gpu: queue AGC suspend points

Port the static instruction patcher from shadPS4.

.github/workflows/build.yml

@@ -6,9 +6,27 @@ on:
     branches: [ main, master ]
   pull_request:
 
+env:
+  KYTY_BUILD_ORIGIN: ${{ github.repository != 'KytyPS5/KytyPS5' && 'Fork' || github.event_name == 'push' && 'Official' || 'Source' }}
+  KYTY_BUILD_REPOSITORY: ${{ github.repository }}
+
 jobs:
+  metadata:
+    name: Prepare build metadata
+    runs-on: ubuntu-latest
+    outputs:
+      release_tag: ${{ steps.release_tag.outputs.value }}
+
+    steps:
+      - name: Generate release tag
+        id: release_tag
+        shell: bash
+        run: |
+          echo "value=KytyPS5-$(date -u +'%Y-%m-%d')-${GITHUB_SHA::7}" >> "$GITHUB_OUTPUT"
+
   windows:
     name: Build KytyPS5 (Windows)
+    needs: metadata
     runs-on: windows-2022
 
     steps:
@@ -73,9 +91,13 @@ jobs:
       - name: Configure
         shell: cmd
         run: |
-          cmake -S src -B _Build/windows ^
+          cmake -S . -B _Build/windows ^
             -G Ninja ^
             -DCMAKE_BUILD_TYPE=Release ^
+            -DCMAKE_INTERPROCEDURAL_OPTIMIZATION=ON ^
+            -DKYTY_BUILD_ORIGIN=%KYTY_BUILD_ORIGIN% ^
+            -DKYTY_BUILD_REPOSITORY=%KYTY_BUILD_REPOSITORY% ^
+            -DKYTY_RELEASE_TAG=${{ needs.metadata.outputs.release_tag }} ^
             -DCMAKE_C_COMPILER=clang-cl ^
             -DCMAKE_CXX_COMPILER=clang-cl ^
             -DCMAKE_PREFIX_PATH="%Qt6_DIR%"
@@ -83,7 +105,12 @@ jobs:
       - name: Build
         shell: cmd
         run: |
-          cmake --build _Build/windows --target launcher --parallel
+          cmake --build _Build/windows --target launcher audio_out2_port_tests ime_dialog_tests virtual_memory_allocation_tests --parallel
+
+      - name: Test
+        shell: cmd
+        run: |
+          ctest --test-dir _Build/windows --output-on-failure -R "^(audio_out2_port|ime_dialog|virtual_memory_allocation)$"
 
       - name: Install
         shell: cmd
@@ -99,6 +126,7 @@ jobs:
 
   macos:
     name: Build KytyPS5 (macOS)
+    needs: metadata
     runs-on: macos-15
 
     steps:
@@ -142,9 +170,13 @@ jobs:
       - name: Configure
         shell: bash
         run: |
-          cmake -S src -B _Build/macos \
+          cmake -S . -B _Build/macos \
             -G Ninja \
             -DCMAKE_BUILD_TYPE=Release \
+            -DCMAKE_INTERPROCEDURAL_OPTIMIZATION=ON \
+            -DKYTY_BUILD_ORIGIN="$KYTY_BUILD_ORIGIN" \
+            -DKYTY_BUILD_REPOSITORY="$KYTY_BUILD_REPOSITORY" \
+            -DKYTY_RELEASE_TAG="${{ needs.metadata.outputs.release_tag }}" \
             -DCMAKE_OSX_ARCHITECTURES=x86_64 \
             -DCMAKE_C_COMPILER=clang \
             -DCMAKE_CXX_COMPILER=clang++ \
@@ -153,7 +185,15 @@ jobs:
       - name: Build
         shell: bash
         run: |
-          cmake --build _Build/macos --target launcher --parallel
+          cmake --build _Build/macos \
+            --target launcher audio_out2_port_tests ime_dialog_tests virtual_memory_allocation_tests \
+            --parallel
+
+      - name: Test
+        shell: bash
+        run: |
+          ctest --test-dir _Build/macos --output-on-failure \
+            -R '^(audio_out2_port|ime_dialog|virtual_memory_allocation)$'
 
       - name: Install
         shell: bash
@@ -200,6 +240,7 @@ jobs:
 
   linux:
     name: Build KytyPS5 (Linux)
+    needs: metadata
     runs-on: ubuntu-24.04
 
     steps:
@@ -257,9 +298,13 @@ jobs:
         shell: bash
         run: |
           mkdir -p _Build
-          cmake -S src -B _Build/linux \
+          cmake -S . -B _Build/linux \
             -G Ninja \
             -DCMAKE_BUILD_TYPE=Release \
+            -DCMAKE_INTERPROCEDURAL_OPTIMIZATION=ON \
+            -DKYTY_BUILD_ORIGIN="$KYTY_BUILD_ORIGIN" \
+            -DKYTY_BUILD_REPOSITORY="$KYTY_BUILD_REPOSITORY" \
+            -DKYTY_RELEASE_TAG="${{ needs.metadata.outputs.release_tag }}" \
             -DCMAKE_C_COMPILER=clang \
             -DCMAKE_CXX_COMPILER=clang++ \
             -DCMAKE_PREFIX_PATH="$Qt6_DIR" 2>&1 | tee _Build/configure.log
@@ -284,13 +329,14 @@ jobs:
         run: |
           cmake --build _Build/linux \
             --target launcher page_manager_tests memory_tracker_tests \
+              audio_out2_port_tests ime_dialog_tests virtual_memory_allocation_tests \
             --parallel
 
       - name: Test
         shell: bash
         run: |
           ctest --test-dir _Build/linux --output-on-failure \
-            -R '^(page_manager|memory_tracker)$'
+            -R '^(audio_out2_port|ime_dialog|page_manager|memory_tracker|virtual_memory_allocation)$'
 
       - name: Install
         shell: bash
@@ -327,46 +373,77 @@ jobs:
 
   release:
     name: Release KytyPS5
-    if: github.event_name == 'push' && github.repository == 'KytyPS5/KytyPS5'
-    needs: [windows, macos, linux]
+    if: >-
+      always() &&
+      github.event_name == 'push' &&
+      github.repository == 'KytyPS5/KytyPS5' &&
+      (needs.windows.result == 'success' ||
+       needs.macos.result == 'success' ||
+       needs.linux.result == 'success')
+    needs: [metadata, windows, macos, linux]
     runs-on: ubuntu-latest
     permissions:
       contents: write
+    env:
+      RELEASE_NAME: ${{ needs.metadata.outputs.release_tag }}
 
     steps:
-      - name: Download builds
+      - name: Download Windows build
+        if: needs.windows.result == 'success'
         uses: actions/download-artifact@v4
         with:
-          path: artifacts
+          name: KytyPS5-Windows-x64
+          path: artifacts/KytyPS5-Windows-x64
 
-      - name: Set release name
-        shell: bash
-        run: |
-          echo "RELEASE_NAME=KytyPS5-$(date -u +'%Y-%m-%d')-${GITHUB_SHA::7}" >> "$GITHUB_ENV"
+      - name: Download macOS build
+        if: needs.macos.result == 'success'
+        uses: actions/download-artifact@v4
+        with:
+          name: KytyPS5-macOS-x86_64
+          path: artifacts/KytyPS5-macOS-x86_64
 
-      - name: Package builds
+      - name: Download Linux build
+        if: needs.linux.result == 'success'
+        uses: actions/download-artifact@v4
+        with:
+          name: KytyPS5-Linux-x86_64
+          path: artifacts/KytyPS5-Linux-x86_64
+
+      - name: Package Windows build
+        if: needs.windows.result == 'success'
         shell: bash
         run: |
           windows_dir="artifacts/KytyPS5-Windows-x64"
-          macos_dir="artifacts/KytyPS5-macOS-x86_64"
-          linux_dir="artifacts/KytyPS5-Linux-x86_64"
           test -d "$windows_dir"
-          test -d "$macos_dir"
-          test -d "$linux_dir"
-          chmod a+x \
-            "$macos_dir/launcher" \
-            "$macos_dir/kyty_emulator" \
-            "$macos_dir/libMoltenVK.dylib" \
-            "$linux_dir/launcher" \
-            "$linux_dir/kyty_emulator"
           (
             cd "$windows_dir"
             zip -r "$GITHUB_WORKSPACE/$RELEASE_NAME-Windows-x64.zip" .
           )
+
+      - name: Package macOS build
+        if: needs.macos.result == 'success'
+        shell: bash
+        run: |
+          macos_dir="artifacts/KytyPS5-macOS-x86_64"
+          test -d "$macos_dir"
+          chmod a+x \
+            "$macos_dir/launcher" \
+            "$macos_dir/kyty_emulator" \
+            "$macos_dir/libMoltenVK.dylib"
           (
             cd "$macos_dir"
             zip -r "$GITHUB_WORKSPACE/$RELEASE_NAME-macOS-x86_64.zip" .
           )
+
+      - name: Package Linux build
+        if: needs.linux.result == 'success'
+        shell: bash
+        run: |
+          linux_dir="artifacts/KytyPS5-Linux-x86_64"
+          test -d "$linux_dir"
+          chmod a+x \
+            "$linux_dir/launcher" \
+            "$linux_dir/kyty_emulator"
           tar -C "$linux_dir" -czf \
             "$GITHUB_WORKSPACE/$RELEASE_NAME-Linux-x86_64.tar.gz" .
 
@@ -375,11 +452,16 @@ jobs:
         env:
           GH_TOKEN: ${{ github.token }}
         run: |
-          assets=(
-            "$RELEASE_NAME-Windows-x64.zip"
-            "$RELEASE_NAME-macOS-x86_64.zip"
-            "$RELEASE_NAME-Linux-x86_64.tar.gz"
-          )
+          assets=()
+          for artifact in \
+            "$RELEASE_NAME-Windows-x64.zip" \
+            "$RELEASE_NAME-macOS-x86_64.zip" \
+            "$RELEASE_NAME-Linux-x86_64.tar.gz"; do
+            if [[ -f "$artifact" ]]; then
+              assets+=("$artifact")
+            fi
+          done
+          test "${#assets[@]}" -gt 0
           if gh release view "$RELEASE_NAME" --repo "$GITHUB_REPOSITORY" > /dev/null 2>&1; then
             gh release upload "$RELEASE_NAME" "${assets[@]}" \
               --clobber \

.gitignore

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

.gitmodules

@@ -39,3 +39,6 @@
 [submodule "3rdparty/nlohmann_json"]
 	path = 3rdparty/nlohmann_json
 	url = https://github.com/nlohmann/json.git
+[submodule "3rdparty/imgui"]
+	path = 3rdparty/imgui
+	url = https://github.com/ocornut/imgui.git

.vscode/settings.json

@@ -1,5 +1,5 @@
 {
-    "cmake.sourceDirectory": "${workspaceFolder}/src",
+    "cmake.sourceDirectory": "${workspaceFolder}",
     "cmake.buildDirectory": "${workspaceFolder}/_Build/vscode-clang",
     "cmake.generator": "Ninja",
     "cmake.environment": {

3rdparty/CMakeLists.txt

@@ -2,6 +2,32 @@ set(BUILD_SHARED_LIBS OFF CACHE BOOL "" FORCE)
 
 add_subdirectory(cpuinfo)
 
+# The Windows guest red-zone patcher needs a decoder/encoder and a small
+# trampoline assembler. Pin the same revisions used by shadPS4's implementation.
+include(FetchContent)
+
+set(XBYAK_BUILD_TESTS OFF CACHE BOOL "" FORCE)
+FetchContent_Declare(
+	xbyak
+	GIT_REPOSITORY https://github.com/herumi/xbyak.git
+	GIT_TAG 44a72f369268f7d552650891b296693e91db86bb
+	GIT_SHALLOW FALSE
+)
+
+set(ZYDIS_BUILD_EXAMPLES OFF CACHE BOOL "" FORCE)
+set(ZYDIS_BUILD_TOOLS OFF CACHE BOOL "" FORCE)
+set(ZYDIS_BUILD_TESTS OFF CACHE BOOL "" FORCE)
+set(ZYDIS_BUILD_DOXYGEN OFF CACHE BOOL "" FORCE)
+FetchContent_Declare(
+	zydis
+	GIT_REPOSITORY https://github.com/zyantific/zydis.git
+	GIT_TAG 120e0e705f8e3b507dc49377ac2879979f0d545c
+	GIT_SUBMODULES_RECURSE TRUE
+	GIT_SHALLOW FALSE
+)
+
+FetchContent_MakeAvailable(xbyak zydis)
+
 set(SDL_SHARED OFF CACHE BOOL "" FORCE)
 add_subdirectory(SDL2)
 
@@ -11,6 +37,25 @@ target_include_directories(Vulkan-Headers INTERFACE
 	"${CMAKE_CURRENT_SOURCE_DIR}/Vulkan-Headers/include"
 )
 
+add_library(imgui STATIC
+	imgui/imgui.cpp
+	imgui/imgui_draw.cpp
+	imgui/imgui_tables.cpp
+	imgui/imgui_widgets.cpp
+	imgui/backends/imgui_impl_vulkan.cpp
+)
+target_include_directories(imgui PUBLIC
+	"${CMAKE_CURRENT_SOURCE_DIR}/imgui"
+	"${CMAKE_CURRENT_SOURCE_DIR}/imgui/backends"
+)
+target_compile_definitions(imgui PRIVATE IMGUI_IMPL_VULKAN_NO_PROTOTYPES)
+target_link_libraries(imgui PUBLIC Vulkan::Headers)
+if(MSVC)
+	target_compile_options(imgui PRIVATE /W0)
+else()
+	target_compile_options(imgui PRIVATE -w)
+endif()
+
 set(SPIRV_SKIP_TESTS ON CACHE BOOL "" FORCE)
 set(SPIRV_SKIP_EXECUTABLES ON CACHE BOOL "" FORCE)
 set(SKIP_SPIRV_TOOLS_INSTALL ON CACHE BOOL "" FORCE)

CMakeLists.txt

@@ -4,8 +4,18 @@ if(CMAKE_VERSION VERSION_GREATER_EQUAL 4.0)
 	set(CMAKE_POLICY_VERSION_MINIMUM 3.5 CACHE STRING "Minimum policy version for bundled dependencies")
 endif()
 
+get_property(KYTY_GENERATOR_IS_MULTI_CONFIG GLOBAL PROPERTY GENERATOR_IS_MULTI_CONFIG)
+if(NOT KYTY_GENERATOR_IS_MULTI_CONFIG AND NOT CMAKE_BUILD_TYPE)
+	set(CMAKE_BUILD_TYPE Release CACHE STRING "Choose the type of build" FORCE)
+	set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS Debug Release RelWithDebInfo MinSizeRel)
+endif()
+
 project(Kyty)
 
+set(KYTY_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/src")
+set(KYTY_TESTS_DIR "${CMAKE_CURRENT_SOURCE_DIR}/tests")
+set(KYTY_THIRD_PARTY_DIR "${CMAKE_CURRENT_SOURCE_DIR}/3rdparty")
+
 if(CMAKE_SYSTEM_NAME MATCHES ".*Linux")
 	set(LINUX TRUE)
 endif()
@@ -19,12 +29,33 @@ set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 set(CMAKE_CXX_SCAN_FOR_MODULES OFF)
 
-include(utils.cmake)
+include("${KYTY_SOURCE_DIR}/utils.cmake")
 include(CTest)
 
 option(KYTY_ENABLE_CLANG_TIDY "Run clang-tidy checks during builds" OFF)
 
-set(KYTY_THIRD_PARTY_DIR "${CMAKE_CURRENT_SOURCE_DIR}/../3rdparty")
+set(KYTY_BUILD_ORIGIN "Source" CACHE STRING "Build provenance shown in diagnostics")
+set_property(CACHE KYTY_BUILD_ORIGIN PROPERTY STRINGS Source Official Fork)
+if(NOT KYTY_BUILD_ORIGIN STREQUAL "Source" AND NOT KYTY_BUILD_ORIGIN STREQUAL "Official" AND
+   NOT KYTY_BUILD_ORIGIN STREQUAL "Fork")
+	message(FATAL_ERROR "KYTY_BUILD_ORIGIN must be Source, Official, or Fork")
+endif()
+
+set(KYTY_RELEASE_TAG "" CACHE STRING "Release tag embedded in official builds")
+set(KYTY_BUILD_REPOSITORY "" CACHE STRING "GitHub repository embedded in fork builds")
+set(KYTY_OFFICIAL_BUILD OFF)
+set(KYTY_FORK_BUILD OFF)
+if(KYTY_BUILD_ORIGIN STREQUAL "Official")
+	if(KYTY_RELEASE_TAG STREQUAL "")
+		message(FATAL_ERROR "Official builds require KYTY_RELEASE_TAG")
+	endif()
+	set(KYTY_OFFICIAL_BUILD ON)
+elseif(KYTY_BUILD_ORIGIN STREQUAL "Fork")
+	if(KYTY_BUILD_REPOSITORY STREQUAL "")
+		message(FATAL_ERROR "Fork builds require KYTY_BUILD_REPOSITORY")
+	endif()
+	set(KYTY_FORK_BUILD ON)
+endif()
 
 set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON)
 
@@ -97,36 +128,40 @@ endif()
 
 project(Kyty${KYTY_PROJECT_NAME}${CMAKE_BUILD_TYPE}${KYTY_COMPILER} VERSION 0.2.2)
 
-include_directories(	
+include_directories(
 	${KYTY_THIRD_PARTY_DIR}/gtest/include
 	${KYTY_THIRD_PARTY_DIR}/gtest
 	${KYTY_THIRD_PARTY_DIR}/fmt/include
 	${KYTY_THIRD_PARTY_DIR}/magic_enum/include/magic_enum
 	${PROJECT_BINARY_DIR}
-	${CMAKE_CURRENT_SOURCE_DIR}
+	${KYTY_SOURCE_DIR}
 )
 
 set(KYTY_VERSION "${PROJECT_VERSION}")
 
 configure_file(
-  ${PROJECT_SOURCE_DIR}/cmake_config.h.in
+  ${KYTY_SOURCE_DIR}/cmake_config.h.in
   ${PROJECT_BINARY_DIR}/cmake_config.h
   )
 
 find_package(Git)
-  
-add_custom_target( KytyGitVersion
-    COMMAND ${CMAKE_COMMAND} 
-	-D INPUT_FILE=${CMAKE_CURRENT_SOURCE_DIR}/kytyGitVersion.h.in
-    -D OUTPUT_FILE=${CMAKE_CURRENT_BINARY_DIR}/kytyGitVersion.h
-	-D GIT_EXECUTABLE=${GIT_EXECUTABLE}
-	-D GIT_WORKING_DIRECTORY=${CMAKE_CURRENT_SOURCE_DIR}
-	-P ${CMAKE_CURRENT_SOURCE_DIR}/generate_version.cmake
+
+add_custom_target(KytyGitVersion
+	COMMAND ${CMAKE_COMMAND}
+		-D INPUT_FILE=${KYTY_SOURCE_DIR}/kytyGitVersion.h.in
+		-D OUTPUT_FILE=${PROJECT_BINARY_DIR}/kytyGitVersion.h
+		-D GIT_EXECUTABLE=${GIT_EXECUTABLE}
+		-D GIT_WORKING_DIRECTORY=${CMAKE_CURRENT_SOURCE_DIR}
+		-P ${KYTY_SOURCE_DIR}/generate_version.cmake
+	BYPRODUCTS ${PROJECT_BINARY_DIR}/kytyGitVersion.h
 	COMMENT "Generate kytyGitVersion.h"
 )
 
+add_library(kyty_git_version INTERFACE)
+add_dependencies(kyty_git_version KytyGitVersion)
+
 if (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 12.0.0)
-	list(APPEND KYTY_IWYU 
+	list(APPEND KYTY_IWYU
 		kyty_emulator
 		common
 		#launcher
@@ -139,55 +174,55 @@ if (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 12.0.0)
 endif()
 
 option(KYTY_BUILD_LAUNCHER "Build Qt launcher" ON)
-  
-config_compiler_and_linker()  
-  
+
+config_compiler_and_linker()
+
 add_subdirectory("${KYTY_THIRD_PARTY_DIR}" "${CMAKE_CURRENT_BINARY_DIR}/3rdparty")
-add_subdirectory(common)
+add_subdirectory("${KYTY_SOURCE_DIR}/common" "${CMAKE_CURRENT_BINARY_DIR}/common")
 
 file(GLOB kyty_emulator_src CONFIGURE_DEPENDS
-	libs/*.cpp
-	libs/*.h
-	graphics/*.cpp
-	graphics/*.h
-	graphics/guest_gpu/*.cpp
-	graphics/guest_gpu/*.h
-	graphics/guest_gpu/command_processor/*.cpp
-	graphics/guest_gpu/command_processor/*.h
-	graphics/host_gpu/*.cpp
-	graphics/host_gpu/*.h
-	graphics/host_gpu/renderer/*.cpp
-	graphics/host_gpu/renderer/*.h
-	graphics/host_gpu/renderer/cache/*.cpp
-	graphics/host_gpu/renderer/cache/*.h
-	graphics/host_gpu/renderer/image/*.cpp
-	graphics/host_gpu/renderer/image/*.h
-	graphics/host_gpu/renderer/pipeline/*.cpp
-	graphics/host_gpu/renderer/pipeline/*.h
-	graphics/shader/*.cpp
-	graphics/shader/*.h
-	graphics/shader/recompiler/*.cpp
-	graphics/shader/recompiler/*.h
-	graphics/shader/recompiler/cfg/*.cpp
-	graphics/shader/recompiler/cfg/*.h
-	graphics/shader/recompiler/decompiler/*.cpp
-	graphics/shader/recompiler/decompiler/*.h
-	graphics/shader/recompiler/emitter/*.cpp
-	graphics/shader/recompiler/emitter/*.h
-	graphics/shader/recompiler/ir/*.cpp
-	graphics/shader/recompiler/ir/*.h
-	graphics/presentation/*.cpp
-	graphics/presentation/*.h
-	graphics/presentation/window/*.cpp
-	graphics/presentation/window/*.h
-	kernel/*.cpp
-	kernel/*.h
-	loader/*.cpp
-	loader/*.h
+	"${KYTY_SOURCE_DIR}/libs/*.cpp"
+	"${KYTY_SOURCE_DIR}/libs/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/guest_gpu/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/guest_gpu/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/guest_gpu/command_processor/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/guest_gpu/command_processor/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/renderer/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/renderer/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/renderer/cache/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/renderer/cache/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/renderer/image/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/renderer/image/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/renderer/pipeline/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/renderer/pipeline/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/shader/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/cfg/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/cfg/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/decompiler/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/decompiler/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/emitter/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/emitter/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/presentation/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/presentation/*.h"
+	"${KYTY_SOURCE_DIR}/graphics/presentation/window/*.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/presentation/window/*.h"
+	"${KYTY_SOURCE_DIR}/kernel/*.cpp"
+	"${KYTY_SOURCE_DIR}/kernel/*.h"
+	"${KYTY_SOURCE_DIR}/loader/*.cpp"
+	"${KYTY_SOURCE_DIR}/loader/*.h"
 )
 
 find_program(KYTY_GLSLANG_VALIDATOR glslangValidator REQUIRED)
-set(gpu_tiler_shader_dir "${CMAKE_CURRENT_SOURCE_DIR}/graphics/host_gpu/shaders")
+set(gpu_tiler_shader_dir "${KYTY_SOURCE_DIR}/graphics/host_gpu/shaders")
 set(gpu_tiler_generated_dir "${PROJECT_BINARY_DIR}/gpu_tiler_shaders")
 set(gpu_tiler_shader_names
 	standard256
@@ -215,7 +250,7 @@ foreach(shader_name IN LISTS gpu_tiler_shader_names)
 		COMMAND "${KYTY_GLSLANG_VALIDATOR}" -V --target-env vulkan1.0 -Os
 		        "-I${gpu_tiler_shader_dir}" -o "${shader_spv}" "${shader_source}"
 		COMMAND ${CMAKE_COMMAND} -DINPUT=${shader_spv} -DOUTPUT=${shader_header}
-		        -DSYMBOL=${shader_symbol} -P "${CMAKE_CURRENT_SOURCE_DIR}/embed_spirv.cmake"
+		        -DSYMBOL=${shader_symbol} -P "${KYTY_SOURCE_DIR}/embed_spirv.cmake"
 		DEPENDS "${shader_source}" ${gpu_tiler_shader_includes}
 		VERBATIM
 	)
@@ -239,7 +274,7 @@ foreach(shader_source IN LISTS gpu_blit_shader_sources)
 		COMMAND "${KYTY_GLSLANG_VALIDATOR}" -V --target-env vulkan1.0 -Os
 		        "-I${gpu_tiler_shader_dir}" -o "${shader_spv}" "${shader_source}"
 		COMMAND ${CMAKE_COMMAND} -DINPUT=${shader_spv} -DOUTPUT=${shader_header}
-		        -DSYMBOL=${shader_symbol} -P "${CMAKE_CURRENT_SOURCE_DIR}/embed_spirv.cmake"
+		        -DSYMBOL=${shader_symbol} -P "${KYTY_SOURCE_DIR}/embed_spirv.cmake"
 		DEPENDS "${shader_source}"
 		VERBATIM
 	)
@@ -248,8 +283,8 @@ endforeach()
 list(APPEND kyty_emulator_src ${gpu_blit_shader_headers})
 
 list(APPEND kyty_emulator_src
-	emulator.h
-	emulator.cpp
+	"${KYTY_SOURCE_DIR}/emulator.h"
+	"${KYTY_SOURCE_DIR}/emulator.cpp"
 )
 list(REMOVE_DUPLICATES kyty_emulator_src)
 
@@ -257,7 +292,7 @@ if (KYTY_CLANG_CL)
 	set_source_files_properties(${kyty_emulator_src} PROPERTIES COMPILE_FLAGS "-Wno-pragma-pack -Wno-deprecated-declarations -D_TIMESPEC_DEFINED")
 endif()
 
-set(kyty_emulator_link_libraries common Vulkan::Headers spirv-tools-opt spirv-tools SDL2-static xxhash FFmpeg::ffmpeg fmt::fmt nlohmann_json::nlohmann_json LibAtrac9)
+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.
 if(LINUX)
@@ -280,7 +315,7 @@ if(LINUX)
 endif()
 
 set(inc_headers
-	${CMAKE_CURRENT_SOURCE_DIR}
+	${KYTY_SOURCE_DIR}
 	${KYTY_THIRD_PARTY_DIR}/SDL2/include
 	${KYTY_THIRD_PARTY_DIR}/VulkanMemoryAllocator/include
 	${KYTY_THIRD_PARTY_DIR}/SPIRV-Tools/include
@@ -292,13 +327,13 @@ set(inc_headers
 
 if (KYTY_CLANG_CL)
 	list(APPEND kyty_emulator_link_libraries winpthread)
-	list(APPEND inc_headers 
+	list(APPEND inc_headers
 		${KYTY_THIRD_PARTY_DIR}/winpthread/include
 	)
 endif()
 
 list(APPEND check_headers
-	${CMAKE_CURRENT_SOURCE_DIR}
+	${KYTY_SOURCE_DIR}
 )
 
 function(add_kyty_full_emulator_test target source)
@@ -312,90 +347,122 @@ function(add_kyty_full_emulator_test target source)
 		target_link_libraries(${target} onecore)
 		add_custom_command(TARGET ${target} POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_if_different "${KYTY_THIRD_PARTY_DIR}/winpthread/bin/libwinpthread-1.dll" $<TARGET_FILE_DIR:${target}>/libwinpthread-1.dll)
 	endif()
+	# The macOS x86_64 guest address space needs its .zerofill segments anchored
+	# by linker flags, or the kernel kills the binary on load (posix_spawn EIO).
+	configure_macos_guest_address_space(${target})
 endfunction()
 
-add_kyty_full_emulator_test(shader_cfg_tests ../tests/shaderCfgTests.cpp)
+function(configure_macos_guest_address_space target)
+	if(APPLE AND (CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" OR
+			(NOT CMAKE_OSX_ARCHITECTURES AND CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|AMD64)$")))
+		target_sources(${target} PRIVATE "${KYTY_SOURCE_DIR}/kernel/macosGuestAddressSpace.cpp")
+		target_compile_definitions(${target} PRIVATE KYTY_LINKED_GUEST_ADDRESS_SPACE=1)
+		target_link_options(${target} PRIVATE
+			-Wl,-ld_classic,-no_pie,-no_fixup_chains,-no_huge,-pagezero_size,0x40000,-segaddr,SYSTEM_MANAGED,0x40000,-segaddr,SYSTEM_RESERVED,0x7ffffc000,-segaddr,USER_AREA,0x7000000000,-image_base,0x700000000000)
+	endif()
+endfunction()
+
+add_kyty_full_emulator_test(shader_cfg_tests "${KYTY_TESTS_DIR}/shaderCfgTests.cpp")
 
 add_executable(scalar_provenance_tests EXCLUDE_FROM_ALL
-	../tests/ScalarProvenanceTests.cpp
-	graphics/host_gpu/hostMemory.cpp
-	graphics/shader/recompiler/ir/ScalarProvenance.cpp
-	graphics/shader/recompiler/ir/SrtWalker.cpp
+	"${KYTY_TESTS_DIR}/ScalarProvenanceTests.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/hostMemory.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/ReadLaneElimination.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/ScalarProvenance.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/SrtWalker.cpp"
 )
 target_link_libraries(scalar_provenance_tests fmt::fmt)
 target_include_directories(scalar_provenance_tests PRIVATE ${inc_headers})
 
 add_executable(page_manager_tests EXCLUDE_FROM_ALL
-	../tests/PageManagerTests.cpp
-	graphics/host_gpu/pageManager.cpp
+	"${KYTY_TESTS_DIR}/PageManagerTests.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/pageManager.cpp"
 )
 target_include_directories(page_manager_tests PRIVATE ${inc_headers})
 
+add_executable(bit_array_tests EXCLUDE_FROM_ALL
+	"${KYTY_TESTS_DIR}/BitArrayTests.cpp"
+)
+target_include_directories(bit_array_tests PRIVATE ${inc_headers})
+
+add_executable(ime_dialog_tests EXCLUDE_FROM_ALL
+	"${KYTY_TESTS_DIR}/ImeDialogTests.cpp"
+	"${KYTY_SOURCE_DIR}/libs/imeDialog.cpp"
+)
+target_include_directories(ime_dialog_tests PRIVATE ${inc_headers})
+
 add_executable(memory_tracker_tests EXCLUDE_FROM_ALL
-	../tests/MemoryTrackerTests.cpp
-	graphics/host_gpu/pageManager.cpp
-	graphics/host_gpu/memoryTracker.cpp
+	"${KYTY_TESTS_DIR}/MemoryTrackerTests.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/pageManager.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/memoryTracker.cpp"
 )
 target_link_libraries(memory_tracker_tests fmt::fmt common)
 target_include_directories(memory_tracker_tests PRIVATE ${inc_headers})
-target_compile_definitions(memory_tracker_tests PRIVATE KYTY_MEMORY_TRACKER_TESTS=1)
 
 add_executable(shader_vertex_metadata_tests EXCLUDE_FROM_ALL
-	../tests/ShaderVertexMetadataTests.cpp
-	graphics/host_gpu/hostMemory.cpp
-	graphics/shader/shaderVertexMetadata.cpp
+	"${KYTY_TESTS_DIR}/ShaderVertexMetadataTests.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/hostMemory.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/shaderVertexMetadata.cpp"
 )
 target_include_directories(shader_vertex_metadata_tests PRIVATE ${inc_headers})
 
 add_executable(shader_stage_runtime_tests EXCLUDE_FROM_ALL
-	../tests/ShaderStageRuntimeTests.cpp
-	graphics/guest_gpu/gpu_format.cpp
-	graphics/host_gpu/hostMemory.cpp
-	graphics/shader/shaderStageRuntime.cpp
-	graphics/shader/recompiler/ir/ResourceMaterialization.cpp
-	graphics/shader/recompiler/ir/ScalarProvenance.cpp
-	graphics/shader/recompiler/ir/SrtWalker.cpp
+	"${KYTY_TESTS_DIR}/ShaderStageRuntimeTests.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/guest_gpu/gpu_format.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/hostMemory.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/shaderStageRuntime.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/ResourceMaterialization.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/ScalarProvenance.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/SrtWalker.cpp"
 )
 target_link_libraries(shader_stage_runtime_tests fmt::fmt)
 target_include_directories(shader_stage_runtime_tests PRIVATE ${inc_headers})
 
 add_executable(resource_tracking_tests EXCLUDE_FROM_ALL
-	../tests/ResourceTrackingTests.cpp
-	graphics/guest_gpu/gpu_format.cpp
-	graphics/host_gpu/hostMemory.cpp
-	graphics/shader/recompiler/ir/ScalarProvenance.cpp
-	graphics/shader/recompiler/ir/SrtWalker.cpp
-	graphics/shader/recompiler/ir/SrtPatcher.cpp
-	graphics/shader/recompiler/ir/ResourceTracking.cpp
-	graphics/shader/recompiler/ir/ResourceMaterialization.cpp
-	graphics/shader/recompiler/ir/ShaderInfoCollection.cpp
-	graphics/shader/recompiler/ir/BindingLayout.cpp
+	"${KYTY_TESTS_DIR}/ResourceTrackingTests.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/guest_gpu/gpu_format.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/host_gpu/hostMemory.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/ScalarProvenance.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/SrtWalker.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/SrtPatcher.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/ResourceTracking.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/ResourceMaterialization.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/ShaderInfoCollection.cpp"
+	"${KYTY_SOURCE_DIR}/graphics/shader/recompiler/ir/BindingLayout.cpp"
 )
 target_link_libraries(resource_tracking_tests fmt::fmt)
 target_include_directories(resource_tracking_tests PRIVATE ${inc_headers})
 
-add_executable(resource_mutex_tests EXCLUDE_FROM_ALL
-	../tests/ResourceMutexTests.cpp
-	graphics/host_gpu/renderer/cache/resourceMutex.cpp
+add_executable(audio_out2_port_tests EXCLUDE_FROM_ALL
+	"${KYTY_TESTS_DIR}/AudioOut2PortTests.cpp"
+	"${KYTY_SOURCE_DIR}/libs/libAudio2.cpp"
+	"${KYTY_SOURCE_DIR}/loader/timer.cpp"
 )
-target_link_libraries(resource_mutex_tests common)
-target_include_directories(resource_mutex_tests PRIVATE ${inc_headers})
+target_link_libraries(audio_out2_port_tests common fmt::fmt)
+target_include_directories(audio_out2_port_tests PRIVATE ${inc_headers})
 
 add_executable(event_queue_lifetime_tests EXCLUDE_FROM_ALL
-	../tests/EventQueueLifetimeTests.cpp
-	kernel/eventQueue.cpp
-	loader/timer.cpp
+	"${KYTY_TESTS_DIR}/EventQueueLifetimeTests.cpp"
+	"${KYTY_SOURCE_DIR}/kernel/eventQueue.cpp"
+	"${KYTY_SOURCE_DIR}/loader/timer.cpp"
 )
 target_link_libraries(event_queue_lifetime_tests common fmt::fmt)
 target_include_directories(event_queue_lifetime_tests PRIVATE ${inc_headers})
 
+add_executable(sync_on_address_tests EXCLUDE_FROM_ALL
+	"${KYTY_TESTS_DIR}/SyncOnAddressTests.cpp"
+	"${KYTY_SOURCE_DIR}/kernel/syncOnAddress.cpp"
+)
+target_link_libraries(sync_on_address_tests common)
+target_include_directories(sync_on_address_tests PRIVATE ${inc_headers})
+
 add_executable(image_page_table_tests EXCLUDE_FROM_ALL
-	../tests/ImagePageTableTests.cpp
+	"${KYTY_TESTS_DIR}/ImagePageTableTests.cpp"
 )
 target_link_libraries(image_page_table_tests common fmt::fmt)
 target_include_directories(image_page_table_tests PRIVATE ${inc_headers})
 
-add_kyty_full_emulator_test(shader_recompiler_compute_tests ../tests/ShaderRecompilerComputeTests.cpp)
+add_kyty_full_emulator_test(shader_recompiler_compute_tests "${KYTY_TESTS_DIR}/ShaderRecompilerComputeTests.cpp")
 
 set(gpu_test_generated_dir "${PROJECT_BINARY_DIR}/gpu_test_shaders")
 set(gpu_test_ms_depth_source
@@ -411,14 +478,14 @@ add_custom_command(
 	        -o "${gpu_test_ms_depth_spv}" "${gpu_test_ms_depth_source}"
 	COMMAND ${CMAKE_COMMAND} -DINPUT=${gpu_test_ms_depth_spv}
 	        -DOUTPUT=${gpu_test_ms_depth_header} -DSYMBOL=GPU_TEST_MS_DEPTH_SPV
-	        -P "${CMAKE_CURRENT_SOURCE_DIR}/embed_spirv.cmake"
+	        -P "${KYTY_SOURCE_DIR}/embed_spirv.cmake"
 	DEPENDS "${gpu_test_ms_depth_source}"
 	VERBATIM
 )
 target_sources(shader_recompiler_compute_tests PRIVATE
 	"${gpu_test_ms_depth_header}")
 
-add_kyty_full_emulator_test(virtual_memory_allocation_tests ../tests/VirtualMemoryAllocationTests.cpp)
+add_kyty_full_emulator_test(virtual_memory_allocation_tests "${KYTY_TESTS_DIR}/VirtualMemoryAllocationTests.cpp")
 target_compile_definitions(virtual_memory_allocation_tests PRIVATE
 	KYTY_VIRTUAL_MEMORY_ALLOCATION_TESTS=1)
 
@@ -431,18 +498,32 @@ if(NOT KYTY_CLANG_CL)
 endif()
 
 if(BUILD_TESTING)
+	add_test(NAME ime_dialog COMMAND $<TARGET_FILE:ime_dialog_tests>)
+	add_test(NAME shader_cfg COMMAND $<TARGET_FILE:shader_cfg_tests>)
+	add_test(NAME scalar_provenance COMMAND $<TARGET_FILE:scalar_provenance_tests>)
 	add_test(NAME image_page_table COMMAND $<TARGET_FILE:image_page_table_tests>)
 	add_test(NAME memory_tracker COMMAND $<TARGET_FILE:memory_tracker_tests>)
 	add_test(NAME page_manager COMMAND $<TARGET_FILE:page_manager_tests>)
-	add_test(NAME resource_mutex COMMAND $<TARGET_FILE:resource_mutex_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_tracking COMMAND $<TARGET_FILE:resource_tracking_tests>)
 	add_test(NAME event_queue_lifetime COMMAND $<TARGET_FILE:event_queue_lifetime_tests>)
+	add_test(NAME sync_on_address COMMAND $<TARGET_FILE:sync_on_address_tests>)
+	add_test(NAME audio_out2_port COMMAND $<TARGET_FILE:audio_out2_port_tests>)
 	add_test(NAME shader_recompiler_compute COMMAND $<TARGET_FILE:shader_recompiler_compute_tests>)
+	add_test(NAME virtual_memory_allocation
+		COMMAND $<TARGET_FILE:virtual_memory_allocation_tests>)
+	add_test(NAME guest_red_zone_patcher
+		COMMAND $<TARGET_FILE:virtual_memory_allocation_tests> --red-zone-patcher-only)
 	add_test(NAME command_scheduler_timeline
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --scheduler-only)
 	add_test(NAME stream_buffer_ring
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --stream-buffer-only)
 	add_test(NAME gpu_command_lane
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --gpu-command-lane-only)
+	add_test(NAME pm4_context_state
+		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --context-state-only)
 	add_test(NAME gpu_tiler
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --gpu-tiler-only)
 
@@ -466,10 +547,28 @@ if(BUILD_TESTING)
 		add_test(NAME buffer_cache_ranges
 			COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --buffer-cache-range-only)
 	endif()
+
+	add_custom_target(kyty_tests DEPENDS
+		ime_dialog_tests
+		shader_cfg_tests
+		scalar_provenance_tests
+		image_page_table_tests
+		memory_tracker_tests
+		page_manager_tests
+		bit_array_tests
+		shader_vertex_metadata_tests
+		shader_stage_runtime_tests
+		resource_tracking_tests
+		event_queue_lifetime_tests
+		sync_on_address_tests
+		shader_recompiler_compute_tests
+		virtual_memory_allocation_tests
+	)
 endif()
 
 
-add_executable(kyty_emulator main.cpp ${kyty_emulator_src})
+add_executable(kyty_emulator "${KYTY_SOURCE_DIR}/main.cpp" ${kyty_emulator_src})
+configure_macos_guest_address_space(kyty_emulator)
 
 target_link_libraries(kyty_emulator ${kyty_emulator_link_libraries})
 if (WIN32)
@@ -498,14 +597,17 @@ set(KYTY_EMULATOR_MAP_LINK_PATH "${CMAKE_CURRENT_BINARY_DIR}/${KYTY_EMULATOR_MAP
 set(KYTY_EMULATOR_PDB_LINK_PATH "${CMAKE_CURRENT_BINARY_DIR}/kyty_emulator.pdb")
 
 if(KYTY_CLANG_CL)
-	set_target_properties(kyty_emulator PROPERTIES LINK_FLAGS "/DYNAMICBASE:NO /DEBUG:FULL /PDB:${KYTY_EMULATOR_PDB_LINK_PATH} /lldmap:${KYTY_EMULATOR_MAP_LINK_PATH}")
+	target_link_options(kyty_emulator PRIVATE
+		"/DYNAMICBASE:NO"
+		"/DEBUG:FULL"
+		"/PDB:${KYTY_EMULATOR_PDB_LINK_PATH}"
+		"/lldmap:${KYTY_EMULATOR_MAP_LINK_PATH}"
+	)
 	add_custom_command(TARGET kyty_emulator POST_BUILD COMMAND ${CMAKE_COMMAND} -E copy_if_different "${KYTY_THIRD_PARTY_DIR}/winpthread/bin/libwinpthread-1.dll" $<TARGET_FILE_DIR:kyty_emulator>/libwinpthread-1.dll)
 elseif(WIN32 OR LINUX)
 	set_target_properties(kyty_emulator PROPERTIES LINK_FLAGS "${KYTY_LD_OPTIONS} -Wl,-Map=${KYTY_EMULATOR_MAP_LINK_PATH}")
 endif()
 
-add_dependencies(kyty_emulator KytyGitVersion)
-
 if(APPLE)
 	# The emulator writes x86-64 trampolines/PLT stubs into memory and executes them, and
 	# runs guest code under Rosetta. That requires the JIT / unsigned-executable-memory
@@ -513,14 +615,14 @@ if(APPLE)
 	# reverts to the hardened defaults and aborts when it first executes written code).
 	add_custom_command(TARGET kyty_emulator POST_BUILD
 		COMMAND codesign -s - --force --options runtime
-			--entitlements "${CMAKE_CURRENT_SOURCE_DIR}/macos_jit.entitlements"
+			--entitlements "${KYTY_SOURCE_DIR}/macos_jit.entitlements"
 			$<TARGET_FILE:kyty_emulator>
 		COMMENT "Codesign kyty_emulator with JIT entitlements (macOS)")
 endif()
 
 install(TARGETS kyty_emulator DESTINATION .)
 if(KYTY_BUILD_LAUNCHER)
-	add_subdirectory(launcher)
+	add_subdirectory("${KYTY_SOURCE_DIR}/launcher" "${CMAKE_CURRENT_BINARY_DIR}/launcher")
 endif()
 if(KYTY_CLANG_CL)
 	install(FILES "${KYTY_THIRD_PARTY_DIR}/winpthread/bin/libwinpthread-1.dll" DESTINATION .)

README.md

@@ -27,8 +27,9 @@ Development is focused on compatibility and boot reliability.
 Windows is the primary platform and receives the most testing. Linux builds and runs; see
 [Building on Linux](#building-on-linux).
 
-macOS support is experimental. Compatibility with the same games on Windows and macOS has not yet
-been tested.
+macOS support is experimental. The emulator is built for x86-64 and runs on Apple Silicon under
+Rosetta 2, with Vulkan provided by MoltenVK. A small number of titles have been verified in-game
+on Apple Silicon hardware; see [Building on macOS](#building-on-macos).
 
 ## Bugs and Issues
 
@@ -50,7 +51,7 @@ graphical glitches, low compatibility, and poor performance.
   </tr>
   <tr>
     <td align="center">
-      <strong>Minecraft Legends</strong><br>
+      <strong>Neptunia ReVerse</strong><br>
       <img src="docs/screenshots/ps5-04.png" width="300" alt="Minecraft Legends running in KytyPS5">
     </td>
     <td align="center">
@@ -58,8 +59,20 @@ graphical glitches, low compatibility, and poor performance.
       <img src="docs/screenshots/ps5-05.png" width="300" alt="SILENT HILL: The Short Message running in KytyPS5">
     </td>
   </tr>
+  <tr>
+    <td align="center">
+      <strong>Hellboy</strong><br>
+      <img src="docs/screenshots/ps5-02.png" width="300" alt="Disgaea 6 running in KytyPS5">
+    </td>
+    <td align="center">
+      <strong>Paleo Pines</strong><br>
+      <img src="docs/screenshots/ps5-06.png" width="300" alt="Dreaming Sarah running in KytyPS5">
+    </td>
+  </tr>
 </table>
 
+<p align="center"><em>And many more...</em></p>
+
 ## Contributing
 
 Testing games and submitting detailed bug reports are useful ways to contribute. Search existing
@@ -102,9 +115,10 @@ the Vulkan/SPIR-V validation rules.
 
 ### System requirements
 
-- Windows 10 version 1803, or a current Linux distribution
-- A 64-bit x86 processor
-- A Vulkan 1.3-capable GPU with current drivers
+- Windows 10 version 1803, a current Linux distribution, or macOS on Apple Silicon
+- A 64-bit x86 processor (on macOS, an Apple Silicon processor with Rosetta 2)
+- A Vulkan 1.3-capable GPU with current drivers (on macOS, Vulkan is provided by the bundled
+  MoltenVK)
 
 ### Build requirements (Windows)
 
@@ -127,7 +141,7 @@ git submodule update --init --recursive
 Configure the project. Replace the Qt path with the version installed on your system:
 
 ```powershell
-cmake -S src -B _Build/windows -G Ninja -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_PREFIX_PATH="C:/Qt/6.x.x/msvc2022_64"
+cmake -S . -B _Build/windows -G Ninja -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_PREFIX_PATH="C:/Qt/6.x.x/msvc2022_64"
 ```
 
 Build the launcher and stage a runnable installation:
@@ -160,7 +174,7 @@ Qt 6 (Concurrent, Network, Widgets) is also required — either the distribution
 ```bash
 git submodule update --init --recursive
 
-cmake -S src -B _Build/linux -G Ninja -DCMAKE_BUILD_TYPE=Release \
+cmake -S . -B _Build/linux -G Ninja -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ \
   -DCMAKE_PREFIX_PATH="$Qt6_DIR"
 
@@ -174,7 +188,57 @@ The install step copies the Qt libraries and plugins next to the binaries, so
 As on Windows, the MSVC compiler is not used; Clang is required. `cl.exe` is rejected at configure
 time.
 
-Note that the CMake source root is `src`, not the repository root.
+The CMake source root is the repository root.
+
+### Building on macOS
+
+macOS builds target x86-64 and run under Rosetta 2 on Apple Silicon, so the PS5's x86-64 game
+code executes through the same translation layer as the emulator itself. Prebuilt archives are
+attached to releases; the steps below are for building from source.
+
+Requirements:
+
+- An Apple Silicon Mac with Rosetta 2 installed (`softwareupdate --install-rosetta`)
+- Xcode (or the Command Line Tools)
+- Homebrew packages: `brew install cmake ninja glslang`
+- Qt 6 (Concurrent, Network, Widgets) with x86-64 support. The official Qt installation is
+  universal and works; Homebrew's Qt is arm64-only and will not link
+
+```bash
+git submodule update --init --recursive
+
+cmake -S . -B _Build/macos -G Ninja -DCMAKE_BUILD_TYPE=Release \
+  -DCMAKE_OSX_ARCHITECTURES=x86_64 \
+  -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ \
+  -DCMAKE_PREFIX_PATH="$Qt6_DIR"
+
+cmake --build _Build/macos --target launcher --parallel
+cmake --install _Build/macos --prefix _Build/macos/install
+```
+
+The build re-signs `kyty_emulator` with the JIT entitlements it needs to execute translated
+guest code; no manual signing step is required.
+
+Vulkan comes from MoltenVK. Download `MoltenVK-macos.tar` from the
+[MoltenVK releases](https://github.com/KhronosGroup/MoltenVK/releases), then copy
+`MoltenVK/dynamic/dylib/macOS/libMoltenVK.dylib` next to `kyty_emulator` and ad-hoc sign it:
+
+```bash
+codesign --force --sign - _Build/macos/install/libMoltenVK.dylib
+```
+
+Release archives already include a signed `libMoltenVK.dylib`.
+
+### Regression tests
+
+Build every regression executable and run the registered tests with:
+
+```powershell
+cmake --build _Build/windows --target kyty_tests
+ctest --test-dir _Build/windows --output-on-failure
+```
+
+Use `_Build/linux` instead of `_Build/windows` for a Linux build.
 
 ### Visual Studio Code
 
@@ -221,6 +285,14 @@ The emulator can also be started directly with a legally obtained game directory
 ./_Build/linux/install/kyty_emulator --game "/games/ExampleGame"
 ```
 
+On macOS, point SDL at the MoltenVK library explicitly; the hardened runtime prevents it from
+being picked up from the executable's directory:
+
+```bash
+cd _Build/macos/install
+SDL_VULKAN_LIBRARY="$PWD/libMoltenVK.dylib" ./kyty_emulator --game "/games/ExampleGame"
+```
+
 Run `kyty_emulator --help` to see the available graphics, logging, validation, profiling, and
 debugging options.
 

src/cmake_config.h.in

@@ -1,4 +1,8 @@
 #define KYTY_VERSION "@KYTY_VERSION@"
+#define KYTY_BUILD_REPOSITORY "@KYTY_BUILD_REPOSITORY@"
+#define KYTY_RELEASE_TAG "@KYTY_RELEASE_TAG@"
+#cmakedefine KYTY_OFFICIAL_BUILD
+#cmakedefine KYTY_FORK_BUILD
 #define KYTY_ENDIAN @KYTY_ENDIAN@
 #define KYTY_PLATFORM @KYTY_PLATFORM@
 #define KYTY_COMPILER KYTY_COMPILER_@KYTY_COMPILER@

src/common/CMakeLists.txt

@@ -13,7 +13,7 @@ endif()
 add_library(common STATIC ${common_src} ${common_headers})
 
 target_include_directories(common
-	PUBLIC ${PROJECT_SOURCE_DIR}
+	PUBLIC ${KYTY_SOURCE_DIR}
 	PRIVATE
 		${KYTY_THIRD_PARTY_DIR}/cpuinfo/include
 )

src/common/bitArray.h

@@ -0,0 +1,260 @@
+#ifndef EMULATOR_SRC_COMMON_BITARRAY_H_
+#define EMULATOR_SRC_COMMON_BITARRAY_H_
+
+#include <array>
+#include <bit>
+#include <cstddef>
+#include <cstdint>
+#include <iterator>
+#include <utility>
+
+namespace Common {
+
+template <size_t N>
+class BitArray final {
+	static_assert(N != 0, "BitArray size must be nonzero");
+	static_assert(N % 64 == 0, "BitArray size must be a multiple of 64 bits");
+
+	static constexpr size_t BITS_PER_WORD = 64;
+	static constexpr size_t WORD_COUNT    = N / BITS_PER_WORD;
+
+public:
+	using Range = std::pair<size_t, size_t>;
+
+	class Iterator final {
+	public:
+		using iterator_category = std::forward_iterator_tag;
+		using value_type        = Range;
+		using difference_type   = std::ptrdiff_t;
+		using pointer           = const Range*;
+		using reference         = const Range&;
+
+		Iterator(const BitArray& bits, size_t start)
+		    : m_bits(bits), m_range(bits.FirstRangeFrom(start)) {}
+
+		Iterator& operator++() {
+			m_range = m_bits.FirstRangeFrom(m_range.second);
+			return *this;
+		}
+
+		[[nodiscard]] bool operator==(const Iterator& other) const {
+			return &m_bits == &other.m_bits && m_range == other.m_range;
+		}
+
+		[[nodiscard]] bool operator!=(const Iterator& other) const { return !(*this == other); }
+
+		[[nodiscard]] reference operator*() const { return m_range; }
+		[[nodiscard]] pointer   operator->() const { return &m_range; }
+
+	private:
+		const BitArray& m_bits;
+		Range           m_range;
+	};
+
+	using const_iterator = Iterator;
+
+	constexpr BitArray() = default;
+
+	constexpr BitArray(const BitArray& other, size_t start, size_t end) {
+		if (start >= end || end > N) {
+			return;
+		}
+
+		const auto first_word = start / BITS_PER_WORD;
+		const auto last_word  = (end - 1) / BITS_PER_WORD;
+		const auto start_bit  = start % BITS_PER_WORD;
+		const auto end_bit    = (end - 1) % BITS_PER_WORD;
+		const auto start_mask = ~uint64_t {0} << start_bit;
+		const auto end_mask =
+		    end_bit == BITS_PER_WORD - 1 ? ~uint64_t {0} : (uint64_t {1} << (end_bit + 1)) - 1;
+
+		if (first_word == last_word) {
+			m_data[first_word] = other.m_data[first_word] & start_mask & end_mask;
+			return;
+		}
+
+		m_data[first_word] = other.m_data[first_word] & start_mask;
+		for (auto word = first_word + 1; word < last_word; word++) {
+			m_data[word] = other.m_data[word];
+		}
+		m_data[last_word] = other.m_data[last_word] & end_mask;
+	}
+
+	[[nodiscard]] constexpr bool Get(size_t index) const {
+		return (m_data[index / BITS_PER_WORD] & (uint64_t {1} << (index % BITS_PER_WORD))) != 0;
+	}
+
+	constexpr void Set(size_t index) {
+		m_data[index / BITS_PER_WORD] |= uint64_t {1} << (index % BITS_PER_WORD);
+	}
+
+	constexpr void Unset(size_t index) {
+		m_data[index / BITS_PER_WORD] &= ~(uint64_t {1} << (index % BITS_PER_WORD));
+	}
+
+	constexpr void SetRange(size_t start, size_t end) {
+		if (start >= end || end > N) {
+			return;
+		}
+
+		const auto first_word = start / BITS_PER_WORD;
+		const auto last_word  = (end - 1) / BITS_PER_WORD;
+		const auto start_bit  = start % BITS_PER_WORD;
+		const auto end_bit    = (end - 1) % BITS_PER_WORD;
+		const auto start_mask = ~uint64_t {0} << start_bit;
+		const auto end_mask =
+		    end_bit == BITS_PER_WORD - 1 ? ~uint64_t {0} : (uint64_t {1} << (end_bit + 1)) - 1;
+
+		if (first_word == last_word) {
+			m_data[first_word] |= start_mask & end_mask;
+			return;
+		}
+
+		m_data[first_word] |= start_mask;
+		for (auto word = first_word + 1; word < last_word; word++) {
+			m_data[word] = ~uint64_t {0};
+		}
+		m_data[last_word] |= end_mask;
+	}
+
+	constexpr void UnsetRange(size_t start, size_t end) {
+		if (start >= end || end > N) {
+			return;
+		}
+
+		const auto first_word = start / BITS_PER_WORD;
+		const auto last_word  = (end - 1) / BITS_PER_WORD;
+		const auto start_bit  = start % BITS_PER_WORD;
+		const auto end_bit    = (end - 1) % BITS_PER_WORD;
+		const auto start_mask = (uint64_t {1} << start_bit) - 1;
+		const auto end_mask =
+		    end_bit == BITS_PER_WORD - 1 ? uint64_t {0} : ~((uint64_t {1} << (end_bit + 1)) - 1);
+
+		if (first_word == last_word) {
+			m_data[first_word] &= start_mask | end_mask;
+			return;
+		}
+
+		m_data[first_word] &= start_mask;
+		for (auto word = first_word + 1; word < last_word; word++) {
+			m_data[word] = 0;
+		}
+		m_data[last_word] &= end_mask;
+	}
+
+	constexpr void Clear() { m_data.fill(0); }
+	constexpr void Fill() { m_data.fill(~uint64_t {0}); }
+
+	[[nodiscard]] constexpr bool None() const {
+		uint64_t combined = 0;
+		for (const auto word: m_data) {
+			combined |= word;
+		}
+		return combined == 0;
+	}
+
+	[[nodiscard]] constexpr bool Any() const { return !None(); }
+
+	[[nodiscard]] constexpr Range FirstRangeFrom(size_t start) const {
+		if (start >= N) {
+			return {N, N};
+		}
+
+		auto word_index = start / BITS_PER_WORD;
+		auto word       = m_data[word_index] & (~uint64_t {0} << (start % BITS_PER_WORD));
+		while (word == 0) {
+			word_index++;
+			if (word_index == WORD_COUNT) {
+				return {N, N};
+			}
+			word = m_data[word_index];
+		}
+
+		const auto first     = word_index * BITS_PER_WORD + std::countr_zero(word);
+		const auto first_bit = first % BITS_PER_WORD;
+		const auto first_ones =
+		    static_cast<size_t>(std::countr_one(m_data[word_index] >> first_bit));
+		if (first_bit + first_ones < BITS_PER_WORD) {
+			return {first, first + first_ones};
+		}
+
+		for (word_index++; word_index < WORD_COUNT; word_index++) {
+			word = m_data[word_index];
+			if (word != ~uint64_t {0}) {
+				return {first, word_index * BITS_PER_WORD + std::countr_one(word)};
+			}
+		}
+		return {first, N};
+	}
+
+	[[nodiscard]] constexpr Range FirstRange() const { return FirstRangeFrom(0); }
+
+	[[nodiscard]] constexpr Range LastRangeFrom(size_t end) const {
+		if (end == 0) {
+			return {0, 0};
+		}
+		if (end > N) {
+			end = N;
+		}
+
+		auto       word_index = (end - 1) / BITS_PER_WORD;
+		const auto end_bit    = (end - 1) % BITS_PER_WORD;
+		const auto end_mask =
+		    end_bit == BITS_PER_WORD - 1 ? ~uint64_t {0} : (uint64_t {1} << (end_bit + 1)) - 1;
+		auto word = m_data[word_index] & end_mask;
+		while (word == 0) {
+			if (word_index == 0) {
+				return {0, 0};
+			}
+			word = m_data[--word_index];
+		}
+
+		const auto empty_bits = static_cast<size_t>(std::countl_zero(word));
+		const auto ones       = static_cast<size_t>(std::countl_one(word << empty_bits));
+		const auto last       = (word_index + 1) * BITS_PER_WORD - empty_bits;
+		if (empty_bits + ones < BITS_PER_WORD) {
+			return {last - ones, last};
+		}
+
+		while (word_index != 0) {
+			word = m_data[--word_index];
+			if (word != ~uint64_t {0}) {
+				return {(word_index + 1) * BITS_PER_WORD - std::countl_one(word), last};
+			}
+		}
+		return {0, last};
+	}
+
+	[[nodiscard]] constexpr Range LastRange() const { return LastRangeFrom(N); }
+
+	[[nodiscard]] const_iterator begin() const { return Iterator(*this, 0); }
+	[[nodiscard]] const_iterator end() const { return Iterator(*this, N); }
+
+	constexpr BitArray& operator^=(const BitArray& other) {
+		for (size_t word = 0; word < WORD_COUNT; word++) {
+			m_data[word] ^= other.m_data[word];
+		}
+		return *this;
+	}
+
+	[[nodiscard]] constexpr BitArray operator^(const BitArray& other) const {
+		auto result = *this;
+		result ^= other;
+		return result;
+	}
+
+	[[nodiscard]] constexpr BitArray operator~() const {
+		auto result = *this;
+		for (auto& word: result.m_data) {
+			word = ~word;
+		}
+		return result;
+	}
+
+private:
+	std::array<uint64_t, WORD_COUNT> m_data {};
+};
+
+} // namespace Common
+
+#endif // EMULATOR_SRC_COMMON_BITARRAY_H_

src/common/emulatorConfig.cpp

@@ -33,10 +33,18 @@ uint32_t GetScreenHeight() {
 	return g_config->screen_height;
 }
 
+bool FullscreenEnabled() {
+	return g_config->fullscreen_enabled;
+}
+
 uint32_t GetVblankFrequency() {
 	return std::clamp(g_config->vblank_frequency, 30u, 360u);
 }
 
+uint32_t GetConsoleLanguage() {
+	return g_config->console_language;
+}
+
 bool VulkanValidationEnabled() {
 	return g_config->vulkan_validation_enabled;
 }
@@ -89,12 +97,18 @@ bool RenderDocEnabled() {
 	return g_config->renderdoc_enabled;
 }
 
-bool NggRectlistDrawEnabled() {
-	return g_config->ngg_rectlist_draw_enabled;
-}
-
 bool ReadbackLinearImagesEnabled() {
 	return g_config->readback_linear_images;
 }
 
+#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
+bool RedZoneProtectionEnabled() {
+	return g_config->red_zone_protection_enabled;
+}
+#endif
+
+const Keymap& GetKeymap() {
+	return g_config->keymap;
+}
+
 } // namespace Config

src/common/emulatorConfig.h

@@ -5,6 +5,8 @@
 #include "common/subsystems.h"
 
 #include <filesystem>
+#include <string>
+#include <vector>
 
 namespace Config {
 
@@ -18,10 +20,17 @@ enum class ProfilerDirection { None, Network };
 
 enum class OutputDirection { Silent, Console, File };
 
+using Keymap = std::vector<std::string>;
+
+constexpr uint32_t DEFAULT_CONSOLE_LANGUAGE = 1;
+constexpr uint32_t MAX_CONSOLE_LANGUAGE     = 29;
+
 struct ConfigOptions {
 	uint32_t               screen_width                = 1280;
 	uint32_t               screen_height               = 720;
+	bool                   fullscreen_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;
@@ -35,15 +44,20 @@ struct ConfigOptions {
 	ProfilerDirection      profiler_direction          = ProfilerDirection::None;
 	bool                   spirv_debug_printf_enabled  = false;
 	bool                   renderdoc_enabled           = false;
-	bool                   ngg_rectlist_draw_enabled   = true;
 	bool                   readback_linear_images      = false;
+#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
+	bool red_zone_protection_enabled = false;
+#endif
+	Keymap keymap;
 };
 
 void Load(const ConfigOptions& cfg);
 
 uint32_t GetScreenWidth();
 uint32_t GetScreenHeight();
+bool     FullscreenEnabled();
 uint32_t GetVblankFrequency();
+uint32_t GetConsoleLanguage();
 bool     VulkanValidationEnabled();
 
 bool                   ShaderValidationEnabled();
@@ -64,8 +78,12 @@ ProfilerDirection GetProfilerDirection();
 bool SpirvDebugPrintfEnabled();
 
 bool RenderDocEnabled();
-bool NggRectlistDrawEnabled();
 bool ReadbackLinearImagesEnabled();
+#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
+bool RedZoneProtectionEnabled();
+#endif
+
+const Keymap& GetKeymap();
 
 } // namespace Config
 

src/common/hostException.cpp

@@ -175,9 +175,9 @@ static void SignalHandler(int sig, siginfo_t* si, void* uctx) {
 	}
 	g_in_exception_filter = true;
 
-	auto*       uc  = static_cast<ucontext_t*>(uctx);
-	const auto* mc  = uc->uc_mcontext;
-	const auto& ss  = mc->__ss;
+	auto*       uc = static_cast<ucontext_t*>(uctx);
+	const auto* mc = uc->uc_mcontext;
+	const auto& ss = mc->__ss;
 
 	ExceptionInfo info {};
 	info.exception_address = ss.__rip;
@@ -214,7 +214,7 @@ static void SignalHandler(int sig, siginfo_t* si, void* uctx) {
 		FailFast("host exception callback is null");
 	}
 
-	const bool resolved = handler(info);
+	const bool resolved   = handler(info);
 	g_in_exception_filter = false;
 
 	if (resolved) {
@@ -255,8 +255,8 @@ static void SignalHandler(int signal_number, siginfo_t* signal_info, void* nativ
 	info.native_context    = context;
 
 	if (signal_number == SIGSEGV || signal_number == SIGBUS) {
-		info.type              = ExceptionType::AccessViolation;
-		const auto error_code  = static_cast<uint64_t>(gregs[REG_ERR]);
+		info.type             = ExceptionType::AccessViolation;
+		const auto error_code = static_cast<uint64_t>(gregs[REG_ERR]);
 		if ((error_code & PAGE_FAULT_ERROR_INSTRUCTION) != 0) {
 			info.access_violation_type = AccessViolationType::Execute;
 		} else if ((error_code & PAGE_FAULT_ERROR_WRITE) != 0) {
@@ -324,6 +324,10 @@ bool InstallHandler(Handler handler) {
 	sa.sa_sigaction = SignalHandler;
 	sa.sa_flags     = SA_SIGINFO;
 	sigemptyset(&sa.sa_mask);
+	// The guest signal-dispatch path (KernelRaiseException) interrupts threads with
+	// SIGUSR1; block it while a fault is being resolved so a stop-the-world request
+	// cannot preempt the handler between the protection fix and the retry.
+	sigaddset(&sa.sa_mask, SIGUSR1);
 
 	// macOS raises SIGBUS for protection faults on some paths and SIGSEGV on others;
 	// SIGILL covers instructions the host cannot execute (routed to the x64 emulator).

