tests: add LeastRecentlyUsedCache regression tests (#671 )

* launcher: keep primary update feed answer when the fallback request fails When the primary feed is reachable but its tag differs from the build tag, the result was discarded and the GitHub API fallback was consulted. If the fallback request failed (api.github.com rate-limits unauthenticated requests), the launcher reported an error even though the primary feed had already returned a usable update answer with a working release URL. Keep the primary feed answer and show it if the fallback fails. * emulator: validate numeric CLI arguments --screen-width, --screen-height, --gpu, and --vblank-frequency were parsed with Common::ToInt32/ToInt64, which ignores std::from_chars errors and returns 0, so values like "--screen-width abc" or "--gpu xyz" were silently accepted: zero screen size, or GPU 0 selected instead of auto (-1). Reject invalid values with an error message like the other options, using from_chars-based helpers (ParseUint32/ParseInt32); reject zero screen dimensions. * tests: add LeastRecentlyUsedCache regression tests Cover the LRU cache walk order, tick filtering in ForEachItemBelow, early-exit boolean callbacks, Touch reordering and same-tick skips, Free, and freed-id reuse by later inserts. Modeled on BitArrayTests. * tests: use a tick cutoff that detects a stale newest-item tick Collect(cache, 35) also visits c when Touch(c, 35) failed to update its tick (30 <= 35). Use a cutoff of 30 and expect only b. * Run LRU cache tests in CI --------- Co-authored-by: nmzik <Nmzik@mail.ru>
Build save_data_memory_tests from the kyty_tests aggregate (#798 )
2026-10-27 16:02:37 +00:00 · 2026-09-26 02:42:31 +04:00 · 2026-09-26 02:39:20 +04:00 · 2026-09-26 02:19:49 +04:00 · 2026-09-26 01:58:55 +04:00 · 2026-09-26 01:50:06 +04:00
211 changed files with 16810 additions and 9151 deletions
@@ -0,0 +1,37 @@
+# yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json
+
+# Keep inherited review profiles, automatic reviews, tools, and pre-merge checks.
+inheritance: true
+
+reviews:
+  # Trim supplemental output and keep the remaining walkthrough collapsed.
+  high_level_summary: false
+  review_status: false
+  review_details: false
+  collapse_walkthrough: true
+  changed_files_summary: false
+  sequence_diagrams: false
+  estimate_code_review_effort: false
+  related_issues: false
+  related_prs: false
+  in_progress_fortune: false
+  poem: false
+  enable_prompt_for_ai_agents: false
+
+  # Hide standard finishing-touch controls; this also disables their commands.
+  finishing_touches:
+    docstrings:
+      enabled: false
+    unit_tests:
+      enabled: false
+    simplify:
+      enabled: false
+    autofix:
+      enabled: false
+    fix_ci:
+      enabled: false
+    resolve_merge_conflict:
+      enabled: false
+
+chat:
+  art: false
@@ -43,9 +43,6 @@ jobs:
      - name: Setup Visual Studio environment
        uses: ilammy/msvc-dev-cmd@v1

-      - name: Setup Ninja
-        uses: seanmiddleditch/gha-setup-ninja@v5
-
      - name: Locate vcpkg
        id: vcpkg
        shell: pwsh
@@ -115,12 +112,12 @@ jobs:
      - name: Build
        shell: cmd
        run: |
-          cmake --build _Build/windows --target launcher audio_out2_port_tests ime_dialog_tests virtual_memory_allocation_tests --parallel
+          cmake --build _Build/windows --target launcher audio_out2_port_tests ime_dialog_tests lru_cache_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)$"
+          ctest --test-dir _Build/windows --output-on-failure -R "^(audio_out2_port|ime_dialog|lru_cache|virtual_memory_allocation)$"

      - name: Install
        shell: cmd
@@ -155,7 +152,7 @@ jobs:

      - name: Install build dependencies
        shell: bash
-        run: brew install glslang ninja
+        run: brew install glslang

      - name: Install Qt 6.10.3
        uses: jurplel/install-qt-action@v4
@@ -201,14 +198,14 @@ jobs:
        shell: bash
        run: |
          cmake --build _Build/macos \
-            --target launcher audio_out2_port_tests ime_dialog_tests virtual_memory_allocation_tests \
+            --target launcher audio_out2_port_tests ime_dialog_tests lru_cache_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)$'
+            -R '^(audio_out2_port|ime_dialog|lru_cache|virtual_memory_allocation)$'

      - name: Install
        shell: bash
@@ -317,8 +314,8 @@ jobs:
            libxkbcommon-dev \
            libxrandr-dev \
            libxss-dev \
+            libxtst-dev \
            lld \
-            ninja-build \
            pkg-config \
            wayland-protocols

@@ -363,7 +360,7 @@ jobs:
            -DCMAKE_PREFIX_PATH="$Qt6_DIR" 2>&1 | tee _Build/configure.log
          exit "${PIPESTATUS[0]}"

-      - name: Verify SDL2 backends
+      - name: Verify SDL3 backends
        shell: bash
        run: |
          status=0
@@ -382,14 +379,14 @@ 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 \
+              audio_out2_port_tests ime_dialog_tests lru_cache_tests virtual_memory_allocation_tests \
            --parallel

      - name: Test
        shell: bash
        run: |
          ctest --test-dir _Build/linux --output-on-failure \
-            -R '^(audio_out2_port|ime_dialog|page_manager|memory_tracker|virtual_memory_allocation)$'
+            -R '^(audio_out2_port|ime_dialog|lru_cache|page_manager|memory_tracker|virtual_memory_allocation)$'

      - name: Install
        shell: bash
@@ -1,7 +1,7 @@
-[submodule "3rdparty/SDL2"]
-	path = 3rdparty/SDL2
+[submodule "3rdparty/SDL3"]
+	path = 3rdparty/SDL3
 	url = https://github.com/libsdl-org/SDL.git
-	branch = SDL2
+	branch = release-3.4.x
 [submodule "3rdparty/magic_enum"]
 	path = 3rdparty/magic_enum
 	url = https://github.com/Neargye/magic_enum.git
@@ -2,15 +2,17 @@
    "cmake.sourceDirectory": "${workspaceFolder}",
    "cmake.buildDirectory": "${workspaceFolder}/_Build/vscode-clang",
    "cmake.generator": "Ninja",
+    "cmake.useVsDeveloperEnvironment": "always",
    "cmake.environment": {
        "CC": "clang-cl",
-        "CXX": "clang-cl",
-        "PATH": "${env:ProgramFiles}/Microsoft Visual Studio/18/Community/VC/Tools/Llvm/x64/bin;${env:ProgramFiles}/Microsoft Visual Studio/18/Community/Common7/IDE/CommonExtensions/Microsoft/CMake/Ninja;${env:PATH}"
+        "CXX": "clang-cl"
    },
    "cmake.configureSettings": {
        "CMAKE_C_COMPILER": "clang-cl",
        "CMAKE_CXX_COMPILER": "clang-cl",
        "CMAKE_PREFIX_PATH": "C:/Qt/6.11.0/msvc2022_64"
-    }
+    },
+    "C_Cpp.default.intelliSenseMode": "windows-clang-x64",
+    "C_Cpp.default.compileCommands": "${workspaceFolder}/_Build/vscode-clang/compile_commands.json",
+    "C_Cpp.default.cppStandard": "c++20"
 }
-
@@ -29,7 +29,7 @@ FetchContent_Declare(
 FetchContent_MakeAvailable(xbyak zydis)

