		static std::atomic<uint32_t> log_count {0};
+		if (log_count.fetch_add(1, std::memory_order_relaxed) < 32) {
+			LOGF("GraphicsRenderDispatchDirect: skipping zero-sized dispatch groups=%ux%ux%u "
+			     "mode=0x%08" PRIx32 " shader=0x%016" PRIx64 "\n",
+			     thread_group_x, thread_group_y, thread_group_z, mode,
+			     sh_ctx.GetCs().cs_regs.data_addr);
+		}
+		return;
+	}
+
 	Common::LockGuard lock(m_context.GetMutex());
 	if (sh_ctx.GetCs().cs_regs.data_addr == 0) {
 		LOGF("GraphicsRenderDispatchDirect: temporary: ignoring dispatch with null CS shader, "
@@ -218,11 +261,8 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 		input_info.dispatch_threads_num[2]    = thread_group_z;
 	}
 
-	const uint32_t frame_num = static_cast<uint32_t>(m_context.GetGpu().GetFrameNum());
-	const bool     large_workgroup =
-	    (input_info.threads_num[0] * input_info.threads_num[1] * input_info.threads_num[2] >= 512);
 	const auto& program   = *input_info.stage.program;
-	const auto& resources = *input_info.stage.resources;
+	const auto& resources = input_info.stage.resources;
 	if (TryConsumeComputeMetaClear(input_info, buffer)) {
 		ResetBindings();
 		return;
@@ -232,15 +272,17 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 		ResetBindings();
 		return;
 	}
-	const auto sampled_images = std::count_if(
-	    program.info.images.begin(), program.info.images.end(), [](const auto& image) {
-		    return image.kind == ShaderRecompiler::IR::ResourceKind::Image ||
-		           image.kind == ShaderRecompiler::IR::ResourceKind::ImageUint;
-	    });
+	const bool large_workgroup =
+	    (input_info.threads_num[0] * input_info.threads_num[1] * input_info.threads_num[2] >= 512);
 	const bool                   has_sampler = !program.info.samplers.empty();
 	static std::atomic<uint32_t> dispatch_log_count {0};
 	if ((large_workgroup || has_sampler) &&
 	    dispatch_log_count.fetch_add(1, std::memory_order_relaxed) < 512) {
+		const auto sampled_images = std::count_if(
+		    program.info.images.begin(), program.info.images.end(), [](const auto& image) {
+			    return image.resource_class == ShaderRecompiler::IR::ImageResourceClass::Sampled;
+		    });
+		const uint32_t frame_num = static_cast<uint32_t>(m_context.GetGpu().GetFrameNum());
 		LOGF("GraphicsRenderDispatchDirect: frame=%u shader=0x%016" PRIx64
 		     " groups=%ux%ux%u mode=0x%08" PRIx32 " local=%ux%ux%u "
 		     "buffers=%zu textures=%zu sampled=%zu storage=%zu samplers=%zu push=%u\n",
@@ -248,7 +290,10 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 		     thread_group_z, mode, input_info.threads_num[0], input_info.threads_num[1],
 		     input_info.threads_num[2], program.info.buffers.size(), program.info.images.size(),
 		     sampled_images, program.info.images.size() - sampled_images,
-		     program.info.samplers.size(), program.bindings.push_constant_size);
+		     program.info.samplers.size(),
+		     program.bindings.UsesPushData()
+		         ? static_cast<uint32_t>(sizeof(ShaderRecompiler::IR::PushData))
+		         : 0u);
 		for (uint32_t i = 0; i < program.info.buffers.size(); i++) {
 			const auto& buffer = program.info.buffers[i];
 			const auto  r      = DecodeNativeDescriptor<ShaderBufferResource>(resources.buffers[i]);
@@ -263,8 +308,7 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 			LOGF("  CS texture[%u]: source=%u usage=%s sampled=%s addr=0x%010" PRIx64
 			     " type=%u fmt=%u extent=%ux%u depth=%u levels=%u tile=%u\n",
 			     i, image.source, image.written ? "read-write" : "read-only",
-			     (image.kind == ShaderRecompiler::IR::ResourceKind::Image ||
-			      image.kind == ShaderRecompiler::IR::ResourceKind::ImageUint)
+			     image.resource_class == ShaderRecompiler::IR::ImageResourceClass::Sampled
 			         ? "true"
 			         : "false",
 			     r.Base40(), static_cast<uint32_t>(r.Type()), static_cast<uint32_t>(r.Format()),
@@ -314,27 +358,16 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 		}
 	}
 
-	if (thread_group_x == 0 || thread_group_y == 0 || thread_group_z == 0) {
-		static std::atomic<uint32_t> log_count {0};
-		if (log_count.fetch_add(1, std::memory_order_relaxed) < 32) {
-			LOGF("GraphicsRenderDispatchDirect: skipping zero-sized dispatch groups=%ux%ux%u "
-			     "mode=0x%08" PRIx32 " shader=0x%016" PRIx64 "\n",
-			     thread_group_x, thread_group_y, thread_group_z, mode,
-			     sh_ctx.GetCs().cs_regs.data_addr);
-		}
-		return;
-	}
-
 	buffer.EndRendering();
 	auto& pipeline =
-	    m_context.GetPipelineCache().CreateComputePipeline(input_info, compute_program);
+	    m_context.GetPipelineCache().GetComputePipeline(input_info, compute_program);
 	auto bindings = PrepareBindings(input_info.stage);
 	FindBuffers(bindings);
 	if (program.info.uses_dma) {
-		m_context.GetGpuResources().PrepareBda();
+		m_context.PrepareBda();
 	}
-	RebindBuffers(bindings);
 	RebindImages(bindings);
+	RebindBuffers(bindings);
 
 	auto              vk_buffer        = buffer.Handle();
 	PreparedBindings* descriptor_stage = &bindings;
@@ -345,8 +378,8 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 	    std::any_of(program.info.images.begin(), program.info.images.end(),
 	                [](const auto& image) {
 		                return image.written &&
-		                       (image.kind == ShaderRecompiler::IR::ResourceKind::StorageImage ||
-		                        image.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint);
+		                       image.resource_class ==
+		                           ShaderRecompiler::IR::ImageResourceClass::Storage;
 	                }) ||
 	    has_storage_writes;
 	if (has_storage_writes) {