src/common/lruCache.h

@@ -19,10 +19,10 @@ class LeastRecentlyUsedCache {
 
 public:
 	[[nodiscard]] size_t Insert(Object object, Tick tick) {
-		const auto id = Build();
+		const auto id   = Build();
 		auto&      item = m_items[id];
-		item.object = std::move(object);
-		item.tick   = tick;
+		item.object     = std::move(object);
+		item.tick       = tick;
 		Attach(item);
 		return id;
 	}
@@ -49,8 +49,7 @@ public:
 
 	template <typename Function>
 	void ForEachItemBelow(Tick tick, Function&& function) {
-		constexpr bool ReturnsBool =
-		    std::is_same_v<std::invoke_result_t<Function, Object>, bool>;
+		constexpr bool ReturnsBool = std::is_same_v<std::invoke_result_t<Function, Object>, bool>;
 		for (auto* item = m_first; item != nullptr;) {
 			if (item->tick > tick) {
 				return;
@@ -87,10 +86,10 @@ private:
 			m_last = &item;
 			return;
 		}
-		item.prev   = m_last;
+		item.prev    = m_last;
 		m_last->next = &item;
-		item.next   = nullptr;
-		m_last      = &item;
+		item.next    = nullptr;
+		m_last       = &item;
 	}
 
 	void Detach(Item& item) {

src/common/platform/sysLinuxDbg.cpp

@@ -31,10 +31,9 @@ static bool OnOwnStack() {
 	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;
+	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;

src/common/platform/sysLinuxFileIO.cpp

@@ -172,8 +172,7 @@ sys_file_t* SysFileCreate(const std::filesystem::path& file_name) {
 	return ret;
 }
 
-sys_file_t* SysFileOpenR(const std::filesystem::path& file_name,
-                         sys_file_cache_type_t cache_type) {
+sys_file_t* SysFileOpenR(const std::filesystem::path& file_name, sys_file_cache_type_t cache_type) {
 	auto* ret = new sys_file_t;
 
 	ret->type = SYS_FILE_FILE;
@@ -218,8 +217,7 @@ sys_file_t* SysFileCreate() {
 	return ret;
 }
 
-sys_file_t* SysFileOpenW(const std::filesystem::path& file_name,
-                         sys_file_cache_type_t cache_type) {
+sys_file_t* SysFileOpenW(const std::filesystem::path& file_name, sys_file_cache_type_t cache_type) {
 	auto* ret = new sys_file_t;
 
 	auto real_name     = get_internal_name(file_name);
@@ -241,7 +239,7 @@ sys_file_t* SysFileOpenW(const std::filesystem::path& file_name,
 }
 
 sys_file_t* SysFileOpenRw(const std::filesystem::path& file_name,
-                          sys_file_cache_type_t cache_type) {
+                          sys_file_cache_type_t        cache_type) {
 	auto* ret = new sys_file_t;
 
 	auto real_name     = get_internal_name(file_name);

src/common/platform/sysLinuxVirtual.cpp

@@ -136,8 +136,8 @@ static void* map_anonymous(uintptr_t addr, size_t size, int protect, int flags)
 			break;
 		}
 		const auto hint = (top - step) & ~(LOW_ARENA_GRAIN - 1);
-		void*      ptr  = mmap(reinterpret_cast<void*>(hint), size, protect,
-		                       flags | MAP_FIXED_NOREPLACE, -1, 0); // NOLINT
+		void* ptr = mmap(reinterpret_cast<void*>(hint), size, protect, flags | MAP_FIXED_NOREPLACE,
+		                 -1, 0); // NOLINT
 		if (ptr != MAP_FAILED) {
 			return ptr;
 		}
@@ -161,8 +161,8 @@ 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;
+		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;
 		}
@@ -194,8 +194,8 @@ uint64_t SysVirtualAllocAligned(uint64_t address, uint64_t size, VirtualMemory::
 	if (ptr != MAP_FAILED && ((ret_addr & (alignment - 1)) != 0)) {
 		munmap(ptr, size);
 
-		ptr      = map_anonymous(addr, size + alignment, protect,
-		                         MAP_PRIVATE | MAP_ANON | MAP_NORESERVE);
+		ptr =
+		    map_anonymous(addr, size + alignment, protect, MAP_PRIVATE | MAP_ANON | MAP_NORESERVE);
 		ret_addr = reinterpret_cast<uintptr_t>(ptr);
 		if (ptr != MAP_FAILED) {
 #if defined(__APPLE__)
@@ -251,8 +251,8 @@ 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;
+	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;
 	}
@@ -266,9 +266,9 @@ uint64_t SysVirtualAllocAligned(uint64_t address, uint64_t size, VirtualMemory::
 // the first mapped region at or above `region_addr`; if it begins before the end of the
 // requested range, the range overlaps an existing mapping.
 static bool is_mapped(void* ptr, size_t length) {
-	auto              query_addr  = reinterpret_cast<mach_vm_address_t>(ptr);
-	mach_vm_address_t region_addr = query_addr;
-	mach_vm_size_t    region_size = 0;
+	auto                           query_addr  = reinterpret_cast<mach_vm_address_t>(ptr);
+	mach_vm_address_t              region_addr = query_addr;
+	mach_vm_size_t                 region_size = 0;
 	vm_region_basic_info_data_64_t info {};
 	mach_msg_type_number_t         count       = VM_REGION_BASIC_INFO_COUNT_64;
 	mach_port_t                    object_name = MACH_PORT_NULL;
@@ -337,8 +337,8 @@ 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;
+		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;
 		}

src/common/threads.cpp

@@ -5,9 +5,9 @@
 
 #include <algorithm>
 #include <atomic>
+#include <cerrno>
 #include <chrono>             // IWYU pragma: keep
 #include <condition_variable> // IWYU pragma: keep
-#include <cerrno>
 #include <mutex>
 #include <vector>
 
@@ -28,7 +28,8 @@
 #ifdef KYTY_WIN_CS
 #include <windows.h> // IWYU pragma: keep
 // IWYU pragma: no_include <winbase.h>
-constexpr DWORD KYTY_CS_SPIN_COUNT = 4000;
+constexpr DWORD    KYTY_CS_SPIN_COUNT          = 4000;
+constexpr uint64_t KYTY_SLEEP_SPIN_LIMIT_100NS = 500; // 50 us
 
 #ifndef CREATE_WAITABLE_TIMER_HIGH_RESOLUTION
 #define CREATE_WAITABLE_TIMER_HIGH_RESOLUTION 0x00000002
@@ -39,7 +40,10 @@ static void SleepHighResolution100ns(uint64_t units_100ns) {
 		return;
 	}
 
-	if (units_100ns <= 10000) {
+	// Keep spinning only where a kernel transition is
+	// likely to cost more than the requested delay; ordinary millisecond sleeps use the
+	// per-thread high-resolution waitable timer below.
+	if (units_100ns <= KYTY_SLEEP_SPIN_LIMIT_100NS) {
 		LARGE_INTEGER frequency {};
 		LARGE_INTEGER start {};
 		if (QueryPerformanceFrequency(&frequency) != 0 && QueryPerformanceCounter(&start) != 0 &&

src/common/uniqueFunction.h

@@ -11,7 +11,7 @@ template <typename Result, typename... Args>
 class UniqueFunction {
 	class CallableBase {
 	public:
-		virtual ~CallableBase() = default;
+		virtual ~CallableBase()               = default;
 		virtual Result Invoke(Args&&... args) = 0;
 	};
 
@@ -20,9 +20,7 @@ class UniqueFunction {
 	public:
 		explicit Callable(Function function): m_function(std::move(function)) {}
 
-		Result Invoke(Args&&... args) override {
-			return m_function(std::forward<Args>(args)...);
-		}
+		Result Invoke(Args&&... args) override { return m_function(std::forward<Args>(args)...); }
 
 	private:
 		Function m_function;

src/common/virtualMemory.cpp

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

src/common/virtualMemory.h

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

src/emulator.cpp

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

src/generate_version.cmake

@@ -1,7 +1,8 @@
 set(KYTY_GIT_VERSION "unknown")
+set(KYTY_GIT_HASH "unknown")
 if(GIT_EXECUTABLE)
 	execute_process(
-		COMMAND "${GIT_EXECUTABLE}" describe --tags --always
+		COMMAND "${GIT_EXECUTABLE}" describe --tags --always --dirty
 		WORKING_DIRECTORY "${GIT_WORKING_DIRECTORY}"
 		OUTPUT_VARIABLE KYTY_GIT_VERSION
 		OUTPUT_STRIP_TRAILING_WHITESPACE
@@ -11,5 +12,27 @@ if(GIT_EXECUTABLE)
 	if(NOT GIT_RESULT EQUAL 0)
 		set(KYTY_GIT_VERSION "unknown")
 	endif()
+
+	execute_process(
+		COMMAND "${GIT_EXECUTABLE}" rev-parse --short=7 HEAD
+		WORKING_DIRECTORY "${GIT_WORKING_DIRECTORY}"
+		OUTPUT_VARIABLE KYTY_GIT_HASH
+		OUTPUT_STRIP_TRAILING_WHITESPACE
+		RESULT_VARIABLE GIT_HASH_RESULT
+		ERROR_QUIET
+	)
+	if(NOT GIT_HASH_RESULT EQUAL 0)
+		set(KYTY_GIT_HASH "unknown")
+	else()
+		execute_process(
+			COMMAND "${GIT_EXECUTABLE}" diff-index --quiet HEAD --
+			WORKING_DIRECTORY "${GIT_WORKING_DIRECTORY}"
+			RESULT_VARIABLE GIT_DIRTY_RESULT
+			ERROR_QUIET
+		)
+		if(NOT GIT_DIRTY_RESULT EQUAL 0)
+			string(APPEND KYTY_GIT_HASH "-dirty")
+		endif()
+	endif()
 endif()
 configure_file("${INPUT_FILE}" "${OUTPUT_FILE}")

src/graphics/guest_gpu/command_processor/commandProcessor.h

@@ -15,6 +15,13 @@ bool TestWaitRegMemValue(uint64_t value, uint64_t ref, uint64_t mask, uint32_t f
 
 enum class Pm4ProcessResult { Complete, Blocked };
 
+enum class ContextStateOperation : uint32_t {
+	Clear     = 0,
+	Push      = 1,
+	Pop       = 2,
+	PushClear = 3,
+};
+
 class Pm4Execution {
 public:
 	[[nodiscard]] bool MadeProgress() const noexcept { return m_made_progress; }
@@ -49,24 +56,12 @@ public:
 	KYTY_CLASS_NO_COPY(CommandProcessor);
 
 	void Reset();
+	void ApplyContextStateOperation(ContextStateOperation operation);
 
-	void BufferInit();
-	void BufferFlush();
-	void BufferFlushAndWait();
-	void BufferWait();
-	void BeginReadbackTransaction() {
-		if (m_readback_active) {
-			EXIT("nested command-processor readback transaction\n");
-		}
-		m_readback_active = true;
-	}
-	void EndReadbackTransaction() {
-		if (!m_readback_active) {
-			EXIT("command-processor readback transaction is not active\n");
-		}
-		m_readback_active = false;
-	}
-
+	void            BufferInit();
+	void            BufferFlush();
+	void            BufferFlushAndWait();
+	void            BufferWait();
 	HW::Context&    GetCtx() { return m_ctx; }
 	HW::UserConfig& GetUcfg() { return m_ucfg; }
 	HW::Shader&     GetShCtx() { return m_sh_ctx; }
@@ -83,8 +78,7 @@ public:
 	               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, uint32_t instance_count = 1,
-	                   uint32_t first_vertex = 0, uint32_t first_instance = 0);
+	                   uint32_t render_target_slice_offset = 0);
 	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,
@@ -152,14 +146,19 @@ private:
 	                      uint32_t interrupt_context_id);
 	void ProcessPm4(Pm4Execution& execution, size_t stop_depth);
 	void SuspendPm4();
+	void SubmitNonIndexedDraw(uint32_t vertex_count, uint32_t flags,
+	                          uint32_t render_target_slice_offset, uint32_t first_vertex,
+	                          uint32_t first_instance);
 
 	CommandScheduler&    GetScheduler() const { return m_renderer.GetCommandScheduler(); }
 	RenderCommandBuffer& CurrentBuffer() { return GetScheduler().Current(); }
 	void                 CheckBuffer() const { GetScheduler().CheckActive(); }
 	GpuResourceManager&  GetGpuResources() const { return m_renderer.GetGpuResources(); }
 
-	RenderContext&    m_renderer;
+	RenderContext&   m_renderer;
 	HW::Context      m_ctx;
+	HW::Context      m_saved_ctx;
+	bool             m_context_state_pushed = false;
 	HW::UserConfig   m_ucfg;
 	HW::Shader       m_sh_ctx;
 	HW::UserSgprType m_user_data_marker                 = HW::UserSgprType::Unknown;
@@ -168,12 +167,12 @@ private:
 	uint64_t         m_index_base_addr                  = 0;
 	uint64_t         m_draw_indirect_args_base_addr     = 0;
 	uint64_t         m_dispatch_indirect_args_base_addr = 0;
-	uint32_t         m_num_instances                    = 1;
+	// Persistent draw state: indirect draws update it for subsequent draws.
+	uint32_t m_num_instances = 1;
 
-	uint32_t m_de_count          = 0;
-	uint32_t m_ce_count          = 0;
-	bool     m_ce_complete       = false;
-	bool     m_readback_active = false;
+	uint32_t m_de_count    = 0;
+	uint32_t m_ce_count    = 0;
+	bool     m_ce_complete = false;
 
 	uint32_t m_const_ram[0x3000] = {0};
 

src/graphics/guest_gpu/command_processor/pm4Dispatch.cpp

@@ -254,6 +254,7 @@ constexpr auto MakeCustomOpcodeDispatchTable() {
 	g_cp_op_custom_func[Pm4::R_WAIT_MEM_64]    = CpOpWaitRegMem64;
 	g_cp_op_custom_func[Pm4::R_FLIP]           = CpOpFlip;
 	g_cp_op_custom_func[Pm4::R_RELEASE_MEM]    = CpOpReleaseMem;
+	g_cp_op_custom_func[Pm4::R_CONTEXT_STATE]  = CpOpContextState;
 
 	return g_cp_op_custom_func;
 }

src/graphics/guest_gpu/command_processor/pm4Dispatch.h

@@ -121,6 +121,7 @@ uint32_t HwUcSetGeIndexOffset(CommandProcessor&, uint32_t, uint32_t, const uint3
 uint32_t HwUcSetGdsOaRegisters(CommandProcessor&, uint32_t, uint32_t, const uint32_t*, uint32_t);
 
 uint32_t CpOpNop(CommandProcessor&, uint32_t, const uint32_t*, uint32_t, uint32_t);
+uint32_t CpOpContextState(CommandProcessor&, uint32_t, const uint32_t*, uint32_t, uint32_t);
 uint32_t CpOpSetBase(CommandProcessor&, uint32_t, const uint32_t*, uint32_t, uint32_t);
 uint32_t CpOpClearState(CommandProcessor&, uint32_t, const uint32_t*, uint32_t, uint32_t);
 uint32_t CpOpIndexBase(CommandProcessor&, uint32_t, const uint32_t*, uint32_t, uint32_t);

src/graphics/guest_gpu/command_processor/pm4Handlers.cpp

@@ -18,7 +18,9 @@
 #include <algorithm>
 #include <array>
 #include <atomic>
+#include <bit>
 #include <cstdio>
+#include <cstring>
 #include <vector>
 
 #define KYTY_HW_CTX_PARSER_ARGS                                                                    \
@@ -895,23 +897,44 @@ KYTY_HW_CTX_PARSER(HwCtxSetModeControl) {
 	return 1;
 }
 
-static void HwCtxIgnorePolyOffsetRegister(uint32_t cmd_offset, uint32_t value) {
-	static std::atomic<uint32_t> log_count = 0;
+static void HwCtxSetPolyOffsetRegister(CommandProcessor& cp, uint32_t cmd_offset, uint32_t value) {
+	HW::PolyOffset offset = cp.GetCtx().GetPolyOffset();
 
-	auto count = log_count.fetch_add(1);
-	if (count < 8) {
-		LOGF_COLOR(Log::Color::Red,
-		           "\t temporary: ignoring polygon offset context register 0x%03" PRIx32
-		           " = 0x%08" PRIx32 " (depth bias not implemented)\n",
-		           cmd_offset, value);
+	switch (cmd_offset) {
+		case Pm4::PA_SU_POLY_OFFSET_DB_FMT_CNTL: {
+			const auto neg_bits =
+			    KYTY_PM4_GET(value, PA_SU_POLY_OFFSET_DB_FMT_CNTL, NEG_NUM_DB_BITS);
+			offset.neg_num_db_bits = std::bit_cast<int8_t>(static_cast<uint8_t>(neg_bits));
+			offset.db_is_float_fmt =
+			    KYTY_PM4_GET(value, PA_SU_POLY_OFFSET_DB_FMT_CNTL, DB_IS_FLOAT_FMT) != 0;
+			break;
+		}
+		case Pm4::PA_SU_POLY_OFFSET_CLAMP: offset.clamp = std::bit_cast<float>(value); break;
+		case Pm4::PA_SU_POLY_OFFSET_FRONT_SCALE:
+			offset.front_scale = std::bit_cast<float>(value);
+			break;
+		case Pm4::PA_SU_POLY_OFFSET_FRONT_OFFSET:
+			offset.front_offset = std::bit_cast<float>(value);
+			break;
+		case Pm4::PA_SU_POLY_OFFSET_BACK_SCALE:
+			offset.back_scale = std::bit_cast<float>(value);
+			break;
+		case Pm4::PA_SU_POLY_OFFSET_BACK_OFFSET:
+			offset.back_offset = std::bit_cast<float>(value);
+			break;
+		default:
+			EXIT("unknown polygon offset context register 0x%03" PRIx32 " = 0x%08" PRIx32 "\n",
+			     cmd_offset, value);
 	}
+
+	cp.GetCtx().SetPolyOffset(offset);
 }
 
 KYTY_HW_CTX_PARSER(HwCtxSetPolyOffsetRegisters) {
 	auto num_values = KYTY_PM4_LEN(cmd_id) - 2u;
 
 	for (uint32_t i = 0; i < num_values; i++) {
-		HwCtxIgnorePolyOffsetRegister(cmd_offset + i, buffer[i]);
+		HwCtxSetPolyOffsetRegister(cp, cmd_offset + i, buffer[i]);
 	}
 
 	return num_values;
@@ -1312,26 +1335,36 @@ static void HwShSetCsRegister(CommandProcessor& cp, uint32_t cmd_offset, uint32_
 			    (value >> Pm4::COMPUTE_PGM_RSRC1_VGPRS_SHIFT) & Pm4::COMPUTE_PGM_RSRC1_VGPRS_MASK;
 			cs_regs.sgprs =
 			    (value >> Pm4::COMPUTE_PGM_RSRC1_SGPRS_SHIFT) & Pm4::COMPUTE_PGM_RSRC1_SGPRS_MASK;
-			cs_regs.bulky =
-			    (value >> Pm4::COMPUTE_PGM_RSRC1_BULKY_SHIFT) & Pm4::COMPUTE_PGM_RSRC1_BULKY_MASK;
-			cs_regs.wave_size = (((value >> Pm4::COMPUTE_PGM_RSRC1_W32_EN_SHIFT) &
-			                      Pm4::COMPUTE_PGM_RSRC1_W32_EN_MASK) != 0u)
-			                        ? 32u
-			                        : 64u;
+			cs_regs.float_mode    = (value >> Pm4::COMPUTE_PGM_RSRC1_FLOAT_MODE_SHIFT) &
+			                        Pm4::COMPUTE_PGM_RSRC1_FLOAT_MODE_MASK;
+			cs_regs.dx10_clamp    = ((value >> Pm4::COMPUTE_PGM_RSRC1_DX10_CLAMP_SHIFT) &
+			                         Pm4::COMPUTE_PGM_RSRC1_DX10_CLAMP_MASK) != 0u;
+			cs_regs.ieee_mode     = ((value >> Pm4::COMPUTE_PGM_RSRC1_IEEE_MODE_SHIFT) &
+			                         Pm4::COMPUTE_PGM_RSRC1_IEEE_MODE_MASK) != 0u;
+			cs_regs.fp16_overflow = ((value >> Pm4::COMPUTE_PGM_RSRC1_FP16_OVFL_SHIFT) &
+			                         Pm4::COMPUTE_PGM_RSRC1_FP16_OVFL_MASK) != 0u;
+			cs_regs.bulky         = ((value >> Pm4::COMPUTE_PGM_RSRC1_BULKY_SHIFT) &
+			                         Pm4::COMPUTE_PGM_RSRC1_BULKY_MASK) != 0u;
+			cs_regs.threadgroup_configuration = ((value >> Pm4::COMPUTE_PGM_RSRC1_WGP_MODE_SHIFT) &
+			                                     Pm4::COMPUTE_PGM_RSRC1_WGP_MODE_MASK) != 0u;
+			cs_regs.wave_size                 = (((value >> Pm4::COMPUTE_PGM_RSRC1_W32_EN_SHIFT) &
+			                                      Pm4::COMPUTE_PGM_RSRC1_W32_EN_MASK) != 0u)
+			                                        ? 32u
+			                                        : 64u;
 			break;
 		case Pm4::COMPUTE_PGM_RSRC2:
-			cs_regs.scratch_en     = (value >> Pm4::COMPUTE_PGM_RSRC2_SCRATCH_EN_SHIFT) &
-			                         Pm4::COMPUTE_PGM_RSRC2_SCRATCH_EN_MASK;
+			cs_regs.scratch_en     = ((value >> Pm4::COMPUTE_PGM_RSRC2_SCRATCH_EN_SHIFT) &
+			                          Pm4::COMPUTE_PGM_RSRC2_SCRATCH_EN_MASK) != 0u;
 			cs_regs.user_sgpr      = (value >> Pm4::COMPUTE_PGM_RSRC2_USER_SGPR_SHIFT) &
 			                         Pm4::COMPUTE_PGM_RSRC2_USER_SGPR_MASK;
-			cs_regs.tgid_x_en      = (value >> Pm4::COMPUTE_PGM_RSRC2_TGID_X_EN_SHIFT) &
-			                         Pm4::COMPUTE_PGM_RSRC2_TGID_X_EN_MASK;
-			cs_regs.tgid_y_en      = (value >> Pm4::COMPUTE_PGM_RSRC2_TGID_Y_EN_SHIFT) &
-			                         Pm4::COMPUTE_PGM_RSRC2_TGID_Y_EN_MASK;
-			cs_regs.tgid_z_en      = (value >> Pm4::COMPUTE_PGM_RSRC2_TGID_Z_EN_SHIFT) &
-			                         Pm4::COMPUTE_PGM_RSRC2_TGID_Z_EN_MASK;
-			cs_regs.tg_size_en     = (value >> Pm4::COMPUTE_PGM_RSRC2_TG_SIZE_EN_SHIFT) &
-			                         Pm4::COMPUTE_PGM_RSRC2_TG_SIZE_EN_MASK;
+			cs_regs.tgid_x_en      = ((value >> Pm4::COMPUTE_PGM_RSRC2_TGID_X_EN_SHIFT) &
+			                          Pm4::COMPUTE_PGM_RSRC2_TGID_X_EN_MASK) != 0u;
+			cs_regs.tgid_y_en      = ((value >> Pm4::COMPUTE_PGM_RSRC2_TGID_Y_EN_SHIFT) &
+			                          Pm4::COMPUTE_PGM_RSRC2_TGID_Y_EN_MASK) != 0u;
+			cs_regs.tgid_z_en      = ((value >> Pm4::COMPUTE_PGM_RSRC2_TGID_Z_EN_SHIFT) &
+			                          Pm4::COMPUTE_PGM_RSRC2_TGID_Z_EN_MASK) != 0u;
+			cs_regs.tg_size_en     = ((value >> Pm4::COMPUTE_PGM_RSRC2_TG_SIZE_EN_SHIFT) &
+			                          Pm4::COMPUTE_PGM_RSRC2_TG_SIZE_EN_MASK) != 0u;
 			cs_regs.tidig_comp_cnt = (value >> Pm4::COMPUTE_PGM_RSRC2_TIDIG_COMP_CNT_SHIFT) &
 			                         Pm4::COMPUTE_PGM_RSRC2_TIDIG_COMP_CNT_MASK;
 			cs_regs.lds_size       = (value >> Pm4::COMPUTE_PGM_RSRC2_LDS_SIZE_SHIFT) &
@@ -1917,17 +1950,23 @@ KYTY_CP_OP_PARSER(CpOpCopyData) {
 
 	EXIT_NOT_IMPLEMENTED(cmd_id != KYTY_PM4(6, Pm4::IT_COPY_DATA, 0u));
 
-	const uint32_t control       = buffer[0];
-	const uint32_t src_sel       = ((control & 0xfu) << 1u) | ((control >> 30u) & 0x1u);
-	const uint32_t dst_sel       = ((control >> 8u) & 0xfu) << 1u;
-	const uint8_t  src_cache     = static_cast<uint8_t>((control >> 13u) & 0x3u);
-	const uint8_t  dst_cache     = static_cast<uint8_t>((control >> 25u) & 0x3u);
-	const uint8_t  write_confirm = static_cast<uint8_t>((control >> 20u) & 0x1u);
-	const uint32_t num_bytes     = ((control >> 16u) & 0x1u) != 0 ? 8u : 4u;
-	const uint64_t src           = buffer[1] | (static_cast<uint64_t>(buffer[2]) << 32u);
-	const uint64_t dst           = buffer[3] | (static_cast<uint64_t>(buffer[4]) << 32u);
-	if (src_sel == (9u << 1u)) {
-		if (dst_sel != (2u << 1u) || dst == 0 || (dst & (num_bytes - 1u)) != 0) {
+	const uint32_t control             = buffer[0];
+	const uint32_t src_sel             = ((control & 0xfu) << 1u) | ((control >> 30u) & 0x1u);
+	const uint32_t dst_sel             = ((control >> 8u) & 0xfu) << 1u;
+	const uint8_t  src_cache           = static_cast<uint8_t>((control >> 13u) & 0x3u);
+	const uint8_t  dst_cache           = static_cast<uint8_t>((control >> 25u) & 0x3u);
+	const uint8_t  write_confirm       = static_cast<uint8_t>((control >> 20u) & 0x1u);
+	const uint32_t num_bytes           = ((control >> 16u) & 0x1u) != 0 ? 8u : 4u;
+	const uint64_t src                 = buffer[1] | (static_cast<uint64_t>(buffer[2]) << 32u);
+	const uint64_t dst                 = buffer[3] | (static_cast<uint64_t>(buffer[4]) << 32u);
+	uint32_t       reference_clock_dst = 0;
+	switch (src_sel) {
+		case 9u: reference_clock_dst = 2u; break;
+		case 18u: reference_clock_dst = 4u; break;
+		default: break;
+	}
+	if (reference_clock_dst != 0) {
+		if (dst_sel != reference_clock_dst || dst == 0 || (dst & (num_bytes - 1u)) != 0) {
 			EXIT("unsupported reference-clock copyData, src_sel=0x%02" PRIx32
 			     " dst_sel=0x%02" PRIx32 " dst=0x%016" PRIx64 " size=%u\n",
 			     src_sel, dst_sel, dst, num_bytes);
@@ -2096,11 +2135,24 @@ KYTY_CP_OP_PARSER(CpOpClearState) {
 	EXIT_NOT_IMPLEMENTED(cmd_id != 0xc0001200);
 	EXIT_NOT_IMPLEMENTED((buffer[0] & ~0xfu) != 0);
 
-	cp.Reset();
+	cp.ApplyContextStateOperation(ContextStateOperation::Clear);
 
 	return 1;
 }
 
+KYTY_CP_OP_PARSER(CpOpContextState) {
+	KYTY_PROFILER_FUNCTION();
+
+	const auto packet_size_dw = KYTY_PM4_LEN(cmd_id);
+	EXIT_NOT_IMPLEMENTED(KYTY_PM4_R(cmd_id) != Pm4::R_CONTEXT_STATE);
+	EXIT_NOT_IMPLEMENTED(packet_size_dw != 3u && packet_size_dw != 5u);
+	EXIT_NOT_IMPLEMENTED(buffer[0] > static_cast<uint32_t>(ContextStateOperation::PushClear));
+
+	cp.ApplyContextStateOperation(static_cast<ContextStateOperation>(buffer[0]));
+
+	return packet_size_dw - 1u;
+}
+
 KYTY_CP_OP_PARSER(CpOpDumpConstRam) {
 	KYTY_PROFILER_FUNCTION();
 
@@ -3390,6 +3442,12 @@ void GraphicsInitJmpTablesCxIndirect() {
 	g_hw_ctx_indirect_func[Pm4::DB_COUNT_CONTROL] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
 		HwCtxIgnoreDepthMetadataRegister(cmd_offset, value);
 	};
+	for (auto cmd_offset = Pm4::DB_SRESULTS_COMPARE_STATE0;
+	     cmd_offset <= Pm4::DB_SRESULTS_COMPARE_STATE1; cmd_offset++) {
+		g_hw_ctx_indirect_func[cmd_offset] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
+			HwCtxIgnoreDepthMetadataRegister(cmd_offset, value);
+		};
+	}
 	g_hw_ctx_indirect_func[Pm4::DB_RENDER_OVERRIDE] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
 		HwCtxIgnoreDepthMetadataRegister(cmd_offset, value);
 	};
@@ -3670,22 +3728,22 @@ void GraphicsInitJmpTablesCxIndirect() {
 	};
 
 	g_hw_ctx_indirect_func[Pm4::PA_SU_POLY_OFFSET_DB_FMT_CNTL] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
-		HwCtxIgnorePolyOffsetRegister(cmd_offset, value);
+		HwCtxSetPolyOffsetRegister(cp, cmd_offset, value);
 	};
 	g_hw_ctx_indirect_func[Pm4::PA_SU_POLY_OFFSET_CLAMP] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
-		HwCtxIgnorePolyOffsetRegister(cmd_offset, value);
+		HwCtxSetPolyOffsetRegister(cp, cmd_offset, value);
 	};
 	g_hw_ctx_indirect_func[Pm4::PA_SU_POLY_OFFSET_FRONT_SCALE] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
-		HwCtxIgnorePolyOffsetRegister(cmd_offset, value);
+		HwCtxSetPolyOffsetRegister(cp, cmd_offset, value);
 	};
 	g_hw_ctx_indirect_func[Pm4::PA_SU_POLY_OFFSET_FRONT_OFFSET] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
-		HwCtxIgnorePolyOffsetRegister(cmd_offset, value);
+		HwCtxSetPolyOffsetRegister(cp, cmd_offset, value);
 	};
 	g_hw_ctx_indirect_func[Pm4::PA_SU_POLY_OFFSET_BACK_SCALE] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
-		HwCtxIgnorePolyOffsetRegister(cmd_offset, value);
+		HwCtxSetPolyOffsetRegister(cp, cmd_offset, value);
 	};
 	g_hw_ctx_indirect_func[Pm4::PA_SU_POLY_OFFSET_BACK_OFFSET] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
-		HwCtxIgnorePolyOffsetRegister(cmd_offset, value);
+		HwCtxSetPolyOffsetRegister(cp, cmd_offset, value);
 	};
 
 	g_hw_ctx_indirect_func[Pm4::DB_Z_INFO] = [](KYTY_HW_CTX_INDIRECT_ARGS) {
@@ -4208,29 +4266,28 @@ void GraphicsInitJmpTablesShIndirect() {
 	};
 
 	g_hw_sh_indirect_func[Pm4::SPI_SHADER_PGM_RSRC1_HS] = [](KYTY_HW_SH_INDIRECT_ARGS) {
-		HW::VsShaderResource1 r1;
-		r1.vgprs                = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, VGPRS);
-		r1.sgprs                = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, SGPRS);
-		r1.priority             = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, PRIORITY);
-		r1.float_mode           = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, FLOAT_MODE);
-		r1.dx10_clamp           = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, DX10_CLAMP) != 0;
-		r1.ieee_mode            = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, IEEE_MODE) != 0;
-		r1.vgpr_component_count = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, VGPR_COMP_CNT);
-		r1.cu_group_enable = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, CU_GROUP_ENABLE) != 0;
+		HW::HsShaderResource1 r1;
+		r1.vgprs      = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, VGPRS);
+		r1.priority   = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, PRIORITY);
+		r1.float_mode = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, FLOAT_MODE);
+		r1.dx10_clamp = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, DX10_CLAMP) != 0;
+		r1.debug_mode = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, DEBUG_MODE) != 0;
+		r1.ieee_mode  = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, IEEE_MODE) != 0;
 		r1.require_forward_progress =
-		    KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, FWD_PROGRESS) != 0;
-		r1.fp16_overflow = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_VS, FP16_OVFL) != 0;
+		    KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, FWD_PROGRESS) != 0;
+		r1.threadgroup_configuration = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, WGP_MODE) != 0;
+		r1.ls_vgpr_component_count = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, LS_VGPR_COMP_CNT);
+		r1.fp16_overflow           = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_HS, FP16_OVFL) != 0;
 		cp.GetShCtx().SetHsShaderResource1(r1);
 	};
 
 	g_hw_sh_indirect_func[Pm4::SPI_SHADER_PGM_RSRC2_HS] = [](KYTY_HW_SH_INDIRECT_ARGS) {
-		HW::VsShaderResource2 r2;
-		r2.scratch_en        = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_VS, SCRATCH_EN) != 0;
-		r2.user_sgpr         = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_VS, USER_SGPR) +
-		                       (KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_VS, USER_SGPR_MSB) << 5u);
-		r2.offchip_lds       = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_VS, OC_LDS_EN) != 0;
-		r2.streamout_enabled = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_VS, SO_EN) != 0;
-		r2.shared_vgprs      = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_VS, SHARED_VGPR_CNT);
+		HW::HsShaderResource2 r2;
+		r2.scratch_en   = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_HS, SCRATCH_EN) != 0;
+		r2.user_sgpr    = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_HS, USER_SGPR) +
+		                  (KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_HS, USER_SGPR_MSB) << 5u);
+		r2.lds_size     = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_HS, LDS_SIZE);
+		r2.shared_vgprs = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_HS, SHARED_VGPR_CNT);
 		cp.GetShCtx().SetHsShaderResource2(r2);
 	};
 
@@ -4318,7 +4375,7 @@ void GraphicsInitJmpTablesShIndirect() {
 		r1.cu_group_enable = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_GS, CU_GROUP_ENABLE) != 0;
 		r1.require_forward_progress =
 		    KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_GS, FWD_PROGRESS) != 0;
-		r1.lds_configuration       = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_GS, WGP_MODE) != 0;
+		r1.threadgroup_configuration = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_GS, WGP_MODE) != 0;
 		r1.gs_vgpr_component_count = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_GS, GS_VGPR_COMP_CNT);
 		r1.fp16_overflow           = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC1_GS, FP16_OVFL) != 0;
 		cp.GetShCtx().SetGsShaderResource1(r1);
@@ -4378,7 +4435,7 @@ void GraphicsInitJmpTablesShIndirect() {
 		r2.wave_cnt_en    = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_PS, WAVE_CNT_EN);
 		r2.extra_lds_size = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_PS, EXTRA_LDS_SIZE);
 		r2.raster_ordered_shading =
-		    KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_PS, LOAD_INTRAWAVE_COLLISION);
+		    KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_PS, RASTER_ORDERED_SHADING);
 		r2.shared_vgprs = KYTY_PM4_GET(value, SPI_SHADER_PGM_RSRC2_PS, SHARED_VGPR_CNT);
 		cp.GetShCtx().SetPsShaderResource2(r2);
 	};

src/graphics/guest_gpu/gpu_defs.h

@@ -60,23 +60,34 @@ enum class GsOutputPrimitiveType : uint32_t {
 };
 
 enum class ChannelLayout : uint32_t {
-	kInvalid     = 0,
-	k8           = 1,
-	k16          = 2,
-	k8_8         = 3,
-	k32          = 4,
-	k16_16       = 5,
-	k11_11_10    = 6,
-	k10_10_10_2  = 9,
-	k8_8_8_8     = 10,
-	k32_32       = 11,
-	k16_16_16_16 = 12,
-	k32_32_32_32 = 14,
-	k5_5_5_1     = 17,
-	k4_4_4_4     = 19,
-	kBc1         = 35,
-	kBc2         = 36,
-	kBc3         = 37,
+	kInvalid         = 0,
+	k8               = 1,
+	k16              = 2,
+	k8_8             = 3,
+	k32              = 4,
+	k16_16           = 5,
+	k11_11_10        = 6,
+	k10_11_11        = 7,
+	k2_10_10_10      = 8,
+	k10_10_10_2      = 9,
+	k8_8_8_8         = 10,
+	k32_32           = 11,
+	k16_16_16_16     = 12,
+	k32_32_32        = 13,
+	k32_32_32_32     = 14,
+	k5_6_5           = 16,
+	k5_5_5_1         = 17,
+	k1_5_5_5         = 18,
+	k4_4_4_4         = 19,
+	k10_10_10_2Float = 31,
+	k9_9_9_5         = 34,
+	kBc1             = 35,
+	kBc2             = 36,
+	kBc3             = 37,
+	kBc4             = 38,
+	kBc5             = 39,
+	kBc6             = 40,
+	kBc7             = 41,
 };
 
 enum class ChannelType : uint32_t {
@@ -374,6 +385,8 @@ enum class BufferFormat : uint32_t {
 	k32_32_32_32UInt    = 75,
 	k32_32_32_32SInt    = 76,
 	k32_32_32_32Float   = 77,
+	k8Srgb              = 128,
+	k8_8Srgb            = 129,
 	k8_8_8_8Srgb        = 130,
 	k9_9_9_5Float       = 132,
 	k5_6_5UNorm         = 133,

src/graphics/guest_gpu/gpu_format.cpp

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

src/graphics/guest_gpu/graphicsRun.cpp

@@ -32,11 +32,11 @@
 
 namespace Libs::Graphics {
 
-static thread_local CommandProcessor* g_current_processor      = nullptr;
-static thread_local Pm4Execution*     g_current_execution      = nullptr;
-static thread_local uint32_t          g_submission_pause_depth = 0;
-static thread_local bool              g_gpu_mutex_owned        = false;
-static thread_local bool              g_gpu_thread             = false;
+static thread_local CommandProcessor* g_current_processor = nullptr;
+static thread_local Pm4Execution*     g_current_execution = nullptr;
+static thread_local bool              g_gpu_mutex_owned   = false;
+static thread_local bool              g_gpu_thread        = false;
+static thread_local GpuState*         g_gpu_state         = nullptr;
 
 class GpuMutexLock final {
 public:
@@ -97,19 +97,16 @@ public:
 	void SubmitCompute(uint32_t queue, uint32_t* cmd_buffer, uint32_t num_dw,
 	                   bool trigger_agc_interrupt_on_done);
 	void SubmitFlipPreparation(uint64_t request_id);
-	void Done();
-	void PauseSubmissions();
-	void ResumeSubmissions();
+	void SuspendPoint();
 	void Shutdown();
-	[[nodiscard]] bool IsStopping();
-	void               SendCommand(Common::UniqueFunction<void>&& command);
-	void               SendCommandSync(Common::UniqueFunction<void>&& command);
-	void SendCommandSyncWithProcessor(Common::UniqueFunction<void, CommandProcessor&>&& command);
-	int  GetFrameNum();
+	[[nodiscard]] bool        IsStopping();
+	void                      SendCommand(Common::UniqueFunction<void>&& command);
+	void                      SendCommandSync(Common::UniqueFunction<void>&& command);
+	int                       GetFrameNum();
 	[[nodiscard]] static bool IsGpuThread() noexcept { return g_gpu_thread; }
 
 private:
-	enum class SubmissionType { Graphics, Compute, FlipPreparation };
+	enum class SubmissionType { Graphics, Compute, FlipPreparation, SuspendPoint };
 
 	struct Submission {
 		SubmissionType type     = SubmissionType::Graphics;
@@ -127,9 +124,8 @@ private:
 		uint64_t       flip_request_id               = 0;
 	};
 
-	void              WaitLocked();
 	void              Enqueue(Submission submission);
-	void              WaitForIdle();
+	void              ProcessCommands();
 	bool              Process(Submission& submission);
 	static void       ThreadRun(void* data);
 	CommandProcessor& GetProcessor(uint32_t queue_id);
@@ -139,16 +135,16 @@ private:
 	Common::Mutex                                  m_queue_mutex;
 	std::mutex                                     m_shutdown_mutex;
 	Common::CondVar                                m_work_available;
-	Common::CondVar                                m_idle;
 	std::array<std::deque<Submission>, QueueCount> m_queues;
 	std::deque<Common::UniqueFunction<void>>       m_commands;
+	std::atomic_uint32_t                           m_pending_commands {0};
 	uint32_t                                       m_next_queue        = 0;
 	uint32_t                                       m_submission_count  = 0;
-	bool                                           m_processing        = false;
-	bool                                           m_graphics_done     = true;
+	bool                                           m_reset_graphics    = true;
 	bool                                           m_accepting         = true;
 	bool                                           m_stopping          = false;
 	bool                                           m_shutdown_complete = false;
+	std::binary_semaphore                          m_suspend_point_slot {1};
 
 	std::unique_ptr<CommandProcessor>                                m_gfx_cp;
 	std::array<std::unique_ptr<CommandProcessor>, ComputeQueueCount> m_compute_cp;
@@ -156,6 +152,8 @@ private:
 	uint64_t        m_submit_id = 0;
 	std::atomic_int m_done_num  = 0;
 	std::jthread    m_thread;
+
+	friend class CommandProcessor;
 };
 
 static bool GraphicsRunDebugDumpEnabled() {
@@ -198,9 +196,25 @@ void GpuState::SendCommand(Common::UniqueFunction<void>&& command) {
 	Common::LockGuard lock(m_queue_mutex);
 	EXIT_IF(!m_accepting);
 	m_commands.push_back(std::move(command));
+	m_pending_commands.fetch_add(1, std::memory_order_release);
 	m_work_available.Signal();
 }
 
+void GpuState::ProcessCommands() {
+	EXIT_IF(!IsGpuThread());
+	while (m_pending_commands.load(std::memory_order_acquire) != 0) {
+		Common::UniqueFunction<void> command;
+		{
+			Common::LockGuard lock(m_queue_mutex);
+			EXIT_IF(m_commands.empty());
+			command = std::move(m_commands.front());
+			m_commands.pop_front();
+			EXIT_IF(m_pending_commands.fetch_sub(1, std::memory_order_acq_rel) == 0);
+		}
+		command();
+	}
+}
+
 void GpuState::SendCommandSync(Common::UniqueFunction<void>&& command) {
 	EXIT_IF(!command);
 	if (IsGpuThread()) {
@@ -215,17 +229,6 @@ void GpuState::SendCommandSync(Common::UniqueFunction<void>&& command) {
 	done.acquire();
 }
 
-void GpuState::SendCommandSyncWithProcessor(
-    Common::UniqueFunction<void, CommandProcessor&>&& command) {
-	EXIT_IF(!command);
-	SendCommandSync([this, operation = std::move(command)]() mutable {
-		EXIT_IF(g_current_processor != nullptr);
-		g_current_processor = m_gfx_cp.get();
-		operation(*m_gfx_cp);
-		g_current_processor = nullptr;
-	});
-}
-
 void GpuState::Submit(uint32_t* cmd_draw_buffer, uint32_t num_draw_dw, uint32_t* cmd_const_buffer,
                       uint32_t num_const_dw, bool trigger_agc_interrupt_on_done) {
 	GpuMutexLock lock(m_submission_mutex);
@@ -235,8 +238,8 @@ void GpuState::Submit(uint32_t* cmd_draw_buffer, uint32_t num_draw_dw, uint32_t*
 	submission.commands                      = OwnedCmdBuffer(cmd_draw_buffer, num_draw_dw);
 	submission.constant_commands             = OwnedCmdBuffer(cmd_const_buffer, num_const_dw);
 	submission.trigger_agc_interrupt_on_done = trigger_agc_interrupt_on_done;
-	submission.reset_processor               = m_graphics_done;
-	m_graphics_done                          = false;
+	submission.reset_processor               = m_reset_graphics;
+	m_reset_graphics                         = false;
 	Enqueue(std::move(submission));
 }
 
@@ -262,32 +265,27 @@ void GpuState::SubmitFlipPreparation(uint64_t request_id) {
 	Submission   submission;
 	submission.type            = SubmissionType::FlipPreparation;
 	submission.queue_id        = 0;
-	submission.reset_processor = m_graphics_done;
+	submission.reset_processor = m_reset_graphics;
 	submission.flip_request_id = request_id;
-	m_graphics_done            = false;
+	m_reset_graphics           = false;
 	Enqueue(std::move(submission));
 }
 
-void GpuState::Done() {
+void GpuState::SuspendPoint() {
+	EXIT_IF(IsGpuThread());
+	m_suspend_point_slot.acquire();
 	GpuMutexLock lock(m_submission_mutex);
-	if (IsGpuThread()) {
-		m_gfx_cp->BufferWait();
-	} else {
-		WaitLocked();
-	}
-	m_graphics_done = true;
-	m_done_num++;
+	Submission   submission;
+	submission.type     = SubmissionType::SuspendPoint;
+	submission.queue_id = 0;
+	m_reset_graphics    = false;
+	Enqueue(std::move(submission));
 }
 
 int GpuState::GetFrameNum() {
 	return m_done_num;
 }
 
-void GpuState::WaitLocked() {
-	WaitForIdle();
-	SendCommandSync([this] { m_gfx_cp->BufferWait(); });
-}
-
 CommandProcessor& GpuState::GetProcessor(uint32_t queue_id) {
 	EXIT_IF(queue_id >= QueueCount);
 	if (queue_id == 0) {
@@ -306,15 +304,44 @@ void CommandProcessor::Reset() {
 	m_sh_ctx.Reset();
 	m_ucfg.Reset();
 	m_ctx.Reset();
+	m_saved_ctx.Reset();
+	m_context_state_pushed             = false;
 	m_index_type_and_size              = 0;
 	m_index_buffer_size                = 0;
+	m_index_base_addr                  = 0;
 	m_user_data_marker                 = HW::UserSgprType::Unknown;
 	m_draw_indirect_args_base_addr     = 0;
 	m_dispatch_indirect_args_base_addr = 0;
+	m_num_instances                    = 1;
+	m_predicate_skip                   = false;
 
 	std::memset(m_const_ram, 0, sizeof(m_const_ram));
 }
 
+void CommandProcessor::ApplyContextStateOperation(ContextStateOperation operation) {
+	switch (operation) {
+		case ContextStateOperation::Clear: m_ctx.Reset(); break;
+		case ContextStateOperation::Push:
+			EXIT_IF(m_context_state_pushed);
+			m_saved_ctx            = m_ctx;
+			m_context_state_pushed = true;
+			break;
+		case ContextStateOperation::Pop:
+			EXIT_IF(!m_context_state_pushed);
+			m_ctx                  = m_saved_ctx;
+			m_saved_ctx            = {};
+			m_context_state_pushed = false;
+			break;
+		case ContextStateOperation::PushClear:
+			EXIT_IF(m_context_state_pushed);
+			m_saved_ctx            = m_ctx;
+			m_context_state_pushed = true;
+			m_ctx.Reset();
+			break;
+		default: EXIT("unknown context state operation: %u\n", static_cast<uint32_t>(operation));
+	}
+}
+
 void CommandProcessor::BufferInit() {
 	GetScheduler().Begin(m_ctx, m_ucfg, m_sh_ctx);
 }
@@ -410,8 +437,13 @@ void CommandProcessor::WriteData(uint32_t* dst, const uint32_t* src, uint32_t dw
 	const uint32_t increment    = (write_control >> 16u) & 0x1u;
 	const uint32_t write_confirm = (write_control >> 20u) & 0x1u;
 
-	if (dst_sel != 0 && dst_sel != 2 && dst_sel != 4 && dst_sel != 5) {
-		EXIT("unsupported writeData destination selector 0x%02" PRIx32 "\n", dst_sel);
+	switch (dst_sel) {
+		case 0:
+		case 2:
+		case 4:
+		case 5:
+		case 6: break;
+		default: EXIT("unsupported writeData destination selector 0x%02" PRIx32 "\n", dst_sel);
 	}
 	EXIT_NOT_IMPLEMENTED(increment != 0);
 
@@ -495,18 +527,12 @@ void GpuState::Enqueue(Submission submission) {
 	m_work_available.Signal();
 }
 
-void GpuState::WaitForIdle() {
-	Common::LockGuard lock(m_queue_mutex);
-	while (m_processing || !m_commands.empty() || m_submission_count != 0) {
-		m_idle.Wait(&m_queue_mutex);
-	}
-}
-
 void GpuState::ThreadRun(void* data) {
 	auto* gpu = static_cast<GpuState*>(data);
 	EXIT_IF(gpu == nullptr);
 	KYTY_PROFILER_THREAD("Thread_Gpu");
 	g_gpu_thread = true;
+	g_gpu_state  = gpu;
 
 	for (;;) {
 		Submission                   submission;
@@ -516,18 +542,14 @@ void GpuState::ThreadRun(void* data) {
 		{
 			Common::LockGuard lock(gpu->m_queue_mutex);
 			while (gpu->m_commands.empty() && gpu->m_submission_count == 0 && !gpu->m_stopping) {
-				gpu->m_processing = false;
-				gpu->m_idle.Signal();
 				gpu->m_work_available.Wait(&gpu->m_queue_mutex);
 			}
 			if (gpu->m_stopping && gpu->m_commands.empty() && gpu->m_submission_count == 0) {
-				gpu->m_processing = false;
-				gpu->m_idle.SignalAll();
 				should_stop = true;
 			} else if (!gpu->m_commands.empty()) {
 				command = std::move(gpu->m_commands.front());
 				gpu->m_commands.pop_front();
-				gpu->m_processing = true;
+				EXIT_IF(gpu->m_pending_commands.fetch_sub(1, std::memory_order_acq_rel) == 0);
 			} else {
 				int selected_queue = -1;
 				for (uint32_t offset = 0; offset < QueueCount; offset++) {
@@ -538,7 +560,6 @@ void GpuState::ThreadRun(void* data) {
 					}
 				}
 				if (selected_queue < 0) {
-					gpu->m_processing = false;
 					gpu->m_work_available.WaitFor(&gpu->m_queue_mutex, 100);
 					for (auto& queue: gpu->m_queues) {
 						if (!queue.empty()) {
@@ -552,12 +573,12 @@ void GpuState::ThreadRun(void* data) {
 				queue.pop_front();
 				gpu->m_submission_count--;
 				gpu->m_next_queue = (static_cast<uint32_t>(selected_queue) + 1) % QueueCount;
-				gpu->m_processing = true;
 				has_submission    = true;
 			}
 		}
 		if (should_stop) {
 			gpu->m_gfx_cp->BufferWait();
+			g_gpu_state  = nullptr;
 			g_gpu_thread = false;
 			return;
 		}
@@ -566,11 +587,6 @@ void GpuState::ThreadRun(void* data) {
 			EXIT_IF(g_current_processor != nullptr);
 			command();
 
-			Common::LockGuard lock(gpu->m_queue_mutex);
-			gpu->m_processing = false;
-			if (gpu->m_commands.empty() && gpu->m_submission_count == 0) {
-				gpu->m_idle.SignalAll();
-			}
 			continue;
 		}
 
@@ -589,16 +605,19 @@ void GpuState::ThreadRun(void* data) {
 				}
 			}
 		}
-		gpu->m_processing = false;
-		if (gpu->m_commands.empty() && gpu->m_submission_count == 0) {
-			gpu->m_idle.SignalAll();
-		}
 	}
 }
 
 bool GpuState::Process(Submission& submission) {
 	auto&      cp          = GetProcessor(submission.queue_id);
 	const bool first_slice = !submission.started;
+	if (submission.type == SubmissionType::SuspendPoint) {
+		EXIT_IF(!first_slice || submission.queue_id != 0);
+		cp.Reset();
+		m_done_num++;
+		m_suspend_point_slot.release();
+		return true;
+	}
 
 	if (first_slice && submission.reset_processor) {
 		cp.Reset();
@@ -686,31 +705,12 @@ bool GpuState::Process(Submission& submission) {
 			m_renderer.GetGpuResources().RunGarbageCollector();
 			cp.PrepareCpuFlip(submission.flip_request_id);
 			break;
+		case SubmissionType::SuspendPoint: EXIT("unreachable suspend point submission\n");
 	}
 
 	return complete;
 }
 
-void GpuState::PauseSubmissions() {
-	if (g_gpu_mutex_owned) {
-		EXIT("GPU submissions are already paused by this thread\n");
-	}
-	g_gpu_mutex_owned = true;
-	m_submission_mutex.Lock();
-	if (!IsGpuThread()) {
-		WaitLocked();
-	}
-	m_renderer.GetCommandScheduler().DrainPriorityOperations();
-}
-
-void GpuState::ResumeSubmissions() {
-	if (!g_gpu_mutex_owned) {
-		EXIT("GPU submissions resumed without an active pause\n");
-	}
-	m_submission_mutex.Unlock();
-	g_gpu_mutex_owned = false;
-}
-
 Pm4ProcessResult CommandProcessor::Process(Pm4Execution& execution, uint32_t* buffer,
                                            uint32_t size_dw) {
 	KYTY_PROFILER_BLOCK("CommandProcessor::Process");
@@ -759,6 +759,9 @@ void CommandProcessor::SuspendPm4() {
 
 void CommandProcessor::ProcessPm4(Pm4Execution& execution, size_t stop_depth) {
 	while (execution.m_buffer_stack.size() > stop_depth) {
+		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];
 		if (cursor.deferred_advance_dw != 0) {
@@ -962,9 +965,8 @@ void CommandProcessor::DrawIndexOffset(uint32_t index_offset, uint32_t index_cou
 	auto* index_addr = reinterpret_cast<const void*>(
 	    m_index_base_addr + static_cast<uint64_t>(index_offset) * index_size);
 
-	m_renderer.GetRenderExecutor().DrawIndex(m_submit_id, CurrentBuffer(),
-	                                             m_index_type_and_size, index_count, index_addr,
-	                                             flags, 1, m_num_instances);
+	m_renderer.GetRenderExecutor().DrawIndex(m_submit_id, CurrentBuffer(), m_index_type_and_size,
+	                                         index_count, index_addr, flags, 1, m_num_instances);
 }
 
 void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiator, bool indexed) {
@@ -1002,8 +1004,9 @@ void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiato
 				     args.start_vertex_location, args.start_instance_location);
 			}
 		}
-		DrawIndexAuto(args.vertex_count_per_instance, 0, 0, args.instance_count,
-		              args.start_vertex_location, args.start_instance_location);
+		m_num_instances = args.instance_count;
+		SubmitNonIndexedDraw(args.vertex_count_per_instance, 0, 0, args.start_vertex_location,
+		                     args.start_instance_location);
 		return;
 	}
 
@@ -1043,6 +1046,7 @@ 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);
 }
@@ -1100,8 +1104,9 @@ void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_coun
 					     args->start_vertex_location, args->start_instance_location);
 				}
 			}
-			DrawIndexAuto(args->vertex_count_per_instance, 0, 0, args->instance_count,
-			              args->start_vertex_location, args->start_instance_location);
+			m_num_instances = args->instance_count;
+			SubmitNonIndexedDraw(args->vertex_count_per_instance, 0, 0, args->start_vertex_location,
+			                     args->start_instance_location);
 			continue;
 		}
 
@@ -1142,6 +1147,7 @@ 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);
 	}
@@ -1150,9 +1156,9 @@ void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_coun
 void CommandProcessor::DispatchDirect(uint32_t thread_group_x, uint32_t thread_group_y,
                                       uint32_t thread_group_z, uint32_t mode) {
 	uint32_t frame_num = 0;
-	uint32_t local_x   = 1;
-	uint32_t local_y   = 1;
-	uint32_t local_z   = 1;
+	// uint32_t local_x   = 1;
+	// uint32_t local_y   = 1;
+	// uint32_t local_z   = 1;
 
 	{
 		CheckBuffer();
@@ -1176,9 +1182,9 @@ void CommandProcessor::DispatchDirect(uint32_t thread_group_x, uint32_t thread_g
 		}
 
 		const auto& cs = m_sh_ctx.GetCs().cs_regs;
-		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);
+		// 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)) {
@@ -1190,31 +1196,31 @@ void CommandProcessor::DispatchDirect(uint32_t thread_group_x, uint32_t thread_g
 			}
 		}
 
-		m_renderer.GetRenderExecutor().DispatchDirect(
-		    m_submit_id, CurrentBuffer(), thread_group_x, thread_group_y, thread_group_z, mode);
+		m_renderer.GetRenderExecutor().DispatchDirect(m_submit_id, CurrentBuffer(), thread_group_x,
+		                                              thread_group_y, thread_group_z, mode);
 	}
 
-	constexpr uint32_t DispatchInitiatorUseThreadDimensions = 1u << 5u;
+	/*constexpr uint32_t DispatchInitiatorUseThreadDimensions = 1u << 5u;
 	auto               group_count = [](uint32_t threads, uint32_t group_size) {
-		return (threads == 0
-		            ? 0u
-		            : (threads + std::max(group_size, 1u) - 1u) / std::max(group_size, 1u));
+	    return (threads == 0
+	                ? 0u
+	                : (threads + std::max(group_size, 1u) - 1u) / std::max(group_size, 1u));
 	};
 
 	auto groups_x = thread_group_x;
 	auto groups_y = thread_group_y;
 	auto groups_z = thread_group_z;
 	if ((mode & DispatchInitiatorUseThreadDimensions) != 0) {
-		groups_x = group_count(thread_group_x, local_x);
-		groups_y = group_count(thread_group_y, local_y);
-		groups_z = group_count(thread_group_z, local_z);
+	    groups_x = group_count(thread_group_x, local_x);
+	    groups_y = group_count(thread_group_y, local_y);
+	    groups_z = group_count(thread_group_z, local_z);
 	}
 
 	const uint64_t invocations =
 	    static_cast<uint64_t>(groups_x) * groups_y * groups_z * local_x * local_y * local_z;
 	if (invocations != 0) {
-		BufferFlushAndWait();
-	}
+	    BufferFlushAndWait();
+	}*/
 }
 
 void CommandProcessor::DispatchIndirect(uint32_t data_offset, uint32_t mode) {
@@ -1233,20 +1239,25 @@ void CommandProcessor::DispatchIndirect(uint32_t data_offset, uint32_t mode) {
 }
 
 void CommandProcessor::DrawIndexAuto(uint32_t index_count, uint32_t flags,
-                                     uint32_t render_target_slice_offset, uint32_t instance_count,
-                                     uint32_t first_vertex, uint32_t first_instance) {
+                                     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) {
 	CheckBuffer();
 
-	m_renderer.GetRenderExecutor().DrawAuto(
-	    m_submit_id, CurrentBuffer(), index_count, flags, render_target_slice_offset,
-	    instance_count, first_vertex, first_instance);
+	m_renderer.GetRenderExecutor().DrawAuto(m_submit_id, CurrentBuffer(), vertex_count, flags,
+	                                        render_target_slice_offset, m_num_instances,
+	                                        first_vertex, first_instance);
 }
 
 void CommandProcessor::WaitFlipDone(uint32_t video_out_handle, uint32_t display_buffer_index) {
 	BufferFlush();
 
 	m_renderer.GetVideoOut().WaitFlipDone(static_cast<int>(video_out_handle),
-	                                              static_cast<int>(display_buffer_index));
+	                                      static_cast<int>(display_buffer_index));
 }
 
 template <typename T>
@@ -1317,8 +1328,8 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 				if (eop_event_type == 0x2f && cache_action == 0x00 && event_index == 0x06) {
 					auto* dst = static_cast<uint32_t*>(dst_gpu_addr);
 					SynchronizeGpu();
-					Sync::ReadGds(m_renderer.GetBufferCache().GetGdsBuffer(), dst,
-					              value & 0xffffu, value >> 16u);
+					Sync::ReadGds(m_renderer.GetBufferCache().GetGdsBuffer(), dst, value & 0xffffu,
+					              value >> 16u);
 					Sync::WriteAtEndOfPipeGds32(m_submit_id, CurrentBuffer(), dst, value & 0xffffu,
 					                            value >> 16u);
 					return;
@@ -1486,8 +1497,7 @@ void CommandProcessor::EmitGlobalBarrier() {
 	barrier.srcStageMask  = vk::PipelineStageFlagBits2::eAllCommands;
 	barrier.srcAccessMask = vk::AccessFlagBits2::eMemoryWrite;
 	barrier.dstStageMask  = vk::PipelineStageFlagBits2::eAllCommands;
-	barrier.dstAccessMask =
-	    vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite;
+	barrier.dstAccessMask = vk::AccessFlagBits2::eMemoryRead | vk::AccessFlagBits2::eMemoryWrite;
 
 	vk::DependencyInfo dependency {};
 	dependency.memoryBarrierCount = 1;
@@ -1661,10 +1671,6 @@ void Gpu::SendCommandSync(Common::UniqueFunction<void>&& command) {
 	m_state->SendCommandSync(std::move(command));
 }
 
-void Gpu::SendCommandSyncWithProcessor(Common::UniqueFunction<void, CommandProcessor&>&& command) {
-	m_state->SendCommandSyncWithProcessor(std::move(command));
-}
-
 void Gpu::Submit(uint32_t* draw_commands, uint32_t draw_size_dw, uint32_t* constant_commands,
                  uint32_t constant_size_dw, bool trigger_agc_interrupt_on_done) {
 	EXIT_IF(draw_commands == nullptr || draw_size_dw == 0);
@@ -1682,54 +1688,16 @@ void Gpu::SubmitFlipPreparation(uint64_t request_id) {
 	m_state->SubmitFlipPreparation(request_id);
 }
 
-void Gpu::Done() {
-	m_state->Done();
+void Gpu::SuspendPoint() {
+	m_state->SuspendPoint();
 }
 
 int Gpu::GetFrameNum() const {
 	return m_state->GetFrameNum();
 }
 
-void Gpu::PauseSubmissions() {
-	m_state->PauseSubmissions();
-}
-
-void Gpu::ResumeSubmissions() {
-	m_state->ResumeSubmissions();
-}
-
-Gpu::SubmissionLock::SubmissionLock(Gpu& gpu): m_gpu(gpu) {
-	if (g_current_processor != nullptr || g_submission_pause_depth == UINT32_MAX) {
-		EXIT("cannot acquire GPU submission lock in the current state\n");
-	}
-	if (g_submission_pause_depth++ == 0) {
-		m_gpu.PauseSubmissions();
-	}
-}
-
-Gpu::SubmissionLock::~SubmissionLock() {
-	if (g_submission_pause_depth == 0) {
-		EXIT("GPU submission lock released without ownership\n");
-	}
-	if (--g_submission_pause_depth == 0) {
-		m_gpu.ResumeSubmissions();
-	}
-}
-
-bool Gpu::IsCommandProcessorThread() noexcept {
-	return g_current_processor != nullptr;
-}
-
-CommandProcessor* Gpu::CurrentCommandProcessor() noexcept {
-	return g_current_processor;
-}
-
-bool Gpu::SubmissionLockHeld() noexcept {
-	return g_submission_pause_depth != 0;
-}
-
-bool Gpu::MutexHeld() noexcept {
-	return g_gpu_mutex_owned;
+bool Gpu::IsGpuThread() noexcept {
+	return GpuState::IsGpuThread();
 }
 
 } // namespace Libs::Graphics

src/graphics/guest_gpu/graphicsRun.h

@@ -9,7 +9,6 @@
 
 namespace Libs::Graphics {
 
-class CommandProcessor;
 class GpuState;
 class RenderContext;
 
@@ -23,37 +22,18 @@ public:
 	[[nodiscard]] bool IsStopping();
 	void               SendCommand(Common::UniqueFunction<void>&& command);
 	void               SendCommandSync(Common::UniqueFunction<void>&& command);
-	void SendCommandSyncWithProcessor(Common::UniqueFunction<void, CommandProcessor&>&& command);
 
 	void Submit(uint32_t* draw_commands, uint32_t draw_size_dw, uint32_t* constant_commands,
 	            uint32_t constant_size_dw, bool trigger_agc_interrupt_on_done = false);
 	void SubmitCompute(uint32_t queue, uint32_t* commands, uint32_t size_dw,
 	                   bool trigger_agc_interrupt_on_done = false);
 	void SubmitFlipPreparation(uint64_t request_id);
-	void Done();
+	void SuspendPoint();
 	[[nodiscard]] int GetFrameNum() const;
 
-	[[nodiscard]] static bool              IsCommandProcessorThread() noexcept;
-	[[nodiscard]] static CommandProcessor* CurrentCommandProcessor() noexcept;
-	[[nodiscard]] static bool              SubmissionLockHeld() noexcept;
-	[[nodiscard]] static bool              MutexHeld() noexcept;
-
-	class SubmissionLock final {
-	public:
-		explicit SubmissionLock(Gpu& gpu);
-		~SubmissionLock();
-		KYTY_CLASS_NO_COPY(SubmissionLock);
-
-	private:
-		Gpu& m_gpu;
-	};
+	[[nodiscard]] static bool IsGpuThread() noexcept;
 
 private:
-	friend class SubmissionLock;
-
-	void PauseSubmissions();
-	void ResumeSubmissions();
-
 	std::unique_ptr<GpuState> m_state;
 };
 } // namespace Libs::Graphics

src/graphics/guest_gpu/hardwareContext.h

@@ -348,6 +348,16 @@ struct ModeControl {
 	bool    persp_corr_dis           = false;
 };
 
+struct PolyOffset {
+	int8_t neg_num_db_bits = -23;
+	bool   db_is_float_fmt = true;
+	float  clamp           = 0.0f;
+	float  front_scale     = 0.0f;
+	float  front_offset    = 0.0f;
+	float  back_scale      = 0.0f;
+	float  back_offset     = 0.0f;
+};
+
 struct BlendControl {
 	uint8_t color_srcblend       = 1;
 	uint8_t color_comb_fcn       = 0;
@@ -482,6 +492,26 @@ struct VsShaderResource2 {
 	uint8_t shared_vgprs      = 0;
 };
 