 set(SDL_SHARED OFF CACHE BOOL "" FORCE)
-add_subdirectory(SDL2)
+add_subdirectory(SDL3)

 add_library(Vulkan-Headers INTERFACE)
 add_library(Vulkan::Headers ALIAS Vulkan-Headers)
@@ -86,6 +86,9 @@ set(SPDLOG_BUILD_WARNINGS OFF CACHE BOOL "" FORCE)
 set(SPDLOG_INSTALL OFF CACHE BOOL "" FORCE)
 set(SPDLOG_FMT_EXTERNAL ON CACHE BOOL "" FORCE)
 set(SPDLOG_NO_EXCEPTIONS ON CACHE BOOL "" FORCE)
+if(WIN32)
+	set(SPDLOG_WCHAR_FILENAMES ON CACHE BOOL "" FORCE)
+endif()
 add_subdirectory(spdlog EXCLUDE_FROM_ALL)

 if (NOT TARGET FFmpeg::ffmpeg)
@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.12...4.3)
+cmake_minimum_required(VERSION 3.16...4.3)

 if(CMAKE_VERSION VERSION_GREATER_EQUAL 4.0)
 	set(CMAKE_POLICY_VERSION_MINIMUM 3.5 CACHE STRING "Minimum policy version for bundled dependencies")
@@ -12,6 +12,22 @@ endif()

 project(Kyty)

+find_program(SCCACHE_PROGRAM sccache)
+find_program(CCACHE_PROGRAM ccache)
+
+foreach(lang IN ITEMS C CXX)
+    # An explicitly empty launcher disables caching for this language.
+    if(NOT DEFINED CMAKE_${lang}_COMPILER_LAUNCHER)
+        if(SCCACHE_PROGRAM)
+            set(CMAKE_${lang}_COMPILER_LAUNCHER "${SCCACHE_PROGRAM}")
+            message(STATUS "sccache enabled for ${lang}: ${SCCACHE_PROGRAM}")
+        elseif(CCACHE_PROGRAM)
+            set(CMAKE_${lang}_COMPILER_LAUNCHER "${CCACHE_PROGRAM}")
+            message(STATUS "ccache enabled for ${lang}: ${CCACHE_PROGRAM}")
+        endif()
+    endif()
+endforeach()
+
 set(KYTY_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/src")
 set(KYTY_TESTS_DIR "${CMAKE_CURRENT_SOURCE_DIR}/tests")
 set(KYTY_THIRD_PARTY_DIR "${CMAKE_CURRENT_SOURCE_DIR}/3rdparty")
@@ -315,7 +331,7 @@ 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 imgui xxhash FFmpeg::ffmpeg fmt::fmt nlohmann_json::nlohmann_json LibAtrac9 kyty_git_version Zydis::Zydis xbyak::xbyak)
+set(kyty_emulator_link_libraries common Vulkan::Headers spirv-tools-opt spirv-tools SDL3::SDL3-static imgui xxhash FFmpeg::ffmpeg fmt::fmt nlohmann_json::nlohmann_json LibAtrac9 kyty_git_version Zydis::Zydis xbyak::xbyak)

 # Linux system libraries.
 if(LINUX)
@@ -325,7 +341,6 @@ endif()

 set(inc_headers
 	${KYTY_SOURCE_DIR}
-	${KYTY_THIRD_PARTY_DIR}/SDL2/include
 	${KYTY_THIRD_PARTY_DIR}/VulkanMemoryAllocator/include
 	${KYTY_THIRD_PARTY_DIR}/SPIRV-Tools/include
 	${KYTY_THIRD_PARTY_DIR}/SPIRV-Headers/include
@@ -396,6 +411,11 @@ add_executable(bit_array_tests EXCLUDE_FROM_ALL
 )
 target_include_directories(bit_array_tests PRIVATE ${inc_headers})

+add_executable(lru_cache_tests EXCLUDE_FROM_ALL
+	"${KYTY_TESTS_DIR}/LruCacheTests.cpp"
+)
+target_include_directories(lru_cache_tests PRIVATE ${inc_headers})
+
 add_executable(ime_dialog_tests EXCLUDE_FROM_ALL
 	"${KYTY_TESTS_DIR}/ImeDialogTests.cpp"
 	"${KYTY_SOURCE_DIR}/libs/imeCommon.cpp"
@@ -450,6 +470,19 @@ add_executable(audio_out2_port_tests EXCLUDE_FROM_ALL
 target_link_libraries(audio_out2_port_tests common fmt::fmt)
 target_include_directories(audio_out2_port_tests PRIVATE ${inc_headers})

+add_executable(ngs2_sampler_tests EXCLUDE_FROM_ALL
+	"${KYTY_TESTS_DIR}/Ngs2SamplerTests.cpp"
+	"${KYTY_SOURCE_DIR}/loader/timer.cpp"
+)
+target_link_libraries(ngs2_sampler_tests common fmt::fmt LibAtrac9)
+target_include_directories(ngs2_sampler_tests PRIVATE ${inc_headers})
+
+add_executable(save_data_memory_tests EXCLUDE_FROM_ALL
+	"${KYTY_TESTS_DIR}/SaveDataMemoryTests.cpp"
+)
+target_link_libraries(save_data_memory_tests common fmt::fmt)
+target_include_directories(save_data_memory_tests PRIVATE ${inc_headers})
+
 add_executable(event_queue_lifetime_tests EXCLUDE_FROM_ALL
 	"${KYTY_TESTS_DIR}/EventQueueLifetimeTests.cpp"
 	"${KYTY_SOURCE_DIR}/kernel/eventQueue.cpp"
@@ -527,6 +560,7 @@ if(BUILD_TESTING)
 	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 bit_array COMMAND $<TARGET_FILE:bit_array_tests>)
+	add_test(NAME lru_cache COMMAND $<TARGET_FILE:lru_cache_tests>)
 	add_test(NAME shader_vertex_metadata COMMAND $<TARGET_FILE:shader_vertex_metadata_tests>)
 	add_test(NAME resource_materialization COMMAND $<TARGET_FILE:resource_materialization_tests>)
 	add_test(NAME resource_tracking COMMAND $<TARGET_FILE:resource_tracking_tests>)
@@ -534,6 +568,8 @@ if(BUILD_TESTING)
 	add_test(NAME kernel_file_system COMMAND $<TARGET_FILE:kernel_file_system_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 ngs2_sampler COMMAND $<TARGET_FILE:ngs2_sampler_tests>)
+	add_test(NAME save_data_memory COMMAND $<TARGET_FILE:save_data_memory_tests>)
 	add_test(NAME shader_recompiler_compute COMMAND $<TARGET_FILE:shader_recompiler_compute_tests>)
 	add_test(NAME shader_recompiler_alignbyte
 		COMMAND $<TARGET_FILE:shader_recompiler_compute_tests> --alignbyte-only)
@@ -589,6 +625,7 @@ if(BUILD_TESTING)
 		memory_tracker_tests
 		page_manager_tests
 		bit_array_tests
+		lru_cache_tests
 		shader_vertex_metadata_tests
 		resource_materialization_tests
 		resource_tracking_tests
@@ -598,9 +635,26 @@ if(BUILD_TESTING)
 		shader_recompiler_compute_tests
 		virtual_memory_allocation_tests
 		audio_out2_port_tests
+		ngs2_sampler_tests
+		save_data_memory_tests
 	)
 endif()

+# Test Darwin signal-context restoration for SSE4a instructions.
+if(APPLE AND (CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" OR
+		(NOT CMAKE_OSX_ARCHITECTURES AND CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|AMD64)$")))
+	add_executable(macos_sse4a_tests EXCLUDE_FROM_ALL
+		"${KYTY_TESTS_DIR}/macosSse4aTests.cpp"
+		"${KYTY_SOURCE_DIR}/loader/x64InstructionEmulator.cpp"
+	)
+	target_link_libraries(macos_sse4a_tests common fmt::fmt Zydis::Zydis)
+	target_include_directories(macos_sse4a_tests PRIVATE ${inc_headers})
+	if(BUILD_TESTING)
+		add_test(NAME macos_sse4a COMMAND $<TARGET_FILE:macos_sse4a_tests>)
+		set_tests_properties(macos_sse4a PROPERTIES TIMEOUT 15)
+		add_dependencies(kyty_tests macos_sse4a_tests)
+	endif()
+endif()

 add_executable(kyty_emulator "${KYTY_SOURCE_DIR}/main.cpp" ${kyty_emulator_src})
 configure_macos_guest_address_space(kyty_emulator)
@@ -640,7 +694,16 @@ if(KYTY_CLANG_CL)
 	)
 	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=${KYTY_EMULATOR_MAP_LINK_PATH}")
+	set_target_properties(kyty_emulator PROPERTIES LINK_FLAGS "${KYTY_LD_OPTIONS}")
+	# LINK_FLAGS is a space-delimited string: a -Map path containing spaces would
+	# be split into bogus arguments, so pass it as a discrete link option instead.
+	target_link_options(kyty_emulator PRIVATE "-Wl,-Map=${KYTY_EMULATOR_MAP_LINK_PATH}")
+endif()
+
+if(LINUX)
+    target_link_options(kyty_emulator PRIVATE
+        $<$<CONFIG:Release>:-Wl,--strip-all>
+    )
 endif()

 if(APPLE)
@@ -4,13 +4,13 @@
 [![Build KytyPS5 (Linux)](https://img.shields.io/github/actions/workflow/status/KytyPS5/KytyPS5/build.yml?branch=main&event=push&label=Build%20KytyPS5%20%28Linux%29)](https://github.com/KytyPS5/KytyPS5/actions/workflows/build.yml)
 [![Build KytyPS5 (macOS)](https://img.shields.io/github/actions/workflow/status/KytyPS5/KytyPS5/build.yml?branch=main&event=push&label=Build%20KytyPS5%20%28macOS%29)](https://github.com/KytyPS5/KytyPS5/actions/workflows/build.yml)
 [![Platform](https://img.shields.io/badge/platform-Windows%20x64%20%7C%20Linux%20x64%20%7C%20macOS%20x86__64-0078D4.svg)](#system-requirements)
-[![Status](https://img.shields.io/badge/status-early%20development-orange.svg)](#current-status)
+[![Status](https://img.shields.io/badge/status-active%20development-orange.svg)](#current-status)
 [![License](https://img.shields.io/badge/license-GPL--2.0-blue.svg)](LICENSE)

 KytyPS5 is a free and open-source PlayStation 5 emulator written in C++ for Windows and Linux,
 with experimental macOS support. It is based on a heavily modified version of
-[Kyty](https://github.com/InoriRus/Kyty). The project is in an early stage of development, so
-compatibility is limited and behavior may change significantly between builds.
+[Kyty](https://github.com/InoriRus/Kyty). The project is in active development, and behavior
+can change significantly between builds.

 > [!IMPORTANT]
 > KytyPS5 is not affiliated with Sony Interactive Entertainment or PlayStation. The project does
@@ -20,12 +20,12 @@ compatibility is limited and behavior may change significantly between builds.
 ## Current Status

 KytyPS5 can boot 2D games and a selection of 3D games, including titles built with Unreal Engine
-4/5, Unity, and custom engines. No external low-level emulation modules are currently required.
+4/5, Unity, and custom engines. External low-level emulation modules are neither required nor
+planned.

-Development is focused on compatibility and boot reliability.
+Development is currently focused on expanding game compatibility and improving boot reliability.

-Windows is the primary platform and receives the most testing. Linux builds and runs; see
-[Building on Linux](#building-on-linux).
+Windows and Linux are the primary platforms and receive the most testing.

 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
@@ -36,16 +36,16 @@ Community game test results are available in the

 ## Bugs and Issues

-The project is in an early stage, so please be mindful when opening new issues. Expect crashes,
-graphical glitches, low compatibility, and poor performance.
+Compatibility, stability, and performance can vary between versions. You may encounter crashes
+or graphical glitches, so please include the version you tested when reporting an issue.

 ## Screenshots

 <table align="center">
  <tr>
    <td align="center">
-      <strong>Disgaea 6</strong><br>
-      <img src="docs/screenshots/ps5-01.png" width="300" alt="Disgaea 6 running in KytyPS5">
+      <strong>Astro Bot</strong><br>
+      <img src="docs/screenshots/ps5-01.png" width="300" alt="Astro Bot running in KytyPS5">
    </td>
    <td align="center">
      <strong>Dreaming Sarah</strong><br>
@@ -64,12 +64,12 @@ graphical glitches, low compatibility, and poor performance.
  </tr>
  <tr>
    <td align="center">
-      <strong>Hellboy</strong><br>
-      <img src="docs/screenshots/ps5-02.png" width="300" alt="Hellboy running in KytyPS5">
+      <strong>Demon's Souls</strong><br>
+      <img src="docs/screenshots/ps5-02.png" width="300" alt="Demon's Souls running in KytyPS5">
    </td>
    <td align="center">
-      <strong>Paleo Pines</strong><br>
-      <img src="docs/screenshots/ps5-06.png" width="300" alt="Paleo Pines running in KytyPS5">
+      <strong>Hellboy</strong><br>
+      <img src="docs/screenshots/ps5-06.png" width="300" alt="Hellboy running in KytyPS5">
    </td>
  </tr>
 </table>
@@ -79,7 +79,7 @@ graphical glitches, low compatibility, and poor performance.
 ## Contributing

 Testing games and submitting detailed bug reports are useful ways to contribute. Search existing
-issues first, then use the **Game Emulation Bug Report** template and attach the complete log file.
+issues first, then use the **Game Emulation Status Report** template and attach the complete log file.