+struct HsShaderResource1 {
+	uint8_t vgprs                     = 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    threadgroup_configuration = false;
+	uint8_t ls_vgpr_component_count   = 0;
+	bool    fp16_overflow             = false;
+};
+
+struct HsShaderResource2 {
+	bool     scratch_en   = false;
+	uint8_t  user_sgpr    = 0;
+	uint16_t lds_size     = 0;
+	uint8_t  shared_vgprs = 0;
+};
+
 struct PsShaderResource1 {
 	uint8_t vgprs                    = 0;
 	uint8_t sgprs                    = 0;
@@ -500,7 +530,7 @@ struct PsShaderResource2 {
 	uint8_t user_sgpr              = 0;
 	bool    wave_cnt_en            = false;
 	uint8_t extra_lds_size         = 0;
-	bool    raster_ordered_shading = false;
+	uint8_t raster_ordered_shading = 0;
 	uint8_t shared_vgprs           = 0;
 };
 
@@ -513,22 +543,27 @@ struct PsStageRegisters {
 
 struct CsStageRegisters {
 
-	uint64_t data_addr      = 0;
-	uint32_t num_thread_x   = 0;
-	uint32_t num_thread_y   = 0;
-	uint32_t num_thread_z   = 0;
-	uint8_t  vgprs          = 0;
-	uint8_t  sgprs          = 0;
-	uint8_t  bulky          = 0;
-	uint8_t  wave_size      = 64;
-	uint8_t  scratch_en     = 0;
-	uint8_t  user_sgpr      = 0;
-	uint8_t  tgid_x_en      = 0;
-	uint8_t  tgid_y_en      = 0;
-	uint8_t  tgid_z_en      = 0;
-	uint8_t  tg_size_en     = 0;
-	uint8_t  tidig_comp_cnt = 0;
-	uint8_t  lds_size       = 0;
+	uint64_t data_addr                 = 0;
+	uint32_t num_thread_x              = 0;
+	uint32_t num_thread_y              = 0;
+	uint32_t num_thread_z              = 0;
+	uint8_t  vgprs                     = 0;
+	uint8_t  sgprs                     = 0;
+	uint8_t  float_mode                = 0;
+	bool     dx10_clamp                = false;
+	bool     ieee_mode                 = false;
+	bool     fp16_overflow             = false;
+	bool     bulky                     = false;
+	bool     threadgroup_configuration = false;
+	uint8_t  wave_size                 = 64;
+	bool     scratch_en                = false;
+	uint8_t  user_sgpr                 = 0;
+	bool     tgid_x_en                 = false;
+	bool     tgid_y_en                 = false;
+	bool     tgid_z_en                 = false;
+	bool     tg_size_en                = false;
+	uint8_t  tidig_comp_cnt            = 0;
+	uint16_t lds_size                  = 0;
 };
 
 struct EsStageRegisters {
@@ -543,24 +578,24 @@ struct LsStageRegisters {
 
 struct HsStageRegisters {
 	uint64_t          data_addr = 0;
-	VsShaderResource1 rsrc1;
-	VsShaderResource2 rsrc2;
+	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;
-	bool    debug_mode               = false;
-	bool    ieee_mode                = false;
-	bool    cu_group_enable          = false;
-	bool    require_forward_progress = false;
-	bool    lds_configuration        = false;
-	uint8_t gs_vgpr_component_count  = 0;
-	bool    fp16_overflow            = false;
+	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    cu_group_enable           = false;
+	bool    require_forward_progress  = false;
+	bool    threadgroup_configuration = false;
+	uint8_t gs_vgpr_component_count   = 0;
+	bool    fp16_overflow             = false;
 };
 
 struct GsShaderResource2 {
@@ -880,6 +915,8 @@ public:
 	void SetDepthControl(const DepthControl& control) { m_depth_control = control; }
 	[[nodiscard]] const ModeControl& GetModeControl() const { return m_mode_control; }
 	void SetModeControl(const ModeControl& control) { m_mode_control = control; }
+	[[nodiscard]] const PolyOffset& GetPolyOffset() const { return m_poly_offset; }
+	void SetPolyOffset(const PolyOffset& offset) { m_poly_offset = offset; }
 	[[nodiscard]] const EqaaControl& GetEqaaControl() const { return m_eqaa_control; }
 	void SetEqaaControl(const EqaaControl& control) { m_eqaa_control = control; }
 	[[nodiscard]] const StencilControl& GetStencilControl() const { return m_stencil_control; }
@@ -977,6 +1014,7 @@ private:
 	uint8_t           m_stencil_clear_value = 0;
 
 	ModeControl m_mode_control;
+	PolyOffset  m_poly_offset;
 	EqaaControl m_eqaa_control;
 
 	ShaderRegisters m_sh_regs;
@@ -1034,8 +1072,8 @@ public:
 	void SetLsShaderResource1(const VsShaderResource1& rsrc1) { m_vs.ls_regs.rsrc1 = rsrc1; }
 	void SetLsShaderResource2(const VsShaderResource2& rsrc2) { m_vs.ls_regs.rsrc2 = rsrc2; }
 	void SetHsShaderBase(uint64_t addr) { m_vs.hs_regs.data_addr = addr; }
-	void SetHsShaderResource1(const VsShaderResource1& rsrc1) { m_vs.hs_regs.rsrc1 = rsrc1; }
-	void SetHsShaderResource2(const VsShaderResource2& rsrc2) { m_vs.hs_regs.rsrc2 = rsrc2; }
+	void SetHsShaderResource1(const HsShaderResource1& rsrc1) { m_vs.hs_regs.rsrc1 = rsrc1; }
+	void SetHsShaderResource2(const HsShaderResource2& rsrc2) { m_vs.hs_regs.rsrc2 = rsrc2; }
 	void SetHsShaderChksum(uint32_t value) { m_vs.hs_regs.chksum = value; }
 	void SetGsShaderBase(uint64_t addr) { m_vs.gs_regs.data_addr = addr; }
 	void SetGsShaderResource1(const GsShaderResource1& rsrc1) { m_vs.gs_regs.rsrc1 = rsrc1; }

src/graphics/guest_gpu/pm4.cpp

@@ -30,6 +30,7 @@ constexpr auto MakeRegisterNames() {
 	names[R_FLIP]           = "R_FLIP";
 	names[R_RELEASE_MEM]    = "R_RELEASE_MEM";
 	names[R_DMA_DATA]       = "R_DMA_DATA";
+	names[R_CONTEXT_STATE]  = "R_CONTEXT_STATE";
 	return names;
 }
 
@@ -110,7 +111,6 @@ void DumpPm4PacketStream(Common::File* file, uint32_t* cmd_buffer, uint32_t star
 	auto* cmd = cmd_buffer + start_dw;
 	auto  dw  = num_dw;
 	while (dw != 0) {
-		EXIT_NOT_IMPLEMENTED(dw < 2);
 		EXIT_NOT_IMPLEMENTED(dw > num_dw);
 
 		auto cmd_id = *cmd++;
@@ -120,6 +120,9 @@ void DumpPm4PacketStream(Common::File* file, uint32_t* cmd_buffer, uint32_t star
 		uint32_t len = 0;
 
 		const auto packet_type = static_cast<PacketType>(cmd_id >> 30u);
+		// Type-2 packets are header-only padding; every other packet type requires a body.
+		EXIT_NOT_IMPLEMENTED(dw < 2 && packet_type != PacketType::Type2);
+
 		switch (packet_type) {
 			case PacketType::Type3: {
 				const bool    sh_gx = (cmd_id & 0x2u) == 0;

src/graphics/guest_gpu/pm4.h

@@ -87,6 +87,7 @@ constexpr uint32_t R_WAIT_MEM_64    = 0x16;
 constexpr uint32_t R_FLIP           = 0x17;
 constexpr uint32_t R_RELEASE_MEM    = 0x18;
 constexpr uint32_t R_DMA_DATA       = 0x19;
+constexpr uint32_t R_CONTEXT_STATE  = 0x1A;
 
 constexpr uint32_t R_NUM = 0x3F + 1;
 
@@ -577,24 +578,28 @@ 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_SRESULTS_COMPARE_STATE0     = 0x2B0;
-constexpr uint32_t DB_SRESULTS_COMPARE_STATE1     = 0x2B1;
-constexpr uint32_t VGT_GS_MAX_VERT_OUT            = 0x2CE;
-constexpr uint32_t GE_NGG_SUBGRP_CNTL             = 0x2D3;
-constexpr uint32_t VGT_TESS_DISTRIBUTION          = 0x2D4;
-constexpr uint32_t VGT_SHADER_STAGES_EN           = 0x2D5;
-constexpr uint32_t VGT_LS_HS_CONFIG               = 0x2D6;
-constexpr uint32_t VGT_TF_PARAM                   = 0x2DB;
-constexpr uint32_t DB_ALPHA_TO_MASK               = 0x2DC;
-constexpr uint32_t PA_SU_POLY_OFFSET_DB_FMT_CNTL  = 0x2DE;
-constexpr uint32_t PA_SU_POLY_OFFSET_CLAMP        = 0x2DF;
-constexpr uint32_t PA_SU_POLY_OFFSET_FRONT_SCALE  = 0x2E0;
-constexpr uint32_t PA_SU_POLY_OFFSET_FRONT_OFFSET = 0x2E1;
-constexpr uint32_t PA_SU_POLY_OFFSET_BACK_SCALE   = 0x2E2;
-constexpr uint32_t PA_SU_POLY_OFFSET_BACK_OFFSET  = 0x2E3;
-constexpr uint32_t VGT_GS_INSTANCE_CNT            = 0x2E4;
-constexpr uint32_t PA_SC_CENTROID_PRIORITY_0      = 0x2F5;
-constexpr uint32_t PA_SC_CENTROID_PRIORITY_1      = 0x2F6;
+constexpr uint32_t DB_SRESULTS_COMPARE_STATE0                          = 0x2B0;
+constexpr uint32_t DB_SRESULTS_COMPARE_STATE1                          = 0x2B1;
+constexpr uint32_t VGT_GS_MAX_VERT_OUT                                 = 0x2CE;
+constexpr uint32_t GE_NGG_SUBGRP_CNTL                                  = 0x2D3;
+constexpr uint32_t VGT_TESS_DISTRIBUTION                               = 0x2D4;
+constexpr uint32_t VGT_SHADER_STAGES_EN                                = 0x2D5;
+constexpr uint32_t VGT_LS_HS_CONFIG                                    = 0x2D6;
+constexpr uint32_t VGT_TF_PARAM                                        = 0x2DB;
+constexpr uint32_t DB_ALPHA_TO_MASK                                    = 0x2DC;
+constexpr uint32_t PA_SU_POLY_OFFSET_DB_FMT_CNTL                       = 0x2DE;
+constexpr uint32_t PA_SU_POLY_OFFSET_DB_FMT_CNTL_NEG_NUM_DB_BITS_SHIFT = 0;
+constexpr uint32_t PA_SU_POLY_OFFSET_DB_FMT_CNTL_NEG_NUM_DB_BITS_MASK  = 0xFF;
+constexpr uint32_t PA_SU_POLY_OFFSET_DB_FMT_CNTL_DB_IS_FLOAT_FMT_SHIFT = 8;
+constexpr uint32_t PA_SU_POLY_OFFSET_DB_FMT_CNTL_DB_IS_FLOAT_FMT_MASK  = 0x1;
+constexpr uint32_t PA_SU_POLY_OFFSET_CLAMP                             = 0x2DF;
+constexpr uint32_t PA_SU_POLY_OFFSET_FRONT_SCALE                       = 0x2E0;
+constexpr uint32_t PA_SU_POLY_OFFSET_FRONT_OFFSET                      = 0x2E1;
+constexpr uint32_t PA_SU_POLY_OFFSET_BACK_SCALE                        = 0x2E2;
+constexpr uint32_t PA_SU_POLY_OFFSET_BACK_OFFSET                       = 0x2E3;
+constexpr uint32_t VGT_GS_INSTANCE_CNT                                 = 0x2E4;
+constexpr uint32_t PA_SC_CENTROID_PRIORITY_0                           = 0x2F5;
+constexpr uint32_t PA_SC_CENTROID_PRIORITY_1                           = 0x2F6;
 
 constexpr uint32_t PA_SC_AA_CONFIG                             = 0x2F8;
 constexpr uint32_t PA_SC_AA_CONFIG_MSAA_NUM_SAMPLES_SHIFT      = 0;
@@ -778,21 +783,21 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_FWD_PROGRESS_MASK      = 0x1;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_FP16_OVFL_SHIFT        = 29;
 constexpr uint32_t SPI_SHADER_PGM_RSRC1_PS_FP16_OVFL_MASK         = 0x1;
 
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS                                = 0xB;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_SCRATCH_EN_SHIFT               = 0;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_SCRATCH_EN_MASK                = 0x1;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_USER_SGPR_SHIFT                = 1;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_USER_SGPR_MASK                 = 0x1F;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_WAVE_CNT_EN_SHIFT              = 7;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_WAVE_CNT_EN_MASK               = 0x1;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_EXTRA_LDS_SIZE_SHIFT           = 8;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_EXTRA_LDS_SIZE_MASK            = 0xFF;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_LOAD_INTRAWAVE_COLLISION_SHIFT = 26;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_LOAD_INTRAWAVE_COLLISION_MASK  = 0x1;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_USER_SGPR_MSB_SHIFT            = 27;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_USER_SGPR_MSB_MASK             = 0x1;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_SHARED_VGPR_CNT_SHIFT          = 28;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_SHARED_VGPR_CNT_MASK           = 0xF;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS                              = 0xB;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_SCRATCH_EN_SHIFT             = 0;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_SCRATCH_EN_MASK              = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_USER_SGPR_SHIFT              = 1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_USER_SGPR_MASK               = 0x1F;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_WAVE_CNT_EN_SHIFT            = 7;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_WAVE_CNT_EN_MASK             = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_EXTRA_LDS_SIZE_SHIFT         = 8;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_EXTRA_LDS_SIZE_MASK          = 0xFF;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_RASTER_ORDERED_SHADING_SHIFT = 25;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_RASTER_ORDERED_SHADING_MASK  = 0x3;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_USER_SGPR_MSB_SHIFT          = 27;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_USER_SGPR_MSB_MASK           = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_PS_SHARED_VGPR_CNT_SHIFT        = 28;
+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;
@@ -885,61 +890,101 @@ constexpr uint32_t SPI_SHADER_PGM_RSRC2_GS_USER_SGPR_MSB_MASK     = 0x1;
 constexpr uint32_t SPI_SHADER_PGM_RSRC2_GS_SHARED_VGPR_CNT_SHIFT  = 28;
 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;
-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_SHADER_PGM_LO_HS            = 0x108;
-constexpr uint32_t SPI_SHADER_PGM_HI_HS            = 0x109;
-constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS         = 0x10A;
-constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS         = 0x10B;
-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;
-constexpr uint32_t COMPUTE_NUM_THREAD_X            = 0x207;
-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 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;
+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_SHADER_PGM_LO_HS                           = 0x108;
+constexpr uint32_t SPI_SHADER_PGM_HI_HS                           = 0x109;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS                        = 0x10A;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS                        = 0x10B;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_VGPRS_SHIFT            = 0;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_VGPRS_MASK             = 0x3F;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_PRIORITY_SHIFT         = 10;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_PRIORITY_MASK          = 0x3;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_FLOAT_MODE_SHIFT       = 12;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_FLOAT_MODE_MASK        = 0xFF;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_DX10_CLAMP_SHIFT       = 21;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_DX10_CLAMP_MASK        = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_DEBUG_MODE_SHIFT       = 22;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_DEBUG_MODE_MASK        = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_IEEE_MODE_SHIFT        = 23;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_IEEE_MODE_MASK         = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_FWD_PROGRESS_SHIFT     = 25;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_FWD_PROGRESS_MASK      = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_WGP_MODE_SHIFT         = 26;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_WGP_MODE_MASK          = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_LS_VGPR_COMP_CNT_SHIFT = 28;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_LS_VGPR_COMP_CNT_MASK  = 0x3;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_FP16_OVFL_SHIFT        = 30;
+constexpr uint32_t SPI_SHADER_PGM_RSRC1_HS_FP16_OVFL_MASK         = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_SCRATCH_EN_SHIFT       = 0;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_SCRATCH_EN_MASK        = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_USER_SGPR_SHIFT        = 1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_USER_SGPR_MASK         = 0x1F;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_LDS_SIZE_SHIFT         = 18;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_LDS_SIZE_MASK          = 0x1FF;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_USER_SGPR_MSB_SHIFT    = 27;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_USER_SGPR_MSB_MASK     = 0x1;
+constexpr uint32_t SPI_SHADER_PGM_RSRC2_HS_SHARED_VGPR_CNT_SHIFT  = 28;
+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;
+constexpr uint32_t COMPUTE_NUM_THREAD_X                           = 0x207;
+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_BULKY_SHIFT  = 24;
-constexpr uint32_t COMPUTE_PGM_RSRC1_BULKY_MASK   = 0x1;
-constexpr uint32_t COMPUTE_PGM_RSRC1_W32_EN_SHIFT = 30;
-constexpr uint32_t COMPUTE_PGM_RSRC1_W32_EN_MASK  = 0x1;
+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_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_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_W32_EN_SHIFT     = 30;
+constexpr uint32_t COMPUTE_PGM_RSRC1_W32_EN_MASK      = 0x1;
 
 constexpr uint32_t COMPUTE_PGM_RSRC2                      = 0x213;
 constexpr uint32_t COMPUTE_PGM_RSRC2_SCRATCH_EN_SHIFT     = 0;

src/graphics/guest_gpu/tile.h

@@ -39,7 +39,7 @@ enum class TileBlockFamily : uint32_t {
 };
 
 struct TileBlockLayout {
-	TileBlockFamily family            = TileBlockFamily::Standard256B;
+	TileBlockFamily family            = TileBlockFamily::Count;
 	uint32_t        bytes_per_element = 0;
 	uint32_t        block_size        = 0;
 	uint32_t        block_width       = 0;
@@ -47,60 +47,91 @@ struct TileBlockLayout {
 	uint32_t        block_depth       = 0;
 };
 
-struct TileVolumeLayout {
-	TileBlockFamily family            = TileBlockFamily::Count;
-	uint32_t        bytes_per_element = 0;
-	uint32_t        texel_width       = 1;
-	uint32_t        texel_height      = 1;
-	uint32_t        first_tail_level  = 0;
-	uint32_t        block_depth       = 1;
-	uint64_t        block_slice_size  = 0;
-	uint64_t        total_size        = 0;
-	uint64_t        level_offsets[16] = {};
-	uint64_t        level_sizes[16]   = {};
-	uint32_t        tail_x[16]        = {};
-	uint32_t        tail_y[16]        = {};
-	uint32_t        level_widths[16]  = {};
-	uint32_t        level_heights[16] = {};
+struct TileTextureElementLayout {
+	uint32_t bytes        = 0;
+	uint32_t texel_width  = 1;
+	uint32_t texel_height = 1;
+};
+
+struct TileTextureBlockLayout {
+	TileBlockLayout block;
+	uint32_t        texel_width  = 1;
+	uint32_t        texel_height = 1;
+};
+
+enum class TileSurfaceDimension : uint32_t {
+	Dim2D,
+	Dim3D,
+};
+
+struct TileSurfaceDescription {
+	uint32_t             format    = 0;
+	uint32_t             tile_mode = 0;
+	TileSurfaceDimension dimension = TileSurfaceDimension::Dim2D;
+	uint32_t             width     = 0;
+	uint32_t             height    = 0;
+	uint32_t             depth     = 1;
+	uint32_t             levels    = 1;
+	uint32_t             layers    = 1;
+};
+
+// Width/height fields are measured in addressable elements, not texture texels.
+struct TileMipLayout {
+	uint64_t offset        = 0;
+	uint64_t size          = 0;
+	uint32_t width         = 0;
+	uint32_t height        = 0;
+	uint32_t padded_width  = 0;
+	uint32_t padded_height = 0;
+	uint32_t tail_x        = 0;
+	uint32_t tail_y        = 0;
+};
+
+struct TileSurfaceLayout {
+	TileSurfaceDescription description;
+	TileTextureBlockLayout texture;
+	uint32_t               first_tail_level = 0;
+	uint64_t               block_slice_size = 0;
+	uint64_t               total_size       = 0;
+	TileMipLayout          mips[16]         = {};
 };
 
 bool TileGetBlockLayout(TileBlockFamily family, uint32_t bytes_per_element,
-	                    TileBlockLayout& layout);
+                        TileBlockLayout& layout);
+bool TileGetTextureElementLayout(uint32_t format, TileTextureElementLayout& layout);
+bool TileGetTextureBlockLayout(uint32_t format, uint32_t tile, bool volume,
+                               TileTextureBlockLayout& layout);
+bool TileGetTiledTextureLayout(const TileSurfaceDescription& description,
+                               TileSurfaceLayout&            layout);
 bool TileGetBlockOffset(const TileBlockLayout& layout, uint32_t x, uint32_t y, uint32_t z,
-	                    uint32_t& byte_offset);
+                        uint32_t& byte_offset);
 bool TileGetBlockXor(const TileBlockLayout& layout, uint32_t block_x, uint32_t block_y,
-	                 uint32_t& byte_offset);
+                     uint32_t& byte_offset);
 bool TileGetBlockXor(const TileBlockLayout& layout, uint32_t block_x, uint32_t block_y,
-	                 uint32_t block_z, uint32_t& byte_offset);
-bool TileIsStandard256BTextureSupported(uint32_t format);
-bool TileIsStandard4KBTextureSupported(uint32_t format);
-bool TileIsStandard64KBTextureSupported(uint32_t format);
-bool TileGetTextureVolumeLayout(uint32_t format, uint32_t width, uint32_t height, uint32_t depth,
-	                            uint32_t levels, uint32_t tile, TileVolumeLayout& layout);
+                     uint32_t block_z, uint32_t& byte_offset);
 
-bool     TileGetHtileSize(uint32_t width, uint32_t height, TileSizeAlign& htile_size);
 bool     TileGetDepthSize(uint32_t width, uint32_t height, uint32_t pitch, uint32_t z_format,
-	                      uint32_t stencil_format, bool htile, TileSizeAlign& stencil_size,
-	                      TileSizeAlign& htile_size, TileSizeAlign& depth_size,
+                          uint32_t stencil_format, bool htile, TileSizeAlign& stencil_size,
+                          TileSizeAlign& htile_size, TileSizeAlign& depth_size,
                           uint32_t num_fragments_log2 = 0);
 uint32_t TileGetRenderTargetPitch(uint32_t width, uint32_t bytes_per_element,
                                   uint32_t num_fragments_log2 = 0);
 uint32_t TileGetDepthPitch(uint32_t width, uint32_t bytes_per_element,
                            uint32_t num_fragments_log2 = 0);
 bool     TileGetRenderTargetSize(uint32_t width, uint32_t height, uint32_t pitch,
-	                             uint32_t bytes_per_element, TileSizeAlign& total_size,
+                                 uint32_t bytes_per_element, 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,
+                                      uint32_t bytes_per_element, uint32_t levels,
+                                      TileSizeAlign& total_size, TileSizeOffset* level_sizes,
                                       TilePaddedSize* padded_size);
-void     TileGetTextureSize(uint32_t format, uint32_t width, uint32_t height, uint32_t pitch,
-                            uint32_t levels, uint32_t tile, TileSizeAlign* total_size,
-                            TileSizeOffset* level_sizes, TilePaddedSize* padded_size);
-void TileGetTextureTotalSize(uint32_t format, uint32_t width, uint32_t height, uint32_t depth,
-                             uint32_t pitch, uint32_t levels, uint32_t tile, bool volume_texture,
-	                         TileSizeAlign& total_size);
-uint32_t TileGetTexturePitch(uint32_t format, uint32_t width, uint32_t levels, uint32_t tile);
+void     TileGetTextureSize(uint32_t format, uint32_t width, uint32_t height, uint32_t levels,
+                            uint32_t tile, TileSizeAlign* total_size, TileSizeOffset* level_sizes,
+                            TilePaddedSize* padded_size);
+void     TileGetTextureTotalSize(uint32_t format, uint32_t width, uint32_t height, uint32_t depth,
+                                 uint32_t levels, uint32_t tile, bool volume_texture,
+                                 TileSizeAlign& total_size);
+uint32_t TileGetTexturePitch(uint32_t format, uint32_t width, uint32_t tile);
 
 } // namespace Libs::Graphics
 

src/graphics/host_gpu/graphicContext.h

@@ -60,19 +60,19 @@ struct VulkanImage {
 	VulkanImage() = default;
 	KYTY_CLASS_NO_COPY(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;
-	vk::ImageUsageFlags    usage       = {};
-	vk::ImageCreateFlags   flags       = {};
-	vk::Image              image       = nullptr;
-	VulkanImageState       state;
+	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;
+	vk::ImageUsageFlags           usage       = {};
+	vk::ImageCreateFlags          flags       = {};
+	vk::Image                     image       = nullptr;
+	VulkanImageState              state;
 	std::vector<VulkanImageState> subresource_states;
-	Graphics::VulkanMemory memory;
+	Graphics::VulkanMemory        memory;
 };
 
 struct VulkanBuffer {

src/graphics/host_gpu/hostMemory.cpp

@@ -30,7 +30,7 @@ bool IsAccessible(DWORD protect, HostMemoryAccess access) {
 } // namespace
 
 bool HostMemoryQueryRange(uint64_t addr, uint64_t requested_size, HostMemoryAccess access,
-	                      uint64_t& accessible_size) {
+                          uint64_t& accessible_size) {
 	accessible_size = 0;
 	if (addr == 0 || requested_size == 0) {
 		return false;

src/graphics/host_gpu/hostMemory.h

@@ -8,7 +8,7 @@ namespace Libs::Graphics {
 enum class HostMemoryAccess { Read, Mapped };
 
 bool HostMemoryQueryRange(uint64_t addr, uint64_t requested_size, HostMemoryAccess access,
-	                      uint64_t& accessible_size);
+                          uint64_t& accessible_size);
 bool HostMemoryQueryReadable(uint64_t addr, uint64_t requested_size, uint64_t& readable_size);
 bool HostMemoryIsReadable(uint64_t addr);
 bool HostMemoryRangeIsReadable(uint64_t addr, uint64_t size);

src/graphics/host_gpu/memoryTracker.cpp

@@ -4,25 +4,9 @@
 
 namespace Libs::Graphics {
 
-#if defined(KYTY_MEMORY_TRACKER_TESTS)
-namespace {
-std::atomic<MemoryTracker::UnmapContentionHook> g_unmap_contention_hook {nullptr};
-}
-
-void MemoryTracker::SetUnmapContentionHook(UnmapContentionHook hook) noexcept {
-	g_unmap_contention_hook.store(hook, std::memory_order_release);
-}
-#endif
-
 static_assert(std::atomic<void*>::is_always_lock_free);
 
-MemoryTracker::MemoryTracker(PageManager& page_manager, PageWatchMode gpu_watch_mode)
-    : m_page_manager(page_manager), m_gpu_watch_mode(gpu_watch_mode) {
-	switch (m_gpu_watch_mode) {
-		case PageWatchMode::Write:
-		case PageWatchMode::ReadWrite: break;
-		default: EXIT("unsupported memory tracker GPU page-watch mode\n");
-	}
+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);
@@ -31,6 +15,7 @@ MemoryTracker::MemoryTracker(PageManager& page_manager, PageWatchMode gpu_watch_
 
 MemoryTracker::~MemoryTracker() = default;
 
+#if KYTY_BUILD == KYTY_BUILD_DEBUG
 void MemoryTracker::ValidateGpuDirtyPages(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
                                           const char* operation) const noexcept {
 	if (vaddr == 0 || size == 0 || size > UINT64_MAX - vaddr ||
@@ -68,6 +53,7 @@ void MemoryTracker::ValidateGpuDirtyOwnership(const RangeSet& dirty, uint64_t va
 		}
 	}
 }
+#endif
 
 void MemoryTracker::ValidateRange(uint64_t vaddr, uint64_t size) {
 	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
@@ -93,8 +79,6 @@ RegionManager* MemoryTracker::GetOrCreateRegion(uint64_t index) {
 
 bool MemoryTracker::IsRegionCpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
-	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
 	return Iterate<true>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
 		return manager->IsModified<DirtySource::Cpu>(offset, bytes);
@@ -103,8 +87,6 @@ bool MemoryTracker::IsRegionCpuModified(uint64_t vaddr, uint64_t size) {
 
 bool MemoryTracker::IsRegionGpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
-	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
 	return Iterate<false>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
 		return manager->IsModified<DirtySource::Gpu>(offset, bytes);
@@ -113,46 +95,29 @@ bool MemoryTracker::IsRegionGpuModified(uint64_t vaddr, uint64_t size) {
 
 void MemoryTracker::MarkRegionAsCpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
-	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
 	Iterate<true>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
-		const auto       changed =
-		    manager->ChangeState<DirtySource::Cpu, true>(manager->GetCpuAddr() + offset, bytes);
-		manager->ApplyProtection(changed, false);
+		manager->ChangeState<DirtySource::Cpu, true>(manager->GetCpuAddr() + offset, bytes);
 	});
 }
 
 void MemoryTracker::MarkRegionAsGpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
-	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
-	Iterate<true>(vaddr, size, [this](RegionManager* manager, uint64_t offset, uint64_t bytes) {
+	Iterate<true>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
-		const auto       changed =
-		    manager->ChangeState<DirtySource::Gpu, true>(manager->GetCpuAddr() + offset, bytes);
-		manager->ApplyGpuProtection(changed, true, m_gpu_watch_mode);
+		manager->ChangeState<DirtySource::Gpu, true>(manager->GetCpuAddr() + offset, bytes);
 	});
 }
 
 void MemoryTracker::UnmarkRegionAsGpuModified(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
-	std::lock_guard access(m_access_mutex);
-	RequireMapped(vaddr, size);
-	Iterate<true>(vaddr, size, [this](RegionManager* manager, uint64_t offset, uint64_t bytes) {
+	Iterate<false>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 		std::scoped_lock lock(manager->lock);
-		if (!manager->IsFullyModified<DirtySource::Gpu>(offset, bytes)) {
-			EXIT("cannot clear partially GPU-dirty tracking range\n");
-		}
-		const auto changed =
-		    manager->ChangeState<DirtySource::Gpu, false>(manager->GetCpuAddr() + offset, bytes);
-		manager->ApplyGpuProtection(changed, false, m_gpu_watch_mode);
+		manager->ChangeState<DirtySource::Gpu, false>(manager->GetCpuAddr() + offset, bytes);
 	});
 }
 
-void MemoryTracker::UntrackMemoryLocked(uint64_t vaddr, uint64_t size) {
-	RequireMapped(vaddr, size);
-
+void MemoryTracker::UntrackMemoryImpl(uint64_t vaddr, uint64_t size) {
 	std::vector<RegionManager*> managers;
 	managers.reserve((vaddr % TRACKER_REGION_SIZE + size + TRACKER_REGION_SIZE - 1) /
 	                 TRACKER_REGION_SIZE);
@@ -171,123 +136,14 @@ void MemoryTracker::UntrackMemoryLocked(uint64_t vaddr, uint64_t size) {
 		EXIT("cannot untrack GPU-dirty memory\n");
 	}
 	Iterate<false>(vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-		const auto changed =
-		    manager->ChangeState<DirtySource::Cpu, true>(manager->GetCpuAddr() + offset, bytes);
-		manager->ApplyProtection(changed, false);
-		manager->Untrack(manager->GetCpuAddr() + offset, bytes);
+		manager->ChangeState<DirtySource::Cpu, true>(manager->GetCpuAddr() + offset, bytes);
 	});
 	locks.clear();
 }
 
 void MemoryTracker::UntrackMemory(uint64_t vaddr, uint64_t size) {
 	CheckNotInUploadCallback();
-	std::lock_guard access(m_access_mutex);
-	UntrackMemoryLocked(vaddr, size);
-}
-
-void MemoryTracker::UnmapMemory(uint64_t vaddr, uint64_t size) {
-	CheckNotInUploadCallback();
-	std::unique_lock access(m_access_mutex, std::try_to_lock);
-	if (!access.owns_lock()) {
-#if defined(KYTY_MEMORY_TRACKER_TESTS)
-		if (const auto hook = g_unmap_contention_hook.load(std::memory_order_acquire);
-		    hook != nullptr) {
-			hook();
-		}
-#endif
-		access.lock();
-	}
-	UntrackMemoryLocked(vaddr, size);
-	m_page_manager.OnGpuUnmap(vaddr, size);
-}
-
-bool MemoryTracker::InvalidateRegion(uint64_t vaddr, uint64_t size, PageFaultPhase phase) noexcept {
-	switch (phase) {
-		case PageFaultPhase::Release: return true;
-		case PageFaultPhase::Invalidate: {
-			const auto action = BeginCpuFault(vaddr, size);
-			switch (action) {
-				case CpuFaultAction::Untracked: return false;
-				case CpuFaultAction::Continue: return true;
-				case CpuFaultAction::Download:
-					EXIT("generic region invalidation cannot download GPU-dirty memory\n");
-			}
-		}
-		case PageFaultPhase::Complete:
-			return CompleteCpuFault(vaddr, size, PageFaultAccess::Write, false);
-	}
-	EXIT("unsupported region invalidation phase\n");
-}
-
-bool MemoryTracker::InvalidateVirtualGpuWrite(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-                                              PageFaultPhase phase) noexcept {
-	switch (phase) {
-		case PageFaultPhase::Release: return true;
-		case PageFaultPhase::Invalidate: {
-			const bool gpu_modified = Iterate<false>(
-			    vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-				    std::scoped_lock lock(manager->lock);
-				    return manager->IsModified<DirtySource::Gpu>(offset, bytes);
-			    });
-			if (!gpu_modified) {
-				return false;
-			}
-			const auto action = BeginCpuFault(vaddr, size);
-			if (access != PageFaultAccess::Write || action != CpuFaultAction::Download) {
-				EXIT("virtual GPU write fault requires write access to GPU-dirty memory\n");
-			}
-			return true;
-		}
-		case PageFaultPhase::Complete: {
-			if (access != PageFaultAccess::Write) {
-				EXIT("virtual GPU write completion requires write access\n");
-			}
-			bool completed = false;
-			Iterate<false>(
-			    vaddr, size, [&completed](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-				    std::scoped_lock lock(manager->lock);
-				    if (completed) {
-					    EXIT("virtual GPU write fault spans multiple tracked regions\n");
-				    }
-				    completed =
-				        manager->CompleteVirtualGpuWrite(manager->GetCpuAddr() + offset, bytes);
-			    });
-			return completed;
-		}
-	}
-	EXIT("unsupported virtual GPU write invalidation phase\n");
-}
-
-CpuFaultAction MemoryTracker::BeginCpuFault(uint64_t vaddr, uint64_t size,
-                                            PageFaultAccess access) noexcept {
-	CheckNotInUploadCallback();
-	CpuFaultAction action = CpuFaultAction::Untracked;
-	Iterate<false>(
-	    vaddr, size, [&action, access](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-		    std::scoped_lock lock(manager->lock);
-		    if (action != CpuFaultAction::Untracked) {
-			    EXIT("CPU fault spans multiple tracked regions\n");
-		    }
-		    action = manager->BeginCpuFault(manager->GetCpuAddr() + offset, bytes, access);
-	    });
-	return action;
-}
-
-bool MemoryTracker::CompleteCpuFault(uint64_t vaddr, uint64_t size, PageFaultAccess access,
-                                     bool downloaded) noexcept {
-	CheckNotInUploadCallback();
-	bool found = false;
-	Iterate<false>(
-	    vaddr, size,
-	    [&found, access, downloaded](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-		    std::scoped_lock lock(manager->lock);
-		    if (found) {
-			    EXIT("CPU fault completion spans multiple tracked regions\n");
-		    }
-		    found = manager->CompleteCpuFault(manager->GetCpuAddr() + offset, bytes, access,
-		                                      downloaded);
-	    });
-	return found;
+	UntrackMemoryImpl(vaddr, size);
 }
 
 } // namespace Libs::Graphics

src/graphics/host_gpu/memoryTracker.h

@@ -18,8 +18,7 @@ namespace Libs::Graphics {
 
 class MemoryTracker final {
 public:
-	explicit MemoryTracker(PageManager&  page_manager,
-	                       PageWatchMode gpu_watch_mode = PageWatchMode::ReadWrite);
+	explicit MemoryTracker(PageManager& page_manager);
 	~MemoryTracker();
 
 	KYTY_CLASS_NO_COPY(MemoryTracker);
@@ -30,69 +29,45 @@ public:
 	void               MarkRegionAsGpuModified(uint64_t vaddr, uint64_t size);
 	void               UnmarkRegionAsGpuModified(uint64_t vaddr, uint64_t size);
 	void               UntrackMemory(uint64_t vaddr, uint64_t size);
-	void               UnmapMemory(uint64_t vaddr, uint64_t size);
-	[[nodiscard]] CpuFaultAction
-	                   BeginCpuFault(uint64_t vaddr, uint64_t size,
-	                                 PageFaultAccess access = PageFaultAccess::Write) noexcept;
-	[[nodiscard]] bool CompleteCpuFault(uint64_t vaddr, uint64_t size, PageFaultAccess access,
-	                                    bool downloaded) noexcept;
-	[[nodiscard]] bool InvalidateRegion(uint64_t vaddr, uint64_t size,
-	                                    PageFaultPhase phase) noexcept;
+	// Removes protection from a range and flushes GPU-owned data when required.
 	template <typename Flush>
-	void InvalidateRegion(uint64_t vaddr, uint64_t size, Flush&& on_flush) {
+	void InvalidateRegion(uint64_t vaddr, uint64_t size, Flush&& on_flush) noexcept {
 		static_assert(std::is_invocable_v<Flush&>);
 		CheckNotInUploadCallback();
 		ValidateRange(vaddr, size);
 
-		const auto update_cpu_state = [this, vaddr, size] {
-			std::lock_guard             access(m_access_mutex);
-			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 bool should_flush = [&] {
+				// Perform both the GPU modification check and CPU state change with the lock in
+				// case the GPU thread is racing to mark the page modified. If a flush is needed,
+				// on_flush performs the CPU state change.
+				std::scoped_lock lock(manager->lock);
+				if (manager->IsModified<DirtySource::Gpu>(offset, bytes)) {
+					return true;
+				}
+				manager->ChangeState<DirtySource::Cpu, true>(manager->GetCpuAddr() + offset, bytes);
+				return false;
+			}();
+			if (should_flush) {
+				on_flush();
 			}
-			const bool gpu_modified = Iterate<false>(
-			    vaddr, size, [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-				    return manager->IsModified<DirtySource::Gpu>(offset, bytes);
-			    });
-			if (gpu_modified) {
-				return true;
-			}
-			Iterate<false>(vaddr, size,
-			               [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-				               const auto changed = manager->ChangeState<DirtySource::Cpu, true>(
-				                   manager->GetCpuAddr() + offset, bytes);
-				               manager->ApplyProtection(changed, false);
-			               });
-			return false;
-		};
-
-		if (!update_cpu_state()) {
-			return;
-		}
-		std::forward<Flush>(on_flush)();
-		if (update_cpu_state()) {
-			EXIT("memory invalidation retained GPU-owned pages\n");
-		}
+		});
 	}
-	[[nodiscard]] bool InvalidateVirtualGpuWrite(PageFaultAccess access, uint64_t vaddr,
-	                                             uint64_t size, PageFaultPhase phase) noexcept;
-	void               ValidateGpuDirtyPages(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
-	                                         const char* operation) const noexcept;
+#if KYTY_BUILD == KYTY_BUILD_DEBUG
+	void ValidateGpuDirtyPages(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
+	                           const char* operation) const noexcept;
 	void ValidateGpuDirtyOwnership(const RangeSet& dirty, uint64_t vaddr, uint64_t size,
 	                               const char* operation);
+#else
+	void ValidateGpuDirtyPages(const RangeSet&, uint64_t, uint64_t, const char*) const noexcept {}
+	void ValidateGpuDirtyOwnership(const RangeSet&, uint64_t, uint64_t, const char*) {}
+#endif
 
 	template <bool clear, typename Preflight, typename Func>
 	void ForEachDownloadRange(uint64_t vaddr, uint64_t size, Preflight&& preflight, Func&& func) {
 		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::lock_guard access(m_access_mutex);
-		RequireMapped(vaddr, size);
 		std::vector<RegionManager*> managers;
 		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t, uint64_t) {
 			managers.push_back(manager);
@@ -104,9 +79,6 @@ public:
 		}
 		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 			const auto address = manager->GetCpuAddr() + offset;
-			if (manager->HasPendingFault(address, bytes)) {
-				EXIT("GPU download synchronization raced a pending CPU fault\n");
-			}
 			manager->template ForEachModifiedRange<DirtySource::Gpu, false>(address, bytes,
 			                                                                preflight);
 		});
@@ -118,10 +90,8 @@ public:
 			Iterate<false>(vaddr, size,
 			               [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 				               const auto address = manager->GetCpuAddr() + offset;
-				               const auto changed =
-				                   manager->template ForEachModifiedRange<DirtySource::Gpu, true>(
-				                       address, bytes, [](uint64_t, uint64_t) noexcept {});
-				               manager->ApplyGpuProtection(changed, false, m_gpu_watch_mode);
+				               manager->template ForEachModifiedRange<DirtySource::Gpu, true>(
+				                   address, bytes, [](uint64_t, uint64_t) noexcept {});
 			               });
 		}
 	}
@@ -132,24 +102,16 @@ public:
 		    vaddr, size, [](uint64_t, uint64_t) noexcept {}, std::forward<Func>(func));
 	}
 
-#if defined(KYTY_MEMORY_TRACKER_TESTS)
-	using UnmapContentionHook = void (*)() noexcept;
-	static void SetUnmapContentionHook(UnmapContentionHook hook) noexcept;
-#endif
-
 	template <typename RangeFunc, typename UploadFunc>
 	void ForEachUploadRange(uint64_t vaddr, uint64_t size, bool is_written, RangeFunc&& range_func,
 	                        UploadFunc&& upload_func) {
 		static_assert(std::is_nothrow_invocable_v<RangeFunc&, uint64_t, uint64_t>);
 		static_assert(std::is_nothrow_invocable_v<UploadFunc&>);
 		CheckNotInUploadCallback();
-		std::unique_lock access(m_access_mutex);
-		RequireMapped(vaddr, size);
 		Iterate<true>(vaddr, size, [](RegionManager*, uint64_t, uint64_t) {});
 		const auto* previous_upload_owner = std::exchange(s_upload_owner, this);
 		Iterate<false>(vaddr, size, [&](RegionManager* manager, uint64_t offset, uint64_t bytes) {
 			manager->lock.lock();
-			manager->Track(manager->GetCpuAddr() + offset, bytes);
 			manager->ForEachModifiedRange<DirtySource::Cpu, true>(manager->GetCpuAddr() + offset,
 			                                                      bytes, range_func);
 			if (!is_written) {
@@ -158,13 +120,12 @@ public:
 		});
 		upload_func();
 		if (is_written) {
-			Iterate<false>(
-			    vaddr, size, [this](RegionManager* manager, uint64_t offset, uint64_t bytes) {
-				    const auto changed = manager->template ChangeState<DirtySource::Gpu, true>(
-				        manager->GetCpuAddr() + offset, bytes);
-				    manager->ApplyGpuProtection(changed, true, m_gpu_watch_mode);
-				    manager->lock.unlock();
-			    });
+			Iterate<false>(vaddr, size,
+			               [](RegionManager* manager, uint64_t offset, uint64_t bytes) {
+				               manager->template ChangeState<DirtySource::Gpu, true>(
+				                   manager->GetCpuAddr() + offset, bytes);
+				               manager->lock.unlock();
+			               });
 		}
 		s_upload_owner = previous_upload_owner;
 	}
@@ -209,24 +170,14 @@ private:
 		return false;
 	}
 
-	static void ValidateRange(uint64_t vaddr, uint64_t size);
-	void        UntrackMemoryLocked(uint64_t vaddr, uint64_t size);
-	void        RequireMapped(uint64_t vaddr, uint64_t size) const {
-		ValidateRange(vaddr, size);
-		if (!m_page_manager.IsMapped(vaddr, size)) {
-			EXIT("memory tracker range [0x%llx, 0x%llx) is not mapped\n",
-			     static_cast<unsigned long long>(vaddr),
-			     static_cast<unsigned long long>(vaddr + size));
-		}
-	}
+	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;
 	std::vector<std::unique_ptr<RegionManager>>    m_region_storage;
 	std::mutex                                     m_region_mutex;
-	std::mutex                                     m_access_mutex;
 	PageManager&                                   m_page_manager;
-	PageWatchMode                                  m_gpu_watch_mode = PageWatchMode::ReadWrite;
 };
 
 } // namespace Libs::Graphics

src/graphics/host_gpu/pageManager.h

@@ -2,60 +2,32 @@
 #define EMULATOR_SRC_GRAPHICS_HOST_GPU_PAGEMANAGER_H_
 
 #include "common/common.h"
-#include "graphics/host_gpu/rangeSet.h"
+#include "graphics/host_gpu/regionDefinitions.h"
 
 #include <memory>
-#include <span>
-#include <vector>
 
 namespace Libs::Graphics {
 
 enum class PageFaultAccess { Read, Write, Execute, Unknown };
-enum class PageFaultPhase { Invalidate, Complete, Release };
-enum class PageWatchMode { Write, ReadWrite };
-enum class GpuAccess { Read, Write, ReadWrite };
-
-using PageFaultHandler = bool (*)(void* context, PageFaultAccess access, uint64_t vaddr,
-                                  uint64_t size, PageFaultPhase phase) noexcept;
 
 class PageManager final {
 public:
-	class BackingWrite final {
-	public:
-		BackingWrite(PageManager& manager, uint64_t vaddr, uint64_t size) noexcept;
-		~BackingWrite();
-		KYTY_CLASS_NO_COPY(BackingWrite);
-
-	private:
-		PageManager& m_manager;
-		uint64_t     m_vaddr = 0;
-		uint64_t     m_size  = 0;
-	};
-
-	PageManager(PageFaultHandler fault_handler, void* fault_context);
+	PageManager();
 	// The owner must stop all PageManager callers before destruction.
 	~PageManager();
 
 	KYTY_CLASS_NO_COPY(PageManager);
 
 	[[nodiscard]] uint64_t GetPageSize() const;
-	[[nodiscard]] bool     IsTracked(uint64_t vaddr) const noexcept;
-	[[nodiscard]] bool     IsMapped(uint64_t vaddr, uint64_t size) const noexcept;
-	[[nodiscard]] bool HasGpuAccess(uint64_t vaddr, uint64_t size, GpuAccess access) const noexcept;
 
-	void UpdatePageWatchers(bool track, uint64_t vaddr, uint64_t size,
-	                        PageWatchMode mode = PageWatchMode::Write);
-	void OnGpuMap(uint64_t vaddr, uint64_t size, GpuAccess access = GpuAccess::ReadWrite);
-	void OnGpuUnmap(uint64_t vaddr, uint64_t size, GpuAccess access = GpuAccess::ReadWrite);
-
-	[[nodiscard]] bool HandleFault(PageFaultAccess access, uint64_t fault_vaddr) noexcept;
-	[[nodiscard]] std::vector<std::unique_ptr<BackingWrite>>
-	ReserveBackingWrites(std::span<const RangeSet::Range> ranges);
+	template <bool track>
+	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:
-	void BeginBackingWrite(uint64_t vaddr, uint64_t size) noexcept;
-	void EndBackingWrite(uint64_t vaddr, uint64_t size) noexcept;
-
 	struct Impl;
 	std::unique_ptr<Impl> m_impl;
 };

src/graphics/host_gpu/regionDefinitions.h

@@ -1,10 +1,9 @@
 #ifndef EMULATOR_SRC_GRAPHICS_HOST_GPU_REGIONDEFINITIONS_H_
 #define EMULATOR_SRC_GRAPHICS_HOST_GPU_REGIONDEFINITIONS_H_
 
+#include "common/bitArray.h"
 #include "common/common.h"
 
-#include <bitset>
-
 namespace Libs::Graphics {
 
 constexpr uint64_t TRACKER_PAGE_SIZE    = 4ull * 1024ull;
@@ -13,7 +12,8 @@ constexpr uint64_t TRACKER_ADDRESS_SIZE = 1ull << 40u;
 constexpr size_t   TRACKER_REGION_PAGES = TRACKER_REGION_SIZE / TRACKER_PAGE_SIZE;
 
 enum class DirtySource { Cpu, Gpu };
-using RegionBits = std::bitset<TRACKER_REGION_PAGES>;
+using RegionBits = Common::BitArray<TRACKER_REGION_PAGES>;
+static_assert(sizeof(RegionBits) == TRACKER_REGION_PAGES / 8);
 
 } // namespace Libs::Graphics
 

src/graphics/host_gpu/regionManager.h

@@ -25,8 +25,6 @@
 
 namespace Libs::Graphics {
 
-enum class CpuFaultAction { Untracked, Continue, Download };
-
 class TrackingSpinLock final {
 public:
 	void lock() noexcept {
@@ -78,229 +76,93 @@ public:
 		if (m_cpu_addr % TRACKER_REGION_SIZE != 0) {
 			EXIT("invalid region tracking manager construction\n");
 		}
-		m_cpu_dirty.set();
-		m_writable.set();
+		m_cpu_dirty.Fill();
+		m_writable.Fill();
+		m_readable.Fill();
 	}
 
 	KYTY_CLASS_NO_COPY(RegionManager);
 
 	[[nodiscard]] uint64_t GetCpuAddr() const { return m_cpu_addr; }
-	void                   Track(uint64_t vaddr, uint64_t size) {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			m_tracked.set(page);
-		}
-	}
-	void Untrack(uint64_t vaddr, uint64_t size) {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			m_tracked.reset(page);
-		}
-	}
 	template <DirtySource source>
 	[[nodiscard]] bool IsModified(uint64_t offset, uint64_t size) const {
 		const auto [start, end] = GetPageRange(m_cpu_addr + offset, size);
 		const auto& bits        = GetBits<source>();
-		for (auto page = start; page < end; page++) {
-			if (bits.test(page)) {
-				return true;
-			}
-		}
-		return false;
-	}
-
-	template <DirtySource source>
-	[[nodiscard]] bool IsFullyModified(uint64_t offset, uint64_t size) const {
-		const auto [start, end] = GetPageRange(m_cpu_addr + offset, size);
-		const auto& bits        = GetBits<source>();
-		for (auto page = start; page < end; page++) {
-			if (!bits.test(page)) {
-				return false;
-			}
-		}
-		return true;
+		return RegionBits(bits, start, end).Any();
 	}
 
 	template <DirtySource source, bool enable>
-	RegionBits ChangeState(uint64_t vaddr, uint64_t size) {
+	void ChangeState(uint64_t vaddr, uint64_t size) {
 		const auto [start, end] = GetPageRange(vaddr, size);
 		if constexpr (source == DirtySource::Cpu && enable) {
-			for (auto page = start; page < end; page++) {
-				if (m_gpu_dirty.test(page) || m_fault_pending.test(page)) {
-					EXIT("CPU dirty state conflicts with GPU dirty or pending fault state\n");
-				}
+			if (RegionBits(m_gpu_dirty, start, end).Any()) {
+				EXIT("CPU dirty state conflicts with GPU dirty state\n");
 			}
 		}
 		if constexpr (source == DirtySource::Gpu && enable) {
-			for (auto page = start; page < end; page++) {
-				if (m_cpu_dirty.test(page) || m_fault_pending.test(page)) {
-					EXIT("GPU dirty state conflicts with CPU dirty or pending fault state\n");
-				}
+			if (RegionBits(m_cpu_dirty, start, end).Any()) {
+				EXIT("GPU dirty state conflicts with CPU dirty state\n");
 			}
 		}
-		auto& bits    = GetBits<source>();
-		auto  changed = bits;
-		for (auto page = start; page < end; page++) {
-			bits.set(page, enable);
+		auto& bits = GetBits<source>();
+		if constexpr (enable) {
+			bits.SetRange(start, end);
+		} else {
+			bits.UnsetRange(start, end);
 		}
-		changed ^= bits;
 		if constexpr (source == DirtySource::Cpu) {
-			changed    = m_cpu_dirty ^ m_writable;
-			m_writable = m_cpu_dirty;
+			UpdateCpuProtection<!enable>();
+		} else {
+			UpdateGpuProtection<enable>();
 		}
-		return changed;
-	}
-
-	[[nodiscard]] CpuFaultAction BeginCpuFault(uint64_t vaddr, uint64_t size,
-	                                           PageFaultAccess access = PageFaultAccess::Write) {
-		if (access != PageFaultAccess::Read && access != PageFaultAccess::Write) {
-			EXIT("unsupported CPU fault access while beginning ownership transfer\n");
-		}
-		const auto [start, end] = GetPageRange(vaddr, size);
-		const bool tracked      = m_tracked.test(start);
-		for (auto page = start; page < end; page++) {
-			if (m_tracked.test(page) != tracked) {
-				EXIT("CPU fault spans mixed tracked and untracked pages\n");
-			}
-			if (m_fault_pending.test(page)) {
-				return CpuFaultAction::Untracked;
-			}
-			if (m_cpu_dirty.test(page) != m_writable.test(page) ||
-			    (m_gpu_dirty.test(page) && (m_cpu_dirty.test(page) || m_writable.test(page)))) {
-				EXIT("inconsistent CPU fault page state\n");
-			}
-		}
-		if (!tracked) {
-			return CpuFaultAction::Untracked;
-		}
-		bool gpu_dirty = m_gpu_dirty.test(start);
-		bool writable  = m_writable.test(start);
-		for (auto page = start + 1; page < end; page++) {
-			if (m_gpu_dirty.test(page) != gpu_dirty || m_writable.test(page) != writable) {
-				EXIT("CPU fault spans pages with incompatible dirty or writable state\n");
-			}
-		}
-		for (auto page = start; page < end; page++) {
-			if (!gpu_dirty && access == PageFaultAccess::Write) {
-				m_cpu_dirty.set(page);
-				m_writable.set(page);
-			}
-			m_fault_pending.set(page);
-		}
-		return gpu_dirty ? CpuFaultAction::Download : CpuFaultAction::Continue;
-	}
-
-	[[nodiscard]] bool CompleteCpuFault(uint64_t vaddr, uint64_t size, PageFaultAccess access,
-	                                    bool downloaded) {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			if (!m_fault_pending.test(page)) {
-				return false;
-			}
-		}
-		for (auto page = start; page < end; page++) {
-			const bool gpu_dirty = m_gpu_dirty.test(page);
-			if (gpu_dirty != downloaded) {
-				EXIT("CPU fault download result disagrees with GPU dirty state\n");
-			}
-			if (gpu_dirty) {
-				m_gpu_dirty.reset(page);
-				switch (access) {
-					case PageFaultAccess::Read: break;
-					case PageFaultAccess::Write:
-						m_cpu_dirty.set(page);
-						m_writable.set(page);
-						break;
-					default: EXIT("unsupported CPU fault access after GPU download\n");
-				}
-			}
-			m_fault_pending.reset(page);
-		}
-		return true;
-	}
-
-	[[nodiscard]] bool HasPendingFault(uint64_t vaddr, uint64_t size) const {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			if (m_fault_pending.test(page)) {
-				return true;
-			}
-		}
-		return false;
-	}
-
-	[[nodiscard]] bool CompleteVirtualGpuWrite(uint64_t vaddr, uint64_t size) {
-		const auto [start, end] = GetPageRange(vaddr, size);
-		for (auto page = start; page < end; page++) {
-			if (!m_fault_pending.test(page)) {
-				return false;
-			}
-			if (!m_gpu_dirty.test(page)) {
-				EXIT("virtual GPU write completion found a non-GPU-dirty page\n");
-			}
-		}
-		for (auto page = start; page < end; page++) {
-			m_gpu_dirty.reset(page);
-			m_cpu_dirty.set(page);
-			m_writable.set(page);
-			m_fault_pending.reset(page);
-		}
-		return true;
 	}
 
 	template <DirtySource source, bool clear, typename Func>
-	RegionBits ForEachModifiedRange(uint64_t vaddr, uint64_t size, Func&& func) {
+	void ForEachModifiedRange(uint64_t vaddr, uint64_t size, Func&& func) {
 		const auto [start, end] = GetPageRange(vaddr, size);
-		auto mask               = GetBits<source>();
-		if constexpr (source == DirtySource::Cpu) {
-			mask &= ~m_fault_pending;
-		}
-		for (auto page = 0u; page < start; page++) {
-			mask.reset(page);
-		}
-		for (auto page = end; page < TRACKER_REGION_PAGES; page++) {
-			mask.reset(page);
-		}
+		RegionBits mask(GetBits<source>(), start, end);
 		if constexpr (clear) {
-			auto& bits = GetBits<source>();
-			for (auto page = start; page < end; page++) {
-				if (mask.test(page)) {
-					bits.reset(page);
-				}
-			}
+			GetBits<source>().UnsetRange(start, end);
 		}
 		if constexpr (source == DirtySource::Cpu && clear) {
-			auto changed = m_cpu_dirty ^ m_writable;
-			m_writable   = m_cpu_dirty;
-			ApplyProtection(changed, true);
+			UpdateCpuProtection<true>();
 			ForEachRange(mask, std::forward<Func>(func));
-			return changed;
+			return;
+		}
+		if constexpr (source == DirtySource::Gpu && clear) {
+			UpdateGpuProtection<false>();
 		}
 		ForEachRange(mask, std::forward<Func>(func));
-		if constexpr (clear) {
-			return mask;
-		}
-		return {};
-	}
-
-	void ApplyProtection(const RegionBits& changed, bool track) {
-		ForEachRange(changed, [this, track](uint64_t vaddr, uint64_t size) {
-			m_page_manager.UpdatePageWatchers(track, vaddr, size);
-		});
-	}
-
-	void ApplyGpuProtection(const RegionBits& changed, bool track, PageWatchMode mode) {
-		if (mode != PageWatchMode::Write && mode != PageWatchMode::ReadWrite) {
-			EXIT("unsupported GPU page-watch mode\n");
-		}
-		ForEachRange(changed, [this, track, mode](uint64_t vaddr, uint64_t size) {
-			m_page_manager.UpdatePageWatchers(track, vaddr, size, mode);
-		});
 	}
 
 	TrackingSpinLock lock;
 
 private:
+	template <bool track>
+	void UpdateCpuProtection() {
+		auto mask  = m_cpu_dirty ^ m_writable;
+		m_writable = m_cpu_dirty;
+		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);
+		}
+	}
+
 	template <DirtySource source>
 	RegionBits& GetBits() {
 		if constexpr (source == DirtySource::Cpu) {
@@ -331,18 +193,8 @@ private:
 
 	template <typename Func>
 	void ForEachRange(const RegionBits& bits, Func&& func) const {
-		size_t page = 0;
-		while (page < TRACKER_REGION_PAGES) {
-			while (page < TRACKER_REGION_PAGES && !bits.test(page)) {
-				page++;
-			}
-			const auto start = page;
-			while (page < TRACKER_REGION_PAGES && bits.test(page)) {
-				page++;
-			}
-			if (start != page) {
-				func(m_cpu_addr + start * TRACKER_PAGE_SIZE, (page - start) * TRACKER_PAGE_SIZE);
-			}
+		for (const auto [start, end]: bits) {
+			func(m_cpu_addr + start * TRACKER_PAGE_SIZE, (end - start) * TRACKER_PAGE_SIZE);
 		}
 	}
 
@@ -351,8 +203,7 @@ private:
 	RegionBits   m_cpu_dirty;
 	RegionBits   m_gpu_dirty;
 	RegionBits   m_writable;
-	RegionBits   m_fault_pending;
-	RegionBits   m_tracked;
+	RegionBits   m_readable;
 };
 
 } // namespace Libs::Graphics

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

@@ -3,15 +3,17 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/profiler.h"
+#include "graphics/guest_gpu/graphicsRun.h"
 #include "graphics/host_gpu/graphicContext.h"
-#include "graphics/host_gpu/renderer/cache/resourceMutex.h"
 #include "graphics/host_gpu/renderer/cache/textureCache.h"
 #include "graphics/host_gpu/renderer/commandScheduler.h"
 #include "graphics/host_gpu/renderer/render.h"
+#include "graphics/host_gpu/renderer/renderContext.h"
 #include "kernel/memory.h"
 
 #include <algorithm>
 #include <array>
+#include <cinttypes>
 #include <cstring>
 #include <utility>
 #include <vector>
@@ -57,15 +59,6 @@ uint64_t BufferCache::AlignUp(uint64_t value) {
 	return (value + CACHING_PAGE_SIZE - 1) & ~(CACHING_PAGE_SIZE - 1);
 }
 
-bool BufferCache::PageOverlaps(uint64_t left, uint64_t left_size, uint64_t right,
-                               uint64_t right_size) noexcept {
-	const auto left_begin  = left & ~(TRACKER_PAGE_SIZE - 1);
-	const auto left_end    = (left + left_size + TRACKER_PAGE_SIZE - 1) & ~(TRACKER_PAGE_SIZE - 1);
-	const auto right_begin = right & ~(TRACKER_PAGE_SIZE - 1);
-	const auto right_end = (right + right_size + TRACKER_PAGE_SIZE - 1) & ~(TRACKER_PAGE_SIZE - 1);
-	return left_begin < right_end && right_begin < left_end;
-}
-
 void BufferCache::Upload(CommandBuffer& command, Buffer& destination, uint64_t destination_offset,
                          const void* source, uint64_t size) {
 	auto* bytes = static_cast<const uint8_t*>(source);
@@ -123,30 +116,6 @@ struct BufferCache::RetiredBuffer {
 	std::shared_ptr<Buffer> owner;
 };
 
-struct BufferCache::FaultReadback {
-	PageFaultAccess            access = PageFaultAccess::Unknown;
-	uint64_t                   vaddr  = 0;
-	uint64_t                   size   = 0;
-	std::vector<DownloadRange> ranges;
-	bool                       installed = false;
-
-	[[nodiscard]] bool Active() const noexcept { return !ranges.empty(); }
-
-	void Reset() {
-		access    = PageFaultAccess::Unknown;
-		vaddr     = 0;
-		size      = 0;
-		installed = false;
-		ranges.clear();
-	}
-};
-
-struct BufferCache::PendingBackingPublication {
-	uint64_t address = 0;
-	uint64_t size    = 0;
-	uint64_t tick    = 0;
-};
-
 std::pair<uint64_t, uint64_t> BufferCache::DownloadEnvelope(const DownloadCopy& copy) {
 	if (copy.owner == nullptr || copy.size == 0 || copy.source_offset > copy.owner->Size() ||
 	    copy.size > copy.owner->Size() - copy.source_offset) {
@@ -218,48 +187,38 @@ void BufferCache::QueueGarbageDownload(std::span<const DownloadCopy> copies, Ret
 	if (copies.empty()) {
 		return;
 	}
-	auto       downloads = RecordDownloads(copies);
-	const auto tick      = m_scheduler.CurrentTick();
-	BeginBackingPublication(retire.address, retire.size, tick);
-	m_scheduler.DeferOperation([this, downloads = std::move(downloads), retire = std::move(retire),
-	                            tick]() mutable {
-		PublishDownloads(downloads);
-		{
-			FaultSafeCacheLock lock(this, m_mutex);
-			if (m_memory_tracker.IsRegionGpuModified(retire.address, retire.size)) {
-				m_memory_tracker.ForEachDownloadRange<true>(
-				    retire.address, retire.size,
-				    [&](uint64_t address, uint64_t size) noexcept {
-					    m_memory_tracker.ValidateGpuDirtyPages(m_gpu_modified_ranges, address, size,
-					                                           "asynchronous garbage retirement");
-				    },
-				    [](uint64_t, uint64_t) noexcept {});
-			}
-			for (const auto& range: downloads) {
-				m_gpu_modified_ranges.Subtract(range.address, range.size);
-			}
-			if (m_memory_tracker.IsRegionGpuModified(retire.address, retire.size) ||
-			    !m_gpu_modified_ranges.Intersections(retire.address, retire.size).empty()) {
-				EXIT("BufferCache: asynchronous garbage collection retained GPU ownership\n");
-			}
-			m_memory_tracker.UntrackMemory(retire.address, retire.size);
-		}
-		CompleteBackingPublication(retire.address, retire.size, tick);
-	});
+	auto downloads = RecordDownloads(copies);
+	m_scheduler.DeferOperation(
+	    [this, downloads = std::move(downloads), retire = std::move(retire)]() mutable {
+		    PublishDownloads(downloads);
+		    {
+			    FaultSafeCacheLock lock(this, m_mutex);
+			    for (const auto& range: downloads) {
+				    m_gpu_modified_ranges.Subtract(range.address, range.size);
+			    }
+			    // ForEachDownloadRange reports full tracker pages, and every exact GPU-owned
+			    // interval on those pages was downloaded and removed. Clearing the original
+			    // query therefore cannot orphan a dirty sibling on an edge page.
+			    m_memory_tracker.UnmarkRegionAsGpuModified(retire.address, retire.size);
+			    if (m_memory_tracker.IsRegionGpuModified(retire.address, retire.size) ||
+			        !m_gpu_modified_ranges.Intersections(retire.address, retire.size).empty()) {
+				    EXIT("BufferCache: asynchronous garbage collection retained GPU ownership\n");
+			    }
+			    m_memory_tracker.UntrackMemory(retire.address, retire.size);
+		    }
+	    });
 }
 