 Code contributions should be focused, build successfully on the platforms they touch, and include
 relevant tests where practical. Windows is the primary target, so a change that alters shared code
@@ -132,11 +132,12 @@ the Vulkan/SPIR-V validation rules.
 ### Build requirements (Windows)

 - Git
- CMake 3.12 or newer
+- CMake 3.22.1 or newer
 - Ninja
 - Visual Studio 2022 or Build Tools 2022 with the **Desktop development with C++** workload and
  **C++ Clang tools for Windows** component
 - Qt 6 for MSVC 2022 64-bit, including Concurrent, Network, and Widgets
+- [glslang](https://github.com/KhronosGroup/glslang/releases) (`glslangValidator`) on `PATH`

 The Microsoft C++ compiler (`cl.exe`) is not supported; use `clang-cl`.

@@ -165,19 +166,19 @@ The finished application and its runtime dependencies will be placed in

 ### Building on Linux

-Install the toolchain and the libraries the bundled SDL2 needs. Without the audio, Wayland and
-udev development packages SDL2 quietly configures itself without those backends, and the resulting
+Install the toolchain and the libraries the bundled SDL3 needs. Without the audio, Wayland and
+udev development packages SDL3 quietly configures itself without those backends, and the resulting
 build has no working sound and no gamepad hotplug:

 ```bash
 sudo apt-get install --no-install-recommends \
  clang lld ninja-build cmake git glslang-tools pkg-config \
  libgl1-mesa-dev libx11-dev libxcursor-dev libxext-dev libxfixes-dev \
-  libxi-dev libxrandr-dev libxss-dev libxkbcommon-dev \
+  libxi-dev libxrandr-dev libxss-dev libxtst-dev libxkbcommon-dev \
  libasound2-dev libpulse-dev libudev-dev libdbus-1-dev libwayland-dev wayland-protocols
 ```