 BufferCache::BufferCache(GraphicContext& graphics, CommandScheduler& scheduler,
-                         PageManager& page_manager, TextureCache& texture_cache,
-                         ResourceMutex& resource_mutex)
+                         PageManager& page_manager, TextureCache& texture_cache)
     : m_graphics(graphics), m_scheduler(scheduler),
       m_gds_buffer(graphics, scheduler, MemoryUsage::Stream, 0, AllFlags, GdsBufferSize),
-      m_fault_readback(std::make_unique<FaultReadback>()), m_memory_tracker(page_manager),
+      m_memory_tracker(page_manager),
       m_staging_buffer(graphics, scheduler, MemoryUsage::Upload, 512 * MiB),
       m_stream_buffer(graphics, scheduler, MemoryUsage::Stream, 64 * MiB),
       m_download_buffer(graphics, scheduler, MemoryUsage::Download, 32 * MiB),
       m_device_buffer(graphics, scheduler, MemoryUsage::DeviceLocal, 128 * MiB),
-      m_page_manager(page_manager), m_texture_cache(texture_cache),
-      m_resource_mutex(resource_mutex) {
+      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());
 	if (!m_graphics.CanReportMemoryUsage()) {
@@ -277,15 +236,9 @@ BufferCache::BufferCache(GraphicContext& graphics, CommandScheduler& scheduler,
 }
 
 BufferCache::~BufferCache() {
-	if (m_fault_readback->Active()) {
-		EXIT("BufferCache: destroyed with an active fault readback\n");
-	}
 	if (!m_gpu_modified_ranges.Empty()) {
 		EXIT("BufferCache: destroyed with pending GPU-modified ranges\n");
 	}
-	if (!m_pending_backing_publications.empty()) {
-		EXIT("BufferCache: destroyed with pending backing publications\n");
-	}
 	for (const auto& [vaddr, cached]: m_buffers) {
 		(void)vaddr;
 		if (m_memory_tracker.IsRegionGpuModified(cached->vaddr, cached->size)) {
@@ -295,68 +248,6 @@ BufferCache::~BufferCache() {
 	m_buffers.clear();
 }
 
-bool BufferCache::SynchronizeBacking(uint64_t vaddr, uint64_t size) {
-	bool waited = false;
-	for (;;) {
-		uint64_t   tick       = 0;
-		const auto page_begin = vaddr & ~(TRACKER_PAGE_SIZE - 1);
-		const auto page_end   = (vaddr + size + TRACKER_PAGE_SIZE - 1) & ~(TRACKER_PAGE_SIZE - 1);
-		CacheRange affected {.address = page_begin, .size = page_end - page_begin};
-		{
-			FaultSafeCacheLock lock(this, m_mutex);
-			bool               changed = true;
-			while (changed) {
-				changed = false;
-				for (const auto& [address, cached]: m_buffers) {
-					const CacheRange previous = affected;
-					if (ResolveOverlap(affected, {address, cached->size}) &&
-					    (previous.address != affected.address || previous.size != affected.size)) {
-						changed = true;
-					}
-				}
-			}
-		}
-		{
-			std::lock_guard lock(m_publication_mutex);
-			for (const auto& publication: m_pending_backing_publications) {
-				if (publication.address < affected.address + affected.size &&
-				    affected.address < publication.address + publication.size) {
-					tick = std::max(tick, publication.tick);
-				}
-			}
-		}
-		if (tick == 0) {
-			return waited;
-		}
-		waited = true;
-		m_scheduler.Wait(tick);
-		m_scheduler.WaitPriorityOperations(tick);
-	}
-}
-
-void BufferCache::RefreshInvalidatedRanges(CommandBuffer& command, CachedBuffer& cached,
-                                           uint64_t vaddr, uint64_t size, bool upload) {
-	const auto invalidated = m_image_invalidated_ranges.Intersections(vaddr, size);
-	if (upload) {
-		std::array<uint8_t, 64 * 1024> bytes;
-		for (const auto& range: invalidated) {
-			for (uint64_t copied = 0; copied < range.size;) {
-				const auto chunk = std::min<uint64_t>(range.size - copied, bytes.size());
-				if (!Libs::LibKernel::Memory::TryReadBacking(range.address + copied, bytes.data(),
-				                                             chunk)) {
-					EXIT("BufferCache: failed to refresh an invalidated image alias\n");
-				}
-				Upload(command, *cached.buffer, cached.buffer->Offset(range.address + copied),
-				       bytes.data(), chunk);
-				copied += chunk;
-			}
-		}
-	}
-	if (!invalidated.empty()) {
-		m_image_invalidated_ranges.Subtract(vaddr, size);
-	}
-}
-
 StreamBuffer& BufferCache::GetUtilityBuffer(MemoryUsage usage) noexcept {
 	switch (usage) {
 		case MemoryUsage::Upload: return m_staging_buffer;
@@ -385,16 +276,28 @@ void BufferCache::InvalidateMemory(uint64_t vaddr, uint64_t size) {
 	    size > TRACKER_ADDRESS_SIZE - vaddr) {
 		EXIT("BufferCache: invalid memory-invalidation range\n");
 	}
-	(void)SynchronizeBacking(vaddr, size);
-	if (!HasPageOverlap(vaddr, size)) {
+	if (!IsRegionRegistered(vaddr, size)) {
 		return;
 	}
 	m_memory_tracker.InvalidateRegion(vaddr, size,
-	                                  [this, vaddr, size] { ReadMemory(vaddr, size); });
+	                                  [this, vaddr, size] { ReadMemory(vaddr, size, true); });
 }
 
-void BufferCache::ReadMemory(uint64_t vaddr, uint64_t size) {
-	(void)SynchronizeBacking(vaddr, size);
+void BufferCache::ReadMemory(uint64_t vaddr, uint64_t size, bool is_write) {
+	if (Gpu::IsGpuThread()) {
+		ReadMemoryOnGpu(vaddr, size, is_write);
+		return;
+	}
+	if (CommandScheduler::InDeferredOperation()) {
+		EXIT("unsupported buffer readback from an asynchronous GPU completion, "
+		     "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) {
 	std::vector<DownloadCopy> copies;
 	{
 		FaultSafeCacheLock lock(this, m_mutex);
@@ -425,126 +328,40 @@ void BufferCache::ReadMemory(uint64_t vaddr, uint64_t size) {
 				    }
 			    }
 		    });
-	}
-	if (copies.empty()) {
-		return;
+		if (copies.empty()) {
+			if (!is_write) {
+				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;
+		}
 	}
 	auto downloads = RecordDownloads(copies);
 	m_scheduler.FinishCurrent();
 	PublishDownloads(downloads);
 	{
 		FaultSafeCacheLock lock(this, m_mutex);
-		m_memory_tracker.ForEachDownloadRange<true>(
-		    vaddr, size,
-		    [&](uint64_t address, uint64_t bytes) noexcept {
-			    m_memory_tracker.ValidateGpuDirtyPages(m_gpu_modified_ranges, address, bytes,
-			                                           "memory invalidation completion");
-		    },
-		    [](uint64_t, uint64_t) noexcept {});
 		for (const auto& range: downloads) {
 			m_gpu_modified_ranges.Subtract(range.address, range.size);
 		}
-	}
-}
-
-bool BufferCache::InvalidateMemory(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-                                   PageFaultPhase phase) noexcept {
-	const auto page = vaddr & ~(TRACKER_PAGE_SIZE - 1);
-	if (size == 0 || size > page + TRACKER_PAGE_SIZE - vaddr) {
-		EXIT("BufferCache: invalid page-fault range\n");
-	}
-
-	if (phase == PageFaultPhase::Complete) {
-		FaultSafeCacheLock lock(this, m_mutex);
-		auto&              fault = *m_fault_readback;
-		if (!fault.Active()) {
-			return m_memory_tracker.CompleteCpuFault(vaddr, size, access, false);
-		}
-		if (fault.access != access || fault.vaddr != vaddr || fault.size != size ||
-		    fault.installed) {
-			EXIT("BufferCache: mismatched fault readback completion\n");
-		}
-		PublishDownloads(fault.ranges);
-		if (!m_memory_tracker.CompleteCpuFault(vaddr, size, access, true)) {
-			EXIT("BufferCache: failed to complete downloaded CPU fault\n");
-		}
-		fault.installed = true;
-		return true;
-	}
-
-	if (phase == PageFaultPhase::Release) {
-		FaultSafeCacheLock lock(this, m_mutex);
-		auto&              fault = *m_fault_readback;
-		if (fault.Active()) {
-			if (fault.access != access || fault.vaddr != vaddr || fault.size != size ||
-			    !fault.installed) {
-				EXIT("BufferCache: mismatched fault readback release\n");
-			}
-			for (const auto& range: fault.ranges) {
-				m_gpu_modified_ranges.Subtract(range.address, range.size);
-			}
-			fault.Reset();
-		}
-		return true;
-	}
-
-	if (phase != PageFaultPhase::Invalidate) {
-		EXIT("BufferCache: unsupported page-fault phase\n");
-	}
-
-	const auto action = m_memory_tracker.BeginCpuFault(vaddr, size, access);
-	if (action != CpuFaultAction::Download) {
-		return action == CpuFaultAction::Continue;
-	}
-
-	auto&                     fault = *m_fault_readback;
-	std::vector<DownloadCopy> copies;
-	{
-		FaultSafeCacheLock lock(this, m_mutex);
-		if (fault.Active()) {
-			EXIT("BufferCache: nested fault readback\n");
-		}
-		fault.access = access;
-		fault.vaddr  = vaddr;
-		fault.size   = size;
-
-		m_gpu_modified_ranges.ForEachIntersection(
-		    page, TRACKER_PAGE_SIZE, [&](RangeSet::Range range) {
-			    auto owner = m_buffers.upper_bound(range.address);
-			    if (owner == m_buffers.begin()) {
-				    EXIT("BufferCache: fault readback has no buffer owner\n");
-			    }
-			    --owner;
-			    auto& cached = *owner->second;
-			    if (!cached.buffer->IsInBounds(range.address, range.size)) {
-				    EXIT("BufferCache: fault readback is outside its buffer owner\n");
-			    }
-			    copies.push_back({cached.buffer, cached.buffer->Offset(range.address),
-			                      range.address, range.size});
-		    });
-		if (copies.empty()) {
-			EXIT("BufferCache: GPU-dirty fault page has no dirty byte ranges\n");
+		// 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);
 		}
 	}
-	fault.ranges = RecordDownloads(copies);
-	if (!fault.Active()) {
-		EXIT("BufferCache: GPU-dirty fault page has no dirty byte ranges\n");
-	}
-	m_scheduler.FinishCurrent();
-	return true;
 }
 
 void BufferCache::UnmapMemory(uint64_t vaddr, uint64_t size) {
 	if (vaddr == 0 || size == 0 || size > UINT64_MAX - vaddr) {
 		EXIT("BufferCache: invalid unmap range\n");
 	}
-	(void)SynchronizeBacking(vaddr, size);
-
-	std::vector<DownloadCopy>                               copies;
-	std::vector<RangeSet::Range>                            dirty_ranges;
-	std::vector<std::pair<uint64_t, uint64_t>>              modified_buffers;
-	std::vector<std::unique_ptr<PageManager::BackingWrite>> backing_writes;
-	std::vector<std::pair<uint64_t, uint64_t>>              retired_buffers;
+	std::vector<DownloadCopy>                  copies;
+	std::vector<std::pair<uint64_t, uint64_t>> modified_buffers;
+	std::vector<std::pair<uint64_t, uint64_t>> retired_buffers;
 	{
 		FaultSafeCacheLock lock(this, m_mutex);
 		for (const auto& [begin, cached]: m_buffers) {
@@ -561,12 +378,8 @@ void BufferCache::UnmapMemory(uint64_t vaddr, uint64_t size) {
 			if (dirty.empty()) {
 				EXIT("BufferCache: GPU-modified buffer has no dirty ranges\n");
 			}
-			dirty_ranges.insert(dirty_ranges.end(), dirty.begin(), dirty.end());
 			modified_buffers.emplace_back(begin, cached->size);
 		}
-		if (!dirty_ranges.empty()) {
-			backing_writes = m_page_manager.ReserveBackingWrites(dirty_ranges);
-		}
 		for (const auto& [begin, bytes]: modified_buffers) {
 			auto owner = m_buffers.find(begin);
 			if (owner == m_buffers.end() || owner->second->size != bytes) {
@@ -596,24 +409,11 @@ void BufferCache::UnmapMemory(uint64_t vaddr, uint64_t size) {
 		// command stream before removing such backing.
 		m_scheduler.FinishCurrent();
 	}
-	backing_writes.clear();
-
 	{
 		FaultSafeCacheLock lock(this, m_mutex);
-		for (const auto& [begin, bytes]: modified_buffers) {
-			if (!m_memory_tracker.IsRegionGpuModified(begin, bytes)) {
-				continue;
-			}
-			m_memory_tracker.ForEachDownloadRange<true>(
-			    begin, bytes,
-			    [&](uint64_t address, uint64_t download_size) noexcept {
-				    m_memory_tracker.ValidateGpuDirtyPages(m_gpu_modified_ranges, address,
-				                                           download_size, "unmap retirement");
-			    },
-			    [](uint64_t, uint64_t) noexcept {});
-		}
 		for (const auto& [begin, bytes]: modified_buffers) {
 			m_gpu_modified_ranges.Subtract(begin, bytes);
+			m_memory_tracker.UnmarkRegionAsGpuModified(begin, bytes);
 		}
 		for (const auto& [begin, bytes]: retired_buffers) {
 			m_memory_tracker.MarkRegionAsCpuModified(begin, bytes);
@@ -621,7 +421,6 @@ void BufferCache::UnmapMemory(uint64_t vaddr, uint64_t size) {
 		if (!m_gpu_modified_ranges.Intersections(vaddr, size).empty()) {
 			EXIT("BufferCache: unmap retained dirty byte ranges\n");
 		}
-		m_image_invalidated_ranges.Subtract(vaddr, size);
 		m_memory_tracker.UntrackMemory(vaddr, size);
 		for (auto it = m_buffers.begin(); it != m_buffers.end();) {
 			if (vaddr < it->first + it->second->size && it->first < vaddr + size) {
@@ -714,42 +513,59 @@ BufferBinding BufferCache::ObtainBuffer(CommandBuffer& command, uint64_t vaddr,
 	if (command.IsInvalid() || command.IsExecute()) {
 		EXIT("BufferCache: buffer request requires a recording command buffer\n");
 	}
-	ValidateGpuAccess(vaddr, size, is_read, is_written);
-	std::lock_guard transaction(m_resource_mutex);
-	(void)SynchronizeBacking(vaddr, size);
 
 	if (is_read && !is_written && size <= CACHING_PAGE_SIZE &&
 	    !m_memory_tracker.IsRegionGpuModified(vaddr, size) &&
 	    m_memory_tracker.IsRegionCpuModified(vaddr, size)) {
-		std::vector<uint8_t> data(size);
-		if (Libs::LibKernel::Memory::TryReadBacking(vaddr, data.data(), size)) {
-			return UploadTransient(data.data(), size, 16);
+		const auto alignment = std::max<uint64_t>(
+		    m_graphics.physical_device_properties.limits.minUniformBufferOffsetAlignment, 1);
+		if (auto [mapped, offset] = m_stream_buffer.Map(size, alignment, false);
+		    mapped != nullptr) {
+			if (Libs::LibKernel::Memory::TryReadBacking(vaddr, mapped, size)) {
+				m_stream_buffer.Commit();
+				return {{}, m_stream_buffer.Handle(), offset};
+			}
+		} else {
+			auto owner = std::make_shared<Buffer>(m_graphics, m_scheduler, MemoryUsage::Upload, 0,
+			                                      AllFlags, size);
+			if (Libs::LibKernel::Memory::TryReadBacking(vaddr, owner->Mapped().data(), size)) {
+				owner->Flush(0, size);
+				return {owner, owner->Handle(), 0};
+			}
 		}
 	}
 
-	if (is_formatted && is_read && !is_written) {
-		(void)m_texture_cache.SynchronizeImageToBuffer(vaddr, size);
-	} else if (is_formatted && is_written) {
+	if (is_formatted && is_written) {
 		(void)m_texture_cache.InvalidateMemoryFromGPU(vaddr, size, true);
 	}
 
-	FaultSafeCacheLock                         lock(this, m_mutex);
-	auto&                                      cached = GetOrCreateBuffer(command, vaddr, size);
-	std::vector<std::pair<uint64_t, uint64_t>> uploads;
-	m_memory_tracker.ForEachUploadRange(
-	    vaddr, size, is_written,
-	    [&](uint64_t address, uint64_t bytes) noexcept { uploads.emplace_back(address, bytes); },
-	    [&]() noexcept {
-		    for (const auto& [address, bytes]: uploads) {
-			    Upload(command, *cached.buffer, cached.buffer->Offset(address),
-			           reinterpret_cast<const void*>(address), bytes);
-		    }
-	    });
-	RefreshInvalidatedRanges(command, cached, vaddr, size, is_read);
-	if (is_written) {
-		m_gpu_modified_ranges.Add(vaddr, size);
+	std::shared_ptr<Buffer> owner;
+	uint64_t                offset = 0;
+	{
+		FaultSafeCacheLock                         lock(this, m_mutex);
+		auto&                                      cached = GetOrCreateBuffer(command, vaddr, size);
+		std::vector<std::pair<uint64_t, uint64_t>> uploads;
+		m_memory_tracker.ForEachUploadRange(
+		    vaddr, size, is_written,
+		    [&](uint64_t address, uint64_t bytes) noexcept {
+			    uploads.emplace_back(address, bytes);
+		    },
+		    [&]() noexcept {
+			    for (const auto& [address, bytes]: uploads) {
+				    Upload(command, *cached.buffer, cached.buffer->Offset(address),
+				           reinterpret_cast<const void*>(address), bytes);
+			    }
+		    });
+		if (is_written) {
+			m_gpu_modified_ranges.Add(vaddr, size);
+		}
+		owner  = cached.buffer;
+		offset = cached.buffer->Offset(vaddr);
 	}
-	return {cached.buffer, cached.buffer->Handle(), cached.buffer->Offset(vaddr)};
+	if (is_formatted && is_read && !is_written) {
+		(void)SynchronizeBufferFromImage(*owner, vaddr, size);
+	}
+	return {owner, owner->Handle(), offset};
 }
 
 std::shared_ptr<Buffer> BufferCache::ObtainNullBuffer() {
@@ -773,7 +589,6 @@ ImageBufferSource BufferCache::ObtainBufferForImage(uint64_t vaddr, uint64_t siz
 	    size > TRACKER_ADDRESS_SIZE - vaddr) {
 		EXIT("BufferCache: invalid image source\n");
 	}
-	(void)SynchronizeBacking(vaddr, size);
 	auto find_owner = [&]() {
 		auto owner = m_buffers.upper_bound(vaddr);
 		if (owner == m_buffers.begin()) {
@@ -788,13 +603,12 @@ ImageBufferSource BufferCache::ObtainBufferForImage(uint64_t vaddr, uint64_t siz
 		const bool         cpu_modified = m_memory_tracker.IsRegionCpuModified(vaddr, size);
 		const bool         gpu_modified = m_memory_tracker.IsRegionGpuModified(vaddr, size);
 		const auto         dirty        = m_gpu_modified_ranges.Intersections(vaddr, size);
-		const bool invalidated = !m_image_invalidated_ranges.Intersections(vaddr, size).empty();
-		const bool requested_gpu_owned = !dirty.empty();
+		const bool         has_dirty_buffer_source = !dirty.empty();
 		m_memory_tracker.ValidateGpuDirtyOwnership(m_gpu_modified_ranges, vaddr, size,
 		                                           "image source");
 
 		auto owner = find_owner();
-		if (requested_gpu_owned && owner == m_buffers.end()) {
+		if (has_dirty_buffer_source && owner == m_buffers.end()) {
 			CacheRange merged {.address = AlignDown(vaddr),
 			                   .size    = AlignUp(vaddr + size) - AlignDown(vaddr)};
 			using Iterator = decltype(m_buffers.begin());
@@ -844,42 +658,32 @@ ImageBufferSource BufferCache::ObtainBufferForImage(uint64_t vaddr, uint64_t siz
 				EXIT("BufferCache: merged image source does not contain the requested range\n");
 			}
 		}
-		if (owner != m_buffers.end() && !cpu_modified && !invalidated &&
-		    (!gpu_modified || requested_gpu_owned)) {
-			DiscardGpuDirtyBytesLocked(vaddr, size, "image source transfer");
+		if (owner != m_buffers.end() && !cpu_modified &&
+		    (!gpu_modified || has_dirty_buffer_source)) {
 			owner->second->tick_accessed_last = m_gc_tick;
-			return {owner->second->buffer.get(), owner->second->buffer->Offset(vaddr),
-			        requested_gpu_owned};
+			return {owner->second->buffer.get(), owner->second->buffer->Offset(vaddr)};
 		}
-		if (requested_gpu_owned && owner == m_buffers.end()) {
+		if (has_dirty_buffer_source && owner == m_buffers.end()) {
 			EXIT("BufferCache: GPU-dirty image source could not resolve its native owner\n");
 		}
 	}
 
-	// Direct-memory backing remains readable while PageManager protects the guest mapping. The
-	// fallback exists for plain host mappings used by standalone renderer tests and is deliberately
-	// performed outside the cache lock so a page fault cannot recurse into BufferCache.
-	const auto stage_address = vaddr & ~(TRACKER_PAGE_SIZE - 1);
-	const auto stage_end     = (vaddr + size + TRACKER_PAGE_SIZE - 1) & ~(TRACKER_PAGE_SIZE - 1);
-	const auto stage_size    = stage_end - stage_address;
-	(void)SynchronizeBacking(stage_address, stage_size);
-	std::vector<uint8_t> bytes(stage_size);
-	if (!Libs::LibKernel::Memory::TryReadBacking(stage_address, bytes.data(), stage_size)) {
+	auto [staging, stage_offset] = m_staging_buffer.Map(size, 16);
+	if (staging == nullptr || !Libs::LibKernel::Memory::TryReadBacking(vaddr, staging, size)) {
 		EXIT("BufferCache: failed to read mapped guest image backing\n");
 	}
+	m_staging_buffer.Commit();
 
 	FaultSafeCacheLock lock(this, m_mutex);
-	const auto         dirty       = m_gpu_modified_ranges.Intersections(vaddr, size);
-	const bool         invalidated = !m_image_invalidated_ranges.Intersections(vaddr, size).empty();
-	const bool         requested_gpu_owned = !dirty.empty();
-	auto               owner               = find_owner();
-	if (requested_gpu_owned && owner == m_buffers.end()) {
+	const auto         dirty                   = m_gpu_modified_ranges.Intersections(vaddr, size);
+	const bool         has_dirty_buffer_source = !dirty.empty();
+	auto               owner                   = find_owner();
+	if (has_dirty_buffer_source && owner == m_buffers.end()) {
 		EXIT("BufferCache: GPU-dirty image source lost its native owner\n");
 	}
-	const auto stage_offset = m_staging_buffer.Copy(bytes.data(), stage_size, 16);
-	if (owner == m_buffers.end() || invalidated ||
-	    (m_memory_tracker.IsRegionGpuModified(vaddr, size) && !requested_gpu_owned)) {
-		return {&m_staging_buffer, stage_offset + vaddr - stage_address, false};
+	if (owner == m_buffers.end() ||
+	    (m_memory_tracker.IsRegionGpuModified(vaddr, size) && !has_dirty_buffer_source)) {
+		return {&m_staging_buffer, stage_offset};
 	}
 
 	auto& cached              = *owner->second;
@@ -893,42 +697,17 @@ ImageBufferSource BufferCache::ObtainBufferForImage(uint64_t vaddr, uint64_t siz
 	    [&]() noexcept {
 		    for (const auto& [address, upload_size]: uploads) {
 			    cached.buffer->CopyFrom(
-			        m_scheduler.Current(), m_staging_buffer, stage_offset + address - stage_address,
+			        m_scheduler.Current(), m_staging_buffer, stage_offset + address - vaddr,
 			        cached.buffer->Offset(address), upload_size, vk::AccessFlagBits::eHostWrite);
 		    }
 	    });
-	DiscardGpuDirtyBytesLocked(vaddr, size, "staged image source transfer");
-	return {cached.buffer.get(), cached.buffer->Offset(vaddr), requested_gpu_owned};
-}
-
-void BufferCache::DiscardGpuDirtyBytesLocked(uint64_t vaddr, uint64_t size, const char* operation) {
-	m_memory_tracker.ValidateGpuDirtyOwnership(m_gpu_modified_ranges, vaddr, size, operation);
-	m_gpu_modified_ranges.Subtract(vaddr, size);
-	const auto page_begin = vaddr & ~(TRACKER_PAGE_SIZE - 1);
-	const auto page_end   = (vaddr + size + TRACKER_PAGE_SIZE - 1) & ~(TRACKER_PAGE_SIZE - 1);
-	for (auto page = page_begin; page < page_end; page += TRACKER_PAGE_SIZE) {
-		if (m_gpu_modified_ranges.Intersections(page, TRACKER_PAGE_SIZE).empty() &&
-		    m_memory_tracker.IsRegionGpuModified(page, TRACKER_PAGE_SIZE)) {
-			m_memory_tracker.UnmarkRegionAsGpuModified(page, TRACKER_PAGE_SIZE);
-		}
-	}
-	m_memory_tracker.ValidateGpuDirtyOwnership(m_gpu_modified_ranges, vaddr, size, operation);
-}
-
-void BufferCache::DiscardGpuDirtyBytes(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
-		EXIT("BufferCache: invalid dirty-byte discard range\n");
-	}
-	FaultSafeCacheLock lock(this, m_mutex);
-	DiscardGpuDirtyBytesLocked(vaddr, size, "image output supersession");
+	return {cached.buffer.get(), cached.buffer->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");
 	}
-	(void)SynchronizeBacking(vaddr, data.size());
 	Libs::LibKernel::Memory::WriteBacking(vaddr, data.data(), data.size());
 
 	FaultSafeCacheLock lock(this, m_mutex);
@@ -944,45 +723,6 @@ void BufferCache::WriteHostMemory(uint64_t vaddr, std::span<const uint8_t> data)
 		       data.data() + begin - vaddr, range_end - begin);
 		cached->tick_accessed_last = m_gc_tick;
 	}
-	m_image_invalidated_ranges.Subtract(vaddr, data.size());
-}
-
-std::pair<std::shared_ptr<Buffer>, uint64_t> BufferCache::ObtainBufferForImageWrite(uint64_t vaddr,
-                                                                                    uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
-		EXIT("BufferCache: invalid image destination\n");
-	}
-	const auto stage_address = vaddr & ~(TRACKER_PAGE_SIZE - 1);
-	const auto stage_end     = (vaddr + size + TRACKER_PAGE_SIZE - 1) & ~(TRACKER_PAGE_SIZE - 1);
-	const auto stage_size    = stage_end - stage_address;
-	(void)SynchronizeBacking(stage_address, stage_size);
-	std::vector<uint8_t> bytes(stage_size);
-	if (!Libs::LibKernel::Memory::TryReadBacking(stage_address, bytes.data(), stage_size)) {
-		EXIT("BufferCache: failed to preserve guest bytes around an image mirror\n");
-	}
-	FaultSafeCacheLock lock(this, m_mutex);
-	auto&              cached = GetOrCreateBuffer(m_scheduler.Current(), vaddr, size);
-	m_memory_tracker.ValidateGpuDirtyOwnership(m_gpu_modified_ranges, vaddr, size,
-	                                           "image destination");
-	if (!m_gpu_modified_ranges.Intersections(vaddr, size).empty()) {
-		EXIT("BufferCache: image destination aliases GPU-owned buffer bytes\n");
-	}
-	const auto stage_offset = m_staging_buffer.Copy(bytes.data(), stage_size, 16);
-	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) {
-			    cached.buffer->CopyFrom(
-			        m_scheduler.Current(), m_staging_buffer, stage_offset + address - stage_address,
-			        cached.buffer->Offset(address), upload_size, vk::AccessFlagBits::eHostWrite);
-		    }
-	    });
-	return {cached.buffer, cached.buffer->Offset(vaddr)};
 }
 
 void BufferCache::FillBuffer(uint64_t vaddr, uint64_t size, uint32_t value, bool is_gds) {
@@ -999,11 +739,9 @@ void BufferCache::FillBuffer(uint64_t vaddr, uint64_t size, uint32_t value, bool
 	if (vaddr == 0) {
 		EXIT("BufferCache: invalid fill memory address\n");
 	}
-	ValidateGpuAccess(vaddr, size, false, true);
 	(void)m_texture_cache.ClearMeta(vaddr);
 	{
-		std::lock_guard transaction(m_resource_mutex);
-		const auto      region = m_texture_cache.QueryRegion(vaddr, size);
+		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);
@@ -1041,25 +779,11 @@ void BufferCache::CopyBuffer(uint64_t dst_vaddr, uint64_t src_vaddr, uint64_t si
 	    (src_gds && (src_vaddr > m_gds_buffer.Size() || size > m_gds_buffer.Size() - src_vaddr))) {
 		EXIT("BufferCache: invalid or overlapping copy range\n");
 	}
-	if (src_memory) {
-		ValidateGpuAccess(src_vaddr, size, true, false);
-	}
-	if (dst_memory) {
-		ValidateGpuAccess(dst_vaddr, size, false, true);
-	}
 	if (src_memory || dst_memory) {
-		std::lock_guard transaction(m_resource_mutex);
-		if (src_memory) {
-			(void)SynchronizeBacking(src_vaddr, size);
-		}
 		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 && src_region.gpu_image_bytes &&
-		    !m_texture_cache.SynchronizeImageToBuffer(src_vaddr, size)) {
-			EXIT("BufferCache: GPU copy source image could not be synchronized\n");
-		}
 		if (src_memory && dst_memory && !HasGpuDirtyBytes(src_vaddr, size) &&
 		    !HasGpuDirtyBytes(dst_vaddr, size) && !src_region.gpu_image_bytes &&
 		    !dst_region.gpu_image_bytes) {
@@ -1081,7 +805,7 @@ void BufferCache::CopyBuffer(uint64_t dst_vaddr, uint64_t src_vaddr, uint64_t si
 	}
 
 	auto& command = m_scheduler.Current();
-	auto  src = src_memory ? ObtainBuffer(command, src_vaddr, size, false, true)
+	auto  src = src_memory ? ObtainBuffer(command, src_vaddr, size, false, true, true)
 	                       : BufferBinding {.buffer = m_gds_buffer.Handle(), .offset = src_vaddr};
 	auto  dst = dst_memory ? ObtainBuffer(command, dst_vaddr, size, true, false, true)
 	                       : BufferBinding {.buffer = m_gds_buffer.Handle(), .offset = dst_vaddr};
@@ -1107,18 +831,20 @@ void BufferCache::CopyBuffer(uint64_t dst_vaddr, uint64_t src_vaddr, uint64_t si
 	destination.CopyFrom(command, source, src.offset, dst.offset, size);
 }
 
-bool BufferCache::HasPageOverlap(uint64_t vaddr, uint64_t size) {
+bool BufferCache::IsRegionRegistered(uint64_t vaddr, uint64_t size) {
 	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
 	    size > TRACKER_ADDRESS_SIZE - vaddr) {
-		EXIT("BufferCache: invalid page-overlap query\n");
+		EXIT("BufferCache: invalid registered-region query\n");
 	}
 	FaultSafeCacheLock lock(this, m_mutex);
-	for (const auto& [address, cached]: m_buffers) {
-		if (PageOverlaps(vaddr, size, address, cached->size)) {
-			return true;
-		}
+	// Cached buffers are ordered and non-overlapping. The last buffer beginning before the query
+	// end is therefore the only possible intersection.
+	const auto candidate = m_buffers.lower_bound(vaddr + size);
+	if (candidate == m_buffers.begin()) {
+		return false;
 	}
-	return false;
+	const auto& [address, cached] = *std::prev(candidate);
+	return address + cached->size > vaddr;
 }
 
 bool BufferCache::IsRegionGpuModified(uint64_t vaddr, uint64_t size) {
@@ -1134,95 +860,12 @@ bool BufferCache::IsRegionCpuModified(uint64_t vaddr, uint64_t size) {
 	return m_memory_tracker.IsRegionCpuModified(vaddr, size);
 }
 
-void BufferCache::InvalidateImageAliases(uint64_t vaddr, uint64_t size) {
-	if (vaddr == 0 || size == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
-		EXIT("BufferCache: invalid image-alias invalidation\n");
-	}
-	FaultSafeCacheLock lock(this, m_mutex);
-	const auto         end = vaddr + size;
-	for (const auto& [address, cached]: m_buffers) {
-		const auto cached_end = address + cached->size;
-		const auto begin      = std::max(vaddr, address);
-		const auto range_end  = std::min(end, cached_end);
-		if (begin >= range_end) {
-			continue;
-		}
-		const auto bytes = range_end - begin;
-		if (!m_gpu_modified_ranges.Intersections(begin, bytes).empty()) {
-			EXIT("BufferCache: image ownership overlaps exact dirty buffer bytes\n");
-		}
-		m_image_invalidated_ranges.Add(begin, bytes);
-	}
-}
-
-void BufferCache::BeginBackingPublication(uint64_t vaddr, uint64_t size, uint64_t tick) {
-	if (vaddr == 0 || size == 0 || tick == 0 || vaddr >= TRACKER_ADDRESS_SIZE ||
-	    size > TRACKER_ADDRESS_SIZE - vaddr) {
-		EXIT("BufferCache: invalid pending backing publication\n");
-	}
-	std::lock_guard lock(m_publication_mutex);
-	m_pending_backing_publications.push_back({vaddr, size, tick});
-}
-
-void BufferCache::CompleteBackingPublication(uint64_t vaddr, uint64_t size, uint64_t tick) {
-	std::lock_guard lock(m_publication_mutex);
-	const auto      publication =
-	    std::ranges::find_if(m_pending_backing_publications, [&](const auto& pending) {
-		    return pending.address == vaddr && pending.size == size && pending.tick == tick;
-	    });
-	if (publication == m_pending_backing_publications.end()) {
-		EXIT("BufferCache: completed an unknown backing publication\n");
-	}
-	m_pending_backing_publications.erase(publication);
-}
-
-void BufferCache::PublishImageBuffer(uint64_t vaddr, uint64_t size) {
-	FaultSafeCacheLock lock(this, m_mutex);
-	auto               owner = m_buffers.end();
-	for (auto it = m_buffers.begin(); it != m_buffers.end(); ++it) {
-		if (!PageOverlaps(vaddr, size, it->second->vaddr, it->second->size)) {
-			continue;
-		}
-		if (owner != m_buffers.end() || !it->second->buffer->IsInBounds(vaddr, size)) {
-			EXIT("BufferCache: image destination aliases a non-containing cached buffer\n");
-		}
-		owner = it;
-	}
-	m_memory_tracker.ValidateGpuDirtyOwnership(m_gpu_modified_ranges, vaddr, size,
-	                                           "image destination publication");
-	if (owner == m_buffers.end() || m_memory_tracker.IsRegionCpuModified(vaddr, size) ||
-	    !m_gpu_modified_ranges.Intersections(vaddr, size).empty()) {
-		EXIT("BufferCache: image destination requires clean buffer ownership\n");
-	}
-	m_memory_tracker.MarkRegionAsGpuModified(vaddr, size);
-	m_gpu_modified_ranges.Add(vaddr, size);
-	m_image_invalidated_ranges.Subtract(vaddr, size);
-	m_memory_tracker.ValidateGpuDirtyOwnership(m_gpu_modified_ranges, vaddr, size,
-	                                           "published image destination");
-	owner->second->tick_accessed_last = m_gc_tick;
-}
-
-void BufferCache::ValidateGpuAccess(uint64_t vaddr, uint64_t size, bool is_read,
-                                    bool is_written) const {
-	if ((!is_read && !is_written) || vaddr == 0 || size == 0 || size > UINT64_MAX - vaddr) {
-		EXIT("BufferCache: invalid GPU access request\n");
-	}
-	if (is_read && !m_page_manager.HasGpuAccess(vaddr, size, GpuAccess::Read)) {
-		EXIT("BufferCache: GPU-read access denied\n");
-	}
-	if (is_written && !m_page_manager.HasGpuAccess(vaddr, size, GpuAccess::Write)) {
-		EXIT("BufferCache: GPU-write access denied\n");
-	}
-}
-
 void BufferCache::RunGarbageCollector() {
-	std::lock_guard transaction(m_resource_mutex);
-	const auto      tick = m_gc_tick++;
+	const auto tick = m_gc_tick++;
 	if (m_graphics.CanReportMemoryUsage()) {
 		m_total_used_memory = m_graphics.GetDeviceMemoryUsage();
 	}
-	if (m_total_used_memory < m_trigger_gc_memory || m_fault_readback->Active()) {
+	if (m_total_used_memory < m_trigger_gc_memory) {
 		return;
 	}
 
@@ -1291,7 +934,6 @@ void BufferCache::RunGarbageCollector() {
 		if (!m_memory_tracker.IsRegionGpuModified(retire.address, retire.size)) {
 			m_memory_tracker.UntrackMemory(retire.address, retire.size);
 		}
-		m_image_invalidated_ranges.Subtract(retire.address, retire.size);
 		if (retire.size > m_total_used_memory) {
 			EXIT("BufferCache: allocation accounting underflow\n");
 		}

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

@@ -10,7 +10,6 @@
 
 #include <map>
 #include <memory>
-#include <mutex>
 #include <span>
 #include <utility>
 #include <vector>
@@ -21,7 +20,6 @@ struct GraphicContext;
 class CommandBuffer;
 class CommandScheduler;
 class TextureCache;
-class ResourceMutex;
 
 struct BufferBinding {
 	std::shared_ptr<void> owner;
@@ -30,9 +28,8 @@ struct BufferBinding {
 };
 
 struct ImageBufferSource {
-	Buffer*  buffer    = nullptr;
-	uint64_t offset    = 0;
-	bool     gpu_owned = false;
+	Buffer*  buffer = nullptr;
+	uint64_t offset = 0;
 };
 
 class BufferCache {
@@ -43,15 +40,13 @@ public:
 	}
 
 	BufferCache(GraphicContext& graphics, CommandScheduler& scheduler, PageManager& page_manager,
-	            TextureCache& texture_cache, ResourceMutex& resource_mutex);
+	            TextureCache& texture_cache);
 	~BufferCache();
 	KYTY_CLASS_NO_COPY(BufferCache);
 
-	[[nodiscard]] bool InvalidateMemory(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-	                                    PageFaultPhase phase) noexcept;
-	void               InvalidateMemory(uint64_t vaddr, uint64_t size);
-	void               ReadMemory(uint64_t vaddr, uint64_t size);
-	void               UnmapMemory(uint64_t vaddr, uint64_t size);
+	void                        InvalidateMemory(uint64_t vaddr, uint64_t size);
+	void                        ReadMemory(uint64_t vaddr, uint64_t size, bool is_write = false);
+	void                        UnmapMemory(uint64_t vaddr, uint64_t size);
 	[[nodiscard]] BufferBinding ObtainBuffer(CommandBuffer& command, uint64_t vaddr, uint64_t size,
 	                                         bool is_written = false, bool is_read = true,
 	                                         bool is_formatted = false);
@@ -62,23 +57,14 @@ public:
 	                                            uint64_t alignment);
 	[[nodiscard]] std::shared_ptr<Buffer> ObtainNullBuffer();
 	[[nodiscard]] ImageBufferSource       ObtainBufferForImage(uint64_t vaddr, uint64_t size);
-	[[nodiscard]] std::pair<std::shared_ptr<Buffer>, uint64_t>
-	     ObtainBufferForImageWrite(uint64_t vaddr, uint64_t size);
-	void DiscardGpuDirtyBytes(uint64_t vaddr, uint64_t size);
 	void FillBuffer(uint64_t vaddr, uint64_t size, uint32_t value, bool is_gds = false);
 	void CopyBuffer(uint64_t dst_vaddr, uint64_t src_vaddr, uint64_t size, bool dst_gds = false,
 	                bool src_gds = false);
-	[[nodiscard]] bool HasPageOverlap(uint64_t vaddr, uint64_t size);
+	[[nodiscard]] bool IsRegionRegistered(uint64_t vaddr, uint64_t size);
 	[[nodiscard]] bool HasGpuDirtyBytes(uint64_t vaddr, uint64_t size);
 	[[nodiscard]] bool IsRegionCpuModified(uint64_t vaddr, uint64_t size);
 	[[nodiscard]] bool IsRegionGpuModified(uint64_t vaddr, uint64_t size);
-	void               InvalidateImageAliases(uint64_t vaddr, uint64_t size);
-	void               BeginBackingPublication(uint64_t vaddr, uint64_t size, uint64_t tick);
-	void               CompleteBackingPublication(uint64_t vaddr, uint64_t size, uint64_t tick);
-	[[nodiscard]] bool SynchronizeBacking(uint64_t vaddr, uint64_t size);
-	void               PublishImageBuffer(uint64_t vaddr, uint64_t size);
-	void ValidateGpuAccess(uint64_t vaddr, uint64_t size, bool is_read, bool is_written) const;
-	void RunGarbageCollector();
+	void               RunGarbageCollector();
 
 private:
 	friend struct BufferCacheTestAccess;
@@ -91,29 +77,24 @@ private:
 	struct DownloadCopy;
 	struct DownloadRange;
 	struct RetiredBuffer;
-	struct FaultReadback;
-	struct PendingBackingPublication;
 	static constexpr uint64_t               DOWNLOAD_ALIGNMENT = 64;
 	[[nodiscard]] static uint64_t           AlignDown(uint64_t value) noexcept;
 	[[nodiscard]] static uint64_t           AlignUp(uint64_t value);
 	[[nodiscard]] static constexpr uint64_t AlignDownload(uint64_t size) noexcept {
 		return (size + DOWNLOAD_ALIGNMENT - 1) & ~(DOWNLOAD_ALIGNMENT - 1);
 	}
-	[[nodiscard]] static bool PageOverlaps(uint64_t left, uint64_t left_size, uint64_t right,
-	                                       uint64_t right_size) noexcept;
 	[[nodiscard]] static std::pair<uint64_t, uint64_t> DownloadEnvelope(const DownloadCopy& copy);
 	[[nodiscard]] static bool ResolveOverlap(CacheRange& merged, CacheRange candidate) noexcept;
 	void Upload(CommandBuffer& command, Buffer& destination, uint64_t destination_offset,
 	            const void* source, uint64_t size);
 	[[nodiscard]] CachedBuffer& GetOrCreateBuffer(CommandBuffer& command, uint64_t vaddr,
 	                                              uint64_t size);
+	[[nodiscard]] bool SynchronizeBufferFromImage(Buffer& buffer, uint64_t vaddr, uint64_t size);
 	[[nodiscard]] std::vector<DownloadRange> RecordDownloads(std::span<const DownloadCopy> copies);
 	void PublishDownloads(std::span<const DownloadRange> downloads);
 	void QueueGarbageDownload(std::span<const DownloadCopy> copies, RetiredBuffer retire);
-	void RefreshInvalidatedRanges(CommandBuffer& command, CachedBuffer& cached, uint64_t vaddr,
-	                              uint64_t size, bool upload);
-	void DiscardGpuDirtyBytesLocked(uint64_t vaddr, uint64_t size, const char* operation);
 	void WriteHostMemory(uint64_t vaddr, std::span<const uint8_t> data);
+	void ReadMemoryOnGpu(uint64_t vaddr, uint64_t size, bool is_write);
 
 	GraphicContext&                                   m_graphics;
 	CommandScheduler&                                 m_scheduler;
@@ -121,19 +102,13 @@ private:
 	Common::Mutex                                     m_mutex;
 	std::shared_ptr<Buffer>                           m_null_buffer;
 	std::map<uint64_t, std::unique_ptr<CachedBuffer>> m_buffers;
-	std::unique_ptr<FaultReadback>                    m_fault_readback;
 	RangeSet                                          m_gpu_modified_ranges;
-	RangeSet                                          m_image_invalidated_ranges;
-	std::mutex                                        m_publication_mutex;
-	std::vector<PendingBackingPublication>            m_pending_backing_publications;
 	MemoryTracker                                     m_memory_tracker;
 	StreamBuffer                                      m_staging_buffer;
 	StreamBuffer                                      m_stream_buffer;
 	StreamBuffer                                      m_download_buffer;
 	StreamBuffer                                      m_device_buffer;
-	PageManager&                                      m_page_manager;
 	TextureCache&                                     m_texture_cache;
-	ResourceMutex&                                    m_resource_mutex;
 	uint64_t                                          m_total_used_memory = 0;
 	uint64_t m_trigger_gc_memory  = 1ull * 1024 * 1024 * 1024;
 	uint64_t m_critical_gc_memory = 2ull * 1024 * 1024 * 1024;

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

@@ -1,108 +1,36 @@
 #include "graphics/host_gpu/renderer/cache/gpuResourceManager.h"
 
 #include "common/assert.h"
-#include "graphics/guest_gpu/command_processor/commandProcessor.h"
 #include "graphics/guest_gpu/graphicsRun.h"
 #include "graphics/host_gpu/renderer/commandScheduler.h"
-
 namespace Libs::Graphics {
 
 GpuResourceManager::GpuResourceManager(GraphicContext& graphics, CommandScheduler& scheduler)
-    : m_page_manager(FaultThunk, this),
-      m_buffer_cache(graphics, scheduler, m_page_manager, m_texture_cache, m_resource_mutex),
-      m_texture_cache(graphics, scheduler, m_page_manager, m_buffer_cache, m_resource_mutex) {}
+    : 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::FaultThunk(void* context, PageFaultAccess access, uint64_t vaddr,
-                                    uint64_t size, PageFaultPhase phase) noexcept {
-	return static_cast<GpuResourceManager*>(context)->InvalidateMemory(access, vaddr, size, phase);
-}
-
-bool GpuResourceManager::InvalidateMemory(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-                                          PageFaultPhase phase) noexcept {
-	// Let the authoritative image materialize first. A clean overlapping buffer marks a write
-	// fault CPU-dirty when it begins ownership transfer; doing that before image preflight would
-	// make the image appear to race a real CPU write. Completion and release retain buffer-first
-	// ordering so its pending fault is gone before TextureCache publishes the downloaded backing.
-	if (phase == PageFaultPhase::Invalidate) {
-		const bool image_handled  = m_texture_cache.InvalidateMemory(access, vaddr, size, phase);
-		const bool buffer_handled = m_buffer_cache.InvalidateMemory(access, vaddr, size, phase);
-		return buffer_handled || image_handled;
-	}
-	const bool buffer_handled = m_buffer_cache.InvalidateMemory(access, vaddr, size, phase);
-	const bool image_handled  = m_texture_cache.InvalidateMemory(access, vaddr, size, phase);
-	return buffer_handled || image_handled;
-}
-
 bool GpuResourceManager::HandleFault(PageFaultAccess access, uint64_t fault_vaddr) noexcept {
 	constexpr uint64_t fault_size = 8;
 	if (!IsMapped(fault_vaddr, fault_size)) {
 		return false;
 	}
-	if (CommandScheduler::InDeferredOperation()) {
-		EXIT("unsupported guest-memory fault from an asynchronous GPU completion, "
-		     "addr=0x%016" PRIx64 " access=%u\n",
-		     fault_vaddr, static_cast<uint32_t>(access));
+	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);
 	}
-	bool       handled = false;
-	const auto resolve = [this, access, fault_vaddr, &handled](CommandProcessor& cp) {
-		cp.BeginReadbackTransaction();
-		{
-			ResourceMutex::FaultScope fault(m_resource_mutex);
-			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);
-			}
-			handled = true;
-		}
-		cp.EndReadbackTransaction();
-	};
-	if (auto* cp = Gpu::CurrentCommandProcessor(); cp != nullptr) {
-		resolve(*cp);
-		return handled;
-	}
-	if (m_resource_mutex.IsOwnedByCurrentThread()) {
-		EXIT("unsupported page fault from a pre-owned resource transaction, addr=0x%016" PRIx64
-		     " access=%u\n",
-		     fault_vaddr, static_cast<uint32_t>(access));
-	}
-	EXIT_IF(m_gpu == nullptr);
-	m_gpu->SendCommandSyncWithProcessor(resolve);
-	return handled;
+	return true;
 }
 
 bool GpuResourceManager::InvalidateMemory(uint64_t vaddr, uint64_t size) {
 	if (!IsMapped(vaddr, size)) {
 		return false;
 	}
-	if (CommandScheduler::InDeferredOperation()) {
-		EXIT("unsupported memory invalidation from an asynchronous GPU completion, "
-		     "addr=0x%016" PRIx64 " size=0x%016" PRIx64 "\n",
-		     vaddr, size);
-	}
-	const auto resolve = [this, vaddr, size](CommandProcessor& cp) {
-		cp.BeginReadbackTransaction();
-		{
-			ResourceMutex::FaultScope fault(m_resource_mutex);
-			m_buffer_cache.InvalidateMemory(vaddr, size);
-			m_texture_cache.InvalidateMemory(vaddr, size);
-		}
-		cp.EndReadbackTransaction();
-	};
-	if (auto* cp = Gpu::CurrentCommandProcessor(); cp != nullptr) {
-		resolve(*cp);
-		return true;
-	}
-	if (m_resource_mutex.IsOwnedByCurrentThread()) {
-		EXIT("unsupported memory invalidation from a pre-owned resource transaction, "
-		     "addr=0x%016" PRIx64 " size=0x%016" PRIx64 "\n",
-		     vaddr, size);
-	}
-	EXIT_IF(m_gpu == nullptr);
-	m_gpu->SendCommandSyncWithProcessor(resolve);
+	m_buffer_cache.InvalidateMemory(vaddr, size);
+	m_texture_cache.InvalidateMemory(vaddr, size);
 	return true;
 }
 
@@ -115,33 +43,36 @@ bool GpuResourceManager::IsMapped(uint64_t vaddr, uint64_t size) const noexcept
 	return m_mapped_ranges.Contains(vaddr, size);
 }
 
-void GpuResourceManager::MapMemory(uint64_t vaddr, uint64_t size, GpuAccess access) {
+void GpuResourceManager::MapMemory(uint64_t vaddr, uint64_t size) {
 	{
 		std::lock_guard lock(m_mapped_ranges_mutex);
 		m_mapped_ranges.Add(vaddr, size);
 	}
-	m_page_manager.OnGpuMap(vaddr, size, access);
+	m_page_manager.OnGpuMap(vaddr, size);
 }
 
-void GpuResourceManager::UnmapMemory(uint64_t vaddr, uint64_t size, GpuAccess access) {
-	if (!IsMapped(vaddr, size)) {
-		EXIT("cannot unmap an unmapped GPU resource range\n");
+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, access] {
-		m_texture_cache.UnmapMemory(vaddr, size);
+	const auto unmap = [this, vaddr, size] {
+		if (m_scheduler.Active()) {
+			const auto tick = m_scheduler.CurrentTick();
+			m_scheduler.FinishCurrent();
+			m_scheduler.WaitPriorityOperations(tick);
+		}
 		m_buffer_cache.UnmapMemory(vaddr, size);
-		m_page_manager.OnGpuUnmap(vaddr, size, access);
+		m_texture_cache.UnmapMemory(vaddr, size);
+		m_page_manager.OnGpuUnmap(vaddr, size);
 		std::lock_guard lock(m_mapped_ranges_mutex);
 		m_mapped_ranges.Subtract(vaddr, size);
 	};
 	if (m_gpu == nullptr) {
-		if (m_resource_mutex.IsOwnedByCurrentThread()) {
-			EXIT("cannot synchronously unmap from a resource transaction\n");
-		}
 		unmap();
 		return;
 	}
-	Gpu::SubmissionLock submissions(*m_gpu);
 	m_gpu->SendCommandSync(unmap);
 }
 

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

@@ -5,7 +5,6 @@
 #include "common/common.h"
 #include "graphics/host_gpu/pageManager.h"
 #include "graphics/host_gpu/renderer/cache/bufferCache.h"
-#include "graphics/host_gpu/renderer/cache/resourceMutex.h"
 #include "graphics/host_gpu/renderer/cache/textureCache.h"
 
 #include <cstdint>
@@ -29,18 +28,13 @@ public:
 	[[nodiscard]] bool HandleFault(PageFaultAccess access, uint64_t fault_vaddr) noexcept;
 	[[nodiscard]] bool InvalidateMemory(uint64_t vaddr, uint64_t size);
 	[[nodiscard]] bool IsMapped(uint64_t vaddr, uint64_t size) const noexcept;
-	void               MapMemory(uint64_t vaddr, uint64_t size, GpuAccess access);
-	void               UnmapMemory(uint64_t vaddr, uint64_t size, GpuAccess access);
+	void               MapMemory(uint64_t vaddr, uint64_t size);
+	void               UnmapMemory(uint64_t vaddr, uint64_t size);
 	void               RunGarbageCollector();
 
 private:
-	static bool FaultThunk(void* context, PageFaultAccess access, uint64_t vaddr, uint64_t size,
-	                       PageFaultPhase phase) noexcept;
-	[[nodiscard]] bool InvalidateMemory(PageFaultAccess access, uint64_t vaddr, uint64_t size,
-	                                    PageFaultPhase phase) noexcept;
-
 	PageManager               m_page_manager;
-	ResourceMutex             m_resource_mutex;
+	CommandScheduler&         m_scheduler;
 	BufferCache               m_buffer_cache;
 	TextureCache              m_texture_cache;
 	mutable std::shared_mutex m_mapped_ranges_mutex;

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

@@ -7,8 +7,9 @@
 #include <array>
 #include <cstddef>
 #include <cstdint>
-#include <list>
+#include <cstring>
 #include <memory>
+#include <type_traits>
 #include <utility>
 #include <vector>
 
@@ -103,6 +104,158 @@ private:
 	size_t                               m_allocated_buckets = 0;
 };
 
+// Inline owner storage for page-table entries. Texture pages normally have only a
+// handful of owners, so avoid a heap allocation on the first 16 registrations.
+//
+// Owners are packed into live uint64_t objects instead of being constructed in
+// raw storage. This keeps sparse bucket construction cheap without relying on
+// subtle implicit-lifetime or pointer-provenance rules.
+template <typename OwnerT, size_t InlineCapacity = 16>
+class InlinePageOwnerList final {
+public:
+	using value_type = OwnerT;
+	using size_type  = size_t;
+
+	class const_iterator final {
+	public:
+		[[nodiscard]] OwnerT operator*() const noexcept { return m_owners->At(m_index); }
+		const_iterator&      operator++() noexcept {
+			++m_index;
+			return *this;
+		}
+		bool operator==(const const_iterator&) const = default;
+
+	private:
+		friend class InlinePageOwnerList;
+		const_iterator(const InlinePageOwnerList* owners, size_type index) noexcept
+		    : m_owners(owners), m_index(index) {}
+
+		const InlinePageOwnerList* m_owners = nullptr;
+		size_type                  m_index  = 0;
+	};
+
+	static_assert(InlineCapacity > 0);
+	static_assert(std::is_default_constructible_v<OwnerT>);
+	static_assert(std::is_trivially_copyable_v<OwnerT>);
+	static_assert(std::has_unique_object_representations_v<OwnerT>);
+	static_assert(sizeof(OwnerT) <= sizeof(uint64_t));
+
+	InlinePageOwnerList() noexcept {}
+	InlinePageOwnerList(const InlinePageOwnerList&)            = delete;
+	InlinePageOwnerList& operator=(const InlinePageOwnerList&) = delete;
+	InlinePageOwnerList(InlinePageOwnerList&& other) noexcept
+	    : m_overflow(std::move(other.m_overflow)),
+	      m_inline_size(std::exchange(other.m_inline_size, 0)) {
+		if (m_overflow == nullptr) {
+			std::copy_n(other.m_inline.begin(), m_inline_size, m_inline.begin());
+		}
+	}
+	InlinePageOwnerList& operator=(InlinePageOwnerList&& other) noexcept {
+		if (this != &other) {
+			m_overflow    = std::move(other.m_overflow);
+			m_inline_size = std::exchange(other.m_inline_size, 0);
+			if (m_overflow == nullptr) {
+				std::copy_n(other.m_inline.begin(), m_inline_size, m_inline.begin());
+			}
+		}
+		return *this;
+	}
+
+	[[nodiscard]] const_iterator begin() const noexcept { return {this, 0}; }
+	[[nodiscard]] const_iterator end() const noexcept { return {this, size()}; }
+
+	[[nodiscard]] bool      empty() const noexcept { return size() == 0; }
+	[[nodiscard]] size_type size() const noexcept {
+		return m_overflow == nullptr ? m_inline_size : m_overflow->size();
+	}
+	[[nodiscard]] OwnerT front() const noexcept { return At(0); }
+	[[nodiscard]] OwnerT operator[](size_type index) const noexcept { return At(index); }
+
+	void push_back(const OwnerT& owner) {
+		const uint64_t packed = Pack(owner);
+		if (m_overflow != nullptr) {
+			m_overflow->push_back(packed);
+			return;
+		}
+		if (m_inline_size < InlineCapacity) {
+			m_inline[m_inline_size] = packed;
+			++m_inline_size;
+			return;
+		}
+		auto overflow = std::make_unique<std::vector<uint64_t>>();
+		overflow->reserve(InlineCapacity * 2);
+		overflow->assign(m_inline.begin(), m_inline.end());
+		overflow->push_back(packed);
+		m_overflow = std::move(overflow);
+	}
+
+	[[nodiscard]] bool Contains(const OwnerT& owner) const noexcept {
+		const uint64_t packed = Pack(owner);
+		if (m_overflow != nullptr) {
+			return std::find(m_overflow->begin(), m_overflow->end(), packed) != m_overflow->end();
+		}
+		return std::find(m_inline.begin(), m_inline.begin() + static_cast<ptrdiff_t>(m_inline_size),
+		                 packed) != m_inline.begin() + static_cast<ptrdiff_t>(m_inline_size);
+	}
+
+	[[nodiscard]] bool Erase(const OwnerT& owner) noexcept {
+		const uint64_t packed = Pack(owner);
+		if (m_overflow != nullptr) {
+			const auto found = std::find(m_overflow->begin(), m_overflow->end(), packed);
+			if (found == m_overflow->end()) {
+				return false;
+			}
+			m_overflow->erase(found);
+			if (m_overflow->size() == InlineCapacity) {
+				std::copy(m_overflow->begin(), m_overflow->end(), m_inline.begin());
+				m_inline_size = InlineCapacity;
+				m_overflow.reset();
+			}
+			return true;
+		}
+		const auto end   = m_inline.begin() + static_cast<ptrdiff_t>(m_inline_size);
+		const auto found = std::find(m_inline.begin(), end, packed);
+		if (found == end) {
+			return false;
+		}
+		std::move(found + 1, end, found);
+		--m_inline_size;
+		return true;
+	}
+
+	template <typename Func>
+	void ForEach(Func&& func) const {
+		if (m_overflow != nullptr) {
+			for (const uint64_t owner: *m_overflow) {
+				func(Unpack(owner));
+			}
+			return;
+		}
+		for (size_type index = 0; index < m_inline_size; ++index) {
+			func(Unpack(m_inline[index]));
+		}
+	}
+
+private:
+	[[nodiscard]] OwnerT At(size_type index) const noexcept {
+		return Unpack(m_overflow == nullptr ? m_inline[index] : (*m_overflow)[index]);
+	}
+	[[nodiscard]] static uint64_t Pack(const OwnerT& owner) noexcept {
+		uint64_t packed = 0;
+		std::memcpy(&packed, &owner, sizeof(owner));
+		return packed;
+	}
+	[[nodiscard]] static OwnerT Unpack(uint64_t packed) noexcept {
+		OwnerT owner {};
+		std::memcpy(&owner, &packed, sizeof(owner));
+		return owner;
+	}
+
+	std::array<uint64_t, InlineCapacity>   m_inline;
+	std::unique_ptr<std::vector<uint64_t>> m_overflow;
+	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) {
@@ -114,247 +267,6 @@ template <typename Container, typename Value>
 	return true;
 }
 
-// Multi-range ownership with 1 MiB candidate buckets and precise 4 KiB
-// lifetime accounting. OwnerT only needs equality.
-template <typename OwnerT>
-class MultiRangePageOwnerIndex final {
-private:
-	struct Registration;
-	using MembershipList = std::vector<const Registration*>;
-
-public:
-	struct ByteRange final {
-		uint64_t address = 0;
-		uint64_t size    = 0;
-	};
-
-	using CoarseTable   = MultiLevelPageTable<MembershipList, 20, 40, 10>;
-	using TrackingTable = MultiLevelPageTable<MembershipList, 12, 40, 18>;
-
-	[[nodiscard]] bool Register(const OwnerT& owner, const std::vector<ByteRange>& ranges) {
-		if (FindRegistration(owner) != m_registrations.end()) {
-			return false;
-		}
-		auto normalized = Normalize(ranges);
-		if (normalized.empty()) {
-			return false;
-		}
-		m_registrations.push_back({owner, std::move(normalized)});
-		const Registration* registration = &m_registrations.back();
-		for (const size_t page: CollectPages<20>(registration->ranges)) {
-			m_coarse_pages[page].push_back(registration);
-		}
-		for (const size_t page: CollectPages<12>(registration->ranges)) {
-			m_tracking_pages[page].push_back(registration);
-		}
-		return true;
-	}
-
-	// No state changes if an expected membership is absent. Final releases are
-	// sorted and coalesced 4 KiB spans whose final owner disappeared.
-	[[nodiscard]] bool Unregister(const OwnerT& owner, std::vector<ByteRange>& final_releases) {
-		final_releases.clear();
-		auto registration = FindRegistration(owner);
-		if (registration == m_registrations.end()) {
-			return false;
-		}
-		const auto          coarse_pages     = CollectPages<20>(registration->ranges);
-		const auto          tracking_pages   = CollectPages<12>(registration->ranges);
-		const Registration* registration_ptr = &*registration;
-		if (!HasAllMemberships(m_coarse_pages, coarse_pages, registration_ptr) ||
-		    !HasAllMemberships(m_tracking_pages, tracking_pages, registration_ptr)) {
-			return false;
-		}
-		std::vector<size_t> final_pages;
-		for (const size_t page: coarse_pages) {
-			(void)EraseExact(*m_coarse_pages.Find(page), registration_ptr);
-		}
-		for (const size_t page: tracking_pages) {
-			auto& owners = *m_tracking_pages.Find(page);
-			if (owners.size() == 1) {
-				final_pages.push_back(page);
-			}
-			(void)EraseExact(owners, registration_ptr);
-		}
-		m_registrations.erase(registration);
-		final_releases = CoalesceTrackingPages(final_pages);
-		return true;
-	}
-
-	[[nodiscard]] std::vector<OwnerT> Query(uint64_t address, uint64_t size) const {
-		return Query(address, size, [](const OwnerT&) { return true; });
-	}
-
-	template <typename Predicate>
-	[[nodiscard]] std::vector<OwnerT> Query(uint64_t address, uint64_t size,
-	                                        Predicate&& predicate) const {
-		return QueryImpl(address, size, true, std::forward<Predicate>(predicate));
-	}
-
-	// Fault paths use page candidates: exact byte-disjoint owners sharing a
-	// touched 4 KiB page are intentionally retained.
-	[[nodiscard]] std::vector<OwnerT> QueryCandidates(uint64_t address, uint64_t size) const {
-		return QueryCandidates(address, size, [](const OwnerT&) { return true; });
-	}
-
-	template <typename Predicate>
-	[[nodiscard]] std::vector<OwnerT> QueryCandidates(uint64_t address, uint64_t size,
-	                                                  Predicate&& predicate) const {
-		return QueryImpl(address, size, false, std::forward<Predicate>(predicate));
-	}
-
-	[[nodiscard]] size_t CoarseMembershipCount(size_t page) const {
-		const auto* owners = m_coarse_pages.Find(page);
-		return owners == nullptr ? 0 : owners->size();
-	}
-	[[nodiscard]] size_t TrackingMembershipCount(size_t page) const {
-		const auto* owners = m_tracking_pages.Find(page);
-		return owners == nullptr ? 0 : owners->size();
-	}
-
-private:
-	template <typename Predicate>
-	[[nodiscard]] std::vector<OwnerT> QueryImpl(uint64_t address, uint64_t size, bool strict_bytes,
-	                                            Predicate&& predicate) const {
-		typename CoarseTable::PageRange   coarse_range {};
-		typename TrackingTable::PageRange tracking_range {};
-		if (!CoarseTable::TryGetPageRange(address, size, coarse_range) ||
-		    !TrackingTable::TryGetPageRange(address, size, tracking_range)) {
-			return {};
-		}
-		MembershipList candidates;
-		for (size_t page = coarse_range.first; page < coarse_range.last_exclusive; ++page) {
-			if (const auto* owners = m_coarse_pages.Find(page); owners != nullptr) {
-				AppendUnique(candidates, *owners);
-			}
-		}
-		std::vector<OwnerT> result;
-		for (const Registration* registration: candidates) {
-			if ((!strict_bytes || Overlaps(registration->ranges, address, size)) &&
-			    HasTrackingMembership(registration, tracking_range) &&
-			    predicate(registration->owner)) {
-				result.push_back(registration->owner);
-			}
-		}
-		return result;
-	}
-
-	struct Registration final {
-		OwnerT                 owner;
-		std::vector<ByteRange> ranges;
-	};
-	using RegistrationIterator      = typename std::list<Registration>::iterator;
-	using ConstRegistrationIterator = typename std::list<Registration>::const_iterator;
-
-	[[nodiscard]] RegistrationIterator FindRegistration(const OwnerT& owner) {
-		return std::find_if(m_registrations.begin(), m_registrations.end(),
-		                    [&](const Registration& item) { return item.owner == owner; });
-	}
-	[[nodiscard]] ConstRegistrationIterator FindRegistration(const OwnerT& owner) const {
-		return std::find_if(m_registrations.begin(), m_registrations.end(),
-		                    [&](const Registration& item) { return item.owner == owner; });
-	}
-
-	[[nodiscard]] static std::vector<ByteRange> Normalize(const std::vector<ByteRange>& ranges) {
-		std::vector<ByteRange> sorted;
-		for (const auto& range: ranges) {
-			typename CoarseTable::PageRange ignored {};
-			if (!CoarseTable::TryGetPageRange(range.address, range.size, ignored)) {
-				return {};
-			}
-			sorted.push_back(range);
-		}
-		std::sort(sorted.begin(), sorted.end(), [](const ByteRange& lhs, const ByteRange& rhs) {
-			return lhs.address < rhs.address;
-		});
-		std::vector<ByteRange> merged;
-		for (const auto& range: sorted) {
-			if (merged.empty() || range.address > merged.back().address + merged.back().size) {
-				merged.push_back(range);
-			} else {
-				const uint64_t end = std::max(merged.back().address + merged.back().size,
-				                              range.address + range.size);
-				merged.back().size = end - merged.back().address;
-			}
-		}
-		return merged;
-	}
-
-	template <size_t Bits>
-	[[nodiscard]] static std::vector<size_t> CollectPages(const std::vector<ByteRange>& ranges) {
-		std::vector<size_t> pages;
-		for (const auto& range: ranges) {
-			const size_t first = static_cast<size_t>(range.address >> Bits);
-			const size_t last  = static_cast<size_t>((range.address + range.size - 1) >> Bits);
-			for (size_t page = first; page <= last; ++page) {
-				pages.push_back(page);
-			}
-		}
-		std::sort(pages.begin(), pages.end());
-		pages.erase(std::unique(pages.begin(), pages.end()), pages.end());
-		return pages;
-	}
-
-	template <typename Table>
-	[[nodiscard]] static bool HasAllMemberships(const Table&               table,
-	                                            const std::vector<size_t>& pages,
-	                                            const Registration*        registration) {
-		for (const size_t page: pages) {
-			const auto* owners = table.Find(page);
-			if (owners == nullptr ||
-			    std::find(owners->begin(), owners->end(), registration) == owners->end()) {
-				return false;
-			}
-		}
-		return true;
-	}
-
-	[[nodiscard]] bool HasTrackingMembership(const Registration*                      registration,
-	                                         const typename TrackingTable::PageRange& range) const {
-		for (size_t page = range.first; page < range.last_exclusive; ++page) {
-			const auto* owners = m_tracking_pages.Find(page);
-			if (owners != nullptr &&
-			    std::find(owners->begin(), owners->end(), registration) != owners->end()) {
-				return true;
-			}
-		}
-		return false;
-	}
-
-	[[nodiscard]] static bool Overlaps(const std::vector<ByteRange>& ranges, uint64_t address,
-	                                   uint64_t size) {
-		const uint64_t end = address + size;
-		return std::any_of(ranges.begin(), ranges.end(), [&](const ByteRange& range) {
-			return range.address < end && address < range.address + range.size;
-		});
-	}
-	static void AppendUnique(MembershipList& destination, const MembershipList& source) {
-		for (const Registration* registration: source) {
-			if (std::find(destination.begin(), destination.end(), registration) ==
-			    destination.end()) {
-				destination.push_back(registration);
-			}
-		}
-	}
-	[[nodiscard]] static std::vector<ByteRange>
-	CoalesceTrackingPages(const std::vector<size_t>& pages) {
-		std::vector<ByteRange> result;
-		for (const size_t page: pages) {
-			const uint64_t address = static_cast<uint64_t>(page) << 12;
-			if (!result.empty() && result.back().address + result.back().size == address) {
-				result.back().size += uint64_t {1} << 12;
-			} else {
-				result.push_back({address, uint64_t {1} << 12});
-			}
-		}
-		return result;
-	}
-
-	CoarseTable             m_coarse_pages;
-	TrackingTable           m_tracking_pages;
-	std::list<Registration> m_registrations;
-};
-
 } // namespace Libs::Graphics
 
 #endif // EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_MULTILEVELPAGETABLE_H_

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

@@ -1,93 +0,0 @@
-#include "graphics/host_gpu/renderer/cache/resourceMutex.h"
-
-#include "common/assert.h"
-
-namespace Libs::Graphics {
-
-ResourceMutex::FaultScope::FaultScope(ResourceMutex& mutex): m_mutex(mutex) {
-	m_resource_preowned = m_mutex.BeginFault();
-}
-
-ResourceMutex::FaultScope::~FaultScope() {
-	m_mutex.EndFault(m_resource_preowned);
-}
-
-ResourceMutex::~ResourceMutex() {
-	std::lock_guard state(m_state);
-	if (m_resource_owner != std::thread::id {} || m_fault_owner != std::thread::id {}) {
-		EXIT("ResourceMutex destroyed with an active owner or fault transaction\n");
-	}
-}
-
-void ResourceMutex::lock() {
-	const auto current = std::this_thread::get_id();
-	{
-		std::lock_guard state(m_state);
-		if (m_resource_owner == current) {
-			EXIT("recursive resource transaction\n");
-		}
-		if (m_fault_owner == current) {
-			EXIT("resource transaction re-entered from a page-fault callback\n");
-		}
-	}
-	m_resource.lock();
-	std::lock_guard state(m_state);
-	if (m_resource_owner != std::thread::id {}) {
-		EXIT("resource mutex acquired with a stale owner\n");
-	}
-	m_resource_owner = current;
-}
-
-void ResourceMutex::unlock() {
-	const auto current = std::this_thread::get_id();
-	{
-		std::lock_guard state(m_state);
-		if (m_resource_owner != current) {
-			EXIT("resource transaction released without ownership\n");
-		}
-		m_resource_owner = {};
-	}
-	m_resource.unlock();
-}
-
-bool ResourceMutex::IsOwnedByCurrentThread() {
-	std::lock_guard state(m_state);
-	return m_resource_owner == std::this_thread::get_id();
-}
-
-bool ResourceMutex::BeginFault() {
-	const auto current = std::this_thread::get_id();
-	{
-		std::lock_guard state(m_state);
-		if (m_fault_owner == current) {
-			EXIT("nested resource page-fault transaction\n");
-		}
-		if (m_resource_owner == current) {
-			m_fault_owner = current;
-			return true;
-		}
-	}
-	lock();
-	std::lock_guard state(m_state);
-	if (m_resource_owner != current || m_fault_owner != std::thread::id {}) {
-		EXIT("resource fault transaction acquired inconsistent ownership\n");
-	}
-	m_fault_owner = current;
-	return false;
-}
-
-void ResourceMutex::EndFault(bool resource_preowned) {
-	const auto current = std::this_thread::get_id();
-	{
-		std::lock_guard state(m_state);
-		if (m_fault_owner != current || m_resource_owner != current) {
-			EXIT("resource page-fault transaction ended with inconsistent ownership\n");
-		}
-		m_fault_owner = {};
-	}
-	if (!resource_preowned) {
-		unlock();
-	}
-}
-
-} // namespace Libs::Graphics

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

@@ -1,48 +0,0 @@
-#ifndef EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_RESOURCEMUTEX_H_
-#define EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_RESOURCEMUTEX_H_
-
-#include "common/common.h"
-
-#include <mutex>
-#include <thread>
-
-namespace Libs::Graphics {
-
-// Owner-tracked shared buffer/image transaction. External faults pause GPU submissions first;
-// command-processor faults drain pending guest processors before entering this transaction.
-class ResourceMutex final {
-public:
-	class FaultScope final {
-	public:
-		explicit FaultScope(ResourceMutex& mutex);
-		~FaultScope();
-		KYTY_CLASS_NO_COPY(FaultScope);
-
-	private:
-		ResourceMutex& m_mutex;
-		bool           m_resource_preowned = false;
-	};
-
-	ResourceMutex() = default;
-	~ResourceMutex();
-	KYTY_CLASS_NO_COPY(ResourceMutex);
-
-	void               lock();
-	void               unlock();
-	[[nodiscard]] bool IsOwnedByCurrentThread();
-
-private:
-	friend class FaultScope;
-
-	[[nodiscard]] bool BeginFault();
-	void               EndFault(bool resource_preowned);
-
-	std::mutex      m_resource;
-	std::mutex      m_state;
-	std::thread::id m_resource_owner;
-	std::thread::id m_fault_owner;
-};
-
-} // namespace Libs::Graphics
-
-#endif // EMULATOR_SRC_GRAPHICS_HOST_GPU_RENDERER_RESOURCEMUTEX_H_

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

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

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

@@ -135,8 +135,8 @@ void Buffer::Write(uint64_t offset, const void* source, uint64_t size) {
 void Buffer::Flush(uint64_t offset, uint64_t size) {
 	EXIT_IF(m_mapped.empty() || offset > m_size || size > m_size - offset);
 	if (!m_is_coherent && size != 0) {
-		const auto result = vmaFlushAllocation(m_graphics->allocator, m_buffer->memory.allocation,
-		                                      offset, size);
+		const auto result =
+		    vmaFlushAllocation(m_graphics->allocator, m_buffer->memory.allocation, offset, size);
 		EXIT_NOT_IMPLEMENTED(static_cast<vk::Result>(result) != vk::Result::eSuccess);
 	}
 }
@@ -144,8 +144,8 @@ void Buffer::Flush(uint64_t offset, uint64_t size) {
 vk::BufferMemoryBarrier Buffer::Barrier(uint64_t offset, uint64_t size, vk::AccessFlags source,
                                         vk::AccessFlags destination) const {
 	if (Handle() == nullptr || size == 0 || offset > m_size || size > m_size - offset) {
-		EXIT("Buffer: invalid DMA barrier, handle=%p offset=0x%016" PRIx64
-		     " size=0x%016" PRIx64 " capacity=0x%016" PRIx64 "\n",
+		EXIT("Buffer: invalid DMA barrier, handle=%p offset=0x%016" PRIx64 " size=0x%016" PRIx64
+		     " capacity=0x%016" PRIx64 "\n",
 		     static_cast<const void*>(Handle()), offset, size, m_size);
 	}
 	vk::BufferMemoryBarrier barrier {};
@@ -175,10 +175,9 @@ void Buffer::CopyFrom(CommandBuffer& command, const Buffer& source, uint64_t sou
 	command.EndRendering();
 	const vk::BufferMemoryBarrier before[] = {
 	    source.Barrier(source_offset, size, source_before, vk::AccessFlagBits::eTransferRead),
-	    Barrier(destination_offset, size, destination_before,
-	            vk::AccessFlagBits::eTransferWrite),
+	    Barrier(destination_offset, size, destination_before, vk::AccessFlagBits::eTransferWrite),
 	};
-	const auto host_access = vk::AccessFlagBits::eHostRead | vk::AccessFlagBits::eHostWrite;
+	const auto host_access  = vk::AccessFlagBits::eHostRead | vk::AccessFlagBits::eHostWrite;
 	auto       before_stage = vk::PipelineStageFlags {vk::PipelineStageFlagBits::eAllCommands};
 	if (static_cast<bool>((source_before | destination_before) & host_access)) {
 		before_stage |= vk::PipelineStageFlagBits::eHost;
@@ -214,9 +213,8 @@ void Buffer::Fill(uint64_t offset, uint64_t size, uint32_t value) {
 	                       vk::PipelineStageFlagBits::eTransfer, vk::DependencyFlagBits::eByRegion,
 	                       0, nullptr, 1, &before, 0, nullptr);
 	native.fillBuffer(Handle(), offset, size, value);
-	const auto after =
-	    Barrier(offset, size, vk::AccessFlagBits::eTransferWrite,
-	            vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite);
+	const auto after = Barrier(offset, size, vk::AccessFlagBits::eTransferWrite,
+	                           vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite);
 	native.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer,
 	                       vk::PipelineStageFlagBits::eAllCommands,
 	                       vk::DependencyFlagBits::eByRegion, 0, nullptr, 1, &after, 0, nullptr);
@@ -250,8 +248,8 @@ std::pair<uint8_t*, uint64_t> StreamBuffer::Map(uint64_t size, uint64_t alignmen
 	if (Mapped().empty()) {
 		return {nullptr, 0};
 	}
-	uint64_t mapped_size = size;
-	const auto atom = Graphics().physical_device_properties.limits.nonCoherentAtomSize;
+	uint64_t   mapped_size = size;
+	const auto atom        = Graphics().physical_device_properties.limits.nonCoherentAtomSize;
 	if (!NormalizeReservation(IsCoherent(), atom, mapped_size, alignment)) {
 		return {nullptr, 0};
 	}

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

@@ -54,16 +54,15 @@ public:
 	[[nodiscard]] bool IsInBounds(uint64_t address, uint64_t size) const noexcept;
 	void               Write(uint64_t offset, const void* source, uint64_t size);
 	void               Flush(uint64_t offset, uint64_t size);
-	void               CopyFrom(
-	                  CommandBuffer& command, const Buffer& source, uint64_t source_offset,
-	                  uint64_t destination_offset, uint64_t size,
-	                  vk::AccessFlags source_before = vk::AccessFlagBits::eMemoryWrite,
-	                  vk::AccessFlags destination_before =
-	                      vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite,
-	                  vk::AccessFlags source_after =
-	                      vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite,
-	                  vk::AccessFlags destination_after =
-	                      vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite);
+	void CopyFrom(CommandBuffer& command, const Buffer& source, uint64_t source_offset,
+	              uint64_t destination_offset, uint64_t size,
+	              vk::AccessFlags source_before      = vk::AccessFlagBits::eMemoryWrite,
+	              vk::AccessFlags destination_before = vk::AccessFlagBits::eMemoryRead |
+	                                                   vk::AccessFlagBits::eMemoryWrite,
+	              vk::AccessFlags source_after       = vk::AccessFlagBits::eMemoryRead |
+	                                                   vk::AccessFlagBits::eMemoryWrite,
+	              vk::AccessFlags destination_after  = vk::AccessFlagBits::eMemoryRead |
+	                                                   vk::AccessFlagBits::eMemoryWrite);
 	void Fill(uint64_t offset, uint64_t size, uint32_t value);
 
 protected:
@@ -107,13 +106,13 @@ private:
 		uint64_t upper_bound = 0;
 	};
 
-	void               ReserveWatches(std::vector<Watch>& watches, size_t grow_size);
+	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,
-	                                         std::optional<size_t>     invalidation_mark,
-	                                         uint64_t requested_upper_bound, bool allow_wait,
-	                                         size_t& wait_cursor, uint64_t& wait_bound);
+	[[nodiscard]] bool        WaitPendingOperations(const std::vector<Watch>& watches,
+	                                                std::optional<size_t>     invalidation_mark,
+	                                                uint64_t requested_upper_bound, bool allow_wait,
+	                                                size_t& wait_cursor, uint64_t& wait_bound);
 
 	uint64_t              m_offset      = 0;
 	uint64_t              m_mapped_size = 0;

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

@@ -8,7 +8,6 @@
 #include "graphics/guest_gpu/tile.h"
 #include "graphics/host_gpu/graphicContext.h"
 #include "graphics/host_gpu/renderer/cache/bufferCache.h"
-#include "graphics/host_gpu/renderer/cache/resourceMutex.h"
 #include "graphics/host_gpu/renderer/commandScheduler.h"
 #include "graphics/host_gpu/renderer/image/imageView.h"
 #include "graphics/host_gpu/renderer/image/textureCommon.h"
@@ -56,13 +55,12 @@ private:
 } // namespace
 
 TextureCache::TextureCache(GraphicContext& graphics, CommandScheduler& scheduler,
-                           PageManager& page_manager, BufferCache& buffer_cache,
-                           ResourceMutex& resource_mutex)
+                           PageManager& page_manager, BufferCache& buffer_cache)
     : m_graphics(graphics), m_scheduler(scheduler), m_page_manager(page_manager),
       m_blit_helper(graphics, scheduler),
       m_tiler(std::make_unique<TileManager>(graphics, scheduler,
                                             buffer_cache.GetUtilityBuffer(MemoryUsage::Stream))),
-      m_buffer_cache(buffer_cache), m_resource_mutex(resource_mutex),
+      m_buffer_cache(buffer_cache),
       m_readback_linear_images(Config::ReadbackLinearImagesEnabled()) {
 	if (m_graphics.CanReportMemoryUsage()) {
 		constexpr int64_t GiB = 1024ll * 1024 * 1024;
@@ -88,15 +86,34 @@ TextureCache::~TextureCache() {
 
 bool TextureCache::SameBacking(const ImageInfo& cached, const ImageInfo& requested,
                                bool exact_format) {
-	const bool unit_extent =
-	    requested.extent.width == 1 && requested.extent.height == 1 && requested.extent.depth == 1;
-	return cached.data == requested.data && cached.extent == requested.extent &&
-	       cached.samples == requested.samples &&
-	       cached.bytes_per_block == requested.bytes_per_block &&
-	       (cached.type == requested.type || unit_extent) &&
-	       (exact_format
-	            ? cached.pixel_format == requested.pixel_format
-	            : ImageViewOps::FormatsCompatible(cached.pixel_format, requested.pixel_format));
+	if (cached.data.address != requested.data.address) {
+		return false;
+	}
+	if (cached.data.size != requested.data.size) {
+		return false;
+	}
+	if (cached.extent != requested.extent) {
+		return false;
+	}
+	if (cached.samples != requested.samples) {
+		return false;
+	}
+	if (cached.bytes_per_block != requested.bytes_per_block) {
+		return false;
+	}
+	if (cached.tile_mode != requested.tile_mode) {
+		return false;
+	}
+	if (!ImageViewOps::FormatsCompatible(cached.pixel_format, requested.pixel_format)) {
+		return false;
+	}
+	if (cached.type != requested.type && requested.extent != vk::Extent3D {1, 1, 1}) {
+		return false;
+	}
+	if (exact_format && cached.pixel_format != requested.pixel_format) {
+		return false;
+	}
+	return true;
 }
 
 TextureCache::BindingType TextureCache::UploadBinding(const Image& image) {
@@ -176,10 +193,12 @@ void TextureCache::RegisterImage(ImageId id) {
 	if (image.registered || image.info.data.Empty()) {
 		EXIT("TextureCache: invalid image registration\n");
 	}
-	std::vector<ImageOwnerIndex::ByteRange> ranges;
-	ranges.push_back({image.info.data.address, image.info.data.size});
-	if (!m_image_owner_index.Register(id, ranges)) {
-		EXIT("TextureCache: duplicate or invalid image registration\n");
+	ImagePageTable::PageRange pages {};
+	if (!ImagePageTable::TryGetPageRange(image.info.data.address, image.info.data.size, pages)) {
+		EXIT("TextureCache: image registration is outside the guest address space\n");
+	}
+	for (size_t page = pages.first; page < pages.last_exclusive; ++page) {
+		m_image_page_table[page].push_back(id);
 	}
 	image.registered = true;
 	image.lru_id     = m_lru_cache.Insert(id, m_gc_tick);
@@ -192,9 +211,15 @@ void TextureCache::UnregisterImage(ImageId id) {
 		return;
 	}
 	UntrackImage(id);
-	std::vector<ImageOwnerIndex::ByteRange> releases;
-	if (!m_image_owner_index.Unregister(id, releases)) {
-		EXIT("TextureCache: image missing from owner index\n");
+	ImagePageTable::PageRange pages {};
+	if (!ImagePageTable::TryGetPageRange(image.info.data.address, image.info.data.size, pages)) {
+		EXIT("TextureCache: registered image is outside the guest address space\n");
+	}
+	for (size_t page = pages.first; page < pages.last_exclusive; ++page) {
+		auto* owners = m_image_page_table.Find(page);
+		if (owners == nullptr || !owners->Erase(id)) {
+			EXIT("TextureCache: image missing from page owner index\n");
+		}
 	}
 	m_lru_cache.Free(image.lru_id);
 	const auto accounted = image.AccountedSize();
@@ -304,7 +329,7 @@ void TextureCache::TrackImage(ImageId id) {
 	if (!image.IsTracked()) {
 		image.track_addr     = image_begin;
 		image.track_addr_end = image_end;
-		m_page_manager.UpdatePageWatchers(true, image_begin, image.info.data.size);
+		m_page_manager.UpdatePageWatchers<true>(image_begin, image.info.data.size);
 		return;
 	}
 	if (image_begin < image.track_addr) {
@@ -329,7 +354,7 @@ void TextureCache::TrackImageHead(ImageId id) {
 	}
 	const auto size  = image.track_addr - image_begin;
 	image.track_addr = image_begin;
-	m_page_manager.UpdatePageWatchers(true, image_begin, size);
+	m_page_manager.UpdatePageWatchers<true>(image_begin, size);
 }
 
 void TextureCache::TrackImageTail(ImageId id) {
@@ -347,7 +372,7 @@ void TextureCache::TrackImageTail(ImageId id) {
 	const auto address   = image.track_addr_end;
 	const auto size      = image_end - address;
 	image.track_addr_end = image_end;
-	m_page_manager.UpdatePageWatchers(true, address, size);
+	m_page_manager.UpdatePageWatchers<true>(address, size);
 }
 
 void TextureCache::UntrackImage(ImageId id) {
@@ -360,7 +385,7 @@ void TextureCache::UntrackImage(ImageId id) {
 	image.track_addr     = 0;
 	image.track_addr_end = 0;
 	if (size != 0) {
-		m_page_manager.UpdatePageWatchers(false, address, size);
+		m_page_manager.UpdatePageWatchers<false>(address, size);
 	}
 }
 
@@ -381,7 +406,7 @@ void TextureCache::UntrackImageHead(ImageId id) {
 		UntrackImage(id);
 	}
 	if (size != 0) {
-		m_page_manager.UpdatePageWatchers(false, begin, size);
+		m_page_manager.UpdatePageWatchers<false>(begin, size);
 	}
 }
 
@@ -402,14 +427,13 @@ void TextureCache::UntrackImageTail(ImageId id) {
 		UntrackImage(id);
 	}
 	if (size != 0) {
-		m_page_manager.UpdatePageWatchers(false, address, size);
+		m_page_manager.UpdatePageWatchers<false>(address, size);
 	}
 }
 
 void TextureCache::TrackImageDownload(ImageId id) {
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
-	auto&           image = ResolveImage(id);
+	CacheLock lock(*this, m_lock);
+	auto&     image = ResolveImage(id);
 	TrackImageDownloadLocked(id, image);
 }
 
@@ -432,10 +456,48 @@ const Image& TextureCache::GetImage(ImageId id) const {
 	return ResolveImage(id);
 }
 
-std::vector<ImageId> TextureCache::FindImagesInRegion(uint64_t address, uint64_t size,
-                                                      bool page_overlap) const {
-	return page_overlap ? m_image_owner_index.QueryCandidates(address, size)
-	                    : m_image_owner_index.Query(address, size);
+TextureCache::ImageIds TextureCache::FindImagesInRegion(uint64_t address, uint64_t size,
+                                                        bool page_overlap) const {
+	ImagePageTable::PageRange pages {};
+	if (!ImagePageTable::TryGetPageRange(address, size, pages)) {
+		return {};
+	}
+
+	uint32_t query_epoch = ++m_image_query_epoch;
+	if (query_epoch == 0) {
+		for (const auto& slot: m_slots) {
+			if (slot.image != nullptr) {
+				slot.image->query_epoch = 0;
+			}
+		}
+		query_epoch = ++m_image_query_epoch;
+	}
+
+	ImageIds result;
+	for (size_t page = pages.first; page < pages.last_exclusive; ++page) {
+		const auto* owners = m_image_page_table.Find(page);
+		if (owners == nullptr) {
+			continue;
+		}
+		owners->ForEach([&](ImageId id) {
+			if (!id || id.index >= m_slots.size()) {
+				return;
+			}
+			const auto& slot = m_slots[id.index];
+			if (slot.generation != id.generation || slot.image == nullptr) {
+				return;
+			}
+			auto& image = *slot.image;
+			if (image.query_epoch == query_epoch) {
+				return;
+			}
+			image.query_epoch = query_epoch;
+			if (image.Overlaps(address, size, page_overlap)) {
+				result.push_back(id);
+			}
+		});
+	}
+	return result;
 }
 
 ImageId TextureCache::GetNullImage(const ImageDesc& desc) {
@@ -735,6 +797,12 @@ TextureCache::OverlapResult TextureCache::ResolveOverlap(const ImageInfo& reques
 		    (requested.IsVolume() || cached.info.IsVolume())) {
 			return {ExpandImage(requested, cached_id)};
 		}
+		if (requested.tile_mode != cached.info.tile_mode) {
+			if (safe_to_delete) {
+				DeleteImages(std::array {cached_id}, cached_id);
+			}
+			return {merged_id};
+		}
 		if (requested.pixel_format != cached.info.pixel_format ||
 		    requested.data.size <= cached.info.data.size) {
 			const auto result_id = merged_id ? merged_id : cached_id;
@@ -747,12 +815,6 @@ TextureCache::OverlapResult TextureCache::ResolveOverlap(const ImageInfo& reques
 		if (requested.type == cached.info.type && requested.resources > cached.info.resources) {
 			return {ExpandImage(requested, cached_id)};
 		}
-		if (requested.tile_mode != cached.info.tile_mode) {
-			if (safe_to_delete) {
-				DeleteImages(std::array {cached_id}, cached_id);
-			}
-			return {merged_id};
-		}
 		EXIT("TextureCache: unresolvable equal-address image overlap, address=0x%016" PRIx64
 		     " requested=%ux%u "
 		     "cached=%ux%u requested_size=0x%016" PRIx64 " cached_size=0x%016" PRIx64
@@ -819,11 +881,20 @@ struct TextureCache::ColorTransferPlan {
 	TextureUploadLayout              layout;
 	std::vector<vk::BufferImageCopy> regions;
 	std::vector<GpuTileInfo>         tiles;
+	uint64_t                         linear_size = 0;
 	bool                             tiled       = false;
 	bool                             swap_bgra16 = false;
 	bool                             valid       = false;
 };
 
+static uint64_t GetLinearSize(std::span<const GpuTileInfo> tiles) {
+	uint64_t size = 0;
+	for (const auto& tile: tiles) {
+		size = std::max(size, tile.linear_offset + tile.linear_size);
+	}
+	return size;
+}
+
 struct TextureCache::DownloadPlan {
 	ColorTransferPlan color;
 	bool              depth = false;
@@ -837,7 +908,6 @@ TextureCache::BuildColorTransfer(const Image& image, BindingType binding,
 	uint32_t    format           = info.guest_format;
 	uint32_t    layers           = info.TransferLayers();
 	bool        volume           = info.IsVolume();
-	bool        layered          = info.IsLayered();
 	bool        allow_depth_tile = direction == TransferDirection::Upload;
 	const char* owner =
 	    direction == TransferDirection::Upload ? "TextureCache" : "TextureCache readback";
@@ -848,22 +918,27 @@ TextureCache::BuildColorTransfer(const Image& image, BindingType binding,
 			case BindingType::Texture: break;
 			case BindingType::Storage: owner = "StorageTextureCache"; break;
 			case BindingType::RenderTarget:
+				if (info.resources.layers == 0 || info.data.size % info.resources.layers != 0 ||
+				    info.samples != 1 || image.backing.samples != 1) {
+					EXIT("TextureCache: invalid color-attachment upload\n");
+				}
+				format           = ImageOps::RenderTargetTransferFormat(info.bytes_per_block);
+				allow_depth_tile = false;
+				plan.swap_bgra16 = info.bgra16;
+				owner            = "RenderTarget";
+				break;
 			case BindingType::VideoOut:
 				if (info.resources.layers == 0 || info.data.size % info.resources.layers != 0 ||
 				    info.samples != 1 || image.backing.samples != 1 ||
-				    (binding == BindingType::VideoOut &&
-				     info.metadata.compression != VideoOutCompression::Uncompressed)) {
+				    info.metadata.compression != VideoOutCompression::Uncompressed) {
 					EXIT("TextureCache: invalid color-attachment upload\n");
 				}
-				format           = binding == BindingType::RenderTarget
-				                       ? ImageOps::RenderTargetTransferFormat(info.bytes_per_block)
-				                       : info.guest_format;
+				format           = info.guest_format;
 				layers           = info.resources.layers;
 				volume           = false;
-				layered          = layers > 1;
 				allow_depth_tile = false;
 				plan.swap_bgra16 = info.bgra16;
-				owner = binding == BindingType::RenderTarget ? "RenderTarget" : "VideoOut";
+				owner            = "VideoOut";
 				break;
 			case BindingType::DepthTarget: return plan;
 		}
@@ -878,22 +953,18 @@ TextureCache::BuildColorTransfer(const Image& image, BindingType binding,
 		plan.swap_bgra16 = info.bgra16;
 	}
 
-	plan.layout = TextureCalcUploadLayout(format, info.extent.width, info.extent.height,
-	                                      info.resources.levels, layers, info.pitch, info.tile_mode,
-	                                      info.data.size, allow_depth_tile, volume, owner);
-	plan.regions = TextureBuildImageCopies(plan.layout, info.extent.width, info.extent.height,
-	                                       layers, info.resources.levels, layered, volume);
-	plan.tiled   = static_cast<Prospero::TileMode>(plan.layout.tile) != Prospero::TileMode::kLinear;
+	plan.layout  = TextureCalcUploadLayout(format, info.extent.width, info.extent.height,
+	                                       info.resources.levels, layers, info.tile_mode,
+	                                       info.data.size, allow_depth_tile, volume, owner);
+	plan.regions = TextureBuildImageCopies(plan.layout);
+	plan.tiled   = static_cast<Prospero::TileMode>(plan.layout.surface.description.tile_mode) !=
+	               Prospero::TileMode::kLinear;
 	if (plan.tiled) {
-		// CMASK/FMASK decoding is intentionally deferred.
-		if (plan.layout.tile_family == TileBlockFamily::Depth64KB &&
-		    Prospero::IsFmaskTextureFormat(format)) {
-			return plan;
-		}
-		if (!TextureBuildGpuTileInfos(info.data.size, plan.regions, plan.layout, format, layers,
+		if (!TextureBuildGpuTileInfos(info.data.size, plan.regions, plan.layout,
 		                              info.resources.levels, plan.tiles)) {
 			return plan;
 		}
+		plan.linear_size = GetLinearSize(plan.tiles);
 	}
 	plan.valid = true;
 	return plan;
@@ -931,11 +1002,20 @@ void TextureCache::UploadImage(Image& image, const ImageDesc& desc, Buffer& sour
 
 	if (desc.type != BindingType::DepthTarget) {
 		auto plan = BuildColorTransfer(image, desc.type, TransferDirection::Upload);
-		EXIT_NOT_IMPLEMENTED(!plan.valid);
+		if (!plan.valid) {
+			EXIT("TextureCache: invalid color upload: binding=%u addr=0x%016" PRIx64
+			     " size=0x%016" PRIx64 " format=%u tile=%u family=%u extent=%ux%ux%u "
+			     "pitch=%u levels=%u layers=%u samples=%u\n",
+			     static_cast<uint32_t>(desc.type), info.data.address, info.data.size,
+			     info.guest_format, info.tile_mode,
+			     static_cast<uint32_t>(plan.layout.surface.texture.block.family), info.extent.width,
+			     info.extent.height, info.extent.depth, info.pitch, info.resources.levels,
+			     info.resources.layers, info.samples);
+		}
 		TileManager::Result linear {source.Handle(), source_offset, info.data.size};
 		if (plan.tiled) {
-			linear = m_tiler->Detile(source.Handle(), source_offset, info.data.size, info.data.size,
-			                         plan.tiles);
+			linear = m_tiler->Detile(source.Handle(), source_offset, info.data.size,
+			                         plan.linear_size, plan.tiles);
 		}
 		if (plan.swap_bgra16) {
 			linear = m_tiler->SwapBgra16(linear);
@@ -1019,25 +1099,20 @@ void TextureCache::InitializeImage(ImageId id, const ImageDesc& desc) {
 	if (image.info.samples > 1) {
 		return;
 	}
-	bool       data_gpu_owned = false;
-	bool       data_imported  = false;
-	const bool upload         = image.IsBufferModified() || image.IsCpuDirty();
+	bool       data_imported = false;
+	const bool upload        = image.IsBufferModified() || image.IsCpuDirty();
 	if (upload) {
 		const auto source =
 		    m_buffer_cache.ObtainBufferForImage(image.info.data.address, image.info.data.size);
 		if (source.buffer == nullptr) {
 			EXIT("TextureCache: failed to obtain image upload source\n");
 		}
-		data_gpu_owned |= source.gpu_owned;
 		data_imported = true;
 		UploadImage(image, desc, *source.buffer, source.offset);
 	}
 	if (data_imported) {
 		image.ClearBufferModified();
 	}
-	if (data_gpu_owned) {
-		image.MarkGpuModified();
-	}
 	if (image.IsCpuDirty()) {
 		image.RefreshComplete();
 	}
@@ -1068,8 +1143,7 @@ void TextureCache::RefreshImage(ImageId id, const ImageDesc& desc) {
 }
 
 void TextureCache::AssociateStencil(ImageId depth_id, GuestRange stencil) {
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
+	CacheLock lock(*this, m_lock);
 	AssociateStencilLocked(depth_id, stencil);
 }
 
@@ -1112,17 +1186,14 @@ ImageId TextureCache::FindImage(ImageDesc& desc, bool exact_format) {
 	}
 
 	ImageId result {};
-	bool    replacement_buffer = false;
-	bool    inserted_new       = false;
 	{
-		std::lock_guard            transaction(m_resource_mutex);
-		CacheLock                  lock(*this, m_lock);
-		const std::vector<ImageId> candidates =
+		CacheLock  lock(*this, m_lock);
+		const auto candidates =
 		    FindImagesInRegion(desc.info.data.address, desc.info.data.size, false);
 
 		for (const auto id: candidates) {
 			const auto owner = ResolveOwner(id);
-			if (owner == nullptr || owner->info.data != desc.info.data) {
+			if (owner == nullptr) {
 				continue;
 			}
 			if (SameBacking(owner->info, desc.info, exact_format)) {
@@ -1140,8 +1211,8 @@ ImageId TextureCache::FindImage(ImageDesc& desc, bool exact_format) {
 				if (owner == nullptr) {
 					continue;
 				}
-				const auto merged_info = result ? ResolveImage(result).info : desc.info;
-				const auto overlap     = ResolveOverlap(merged_info, desc.type, candidate, result);
+				const auto& merged_info = result ? ResolveImage(result).info : desc.info;
+				const auto  overlap     = ResolveOverlap(merged_info, desc.type, candidate, result);
 				if (overlap.image) {
 					result     = overlap.image;
 					view_mip   = overlap.mip;
@@ -1155,32 +1226,17 @@ ImageId TextureCache::FindImage(ImageDesc& desc, bool exact_format) {
 			if (exact_format && resolved.info.pixel_format != desc.info.pixel_format) {
 				result = {};
 			} else if (resolved.info.resources < desc.info.resources) {
-				ImageDesc refresh {
-				    .info = resolved.info, .view_info = {}, .type = UploadBinding(resolved)};
-				RefreshImage(result, refresh);
-				if (resolved.IsGpuModified() && !SynchronizeImageToBuffer(result)) {
-					EXIT("TextureCache: cannot preserve an unsupported replacement image\n");
-				}
-				replacement_buffer = resolved.IsBufferModified();
-				DeleteImage(result);
-				result = {};
+				result = ExpandImage(desc.info, result);
 			}
 		}
 		if (!result) {
 			result         = InsertImage(desc.info);
-			inserted_new   = true;
 			auto& inserted = ResolveImage(result);
-			if (replacement_buffer || m_buffer_cache.HasGpuDirtyBytes(inserted.info.data.address,
-			                                                          inserted.info.data.size)) {
+			if (m_buffer_cache.HasGpuDirtyBytes(inserted.info.data.address,
+			                                    inserted.info.data.size)) {
 				inserted.MarkBufferModified();
 			}
 		}
-		if (inserted_new) {
-			InitializeImage(result, desc);
-		} else {
-			RefreshImage(result, desc);
-		}
-
 		auto& image = ResolveImage(result);
 		if (desc.type == BindingType::VideoOut &&
 		    desc.info.metadata.compression != VideoOutCompression::Uncompressed) {
@@ -1205,6 +1261,13 @@ ImageId TextureCache::FindImage(ImageDesc& desc, bool exact_format) {
 	return result;
 }
 
+void TextureCache::UpdateImage(ImageId id) {
+	CacheLock lock(*this, m_lock);
+	auto&     image = ResolveImage(id);
+	TouchImage(image);
+	RefreshImage(id, ImageDesc {.info = image.info, .type = UploadBinding(image)});
+}
+
 ImageId TextureCache::FindImageFromRange(uint64_t address, uint64_t size, bool ensure_valid) {
 	if (!GuestRange {address, size}.Valid()) {
 		return {};
@@ -1249,22 +1312,24 @@ ImageId TextureCache::FindImageFromRange(uint64_t address, uint64_t size, bool e
 }
 
 vk::ImageView TextureCache::FindTexture(ImageId id, const ImageDesc& desc) {
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
-	auto&           image = ResolveImage(id);
+	CacheLock lock(*this, m_lock);
+	auto&     image = ResolveImage(id);
 	TouchImage(image);
 	if (!image.info.data.Empty()) {
 		if (!image.registered || image.depth_id || image.binding.needs_rebind) {
 			EXIT("TextureCache: texture requires rediscovery before final acquisition\n");
 		}
+	}
+	if (desc.type == BindingType::Storage) {
+		image.MarkGpuModified();
+	}
+	if (!image.info.data.Empty()) {
 		RefreshImage(id, desc);
 	}
 	switch (desc.type) {
 		case BindingType::Texture: break;
 		case BindingType::Storage:
-			if (image.info.data.Empty()) {
-				image.MarkGpuModified();
-			} else {
+			if (!image.info.data.Empty()) {
 				if (!image.registered || image.depth_id) {
 					EXIT("TextureCache: cannot acquire an unavailable storage image\n");
 				}
@@ -1283,16 +1348,16 @@ vk::ImageView TextureCache::FindRenderTarget(ImageId id, const ImageDesc& desc)
 	if (desc.type != BindingType::RenderTarget) {
 		EXIT("TextureCache: invalid color-target binding\n");
 	}
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
-	auto&           image = ResolveImage(id);
+	CacheLock lock(*this, m_lock);
+	auto&     image = ResolveImage(id);
 	if (!image.registered || image.depth_id || image.binding.needs_rebind) {
 		EXIT("TextureCache: color target requires rediscovery before final acquisition\n");
 	}
 	TouchImage(image);
+	image.MarkGpuModified();
+	image.usage.render_target = true;
 	RefreshImage(id, desc);
 	CommitGpuWrite(image);
-	image.usage.render_target = true;
 	TrackImageDownloadLocked(id, image);
 	const auto view = image.FindView(desc.view_info);
 	RetainImage(m_scheduler.Current(), id);
@@ -1303,13 +1368,14 @@ vk::ImageView TextureCache::FindDepthTarget(ImageId id, const ImageDesc& desc) {
 	if (desc.type != BindingType::DepthTarget) {
 		EXIT("TextureCache: invalid depth-target binding\n");
 	}
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
-	auto&           image = ResolveImage(id);
+	CacheLock lock(*this, m_lock);
+	auto&     image = ResolveImage(id);
 	if (!image.registered || image.depth_id || image.binding.needs_rebind) {
 		EXIT("TextureCache: depth target requires rediscovery before final acquisition\n");
 	}
 	TouchImage(image);
+	image.MarkGpuModified();
+	image.usage.depth_target = true;
 	RefreshImage(id, desc);
 	if (desc.info.HasMetadata()) {
 		image.info.metadata = desc.info.metadata;
@@ -1317,7 +1383,6 @@ vk::ImageView TextureCache::FindDepthTarget(ImageId id, const ImageDesc& desc) {
 		                            MetaDataInfo {.clear_mask = image.info.htile_clear_mask});
 	}
 	CommitGpuWrite(image);
-	image.usage.depth_target = true;
 	if (desc.info.HasStencil()) {
 		AssociateStencilLocked(id, desc.info.stencil);
 	}
@@ -1327,9 +1392,8 @@ vk::ImageView TextureCache::FindDepthTarget(ImageId id, const ImageDesc& desc) {
 }
 
 void TextureCache::MarkGpuWritten(ImageId id) {
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
-	auto&           image = ResolveImage(id);
+	CacheLock lock(*this, m_lock);
+	auto&     image = ResolveImage(id);
 	if (!image.registered || image.depth_id) {
 		EXIT("TextureCache: cannot mark an unavailable image GPU-written\n");
 	}
@@ -1341,11 +1405,6 @@ void TextureCache::CommitGpuWrite(Image& image) {
 	if (image.depth_id || image.backing.image == nullptr) {
 		EXIT("TextureCache: stencil association cannot own image contents\n");
 	}
-	const auto range = image.info.data;
-	if (m_buffer_cache.HasGpuDirtyBytes(range.address, range.size)) {
-		m_buffer_cache.DiscardGpuDirtyBytes(range.address, range.size);
-	}
-	m_buffer_cache.InvalidateImageAliases(range.address, range.size);
 	image.ClearBufferModified();
 	if (image.IsCpuDirty()) {
 		image.RefreshComplete();
@@ -1358,8 +1417,6 @@ bool TextureCache::ClearImageFromBuffer(CommandBuffer& command, uint64_t address
 	if (command.IsInvalid() || !GuestRange {address, size}.Valid()) {
 		EXIT("TextureCache: invalid image clear\n");
 	}
-	m_buffer_cache.ValidateGpuAccess(address, size, false, true);
-	std::lock_guard      transaction(m_resource_mutex);
 	CacheLock            lock(*this, m_lock);
 	ImageId              selected {};
 	vk::ImageAspectFlags aspect {};
@@ -1411,9 +1468,6 @@ bool TextureCache::ClearImageFromBuffer(CommandBuffer& command, uint64_t address
 			return false;
 		}
 	}
-	if (m_buffer_cache.HasGpuDirtyBytes(address, size)) {
-		m_buffer_cache.DiscardGpuDirtyBytes(address, size);
-	}
 	if (image.IsBufferModified() || image.IsCpuDirty()) {
 		ImageDesc refresh {.info = image.info, .view_info = {}, .type = UploadBinding(image)};
 		InitializeImage(selected, refresh);
@@ -1533,11 +1587,14 @@ void TextureCache::DownloadDepth(Image& image, Buffer& destination, uint64_t des
 }
 
 void TextureCache::DownloadImageData(Image& image, Buffer& destination, uint64_t destination_offset,
-                                     DownloadPlan plan) {
+                                     uint64_t destination_size, DownloadPlan plan) {
 	if (!plan.valid) {
 		EXIT("TextureCache: invalid image download plan\n");
 	}
 	if (plan.depth) {
+		if (destination_size != image.info.data.size) {
+			EXIT("TextureCache: partial depth image download is unsupported\n");
+		}
 		DownloadDepth(image, destination, destination_offset);
 		return;
 	}
@@ -1547,22 +1604,114 @@ void TextureCache::DownloadImageData(Image& image, Buffer& destination, uint64_t
 	                                         : TileManager::ColorTransform::None;
 	if (!color.tiled) {
 		if (transform == TileManager::ColorTransform::SwapBgra16) {
-			auto linear = m_tiler->GetScratchBuffer(image.info.data.size);
+			auto linear = m_tiler->GetScratchBuffer(destination_size);
 			image.Download(color.regions, linear.buffer, 0, linear.size);
 			m_tiler->SwapBgra16(linear,
-			                    {destination.Handle(), destination_offset, image.info.data.size});
+			                    {destination.Handle(), destination_offset, destination_size});
 			return;
 		}
 		for (auto& copy: color.regions) {
 			copy.bufferOffset += destination_offset;
 		}
-		image.Download(color.regions, destination.Handle(), destination_offset,
-		               image.info.data.size);
+		image.Download(color.regions, destination.Handle(), destination_offset, destination_size);
 		return;
 	}
 
 	m_tiler->TileImage(image, color.regions, destination.Handle(), destination_offset,
-	                   image.info.data.size, image.info.data.size, color.tiles, transform);
+	                   destination_size, color.linear_size, color.tiles, transform);
+}
+
+bool BufferCache::SynchronizeBufferFromImage(Buffer& buffer, uint64_t vaddr, uint64_t size) {
+	CacheLock            lock(m_texture_cache, m_texture_cache.m_lock);
+	std::vector<ImageId> matches;
+	for (const auto id: m_texture_cache.FindImagesInRegion(vaddr, size, false)) {
+		auto owner = m_texture_cache.ResolveOwner(id);
+		if (owner == nullptr || owner->info.data.address != vaddr) {
+			continue;
+		}
+		if (owner->depth_id) {
+			owner = m_texture_cache.ResolveOwner(owner->depth_id);
+		}
+		if (owner != nullptr && owner->SafeToDownload()) {
+			matches.push_back(id);
+		}
+	}
+
+	ImageId selected {};
+	if (matches.size() == 1) {
+		selected = matches.front();
+	} else {
+		for (const auto id: matches) {
+			const auto& image = m_texture_cache.ResolveImage(id);
+			if (image.info.data.size == size) {
+				selected = id;
+				break;
+			}
+		}
+	}
+	if (!selected) {
+		return false;
+	}
+	if (const auto owner = m_texture_cache.ResolveOwner(selected);
+	    owner != nullptr && owner->depth_id) {
+		selected = owner->depth_id;
+	}
+
+	auto& image = m_texture_cache.ResolveImage(selected);
+	if (!buffer.IsInBounds(image.info.data.address, 1)) {
+		return false;
+	}
+	const auto buf_offset = buffer.Offset(image.info.data.address);
+	const auto available  = buffer.Size() - buf_offset;
+	uint32_t   levels     = 0;
+	uint64_t   copy_size  = 0;
+	if (image.info.IsVolume()) {
+		// Volume mips contain strided block slices, so a mip's linear span cannot prove that
+		// every retained slice fits. Keep volume synchronization whole-image only.
+		if (!buffer.IsInBounds(image.info.data.address, image.info.data.size)) {
+			return false;
+		}
+		levels    = image.info.resources.levels;
+		copy_size = image.info.data.size;
+	} else {
+		for (; levels < image.info.resources.levels; ++levels) {
+			const auto& mip = image.info.mip_layout[levels];
+			if (mip.size == 0 || mip.offset > available || mip.size > available - mip.offset) {
+				break;
+			}
+			copy_size = std::max(copy_size, mip.offset + mip.size);
+		}
+	}
+	if (copy_size == 0) {
+		return false;
+	}
+	auto plan = m_texture_cache.BuildDownload(image);
+	if (!plan.valid) {
+		return false;
+	}
+	if (plan.depth && copy_size != image.info.data.size) {
+		return false;
+	}
+	if (!plan.depth && levels < image.info.resources.levels) {
+		auto& color = plan.color;
+		std::erase_if(color.regions, [levels](const vk::BufferImageCopy& region) {
+			return region.imageSubresource.mipLevel >= levels;
+		});
+		if (color.regions.empty()) {
+			return false;
+		}
+		if (color.tiled) {
+			color.tiles.clear();
+			if (!TextureBuildGpuTileInfos(copy_size, color.regions, color.layout, levels,
+			                              color.tiles)) {
+				return false;
+			}
+			color.linear_size = GetLinearSize(color.tiles);
+		}
+	}
+	m_texture_cache.DownloadImageData(image, buffer, buf_offset, copy_size, std::move(plan));
+	m_texture_cache.RetainImage(m_scheduler.Current(), selected);
+	return true;
 }
 
 std::pair<uint8_t*, uint64_t> TextureCache::MapDownload(uint64_t size, uint64_t alignment) {
@@ -1594,12 +1743,9 @@ void TextureCache::QueueDownload(GuestRange range, StreamBuffer& download, uint8
 	m_scheduler.Current().Handle().pipelineBarrier(vk::PipelineStageFlagBits::eAllCommands,
 	                                               vk::PipelineStageFlagBits::eHost, {}, 0, nullptr,
 	                                               1, &barrier, 0, nullptr);
-	const auto tick = m_scheduler.CurrentTick();
-	m_buffer_cache.BeginBackingPublication(range.address, range.size, tick);
-	m_scheduler.DeferPriorityOperation([this, &download, range, mapped, offset, tick] {
+	m_scheduler.DeferPriorityOperation([&download, range, mapped, offset] {
 		download.Invalidate(offset, range.size);
 		LibKernel::Memory::WriteBacking(range.address, mapped, range.size);
-		m_buffer_cache.CompleteBackingPublication(range.address, range.size, tick);
 	});
 }
 
@@ -1620,7 +1766,7 @@ bool TextureCache::TryDownloadImage(ImageId id) {
 	}
 	download.Flush(offset, range.size);
 
-	DownloadImageData(image, download, offset, std::move(plan));
+	DownloadImageData(image, download, offset, range.size, std::move(plan));
 
 	QueueDownload(range, download, mapped, offset);
 	return true;
@@ -1634,62 +1780,6 @@ void TextureCache::DownloadImage(ImageId id) {
 	m_scheduler.DrainPriorityOperations();
 }
 
-bool TextureCache::SynchronizeImageToBuffer(ImageId id) {
-	auto& image = ResolveImage(id);
-	if (image.depth_id) {
-		return true;
-	}
-	auto plan = BuildDownload(image);
-	if (!plan.valid) {
-		return false;
-	}
-	const auto range = image.info.data;
-	if (image.IsCpuDirty()) {
-		RefreshImage(id,
-		             ImageDesc {.info = image.info, .view_info = {}, .type = UploadBinding(image)});
-	}
-	if (!image.IsGpuModified()) {
-		return true;
-	}
-	if (image.IsDefinitelyCpuDirty() || image.IsBufferModified()) {
-		EXIT("TextureCache: image mirror source is not native-current\n");
-	}
-	auto [destination, offset] =
-	    m_buffer_cache.ObtainBufferForImageWrite(range.address, range.size);
-	if (destination == nullptr) {
-		EXIT("TextureCache: failed to allocate image mirror\n");
-	}
-	DownloadImageData(image, *destination, offset, std::move(plan));
-	m_scheduler.Current().RetainResourceUntilFence(destination);
-	m_buffer_cache.PublishImageBuffer(range.address, range.size);
-	image.MarkBufferModified();
-	RetainImage(m_scheduler.Current(), id);
-	ClearGpuModified(id);
-	return true;
-}
-
-bool TextureCache::SynchronizeImageToBuffer(uint64_t address, uint64_t size) {
-	if (!GuestRange {address, size}.Valid()) {
-		return false;
-	}
-	CacheLock lock(*this, m_lock);
-	ImageId   selected {};
-	for (const auto id: FindImagesInRegion(address, size, true)) {
-		auto owner = ResolveOwner(id);
-		if (owner == nullptr || !owner->GpuOverlaps(address, size)) {
-			continue;
-		}
-		if (selected) {
-			EXIT("TextureCache: ambiguous image-to-buffer synchronization\n");
-		}
-		selected = id;
-	}
-	if (!selected) {
-		return false;
-	}
-	return SynchronizeImageToBuffer(selected);
-}
-
 bool TextureCache::InvalidateMemoryFromGPU(uint64_t address, uint64_t size,
                                            bool formatted_buffer_write) {
 	if (!GuestRange {address, size}.Valid()) {
@@ -1784,9 +1874,8 @@ bool TextureCache::IsMetaCleared(uint64_t address, uint32_t slice) {
 }
 
 bool TextureCache::ClearMeta(uint64_t address) {
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
-	const auto      found = m_surface_metas.find(address);
+	CacheLock  lock(*this, m_lock);
+	const auto found = m_surface_metas.find(address);
 	if (found == m_surface_metas.end()) {
 		return false;
 	}
@@ -1808,44 +1897,12 @@ bool TextureCache::TouchMeta(uint64_t address, uint32_t slice, bool is_clear) {
 	return true;
 }
 
-bool TextureCache::InvalidateMemory(PageFaultAccess access, uint64_t address, uint64_t size,
-                                    PageFaultPhase phase) noexcept {
-	if ((access != PageFaultAccess::Read && access != PageFaultAccess::Write) ||
-	    !GuestRange {address, size}.Valid()) {
-		return false;
-	}
-	if (access == PageFaultAccess::Read) {
-		return false;
-	}
-	if (phase == PageFaultPhase::Invalidate) {
-		CacheLock  lock(*this, m_lock);
-		const bool tracked =
-		    std::ranges::any_of(FindImagesInRegion(address, size, true), [&](ImageId id) {
-			    const auto owner = ResolveOwner(id);
-			    return owner != nullptr && !owner->depth_id && owner->IsTracked();
-		    });
-		if (tracked) {
-			InvalidateCpuAliases(address, size);
-		}
-		return tracked;
-	}
-	if (phase != PageFaultPhase::Complete && phase != PageFaultPhase::Release) {
-		return false;
-	}
-	CacheLock lock(*this, m_lock);
-	return std::ranges::any_of(FindImagesInRegion(address, size, true), [&](ImageId id) {
-		const auto owner = ResolveOwner(id);
-		return owner != nullptr && !owner->depth_id;
-	});
-}
-
 void TextureCache::UnmapMemory(uint64_t address, uint64_t size) {
 	if (!GuestRange {address, size}.Valid()) {
 		EXIT("TextureCache: invalid unmap range\n");
 	}
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
-	auto            images = FindImagesInRegion(address, size, false);
+	CacheLock lock(*this, m_lock);
+	auto      images = FindImagesInRegion(address, size, false);
 	if (!images.empty()) {
 		m_scheduler.Finish();
 	}
@@ -1862,9 +1919,8 @@ void TextureCache::UnmapMemory(uint64_t address, uint64_t size) {
 }
 
 void TextureCache::RunGarbageCollector() {
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
-	const uint64_t  tick = m_gc_tick++;
+	CacheLock      lock(*this, m_lock);
+	const uint64_t tick = m_gc_tick++;
 	if (m_graphics.CanReportMemoryUsage()) {
 		m_total_used_memory = m_graphics.GetDeviceMemoryUsage();
 	}
@@ -1924,8 +1980,7 @@ void TextureCache::RunGarbageCollector() {
 }
 
 void TextureCache::ProcessDownloadImages() {
-	std::lock_guard transaction(m_resource_mutex);
-	CacheLock       lock(*this, m_lock);
+	CacheLock lock(*this, m_lock);
 	for (const auto id: m_download_images) {
 		const auto owner = ResolveOwner(id);
 		if (owner != nullptr && owner->registered && owner->IsGpuModified()) {

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

@@ -26,7 +26,6 @@ class Buffer;
 class BufferCache;
 class CommandBuffer;
 class CommandScheduler;
-class ResourceMutex;
 class RenderExecutor;
 class StreamBuffer;
 class TileManager;
@@ -49,11 +48,12 @@ public:
 	};
 
 	TextureCache(GraphicContext& graphics, CommandScheduler& scheduler, PageManager& page_manager,
-	             BufferCache& buffer_cache, ResourceMutex& resource_mutex);
+	             BufferCache& buffer_cache);
 	~TextureCache();
 	KYTY_CLASS_NO_COPY(TextureCache);
 
 	[[nodiscard]] ImageId       FindImage(ImageDesc& desc, bool exact_format = false);
+	void                        UpdateImage(ImageId id);
 	[[nodiscard]] ImageId       FindImageFromRange(uint64_t address, uint64_t size,
 	                                               bool ensure_valid = true);
 	[[nodiscard]] vk::ImageView FindTexture(ImageId id, const ImageDesc& desc);
@@ -66,7 +66,6 @@ public:
 	[[nodiscard]] bool ClearImageFromBuffer(CommandBuffer& command, uint64_t address, uint64_t size,
 	                                        uint32_t packed_clear);
 	void               InvalidateMemory(uint64_t address, uint64_t size);
-	[[nodiscard]] bool SynchronizeImageToBuffer(uint64_t address, uint64_t size);
 	[[nodiscard]] bool InvalidateMemoryFromGPU(uint64_t address, uint64_t size,
 	                                           bool formatted_buffer_write = false);
 	[[nodiscard]] RegionInfo QueryRegion(uint64_t address, uint64_t size);
@@ -76,11 +75,9 @@ public:
 	[[nodiscard]] bool ClearMeta(uint64_t address);
 	[[nodiscard]] bool TouchMeta(uint64_t address, uint32_t slice, bool is_clear);
 
-	[[nodiscard]] bool InvalidateMemory(PageFaultAccess access, uint64_t address, uint64_t size,
-	                                    PageFaultPhase phase) noexcept;
-	void               UnmapMemory(uint64_t address, uint64_t size);
-	void               ProcessDownloadImages();
-	void               RunGarbageCollector();
+	void UnmapMemory(uint64_t address, uint64_t size);
+	void ProcessDownloadImages();
+	void RunGarbageCollector();
 
 private:
 	enum class TransferDirection { Upload, Download };
@@ -102,7 +99,8 @@ private:
 		int32_t layer = -1;
 	};
 
-	using ImageOwnerIndex = MultiRangePageOwnerIndex<ImageId>;
+	using ImageIds       = InlinePageOwnerList<ImageId, 16>;
+	using ImagePageTable = MultiLevelPageTable<ImageIds, 20, 40, 10>;
 
 	[[nodiscard]] Image&                 ResolveImage(ImageId id);
 	[[nodiscard]] const Image&           ResolveImage(ImageId id) const;
@@ -128,8 +126,9 @@ private:
 	[[nodiscard]] static BindingType UploadBinding(const Image& image);
 	[[nodiscard]] bool               SafeToDownload(const Image& image);
 
-	[[nodiscard]] std::vector<ImageId> FindImagesInRegion(uint64_t address, uint64_t size,
-	                                                      bool page_overlap) const;
+	// Caller holds m_lock; it also serializes the per-image query epoch.
+	[[nodiscard]] ImageIds      FindImagesInRegion(uint64_t address, uint64_t size,
+	                                               bool page_overlap) const;
 	[[nodiscard]] OverlapResult ResolveOverlap(const ImageInfo& requested, BindingType binding,
 	                                           ImageId cached, ImageId merged);
 	[[nodiscard]] ImageId       ResolveDepthOverlap(const ImageInfo& requested, BindingType binding,
@@ -142,7 +141,7 @@ private:
 	[[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,
-	                       DownloadPlan plan);
+	                       uint64_t destination_size, DownloadPlan plan);
 	void DownloadDepth(Image& image, Buffer& destination, uint64_t destination_offset);
 	void CommitGpuWrite(Image& image);
 	void PrepareImageCopy(Image& image);
@@ -157,7 +156,6 @@ private:
 	void InvalidateCpuAliases(uint64_t address, uint64_t size);
 	void ClearGpuModified(ImageId id);
 
-	[[nodiscard]] bool                          SynchronizeImageToBuffer(ImageId id);
 	void                                        DownloadImage(ImageId id);
 	[[nodiscard]] bool                          TryDownloadImage(ImageId id);
 	[[nodiscard]] std::pair<uint8_t*, uint64_t> MapDownload(uint64_t size, uint64_t alignment);
@@ -170,10 +168,9 @@ private:
 	BlitHelper                                        m_blit_helper;
 	std::unique_ptr<TileManager>                      m_tiler;
 	BufferCache&                                      m_buffer_cache;
-	ResourceMutex&                                    m_resource_mutex;
 	std::vector<Slot>                                 m_slots;
 	std::vector<uint32_t>                             m_free_slots;
-	ImageOwnerIndex                                   m_image_owner_index;
+	ImagePageTable                                    m_image_page_table;
 	std::map<vk::Format, ImageId>                     m_null_images;
 	Common::LeastRecentlyUsedCache<ImageId, uint64_t> m_lru_cache;
 	std::set<ImageId>                                 m_download_images;
@@ -181,11 +178,13 @@ private:
 	uint64_t                                          m_total_used_memory  = 0;
 	uint64_t                                          m_trigger_gc_memory  = 0;
 	uint64_t                                          m_pressure_gc_memory = 1536ull * 1024 * 1024;
-	uint64_t m_critical_gc_memory     = 3ull * 1024 * 1024 * 1024;
-	uint64_t m_gc_tick                = 0;
-	bool     m_readback_linear_images = false;
+	uint64_t         m_critical_gc_memory     = 3ull * 1024 * 1024 * 1024;
+	uint64_t         m_gc_tick                = 0;
+	mutable uint32_t m_image_query_epoch      = 0;
+	bool             m_readback_linear_images = false;
 
 	friend struct TextureCacheTestAccess;
+	friend class BufferCache;
 	friend class RenderExecutor;
 };
 

src/graphics/host_gpu/renderer/colorRenderTarget.cpp

@@ -7,8 +7,8 @@
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/guest_gpu/tile.h"
 #include "graphics/host_gpu/graphicContext.h"
-#include "graphics/host_gpu/renderer/image/textureCommon.h"
 #include "graphics/host_gpu/renderer/debug.h"
+#include "graphics/host_gpu/renderer/image/textureCommon.h"
 #include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
@@ -23,10 +23,10 @@ static std::atomic<uint32_t> g_render_color_log_count = 0;
 
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
 void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandBuffer& buffer,
-                                          RenderColorInfo& r,
-                                          uint32_t render_target_slice_offset,
-                                          uint32_t render_target_slot, bool ignore_target_mask,
-                                          bool exact_format) {
+                                              RenderColorInfo& r,
+                                              uint32_t         render_target_slice_offset,
+                                              uint32_t render_target_slot, bool ignore_target_mask,
+                                              bool exact_format) {
 	KYTY_PROFILER_FUNCTION();
 	const auto& hw = buffer.GetRegisters();
 
@@ -79,10 +79,8 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	const auto view = ResolveTargetViewInfo(
 	    rt.view.base_array_slice_index, rt.view.last_array_slice_index, render_target_slice_offset);
 	switch (view.type) {
-		case TargetViewType::Image2D: break;
-		case TargetViewType::Image2DArray:
-			EXIT("layered render-target views are unsupported: base=%u count=%u\n", view.base_layer,
-			     view.layer_count);
+		case TargetViewType::Image2D:
+		case TargetViewType::Image2DArray: break;
 		case TargetViewType::Unsupported:
 			EXIT("invalid render-target view: base=%u last=%u draw_offset=%u\n",
 			     rt.view.base_array_slice_index, rt.view.last_array_slice_index,
@@ -121,7 +119,18 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	uint32_t   pitch  = 0;
 	uint64_t   size   = 0;
 	bool       tile   = false;
-	const bool standard64 =
+	const bool volume = rt.attrib3.dimension == 2;
+	if (rt.attrib3.dimension != 1 && !volume) {
+		EXIT("unsupported render-target dimension: %u\n", rt.attrib3.dimension);
+	}
+	if (!volume && rt.attrib3.depth != 0) {
+		EXIT("2D render target has nonzero depth: %u\n", rt.attrib3.depth);
+	}
+	if (volume && samples != 1) {
+		EXIT("multisampled 3D render targets are unsupported\n");
+	}
+	const uint32_t depth = volume ? rt.attrib3.depth + 1u : 1u;
+	const bool     standard64 =
 	    rt.attrib3.tile_mode == Prospero::GpuEnumValue(Prospero::TileMode::kStandard64KB);
 
 	switch (rt.attrib3.tile_mode) {
@@ -147,6 +156,7 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	if (bytes_per_element == 0) {
 		EXIT("render-target format has no valid element size\n");
 	}
+	const auto transfer_format = ImageOps::RenderTargetTransferFormat(bytes_per_element);
 	if (standard64 &&
 	    (rt.attrib3.dimension != 1 || rt.attrib3.depth != 0 || levels != 1 ||
 	     rt.view.current_mip_level != 0 || view.base_layer != 0 || view.image_layers != 1 ||
@@ -167,10 +177,14 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	if (rt.pitch.pitch_div8_minus1 != 0) {
 		pitch = (rt.pitch.pitch_div8_minus1 + 1u) << 3u;
 	} else if (tile) {
-		pitch = standard64
-		            ? TileGetTexturePitch(Prospero::GpuEnumValue(Prospero::BufferFormat::k32Float),
-		                                  width, levels, rt.attrib3.tile_mode)
-		            : TileGetRenderTargetPitch(width, bytes_per_element, rt.attrib.num_fragments);
+		if (volume) {
+			pitch = TileGetTexturePitch(transfer_format, width, rt.attrib3.tile_mode);
+		} else if (standard64) {
+			pitch = TileGetTexturePitch(Prospero::GpuEnumValue(Prospero::BufferFormat::k32Float),
+			                            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);
 		}
@@ -178,14 +192,31 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 		pitch = width;
 	}
 