-Qt 6 (Concurrent, Network, Widgets) is also required — either the distribution packages
+Qt 6 (Concurrent, Network, Widgets) is required for the launcher — either the distribution packages
 (`qt6-base-dev`) or an official Qt installation.

 ```bash
@@ -196,6 +197,18 @@ The install step copies the Qt libraries and plugins next to the binaries, so
 from the pinned [KytyPS5 FFmpeg core](https://github.com/KytyPS5/ext-ffmpeg-core)
 release, including VP9 and WebM support. System FFmpeg packages are not required.

+To build `kyty_emulator` and the `kyty_tests` target without Qt, use a separate build directory:
+
+```bash
+git submodule update --init --recursive
+
+cmake -S . -B _Build/linux-no-qt -G Ninja -DCMAKE_BUILD_TYPE=Release \
+  -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ \
+  -DKYTY_BUILD_LAUNCHER=OFF
+
+cmake --build _Build/linux-no-qt --target kyty_emulator kyty_tests --parallel
+```
+
 As on Windows, the MSVC compiler is not used; Clang is required. `cl.exe` is rejected at configure
 time.

@@ -203,7 +216,7 @@ The CMake source root is the repository root.

 ### Building on NixOS

-A development shell provides Clang, CMake, Ninja, Qt 6, the Vulkan headers, and the SDL2 backend
+A development shell provides Clang, CMake, Ninja, Qt 6, the Vulkan headers, and the SDL3 backend
 libraries. Enter it and configure exactly as on other Linux distributions; the shell exports
 `CMAKE_PREFIX_PATH` and `QT_PLUGIN_PATH`, so the `-DCMAKE_PREFIX_PATH="$Qt6_DIR"` argument is not
 needed:
@@ -369,5 +382,6 @@ licenses included with those components.

 - [InoriRus/Kyty](https://github.com/InoriRus/Kyty) — KytyPS5 is based on a heavily modified version
  of the original Kyty project.
- [shadps4-emu/shadPS4](https://github.com/shadps4-emu/shadPS4) — reference for memory-model
-  understanding and the AVPlayer implementation.
+- [shadps4-emu/shadPS4](https://github.com/shadps4-emu/shadPS4) — reference for understanding PS4
+  memory behavior, GPU resource aliasing and cache coherency,
+  and the AVPlayer implementation.
@@ -1,8 +1,7 @@
 { pkgs }:

 let
-  # Build-time dependencies. Every -dev package the bundled SDL2 probes for must
-  # be listed here, otherwise SDL2 quietly drops the audio/Wayland/udev backends.
+  # Build-time dependencies for the bundled SDL3 audio, video and input backends.
  buildDeps = with pkgs; [
    qt6.qtbase # Concurrent, Network, Widgets (launcher)
    vulkan-headers
@@ -18,6 +17,7 @@ let
    libXi
    libXrandr
    libXScrnSaver
+    libXtst
    libxkbcommon
    # Wayland
    wayland
@@ -46,6 +46,7 @@ let
    libXi
    libXrandr
    libXScrnSaver
+    libXtst
    libxkbcommon
    wayland
    libdecor
@@ -1,166 +0,0 @@
-#ifndef KYTY_COMMON_BYTEBUFFER_H_
-#define KYTY_COMMON_BYTEBUFFER_H_
-
-#include "common/assert.h"
-#include "common/common.h"
-
-#include <algorithm>
-#include <cstddef> // IWYU pragma: export
-#include <cstring>
-#include <utility>
-#include <vector>
-
-namespace Common {
-
-using Byte = std::byte;
-
-class ByteBuffer {
-public:
-	using DataType = std::vector<Byte>;
-
-	static constexpr uint32_t INVALID_INDEX = static_cast<uint32_t>(-1);
-
-	using iterator       = DataType::iterator;
-	using const_iterator = DataType::const_iterator;
-
-	ByteBuffer()                  = default;
-	ByteBuffer(const ByteBuffer&) = default;
-	ByteBuffer(ByteBuffer&&)      = default;
-
-	ByteBuffer& operator=(const ByteBuffer&) = default;
-	ByteBuffer& operator=(ByteBuffer&&)      = default;
-
-	explicit ByteBuffer(uint32_t size): m_data(size) {}
-
-	ByteBuffer(std::initializer_list<uint8_t> list) {
-		for (const uint8_t& e: list) {
-			m_data.emplace_back(static_cast<Byte>(e));
-		}
-	}
-
-	explicit ByteBuffer(const void* buf, uint32_t size)
-	    : ByteBuffer(size) // @suppress("Ambiguous problem")
-	{
-		EXIT_IF(buf == nullptr && size != 0);
-
-		if (size != 0) {
-			std::memcpy(m_data.data(), buf, size);
-		}
-	}
-
-	[[nodiscard]] uint32_t Size() const { return static_cast<uint32_t>(m_data.size()); }
-	[[nodiscard]] bool     IsEmpty() const { return m_data.empty(); }
-
-	void Clear() { m_data.clear(); }
-
-	void Add(const Byte& val) { m_data.push_back(val); }
-	void Add(Byte&& val) { m_data.push_back(std::move(val)); }
-
-	void Add(const Byte* val, uint32_t num) {
-		EXIT_IF(val == nullptr && num != 0);
-		m_data.insert(m_data.end(), val, val + num);
-	}
-
-	void Add(const ByteBuffer& v) {
-		if (this == &v) {
-			const ByteBuffer copy(v);
-			Add(copy.GetDataConst(), copy.Size());
-			return;
-		}
-		Add(v.GetDataConst(), v.Size());
-	}
-
-	void InsertAt(uint32_t index, const Byte& val) {
-		EXIT_IF(index > Size());
-		m_data.insert(m_data.begin() + index, val);
-	}
-
-	void InsertAt(uint32_t index, Byte&& val) {
-		EXIT_IF(index > Size());
-		m_data.insert(m_data.begin() + index, std::move(val));
-	}
-
-	[[nodiscard]] uint32_t Find(const Byte& t) const {
-		const auto it = std::find(m_data.begin(), m_data.end(), t);
-		return (it == m_data.end() ? INVALID_INDEX
-		                           : static_cast<uint32_t>(std::distance(m_data.begin(), it)));
-	}
-
-	[[nodiscard]] bool Contains(const Byte& t) const { return Find(t) != INVALID_INDEX; }
-
-	bool Remove(const Byte& t) {
-		const auto it = std::find(m_data.begin(), m_data.end(), t);
-		if (it == m_data.end()) {
-			return false;
-		}
-		m_data.erase(it);
-		return true;
-	}
-
-	bool RemoveAt(uint32_t index, uint32_t count = 1) {
-		if (index >= Size() || count == 0) {
-			return false;
-		}
-
-		const uint32_t erase_count = std::min(count, Size() - index);
-		m_data.erase(m_data.begin() + index, m_data.begin() + index + erase_count);
-		return true;
-	}
-
-	[[nodiscard]] bool IndexValid(uint32_t index) const { return index < Size(); }
-
-	Byte& operator[](uint32_t index) {
-		EXIT_IF(index >= Size());
-		return m_data[index];
-	}
-
-	const Byte& operator[](uint32_t index) const {
-		EXIT_IF(index >= Size());
-		return m_data[index];
-	}
-
-	[[nodiscard]] const Byte& At(uint32_t index) const {
-		EXIT_IF(index >= Size());
-		return m_data[index];
-	}
-
-	Byte*                     GetData() { return m_data.data(); }
-	[[nodiscard]] const Byte* GetData() const { return m_data.data(); }
-	[[nodiscard]] const Byte* GetDataConst() const { return m_data.data(); }
-
-	void Sort() {
-		std::sort(m_data.begin(), m_data.end(), [](Byte a, Byte b) {
-			return std::to_integer<uint8_t>(a) < std::to_integer<uint8_t>(b);
-		});
-	}
-
-	template <typename OP>
-	void Sort(OP&& comp_func) {
-		std::sort(m_data.begin(), m_data.end(), std::forward<OP>(comp_func));
-	}
-
-	bool operator==(const ByteBuffer& s) const { return m_data == s.m_data; }
-	bool operator!=(const ByteBuffer& s) const { return !(*this == s); }
-
-	iterator begin() { return m_data.begin(); } // NOLINT(readability-identifier-naming)
-	iterator end() { return m_data.end(); }     // NOLINT(readability-identifier-naming)
-	[[nodiscard]] const_iterator begin() const {
-		return m_data.begin();
-	} // NOLINT(readability-identifier-naming)
-	[[nodiscard]] const_iterator end() const {
-		return m_data.end();
-	} // NOLINT(readability-identifier-naming)
-	[[nodiscard]] const_iterator cbegin() const {
-		return m_data.cbegin();
-	} // NOLINT(readability-identifier-naming)
-	[[nodiscard]] const_iterator cend() const {
-		return m_data.cend();
-	} // NOLINT(readability-identifier-naming)
-
-private:
-	DataType m_data;
-};
-
-} // namespace Common
-
-#endif /* KYTY_COMMON_BYTEBUFFER_H_ */
@@ -24,8 +24,4 @@ std::string Debug::GetLinker() {
 #endif
 }

-std::string Debug::GetBitness() {
-	return "64";
-}
-
 } // namespace Common
@@ -10,7 +10,6 @@ namespace Common {
 namespace Debug {
 std::string GetCompiler();
 std::string GetLinker();
-std::string GetBitness();
 } // namespace Debug

 } // namespace Common
@@ -135,6 +135,10 @@ bool ReadbackLinearImagesEnabled() {
 	return g_config->readback_linear_images;
 }

+bool TessellationEnabled() {
+	return g_config->tessellation_enabled;
+}
+
 bool PlayGoHackEnabled() {
 	return g_config->playgo_hack_enabled;
 }
@@ -66,6 +66,7 @@ struct ConfigOptions {
 	bool                   gpu_assisted_validation_enabled = false;
 	bool                   renderdoc_enabled           = false;
 	bool                   readback_linear_images      = false;
+	bool                   tessellation_enabled        = false;
 	bool                   playgo_hack_enabled         = false;
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
 	bool red_zone_protection_enabled = false;
@@ -110,6 +111,7 @@ bool GpuAssistedValidationEnabled();

 bool RenderDocEnabled();
 bool ReadbackLinearImagesEnabled();
+bool TessellationEnabled();
 bool PlayGoHackEnabled();
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
 bool RedZoneProtectionEnabled();
@@ -136,10 +136,6 @@ bool File::OpenInMem(void* buf, uint32_t buf_size) {
 	return true;
 }

-bool File::OpenInMem(ByteBuffer& buf) {
-	return OpenInMem(buf.GetData(), buf.Size());
-}
-
 bool File::CreateInMem() {
 	EXIT_IF(m_p->f != nullptr);

@@ -249,88 +245,78 @@ void File::Printf(const char* format, ...) {
 	Write(s.data(), static_cast<uint32_t>(s.size()));
 }

+static std::filesystem::path WithoutTrailingSeparator(
+	const std::filesystem::path& path) {
+		if (!path.empty() && !path.has_filename() &&
+			path != path.root_path()) {
+			return path.parent_path();
+		}
+
+		return path;
+}
+
+
 bool File::IsDirectoryExisting(const std::filesystem::path& path) {
-	auto path_str = PathToGenericString(path);
-	return SysFileIsDirectoryExisting(
-	    Common::EndsWith(path_str, "/") ? Common::RemoveLast(path_str, 1) : path_str);
+	return SysFileIsDirectoryExisting(WithoutTrailingSeparator(path));
 }

 bool File::IsFileExisting(const std::filesystem::path& name) {
 	return SysFileIsFileExisting(name);
 }

-bool File::CreateDirectory(
-    const std::filesystem::path& path) // @suppress("Member declaration not found")
-{
-	auto path_str = PathToGenericString(path);
-	return SysFileCreateDirectory(Common::EndsWith(path_str, "/") ? Common::RemoveLast(path_str, 1)
-	                                                              : path_str);
+bool File::CreateDirectory(const std::filesystem::path& path) {
+    return SysFileCreateDirectory(
+        WithoutTrailingSeparator(path));
 }

 bool File::DeleteDirectory(const std::filesystem::path& path) {
-	auto path_str = PathToGenericString(path);
-	return SysFileDeleteDirectory(Common::EndsWith(path_str, "/") ? Common::RemoveLast(path_str, 1)
-	                                                              : path_str);
+    return SysFileDeleteDirectory(
+        WithoutTrailingSeparator(path));
 }

 bool File::CreateDirectories(const std::filesystem::path& path) {
-	std::string real_path = Common::ReplaceChar(PathToGenericString(path), '\\', '/');
+    const auto normalized_path =
+        WithoutTrailingSeparator(path);

-	std::vector<std::string> list = Common::Split(real_path, "/");
+    auto current_path = normalized_path.root_path();

-	std::string p;
+    for (const auto& component:
+         normalized_path.relative_path()) {
+        current_path /= component;

-	for (uint32_t si = 0; si < list.size(); si++) {
-		const std::string& s = list[si];
+        if (IsDirectoryExisting(current_path)) {
+            continue;
+        }

-		if (si != 0 || Common::StartsWith(real_path, "/")) {
-			p += "/";
-		}
+        if (!CreateDirectory(current_path)) {
+            return false;
+        }
+    }

-		p += s;
-
-		if (IsDirectoryExisting(p)) {
-			continue;
-		}
-
-		if (!CreateDirectory(p)) // @suppress("Invalid arguments")
-		{
-			return false;
-		}
-	}
-
-	return true;
+    return true;
 }

 bool File::DeleteDirectories(const std::filesystem::path& path) {
-	std::string real_path = Common::ReplaceChar(PathToGenericString(path), '\\', '/');
+    const auto normalized_path =
+        WithoutTrailingSeparator(path);

-	std::vector<std::string> list = Common::Split(real_path, "/");
+    std::vector<std::filesystem::path> directories;
+    auto current_path = normalized_path.root_path();

-	std::string              p;
-	std::vector<std::string> list2;
+    for (const auto& component:
+         normalized_path.relative_path()) {
+        current_path /= component;
+        directories.push_back(current_path);
+    }

-	for (uint32_t si = 0; si < list.size(); si++) {
-		const std::string& s = list[si];
+    for (auto it = directories.rbegin();
+         it != directories.rend(); ++it) {
+        if (!DeleteDirectory(*it)) {
+            return false;
+        }
+    }

-		if (si != 0 || Common::StartsWith(real_path, "/")) {
-			p += "/";
-		}
-
-		p += s;
-
-		list2.push_back(p);
-	}
-
-	uint32_t num = list2.size();
-
-	for (uint32_t si = num - 1; si < num; si--) {
-		if (!DeleteDirectory(list2[si])) {
-			return false;
-		}
-	}
-
-	return true;
+    return true;
 }

 bool File::DeleteFile(
@@ -345,7 +331,7 @@ bool File::Flush() {
 	return SysFileFlush(*m_p->f);
 }

-ByteBuffer File::ReadWholeBuffer() {
+std::vector<std::byte> File::ReadWholeBuffer() {
 	EXIT_IF(IsInvalid());
 	EXIT_IF(Tell() != 0);

@@ -353,25 +339,14 @@ ByteBuffer File::ReadWholeBuffer() {

 	EXIT_IF((s >> 32u) != 0);

-	ByteBuffer buf(s);
+	const auto            read_size = static_cast<uint32_t>(s);
+	std::vector<std::byte> buf(read_size);

-	Read(buf.GetData(), s);
+	Read(buf.data(), read_size);

 	return buf;
 }

-ByteBuffer File::Read(uint32_t size) {
-	ByteBuffer buf(size);
-	uint32_t   b = 0;
-	Read(buf.GetData(), size, &b);
-	buf.RemoveAt(b, size - b);
-	return buf;
-}
-
-void File::Write(const ByteBuffer& buf, uint32_t* bytes_written) {
-	Write(buf.GetDataConst(), buf.Size(), bytes_written);
-}
-
 DateTime File::GetLastAccessTimeUTC(const std::filesystem::path& name) {
 	SysTimeStruct t {};
 	SysFileToSystemTimeUtc(SysFileGetLastAccessTimeUtc(name), t);
@@ -470,51 +445,6 @@ bool File::SetLastAccessAndWriteTimeUTC(const std::filesystem::path& name, const
 	return SysFileSetLastAccessAndWriteTimeUtc(name, af, wf);
 }

-std::vector<File::FindInfo> File::FindFiles(const std::filesystem::path& path) {
-	std::vector<sys_file_find_t> files;
-
-	SysFileFindFiles(path, files);
-
-	auto     path_str = PathToGenericString(path);
-	uint32_t len      = static_cast<uint32_t>(path_str.size());
-
-	if (!Common::EndsWith(path_str, "/") && !Common::EndsWith(path_str, "\\")) {
-		len++;
-	}
-
-	std::vector<File::FindInfo> ret;
-	ret.reserve(files.size());
-
-	for (const auto& f: files) {
-		File::FindInfo r {};
-
-		r.path_with_name     = f.path_with_name;
-		r.rel_path_with_name = Common::Mid(PathToGenericString(f.path_with_name), len);
-		r.size               = f.size;
-
-		SysTimeStruct at {};
-		SysTimeStruct wt {};
-