-	TileSizeOffset mip_sizes[16] {};
-	TilePaddedSize mip_padded[16] {};
-	if (tile) {
+	TileSizeOffset    mip_sizes[16] {};
+	TilePaddedSize    mip_padded[16] {};
+	TileSurfaceLayout volume_layout {};
+	uint64_t          backing_size = 0;
+	if (volume) {
+		const TileSurfaceDescription description {transfer_format,
+		                                          rt.attrib3.tile_mode,
+		                                          TileSurfaceDimension::Dim3D,
+		                                          width,
+		                                          height,
+		                                          depth,
+		                                          levels,
+		                                          1};
+		if (!tile || !TileGetTiledTextureLayout(description, volume_layout)) {
+			EXIT("unsupported 3D render-target layout: %ux%ux%u levels=%u tile=%u\n", width, height,
+			     depth, levels, rt.attrib3.tile_mode);
+		}
+		size         = volume_layout.block_slice_size;
+		backing_size = volume_layout.total_size;
+	} else if (tile) {
 		TileSizeAlign layout {};
 		bool          valid_layout = false;
 		if (standard64) {
 			TileGetTextureSize(Prospero::GpuEnumValue(Prospero::BufferFormat::k32Float), width,
-			                   height, pitch, levels, rt.attrib3.tile_mode, &layout, mip_sizes,
+			                   height, levels, rt.attrib3.tile_mode, &layout, mip_sizes,
 			                   mip_padded);
 			valid_layout = layout.size != 0 && layout.align == 65536;
 		} else {
@@ -205,12 +236,6 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 			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);
-		}
 	} else {
 		size = static_cast<uint64_t>(pitch) * height * bytes_per_element * samples;
 		if (size > UINT32_MAX) {
@@ -219,23 +244,40 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 		mip_sizes[0]  = {static_cast<uint32_t>(size), 0, 0, 0, 0, 0};
 		mip_padded[0] = {pitch, height};
 	}
-	if (size == 0 || size > UINT64_MAX / view.image_layers) {
+	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");
 	}
-	const auto backing_size = size * view.image_layers;
+	if (!volume) {
+		backing_size = size * view.image_layers;
+	}
+	if (backing_size == 0) {
+		EXIT("render-target backing is empty\n");
+	}
 	if (backing_size > TRACKER_ADDRESS_SIZE - rt.base.addr) {
 		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) {
 		LOGF("RenderColorTarget: slot=%" PRIu32 " addr=0x%010" PRIx64 " size=0x%016" PRIx64
-		     " extent=%ux%u view_mip=%u view_extent=%ux%u levels=%u pitch=%u"
+		     " extent=%ux%ux%u view_mip=%u view_extent=%ux%u levels=%u pitch=%u"
 		     " fmt=0x%08" PRIx32 " nfmt=0x%08" PRIx32 " order=0x%08" PRIx32 " samples=%u tile=%s\n",
-		     rt_slot, rt.base.addr, backing_size, width, height, rt.view.current_mip_level,
+		     rt_slot, rt.base.addr, backing_size, width, height, depth, rt.view.current_mip_level,
 		     view_extent.width, view_extent.height, levels, pitch, rt.info.format,
 		     rt.info.channel_type, rt.info.channel_order, samples, tile ? "tiled" : "linear");
 	}
@@ -244,15 +286,25 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	desc.type              = TextureCache::BindingType::RenderTarget;
 	desc.info.data         = {rt.base.addr, backing_size};
 	desc.info.pixel_format = target_format.format;
-	desc.info.guest_format = ImageOps::RenderTargetTransferFormat(bytes_per_element);
-	desc.info.type   = Prospero::ImageType::kColor2D;
-	desc.info.extent = {width, height, 1};
-	desc.info.resources       = {levels, view.image_layers};
-	desc.info.pitch           = pitch;
+	desc.info.guest_format = transfer_format;
+	desc.info.type         = volume ? Prospero::ImageType::kColor3D : Prospero::ImageType::kColor2D;
+	desc.info.extent       = {width, height, depth};
+	desc.info.resources    = {levels, volume ? 1u : view.image_layers};
+	desc.info.pitch        = pitch;
 	desc.info.bytes_per_block = bytes_per_element;
 	desc.info.samples         = samples;
 	desc.info.tile_mode       = rt.attrib3.tile_mode;
 	for (uint32_t level = 0; level < levels; level++) {
+		if (volume) {
+			const auto& mip             = volume_layout.mips[level];
+			desc.info.mip_layout[level] = {
+			    mip.offset,
+			    mip.size,
+			    mip.padded_width,
+			    mip.padded_height,
+			};
+			continue;
+		}
 		const auto level_offset =
 		    mip_sizes[level].src_size != 0 ? mip_sizes[level].src_offset : mip_sizes[level].offset;
 		const auto level_size =
@@ -275,20 +327,20 @@ void RenderExecutor::ResolveRenderColorTarget(uint64_t submit_id, RenderCommandB
 	desc.view_info.base_layer  = view.base_layer;
 	desc.view_info.layer_count = view.layer_count;
 	desc.view_info.usage       = vk::ImageUsageFlagBits::eColorAttachment;
-	auto& texture_cache = m_context.GetTextureCache();
-	r.desc              = std::move(desc);
-	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;
-	r.color_clear_value = {};
+	auto& texture_cache        = m_context.GetTextureCache();
+	r.desc                     = std::move(desc);
+	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;
+	r.color_clear_value        = {};
 	BindRenderTarget(r.image_id);
 }
 

src/graphics/host_gpu/renderer/colorRenderTarget.h

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

src/graphics/host_gpu/renderer/commandScheduler.cpp

@@ -44,13 +44,13 @@ CommandSlot* CommandScheduler::CommandPool::CreateSlot() {
 	allocate.commandPool        = m_pool;
 	allocate.level              = vk::CommandBufferLevel::ePrimary;
 	allocate.commandBufferCount = 1;
-	vk::CommandBuffer buffer = nullptr;
+	vk::CommandBuffer buffer    = nullptr;
 	EXIT_IF(graphics.device.allocateCommandBuffers(&allocate, &buffer) != vk::Result::eSuccess);
 
 	vk::FenceCreateInfo fence_create {};
 	fence_create.sType = vk::StructureType::eFenceCreateInfo;
 	fence_create.flags = vk::FenceCreateFlagBits::eSignaled;
-	vk::Fence fence = nullptr;
+	vk::Fence fence    = nullptr;
 	if (graphics.device.createFence(&fence_create, nullptr, &fence) != vk::Result::eSuccess) {
 		graphics.device.freeCommandBuffers(m_pool, 1, &buffer);
 		EXIT("failed to create command-buffer fence\n");
@@ -70,9 +70,9 @@ CommandSlot* CommandScheduler::CommandPool::Allocate(GraphicContext& graphics) {
 		Create(graphics);
 	}
 	EXIT_IF(m_graphics != &graphics);
-	auto found = std::ranges::find_if(m_slots, [](const auto& slot) { return !slot.busy; });
-	auto* slot = found != m_slots.end() ? &*found : CreateSlot();
-	slot->busy = true;
+	auto  found = std::ranges::find_if(m_slots, [](const auto& slot) { return !slot.busy; });
+	auto* slot  = found != m_slots.end() ? &*found : CreateSlot();
+	slot->busy  = true;
 	slot->Reset();
 	return slot;
 }
@@ -99,7 +99,10 @@ bool CommandScheduler::InDeferredOperation() noexcept {
 
 CommandScheduler::CommandScheduler(RenderContext& context, GraphicContext& graphics)
     : m_master(graphics), m_context(context), m_graphics(graphics),
-      m_priority_thread([this](std::stop_token stop) { PriorityOperationsThread(stop); }) {}
+      m_priority_thread([this](std::stop_token stop) { PriorityOperationsThread(stop); }) {
+	m_buffers.reserve(CommandBufferGrowStep);
+	m_buffer_ticks.reserve(CommandBufferGrowStep);
+}
 
 CommandScheduler::~CommandScheduler() {
 	Shutdown();
@@ -154,10 +157,7 @@ void CommandScheduler::Begin(HW::Context& registers, HW::UserConfig& user_config
 	m_shaders     = &shaders;
 
 	if (!Active()) {
-		for (auto& buffer: m_buffers) {
-			buffer = std::make_unique<RenderCommandBuffer>(*this);
-		}
-		m_current = 0;
+		m_current = static_cast<int>(GrowCommandBuffers());
 	}
 
 	BindCurrent();
@@ -235,7 +235,13 @@ void CommandScheduler::Wait(uint64_t tick) {
 }
 
 void CommandScheduler::PopPendingOperations() {
-	m_master.Refresh();
+	PopPendingOperations(true);
+}
+
+void CommandScheduler::PopPendingOperations(bool refresh_gpu_tick) {
+	if (refresh_gpu_tick) {
+		m_master.Refresh();
+	}
 	for (;;) {
 		Common::UniqueFunction<void> callback;
 		{
@@ -331,8 +337,7 @@ void CommandScheduler::WaitPriorityOperations(uint64_t tick) {
 	EXIT_IF(g_deferred_callback_scheduler == this);
 	std::unique_lock lock(m_operation_mutex);
 	m_operation_available.wait(lock, [this, tick] {
-		const bool active_before_or_at =
-		    m_priority_active && m_priority_active_tick <= tick;
+		const bool active_before_or_at = m_priority_active && m_priority_active_tick <= tick;
 		const bool queued_before_or_at =
 		    !m_priority_operations.empty() && m_priority_operations.front().tick <= tick;
 		return !active_before_or_at && !queued_before_or_at;
@@ -363,7 +368,7 @@ uint64_t CommandScheduler::NextSubmitSequence() noexcept {
 }
 
 void CommandScheduler::CheckActive() const {
-	EXIT_IF(!Active() || m_current >= BufferCount);
+	EXIT_IF(!Active() || static_cast<size_t>(m_current) >= m_buffers.size());
 }
 
 RenderCommandBuffer& CommandScheduler::Current() const {
@@ -385,15 +390,52 @@ CommandBuffer& CommandScheduler::SubmitCurrent(SubmitInfo& submit) {
 	const auto signal_tick = m_master.NextTick();
 	submit.AddSignal(m_master.Handle(), signal_tick);
 	submitted.Execute(submit);
-	m_recording = false;
+	m_buffer_ticks[static_cast<size_t>(m_current)] = signal_tick;
+	m_recording                                    = false;
 	return submitted;
 }
 
+int CommandScheduler::FindReusableBuffer(uint64_t gpu_tick) const {
+	EXIT_IF(m_buffer_ticks.size() != m_buffers.size());
+	const auto buffer_count = m_buffers.size();
+	for (size_t offset = 1; offset <= buffer_count; ++offset) {
+		const auto candidate = (static_cast<size_t>(m_current) + offset) % buffer_count;
+		if (gpu_tick >= m_buffer_ticks[candidate]) {
+			return static_cast<int>(candidate);
+		}
+	}
+	return -1;
+}
+
+size_t CommandScheduler::GrowCommandBuffers() {
+	const auto first = m_buffers.size();
+	const auto end   = first + CommandBufferGrowStep;
+	m_buffers.reserve(end);
+	m_buffer_ticks.reserve(end);
+	for (size_t i = first; i < end; ++i) {
+		m_buffers.emplace_back(std::make_unique<RenderCommandBuffer>(*this));
+		m_buffer_ticks.push_back(0);
+	}
+	return first;
+}
+
 void CommandScheduler::BeginNext() {
 	EXIT_IF(m_recording);
-	m_current = (m_current + 1) % BufferCount;
+
+	auto candidate = FindReusableBuffer(m_master.KnownGpuTick());
+	bool refreshed = false;
+	if (candidate < 0) {
+		m_master.Refresh();
+		refreshed = true;
+		candidate = FindReusableBuffer(m_master.KnownGpuTick());
+	}
+	if (candidate < 0) {
+		candidate = static_cast<int>(GrowCommandBuffers());
+	}
+
+	m_current = candidate;
 	Current().WaitForFenceAndReset();
-	PopPendingOperations();
+	PopPendingOperations(!refreshed);
 	BindCurrent();
 	Current().Begin();
 	m_recording = true;

src/graphics/host_gpu/renderer/commandScheduler.h

@@ -16,6 +16,7 @@
 #include <queue>
 
 #include <thread>
+#include <vector>
 
 namespace Libs::Graphics {
 
@@ -31,8 +32,6 @@ struct CommandSlot {
 
 class CommandScheduler {
 public:
-	static constexpr int BufferCount = 8;
-
 	CommandScheduler(RenderContext& context, GraphicContext& graphics);
 	~CommandScheduler();
 	KYTY_CLASS_NO_COPY(CommandScheduler);
@@ -47,24 +46,26 @@ public:
 	void           FinishCurrent();
 	// Deferred callbacks can observe an externally owned drain, but cannot initiate shutdown:
 	// the priority runner cannot join itself.
-	void Shutdown();
-	void Wait(uint64_t tick);
-	void PopPendingOperations();
-	void DrainPriorityOperations();
-	void WaitPriorityOperations(uint64_t tick);
-	void DeferOperation(Common::UniqueFunction<void>&& operation);
-	void DeferPriorityOperation(Common::UniqueFunction<void>&& operation);
+	void                      Shutdown();
+	void                      Wait(uint64_t tick);
+	void                      PopPendingOperations();
+	void                      DrainPriorityOperations();
+	void                      WaitPriorityOperations(uint64_t tick);
+	void                      DeferOperation(Common::UniqueFunction<void>&& operation);
+	void                      DeferPriorityOperation(Common::UniqueFunction<void>&& operation);
 	[[nodiscard]] static bool InDeferredOperation() noexcept;
 
-	[[nodiscard]] bool     Active() const noexcept { return m_current >= 0; }
-	void                   CheckActive() const;
-	RenderCommandBuffer&   Current() const;
-	[[nodiscard]] uint64_t CurrentTick() const noexcept { return m_master.CurrentTick(); }
-	[[nodiscard]] bool     IsFree(uint64_t tick);
-	[[nodiscard]] RenderContext& Context() const noexcept { return m_context; }
+	[[nodiscard]] bool            Active() const noexcept { return m_current >= 0; }
+	void                          CheckActive() const;
+	RenderCommandBuffer&          Current() const;
+	[[nodiscard]] uint64_t        CurrentTick() const noexcept { return m_master.CurrentTick(); }
+	[[nodiscard]] bool            IsFree(uint64_t tick);
+	[[nodiscard]] RenderContext&  Context() const noexcept { return m_context; }
 	[[nodiscard]] GraphicContext& Graphics() const noexcept { return m_graphics; }
 
 private:
+	static constexpr size_t CommandBufferGrowStep = 4;
+
 	class CommandPool {
 	public:
 		CommandPool() = default;
@@ -91,33 +92,37 @@ private:
 		uint64_t                     tick = 0;
 	};
 
-	void           BindCurrent() const;
-	CommandBuffer& SubmitCurrent(SubmitInfo& submit);
-	void           BeginNext();
-	void           PriorityOperationsThread(std::stop_token stop);
-	void           RunOperation(Common::UniqueFunction<void>&& operation);
+	void                       BindCurrent() const;
+	CommandBuffer&             SubmitCurrent(SubmitInfo& submit);
+	void                       BeginNext();
+	void                       PopPendingOperations(bool refresh_gpu_tick);
+	[[nodiscard]] int          FindReusableBuffer(uint64_t gpu_tick) const;
+	[[nodiscard]] size_t       GrowCommandBuffers();
+	void                       PriorityOperationsThread(std::stop_token stop);
+	void                       RunOperation(Common::UniqueFunction<void>&& operation);
 	[[nodiscard]] CommandSlot* AllocateCommandBuffer();
 	[[nodiscard]] uint64_t     NextSubmitSequence() noexcept;
 
-	MasterSemaphore                                               m_master;
-	RenderContext&                                                m_context;
-	GraphicContext&                                               m_graphics;
-	CommandPool                                                   m_command_pool;
-	std::array<std::unique_ptr<RenderCommandBuffer>, BufferCount> m_buffers;
-	std::queue<PendingOperation>                                  m_pending_operations;
-	std::queue<PendingOperation>                                  m_priority_operations;
-	std::mutex                                                    m_operation_mutex;
-	std::condition_variable                                       m_operation_available;
-	std::jthread                                                  m_priority_thread;
-	bool                                                          m_priority_active = false;
-	uint64_t                                                      m_priority_active_tick = 0;
-	OperationState  m_operation_state = OperationState::Open;
-	int             m_current         = -1;
-	bool            m_recording       = false;
-	HW::Context*    m_registers       = nullptr;
-	HW::UserConfig* m_user_config     = nullptr;
-	HW::Shader*     m_shaders         = nullptr;
-	std::atomic<uint64_t> m_submit_sequence = 0;
+	MasterSemaphore                                   m_master;
+	RenderContext&                                    m_context;
+	GraphicContext&                                   m_graphics;
+	CommandPool                                       m_command_pool;
+	std::vector<std::unique_ptr<RenderCommandBuffer>> m_buffers;
+	std::vector<uint64_t>                             m_buffer_ticks;
+	std::queue<PendingOperation>                      m_pending_operations;
+	std::queue<PendingOperation>                      m_priority_operations;
+	std::mutex                                        m_operation_mutex;
+	std::condition_variable                           m_operation_available;
+	std::jthread                                      m_priority_thread;
+	bool                                              m_priority_active      = false;
+	uint64_t                                          m_priority_active_tick = 0;
+	OperationState                                    m_operation_state      = OperationState::Open;
+	int                                               m_current              = -1;
+	bool                                              m_recording            = false;
+	HW::Context*                                      m_registers            = nullptr;
+	HW::UserConfig*                                   m_user_config          = nullptr;
+	HW::Shader*                                       m_shaders              = nullptr;
+	std::atomic<uint64_t>                             m_submit_sequence      = 0;
 
 	friend class CommandBuffer;
 };

src/graphics/host_gpu/renderer/context.cpp

@@ -8,8 +8,8 @@
 #include "graphics/host_gpu/renderer/colorRenderTarget.h"
 #include "graphics/host_gpu/renderer/debug.h"
 #include "graphics/host_gpu/renderer/depthRenderTarget.h"
-#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/image/imageView.h"
+#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/vma.h"
@@ -270,30 +270,30 @@ void CommandBuffer::BeginRendering(const RenderState& state) const {
 		colors[i].sType        = vk::StructureType::eRenderingAttachmentInfo;
 		colors[i].imageView    = attachment.image_view;
 		colors[i].imageLayout  = attachment.image_layout;
-		colors[i].loadOp = attachment.is_clear ? vk::AttachmentLoadOp::eClear
-		                                      : vk::AttachmentLoadOp::eLoad;
-		colors[i].storeOp                  = vk::AttachmentStoreOp::eStore;
-		colors[i].clearValue.color.uint32  = attachment.clear_value;
+		colors[i].loadOp =
+		    attachment.is_clear ? vk::AttachmentLoadOp::eClear : vk::AttachmentLoadOp::eLoad;
+		colors[i].storeOp                 = vk::AttachmentStoreOp::eStore;
+		colors[i].clearValue.color.uint32 = attachment.clear_value;
 	}
 
-	const auto& depth_stencil = state.depth_stencil_attachment;
+	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 = depth_stencil.depth_clear ? vk::AttachmentLoadOp::eClear
-	                                        : vk::AttachmentLoadOp::eLoad;
-	depth.storeOp                        = vk::AttachmentStoreOp::eStore;
+	depth.loadOp =
+	    depth_stencil.depth_clear ? vk::AttachmentLoadOp::eClear : vk::AttachmentLoadOp::eLoad;
+	depth.storeOp                       = vk::AttachmentStoreOp::eStore;
 	depth.clearValue.depthStencil.depth = std::bit_cast<float>(depth_stencil.clear_value[0]);
 
 	vk::RenderingAttachmentInfo stencil {};
 	stencil.sType       = vk::StructureType::eRenderingAttachmentInfo;
 	stencil.imageView   = depth_stencil.image_view;
 	stencil.imageLayout = depth_stencil.image_layout;
-	stencil.loadOp = depth_stencil.stencil_clear ? vk::AttachmentLoadOp::eClear
-	                                            : vk::AttachmentLoadOp::eLoad;
-	stencil.storeOp                            = vk::AttachmentStoreOp::eStore;
-	stencil.clearValue.depthStencil.stencil   = depth_stencil.clear_value[1];
+	stencil.loadOp =
+	    depth_stencil.stencil_clear ? vk::AttachmentLoadOp::eClear : vk::AttachmentLoadOp::eLoad;
+	stencil.storeOp                         = vk::AttachmentStoreOp::eStore;
+	stencil.clearValue.depthStencil.stencil = depth_stencil.clear_value[1];
 
 	vk::RenderingInfo rendering {};
 	rendering.sType                = vk::StructureType::eRenderingInfo;

src/graphics/host_gpu/renderer/debug.cpp

@@ -10,6 +10,7 @@
 #include "graphics/host_gpu/renderer/render.h"
 
 #include <algorithm>
+#include <array>
 #include <atomic>
 #include <cmath>
 #include <fmt/format.h>
@@ -359,30 +360,12 @@ static void RtCheck(const HW::RenderTarget& rt) {
 				logged = true;
 			}
 		}
-		if (rt.attrib3.depth != 0x00000000) {
-			static bool logged = false;
-			if (!logged) {
-				LOGF("RenderTarget: temporary: ignoring PS5 color target depth_minus1=0x%08" PRIx32
-				     "\n",
-				     rt.attrib3.depth);
-				logged = true;
-			}
-		}
 		if (!RenderIsColorTileMode(rt.attrib3.tile_mode)) {
 			EXIT("unknown PS5 render-target tile mode: 0x%08" PRIx32 "\n", rt.attrib3.tile_mode);
 		}
 		if (!RenderIsColorDimension(rt.attrib3.dimension)) {
 			EXIT("unknown PS5 render-target dimension: 0x%08" PRIx32 "\n", rt.attrib3.dimension);
 		}
-		if (rt.attrib3.dimension != 0x00000001) {
-			static bool logged = false;
-			if (!logged) {
-				LOGF("RenderTarget: temporary: using 2D fallback for PS5 color "
-				     "dimension=0x%08" PRIx32 "\n",
-				     rt.attrib3.dimension);
-				logged = true;
-			}
-		}
 		if (!rt.attrib3.cmask_pipe_aligned) {
 			static bool logged = false;
 			if (!logged) {
@@ -497,7 +480,15 @@ static void ZPrint(const char* func, const HW::DepthRenderTarget& z) {
 }
 
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
-static void ZCheck(const HW::DepthRenderTarget& z) {
+static void ZCheck(const HW::DepthRenderTarget& z, const HW::DepthControl& dc,
+                   const HW::RenderControl& rc) {
+	const bool depth_active =
+	    dc.z_enable || dc.z_write_enable || dc.depth_bounds_enable || rc.depth_clear_enable;
+	const bool stencil_active = dc.stencil_enable || rc.stencil_clear_enable;
+	if (!depth_active && !stencil_active) {
+		return;
+	}
+
 	EXIT_NOT_IMPLEMENTED(!z.z_info.HasValidTextureCompatibility());
 	EXIT_NOT_IMPLEMENTED(!z.stencil_info.HasValidTextureCompatibility());
 	if (z.z_info.format == 0) {
@@ -548,14 +539,6 @@ static void ZCheck(const HW::DepthRenderTarget& z) {
 		EXIT_NOT_IMPLEMENTED(z.htile_surface.prefetch_height != 0x00000000);
 		EXIT_NOT_IMPLEMENTED(z.htile_surface.dst_outside_zero_to_one != 0x00000000);
 
-		if (z.depth_view.slice_start != 0x00000000 || z.depth_view.slice_max != 0x00000000) {
-			static std::atomic<uint32_t> log_count {0};
-			if (log_count.fetch_add(1, std::memory_order_relaxed) < 16) {
-				LOGF("DepthTarget: temporary: ignoring PS5 array slice view start=0x%08" PRIx32
-				     ", max=0x%08" PRIx32 "\n",
-				     z.depth_view.slice_start, z.depth_view.slice_max);
-			}
-		}
 		if (z.depth_view.current_mip_level != 0x00000000) {
 			static std::atomic<uint32_t> log_count {0};
 			if (log_count.fetch_add(1, std::memory_order_relaxed) < 16) {
@@ -704,14 +687,6 @@ static void McCheck(const HW::ModeControl& c) {
 	}
 	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.poly_offset_front_enable || c.poly_offset_back_enable) {
-		static bool logged = false;
-		if (!logged) {
-			LOGF("\t temporary: PA_SU_SC_MODE_CNTL.POLY_OFFSET_*_ENABLE is not implemented; "
-			     "continuing without depth bias\n");
-			logged = true;
-		}
-	}
 	if (c.vtx_window_offset_enable) {
 		static bool logged = false;
 		if (!logged) {
@@ -1085,18 +1060,24 @@ static ScissorRect ScissorRectClamp(ScissorRect r, uint32_t width, uint32_t heig
 	return r;
 }
 
+static constexpr std::array<uint16_t, 16> MakeScissorIntersectionRules() {
+	std::array<uint16_t, 16> rules {};
+	for (uint32_t candidate = 0; candidate < rules.size(); candidate++) {
+		for (uint32_t combination = 0; combination < 16; combination++) {
+			if ((combination & candidate) == candidate) {
+				rules[candidate] |= static_cast<uint16_t>(1u << combination);
+			}
+		}
+	}
+	return rules;
+}
+
 static bool ScissorClipRuleToIntersectionMask(uint16_t rule, uint8_t* mask) {
 	EXIT_IF(mask == nullptr);
 
-	for (uint32_t candidate = 0; candidate < 16; candidate++) {
-		uint16_t expected = 0;
-		for (uint32_t combination = 0; combination < 16; combination++) {
-			if ((combination & candidate) == candidate) {
-				expected |= static_cast<uint16_t>(1u << combination);
-			}
-		}
-
-		if (expected == rule) {
+	static constexpr auto rules = MakeScissorIntersectionRules();
+	for (uint32_t candidate = 0; candidate < rules.size(); candidate++) {
+		if (rules[candidate] == rule) {
 			*mask = static_cast<uint8_t>(candidate);
 			return true;
 		}
@@ -1214,7 +1195,7 @@ void hw_check(const RenderCommandBuffer& buffer) {
 	log_phase("vp");
 	VpCheck(vp, smc);
 	log_phase("z");
-	ZCheck(z);
+	ZCheck(z, d, rc);
 	log_phase("clip");
 	ClipCheck(c);
 	log_phase("rc");

src/graphics/host_gpu/renderer/depthRenderTarget.cpp

@@ -10,10 +10,10 @@
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/guest_gpu/tile.h"
 #include "graphics/host_gpu/graphicContext.h"
-#include "graphics/host_gpu/renderer/image/textureCommon.h"
 #include "graphics/host_gpu/renderer/debug.h"
-#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/image/imageView.h"
+#include "graphics/host_gpu/renderer/image/textureCommon.h"
+#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/vulkanCommon.h"
@@ -38,7 +38,10 @@ namespace Libs::Graphics {
 	EXIT("unsupported render state; details were printed above\n");
 }
 
-static vk::StencilOp ConvertStencilOp(uint8_t value) {
+static vk::StencilOp ConvertStencilOp(uint8_t value, uint8_t write_mask, uint8_t op_value) {
+	if (write_mask == 0) {
+		return vk::StencilOp::eKeep;
+	}
 	switch (static_cast<Prospero::StencilOp>(value)) {
 		case Prospero::StencilOp::kKeep: return vk::StencilOp::eKeep;
 		case Prospero::StencilOp::kZero: return vk::StencilOp::eZero;
@@ -49,7 +52,17 @@ static vk::StencilOp ConvertStencilOp(uint8_t value) {
 		case Prospero::StencilOp::kInvert: return vk::StencilOp::eInvert;
 		case Prospero::StencilOp::kAddWrap: return vk::StencilOp::eIncrementAndWrap;
 		case Prospero::StencilOp::kSubWrap: return vk::StencilOp::eDecrementAndWrap;
-		default: DepthFatal("unsupported stencil operation");
+		case Prospero::StencilOp::kXor:
+			if ((write_mask & op_value) == 0) {
+				return vk::StencilOp::eKeep;
+			}
+			if ((write_mask & ~op_value) != 0) {
+				DepthFatal("unsupported stencil XOR operands: write mask=0x%02" PRIx8
+				           ", operation value=0x%02" PRIx8,
+				           write_mask, op_value);
+			}
+			return vk::StencilOp::eInvert;
+		default: DepthFatal("unsupported stencil operation: 0x%02" PRIx8, value);
 	}
 }
 
@@ -96,9 +109,11 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, RenderCommandB
 	const auto& dc = hw.GetDepthControl();
 	const auto& sc = hw.GetStencilControl();
 	const auto& sm = hw.GetStencilMask();
-	const bool  depth_active =
+	const bool  has_stencil =
+	    z.stencil_info.format != Prospero::GpuEnumValue(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 = dc.stencil_enable || rc.stencil_clear_enable;
+	const bool stencil_active = has_stencil && (dc.stencil_enable || rc.stencil_clear_enable);
 	if (!depth_active && !stencil_active) {
 		return;
 	}
@@ -139,8 +154,6 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, RenderCommandB
 		}
 		return;
 	}
-	const bool has_stencil =
-	    z.stencil_info.format != Prospero::GpuEnumValue(Prospero::StencilFormat::kInvalid);
 	const bool has_htile = z.z_info.htile_acceleration;
 	const auto samples   = render_sample_count(z.z_info.num_samples);
 	if (samples == 0) {
@@ -150,16 +163,14 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, RenderCommandB
 	    has_stencil, has_htile, z.stencil_info.htile_stencil_disabled);
 	const auto view = ResolveTargetViewInfo(z.depth_view.slice_start, z.depth_view.slice_max);
 	switch (view.type) {
-		case TargetViewType::Image2D: break;
-		case TargetViewType::Image2DArray:
-			DepthFatal("layered depth views are unsupported: base=%u count=%u", view.base_layer,
-			           view.layer_count);
+		case TargetViewType::Image2D:
+		case TargetViewType::Image2DArray: break;
 		case TargetViewType::Unsupported:
 			DepthFatal("invalid depth view: base=%u last=%u", z.depth_view.slice_start,
 			           z.depth_view.slice_max);
 	}
-	if ((stencil_active && !has_stencil) || rc.resummarize_enable || rc.copy_centroid ||
-	    rc.copy_sample != 0 || z.z_info.expclear_enabled || z.stencil_info.expclear_enabled ||
+	if (rc.resummarize_enable || rc.copy_centroid || rc.copy_sample != 0 ||
+	    z.z_info.expclear_enabled || z.stencil_info.expclear_enabled ||
 	    z.z_info.partially_resident || z.stencil_info.partially_resident ||
 	    z.z_info.max_mip_level != 0 || z.depth_view.current_mip_level != 0 ||
 	    z.depth_info.addr5_swizzle_mask != 0 || z.depth_info.array_mode != 0 ||
@@ -281,10 +292,12 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, RenderCommandB
 	r.depth_min_bounds         = hw.GetDepthBoundsMin();
 	r.depth_max_bounds         = hw.GetDepthBoundsMax();
 
-	r.stencil_clear_enable = rc.stencil_clear_enable;
+	r.stencil_clear_enable = has_stencil && rc.stencil_clear_enable;
 	r.stencil_clear_value  = hw.GetStencilClearValue();
-	r.stencil_test_enable  = dc.stencil_enable;
-	if (dc.stencil_enable) {
+	r.stencil_test_enable  = has_stencil && dc.stencil_enable;
+	if (r.stencil_test_enable) {
+		const uint8_t front_write_mask = rc.stencil_clear_enable ? 0 : sm.stencil_writemask;
+		const uint8_t back_write_mask  = rc.stencil_clear_enable ? 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)) ||
@@ -296,19 +309,18 @@ void RenderExecutor::ResolveRenderDepthTarget(uint64_t submit_id, RenderCommandB
 			DepthFatal("unsupported stencil compare or replacement state");
 		}
 		r.stencil_static_front = {
-		    ConvertStencilOp(sc.stencil_fail), ConvertStencilOp(sc.stencil_zpass),
-		    ConvertStencilOp(sc.stencil_zfail), static_cast<vk::CompareOp>(dc.stencilfunc)};
-		r.stencil_dynamic_front = {sm.stencil_mask,
-		                           rc.stencil_clear_enable ? 0u : sm.stencil_writemask,
-		                           sm.stencil_testval};
+		    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};
 		if (dc.backface_enable) {
-			r.stencil_static_back  = {ConvertStencilOp(sc.stencil_fail_bf),
-			                          ConvertStencilOp(sc.stencil_zpass_bf),
-			                          ConvertStencilOp(sc.stencil_zfail_bf),
-			                          static_cast<vk::CompareOp>(dc.stencilfunc_bf)};
-			r.stencil_dynamic_back = {sm.stencil_mask_bf,
-			                          rc.stencil_clear_enable ? 0u : sm.stencil_writemask_bf,
-			                          sm.stencil_testval_bf};
+			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};
 		} else {
 			r.stencil_static_back  = r.stencil_static_front;
 			r.stencil_dynamic_back = r.stencil_dynamic_front;

src/graphics/host_gpu/renderer/depthRenderTarget.h

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

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

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

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

@@ -105,6 +105,10 @@ Image::Barriers Image::GetBarriers(vk::ImageLayout                      destinat
                                    std::optional<ImageSubresourceRange> range) {
 	auto& state              = backing.state;
 	auto& subresource_states = backing.subresource_states;
+	if (range && info.IsVolume()) {
+		range->base_layer  = 0;
+		range->layer_count = 1;
+	}
 
 	const bool partial =
 	    range && (range->base_level != 0 || range->level_count != info.resources.levels ||
@@ -659,7 +663,8 @@ Image::Image(GraphicContext& graphics, CommandScheduler& scheduler, const ImageI
     : info(image_info), m_graphics(&graphics), m_scheduler(&scheduler) {
 	KYTY_PROFILER_FUNCTION();
 	ImageOps::Validate(info);
-	m_cpu_dirty = !info.data.Empty();
+	m_cpu_dirty =
+	    !info.data.Empty() && info.metadata.compression == VideoOutCompression::Uncompressed;
 	if (info.pixel_format == vk::Format::eUndefined) {
 		return;
 	}

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

@@ -154,17 +154,18 @@ public:
 	}
 	[[nodiscard]] uint64_t HashGuestEdges() const;
 
-	ImageInfo      info;
-	VulkanImage    backing;
-	ImageViewCache views;
-	ImageUsage     usage;
-	ImageBinding   binding;
-	bool           registered     = false;
-	uint64_t       track_addr     = 0;
-	uint64_t       track_addr_end = 0;
-	ImageId        depth_id {};
-	uint64_t       tick_accessed_last = 0;
-	size_t         lru_id             = 0;
+	ImageInfo        info;
+	VulkanImage      backing;
+	ImageViewCache   views;
+	ImageUsage       usage;
+	ImageBinding     binding;
+	bool             registered     = false;
+	mutable uint32_t query_epoch    = 0;
+	uint64_t         track_addr     = 0;
+	uint64_t         track_addr_end = 0;
+	ImageId          depth_id {};
+	uint64_t         tick_accessed_last = 0;
+	size_t           lru_id             = 0;
 
 private:
 	friend struct ImageTestAccess;

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

@@ -19,10 +19,9 @@ 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 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;
@@ -47,10 +46,10 @@ struct ImageSubresources {
 };
 
 struct ImageSubresourceRange {
-	uint32_t base_level  = 0;
-	uint32_t level_count = 1;
-	uint32_t base_layer  = 0;
-	uint32_t layer_count = 1;
+	uint32_t base_level                                      = 0;
+	uint32_t level_count                                     = 1;
+	uint32_t base_layer                                      = 0;
+	uint32_t layer_count                                     = 1;
 	auto     operator<=>(const ImageSubresourceRange&) const = default;
 };
 
@@ -67,10 +66,10 @@ struct ImageInfo {
 	GuestRange                   stencil;
 	ImageMetadataInfo            metadata;
 	uint32_t                     htile_clear_mask = UINT32_MAX;
-	vk::Format                   pixel_format = vk::Format::eUndefined;
-	uint32_t                     guest_format = 0;
-	Prospero::ImageType          type         = Prospero::ImageType::kColor2D;
-	vk::Extent3D                 extent       = {1, 1, 1};
+	vk::Format                   pixel_format     = vk::Format::eUndefined;
+	uint32_t                     guest_format     = 0;
+	Prospero::ImageType          type             = Prospero::ImageType::kColor2D;
+	vk::Extent3D                 extent           = {1, 1, 1};
 	ImageSubresources            resources;
 	uint32_t                     pitch           = 0;
 	uint32_t                     bytes_per_block = 0;
@@ -352,8 +351,7 @@ inline bool ImageInfo::IsDepth() const noexcept {
 	}
 	const auto transfer_bytes = DepthAspectTransferBytes(info.pixel_format);
 	return transfer_bytes == info.bytes_per_block ||
-	       (info.bytes_per_block == sizeof(uint16_t) &&
-	        transfer_bytes == sizeof(uint32_t));
+	       (info.bytes_per_block == sizeof(uint16_t) && transfer_bytes == sizeof(uint32_t));
 }
 
 [[nodiscard]] inline VideoOutCompression
@@ -470,18 +468,13 @@ IsSupportedDisplayRenderTargetTileMode(uint32_t tile_mode) noexcept {
                                                  vk::ClearColorValue& clear) {
 	vk::ClearColorValue next {};
 	const auto unorm8 = [](uint32_t value) { return static_cast<float>(value & 0xffu) / 255.0f; };
-	const auto srgb8 = [](uint32_t value) {
+	const auto srgb8  = [](uint32_t value) {
 		const auto encoded = static_cast<float>(value & 0xffu) / 255.0f;
-		return encoded <= 0.04045f ? encoded / 12.92f
-		                           : std::pow((encoded + 0.055f) / 1.055f, 2.4f);
+		return encoded <= 0.04045f ? encoded / 12.92f : std::pow((encoded + 0.055f) / 1.055f, 2.4f);
 	};
 	switch (format) {
-		case vk::Format::eR32Uint:
-			next.uint32[0] = packed;
-			break;
-		case vk::Format::eR32Sint:
-			next.int32[0] = static_cast<int32_t>(packed);
-			break;
+		case vk::Format::eR32Uint: next.uint32[0] = packed; break;
+		case vk::Format::eR32Sint: next.int32[0] = static_cast<int32_t>(packed); break;
 		case vk::Format::eR8G8B8A8Srgb:
 			next.float32[0] = srgb8(packed);
 			next.float32[1] = srgb8(packed >> 8u);

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

@@ -70,15 +70,14 @@ namespace {
 			}
 		case vk::ImageType::e3D:
 			switch (info.type) {
-				case vk::ImageViewType::e3D:
-					return info.base_layer == 0 && info.layer_count == 1;
+				case vk::ImageViewType::e3D: return info.base_layer == 0 && info.layer_count == 1;
 				case vk::ImageViewType::e2D:
-					return static_cast<bool>(
-					           image.flags & vk::ImageCreateFlagBits::e2DArrayCompatible) &&
+					return static_cast<bool>(image.flags &
+					                         vk::ImageCreateFlagBits::e2DArrayCompatible) &&
 					       info.level_count == 1 && info.layer_count == 1;
 				case vk::ImageViewType::e2DArray:
-					return static_cast<bool>(
-					           image.flags & vk::ImageCreateFlagBits::e2DArrayCompatible) &&
+					return static_cast<bool>(image.flags &
+					                         vk::ImageCreateFlagBits::e2DArrayCompatible) &&
 					       info.level_count == 1;
 				default: return false;
 			}
@@ -325,11 +324,10 @@ bool FormatsCompatible(vk::Format base, vk::Format view) noexcept {
 } // namespace ImageViewOps
 
 vk::ImageView Image::FindView(const ImageViewInfo& view_info) {
-	const auto& image = backing;
+	const auto& image      = backing;
 	auto        normalized = view_info;
-	const bool  is_storage =
-	    static_cast<bool>(normalized.usage & vk::ImageUsageFlagBits::eStorage);
-	normalized.aspect = FullAspectMask(image.format);
+	const bool  is_storage = static_cast<bool>(normalized.usage & vk::ImageUsageFlagBits::eStorage);
+	normalized.aspect      = FullAspectMask(image.format);
 	if (normalized.aspect & vk::ImageAspectFlagBits::eDepth &&
 	    IsDepthViewFormat(normalized.format)) {
 		normalized.format = image.format;
@@ -340,28 +338,26 @@ vk::ImageView Image::FindView(const ImageViewInfo& view_info) {
 		normalized.format = image.format;
 		normalized.aspect = vk::ImageAspectFlagBits::eStencil;
 	}
-	normalized.usage =
-	    is_storage ? vk::ImageUsageFlagBits::eStorage : vk::ImageUsageFlags {};
+	normalized.usage = is_storage ? vk::ImageUsageFlagBits::eStorage : vk::ImageUsageFlags {};
 	const bool format_compatible = normalized.format != vk::Format::eUndefined &&
 	                               IsCompatibleViewFormat(image.format, normalized.format);
-	const bool slice_view = image.image_type == vk::ImageType::e3D &&
-	                        (normalized.type == vk::ImageViewType::e2D ||
-	                         normalized.type == vk::ImageViewType::e2DArray);
+	const bool slice_view =
+	    image.image_type == vk::ImageType::e3D && (normalized.type == vk::ImageViewType::e2D ||
+	                                               normalized.type == vk::ImageViewType::e2DArray);
 	const bool levels_valid = normalized.level_count != 0 &&
 	                          normalized.base_level < image.mip_levels &&
 	                          normalized.level_count <= image.mip_levels - normalized.base_level;
-	const auto view_layers = slice_view && levels_valid
-	                             ? std::max(image.extent.depth >> normalized.base_level, 1u)
-	                             : image.layers;
-	const bool ranges_valid = levels_valid &&
-	                          normalized.layer_count != 0 && normalized.base_layer < view_layers &&
+	const auto view_layers  = slice_view && levels_valid
+	                              ? std::max(image.extent.depth >> normalized.base_level, 1u)
+	                              : image.layers;
+	const bool ranges_valid = levels_valid && normalized.layer_count != 0 &&
+	                          normalized.base_layer < view_layers &&
 	                          normalized.layer_count <= view_layers - normalized.base_layer;
 	const bool mapping_valid =
 	    IsComponentSwizzle(normalized.mapping.r) && IsComponentSwizzle(normalized.mapping.g) &&
 	    IsComponentSwizzle(normalized.mapping.b) && IsComponentSwizzle(normalized.mapping.a);
 	if (image.image == nullptr || !format_compatible || !ranges_valid || !mapping_valid ||
-	    !IsValidViewType(image, normalized) ||
-	    !IsValidAspect(image, normalized.aspect)) {
+	    !IsValidViewType(image, normalized) || !IsValidAspect(image, normalized.aspect)) {
 		EXIT("invalid image view: image_format=%d view_format=%d type=%d aspect=0x%x "
 		     "mip=%u+%u layer=%u+%u usage=0x%x image_levels=%u image_layers=%u\n",
 		     static_cast<int>(image.format), static_cast<int>(normalized.format),

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

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

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

@@ -22,6 +22,10 @@ struct RenderTargetFormatMapping {
 };
 
 constexpr RenderTargetFormatMapping kRenderTargetFormats[] = {
+    {Prospero::ChannelLayout::k8,
+     Prospero::ChannelType::kUInt,
+     Prospero::ChannelOrder::kStandard,
+     {vk::Format::eR8Uint, 1}},
     {Prospero::ChannelLayout::k8_8,
      Prospero::ChannelType::kUNorm,
      Prospero::ChannelOrder::kStandard,
@@ -70,6 +74,10 @@ constexpr RenderTargetFormatMapping kRenderTargetFormats[] = {
      Prospero::ChannelType::kFloat,
      Prospero::ChannelOrder::kStandard,
      {vk::Format::eB10G11R11UfloatPack32, 4}},
+    {Prospero::ChannelLayout::k5_6_5,
+     Prospero::ChannelType::kUNorm,
+     Prospero::ChannelOrder::kStandard,
+     {vk::Format::eB5G6R5UnormPack16, 2}},
     {Prospero::ChannelLayout::k16,
      Prospero::ChannelType::kUNorm,
      Prospero::ChannelOrder::kStandard,
@@ -160,14 +168,6 @@ RenderTargetFormatInfo TextureGetRenderTargetFormat(uint32_t raw_layout, uint32_
 
 namespace {
 
-static uint64_t GetLevelSrcOffset(const TileSizeOffset& level_size) {
-	return (level_size.src_size != 0 ? level_size.src_offset : level_size.offset);
-}
-
-static uint64_t GetLevelSrcSize(const TileSizeOffset& level_size) {
-	return (level_size.src_size != 0 ? level_size.src_size : level_size.size);
-}
-
 static uint32_t GetTextureLevelDepth(uint32_t depth, uint32_t level, bool volume_texture) {
 	return volume_texture ? std::max(depth >> level, 1u) : depth;
 }
@@ -180,20 +180,6 @@ static size_t GetTextureRegionCount(uint32_t depth, uint64_t levels, bool volume
 	return count;
 }
 
-uint64_t TextureUploadSliceSourceOffset(const TextureUploadLayout& layout, uint32_t level,
-                                        uint32_t slice) {
-	if (level >= 16 || layout.level_sizes[level].size == 0) {
-		EXIT("invalid texture upload slice source, level=%u slice=%u\n", level, slice);
-	}
-	const auto level_offset = GetLevelSrcOffset(layout.level_sizes[level]);
-	const auto slice_stride =
-	    layout.source_slice_stride != 0 ? layout.source_slice_stride : layout.slice_stride;
-	if (slice_stride != 0 && slice > (UINT64_MAX - level_offset) / slice_stride) {
-		EXIT("texture upload slice source offset overflow, level=%u slice=%u\n", level, slice);
-	}
-	return level_offset + static_cast<uint64_t>(slice) * slice_stride;
-}
-
 vk::ComponentSwizzle TextureGetComponentSwizzle(uint8_t s) {
 	switch (static_cast<Prospero::CompSwizzle>(s)) {
 		case Prospero::CompSwizzle::kZero: return vk::ComponentSwizzle::eZero;
@@ -233,12 +219,11 @@ vk::Format TextureGetFormat(uint32_t fmt) {
 
 namespace {
 
-uint64_t CalcTextureSliceStride(const TileSizeOffset* level_sizes, uint64_t levels,
+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, static_cast<uint64_t>(level_sizes[i].offset) + level_sizes[i].size);
+		stride = std::max(stride, mips[i].offset + mips[i].size);
 	}
 
 	if (depth > 1 && total_size != 0 && total_size % depth == 0) {
@@ -251,39 +236,26 @@ uint64_t CalcTextureSliceStride(const TileSizeOffset* level_sizes, uint64_t leve
 	return stride;
 }
 
-uint64_t CalcLinearUploadLevelSize(uint32_t fmt, uint32_t pitch, uint32_t height) {
-	if (const uint32_t bytes_per_element = Prospero::NumBytesPerElement(fmt);
-	    bytes_per_element != 0) {
-		return static_cast<uint64_t>(pitch) * height * bytes_per_element;
-	}
-
-	if (const uint32_t bytes_per_block = Prospero::BlockCompressedBytesPerBlock(fmt);
-	    bytes_per_block != 0) {
-		const uint32_t blocks_w = std::max((pitch + 3u) / 4u, 1u);
-		const uint32_t blocks_h = std::max((height + 3u) / 4u, 1u);
-		return static_cast<uint64_t>(blocks_w) * blocks_h * bytes_per_block;
-	}
-
-	return 0;
+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(TileSizeOffset* level_sizes, uint32_t fmt, uint64_t height,
+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(fmt, pitch, h);
-		EXIT_NOT_IMPLEMENTED(size > 0xffffffffull);
-		EXIT_NOT_IMPLEMENTED(offset > 0xffffffffull);
-
-		level_sizes[i].size       = static_cast<uint32_t>(size);
-		level_sizes[i].offset     = static_cast<uint32_t>(offset);
-		level_sizes[i].src_size   = 0;
-		level_sizes[i].src_offset = 0;
-		level_sizes[i].x          = 0;
-		level_sizes[i].y          = 0;
+		const auto size = CalcLinearUploadLevelSize(element, pitch, h);
+		mips[i].size    = size;
+		mips[i].offset  = offset;
 
 		offset += size;
 		if (pitch > 1) {
@@ -299,143 +271,123 @@ uint64_t SetLinearUploadLevels(TileSizeOffset* level_sizes, uint32_t fmt, uint64
 
 } // namespace
 
-TextureUploadLayout TextureCalcUploadLayout(uint32_t fmt, uint64_t width, uint64_t height,
-                                            uint64_t levels, uint32_t depth, uint64_t pitch,
-                                            uint64_t tile, uint64_t upload_size,
-                                            bool allow_depth_tile, bool volume_texture,
-                                            const char* owner) {
+TextureUploadLayout TextureCalcUploadLayout(uint32_t fmt, uint32_t width, uint32_t height,
+                                            uint32_t levels, uint32_t depth, uint32_t tile,
+                                            uint64_t upload_size, bool allow_depth_tile,
+                                            bool volume_texture, const char* owner) {
 	TextureUploadLayout layout {};
-	layout.tile           = static_cast<uint32_t>(tile);
-	layout.pitch          = static_cast<uint32_t>(pitch);
-	layout.texel_block    = Prospero::BlockCompressedBytesPerBlock(fmt) != 0 ? 4u : 1u;
-	layout.volume_texture = volume_texture;
+	layout.surface.description = {
+	    fmt,
+	    tile,
+	    volume_texture ? TileSurfaceDimension::Dim3D : TileSurfaceDimension::Dim2D,
+	    width,
+	    height,
+	    volume_texture ? depth : 1u,
+	    levels,
+	    volume_texture ? 1u : depth,
+	};
+	const auto&              description = layout.surface.description;
+	const auto               tile_mode   = static_cast<Prospero::TileMode>(description.tile_mode);
+	TileTextureElementLayout element {};
 
-	if (fmt != 0) {
-		if (layout.tile != 0) {
-			const auto tile_mode = static_cast<Prospero::TileMode>(layout.tile);
-			switch (tile_mode) {
-				case Prospero::TileMode::kStandard256B:
-					if (TileIsStandard256BTextureSupported(fmt)) {
-						layout.tile_family = TileBlockFamily::Standard256B;
-					}
-					break;
-				case Prospero::TileMode::kStandard4KB:
-					if (TileIsStandard4KBTextureSupported(fmt)) {
-						layout.tile_family = TileBlockFamily::Standard4KB;
-					}
-					break;
-				case Prospero::TileMode::kStandard64KB:
-					if (TileIsStandard64KBTextureSupported(fmt)) {
-						layout.tile_family = TileBlockFamily::Standard64KB;
-					}
-					break;
-				case Prospero::TileMode::kPrt:
-					if (TileIsStandard64KBTextureSupported(fmt)) {
-						layout.tile_family = TileBlockFamily::Prt64KB;
-					}
-					break;
-				case Prospero::TileMode::kRenderTarget:
-					if (Prospero::RenderTargetBytesPerElement(fmt) != 0) {
-						layout.tile_family = TileBlockFamily::RenderTarget64KB;
-					}
-					break;
-				case Prospero::TileMode::kDepth:
-					if (allow_depth_tile && Prospero::RenderTargetBytesPerElement(fmt) != 0) {
-						layout.tile_family = TileBlockFamily::Depth64KB;
-					}
-					break;
-				default: break;
-			}
-			if (layout.tile_family == TileBlockFamily::Count) {
-				EXIT("%s: unsupported typed tiled upload: fmt=%u tile=%u "
-				     "size=%" PRIu64 " extent=%" PRIu64 "x%" PRIu64 " pitch=%" PRIu64
-				     " levels=%" PRIu64 "\n",
-				     owner, static_cast<uint32_t>(fmt), layout.tile, upload_size, width, height,
-				     pitch, levels);
-			}
-		}
-
-		layout.pitch = TileGetTexturePitch(fmt, width, levels, layout.tile);
-
-		TileGetTextureSize(fmt, width, height, layout.pitch, levels, layout.tile, nullptr,
-		                   layout.level_sizes, layout.padded_sizes);
-
-		if (static_cast<Prospero::TileMode>(layout.tile) != Prospero::TileMode::kLinear) {
-			if (layout.volume_texture) {
-				layout.slice_stride = SetLinearUploadLevels(layout.level_sizes, fmt, height, levels,
-				                                            static_cast<uint32_t>(width));
-			} else {
-				TileSizeOffset tiled_levels[16] {};
-				std::copy_n(layout.level_sizes, levels, tiled_levels);
-				layout.source_slice_stride =
-				    CalcTextureSliceStride(tiled_levels, levels, upload_size, depth);
-				SetLinearUploadLevels(layout.level_sizes, fmt, height, levels, layout.pitch);
-				for (uint32_t i = 0; i < levels; ++i) {
-					if (tiled_levels[i].src_size > tiled_levels[i].size) {
-						layout.first_tail_level = std::min(layout.first_tail_level, i);
-					}
-					layout.level_sizes[i].src_offset = GetLevelSrcOffset(tiled_levels[i]);
-					layout.level_sizes[i].src_size   = GetLevelSrcSize(tiled_levels[i]);
-					layout.level_sizes[i].x          = tiled_levels[i].x;
-					layout.level_sizes[i].y          = tiled_levels[i].y;
-				}
-			}
-		} else if (layout.volume_texture) {
-			layout.slice_stride =
-			    CalcTextureSliceStride(layout.level_sizes, levels, upload_size, depth);
-		}
-	} else {
+	if (fmt == 0) {
 		EXIT("%s: legacy texture upload format unsupported: fmt=0 tile=%u size=%" PRIu64
-		     " extent=%" PRIu64 "x%" PRIu64 " pitch=%" PRIu64 " levels=%" PRIu64 "\n",
-		     owner, layout.tile, upload_size, width, height, pitch, levels);
+		     " extent=%ux%u levels=%u\n",
+		     owner, description.tile_mode, upload_size, width, height, levels);
 	}
 
-	if (!layout.volume_texture) {
-		layout.slice_stride =
-		    CalcTextureSliceStride(layout.level_sizes, levels, upload_size, depth);
+	switch (tile_mode) {
+		case Prospero::TileMode::kLinear:
+			if (!TileGetTextureElementLayout(fmt, element)) {
+				EXIT("%s: unsupported linear texture format: fmt=%u\n", owner, fmt);
+			}
+			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>(fmt), 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>(fmt), description.tile_mode, upload_size, width,
+				     height, levels);
+			}
+			break;
+	}
+
+	if (tile_mode != Prospero::TileMode::kLinear) {
+		element = {layout.surface.texture.block.bytes_per_element,
+		           layout.surface.texture.texel_width, layout.surface.texture.texel_height};
+	}
+	layout.pitch = TileGetTexturePitch(fmt, width, description.tile_mode);
+	if (tile_mode == Prospero::TileMode::kLinear) {
+		TileSizeOffset level_sizes[16] {};
+		TilePaddedSize padded_sizes[16] {};
+		TileGetTextureSize(fmt, width, height, levels, description.tile_mode, nullptr, level_sizes,
+		                   padded_sizes);
+		for (uint32_t level = 0; level < levels; ++level) {
+			layout.mips[level] = {level_sizes[level].offset, level_sizes[level].size,
+			                      padded_sizes[level].width, padded_sizes[level].height};
+		}
+	} else if (volume_texture) {
+		layout.slice_stride = SetLinearUploadLevels(layout.mips, element, height, levels, width);
+	} else {
+		layout.source_slice_stride = layout.surface.block_slice_size;
+		if (depth > 1 && upload_size != 0 && upload_size % depth == 0) {
+			layout.source_slice_stride = std::max(layout.source_slice_stride, upload_size / depth);
+		}
+		SetLinearUploadLevels(layout.mips, element, height, levels, layout.pitch);
+	}
+
+	if (!volume_texture || tile_mode == Prospero::TileMode::kLinear) {
+		layout.slice_stride = CalcTextureSliceStride(layout.mips, levels, upload_size, depth);
 	}
 	return layout;
 }
 
-std::vector<vk::BufferImageCopy>
-TextureBuildImageCopies(const TextureUploadLayout& layout, uint32_t width, uint32_t height,
-                        uint32_t depth, uint64_t levels, bool array_texture,
-                        bool volume_texture) {
-	uint32_t mip_width  = width;
-	uint32_t mip_height = height;
-	uint32_t mip_pitch  = volume_texture && static_cast<Prospero::TileMode>(layout.tile) !=
-	                                            Prospero::TileMode::kLinear
-	                          ? width
-	                          : layout.pitch;
+std::vector<vk::BufferImageCopy> TextureBuildImageCopies(const TextureUploadLayout& layout) {
+	const auto& description    = layout.surface.description;
+	const bool  volume_texture = description.dimension == TileSurfaceDimension::Dim3D;
+	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 && static_cast<Prospero::TileMode>(description.tile_mode) !=
+	                                           Prospero::TileMode::kLinear
+	                         ? description.width
+	                         : layout.pitch;
 
 	std::vector<vk::BufferImageCopy> regions;
-	regions.reserve(GetTextureRegionCount(depth, levels, volume_texture));
-	for (uint32_t i = 0; i < levels; i++) {
-		EXIT_NOT_IMPLEMENTED(layout.level_sizes[i].size == 0);
+	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;
+			const auto          slice_offset = z * layout.slice_stride;
 			vk::BufferImageCopy region {};
-			region.bufferOffset =
-			    layout.level_sizes[i].offset + slice_offset;
-			region.imageSubresource = {vk::ImageAspectFlagBits::eColor, i,
-			                           array_texture ? z : 0, 1};
-			region.imageOffset.z = volume_texture ? static_cast<int>(z) : 0;
-			region.imageExtent   = {mip_width, mip_height, 1};
-			const bool linear =
-			    static_cast<Prospero::TileMode>(layout.tile) == Prospero::TileMode::kLinear;
+			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       = static_cast<Prospero::TileMode>(description.tile_mode) ==
+			                          Prospero::TileMode::kLinear;
 			if (linear) {
-				region.bufferRowLength   = layout.padded_sizes[i].width;
-				region.bufferImageHeight = layout.padded_sizes[i].height;
+				region.bufferRowLength   = layout.mips[i].row_length;
+				region.bufferImageHeight = layout.mips[i].image_height;
 			} else {
-				const auto align = [](uint32_t value, uint32_t block) {
+				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, layout.texel_block);
-				region.bufferRowLength =
-				    pitch > align(mip_width, layout.texel_block) ? pitch : 0;
+				const auto pitch       = align(mip_pitch, texel_width);
+				region.bufferRowLength = pitch > align(mip_width, texel_width) ? pitch : 0;
 			}
 			regions.push_back(region);
 		}
@@ -454,24 +406,6 @@ TextureBuildImageCopies(const TextureUploadLayout& layout, uint32_t width, uint3
 	return regions;
 }
 
-struct GpuTileElementLayout {
-	uint32_t bytes = 0;
-	uint32_t wide  = 1;
-	uint32_t tall  = 1;
-};
-
-static bool GetGpuTileElementLayout(uint32_t fmt, GpuTileElementLayout& out) {
-	if (const auto bytes = Prospero::NumBytesPerElement(fmt); bytes != 0) {
-		out = {bytes, 1, 1};
-		return true;
-	}
-	if (const auto bytes = Prospero::BlockCompressedBytesPerBlock(fmt); bytes != 0) {
-		out = {bytes, 4, 4};
-		return true;
-	}
-	return false;
-}
-
 static bool SetGpuTileSize(uint64_t offset, uint64_t length, uint64_t capacity, uint64_t& size) {
 	if (offset > capacity || length > capacity - offset) {
 		return false;
@@ -480,148 +414,122 @@ static bool SetGpuTileSize(uint64_t offset, uint64_t length, uint64_t capacity,
 	return true;
 }
 
-bool TextureBuildGpuTileInfos(uint64_t size,
-                              const std::vector<vk::BufferImageCopy>& regions,
-                              const TextureUploadLayout& layout, uint32_t fmt, uint32_t depth,
-                              uint64_t levels, std::vector<GpuTileInfo>& out_infos) {
-	if (size == 0 || levels == 0 || levels > 16 || depth == 0 ||
-	    regions.size() != GetTextureRegionCount(depth, levels, layout.volume_texture) ||
-	    Prospero::IsFmaskTextureFormat(fmt)) {
+bool TextureBuildGpuTileInfos(uint64_t tiled_size, const std::vector<vk::BufferImageCopy>& regions,
+                              const TextureUploadLayout& layout, uint32_t levels,
+                              std::vector<GpuTileInfo>& out_infos) {
+	const auto& description    = layout.surface.description;
+	const bool  volume_texture = description.dimension == TileSurfaceDimension::Dim3D;
+	const auto  depth          = volume_texture ? description.depth : description.layers;
+	if (tiled_size == 0 || levels == 0 || levels > 16 || depth == 0 ||
+	    regions.size() != GetTextureRegionCount(depth, levels, volume_texture) ||
+	    Prospero::IsFmaskTextureFormat(description.format)) {
 		return false;
 	}
 
-	GpuTileElementLayout element {};
-	if (layout.tile_family == TileBlockFamily::RenderTarget64KB ||
-	    layout.tile_family == TileBlockFamily::Depth64KB) {
-		element.bytes = Prospero::RenderTargetBytesPerElement(fmt);
-	} else if (!GetGpuTileElementLayout(fmt, element)) {
-		return false;
-	}
-	if (element.bytes == 0) {
+	const auto& surface = layout.surface;
+	const auto& texture = surface.texture;
+	const auto& block   = texture.block;
+	if (surface.description.levels < levels || block.bytes_per_element == 0 ||
+	    block.family == TileBlockFamily::Count) {
 		return false;
 	}
 
 	std::vector<GpuTileInfo> infos;
 	infos.reserve(regions.size());
-	if (layout.volume_texture) {
-		TileVolumeLayout volume {};
-		if (!TileGetTextureVolumeLayout(fmt, regions[0].imageExtent.width,
-		                                regions[0].imageExtent.height, depth,
-		                                static_cast<uint32_t>(levels), layout.tile, volume)) {
-			return false;
-		}
-		element = {volume.bytes_per_element, volume.texel_width, volume.texel_height};
-		TileBlockLayout block {};
-		if (!TileGetBlockLayout(volume.family, element.bytes, block)) return false;
-
+	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 >= volume.first_tail_level;
+			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 {};
+				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) * volume.block_slice_size +
-				    volume.level_offsets[level];
+				    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.level_sizes[level].size;
-				if (!SetGpuTileSize(info.linear_offset, linear_span, size, info.linear_size) ||
-				    !SetGpuTileSize(info.tiled_offset, volume.level_sizes[level], size,
+				    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 + element.wide - 1u) / element.wide, 1u);
+				info.width  = std::max((region.imageExtent.width + texture.texel_width - 1u) /
+				                           texture.texel_width,
+				                       1u);
 				info.height = std::max(
-				    (logical_height + element.tall - 1u) / element.tall, 1u);
-				info.depth       = copy_depth;
-				info.surface_z = block.block_depth == 1
-				                     ? static_cast<uint32_t>(region.imageOffset.z)
-				                     : 0;
-				info.pitch =
-				    std::max((pitch + element.wide - 1u) / element.wide, 1u);
-				info.tail_x      = tail ? volume.tail_x[level] : 0;
-				info.tail_y      = tail ? volume.tail_y[level] : 0;
-				info.tail        = tail;
-				info.tiled_width = volume.level_widths[level];
-				info.tiled_height = volume.level_heights[level];
+				    (logical_height + texture.texel_height - 1u) / texture.texel_height, 1u);
+				info.depth = copy_depth;
+				info.surface_z =
+				    block.block_depth == 1 ? static_cast<uint32_t>(region.imageOffset.z) : 0;
+				info.pitch = std::max((pitch + 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 {
-		const auto base_family = layout.tile_family;
-		if (base_family == TileBlockFamily::Count) {
-			return false;
-		}
-
 		size_t region_index = 0;
 		for (uint32_t level = 0; level < levels; level++) {
-			const auto&     level_size = layout.level_sizes[level];
-			const bool      tail       = level >= layout.first_tail_level;
-			const auto      family     = base_family;
-			TileBlockLayout block {};
-			if (!TileGetBlockLayout(family, element.bytes, block)) {
-				return false;
-			}
-			const auto level_depth = GetTextureLevelDepth(depth, level, layout.volume_texture);
+			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;
+				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      = TextureUploadSliceSourceOffset(layout, level, z);
-				if (!SetGpuTileSize(info.linear_offset, level_size.size, size, info.linear_size) ||
-				    !SetGpuTileSize(info.tiled_offset, GetLevelSrcSize(level_size), size,
-				                    info.tiled_size)) {
+				info.family              = block.family;
+				info.bytes_per_element   = block.bytes_per_element;
+				info.linear_offset       = region.bufferOffset;
+				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) {
 					return false;
 				}
-				info.width = std::max(
-				    (region.imageExtent.width + element.wide - 1u) / element.wide, 1u);
+				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 + element.tall - 1u) / element.tall, 1u);
-				info.surface_z = base_family == TileBlockFamily::RenderTarget64KB ||
-				                         base_family == TileBlockFamily::Depth64KB
+				    (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 + element.wide - 1u) / element.wide, 1u);
-				info.tail      = tail;
-				info.tail_x    = tail ? level_size.x : 0;
-				info.tail_y    = tail ? level_size.y : 0;
-				info.tiled_width =
-				    layout.padded_sizes[level].width != 0
-				        ? std::max((layout.padded_sizes[level].width + element.wide - 1u) /
-				                       element.wide,
-				                   1u)
-				        : info.pitch;
-				info.tiled_height =
-				    layout.padded_sizes[level].height != 0
-				        ? std::max((layout.padded_sizes[level].height + element.tall - 1u) /
-				                       element.tall,
-				                   1u)
-				        : info.height;
+				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);
 			}
 		}

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

@@ -19,35 +19,32 @@ struct RenderTargetFormatInfo {
 	Prospero::ColorComponentMapping export_mapping;
 };
 
-struct TextureUploadLayout {
-	uint32_t        tile                = 0;
-	uint32_t        pitch               = 0;
-	uint32_t        texel_block         = 1;
-	uint64_t        slice_stride        = 0;
-	uint64_t        source_slice_stride = 0;
-	uint32_t        first_tail_level    = 16;
-	TileBlockFamily tile_family         = TileBlockFamily::Count;
-	bool            volume_texture      = false;
-	TileSizeOffset  level_sizes[16]     = {};
-	TilePaddedSize  padded_sizes[16]    = {};
+struct TextureUploadMipLayout {
+	uint64_t offset       = 0;
+	uint64_t size         = 0;
+	uint32_t row_length   = 0;
+	uint32_t image_height = 0;
 };
 