-		SysFileToSystemTimeUtc(f.last_access_time, at);
-		SysFileToSystemTimeUtc(f.last_write_time, wt);
-
-		if (!at.is_invalid && !wt.is_invalid) {
-			r.last_access_time = DateTime(Date(at.Year, at.Month, at.Day),
-			                              Time(at.Hour, at.Minute, at.Second, at.Milliseconds));
-			r.last_write_time  = DateTime(Date(wt.Year, wt.Month, wt.Day),
-			                              Time(wt.Hour, wt.Minute, wt.Second, wt.Milliseconds));
-		}
-
-		// printf("%s, %s, %" PRIu64", %s, %s\n", r.path_with_name.c_str(),
-		// r.rel_path_with_name.c_str(), r.size, r.last_access_time.ToString().c_str(),
-		// r.last_write_time.ToString().c_str());
-
-		ret.push_back(std::move(r));
-	}
-
-	return ret;
-}
-
 std::vector<File::DirEntry> File::GetDirEntries(const std::filesystem::path& path) {
 	std::vector<sys_dir_entry_t> files;

@@ -523,13 +453,8 @@ std::vector<File::DirEntry> File::GetDirEntries(const std::filesystem::path& pat
 	std::vector<File::DirEntry> ret;
 	ret.reserve(files.size());

-	for (const auto& f: files) {
-		File::DirEntry r {};
-
-		r.name    = f.name;
-		r.is_file = f.is_file;
-
-		ret.push_back(std::move(r));
+	for (auto& f: files) {
+		ret.emplace_back(std::move(f.name), f.is_file);
 	}

 	return ret;
@@ -1,10 +1,10 @@
 #ifndef KYTY_COMMON_FILE_H_
 #define KYTY_COMMON_FILE_H_

-#include "common/byteBuffer.h"
 #include "common/common.h"
 #include "common/dateTime.h"

+#include <cstddef>
 #include <filesystem>
 #include <memory>
 #include <string>
@@ -30,14 +30,6 @@ class File {
 public:
 	enum class Mode { Read, Write, ReadWrite, WriteRead };

-	struct FindInfo {
-		std::filesystem::path path_with_name;
-		std::filesystem::path rel_path_with_name;
-		DateTime              last_access_time;
-		DateTime              last_write_time;
-		uint64_t              size;
-	};
-
 	struct DirEntry {
 		std::string name;
 		bool        is_file;
@@ -51,7 +43,6 @@ public:
 	bool Create(const std::filesystem::path& name);
 	bool Open(const std::filesystem::path& name, Mode mode);
 	bool OpenInMem(void* buf, uint32_t buf_size);
-	bool OpenInMem(ByteBuffer& buf); // NOLINT(google-runtime-references)
 	bool CreateInMem();

 	void Close();
@@ -73,14 +64,12 @@ public:

 	void GetLastAccessAndWriteTimeUTC(DateTime* access, DateTime* write);

-	void       Read(void* data, uint32_t size, uint32_t* bytes_read = nullptr);
-	ByteBuffer Read(uint32_t size);
-	void       Write(const void* data, uint32_t size, uint32_t* bytes_written = nullptr);
-	void       Write(const ByteBuffer& buf, uint32_t* bytes_written = nullptr);
+	void Read(void* data, uint32_t size, uint32_t* bytes_read = nullptr);
+	void Write(const void* data, uint32_t size, uint32_t* bytes_written = nullptr);

 	void Printf(const char* format, ...) KYTY_FORMAT_PRINTF(2, 3);

-	ByteBuffer ReadWholeBuffer();
+	std::vector<std::byte> ReadWholeBuffer();

 	static uint64_t Size(const std::filesystem::path& name);

@@ -101,7 +90,6 @@ public:
 	static bool SetLastAccessAndWriteTimeUTC(const std::filesystem::path& name,
 	                                         const DateTime& access, const DateTime& write);

-	static std::vector<FindInfo> FindFiles(const std::filesystem::path& path);
 	static std::vector<DirEntry> GetDirEntries(const std::filesystem::path& path);
 	static bool CopyFile(const std::filesystem::path& src, const std::filesystem::path& dst);
 	static bool RenameFile(const std::filesystem::path& src, const std::filesystem::path& dst);
@@ -54,103 +54,6 @@ inline uint32_t hash(const void* key, uint32_t key_len) {
 	return hash;
 }

-inline uint32_t hash8(uint8_t key) {
-	uint32_t hash = 0;
-
-	uint8_t* ptr = &key;
-
-	hash += ptr[0];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-
-	hash += (hash << 3u);
-	hash ^= (hash >> 11u);
-	hash += (hash << 15u);
-
-	return hash;
-}
-
-inline uint32_t hash16(uint16_t key) {
-	uint32_t hash = 0;
-
-	auto* ptr = reinterpret_cast<uint8_t*>(&key);
-
-	hash += ptr[0];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[1];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-
-	hash += (hash << 3u);
-	hash ^= (hash >> 11u);
-	hash += (hash << 15u);
-
-	return hash;
-}
-
-inline uint32_t hash32(uint32_t key) {
-	uint32_t hash = 0;
-
-	auto* ptr = reinterpret_cast<uint8_t*>(&key);
-
-	hash += ptr[0];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[1];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[2];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[3];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-
-	hash += (hash << 3u);
-	hash ^= (hash >> 11u);
-	hash += (hash << 15u);
-
-	return hash;
-}
-
-inline uint32_t hash64(uint64_t key) {
-	uint32_t hash = 0;
-
-	auto* ptr = reinterpret_cast<uint8_t*>(&key);
-
-	hash += ptr[0];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[1];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[2];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[3];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[4];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[5];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[6];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-	hash += ptr[7];
-	hash += (hash << 10u);
-	hash ^= (hash >> 6u);
-
-	hash += (hash << 3u);
-	hash ^= (hash >> 11u);
-	hash += (hash << 15u);
-
-	return hash;
-}
-
 } // namespace Common

 #endif /* KYTY_COMMON_HASH_H_ */
@@ -3,7 +3,6 @@

 #include "common/assert.h"
 #include "common/emulatorConfig.h"
-#include "common/stringUtils.h"

 #include <cstdio>
 #include <filesystem>
@@ -47,7 +46,7 @@ std::shared_ptr<spdlog::logger> MakeFileLogger(std::string                  name
 	}

 	auto sink =
-	    std::make_shared<spdlog::sinks::basic_file_sink_mt>(Common::PathToString(path), true);
+	    std::make_shared<spdlog::sinks::basic_file_sink_mt>(path.native(), true);
 	return MakeLogger(std::move(name), std::move(sink));
 }

@@ -1,42 +0,0 @@
-#ifndef KYTY_COMMON_MAGICENUM_H_
-#define KYTY_COMMON_MAGICENUM_H_
-
-#include "common/common.h"
-#include "common/stringUtils.h"
-#include "magic_enum.hpp" // IWYU pragma: export
-
-namespace Common {
-
-template <typename E>
-inline std::string EnumName(E v) {
-	auto str = magic_enum::enum_name(v);
-	return std::string(str.data(), str.length());
-}
-
-template <typename E>
-inline std::string EnumName8(E v) {
-	auto str = magic_enum::enum_name(v);
-	return std::string(str.data(), str.length());
-}
-
-template <typename E>
-inline E EnumValue(const std::string& str, E default_value) {
-	auto v = magic_enum::enum_cast<E>(str.c_str());
-	if (v.has_value()) {
-		return v.value();
-	}
-	return default_value;
-}
-
-#define KYTY_ENUM_RANGE(e, mx, mn)                                                                 \
-	namespace magic_enum::customize {                                                              \
-	template <>                                                                                    \
-	struct enum_range<e> {                                                                         \
-		static constexpr int min = (mx);                                                           \
-		static constexpr int max = (mn);                                                           \
-	};                                                                                             \
-	}
-
-} // namespace Common
-
-#endif /* KYTY_COMMON_MAGICENUM_H_ */