-vk::ComponentMapping TextureGetComponentMapping(uint32_t swizzle);
+struct TextureUploadLayout {
+	uint32_t               pitch               = 0;
+	uint64_t               slice_stride        = 0;
+	uint64_t               source_slice_stride = 0;
+	TileSurfaceLayout      surface;
+	TextureUploadMipLayout mips[16] = {};
+};
+
+vk::ComponentMapping   TextureGetComponentMapping(uint32_t swizzle);
 vk::Format             TextureGetFormat(uint32_t fmt);
 RenderTargetFormatInfo TextureGetRenderTargetFormat(uint32_t layout, uint32_t type, uint32_t order);
-TextureUploadLayout TextureCalcUploadLayout(uint32_t fmt, uint64_t width, uint64_t height,
-                                            uint64_t levels, uint32_t depth, uint64_t pitch,
-                                            uint64_t tile, uint64_t upload_size,
-                                            bool allow_depth_tile, bool volume_texture,
-                                            const char* owner);
-std::vector<vk::BufferImageCopy>
-TextureBuildImageCopies(const TextureUploadLayout& layout, uint32_t width, uint32_t height,
-                        uint32_t depth, uint64_t levels, bool array_texture,
-                        bool volume_texture);
-bool TextureBuildGpuTileInfos(uint64_t size,
-                              const std::vector<vk::BufferImageCopy>& regions,
-                              const TextureUploadLayout& layout, uint32_t fmt, uint32_t depth,
-                              uint64_t levels, std::vector<GpuTileInfo>& infos);
+TextureUploadLayout    TextureCalcUploadLayout(uint32_t fmt, uint32_t width, uint32_t height,
+                                               uint32_t levels, uint32_t depth, uint32_t tile,
+                                               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);
 
 } // namespace Libs::Graphics
 

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

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

src/graphics/host_gpu/renderer/masterSemaphore.cpp

@@ -26,8 +26,8 @@ MasterSemaphore::~MasterSemaphore() {
 }
 
 void MasterSemaphore::Refresh() {
-	uint64_t counter = 0;
-	const auto result = m_graphics.device.getSemaphoreCounterValue(m_semaphore, &counter);
+	uint64_t   counter = 0;
+	const auto result  = m_graphics.device.getSemaphoreCounterValue(m_semaphore, &counter);
 	EXIT_NOT_IMPLEMENTED(result != vk::Result::eSuccess);
 
 	auto known = m_gpu_tick.load(std::memory_order_acquire);

src/graphics/host_gpu/renderer/masterSemaphore.h

@@ -22,7 +22,7 @@ public:
 	[[nodiscard]] uint64_t KnownGpuTick() const noexcept {
 		return m_gpu_tick.load(std::memory_order_acquire);
 	}
-	[[nodiscard]] bool IsFree(uint64_t tick) const noexcept { return KnownGpuTick() >= tick; }
+	[[nodiscard]] bool     IsFree(uint64_t tick) const noexcept { return KnownGpuTick() >= tick; }
 	[[nodiscard]] uint64_t NextTick() noexcept {
 		return m_current_tick.fetch_add(1, std::memory_order_release);
 	}

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

@@ -26,11 +26,15 @@ bool IsSampledImage(BindingKind kind) {
 		case BindingKind::Sampled1DArray:
 		case BindingKind::Sampled2D:
 		case BindingKind::Sampled2DArray:
+		case BindingKind::Sampled2DMsaa:
+		case BindingKind::Sampled2DMsaaArray:
 		case BindingKind::Sampled3D:
 		case BindingKind::SampledUint1D:
 		case BindingKind::SampledUint1DArray:
 		case BindingKind::SampledUint2D:
 		case BindingKind::SampledUint2DArray:
+		case BindingKind::SampledUint2DMsaa:
+		case BindingKind::SampledUint2DMsaaArray:
 		case BindingKind::SampledUint3D: return true;
 		default: return false;
 	}

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

@@ -95,7 +95,7 @@ private:
 	};
 
 	static vk::DescriptorImageInfo MakeImageInfo(const TextureBinding& texture);
-	void                 CreatePool();
+	void                           CreatePool();
 	VulkanDescriptorSet* Allocate(Stage stage, const ShaderRecompiler::IR::Program& program);
 	vk::DescriptorSetLayout
 	GetDescriptorSetLayoutInternal(Stage stage, const ShaderRecompiler::IR::Program& program);

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

@@ -73,6 +73,11 @@ static Prospero::ImageType TextureBaseType(Prospero::ImageType type) {
 	}
 }
 
+static bool IsMultisampledTexture(Prospero::ImageType type) {
+	return type == Prospero::ImageType::kColor2DMsaa ||
+	       type == Prospero::ImageType::kColor2DMsaaArray;
+}
+
 static BufferView NativeStorageBuffer(RenderContext& context, CommandBuffer& command_buffer,
                                       const ShaderBufferResource&                 descriptor,
                                       const ShaderRecompiler::IR::BufferResource& resource,
@@ -159,6 +164,8 @@ static bool IsSupportedSampledColorResource(const ShaderRecompiler::IR::ImageRes
 		case ShaderRecompiler::Decoder::ImageDimension::Dim1DArray:
 		case ShaderRecompiler::Decoder::ImageDimension::Dim2D:
 		case ShaderRecompiler::Decoder::ImageDimension::Dim2DArray:
+		case ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaa:
+		case ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaaArray:
 			supported_dimension = true;
 			break;
 		default: break;
@@ -195,6 +202,22 @@ TargetTextureViewInfo ResolveTargetTextureView(const ShaderRecompiler::IR::Image
 			           ? TargetTextureViewInfo {vk::ImageViewType::e2DArray, base_layer,
 			                                    image_layers - base_layer}
 			           : TargetTextureViewInfo {};
+		case Prospero::ImageType::kColor2DMsaa:
+			return resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaa &&
+			               base_layer == 0 && image_layers == 1
+			           ? TargetTextureViewInfo {vk::ImageViewType::e2D, 0, 1}
+			           : TargetTextureViewInfo {};
+		case Prospero::ImageType::kColor2DMsaaArray:
+			if (resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaa &&
+			    base_layer == 0 && image_layers == 1) {
+				return {vk::ImageViewType::e2D, 0, 1};
+			}
+			return resource.dimension ==
+			                   ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaaArray &&
+			               base_layer < image_layers
+			           ? TargetTextureViewInfo {vk::ImageViewType::e2DArray, base_layer,
+			                                    image_layers - base_layer}
+			           : TargetTextureViewInfo {};
 		default: return {};
 	}
 }
@@ -209,41 +232,64 @@ bool IsSupportedSampledVideoOutView(const ShaderRecompiler::IR::ImageResource& r
 }
 
 bool IsSupportedDepthTargetDescriptor(const ShaderTextureResource& descriptor, const Image& image) {
-	const auto width  = static_cast<uint32_t>(descriptor.Width5()) + 1u;
-	const auto height = static_cast<uint32_t>(descriptor.Height5()) + 1u;
-	const auto pitch  = TileGetTexturePitch(descriptor.Format(), width, 1, descriptor.TileMode());
-	const auto type   = static_cast<Prospero::ImageType>(descriptor.Type());
-	const bool supported_single_layer =
-	    image.info.resources.layers == 1 && descriptor.Depth() == 0 &&
-	    descriptor.BaseArray5() == 0 &&
-	    (type == Prospero::ImageType::kColor2D || type == Prospero::ImageType::kColor2DArray);
+	const auto width        = static_cast<uint32_t>(descriptor.Width5()) + 1u;
+	const auto height       = static_cast<uint32_t>(descriptor.Height5()) + 1u;
+	const auto type         = static_cast<Prospero::ImageType>(descriptor.Type());
+	const bool multisampled = IsMultisampledTexture(type);
+	const auto samples      = multisampled ? 1u << descriptor.LastLevel() : 1u;
+	const auto pitch =
+	    multisampled ? TileGetDepthPitch(width, image.info.bytes_per_block, descriptor.LastLevel())
+	                 : TileGetTexturePitch(descriptor.Format(), width, descriptor.TileMode());
+	const bool supported_2d    = type == Prospero::ImageType::kColor2D &&
+	                             image.info.resources.layers == 1 && descriptor.Depth() == 0 &&
+	                             descriptor.BaseArray5() == 0;
+	const bool supported_array = type == Prospero::ImageType::kColor2DArray &&
+	                             descriptor.BaseArray5() <= descriptor.Depth() &&
+	                             descriptor.Depth() < image.info.resources.layers;
 	const bool supported_cube =
 	    type == Prospero::ImageType::kCube && width == height && image.info.resources.layers >= 6 &&
 	    image.info.resources.layers % 6u == 0 &&
 	    static_cast<uint32_t>(descriptor.Depth()) + 1u == image.info.resources.layers &&
 	    descriptor.BaseArray5() == 0;
+	const bool supported_msaa_2d    = type == Prospero::ImageType::kColor2DMsaa &&
+	                                  image.info.resources.layers == 1 && descriptor.Depth() == 0 &&
+	                                  descriptor.BaseArray5() == 0;
+	const bool supported_msaa_array = type == Prospero::ImageType::kColor2DMsaaArray &&
+	                                  descriptor.BaseArray5() <= descriptor.Depth() &&
+	                                  descriptor.Depth() < image.info.resources.layers;
+	const bool levels_ok =
+	    multisampled
+	        ? descriptor.BaseLevel() == 0 && descriptor.LastLevel() >= 1 &&
+	              descriptor.LastLevel() <= 3 && descriptor.MaxMip() == descriptor.LastLevel() &&
+	              image.info.resources.levels == 1 && image.info.samples == samples
+	        : descriptor.BaseLevel() == 0 && descriptor.LastLevel() == 0 &&
+	              descriptor.MaxMip() == 0 && image.info.samples == 1;
 	return image.info.IsDepth() && width == image.info.extent.width &&
-	       height == image.info.extent.height && (supported_single_layer || supported_cube) &&
-	       descriptor.BaseLevel() == 0 && descriptor.LastLevel() == 0 && descriptor.MaxMip() == 0 &&
-	       descriptor.MinLod() == 0 && descriptor.BaseArray5() == 0 &&
+	       height == image.info.extent.height &&
+	       (supported_2d || supported_array || supported_cube || supported_msaa_2d ||
+	        supported_msaa_array) &&
+	       levels_ok && descriptor.MinLod() == 0 &&
 	       descriptor.TileMode() == Prospero::GpuEnumValue(Prospero::TileMode::kDepth) &&
-	       descriptor.BCSwizzle() == 0 && !descriptor.MsaaDepth() && pitch >= width &&
-	       pitch == image.info.pitch;
+	       descriptor.BCSwizzle() == 0 && (!descriptor.MsaaDepth() || multisampled) &&
+	       pitch >= width && pitch == image.info.pitch;
 }
 
 bool IsSupportedDepthTextureEncoding(const ShaderTextureResource& descriptor, const Image& image) {
 	constexpr uint32_t field1_reserved_mask = 0x200fff00u;
 	constexpr uint32_t field2_reserved_mask = 0xf0003000u;
-	constexpr uint32_t field3_common        = 0x01800000u;
-	constexpr uint32_t field5_expected      = 0x00700000u;
-	const uint32_t     field3_expected =
-	    (descriptor.Type() << 28u) | field3_common | descriptor.DstSelXYZW();
-	const uint32_t field4_expected = descriptor.Depth() | (descriptor.BaseArray5() << 16u);
-	const bool     common          = (descriptor.fields[1] & field1_reserved_mask) == 0 &&
-	                                 (descriptor.fields[2] & field2_reserved_mask) == 0 &&
-	                                 descriptor.fields[3] == field3_expected &&
-	                                 descriptor.fields[4] == field4_expected &&
-	                                 descriptor.fields[5] == field5_expected;
+	const uint32_t     field3_expected = descriptor.DstSelXYZW() |
+	                                     (static_cast<uint32_t>(descriptor.BaseLevel()) << 12u) |
+	                                     (static_cast<uint32_t>(descriptor.LastLevel()) << 16u) |
+	                                     (static_cast<uint32_t>(descriptor.TileMode()) << 20u) |
+	                                     (static_cast<uint32_t>(descriptor.Type()) << 28u);
+	const uint32_t     field4_expected = descriptor.Depth() | (descriptor.BaseArray5() << 16u);
+	const uint32_t     field5_expected =
+	    0x00700000u | (static_cast<uint32_t>(descriptor.MaxMip()) << 4u);
+	const bool common = (descriptor.fields[1] & field1_reserved_mask) == 0 &&
+	                    (descriptor.fields[2] & field2_reserved_mask) == 0 &&
+	                    descriptor.fields[3] == field3_expected &&
+	                    descriptor.fields[4] == field4_expected &&
+	                    descriptor.fields[5] == field5_expected;
 	if (!common || (descriptor.fields[6] == 0 && descriptor.fields[7] != 0)) {
 		return false;
 	}
@@ -251,8 +297,9 @@ bool IsSupportedDepthTextureEncoding(const ShaderTextureResource& descriptor, co
 		return true;
 	}
 	constexpr uint32_t htile_control = 0x00280000u;
-	const auto         metadata_addr = descriptor.MetaAddr() << 8u;
-	return (descriptor.fields[6] & 0x00ffffffu) == htile_control && metadata_addr != 0 &&
+	const uint32_t expected_control  = htile_control | (descriptor.MsaaDepth() ? (1u << 10u) : 0u);
+	const auto     metadata_addr     = descriptor.MetaAddr() << 8u;
+	return (descriptor.fields[6] & 0x00ffffffu) == expected_control && metadata_addr != 0 &&
 	       descriptor.TileMode() == Prospero::GpuEnumValue(Prospero::TileMode::kDepth) &&
 	       image.info.tile_mode == Prospero::GpuEnumValue(Prospero::TileMode::kDepth) &&
 	       image.info.metadata.kind == ImageMetadataKind::Htile &&
@@ -274,7 +321,7 @@ static void ValidateDepthTargetBinding(const ShaderRecompiler::IR::ImageResource
 		return;
 	}
 	const auto descriptor_pitch =
-	    TileGetTexturePitch(descriptor.Format(), static_cast<uint32_t>(descriptor.Width5()) + 1u, 1,
+	    TileGetTexturePitch(descriptor.Format(), static_cast<uint32_t>(descriptor.Width5()) + 1u,
 	                        descriptor.TileMode());
 	EXIT("unsupported sampled depth target: resource=%d descriptor=%d encoding=%d format=%d "
 	     "kind=%u dimension=%u mip_mode=%u read=%d written=%d atomic=%d compare=%d "
@@ -326,27 +373,38 @@ static bool IsSupportedStorageTextureDescriptor(const ShaderRecompiler::IR::Imag
 	const bool is_3d = resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim3D &&
 	                   descriptor.Type() == Prospero::GpuEnumValue(Prospero::ImageType::kColor3D) &&
 	                   descriptor.BaseArray5() == 0;
-	TileBlockLayout depth_block {};
-	const auto      depth_bpe = Prospero::RenderTargetBytesPerElement(descriptor.Format());
-	const bool      supported_depth_tile =
-	    tile == Prospero::GpuEnumValue(Prospero::TileMode::kDepth) && !resource.read &&
-	    !Prospero::IsFmaskTextureFormat(descriptor.Format()) && (is_2d || is_2d_array) &&
-	    TileGetBlockLayout(TileBlockFamily::Depth64KB, depth_bpe, depth_block);
-	const bool supported_standard_tile =
-	    (tile == Prospero::GpuEnumValue(Prospero::TileMode::kStandard4KB) &&
-	     TileIsStandard4KBTextureSupported(descriptor.Format())) ||
-	    (tile == Prospero::GpuEnumValue(Prospero::TileMode::kStandard64KB) &&
-	     TileIsStandard64KBTextureSupported(descriptor.Format()));
-	const bool supported_tile = tile == Prospero::GpuEnumValue(Prospero::TileMode::kLinear) ||
-	                            tile == Prospero::GpuEnumValue(Prospero::TileMode::kRenderTarget) ||
-	                            supported_depth_tile || supported_standard_tile;
-	const auto swizzle        = descriptor.DstSelXYZW();
+	TileTextureBlockLayout tile_layout {};
+	bool                   supported_tile = false;
+	switch (static_cast<Prospero::TileMode>(tile)) {
+		case Prospero::TileMode::kLinear: supported_tile = true; break;
+		case Prospero::TileMode::kDepth:
+			supported_tile =
+			    !resource.read && !Prospero::IsFmaskTextureFormat(descriptor.Format()) &&
+			    (is_2d || is_2d_array) &&
+			    TileGetTextureBlockLayout(descriptor.Format(), tile, false, tile_layout);
+			break;
+		case Prospero::TileMode::kStandard256B:
+			supported_tile =
+			    (is_2d || is_2d_array) &&
+			    TileGetTextureBlockLayout(descriptor.Format(), tile, false, tile_layout);
+			break;
+		case Prospero::TileMode::kStandard4KB:
+		case Prospero::TileMode::kStandard64KB:
+			supported_tile =
+			    TileGetTextureBlockLayout(descriptor.Format(), tile, is_3d, tile_layout);
+			break;
+		case Prospero::TileMode::kRenderTarget:
+			supported_tile =
+			    TileGetTextureBlockLayout(descriptor.Format(), tile, false, tile_layout);
+			break;
+		default: break;
+	}
+	const auto swizzle = descriptor.DstSelXYZW();
 	const bool supported_swizzle =
 	    IsValidImageSwizzle(swizzle) &&
 	    (swizzle == DstSel(4, 5, 6, 7) || !resource.read || resource.atomic);
-	const bool supported_mip_view = descriptor.BaseLevel() == 0 || is_1d || is_2d;
 	return (is_1d || is_1d_array || is_2d || is_2d_array || is_3d) && supported_tile &&
-	       supported_mip_view && descriptor.BaseLevel() == descriptor.LastLevel() &&
+	       descriptor.BaseLevel() == descriptor.LastLevel() &&
 	       descriptor.LastLevel() <= descriptor.MaxMip() && descriptor.MinLod() == 0 &&
 	       supported_swizzle && descriptor.BCSwizzle() == 0 && !descriptor.MsaaDepth();
 }
@@ -377,10 +435,14 @@ void ValidateStorageTexture(const ShaderRecompiler::IR::ImageResource& resource,
 	const bool encoding_ok   = IsSupportedStorageTextureEncoding(descriptor);
 	const bool uint_resource =
 	    resource.kind == ShaderRecompiler::IR::ResourceKind::StorageImageUint;
+	const bool raw_sint_storage =
+	    format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32SInt) && uint_resource &&
+	    resource.written && !resource.read && !resource.atomic;
 	const bool format_ok =
-	    Prospero::IsSupportedTextureFormat(format) &&
-	    uint_resource == Prospero::IsUintTextureFormat(format) &&
-	    (!resource.atomic || format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32UInt));
+	    raw_sint_storage ||
+	    (Prospero::IsSupportedTextureFormat(format) &&
+	     uint_resource == Prospero::IsUintTextureFormat(format) &&
+	     (!resource.atomic || format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32UInt)));
 	if (resource_ok && descriptor_ok && encoding_ok && format_ok && size != 0) {
 		return;
 	}
@@ -443,18 +505,20 @@ static TextureCache::ImageDesc NullTextureDesc(const ShaderRecompiler::IR::Image
 
 static void PopulateTextureMipLayout(ImageInfo& info) {
 	if (info.IsVolume() && info.tile_mode != Prospero::GpuEnumValue(Prospero::TileMode::kLinear)) {
-		TileVolumeLayout volume {};
-		if (!TileGetTextureVolumeLayout(info.guest_format, info.extent.width, info.extent.height,
-		                                info.extent.depth, info.resources.levels, info.tile_mode,
-		                                volume)) {
+		TileSurfaceLayout            surface {};
+		const TileSurfaceDescription description {
+		    info.guest_format,  info.tile_mode,    TileSurfaceDimension::Dim3D, info.extent.width,
+		    info.extent.height, info.extent.depth, info.resources.levels,       1};
+		if (!TileGetTiledTextureLayout(description, surface)) {
 			EXIT("unsupported normalized volume texture layout\n");
 		}
 		for (uint32_t level = 0; level < info.resources.levels; level++) {
+			const auto& mip        = surface.mips[level];
 			info.mip_layout[level] = {
-			    volume.level_offsets[level],
-			    volume.level_sizes[level],
-			    volume.level_widths[level],
-			    volume.level_heights[level],
+			    mip.offset,
+			    mip.size,
+			    mip.padded_width,
+			    mip.padded_height,
 			};
 		}
 		return;
@@ -462,7 +526,7 @@ static void PopulateTextureMipLayout(ImageInfo& info) {
 
 	TileSizeOffset levels[16] {};
 	TilePaddedSize padded[16] {};
-	TileGetTextureSize(info.guest_format, info.extent.width, info.extent.height, info.pitch,
+	TileGetTextureSize(info.guest_format, info.extent.width, info.extent.height,
 	                   info.resources.levels, info.tile_mode, nullptr, levels, padded);
 	for (uint32_t level = 0; level < info.resources.levels; level++) {
 		const auto offset =
@@ -518,6 +582,7 @@ static ImageViewInfo TextureViewInfo(const ShaderRecompiler::IR::ImageResource&
 			view.layer_count = 1;
 			break;
 		case ShaderRecompiler::Decoder::ImageDimension::Dim2DArray:
+		case ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaaArray:
 			view.type       = vk::ImageViewType::e2DArray;
 			view.base_layer = descriptor.BaseArray5();
 			if (view.base_layer >= image_layers) {
@@ -526,6 +591,7 @@ static ImageViewInfo TextureViewInfo(const ShaderRecompiler::IR::ImageResource&
 			view.layer_count = image_layers - view.base_layer;
 			break;
 		case ShaderRecompiler::Decoder::ImageDimension::Dim2D:
+		case ShaderRecompiler::Decoder::ImageDimension::Dim2DMsaa:
 			view.type       = vk::ImageViewType::e2D;
 			view.base_layer = descriptor.BaseArray5();
 			if (view.base_layer >= image_layers) {
@@ -556,25 +622,34 @@ RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageResource&   reso
 		return {id, nullptr, std::move(desc)};
 	}
 
-	const auto address    = descriptor.Base40();
-	const auto width      = static_cast<uint32_t>(descriptor.Width5()) + 1u;
-	const auto height     = static_cast<uint32_t>(descriptor.Height5()) + 1u;
-	const auto base_level = descriptor.BaseLevel();
-	const auto last_level = descriptor.LastLevel();
-	const auto type       = TextureType(descriptor);
-	const bool multisampled =
-	    type == Prospero::ImageType::kColor2DMsaa || type == Prospero::ImageType::kColor2DMsaaArray;
-	const auto levels     = multisampled ? 1u : static_cast<uint32_t>(descriptor.MaxMip()) + 1u;
-	const auto tile       = descriptor.TileMode();
-	const bool msaa_tile  = tile == Prospero::GpuEnumValue(Prospero::TileMode::kRenderTarget);
+	const auto address      = descriptor.Base40();
+	const auto width        = static_cast<uint32_t>(descriptor.Width5()) + 1u;
+	const auto height       = static_cast<uint32_t>(descriptor.Height5()) + 1u;
+	const auto base_level   = descriptor.BaseLevel();
+	const auto last_level   = descriptor.LastLevel();
+	const auto type         = TextureType(descriptor);
+	const bool multisampled = IsMultisampledTexture(type);
+	const auto levels       = multisampled ? 1u : static_cast<uint32_t>(descriptor.MaxMip()) + 1u;
+	const auto tile         = descriptor.TileMode();
+	const bool depth_tile   = tile == Prospero::GpuEnumValue(Prospero::TileMode::kDepth);
+	const bool msaa_tile =
+	    depth_tile || tile == Prospero::GpuEnumValue(Prospero::TileMode::kRenderTarget);
 	const bool msaa_array = type == Prospero::ImageType::kColor2DMsaaArray;
 	if ((!multisampled && (base_level > last_level || last_level >= levels)) ||
 	    (multisampled &&
 	     (base_level != 0 || last_level == 0 || last_level > 3 ||
-	      descriptor.MaxMip() != last_level || !msaa_tile || descriptor.MsaaDepth() ||
+	      descriptor.MaxMip() != last_level || !msaa_tile ||
+	      (descriptor.MsaaDepth() && !depth_tile) ||
 	      (!msaa_array && (descriptor.Depth() != 0 || descriptor.BaseArray5() != 0))))) {
-		EXIT("unsupported texture mip view: base=%u last=%u levels=%u\n", base_level, last_level,
-		     levels);
+		EXIT("unsupported texture mip view: base=%u last=%u levels=%u max=%u type=%u tile=%u "
+		     "kind=%u dimension=%u mip_mode=%u read=%d written=%d "
+		     "dwords=%08x,%08x,%08x,%08x,%08x,%08x,%08x,%08x\n",
+		     base_level, last_level, levels, descriptor.MaxMip(), descriptor.Type(), tile,
+		     static_cast<uint32_t>(resource.kind), static_cast<uint32_t>(resource.dimension),
+		     static_cast<uint32_t>(resource.mip_mode), resource.read, resource.written,
+		     descriptor.fields[0], descriptor.fields[1], descriptor.fields[2], descriptor.fields[3],
+		     descriptor.fields[4], descriptor.fields[5], descriptor.fields[6],
+		     descriptor.fields[7]);
 	}
 	const auto samples = multisampled ? 1u << last_level : 1u;
 	const auto view_levels =
@@ -601,27 +676,29 @@ RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageResource&   reso
 	TileSizeAlign size {};
 	if (multisampled) {
 		const auto bytes = Prospero::NumBytesPerElement(format);
-		pitch            = TileGetRenderTargetPitch(width, bytes, last_level);
+		pitch            = depth_tile ? TileGetDepthPitch(width, bytes, last_level)
+		                              : TileGetRenderTargetPitch(width, bytes, last_level);
 		if (pitch == 0 || !TileGetRenderTargetSize(width, height, pitch, bytes, size, last_level) ||
 		    size.size > UINT32_MAX / image_layers) {
 			EXIT("unsupported multisample texture layout\n");
 		}
 		size.size *= image_layers;
 	} else {
-		pitch = TileGetTexturePitch(format, width, levels, tile);
-		TileGetTextureTotalSize(format, width, height, volume ? depth : image_layers, pitch, levels,
-		                        tile, volume, size);
+		pitch = TileGetTexturePitch(format, width, tile);
+		TileGetTextureTotalSize(format, width, height, volume ? depth : image_layers, levels, tile,
+		                        volume, size);
 	}
 	EXIT_NOT_IMPLEMENTED(size.size == 0 || size.align == 0 ||
 	                     (address & (static_cast<uint64_t>(size.align) - 1u)) != 0);
 	if (storage) {
 		ValidateStorageTexture(resource, descriptor, size.size);
-		m_context.GetBufferCache().ValidateGpuAccess(address, size.size, resource.read,
-		                                             resource.written);
 	}
 
-	const auto              pixel_format        = TextureGetFormat(format);
-	const auto              storage_view_format = SrgbStorageViewFormat(pixel_format);
+	const auto pixel_format = TextureGetFormat(format);
+	const auto storage_view_format =
+	    storage && format == Prospero::GpuEnumValue(Prospero::BufferFormat::k32SInt)
+	        ? vk::Format::eR32Uint
+	        : SrgbStorageViewFormat(pixel_format);
 	const auto              view_format = storage && storage_view_format != vk::Format::eUndefined
 	                                          ? storage_view_format
 	                                          : pixel_format;
@@ -856,22 +933,21 @@ RenderExecutor::PrepareGraphicsBindings(CommandBuffer& buffer, const ShaderStage
 	return bindings;
 }
 
-static void RetainBindings(CommandBuffer&                            buffer,
-                           const DescriptorCache::NativeDescriptors& resources) {
+static void RetainBindings(CommandBuffer& buffer, DescriptorCache::NativeDescriptors& resources) {
 	if (resources.flattened_srt.owner != nullptr) {
-		buffer.RetainResourceUntilFence(resources.flattened_srt.owner);
+		buffer.RetainResourceUntilFence(std::move(resources.flattened_srt.owner));
 	}
 	if (resources.user_data.owner != nullptr) {
-		buffer.RetainResourceUntilFence(resources.user_data.owner);
+		buffer.RetainResourceUntilFence(std::move(resources.user_data.owner));
 	}
-	for (const auto& view: resources.buffers) {
+	for (auto& view: resources.buffers) {
 		if (view.owner != nullptr) {
-			buffer.RetainResourceUntilFence(view.owner);
+			buffer.RetainResourceUntilFence(std::move(view.owner));
 		}
 	}
-	for (const auto& view: resources.addresses) {
+	for (auto& view: resources.addresses) {
 		if (view.owner != nullptr) {
-			buffer.RetainResourceUntilFence(view.owner);
+			buffer.RetainResourceUntilFence(std::move(view.owner));
 		}
 	}
 }

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

@@ -36,7 +36,7 @@ ResolveTargetTextureView(const ShaderRecompiler::IR::ImageResource& resource,
 [[nodiscard]] bool IsSupportedDepthTargetDescriptor(const ShaderTextureResource& descriptor,
                                                     const Image&                 image);
 [[nodiscard]] bool IsSupportedDepthTextureEncoding(const ShaderTextureResource& descriptor,
-                                                    const Image&                 image);
+                                                   const Image&                 image);
 [[nodiscard]] bool
 IsSupportedSampledVideoOutView(const ShaderRecompiler::IR::ImageResource& resource,
                                const ShaderTextureResource& descriptor, const Image& image);

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

@@ -88,12 +88,12 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 
 	PipelineStaticParameters static_params {};
 	GraphicsPipeline         p {};
-	p.ps_shader_id   = ps_id;
-	p.vs_shader_id   = vs_id;
+	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;
+	rendering.color_count       = color_count;
 	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);
@@ -116,8 +116,8 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 		if (attachment_samples == 0) {
 			attachment_samples = depth.samples;
 		} else if (attachment_samples != depth.samples) {
-			EXIT("mixed color/depth sample counts are unsupported: %u and %u\n",
-			     attachment_samples, depth.samples);
+			EXIT("mixed color/depth sample counts are unsupported: %u and %u\n", attachment_samples,
+			     depth.samples);
 		}
 	}
 	EXIT_IF(attachment_samples == 0 ||
@@ -154,8 +154,9 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 	for (uint32_t i = 0; i < RENDER_COLOR_ATTACHMENTS_MAX; i++) {
 		static_params.color_mask[i] = color_mask[i];
 	}
-	static_params.cull_back  = mc.cull_back;
-	static_params.cull_front = mc.cull_front;
+	const bool rect_list     = topology == vk::PrimitiveTopology::ePatchList;
+	static_params.cull_back  = !rect_list && mc.cull_back;
+	static_params.cull_front = !rect_list && mc.cull_front;
 	static_params.face       = mc.face;
 
 	for (uint32_t i = 0; i < color_count; i++) {
@@ -179,10 +180,10 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 	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;
+	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;
@@ -203,9 +204,8 @@ PipelineCache::GraphicsPipeline& PipelineCache::CreateGraphicsPipeline(
 	LogPipelineTrace("CreatePipelineInternal begin", vs_id.hash0, vs_id.crc32, ps_id.hash0,
 	                 ps_id.crc32);
 	CreatePipelineInternal(m_graphics, m_descriptor_cache, *cached, rendering, vs_input_info,
-	                       vs_spirv, ps_input_info,
-	                       ps_spirv, static_params, vs_id.hash0, vs_id.crc32, ps_id.hash0,
-	                       ps_id.crc32, ps_active);
+	                       vs_spirv, ps_input_info, ps_spirv, static_params, vs_id.hash0,
+	                       vs_id.crc32, ps_id.hash0, ps_id.crc32, ps_active);
 	LogPipelineTrace("CreatePipelineInternal done", vs_id.hash0, vs_id.crc32, ps_id.hash0,
 	                 ps_id.crc32);
 

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

@@ -88,9 +88,9 @@ static_assert(sizeof(PipelineStaticParameters) ==
 
 struct PipelineRenderingState {
 	std::array<vk::Format, RENDER_COLOR_ATTACHMENTS_MAX> color_formats {};
-	vk::Format depth_format   = vk::Format::eUndefined;
-	vk::Format stencil_format = vk::Format::eUndefined;
-	uint32_t   color_count    = 0;
+	vk::Format                                           depth_format   = vk::Format::eUndefined;
+	vk::Format                                           stencil_format = vk::Format::eUndefined;
+	uint32_t                                             color_count    = 0;
 
 	bool operator==(const PipelineRenderingState&) const = default;
 };
@@ -118,11 +118,12 @@ public:
 		ShaderId cs_shader_id;
 	};
 
-	GraphicsPipeline& CreateGraphicsPipeline(
-	    RenderColorInfo* colors, uint32_t color_count, RenderDepthInfo& depth,
-	    ShaderVertexInputInfo& vs_input_info, RenderCommandBuffer& command,
-	    ShaderPixelInputInfo* ps_input_info, vk::PrimitiveTopology topology, bool ps_active,
-	    std::span<const uint32_t> vs_spirv, std::span<const uint32_t> ps_spirv);
+	GraphicsPipeline&
+	CreateGraphicsPipeline(RenderColorInfo* colors, uint32_t color_count, RenderDepthInfo& depth,
+	                       ShaderVertexInputInfo& vs_input_info, RenderCommandBuffer& command,
+	                       ShaderPixelInputInfo* ps_input_info, vk::PrimitiveTopology topology,
+	                       bool ps_active, std::span<const uint32_t> vs_spirv,
+	                       std::span<const uint32_t> ps_spirv);
 	ComputePipeline& CreateComputePipeline(ShaderComputeInputInfo&      input_info,
 	                                       const HW::ComputeShaderInfo& cs_regs,
 	                                       std::span<const uint32_t>    cs_spirv);
@@ -199,7 +200,7 @@ private:
 		}
 	};
 
-	GraphicContext& m_graphics;
+	GraphicContext&  m_graphics;
 	DescriptorCache& m_descriptor_cache;
 	std::unordered_map<GraphicsPipelineKey, std::unique_ptr<GraphicsPipeline>,
 	                   GraphicsPipelineKeyHash>
@@ -211,16 +212,13 @@ private:
 
 void LogPipelineTrace(const char* phase, uint32_t vs_hash0, uint32_t vs_crc32, uint32_t ps_hash0,
                       uint32_t ps_crc32);
-void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descriptor_cache,
-                            PipelineCache::GraphicsPipeline& pipeline,
-                            const PipelineRenderingState&   rendering,
-                            const ShaderVertexInputInfo&    vs_input_info,
-                            std::span<const uint32_t>       vs_shader,
-                            const ShaderPixelInputInfo*     ps_input_info,
-                            std::span<const uint32_t>       ps_shader,
-                            const PipelineStaticParameters& static_params, uint32_t vs_hash0,
-                            uint32_t vs_crc32, uint32_t ps_hash0, uint32_t ps_crc32,
-                            bool ps_active);
+void CreatePipelineInternal(
+    GraphicContext& graphics, DescriptorCache& descriptor_cache,
+    PipelineCache::GraphicsPipeline& pipeline, const PipelineRenderingState& rendering,
+    const ShaderVertexInputInfo& vs_input_info, std::span<const uint32_t> vs_shader,
+    const ShaderPixelInputInfo* ps_input_info, std::span<const uint32_t> ps_shader,
+    const PipelineStaticParameters& static_params, uint32_t vs_hash0, uint32_t vs_crc32,
+    uint32_t ps_hash0, uint32_t ps_crc32, bool ps_active);
 void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descriptor_cache,
                             PipelineCache::ComputePipeline& pipeline,
                             const ShaderComputeInputInfo&   input_info,

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

@@ -36,6 +36,22 @@ VulkanMemoryBarrier MakeShaderWriteDependency() {
 	return barrier;
 }
 
+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;
+}
+
+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;
+}
+
 vk::BufferMemoryBarrier MakeGdsDependency(vk::Buffer buffer) {
 	EXIT_IF(buffer == nullptr);
 
@@ -84,6 +100,21 @@ bool HasShaderBufferWrites(const ShaderStageRuntime& runtime) {
 	return !CollectShaderBufferWrites(*runtime.program, *runtime.resources).empty();
 }
 
+void ShaderAccessBarrier(vk::CommandBuffer vk_buffer, vk::PipelineStageFlags source_stages) {
+	EXIT_IF(vk_buffer == nullptr || !source_stages);
+	const auto barrier = MakeShaderAccessDependency();
+	vk_buffer.pipelineBarrier(source_stages, vk::PipelineStageFlagBits::eAllCommands,
+	                          vk::DependencyFlags {}, 1, &barrier, 0, nullptr, 0, nullptr);
+}
+
+void ShaderWriteHazardBarrier(vk::CommandBuffer      vk_buffer,
+                              vk::PipelineStageFlags destination_stages) {
+	EXIT_IF(vk_buffer == nullptr || !destination_stages);
+	const auto barrier = MakeShaderWriteHazardDependency();
+	vk_buffer.pipelineBarrier(vk::PipelineStageFlagBits::eAllCommands, destination_stages,
+	                          vk::DependencyFlags {}, 1, &barrier, 0, nullptr, 0, nullptr);
+}
+
 void ShaderWriteBarrier(vk::CommandBuffer vk_buffer, vk::PipelineStageFlags source_stages) {
 	EXIT_IF(vk_buffer == nullptr || !source_stages);
 	const auto barrier = MakeShaderWriteDependency();

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

@@ -18,12 +18,17 @@ struct ShaderBufferWriteRange {
 };
 
 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,
+                              vk::PipelineStageFlags destination_stages);
 void ShaderWriteBarrier(vk::CommandBuffer vk_buffer, vk::PipelineStageFlags source_stages);
 
 } // namespace Libs::Graphics

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

@@ -8,12 +8,13 @@
 #include "graphics/host_gpu/renderer/debug.h"
 #include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
+#include "graphics/host_gpu/renderer/pipeline/shaderSubgroup.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/renderer/renderTarget.h"
-#include "graphics/host_gpu/renderer/pipeline/shaderSubgroup.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 #include "graphics/shader/recompiler/ir/ShaderIR.h"
+#include "graphics/shader/rectListShader.h"
 #include "graphics/shader/shader.h"
 
 #include <algorithm>
@@ -385,9 +386,8 @@ static vk::BlendOp GetBlendOp(uint32_t op) {
 	return vk::BlendOp::eAdd;
 }
 
-static void CreateLayout(DescriptorCache& descriptor_cache,
-                         std::span<vk::DescriptorSetLayout> set_layouts,
-                         uint32_t& set_layouts_num,
+static void CreateLayout(DescriptorCache&                   descriptor_cache,
+                         std::span<vk::DescriptorSetLayout> set_layouts, uint32_t& set_layouts_num,
                          std::span<vk::PushConstantRange>     push_constant_info,
                          uint32_t&                            push_constant_info_num,
                          const ShaderRecompiler::IR::Program& program,
@@ -412,12 +412,11 @@ static void CreateLayout(DescriptorCache& descriptor_cache,
 	}
 }
 
-static void ConfigureSubgroupSize(const GraphicContext&                                  graphics,
-                                  vk::ShaderStageFlagBits                                vk_stage,
+static void ConfigureSubgroupSize(const GraphicContext& graphics, vk::ShaderStageFlagBits vk_stage,
                                   const ShaderRecompiler::IR::Program&                   program,
                                   vk::PipelineShaderStageRequiredSubgroupSizeCreateInfo& required,
                                   vk::PipelineShaderStageCreateInfo&                     stage) {
-	const auto  config =
+	const auto config =
 	    ConfigureShaderSubgroup(ShaderSubgroupCapabilities {graphics}, vk_stage, program);
 	switch (config.mode) {
 		case ShaderSubgroupMode::Natural: return;
@@ -456,20 +455,21 @@ static void ConfigureSubgroupSize(const GraphicContext&
 }
 
 // NOLINTNEXTLINE(readability-function-cognitive-complexity)
-void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descriptor_cache,
-                            PipelineCache::GraphicsPipeline& pipeline,
-                            const PipelineRenderingState&   rendering,
-                            const ShaderVertexInputInfo&    vs_input_info,
-                            std::span<const uint32_t>       vs_shader,
-                            const ShaderPixelInputInfo*     ps_input_info,
-                            std::span<const uint32_t>       ps_shader,
-                            const PipelineStaticParameters& static_params, uint32_t vs_hash0,
-                            uint32_t vs_crc32, uint32_t ps_hash0, uint32_t ps_crc32,
-                            bool ps_active) {
+void CreatePipelineInternal(
+    GraphicContext& graphics, DescriptorCache& descriptor_cache,
+    PipelineCache::GraphicsPipeline& pipeline, const PipelineRenderingState& rendering,
+    const ShaderVertexInputInfo& vs_input_info, std::span<const uint32_t> vs_shader,
+    const ShaderPixelInputInfo* ps_input_info, std::span<const uint32_t> ps_shader,
+    const PipelineStaticParameters& static_params, uint32_t vs_hash0, uint32_t vs_crc32,
+    uint32_t ps_hash0, uint32_t ps_crc32, bool ps_active) {
 	EXIT_IF(ps_active && ps_input_info == nullptr);
 
-	vk::ShaderModule vert_shader_module = nullptr;
-	vk::ShaderModule frag_shader_module = nullptr;
+	const bool rect_list = static_params.topology == vk::PrimitiveTopology::ePatchList;
+
+	vk::ShaderModule vert_shader_module         = nullptr;
+	vk::ShaderModule tess_control_shader_module = nullptr;
+	vk::ShaderModule tess_eval_shader_module    = nullptr;
+	vk::ShaderModule frag_shader_module         = nullptr;
 
 	vk::ShaderModuleCreateInfo create_info {};
 
@@ -496,8 +496,33 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 		}
 		EXIT_NOT_IMPLEMENTED(result != vk::Result::eSuccess);
 	}
+	if (rect_list) {
+		const auto shaders =
+		    BuildRectListShaders(vs_input_info, ps_active ? ps_input_info : nullptr);
+		create_info.codeSize = shaders.control.size() * 4;
+		create_info.pCode    = shaders.control.data();
+		result =
+		    graphics.device.createShaderModule(&create_info, nullptr, &tess_control_shader_module);
+		if (graphics_debug_dump_enabled()) {
+			LOGF("PipelineTrace: vkCreateShaderModule RectList TCS done result=%s module=%p\n",
+			     VulkanToString(result).c_str(), static_cast<void*>(tess_control_shader_module));
+		}
+		EXIT_NOT_IMPLEMENTED(result != vk::Result::eSuccess);
+
+		create_info.codeSize = shaders.evaluation.size() * 4;
+		create_info.pCode    = shaders.evaluation.data();
+		result =
+		    graphics.device.createShaderModule(&create_info, nullptr, &tess_eval_shader_module);
+		if (graphics_debug_dump_enabled()) {
+			LOGF("PipelineTrace: vkCreateShaderModule RectList TES done result=%s module=%p\n",
+			     VulkanToString(result).c_str(), static_cast<void*>(tess_eval_shader_module));
+		}
+		EXIT_NOT_IMPLEMENTED(result != vk::Result::eSuccess);
+	}
 
 	EXIT_NOT_IMPLEMENTED(vert_shader_module == nullptr);
+	EXIT_NOT_IMPLEMENTED(
+	    rect_list && (tess_control_shader_module == nullptr || tess_eval_shader_module == nullptr));
 	EXIT_NOT_IMPLEMENTED(ps_active && frag_shader_module == nullptr);
 
 	vk::PipelineShaderStageCreateInfo                     vert_shader_stage_info {};
@@ -511,8 +536,7 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	vert_shader_stage_info.pName               = "main";
 	vert_shader_stage_info.pSpecializationInfo = nullptr;
 	EXIT_IF(!vs_input_info.stage);
-	ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eVertex,
-	                      *vs_input_info.stage.program,
+	ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eVertex, *vs_input_info.stage.program,
 	                      vert_subgroup_size, vert_shader_stage_info);
 
 	vk::PipelineShaderStageCreateInfo                     frag_shader_stage_info {};
@@ -527,13 +551,32 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	if (ps_active) {
 		EXIT_IF(!ps_input_info->stage);
 		ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eFragment,
-		                      *ps_input_info->stage.program,
-		                      frag_subgroup_size, frag_shader_stage_info);
+		                      *ps_input_info->stage.program, frag_subgroup_size,
+		                      frag_shader_stage_info);
 	}
 
-	vk::PipelineShaderStageCreateInfo shader_stages[]    = {vert_shader_stage_info,
-	                                                        frag_shader_stage_info};
-	const uint32_t                    shader_stage_count = ps_active ? 2u : 1u;
+	vk::PipelineShaderStageCreateInfo tess_control_shader_stage_info {};
+	tess_control_shader_stage_info.sType  = vk::StructureType::ePipelineShaderStageCreateInfo;
+	tess_control_shader_stage_info.stage  = vk::ShaderStageFlagBits::eTessellationControl;
+	tess_control_shader_stage_info.module = tess_control_shader_module;
+	tess_control_shader_stage_info.pName  = "main";
+
+	vk::PipelineShaderStageCreateInfo tess_eval_shader_stage_info {};
+	tess_eval_shader_stage_info.sType  = vk::StructureType::ePipelineShaderStageCreateInfo;
+	tess_eval_shader_stage_info.stage  = vk::ShaderStageFlagBits::eTessellationEvaluation;
+	tess_eval_shader_stage_info.module = tess_eval_shader_module;
+	tess_eval_shader_stage_info.pName  = "main";
+
+	vk::PipelineShaderStageCreateInfo shader_stages[4]   = {};
+	uint32_t                          shader_stage_count = 0;
+	shader_stages[shader_stage_count++]                  = vert_shader_stage_info;
+	if (rect_list) {
+		shader_stages[shader_stage_count++] = tess_control_shader_stage_info;
+		shader_stages[shader_stage_count++] = tess_eval_shader_stage_info;
+	}
+	if (ps_active) {
+		shader_stages[shader_stage_count++] = frag_shader_stage_info;
+	}
 
 	vk::VertexInputAttributeDescription input_attr[ShaderVertexInputInfo::RES_MAX];
 	vk::VertexInputBindingDescription   input_desc[ShaderVertexInputInfo::RES_MAX];
@@ -728,13 +771,13 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	clip_ext.depthClipEnable = static_params.depth_clip_enable ? VK_TRUE : VK_FALSE;
 
 	vk::PipelineRasterizationStateCreateInfo rasterizer {};
-	rasterizer.sType                   = vk::StructureType::ePipelineRasterizationStateCreateInfo;
+	rasterizer.sType = vk::StructureType::ePipelineRasterizationStateCreateInfo;
 	// MoltenVK lacks VK_EXT_depth_clip_enable; omit the depth-clip struct on macOS and accept
 	// Vulkan's default depth clipping (enabled) instead of the PS5's clamp behavior.
 #if defined(__APPLE__)
-	rasterizer.pNext                   = nullptr;
+	rasterizer.pNext = nullptr;
 #else
-	rasterizer.pNext                   = &clip_ext;
+	rasterizer.pNext = &clip_ext;
 #endif
 	rasterizer.flags                   = {};
 	rasterizer.depthClampEnable        = VK_FALSE;
@@ -812,13 +855,13 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	color_write.pColorWriteEnables = color_write_enable;
 
 	vk::PipelineColorBlendStateCreateInfo color_blending {};
-	color_blending.sType             = vk::StructureType::ePipelineColorBlendStateCreateInfo;
+	color_blending.sType = vk::StructureType::ePipelineColorBlendStateCreateInfo;
 	// MoltenVK lacks VK_EXT_color_write_enable; drop the dynamic color-write struct on macOS
 	// and rely on each attachment's static colorWriteMask (all channels enabled by default).
 #if defined(__APPLE__)
-	color_blending.pNext             = nullptr;
+	color_blending.pNext = nullptr;
 #else
-	color_blending.pNext             = &color_write;
+	color_blending.pNext = &color_write;
 #endif
 	color_blending.flags             = {};
 	color_blending.logicOpEnable     = VK_FALSE;
@@ -838,15 +881,13 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 
 	EXIT_IF(!vs_input_info.stage);
 	CreateLayout(descriptor_cache, set_layouts, set_layouts_num, push_constant_info,
-	             push_constant_info_num,
-	             *vs_input_info.stage.program, vk::ShaderStageFlagBits::eVertex,
-	             DescriptorCache::Stage::Vertex);
+	             push_constant_info_num, *vs_input_info.stage.program,
+	             vk::ShaderStageFlagBits::eVertex, DescriptorCache::Stage::Vertex);
 	if (ps_active) {
 		EXIT_IF(!ps_input_info->stage);
 		CreateLayout(descriptor_cache, set_layouts, set_layouts_num, push_constant_info,
-		             push_constant_info_num,
-		             *ps_input_info->stage.program, vk::ShaderStageFlagBits::eFragment,
-		             DescriptorCache::Stage::Pixel);
+		             push_constant_info_num, *ps_input_info->stage.program,
+		             vk::ShaderStageFlagBits::eFragment, DescriptorCache::Stage::Pixel);
 	}
 
 	vk::PipelineLayoutCreateInfo pipeline_layout_info {};
@@ -906,6 +947,8 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	    vk::DynamicState::eViewport,
 	    vk::DynamicState::eScissor,
 	    vk::DynamicState::eLineWidth,
+	    vk::DynamicState::eDepthBiasEnable,
+	    vk::DynamicState::eDepthBias,
 	    vk::DynamicState::eStencilCompareMask,
 	    vk::DynamicState::eStencilReference,
 	    vk::DynamicState::eStencilWriteMask,
@@ -923,32 +966,35 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	dynamic_state.dynamicStateCount = dynamic_states_count;
 	dynamic_state.pDynamicStates    = dynamic_states;
 
-	vk::GraphicsPipelineCreateInfo pipeline_info {};
+	vk::GraphicsPipelineCreateInfo  pipeline_info {};
 	vk::PipelineRenderingCreateInfo rendering_info {};
 	rendering_info.sType                   = vk::StructureType::ePipelineRenderingCreateInfo;
 	rendering_info.colorAttachmentCount    = rendering.color_count;
 	rendering_info.pColorAttachmentFormats = rendering.color_formats.data();
 	rendering_info.depthAttachmentFormat   = rendering.depth_format;
 	rendering_info.stencilAttachmentFormat = rendering.stencil_format;
-	pipeline_info.sType               = vk::StructureType::eGraphicsPipelineCreateInfo;
-	pipeline_info.pNext               = &rendering_info;
-	pipeline_info.flags               = {};
-	pipeline_info.stageCount          = shader_stage_count;
-	pipeline_info.pStages             = shader_stages;
-	pipeline_info.pVertexInputState   = &vertex_input_info;
-	pipeline_info.pInputAssemblyState = &input_assembly;
-	pipeline_info.pTessellationState  = nullptr;
-	pipeline_info.pViewportState      = &viewport_state;
-	pipeline_info.pRasterizationState = &rasterizer;
-	pipeline_info.pMultisampleState   = &multisampling;
-	pipeline_info.pDepthStencilState  = (static_params.with_depth ? &depth_stencil_info : nullptr);
-	pipeline_info.pColorBlendState    = &color_blending;
-	pipeline_info.pDynamicState       = &dynamic_state;
-	pipeline_info.layout              = pipeline.pipeline_layout;
-	pipeline_info.renderPass          = nullptr;
-	pipeline_info.subpass             = 0;
-	pipeline_info.basePipelineHandle  = nullptr;
-	pipeline_info.basePipelineIndex   = -1;
+	pipeline_info.sType                    = vk::StructureType::eGraphicsPipelineCreateInfo;
+	pipeline_info.pNext                    = &rendering_info;
+	pipeline_info.flags                    = {};
+	pipeline_info.stageCount               = shader_stage_count;
+	pipeline_info.pStages                  = shader_stages;
+	pipeline_info.pVertexInputState        = &vertex_input_info;
+	pipeline_info.pInputAssemblyState      = &input_assembly;
+	vk::PipelineTessellationStateCreateInfo tessellation_state {};
+	tessellation_state.sType              = vk::StructureType::ePipelineTessellationStateCreateInfo;
+	tessellation_state.patchControlPoints = 3;
+	pipeline_info.pTessellationState      = (rect_list ? &tessellation_state : nullptr);
+	pipeline_info.pViewportState          = &viewport_state;
+	pipeline_info.pRasterizationState     = &rasterizer;
+	pipeline_info.pMultisampleState       = &multisampling;
+	pipeline_info.pDepthStencilState = (static_params.with_depth ? &depth_stencil_info : nullptr);
+	pipeline_info.pColorBlendState   = &color_blending;
+	pipeline_info.pDynamicState      = &dynamic_state;
+	pipeline_info.layout             = pipeline.pipeline_layout;
+	pipeline_info.renderPass         = nullptr;
+	pipeline_info.subpass            = 0;
+	pipeline_info.basePipelineHandle = nullptr;
+	pipeline_info.basePipelineIndex  = -1;
 
 	EXIT_IF(pipeline.pipeline != nullptr);
 
@@ -976,6 +1022,12 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	if (frag_shader_module != nullptr) {
 		graphics.device.destroyShaderModule(frag_shader_module, nullptr);
 	}
+	if (tess_control_shader_module != nullptr) {
+		graphics.device.destroyShaderModule(tess_control_shader_module, nullptr);
+	}
+	if (tess_eval_shader_module != nullptr) {
+		graphics.device.destroyShaderModule(tess_eval_shader_module, nullptr);
+	}
 	graphics.device.destroyShaderModule(vert_shader_module, nullptr);
 }
 
@@ -1012,8 +1064,7 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 	comp_shader_stage_info.pName               = "main";
 	comp_shader_stage_info.pSpecializationInfo = nullptr;
 	EXIT_IF(!input_info.stage);
-	ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eCompute,
-	                      *input_info.stage.program,
+	ConfigureSubgroupSize(graphics, vk::ShaderStageFlagBits::eCompute, *input_info.stage.program,
 	                      comp_subgroup_size, comp_shader_stage_info);
 
 	vk::DescriptorSetLayout set_layouts[1]  = {};
@@ -1024,9 +1075,8 @@ void CreatePipelineInternal(GraphicContext& graphics, DescriptorCache& descripto
 
 	EXIT_IF(!input_info.stage);
 	CreateLayout(descriptor_cache, set_layouts, set_layouts_num, push_constant_info,
-	             push_constant_info_num,
-	             *input_info.stage.program, vk::ShaderStageFlagBits::eCompute,
-	             DescriptorCache::Stage::Compute);
+	             push_constant_info_num, *input_info.stage.program,
+	             vk::ShaderStageFlagBits::eCompute, DescriptorCache::Stage::Compute);
 
 	vk::PipelineLayoutCreateInfo pipeline_layout_info {};
 	pipeline_layout_info.sType                  = vk::StructureType::ePipelineLayoutCreateInfo;

src/graphics/host_gpu/renderer/renderCompute.cpp

@@ -10,14 +10,14 @@
 #include "graphics/guest_gpu/graphicsRun.h"
 #include "graphics/guest_gpu/hardwareContext.h"
 #include "graphics/host_gpu/graphicContext.h"
+#include "graphics/host_gpu/renderer/image/imageInfo.h"
 #include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/pipeline/descriptors.h"
-#include "graphics/host_gpu/renderer/image/imageInfo.h"
 #include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
-#include "graphics/host_gpu/renderer/render.h"
-#include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/renderer/pipeline/shaderResourceBarrier.h"
 #include "graphics/host_gpu/renderer/pipeline/shaderSubgroup.h"
+#include "graphics/host_gpu/renderer/render.h"
+#include "graphics/host_gpu/renderer/renderContext.h"
 #include "graphics/host_gpu/vulkanCommon.h"
 #include "graphics/shader/recompiler/ir/ResourceMaterialization.h"
 #include "graphics/shader/recompiler/ir/ShaderIR.h"
@@ -325,9 +325,6 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, RenderCommandBuffer& buf
 
 	auto vk_buffer = buffer.Handle();
 	CommitBindings(buffer, vk::PipelineBindPoint::eCompute, pipeline.pipeline_layout, bindings);
-	vk_buffer.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline.pipeline);
-	vk_buffer.dispatch(thread_group_x, thread_group_y, thread_group_z);
-
 	bool has_storage_writes = HasShaderBufferWrites(input_info.stage);
 	has_storage_writes =
 	    std::any_of(program.info.images.begin(), program.info.images.end(),
@@ -338,8 +335,15 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, RenderCommandBuffer& buf
 	                }) ||
 	    has_storage_writes;
 	if (has_storage_writes) {
-		ShaderWriteBarrier(vk_buffer, vk::PipelineStageFlagBits::eComputeShader);
+		// A host fence used to serialize every dispatch. Preserve its read-before-write ordering
+		// while allowing the queue to execute asynchronously.
+		ShaderWriteHazardBarrier(vk_buffer, vk::PipelineStageFlagBits::eComputeShader);
 	}
+	vk_buffer.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline.pipeline);
+	vk_buffer.dispatch(thread_group_x, thread_group_y, thread_group_z);
+
+	// The removed host fence also ordered read-only dispatches before later writers.
+	ShaderAccessBarrier(vk_buffer, vk::PipelineStageFlagBits::eComputeShader);
 	ResetBindings();
 }
 

src/graphics/host_gpu/renderer/renderContext.cpp

@@ -14,8 +14,7 @@ namespace Libs::Graphics {
 RenderContext::RenderContext(GraphicContext& graphics)
     : m_graphics(graphics), m_render_executor(*this), m_command_scheduler(*this, graphics),
       m_descriptor_cache(graphics), m_pipeline_cache(graphics, m_descriptor_cache),
-      m_sampler_cache(graphics),
-      m_gpu_resources(graphics, m_command_scheduler) {
+      m_sampler_cache(graphics), m_gpu_resources(graphics, m_command_scheduler) {
 	EXIT_NOT_IMPLEMENTED(!Common::Thread::IsMainThread());
 }
 
@@ -27,7 +26,7 @@ RenderContext::~RenderContext() {
 void RenderContext::InitializeGpu(VideoOut::VideoOutDriver* video_out) {
 	EXIT_IF(m_gpu != nullptr);
 	m_video_out = video_out;
-	m_gpu = std::make_unique<Gpu>(*this);
+	m_gpu       = std::make_unique<Gpu>(*this);
 	m_gpu_resources.SetGpu(m_gpu.get());
 }
 
@@ -99,8 +98,7 @@ void RenderContext::TriggerEopEvent(uint32_t context_id) {
 		    registration.eq, static_cast<uintptr_t>(registration.id),
 		    LibKernel::EventQueue::KERNEL_EVFILT_GRAPHICS,
 		    reinterpret_cast<void*>(static_cast<uintptr_t>(context_id)));
-		if (result == LibKernel::KERNEL_ERROR_EBADF ||
-		    result == LibKernel::KERNEL_ERROR_ENOENT) {
+		if (result == LibKernel::KERNEL_ERROR_EBADF || result == LibKernel::KERNEL_ERROR_ENOENT) {
 			DeleteEopEq(registration.eq, registration.id);
 			continue;
 		}

src/graphics/host_gpu/renderer/renderContext.h

@@ -6,12 +6,12 @@
 #include "common/common.h"
 #include "common/threads.h"
 #include "graphics/host_gpu/renderer/cache/bufferCache.h"
-#include "graphics/host_gpu/renderer/commandScheduler.h"
-#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
 #include "graphics/host_gpu/renderer/cache/gpuResourceManager.h"
-#include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
 #include "graphics/host_gpu/renderer/cache/samplerCache.h"
 #include "graphics/host_gpu/renderer/cache/textureCache.h"
+#include "graphics/host_gpu/renderer/commandScheduler.h"
+#include "graphics/host_gpu/renderer/pipeline/descriptorCache.h"
+#include "graphics/host_gpu/renderer/pipeline/pipelineCache.h"
 #include "kernel/eventQueue.h"
 
 #include <memory>
@@ -32,10 +32,10 @@ public:
 	~RenderContext();
 	KYTY_CLASS_NO_COPY(RenderContext);
 
-	[[nodiscard]] GraphicContext& GetGraphics() const noexcept { return m_graphics; }
-	void                          InitializeGpu(VideoOut::VideoOutDriver* video_out);
-	void                          ShutdownGpu();
-	[[nodiscard]] Gpu&            GetGpu() const;
+	[[nodiscard]] GraphicContext&           GetGraphics() const noexcept { return m_graphics; }
+	void                                    InitializeGpu(VideoOut::VideoOutDriver* video_out);
+	void                                    ShutdownGpu();
+	[[nodiscard]] Gpu&                      GetGpu() const;
 	[[nodiscard]] VideoOut::VideoOutDriver& GetVideoOut() const;
 
 	Common::Mutex&      GetMutex() { return m_mutex; }
@@ -56,18 +56,18 @@ private:
 	struct EopEqRegistration {
 		LibKernel::EventQueue::KernelEqueue    eq = LibKernel::EventQueue::KERNEL_EQUEUE_INVALID;
 		LibKernel::EventQueue::KernelEqueueRef queue;
-		int                                   id = 0;
+		int                                    id = 0;
 	};
 
-	GraphicContext&      m_graphics;
-	Common::Mutex        m_mutex;
-	RenderExecutor       m_render_executor;
-	CommandScheduler     m_command_scheduler;
-	DescriptorCache      m_descriptor_cache;
-	PipelineCache        m_pipeline_cache;
-	SamplerCache         m_sampler_cache;
-	GpuResourceManager   m_gpu_resources;
-	std::unique_ptr<Gpu> m_gpu;
+	GraphicContext&           m_graphics;
+	Common::Mutex             m_mutex;
+	RenderExecutor            m_render_executor;
+	CommandScheduler          m_command_scheduler;
+	DescriptorCache           m_descriptor_cache;
+	PipelineCache             m_pipeline_cache;
+	SamplerCache              m_sampler_cache;
+	GpuResourceManager        m_gpu_resources;
+	std::unique_ptr<Gpu>      m_gpu;
 	VideoOut::VideoOutDriver* m_video_out = nullptr;
 
 	Common::Mutex                  m_eop_mutex;

src/graphics/host_gpu/renderer/renderDraw.cpp

@@ -2,7 +2,6 @@
 
 #include "common/assert.h"
 #include "common/common.h"
-#include "common/emulatorConfig.h"
 #include "common/file.h"
 #include "common/logging/log.h"
 #include "common/profiler.h"
@@ -65,12 +64,32 @@ int32_t ResolveVertexOffset(uint32_t index_offset, const ShaderVertexInputInfo&
 	return 0;
 }
 
-static std::atomic<uint32_t> g_draw_state_log_count       = 0;
-static std::atomic<uint32_t> g_draw_input_log_count       = 0;
-static std::atomic<uint32_t> g_mrt_state_log_count        = 0;
-static std::atomic<uint32_t> g_shader_stage_log_count     = 0;
+static std::atomic<uint32_t> g_draw_state_log_count   = 0;
+static std::atomic<uint32_t> g_draw_input_log_count   = 0;
+static std::atomic<uint32_t> g_mrt_state_log_count    = 0;
+static std::atomic<uint32_t> g_shader_stage_log_count = 0;
+
 static std::atomic<uint32_t> g_framebuffer_skip_log_count = 0;
 
+static float ConvertPolygonOffsetConstantFactor(float guest_factor, const HW::PolyOffset& offset,
+                                                vk::Format host_depth_format) {
+	if (offset.db_is_float_fmt) {
+		return guest_factor;
+	}
+
+	int host_depth_bits = 0;
+	switch (host_depth_format) {
+		case vk::Format::eD16Unorm:
+		case vk::Format::eD16UnormS8Uint: host_depth_bits = 16; break;
+		case vk::Format::eD24UnormS8Uint: host_depth_bits = 24; break;
+		default:
+			// A fixed-point guest bias cannot be represented exactly by a floating-point host
+			// attachment without VK_EXT_depth_bias_control.
+			return guest_factor;
+	}
+	return std::ldexp(guest_factor, host_depth_bits + offset.neg_num_db_bits);
+}
+
 static const char* RenderColorTypeName(RenderColorType type) {
 	switch (type) {
 		case RenderColorType::NoColorOutput: return "NoColorOutput";
@@ -336,6 +355,22 @@ static PipelineDynamicParameters BuildGraphicsDynamicParams(const RenderCommandB
 	ret.stencil_front      = depth.stencil_dynamic_front;
 	ret.stencil_back       = depth.stencil_dynamic_back;
 
+	const auto& mode        = ctx.GetModeControl();
+	const auto& poly_offset = ctx.GetPolyOffset();
+	const bool  use_front   = mode.poly_offset_front_enable && !mode.cull_front;
+	const bool  use_back    = mode.poly_offset_back_enable && !mode.cull_back;
+	ret.depth_bias_enable   = use_front || use_back;
+	if (ret.depth_bias_enable) {
+		// Vulkan has one bias for both faces. Prefer a visible front face when both are enabled.
+		const float guest_constant_factor =
+		    use_front ? poly_offset.front_offset : poly_offset.back_offset;
+		ret.depth_bias_constant_factor =
+		    ConvertPolygonOffsetConstantFactor(guest_constant_factor, poly_offset, depth.format);
+		ret.depth_bias_clamp = poly_offset.clamp;
+		ret.depth_bias_slope_factor =
+		    (use_front ? poly_offset.front_scale : poly_offset.back_scale) / 16.0f;
+	}
+
 	// CB_COLOR_CONTROL.operation controls special CB operations, not the normal color component
 	// write mask. Use CB_TARGET_MASK here so scanout passes with mode=Disable do not go black.
 	for (uint32_t i = 0; i < RENDER_COLOR_ATTACHMENTS_MAX; i++) {
@@ -380,6 +415,13 @@ static void SetDynamicParams(const RenderCommandBuffer& buffer, vk::CommandBuffe
 	}
 	vk_buffer.setLineWidth(line_width);
 
+	vk_buffer.setDepthBiasEnable(dynamic_params.depth_bias_enable ? VK_TRUE : VK_FALSE);
+	if (dynamic_params.depth_bias_enable) {
+		vk_buffer.setDepthBias(dynamic_params.depth_bias_constant_factor,
+		                       dynamic_params.depth_bias_clamp,
+		                       dynamic_params.depth_bias_slope_factor);
+	}
+
 	if (dynamic_params.stencil_test_enable) {
 		vk_buffer.setStencilCompareMask(vk::StencilFaceFlagBits::eFront,
 		                                dynamic_params.stencil_front.compareMask);
@@ -650,11 +692,9 @@ static bool ConsumeMetadataColorOperation(const RenderCommandBuffer& buffer) {
 }
 
 struct DrawEmitInfo {
-	bool     indexed           = false;
-	bool     draw_prim7_as_ngg = false;
-	uint32_t draw_vertex_count = 0;
-	int32_t  vertex_offset     = 0;
-	uint32_t first_vertex      = 0;
+	bool     indexed       = false;
+	int32_t  vertex_offset = 0;
+	uint32_t first_vertex  = 0;
 };
 
 struct DrawIndexBufferSource {
@@ -667,12 +707,10 @@ struct DrawIndexBufferSource {
 
 struct PreparedIndexBuffer {
 	std::shared_ptr<void> owner;
-	vk::Buffer            buffer   = nullptr;
-	uint64_t              address  = 0;
-	uint64_t              size     = 0;
-	vk::DeviceSize        offset   = 0;
-	vk::IndexType         type     = vk::IndexType::eUint16;
-	bool                  streamed = false;
+	vk::Buffer            buffer = nullptr;
+	uint64_t              size   = 0;
+	vk::DeviceSize        offset = 0;
+	vk::IndexType         type   = vk::IndexType::eUint16;
 };
 
 static uint64_t VertexBufferDescriptorSize(const ShaderVertexInputBuffer& buffer) {
@@ -689,11 +727,24 @@ struct VertexBufferRange {
 	[[nodiscard]] uint64_t RequestedSize() const { return requested_end - base_address; }
 };
 
-static std::vector<BufferBinding> AcquireVertexBuffers(RenderCommandBuffer&         buffer,
-                                                       const ShaderVertexInputInfo& vs_input_info) {
+struct PreparedVertexBuffers {
+	static constexpr uint32_t MaxBuffers = ShaderVertexInputInfo::RES_MAX;
+
+	std::array<vk::Buffer, MaxBuffers>            buffers {};
+	std::array<vk::DeviceSize, MaxBuffers>        offsets {};
+	std::array<std::shared_ptr<void>, MaxBuffers> owners {};
+	uint32_t                                      count       = 0;
+	uint32_t                                      owner_count = 0;
+};
+
+static PreparedVertexBuffers AcquireVertexBuffers(RenderCommandBuffer&         buffer,
+                                                  const ShaderVertexInputInfo& vs_input_info) {
+	EXIT_IF(vs_input_info.buffers_num < 0 ||
+	        vs_input_info.buffers_num > ShaderVertexInputInfo::RES_MAX);
+
 	// Collect the non-empty guest vertex ranges.
-	std::vector<VertexBufferRange> ranges;
-	ranges.reserve(vs_input_info.buffers_num);
+	std::array<VertexBufferRange, ShaderVertexInputInfo::RES_MAX> ranges {};
+	uint32_t                                                      range_count = 0;
 	for (int i = 0; i < vs_input_info.buffers_num; i++) {
 		const auto& vertex = vs_input_info.buffers[i];
 		const auto  size   = VertexBufferDescriptorSize(vertex);
@@ -704,27 +755,31 @@ static std::vector<BufferBinding> AcquireVertexBuffers(RenderCommandBuffer&
 			EXIT("invalid vertex buffer range: addr=0x%016" PRIx64 " size=0x%016" PRIx64 "\n",
 			     vertex.addr, size);
 		}
-		ranges.push_back({vertex.addr, vertex.addr + size});
+		ranges[range_count++] = {vertex.addr, vertex.addr + size};
 	}
 
-	std::ranges::sort(ranges, [](const VertexBufferRange& left, const VertexBufferRange& right) {
-		return left.base_address < right.base_address;
-	});
+	std::sort(ranges.begin(), ranges.begin() + range_count,
+	          [](const VertexBufferRange& left, const VertexBufferRange& right) {
+		          return left.base_address < right.base_address;
+	          });
 
 	// Merge overlapping or touching ranges before acquiring host buffers.
-	std::vector<VertexBufferRange> merged_ranges;
-	merged_ranges.reserve(ranges.size());
-	for (const auto& range: ranges) {
-		if (!merged_ranges.empty() && merged_ranges.back().requested_end >= range.base_address) {
-			merged_ranges.back().requested_end =
-			    std::max(merged_ranges.back().requested_end, range.requested_end);
+	std::array<VertexBufferRange, ShaderVertexInputInfo::RES_MAX> merged_ranges {};
+	uint32_t                                                      merged_count = 0;
+	for (uint32_t i = 0; i < range_count; i++) {
+		const auto& range = ranges[i];
+		if (merged_count != 0 &&
+		    merged_ranges[merged_count - 1].requested_end >= range.base_address) {
+			merged_ranges[merged_count - 1].requested_end =
+			    std::max(merged_ranges[merged_count - 1].requested_end, range.requested_end);
 			continue;
 		}
-		merged_ranges.push_back(range);
+		merged_ranges[merged_count++] = {range.base_address, range.requested_end};
 	}
 
 	auto& cache = buffer.GetContext().GetBufferCache();
-	for (auto& range: merged_ranges) {
+	for (uint32_t i = 0; i < merged_count; i++) {
+		auto& range = merged_ranges[i];
 		// PPSA20298
 		const auto size =
 		    Libs::LibKernel::Memory::ClampRangeSize(range.base_address, range.RequestedSize());
@@ -733,30 +788,47 @@ static std::vector<BufferBinding> AcquireVertexBuffers(RenderCommandBuffer&
 	}
 
 	// Rebuild slot bindings, offsetting non-empty slots into their acquired merged range.
-	std::vector<BufferBinding> bindings;
-	bindings.reserve(vs_input_info.buffers_num);
+	PreparedVertexBuffers prepared;
+	prepared.count = static_cast<uint32_t>(vs_input_info.buffers_num);
+	std::shared_ptr<Buffer> null_owner;
+	vk::Buffer              null_buffer = nullptr;
 	for (int i = 0; i < vs_input_info.buffers_num; i++) {
 		const auto& vertex = vs_input_info.buffers[i];
 		const auto  size   = VertexBufferDescriptorSize(vertex);
 		if (size == 0) {
-			auto owner = cache.ObtainNullBuffer();
-			bindings.push_back({owner, owner->Handle(), 0});
+			if (null_owner == nullptr) {
+				null_owner  = cache.ObtainNullBuffer();
+				null_buffer = null_owner->Handle();
+			}
+			prepared.buffers[i] = null_buffer;
+			prepared.offsets[i] = 0;
 			continue;
 		}
 
-		const auto range = std::ranges::find_if(merged_ranges, [&](const VertexBufferRange& value) {
-			return vertex.addr >= value.base_address && vertex.addr < value.acquired_end;
-		});
-		if (range == merged_ranges.end()) {
+		const auto range = std::find_if(merged_ranges.begin(), merged_ranges.begin() + merged_count,
+		                                [&](const VertexBufferRange& value) {
+			                                return vertex.addr >= value.base_address &&
+			                                       vertex.addr < value.acquired_end;
+		                                });
+		if (range == merged_ranges.begin() + merged_count) {
 			EXIT("vertex buffer address is outside the acquired range: addr=0x%016" PRIx64 "\n",
 			     vertex.addr);
 		}
 
-		auto binding = range->binding;
-		binding.offset += vertex.addr - range->base_address;
-		bindings.push_back(std::move(binding));
+		prepared.buffers[i] = range->binding.buffer;
+		prepared.offsets[i] = range->binding.offset + vertex.addr - range->base_address;
 	}
-	return bindings;
+
+	if (null_owner != nullptr) {
+		prepared.owners[prepared.owner_count++] = std::move(null_owner);
+	}
+	for (uint32_t i = 0; i < merged_count; i++) {
+		if (merged_ranges[i].binding.owner != nullptr) {
+			EXIT_IF(prepared.owner_count >= PreparedVertexBuffers::MaxBuffers);
+			prepared.owners[prepared.owner_count++] = std::move(merged_ranges[i].binding.owner);
+		}
+	}
+	return prepared;
 }
 
 static void SetDrawDebugPhase(RenderCommandBuffer& buffer, uint64_t submit_id,
@@ -767,7 +839,7 @@ static void SetDrawDebugPhase(RenderCommandBuffer& buffer, uint64_t submit_id,
 	                    draw.flags, draw.instance_count, draw.first_instance);
 }
 
-static bool GetDrawTopology(const HW::UserConfig& ucfg, bool auto_draw, bool use_ngg_rectlist_draw,
+static bool GetDrawTopology(const HW::UserConfig& ucfg, bool auto_draw,
                             vk::PrimitiveTopology& topology) {
 
 	topology = vk::PrimitiveTopology::ePointList;
@@ -791,8 +863,7 @@ static bool GetDrawTopology(const HW::UserConfig& ucfg, bool auto_draw, bool use
 			topology = vk::PrimitiveTopology::eTriangleStrip;
 			break;
 		case Prospero::PrimitiveType::kRectList:
-			topology = (auto_draw && use_ngg_rectlist_draw ? vk::PrimitiveTopology::eTriangleStrip
-			                                               : vk::PrimitiveTopology::eTriangleList);
+			topology = vk::PrimitiveTopology::ePatchList;
 			break;
 		case Prospero::PrimitiveType::kRectListLegacy:
 			if (!auto_draw) {
@@ -889,10 +960,9 @@ static void RefreshShaders(RenderCommandBuffer& buffer, const DrawCallInfo& draw
 	}
 }
 
-static std::vector<BufferBinding> PrepareVertexBuffers(uint64_t                     submit_id,
-                                                       RenderCommandBuffer&         buffer,
-                                                       const DrawCallInfo&          draw,
-                                                       const ShaderVertexInputInfo& vs_input_info) {
+static PreparedVertexBuffers PrepareVertexBuffers(uint64_t submit_id, RenderCommandBuffer& buffer,
+                                                  const DrawCallInfo&          draw,
+                                                  const ShaderVertexInputInfo& vs_input_info) {
 	EXIT_IF(draw.name == nullptr);
 	(void)submit_id;
 
@@ -900,12 +970,6 @@ static std::vector<BufferBinding> PrepareVertexBuffers(uint64_t
 	return AcquireVertexBuffers(buffer, vs_input_info);
 }
 
-static void RebindVertexBuffers(RenderCommandBuffer&         buffer,
-                                const ShaderVertexInputInfo& vs_input_info,
-                                std::vector<BufferBinding>&  bindings) {
-	bindings = AcquireVertexBuffers(buffer, vs_input_info);
-}
-
 static PreparedIndexBuffer PrepareIndexBuffer(RenderCommandBuffer&         buffer,
                                               const DrawIndexBufferSource& source) {
 	PreparedIndexBuffer prepared;
@@ -913,16 +977,14 @@ static PreparedIndexBuffer PrepareIndexBuffer(RenderCommandBuffer&         buffe
 		return prepared;
 	}
 	EXIT_IF(source.size == 0);
-	prepared.address = source.address;
-	prepared.size    = source.size;
-	prepared.type    = source.type;
+	prepared.size = source.size;
+	prepared.type = source.type;
 	if (source.host_data != nullptr) {
 		auto binding =
 		    buffer.GetContext().GetBufferCache().UploadTransient(source.host_data, source.size, 16);
-		prepared.owner    = std::move(binding.owner);
-		prepared.buffer   = binding.buffer;
-		prepared.offset   = binding.offset;
-		prepared.streamed = true;
+		prepared.owner  = std::move(binding.owner);
+		prepared.buffer = binding.buffer;
+		prepared.offset = binding.offset;
 	} else {
 		auto binding =
 		    buffer.GetContext().GetBufferCache().ObtainBuffer(buffer, source.address, source.size);
@@ -933,36 +995,29 @@ static PreparedIndexBuffer PrepareIndexBuffer(RenderCommandBuffer&         buffe
 	return prepared;
 }
 
-static void RebindIndexBuffer(RenderCommandBuffer& buffer, PreparedIndexBuffer& prepared) {
-	if (prepared.size == 0 || prepared.streamed) {
-		return;
-	}
-	auto binding =
-	    buffer.GetContext().GetBufferCache().ObtainBuffer(buffer, prepared.address, prepared.size);
-	prepared.owner  = std::move(binding.owner);
-	prepared.buffer = binding.buffer;
-	prepared.offset = binding.offset;
-}
-
 static void CommitVertexBuffers(RenderCommandBuffer& buffer, vk::CommandBuffer vk_buffer,
-                                std::vector<BufferBinding>& bindings) {
-	for (uint32_t slot = 0; slot < bindings.size(); slot++) {
-		auto& binding = bindings[slot];
-		if (binding.owner != nullptr) {
-			buffer.RetainResourceUntilFence(binding.owner);
+                                PreparedVertexBuffers& prepared) {
+	for (uint32_t i = 0; i < prepared.owner_count; i++) {
+		if (prepared.owners[i] != nullptr) {
+			buffer.RetainResourceUntilFence(std::move(prepared.owners[i]));
 		}
-		EXIT_IF(binding.buffer == nullptr);
-		vk_buffer.bindVertexBuffers(slot, 1, &binding.buffer, &binding.offset);
+	}
+	for (uint32_t i = 0; i < prepared.count; i++) {
+		EXIT_IF(prepared.buffers[i] == nullptr);
+	}
+	if (prepared.count != 0) {
+		vk_buffer.bindVertexBuffers(0, prepared.count, prepared.buffers.data(),
+		                            prepared.offsets.data());
 	}
 }
 
 static void CommitIndexBuffer(RenderCommandBuffer& buffer, vk::CommandBuffer vk_buffer,
-                              const PreparedIndexBuffer& prepared) {
+                              PreparedIndexBuffer& prepared) {
 	if (prepared.size == 0) {
 		return;
 	}
 	if (prepared.owner != nullptr) {
-		buffer.RetainResourceUntilFence(prepared.owner);
+		buffer.RetainResourceUntilFence(std::move(prepared.owner));
 	}
 	EXIT_IF(prepared.buffer == nullptr);
 	vk_buffer.bindIndexBuffer(prepared.buffer, prepared.offset, prepared.type);
@@ -991,20 +1046,6 @@ static void LogDrawStateIfNeeded(const RenderCommandBuffer& buffer, const DrawCa
 	// LogDrawTextureState(draw.name, state.color_info[0], state.ps_input_info);
 }
 
-static bool IsHostExpandedRectListDrawSupported(const ShaderVertexInputInfo& vs_input_info,
-                                                const DrawCallInfo&          draw,
-                                                const DrawEmitInfo&          emit) {
-	if (!emit.draw_prim7_as_ngg) {
-		return true;
-	}
-
-	if (vs_input_info.buffers_num != 0) {
-		return false;
-	}
-
-	return draw.index_count == 3 || draw.index_count == emit.draw_vertex_count;
-}
-
 static void EmitDrawPrimitives(const HW::UserConfig& ucfg, vk::CommandBuffer vk_buffer,
                                const ShaderVertexInputInfo& vs_input_info, const DrawCallInfo& draw,
                                const DrawEmitInfo& emit) {
@@ -1017,22 +1058,12 @@ static void EmitDrawPrimitives(const HW::UserConfig& ucfg, vk::CommandBuffer vk_
 		case Prospero::PrimitiveType::kTriList:
 		case Prospero::PrimitiveType::kTriFan:
 		case Prospero::PrimitiveType::kTriStrip:
-			if (emit.indexed) {
-				vk_buffer.drawIndexed(draw.index_count, draw.instance_count, 0, emit.vertex_offset,
-				                      draw.first_instance);
-			} else {
-				vk_buffer.draw(draw.index_count, draw.instance_count, emit.first_vertex,
-				               draw.first_instance);
-			}
-			break;
 		case Prospero::PrimitiveType::kRectList:
 			if (emit.indexed) {
 				vk_buffer.drawIndexed(draw.index_count, draw.instance_count, 0, emit.vertex_offset,
 				                      draw.first_instance);
 			} else {
-				EXIT_NOT_IMPLEMENTED(
-				    !IsHostExpandedRectListDrawSupported(vs_input_info, draw, emit));
-				vk_buffer.draw(emit.draw_vertex_count, draw.instance_count, emit.first_vertex,
+				vk_buffer.draw(draw.index_count, draw.instance_count, emit.first_vertex,
 				               draw.first_instance);
 			}
 			break;
@@ -1074,8 +1105,6 @@ void RenderExecutor::ExecutePreparedDraw(uint64_t submit_id, RenderCommandBuffer
 	                                               state.ps_input_info.stage, state.ps_active);
 	auto vertex_bindings = PrepareVertexBuffers(submit_id, buffer, draw, state.vs_input_info);
 	auto index_binding   = PrepareIndexBuffer(buffer, index_source);
-	RebindVertexBuffers(buffer, state.vs_input_info, vertex_bindings);
-	RebindIndexBuffer(buffer, index_binding);
 	state.rendering =
 	    AcquireRenderTargets(buffer, state.color_info, state.color_count, state.depth_info);
 
@@ -1159,7 +1188,7 @@ void RenderExecutor::DrawIndex(uint64_t submit_id, RenderCommandBuffer& buffer,
 	                    reinterpret_cast<uint64_t>(index_addr));
 
 	Common::LockGuard lock(m_context.GetMutex());
-	if (index_count == 0) {
+	if (index_count == 0 || instance_count == 0) {
 		return;
 	}
 
@@ -1201,7 +1230,7 @@ void RenderExecutor::DrawIndex(uint64_t submit_id, RenderCommandBuffer& buffer,
 	hw_check(buffer);
 
 	vk::PrimitiveTopology topology = vk::PrimitiveTopology::ePointList;
-	if (!GetDrawTopology(ucfg, false, false, topology)) {
+	if (!GetDrawTopology(ucfg, false, topology)) {
 		return;
 	}
 
@@ -1229,10 +1258,6 @@ void RenderExecutor::DrawIndex(uint64_t submit_id, RenderCommandBuffer& buffer,
 
 	EXIT_NOT_IMPLEMENTED(flags != 0);
 	EXIT_NOT_IMPLEMENTED(type != 1);
-	if (instance_count == 0) {
-		instance_count = 1;
-	}
-
 	const DrawCallInfo    draw {"DrawIndex",    CommandBufferDebugOp::DrawIndex,
 	                            index_count,    flags,
 	                            instance_count, first_instance};
@@ -1292,7 +1317,7 @@ void RenderExecutor::DrawAuto(uint64_t submit_id, RenderCommandBuffer& buffer, u
 	                    index_count, flags, first_vertex, instance_count, first_instance);
 
 	Common::LockGuard lock(m_context.GetMutex());
-	if (index_count == 0) {
+	if (index_count == 0 || instance_count == 0) {
 		return;
 	}
 
@@ -1330,10 +1355,6 @@ void RenderExecutor::DrawAuto(uint64_t submit_id, RenderCommandBuffer& buffer, u
 	hw_check(buffer);
 
 	EXIT_NOT_IMPLEMENTED(flags != 0);
-	if (instance_count == 0) {
-		instance_count = 1;
-	}
-
 	const DrawCallInfo draw {"DrawIndexAuto", CommandBufferDebugOp::DrawIndexAuto,
 	                         index_count,     flags,
 	                         instance_count,  first_instance};
@@ -1345,20 +1366,15 @@ void RenderExecutor::DrawAuto(uint64_t submit_id, RenderCommandBuffer& buffer, u
 		return;
 	}
 
-	vk::PrimitiveTopology topology              = vk::PrimitiveTopology::ePointList;
-	const bool            use_ngg_rectlist_draw = Config::NggRectlistDrawEnabled();
-
-	if (!GetDrawTopology(ucfg, true, use_ngg_rectlist_draw, topology)) {
+	vk::PrimitiveTopology topology = vk::PrimitiveTopology::ePointList;
+	if (!GetDrawTopology(ucfg, true, topology)) {
 		ResetBindings();
 		return;
 	}
-	const bool draw_prim7_as_ngg =
-	    (use_ngg_rectlist_draw &&
-	     ucfg.GetPrimType() == Prospero::GpuEnumValue(Prospero::PrimitiveType::kRectList));
-
 	RefreshShaders(buffer, draw, false, state);
 
-	if (draw_prim7_as_ngg && state.vs_input_info.buffers_num == 0 &&
+	const bool rect_list = topology == vk::PrimitiveTopology::ePatchList;
+	if (rect_list && state.vs_input_info.buffers_num == 0 &&
 	    state.vs_input_info.param_export_mask == 0 && state.ps_input_info.input_num != 0) {
 		if (graphics_debug_dump_enabled()) {
 			LOGF("DrawIndexAuto: skipping rect-list draw with no VS param exports and PS inputs: "
@@ -1375,13 +1391,10 @@ void RenderExecutor::DrawAuto(uint64_t submit_id, RenderCommandBuffer& buffer, u
 	                         Prospero::GpuEnumValue(Prospero::PrimitiveType::kRectListLegacy),
 	                     0, nullptr);
 
-	const uint32_t draw_vertex_count = (draw_prim7_as_ngg ? 4u : index_count);
-	const auto     vertex_offset = ResolveVertexOffset(ucfg.GetIndexOffset(), state.vs_input_info) +
-	                               static_cast<int32_t>(first_vertex);
-	DrawEmitInfo   emit {};
-	emit.draw_prim7_as_ngg = draw_prim7_as_ngg;
-	emit.draw_vertex_count = draw_vertex_count;
-	emit.first_vertex      = static_cast<uint32_t>(vertex_offset);
+	const auto   vertex_offset = ResolveVertexOffset(ucfg.GetIndexOffset(), state.vs_input_info) +
+	                             static_cast<int32_t>(first_vertex);
+	DrawEmitInfo emit {};
+	emit.first_vertex = static_cast<uint32_t>(vertex_offset);
 
 	DrawIndexBufferSource index_source {};
 	ExecutePreparedDraw(submit_id, buffer, draw, state, topology, emit, index_source, false, false,

src/graphics/host_gpu/renderer/renderDraw.h

@@ -15,12 +15,16 @@ struct ShaderVertexInputInfo;
 
 struct PipelineDynamicParameters {
 	bool stencil_test_enable = false;
+	bool depth_bias_enable   = false;
 
 	float viewport_scale[3]  = {};
 	float viewport_offset[3] = {};
 	int   scissor_ltrb[4]    = {0};
 
 	float    line_width                                       = 1.0f;
+	float    depth_bias_constant_factor                       = 0.0f;
+	float    depth_bias_clamp                                 = 0.0f;
+	float    depth_bias_slope_factor                          = 0.0f;
 	uint32_t color_write_count                                = 1;
 	bool     color_write_enable[RENDER_COLOR_ATTACHMENTS_MAX] = {true, true};
 
@@ -33,10 +37,6 @@ struct PipelineDynamicParameters {
 static_assert(std::is_trivially_copyable_v<PipelineDynamicParameters>);
 static_assert(std::is_standard_layout_v<PipelineDynamicParameters>);
 static_assert(alignof(PipelineDynamicParameters) == 1);
-static_assert(sizeof(PipelineDynamicParameters) ==
-              sizeof(bool) + sizeof(float[3]) + sizeof(float[3]) + sizeof(int[4]) + sizeof(float) +
-                  sizeof(uint32_t) + sizeof(bool[RENDER_COLOR_ATTACHMENTS_MAX]) +
-                  sizeof(PipelineStencilDynamicState) * 2);
 
 [[nodiscard]] int32_t ResolveVertexOffset(uint32_t                     index_offset,
                                           const ShaderVertexInputInfo& vs_input_info);

src/graphics/host_gpu/renderer/renderTarget.h

@@ -12,13 +12,13 @@ namespace Libs::Graphics {
 static constexpr uint32_t RENDER_COLOR_ATTACHMENTS_MAX = 8;
 
 struct RenderAttachment {
-	vk::ImageView           image_view   = nullptr;
-	vk::ImageLayout         image_layout = vk::ImageLayout::eUndefined;
-	std::array<uint32_t, 4> clear_value  = {};
+	vk::ImageView           image_view    = nullptr;
+	vk::ImageLayout         image_layout  = vk::ImageLayout::eUndefined;
+	std::array<uint32_t, 4> clear_value   = {};
 	bool                    is_clear      = false;
-	bool                    has_depth    = false;
-	bool                    depth_clear  = false;
-	bool                    has_stencil  = false;
+	bool                    has_depth     = false;
+	bool                    depth_clear   = false;
+	bool                    has_stencil   = false;
 	bool                    stencil_clear = false;
 
 	bool operator==(const RenderAttachment&) const = default;

src/graphics/host_gpu/renderer/sync.cpp

@@ -251,9 +251,9 @@ uint64_t PrepareVideoOutFlip(CommandBuffer& buffer, int handle, int index, int f
                              int64_t flip_arg) {
 	for (;;) {
 		uint64_t   request_id = 0;
-		auto&      video_out = buffer.GetContext().GetVideoOut();
-		const auto result = video_out.SubmitFlipFromGpu(
-		    buffer, handle, index, flip_mode, flip_arg, request_id);
+		auto&      video_out  = buffer.GetContext().GetVideoOut();
+		const auto result =
+		    video_out.SubmitFlipFromGpu(buffer, handle, index, flip_mode, flip_arg, request_id);
 		if (result == OK) {
 			EXIT_IF(request_id == 0);
 			return request_id;

src/graphics/host_gpu/vma.cpp

@@ -122,9 +122,9 @@ uint64_t GraphicContext::GetDeviceMemoryUsage() const {
 	    physical_device_properties.deviceType == vk::PhysicalDeviceType::eDiscreteGpu;
 	uint64_t usage = 0;
 	for (uint32_t heap = 0; heap < physical_device_memory_properties.memoryHeapCount; heap++) {
-		const bool device_local = static_cast<bool>(
-		    physical_device_memory_properties.memoryHeaps[heap].flags &
-		    vk::MemoryHeapFlagBits::eDeviceLocal);
+		const bool device_local =
+		    static_cast<bool>(physical_device_memory_properties.memoryHeaps[heap].flags &
+		                      vk::MemoryHeapFlagBits::eDeviceLocal);
 		if (!discrete || device_local) {
 			usage += budgets[heap].usage;
 		}
@@ -144,7 +144,7 @@ uint64_t GraphicContext::GetTotalMemoryBudget() const {
 	uint64_t local  = 0;
 	uint64_t usage  = 0;
 	for (uint32_t heap = 0; heap < physical_device_memory_properties.memoryHeapCount; heap++) {
-		const auto& properties   = physical_device_memory_properties.memoryHeaps[heap];
+		const auto& properties = physical_device_memory_properties.memoryHeaps[heap];
 		const bool  device_local =
 		    static_cast<bool>(properties.flags & vk::MemoryHeapFlagBits::eDeviceLocal);
 		if (device_local) {
@@ -159,9 +159,8 @@ uint64_t GraphicContext::GetTotalMemoryBudget() const {
 		return budget - std::min<uint64_t>(budget / 8, 1024ull * 1024 * 1024);
 	}
 	constexpr uint64_t system_reserve = 8ull * 1024 * 1024 * 1024;
-	const auto available = budget > usage ? budget - usage : uint64_t {0};
-	return std::max(local,
-	                available > system_reserve ? available - system_reserve : uint64_t {0});
+	const auto         available      = budget > usage ? budget - usage : uint64_t {0};
+	return std::max(local, available > system_reserve ? available - system_reserve : uint64_t {0});
 }
 
 void GraphicContext::CreateBuffer(uint64_t size, VulkanBuffer& buffer) {

src/graphics/host_gpu/vulkanCommon.cpp

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

src/graphics/presentation/imeDialogOverlay.cpp

@@ -0,0 +1,876 @@
+#include "graphics/presentation/imeDialogOverlay.h"
+
+#include "SDL.h"
+#include "common/assert.h"
+#include "common/stringUtils.h"
+#include "graphics/host_gpu/graphicContext.h"
+#include "imgui.h"
+#include "imgui_impl_vulkan.h"
+#include "libs/controller.h"
+#include "libs/imeDialog.h"
+
+#include <algorithm>
+#include <array>
+#include <atomic>
+#include <chrono>
+#include <cmath>
+#include <cstdint>
+#include <deque>
+#include <mutex>
+#include <span>
+#include <string>
+#include <string_view>
+#include <utility>
+
+namespace Libs::Graphics {
+
+namespace {
+
+namespace Ime = Libs::Dialog::ImeDialog;
+
+constexpr size_t INPUT_QUEUE_CAPACITY = 128;
+
+enum class InputKind : uint8_t {
+	Button,
+	Axis,
+	MousePosition,
+	MouseButton,
+	MouseWheel,
+	ResetController
+};
+
+struct InputEvent {
+	InputKind kind;
+	uint64_t  generation;
+	int       id;
+	float     x;
+	float     y;
+};
+
+struct VisibilityUpdate {
+	uint64_t generation;
+	bool     visible;
+	bool     capture_controller;
+	bool     capture_keyboard;
+	bool     multiline;
+};
+
+std::atomic<Uint32>          g_visibility_event {static_cast<Uint32>(-1)};
+std::mutex                   g_visibility_mutex;
+std::mutex                   g_input_mutex;
+std::deque<InputEvent>       g_input_events;
+std::deque<VisibilityUpdate> g_visibility_updates;
+size_t                       g_missing_visibility_wakeups = 0;
+bool                         g_input_reset_requested      = false;
+uint16_t                     g_last_external_keycode      = 0;
+uint32_t                     g_last_external_status       = 0;
+uint64_t                     g_input_generation           = 0;
+bool                         g_input_active               = false;
+bool                         g_input_controller           = false;
+bool                         g_input_keyboard             = false;
+bool                         g_input_multiline            = false;
+bool                         g_input_lifecycle_active     = false;
+bool                         g_active                     = false;
+bool                         g_controller_captured        = false;
+uint64_t                     g_generation                 = 0;
+
+void ClearInputEvents() {
+	std::scoped_lock lock(g_input_mutex);
+	g_input_events.clear();
+	g_input_reset_requested = false;
+}
+
+void QueueInput(InputEvent event) {
+	std::scoped_lock lock(g_input_mutex);
+	const bool       replaceable =
+	    event.kind == InputKind::Axis || event.kind == InputKind::MousePosition;
+	if (replaceable && !g_input_events.empty()) {
+		auto& last = g_input_events.back();
+		if (last.generation == event.generation && last.kind == event.kind && last.id == event.id) {
+			last = event;
+			return;
+		}
+	}
+	if (g_input_events.size() == INPUT_QUEUE_CAPACITY) {
+		g_input_events.clear();
+		g_input_reset_requested = true;
+	}
+	g_input_events.push_back(event);
+}
+
+void RetryVisibilityWakeup() {
+	const Uint32 type = g_visibility_event.load(std::memory_order_acquire);
+	if (type == static_cast<Uint32>(-1)) {
+		return;
+	}
+	std::scoped_lock lock(g_input_mutex);
+	if (g_missing_visibility_wakeups == 0) {
+		return;
+	}
+	SDL_Event event {};
+	event.type = type;
+	if (SDL_PushEvent(&event) > 0) {
+		g_missing_visibility_wakeups--;
+	}
+}
+
+void OnVisibilityChanged(bool visible, uint64_t generation) {
+	std::scoped_lock visibility_lock(g_visibility_mutex);
+	if (!g_input_lifecycle_active) {
+		return;
+	}
+	if (generation < g_generation || (generation == g_generation && visible == g_active)) {
+		return;
+	}
+
+	g_generation            = generation;
+	bool capture_controller = false;
+	bool capture_keyboard   = false;
+	bool multiline          = false;
+	if (visible) {
+		Ime::HostSnapshot snapshot;
+		if (Ime::GetHostSnapshot(&snapshot) && snapshot.generation == generation) {
+			capture_controller = (snapshot.disable_device & Ime::DISABLE_DEVICE_CONTROLLER) == 0;
+			capture_keyboard   = (snapshot.disable_device & Ime::DISABLE_DEVICE_EXT_KEYBOARD) == 0;
+			multiline          = (snapshot.option & Ime::OPTION_MULTILINE) != 0;
+		} else {
+			visible = false;
+		}
+	}
+	const bool was_controller = std::exchange(g_controller_captured, capture_controller);
+	if (capture_controller || was_controller) {
+		Controller::ControllerResetInputState();
+	}
+	g_active = visible;
+
+	const Uint32 type = g_visibility_event.load(std::memory_order_acquire);
+	if (type != static_cast<Uint32>(-1)) {
+		SDL_Event event {};
+		event.type = type;
+		std::scoped_lock input_lock(g_input_mutex);
+		g_visibility_updates.push_back(
+		    {generation, visible, capture_controller, capture_keyboard, multiline});
+		if (g_missing_visibility_wakeups != 0 || SDL_PushEvent(&event) <= 0) {
+			g_missing_visibility_wakeups++;
+		}
+	}
+}
+
+uint32_t ExternalKeyStatus(SDL_Keymod modifiers, bool character_valid) {
+	uint32_t status = 0x00000001 | (character_valid ? 0x00000002 : 0);
+	if ((modifiers & KMOD_LCTRL) != 0) status |= 0x00000100;
+	if ((modifiers & KMOD_LSHIFT) != 0) status |= 0x00000200;
+	if ((modifiers & KMOD_LALT) != 0) status |= 0x00000400;
+	if ((modifiers & KMOD_LGUI) != 0) status |= 0x00000800;
+	if ((modifiers & KMOD_RCTRL) != 0) status |= 0x00001000;
+	if ((modifiers & KMOD_RSHIFT) != 0) status |= 0x00002000;
+	if ((modifiers & KMOD_RALT) != 0) status |= 0x00004000;
+	if ((modifiers & KMOD_RGUI) != 0) status |= 0x00008000;
+	if ((modifiers & KMOD_NUM) != 0) status |= 0x00010000;
+	if ((modifiers & KMOD_CAPS) != 0) status |= 0x00020000;
+	return status;
+}
+
+Ime::ExternalInput MakeExternalInput(Ime::ExternalAction action, uint16_t keycode,
+                                     uint32_t status) {
+	Ime::ExternalInput input {};
+	input.key.keycode = keycode;
+	input.key.status  = status;
+	input.key.type    = 4;
+	input.action      = action;
+	return input;
+}
+
+std::u16string Utf8ToUtf16(std::string_view text) {
+	std::u16string result;
+	result.reserve(text.size());
+	for (size_t i = 0; i < text.size();) {
+		const auto first     = static_cast<uint8_t>(text[i++]);
+		uint32_t   codepoint = 0;
+		uint32_t   remaining = 0;
+		if (first < 0x80) {
+			codepoint = first;
+		} else if ((first & 0xe0) == 0xc0) {
+			codepoint = first & 0x1f;
+			remaining = 1;
+		} else if ((first & 0xf0) == 0xe0) {
+			codepoint = first & 0x0f;
+			remaining = 2;
+		} else if ((first & 0xf8) == 0xf0) {
+			codepoint = first & 0x07;
+			remaining = 3;
+		} else {
+			continue;
+		}
+		if (i + remaining > text.size()) {
+			break;
+		}
+		bool valid = true;
+		for (uint32_t j = 0; j < remaining; j++) {
+			const auto next = static_cast<uint8_t>(text[i++]);
+			if ((next & 0xc0) != 0x80) {
+				valid = false;
+				break;
+			}
+			codepoint = (codepoint << 6) | (next & 0x3f);
+		}
+		if (!valid || codepoint > 0x10ffff || (codepoint >= 0xd800 && codepoint <= 0xdfff)) {
+			continue;
+		}
+		if (codepoint <= 0xffff) {
+			result.push_back(static_cast<char16_t>(codepoint));
+		} else {
+			codepoint -= 0x10000;
+			result.push_back(static_cast<char16_t>(0xd800 + (codepoint >> 10)));
+			result.push_back(static_cast<char16_t>(0xdc00 + (codepoint & 0x3ff)));
+		}
+	}
+	return result;
+}
+
+std::string VisibleText(const Ime::HostSnapshot& snapshot, size_t max_units) {
+	const size_t cursor        = std::min<size_t>(snapshot.cursor, snapshot.text.size());
+	const size_t payload_units = max_units > 4 ? max_units - 4 : 1;
+	size_t       begin         = cursor > payload_units / 2 ? cursor - payload_units / 2 : 0;
+	size_t       end           = std::min(snapshot.text.size(), begin + payload_units);
+	if (end == snapshot.text.size() && end - begin < payload_units) {
+		begin = end > payload_units ? end - payload_units : 0;
+	}
+	if (begin > 0 && snapshot.text[begin] >= 0xdc00 && snapshot.text[begin] <= 0xdfff) {
+		begin--;
+	}
+	if (end < snapshot.text.size() && end > begin && snapshot.text[end - 1] >= 0xd800 &&
+	    snapshot.text[end - 1] <= 0xdbff) {
+		end--;
+	}
+
+	std::u16string visible;
+	if (begin != 0) {
+		visible.push_back(u'\u2026');
+	}
+	const size_t caret = visible.size() + cursor - begin;
+	if ((snapshot.option & Ime::OPTION_PASSWORD) != 0) {
+		visible.append(end - begin, u'*');
+	} else {
+		visible.append(snapshot.text, begin, end - begin);
+		std::replace(visible.begin(), visible.end(), u'\n', u'\u21b5');
+		std::replace(visible.begin(), visible.end(), u'\r', u'\u21b5');
+	}
+	visible.insert(std::min(caret, visible.size()), 1, u'|');
+	if (end != snapshot.text.size()) {
+		visible.push_back(u'\u2026');
+	}
+	return Common::Utf16ToUtf8(visible.c_str());
+}
+
+const char* EnterLabel(Ime::EnterLabel label) {
+	switch (label) {
+		case Ime::EnterLabel::Send: return "Send";
+		case Ime::EnterLabel::Search: return "Search";
+		case Ime::EnterLabel::Go: return "Go";
+		default: return "Done";
+	}
+}
+
+float AlignmentPivot(Ime::Alignment alignment) {
+	switch (alignment) {
+		case Ime::Alignment::Start: return 0.0f;
+		case Ime::Alignment::End: return 1.0f;
+		default: return 0.5f;
+	}
+}
+
+ImGuiKey ControllerButtonToKey(int button) {
+	switch (button) {
+		case SDL_CONTROLLER_BUTTON_A: return ImGuiKey_GamepadFaceDown;
+		case SDL_CONTROLLER_BUTTON_B: return ImGuiKey_GamepadFaceRight;
+		case SDL_CONTROLLER_BUTTON_X: return ImGuiKey_GamepadFaceLeft;
+		case SDL_CONTROLLER_BUTTON_Y: return ImGuiKey_GamepadFaceUp;
+		case SDL_CONTROLLER_BUTTON_DPAD_LEFT: return ImGuiKey_GamepadDpadLeft;
+		case SDL_CONTROLLER_BUTTON_DPAD_RIGHT: return ImGuiKey_GamepadDpadRight;
+		case SDL_CONTROLLER_BUTTON_DPAD_UP: return ImGuiKey_GamepadDpadUp;
+		case SDL_CONTROLLER_BUTTON_DPAD_DOWN: return ImGuiKey_GamepadDpadDown;
+		default: return ImGuiKey_None;
+	}
+}
+
+PFN_vkVoidFunction LoadVulkanFunction(const char* name, void* user_data) {
+	auto& graphics = *static_cast<GraphicContext*>(user_data);
+	return graphics.instance.getProcAddr(name);
+}
+
+void CheckVulkanResult(VkResult result) {
+	EXIT_IF(result != VK_SUCCESS);
+}
+
+} // namespace
+
+void InitializeImeDialogInput() {
+	const Uint32 type = SDL_RegisterEvents(1);
+	EXIT_IF(type == static_cast<Uint32>(-1));
+	{
+		std::scoped_lock lock(g_visibility_mutex);
+		g_input_lifecycle_active = true;
+		g_visibility_event.store(type, std::memory_order_release);
+	}
+	Ime::SetVisibilityCallback(OnVisibilityChanged);
+	Ime::HostSnapshot snapshot;
+	if (Ime::GetHostSnapshot(&snapshot)) {
+		OnVisibilityChanged(true, snapshot.generation);
+	}
+}
+
+void ShutdownImeDialogInput() {
+	Ime::SetVisibilityCallback(nullptr);
+	{
+		std::scoped_lock lock(g_visibility_mutex);
+		g_input_lifecycle_active = false;
+		g_active                 = false;
+		if (std::exchange(g_controller_captured, false)) {
+			Controller::ControllerResetInputState();
+		}
+		g_visibility_event.store(static_cast<Uint32>(-1), std::memory_order_release);
+	}
+	ClearInputEvents();
+	{
+		std::scoped_lock lock(g_input_mutex);
+		g_visibility_updates.clear();
+		g_missing_visibility_wakeups = 0;
+	}
+	g_input_active     = false;
+	g_input_controller = false;
+	g_input_keyboard   = false;
+	g_input_multiline  = false;
+	if (SDL_IsTextInputActive() == SDL_TRUE) {
+		SDL_StopTextInput();
+	}
+}
+
+bool ProcessImeDialogInput(const SDL_Event& event) {
+	RetryVisibilityWakeup();
+	if (event.type == g_visibility_event.load(std::memory_order_acquire)) {
+		VisibilityUpdate update {};
+		{
+			std::scoped_lock lock(g_input_mutex);
+			if (g_visibility_updates.empty()) {
+				return true;
+			}
+			update = g_visibility_updates.front();
+			g_visibility_updates.pop_front();
+		}
+		ClearInputEvents();
+		g_last_external_keycode = 0;
+		g_last_external_status  = 0;
+		g_input_generation      = update.generation;
+		g_input_active          = update.visible;
+		g_input_controller      = update.capture_controller;
+		g_input_keyboard        = update.capture_keyboard;
+		g_input_multiline       = update.multiline;
+		if (g_input_keyboard) {
+			SDL_StartTextInput();
+		} else if (SDL_IsTextInputActive() == SDL_TRUE) {
+			SDL_StopTextInput();
+		}
+		return true;
+	}
+	if (event.type == SDL_CONTROLLERDEVICEREMOVED) {
+		if (g_input_active && g_input_controller) {
+			QueueInput({InputKind::ResetController, g_input_generation, 0, 0.0f, 0.0f});
+		}
+		return false;
+	}
+	if (!g_input_active) {
+		return false;
+	}
+
+	const uint64_t generation       = g_input_generation;
+	const bool     controller_event = event.type == SDL_CONTROLLERBUTTONDOWN ||
+	                                  event.type == SDL_CONTROLLERBUTTONUP ||
+	                                  event.type == SDL_CONTROLLERAXISMOTION;
+	if (controller_event && !g_input_controller) {
+		return false;
+	}
+	const bool keyboard_event = event.type == SDL_TEXTINPUT || event.type == SDL_TEXTEDITING ||
+	                            event.type == SDL_KEYDOWN || event.type == SDL_KEYUP;
+	if (keyboard_event && !g_input_keyboard) {
+		return false;
+	}
+	switch (event.type) {
+		case SDL_TEXTINPUT: {
+			const auto text = Utf8ToUtf16(event.text.text);
+			if (!text.empty()) {
+				auto input = MakeExternalInput(Ime::ExternalAction::Text, g_last_external_keycode,
+				                               g_last_external_status | 0x00000002);
+				input.key.character = text.front();
+				input.text          = text;
+				Ime::HostQueueExternalInput(generation, std::move(input));
+			}
+			return true;
+		}
+		case SDL_TEXTEDITING: return true;
+		case SDL_KEYDOWN: {
+			g_last_external_keycode = static_cast<uint16_t>(event.key.keysym.scancode);
+			g_last_external_status =
+			    ExternalKeyStatus(static_cast<SDL_Keymod>(event.key.keysym.mod), false);
+			auto action = Ime::ExternalAction::Text;
+			bool queue  = true;
+			if (event.key.keysym.sym == SDLK_BACKSPACE) {
+				action = Ime::ExternalAction::Backspace;
+			} else if (event.key.keysym.sym == SDLK_LEFT) {
+				action = Ime::ExternalAction::MoveLeft;
+			} else if (event.key.keysym.sym == SDLK_RIGHT) {
+				action = Ime::ExternalAction::MoveRight;
+			} else if (event.key.keysym.sym == SDLK_ESCAPE) {
+				action = Ime::ExternalAction::Cancel;
+			} else if (event.key.keysym.sym == SDLK_RETURN ||
+			           event.key.keysym.sym == SDLK_KP_ENTER) {
+				action =
+				    g_input_multiline ? Ime::ExternalAction::Newline : Ime::ExternalAction::Accept;
+			} else if (event.key.keysym.sym == SDLK_TAB) {
+				action = Ime::ExternalAction::None;
+			} else {
+				queue = false;
+			}
+			if (queue) {
+				Ime::HostQueueExternalInput(
+				    generation,
+				    MakeExternalInput(action, g_last_external_keycode, g_last_external_status));
+			}
+			return true;
+		}
+		case SDL_KEYUP: return true;
+		case SDL_CONTROLLERBUTTONDOWN:
+		case SDL_CONTROLLERBUTTONUP:
+			QueueInput({InputKind::Button, generation, event.cbutton.button,
+			            event.type == SDL_CONTROLLERBUTTONDOWN ? 1.0f : 0.0f, 0.0f});
+			return true;
+		case SDL_CONTROLLERAXISMOTION:
+			QueueInput({InputKind::Axis, generation, event.caxis.axis,
+			            static_cast<float>(event.caxis.value), 0.0f});
+			return true;
+		case SDL_MOUSEMOTION:
+			QueueInput({InputKind::MousePosition, generation, 0, static_cast<float>(event.motion.x),
+			            static_cast<float>(event.motion.y)});
+			return true;
+		case SDL_MOUSEBUTTONDOWN:
+		case SDL_MOUSEBUTTONUP:
+			QueueInput({InputKind::MouseButton, generation, event.button.button,
+			            event.type == SDL_MOUSEBUTTONDOWN ? 1.0f : 0.0f, 0.0f});
+			return true;
+		case SDL_MOUSEWHEEL:
+			QueueInput({InputKind::MouseWheel, generation, 0, static_cast<float>(event.wheel.x),
+			            static_cast<float>(event.wheel.y)});
+			return true;
+		default: return false;
+	}
+}
+
+struct ImeDialogOverlay::Impl {
+	explicit Impl(GraphicContext& context): graphics(context) {}
+
+	~Impl() {
+		ReleaseVulkan();
+		if (imgui_context != nullptr) {
+			ImGui::DestroyContext(imgui_context);
+		}
+	}
+
+	void EnsureContext() {
+		if (imgui_context != nullptr) {
+			ImGui::SetCurrentContext(imgui_context);
+			return;
+		}
+		IMGUI_CHECKVERSION();
+		imgui_context = ImGui::CreateContext();
+		ImGui::SetCurrentContext(imgui_context);
+		auto& io       = ImGui::GetIO();
+		io.IniFilename = nullptr;
+		io.LogFilename = nullptr;
+		io.ConfigFlags |= ImGuiConfigFlags_NavEnableGamepad;
+		io.BackendFlags |= ImGuiBackendFlags_HasGamepad;
+		io.BackendPlatformName = "Kyty ImeDialog input";
+		ImGui::StyleColorsDark();
+		auto& style          = ImGui::GetStyle();
+		style.WindowRounding = 10.0f;
+		style.FrameRounding  = 6.0f;
+		style.ItemSpacing    = {8.0f, 8.0f};
+	}
+
+	void EnsureVulkan(vk::Format format, uint32_t image_count) {
+		EnsureContext();
+		if (vulkan_initialized) {
+			return;
+		}
+		EXIT_IF(image_count < 2);
+		EXIT_IF(!ImGui_ImplVulkan_LoadFunctions(VK_API_VERSION_1_3, LoadVulkanFunction, &graphics));
+
+		const VkFormat            color_format = static_cast<VkFormat>(format);
+		ImGui_ImplVulkan_InitInfo info {};
+		info.ApiVersion                   = VK_API_VERSION_1_3;
+		info.Instance                     = static_cast<VkInstance>(graphics.instance);
+		info.PhysicalDevice               = static_cast<VkPhysicalDevice>(graphics.physical_device);
+		info.Device                       = static_cast<VkDevice>(graphics.device);
+		info.QueueFamily                  = graphics.queue_family;
+		info.Queue                        = static_cast<VkQueue>(graphics.queue);
+		info.DescriptorPoolSize           = IMGUI_IMPL_VULKAN_MINIMUM_SAMPLED_IMAGE_POOL_SIZE;
+		info.MinImageCount                = image_count;
+		info.ImageCount                   = image_count;
+		info.UseDynamicRendering          = true;
+		info.PipelineInfoMain.MSAASamples = VK_SAMPLE_COUNT_1_BIT;
+		info.PipelineInfoMain.PipelineRenderingCreateInfo.sType =
+		    VK_STRUCTURE_TYPE_PIPELINE_RENDERING_CREATE_INFO;
+		info.PipelineInfoMain.PipelineRenderingCreateInfo.colorAttachmentCount    = 1;
+		info.PipelineInfoMain.PipelineRenderingCreateInfo.pColorAttachmentFormats = &color_format;
+		info.CheckVkResultFn = CheckVulkanResult;
+		EXIT_IF(!ImGui_ImplVulkan_Init(&info));
+		vulkan_initialized = true;
+	}
+
+	void DrainInput(uint64_t generation) {
+		std::deque<InputEvent> events;
+		bool                   reset = false;
+		{
+			std::scoped_lock lock(g_input_mutex);
+			events.swap(g_input_events);
+			reset                   = g_input_reset_requested;
+			g_input_reset_requested = false;
+		}
+		auto& io = ImGui::GetIO();
+		if (reset) {
+			io.ClearEventsQueue();
+			io.ClearInputKeys();
+			io.ClearInputMouse();
+			right_stick = {};
+		}
+		for (const auto& event: events) {
+			if (event.generation != generation) {
+				continue;
+			}
+			switch (event.kind) {
+				case InputKind::Button: {
+					const bool down = event.x != 0.0f;
+					if (down && event.id == SDL_CONTROLLER_BUTTON_B) {
+						Ime::HostCancel(generation);
+					} else if (down && event.id == SDL_CONTROLLER_BUTTON_Y) {
+						Ime::HostBackspace(generation);
+					}
+					const ImGuiKey key = ControllerButtonToKey(event.id);
+					if (key != ImGuiKey_None) {
+						io.AddKeyEvent(key, down);
+					}
+					break;
+				}
+				case InputKind::Axis: {
+					const float value = event.x / (event.x < 0.0f ? 32768.0f : 32767.0f);
+					if (event.id == SDL_CONTROLLER_AXIS_LEFTX) {
+						io.AddKeyAnalogEvent(ImGuiKey_GamepadLStickLeft, value < -0.25f,
+						                     std::max(-value, 0.0f));
+						io.AddKeyAnalogEvent(ImGuiKey_GamepadLStickRight, value > 0.25f,
+						                     std::max(value, 0.0f));
+					} else if (event.id == SDL_CONTROLLER_AXIS_LEFTY) {
+						io.AddKeyAnalogEvent(ImGuiKey_GamepadLStickUp, value < -0.25f,
+						                     std::max(-value, 0.0f));
+						io.AddKeyAnalogEvent(ImGuiKey_GamepadLStickDown, value > 0.25f,
+						                     std::max(value, 0.0f));
+					} else if (event.id == SDL_CONTROLLER_AXIS_RIGHTX) {
+						right_stick.x = std::abs(value) > 0.2f ? value : 0.0f;
+					} else if (event.id == SDL_CONTROLLER_AXIS_RIGHTY) {
+						right_stick.y = std::abs(value) > 0.2f ? value : 0.0f;
+					}
+					break;
+				}
+				case InputKind::MousePosition: io.AddMousePosEvent(event.x, event.y); break;
+				case InputKind::MouseButton: {
+					int button = -1;
+					if (event.id == SDL_BUTTON_LEFT) button = 0;
+					if (event.id == SDL_BUTTON_RIGHT) button = 1;
+					if (event.id == SDL_BUTTON_MIDDLE) button = 2;
+					if (button >= 0) io.AddMouseButtonEvent(button, event.x != 0.0f);
+					break;
+				}
+				case InputKind::MouseWheel: io.AddMouseWheelEvent(event.x, event.y); break;
+				case InputKind::ResetController:
+					io.ClearEventsQueue();
+					io.ClearInputKeys();
+					right_stick = {};
+					break;
+			}
+		}
+	}
+
+	void KeyButton(std::string_view label, char16_t value, uint64_t generation, float width,
+	               bool default_focus) {
+		const bool pressed = ImGui::Button(label.data(), {width, button_height});
+		if (default_focus) {
+			ImGui::SetItemDefaultFocus();
+		}
+		if (pressed) {
+			Ime::HostInsertText(generation, std::u16string_view(&value, 1));
+		}
+	}
+
+	void DrawKeyRows(const Ime::HostSnapshot& snapshot, float width) {
+		static constexpr std::array<std::string_view, 4> lower   = {"1234567890", "qwertyuiop",
+		                                                            "asdfghjkl", "zxcvbnm"};
+		static constexpr std::array<std::string_view, 4> upper   = {"1234567890", "QWERTYUIOP",
+		                                                            "ASDFGHJKL", "ZXCVBNM"};
+		static constexpr std::array<std::string_view, 3> symbols = {"1234567890", "!@#$%^&*()",
+		                                                            "-_=+/:?."};
+		static constexpr std::array<std::string_view, 4> number  = {"123", "456", "789", "-0."};
+
+		std::span<const std::string_view> rows;
+		if (snapshot.type == Ime::Type::Number) {
+			rows = number;
+		} else if (symbol_mode) {
+			rows = symbols;
+		} else {
+			rows = shift ? std::span<const std::string_view>(upper)
+			             : std::span<const std::string_view>(lower);
+		}
+		bool   first     = focus_pending;
+		size_t row_index = 0;
+		for (const auto row: rows) {
+			const float key_width =
+			    std::min(56.0f * ui_scale, (width - 8.0f * (row.size() - 1)) / row.size());
+			const float row_width = key_width * row.size() + 8.0f * (row.size() - 1);
+			ImGui::SetCursorPosX((width - row_width) * 0.5f);
+			ImGui::PushID(static_cast<int>(row_index++));
+			for (size_t i = 0; i < row.size(); i++) {
+				ImGui::PushID(static_cast<int>(i));
+				const char label[2] = {row[i], '\0'};
+				KeyButton(label, static_cast<char16_t>(row[i]), snapshot.generation, key_width,
+				          first);
+				first = false;
+				ImGui::PopID();
+				if (i + 1 < row.size()) ImGui::SameLine();
+			}
+			ImGui::PopID();
+		}
+		focus_pending = false;
+	}
+
+	void DrawDialog(const Ime::HostSnapshot& snapshot, vk::Extent2D extent) {
+		const ImVec2 display(static_cast<float>(extent.width), static_cast<float>(extent.height));
+		const bool   over_2k          = (snapshot.option & Ime::OPTION_USE_OVER_2K) != 0;
+		const float  reference_width  = over_2k ? 3840.0f : 1920.0f;
+		const float  reference_height = over_2k ? 2160.0f : 1080.0f;
+		const float  scale_x          = display.x / reference_width;
+		const float  scale_y          = display.y / reference_height;
+		const float  width            = static_cast<float>(snapshot.panel_width) * scale_x;
+		const float  height           = static_cast<float>(snapshot.panel_height) * scale_y;
+		ui_scale                      = std::min(scale_x, scale_y) * (over_2k ? 2.0f : 1.0f);
+		button_height                 = std::max(28.0f, 42.0f * ui_scale);
+
+		const ImVec2 pivot {AlignmentPivot(snapshot.horizontal_alignment),
+		                    AlignmentPivot(snapshot.vertical_alignment)};
+		if ((snapshot.option & Ime::OPTION_FIXED_POSITION) == 0) {
+			const float movement = 600.0f * ImGui::GetIO().DeltaTime;
+			panel_offset.x += right_stick.x * movement;
+			panel_offset.y += right_stick.y * movement;
+		}
+		const ImVec2 base_position {snapshot.posx * scale_x - pivot.x * width,
+		                            snapshot.posy * scale_y - pivot.y * height};
+		ImVec2       position {base_position.x + panel_offset.x, base_position.y + panel_offset.y};
+		if ((snapshot.option & Ime::OPTION_DISABLE_POSITION_ADJ) == 0) {
+			position.x   = std::clamp(position.x, 0.0f, std::max(display.x - width, 0.0f));
+			position.y   = std::clamp(position.y, 0.0f, std::max(display.y - height, 0.0f));
+			panel_offset = {position.x - base_position.x, position.y - base_position.y};
+		}
+		ImGui::SetNextWindowPos(position, ImGuiCond_Always);
+		ImGui::SetNextWindowSize({width, height}, ImGuiCond_Always);
+		constexpr ImGuiWindowFlags flags = ImGuiWindowFlags_NoDecoration | ImGuiWindowFlags_NoMove |
+		                                   ImGuiWindowFlags_NoSavedSettings;
+		ImGui::Begin("##ImeDialog", nullptr, flags);
+
+		const std::u16string title = snapshot.title.empty() ? u"Enter text" : snapshot.title;
+		if (!snapshot.key_panel_visible) {
+			std::string compact     = Common::Utf16ToUtf8(title.c_str()) + ": ";
+			const float glyph_width = std::max(ImGui::CalcTextSize("M").x, 1.0f);
+			const float text_width =
+			    std::max(ImGui::GetContentRegionAvail().x - ImGui::CalcTextSize(compact.c_str()).x,
+			             glyph_width);
+			const size_t visible_units =
+			    std::max<size_t>(6, static_cast<size_t>(text_width / glyph_width));
+			compact += snapshot.text.empty() && !snapshot.placeholder.empty()
+			               ? Common::Utf16ToUtf8(snapshot.placeholder.c_str())
+			               : VisibleText(snapshot, visible_units);
+			ImGui::TextUnformatted(compact.c_str());
+			const float compact_height = std::min(button_height, 28.0f * ui_scale);
+			if (ImGui::Button("Cancel", {90.0f * ui_scale, compact_height})) {
+				Ime::HostCancel(snapshot.generation);
+			}
+			ImGui::SameLine();
+			if (ImGui::Button(EnterLabel(snapshot.enter_label),
+			                  {90.0f * ui_scale, compact_height})) {
+				Ime::HostAccept(snapshot.generation);
+			}
+			ImGui::End();
+			return;
+		}
+
+		ImGui::TextUnformatted(Common::Utf16ToUtf8(title.c_str()).c_str());
+		ImGui::Separator();
+		const float text_height = std::max(
+		    32.0f, ((snapshot.option & Ime::OPTION_MULTILINE) != 0 ? 80.0f : 48.0f) * ui_scale);
+		ImGui::BeginChild("##ImeText", {0.0f, text_height}, true);
+		if (snapshot.text.empty() && !snapshot.placeholder.empty()) {
+			ImGui::TextDisabled("%s", Common::Utf16ToUtf8(snapshot.placeholder.c_str()).c_str());
+		} else {
+			const float  glyph_width = std::max(ImGui::CalcTextSize("M").x, 1.0f);
+			const size_t columns     = std::max<size_t>(
+			    8, static_cast<size_t>(ImGui::GetContentRegionAvail().x / glyph_width));
+			const size_t lines =
+			    std::max<size_t>(1, static_cast<size_t>(ImGui::GetContentRegionAvail().y /
+			                                            ImGui::GetTextLineHeightWithSpacing()));
+			const auto text = VisibleText(snapshot, columns * lines);
+			ImGui::TextWrapped("%s", text.c_str());
+		}
+		ImGui::EndChild();
+		ImGui::TextDisabled("%zu / %u", snapshot.text.size(), snapshot.max_text_length);
+		ImGui::Separator();
+
+		const float content_width = ImGui::GetContentRegionAvail().x;
+		DrawKeyRows(snapshot, content_width);
+		if (snapshot.type != Ime::Type::Number) {
+			if (ImGui::Button(shift ? "Lower" : "Shift", {84.0f * ui_scale, button_height})) {
+				shift = !shift;
+			}
+			ImGui::SameLine();
+			if (ImGui::Button(symbol_mode ? "ABC" : "Symbols", {84.0f * ui_scale, button_height})) {
+				symbol_mode = !symbol_mode;
+			}
+			ImGui::SameLine();
+			if (ImGui::Button("Space", {120.0f * ui_scale, button_height})) {
+				Ime::HostInsertText(snapshot.generation, u" ");
+			}
+			ImGui::SameLine();
+		}
+		const float action_width = snapshot.type == Ime::Type::Number
+		                               ? std::max((content_width - 16.0f) / 3.0f, 48.0f)
+		                               : 100.0f * ui_scale;
+		if (ImGui::Button("Backspace", {action_width, button_height})) {
+			Ime::HostBackspace(snapshot.generation);
+		}
+		ImGui::SameLine();
+		if ((snapshot.option & Ime::OPTION_MULTILINE) != 0) {
+			if (ImGui::Button("Newline", {action_width, button_height})) {
+				Ime::HostInsertText(snapshot.generation, u"\n");
+			}
+			ImGui::SameLine();
+		}
+		if (ImGui::Button("Cancel", {action_width, button_height})) {
+			Ime::HostCancel(snapshot.generation);
+		}
+		ImGui::SameLine();
+		if (ImGui::Button(EnterLabel(snapshot.enter_label), {action_width, button_height})) {
+			Ime::HostAccept(snapshot.generation);
+		}
+		ImGui::End();
+	}
+
+	bool PrepareFrame(vk::Extent2D frame_extent, vk::Format format, uint32_t image_count) {
+		Ime::HostSnapshot snapshot;
+		if (!Ime::GetHostSnapshot(&snapshot)) {
+			return false;
+		}
+		const uint64_t prepared_generation = snapshot.generation;
+		EnsureVulkan(format, image_count);
+		if (generation != snapshot.generation) {
+			generation    = snapshot.generation;
+			focus_pending = true;
+			shift         = (snapshot.option & Ime::OPTION_NO_AUTO_CAPITALIZE) == 0;
+			symbol_mode   = false;
+			panel_offset  = {};
+			right_stick   = {};
+			auto& io      = ImGui::GetIO();
+			io.ClearEventsQueue();
+			io.ClearInputKeys();
+			io.ClearInputMouse();
+		}
+		DrainInput(snapshot.generation);
+
+		auto& io       = ImGui::GetIO();
+		io.DisplaySize = {static_cast<float>(frame_extent.width),
+		                  static_cast<float>(frame_extent.height)};
+		const auto now = std::chrono::steady_clock::now();
+		io.DeltaTime   = last_frame == std::chrono::steady_clock::time_point {}
+		                     ? 1.0f / 60.0f
+		                     : std::clamp(std::chrono::duration<float>(now - last_frame).count(),
+		                                  1.0f / 1000.0f, 0.1f);
+		last_frame     = now;
+		ImGui_ImplVulkan_NewFrame();
+		ImGui::NewFrame();
+		if (!Ime::GetHostSnapshot(&snapshot) || snapshot.generation != prepared_generation) {
+			ImGui::EndFrame();
+			return false;
+		}
+		DrawDialog(snapshot, frame_extent);
+		ImGui::Render();
+		extent = frame_extent;
+		return true;
+	}
+
+	void Record(vk::CommandBuffer command, vk::ImageView target) {
+		vk::RenderingAttachmentInfo color {};
+		color.sType       = vk::StructureType::eRenderingAttachmentInfo;
+		color.imageView   = target;
+		color.imageLayout = vk::ImageLayout::eColorAttachmentOptimal;
+		color.loadOp      = vk::AttachmentLoadOp::eLoad;
+		color.storeOp     = vk::AttachmentStoreOp::eStore;
+		vk::RenderingInfo rendering {};
+		rendering.sType                = vk::StructureType::eRenderingInfo;
+		rendering.renderArea.extent    = extent;
+		rendering.layerCount           = 1;
+		rendering.colorAttachmentCount = 1;
+		rendering.pColorAttachments    = &color;
+		command.beginRendering(rendering);
+		{
+			Common::LockGuard queue_lock(graphics.queue_mutex);
+			ImGui_ImplVulkan_RenderDrawData(ImGui::GetDrawData(),
+			                                static_cast<VkCommandBuffer>(command));
+		}
+		command.endRendering();
+	}
+
+	void ReleaseVulkan() {
+		if (!vulkan_initialized) {
+			return;
+		}
+		ImGui::SetCurrentContext(imgui_context);
+		ImGui_ImplVulkan_Shutdown();
+		vulkan_initialized = false;
+	}
+
+	GraphicContext&                       graphics;
+	ImGuiContext*                         imgui_context      = nullptr;
+	bool                                  vulkan_initialized = false;
+	bool                                  shift              = false;
+	bool                                  symbol_mode        = false;
+	bool                                  focus_pending      = true;
+	float                                 ui_scale           = 1.0f;
+	float                                 button_height      = 42.0f;
+	ImVec2                                panel_offset {};
+	ImVec2                                right_stick {};
+	uint64_t                              generation = 0;
+	vk::Extent2D                          extent {};
+	std::chrono::steady_clock::time_point last_frame;
+};
+
+ImeDialogOverlay::ImeDialogOverlay(GraphicContext& graphics)
+    : m_impl(std::make_unique<Impl>(graphics)) {}
+
+ImeDialogOverlay::~ImeDialogOverlay() = default;
+
+bool ImeDialogOverlay::PrepareFrame(vk::Extent2D extent, vk::Format format, uint32_t image_count) {
+	return m_impl->PrepareFrame(extent, format, image_count);
+}
+
+void ImeDialogOverlay::Record(vk::CommandBuffer command, vk::ImageView target) {
+	m_impl->Record(command, target);
+}
+
+void ImeDialogOverlay::ReleaseVulkan() {
+	m_impl->ReleaseVulkan();
+}
+
+} // namespace Libs::Graphics