@@ -18,14 +18,6 @@ enum sys_file_cache_type_t {
 	SYS_FILE_CACHE_SEQUENTIAL_SCAN = 2  // NOLINT(readability-identifier-naming)
 };

-// NOLINTNEXTLINE(readability-identifier-naming)
-struct sys_file_find_t {
-	std::filesystem::path path_with_name;
-	SysFileTimeStruct     last_access_time;
-	SysFileTimeStruct     last_write_time;
-	uint64_t              size;
-};
-
 // NOLINTNEXTLINE(readability-identifier-naming)
 struct sys_dir_entry_t {
 	std::string name;
@@ -52,7 +44,6 @@ uint64_t    SysFileTell(sys_file_t& f);                    // NOLINT(google-runt
 bool        SysFileTruncate(sys_file_t& f, uint64_t size); // NOLINT(google-runtime-references)
 bool        SysFileUnlink(sys_file_t&                  f,
                          const std::filesystem::path& name); // NOLINT(google-runtime-references)
-void        SysFileWrite(uint32_t n, sys_file_t& f);          // NOLINT(google-runtime-references)
 uint64_t    SysFileSize(const std::filesystem::path& file_name);
 bool        SysFileIsError(sys_file_t& f); // NOLINT(google-runtime-references)
 bool        SysFileIsDirectoryExisting(const std::filesystem::path& path);
@@ -76,8 +67,6 @@ bool SysFileSetLastWriteTimeUtc(const std::filesystem::path& name, SysFileTimeSt
 bool SysFileSetLastAccessAndWriteTimeUtc(const std::filesystem::path& name,
                                         SysFileTimeStruct& access, SysFileTimeStruct& write);
 // NOLINTNEXTLINE(google-runtime-references)
-void SysFileFindFiles(const std::filesystem::path& path, std::vector<sys_file_find_t>& out);
-// NOLINTNEXTLINE(google-runtime-references)
 void SysFileGetDents(const std::filesystem::path& path, std::vector<sys_dir_entry_t>& out);
 bool SysFileCopyFile(const std::filesystem::path& src, const std::filesystem::path& dst);
 bool SysFileRenameFile(const std::filesystem::path& src, const std::filesystem::path& dst);
@@ -11,6 +11,7 @@

 #include <cerrno>
 #include <cstdlib>
+#include <cstring>
 #include <dirent.h>
 #include <fcntl.h>
 #include <filesystem>
@@ -151,10 +152,6 @@ void SysFileWrite(const void* data, uint32_t size, sys_file_t& f, uint32_t* byte
 	}
 }

-void SysFileWrite(uint32_t n, sys_file_t& f) {
-	SysFileWrite(&n, 4, f);
-}
-
 sys_file_t* SysFileCreate(const std::filesystem::path& file_name) {
 	auto* ret = new sys_file_t;

@@ -613,51 +610,6 @@ bool SysFileSetLastAccessAndWriteTimeUtc(const std::filesystem::path& name,
 	//	}
 }

-// Recursively collect regular files.
-void SysFileFindFiles(const std::filesystem::path& path, std::vector<sys_file_find_t>& out) {
-	auto real_path = get_internal_name(path);
-
-	DIR* dir = opendir(real_path.string().c_str());
-	if (dir == nullptr) {
-		return;
-	}
-
-	for (const dirent* entry = readdir(dir); entry != nullptr; entry = readdir(dir)) {
-		const std::string file_name(entry->d_name);
-
-		if (file_name == "." || file_name == "..") {
-			continue;
-		}
-
-		auto        child = real_path / file_name;
-		struct stat s {};
-
-		// lstat, so a symlink is never followed into a cycle during the recursive walk.
-		if (0 != lstat(child.string().c_str(), &s)) {
-			continue;
-		}
-
-		if (S_ISDIR(s.st_mode)) {
-			SysFileFindFiles(child, out);
-		} else if (S_ISREG(s.st_mode)) {
-			sys_file_find_t r {};
-
-			r.path_with_name              = child;
-			r.size                        = static_cast<uint64_t>(s.st_size);
-			r.last_access_time.is_invalid = false;
-			r.last_access_time.time       = s.st_atime;
-			r.last_access_time.nanos      = KYTY_STAT_ATIME_NS(s);
-			r.last_write_time.is_invalid  = false;
-			r.last_write_time.time        = s.st_mtime;
-			r.last_write_time.nanos       = KYTY_STAT_MTIME_NS(s);
-
-			out.push_back(r);
-		}
-	}
-
-	closedir(dir);
-}
-
 // Keep "." and ".." to match FindFirstFileW.
 void SysFileGetDents(const std::filesystem::path& path, std::vector<sys_dir_entry_t>& out) {
 	auto real_path = get_internal_name(path);
@@ -5,7 +5,6 @@
 #else

 #include "common/assert.h"
-#include "common/platform/sysVirtual.h"
 #include "common/virtualMemory.h"

 #include <atomic>
@@ -28,12 +27,12 @@
 #define KYTY_FIXED_NOREPLACE
 #endif

-namespace Common {
+namespace Common::VirtualMemory {

 static pthread_mutex_t              g_virtual_mutex {};
 static std::map<uintptr_t, size_t>* g_allocs = nullptr;

-void SysVirtualInit() {
+void Init() {
 	pthread_mutexattr_t attr {};

 	pthread_mutexattr_init(&attr);
@@ -48,19 +47,19 @@ void SysVirtualInit() {
 	g_allocs = new std::map<uintptr_t, size_t>;
 }

-static int get_protection_flag(VirtualMemory::Mode mode) {
+static int get_protection_flag(Mode mode) {
 	int protect = PROT_NONE;
 	switch (mode) {
-		case VirtualMemory::Mode::Read: protect = PROT_READ; break;
-		case VirtualMemory::Mode::Write:
-		case VirtualMemory::Mode::ReadWrite: protect = PROT_READ | PROT_WRITE; break; // NOLINT
-		case VirtualMemory::Mode::Execute: protect = PROT_EXEC; break;
-		case VirtualMemory::Mode::ExecuteRead: protect = PROT_EXEC | PROT_READ; break; // NOLINT
-		case VirtualMemory::Mode::ExecuteWrite:
-		case VirtualMemory::Mode::ExecuteReadWrite:
+		case Mode::Read: protect = PROT_READ; break;
+		case Mode::Write:
+		case Mode::ReadWrite: protect = PROT_READ | PROT_WRITE; break; // NOLINT
+		case Mode::Execute: protect = PROT_EXEC; break;
+		case Mode::ExecuteRead: protect = PROT_EXEC | PROT_READ; break; // NOLINT
+		case Mode::ExecuteWrite:
+		case Mode::ExecuteReadWrite:
 			protect = PROT_EXEC | PROT_WRITE | PROT_READ;
 			break; // NOLINT
-		case VirtualMemory::Mode::NoAccess:
+		case Mode::NoAccess:
 		default: protect = PROT_NONE; break;
 	}
 	return protect;
@@ -130,7 +129,7 @@ static void* map_anonymous(uintptr_t addr, size_t size, int protect, int flags)
 	return mmap(nullptr, size, protect, flags, -1, 0); // NOLINT
 }

-uint64_t SysVirtualAlloc(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
+uint64_t Alloc(uint64_t address, uint64_t size, Mode mode) {
 	EXIT_IF(g_allocs == nullptr);

 	auto addr = static_cast<uintptr_t>(address);
@@ -154,8 +153,7 @@ static uintptr_t align_up(uintptr_t addr, uint64_t alignment) {
 	return (addr + alignment - 1) & ~(alignment - 1);
 }

-uint64_t SysVirtualAllocAligned(uint64_t address, uint64_t size, VirtualMemory::Mode mode,
-                                uint64_t alignment) {
+uint64_t AllocAligned(uint64_t address, uint64_t size, Mode mode, uint64_t alignment) {
 	if (alignment == 0) {
 		return 0;
 	}
@@ -224,7 +222,7 @@ uint64_t SysVirtualAllocAligned(uint64_t address, uint64_t size, VirtualMemory::
 	}

 	if (ptr == MAP_FAILED) {
-		return SysVirtualAllocAligned(address, size, mode, alignment << 1u);
+		return AllocAligned(address, size, mode, alignment << 1u);
 	}

 	pthread_mutex_lock(&g_virtual_mutex);
@@ -281,7 +279,7 @@ static bool is_mapped(void* ptr, size_t length) {
 }
 #endif

-bool SysVirtualAllocFixed(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
+bool AllocFixed(uint64_t address, uint64_t size, Mode mode) {
 	EXIT_IF(g_allocs == nullptr);

 	auto addr    = static_cast<uintptr_t>(address);
@@ -318,15 +316,15 @@ bool SysVirtualAllocFixed(uint64_t address, uint64_t size, VirtualMemory::Mode m
 	return false;
 }

-bool SysVirtualCommit(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
-	return SysVirtualProtect(address, size, mode);
+bool Commit(uint64_t address, uint64_t size, Mode mode) {
+	return Protect(address, size, mode);
 }

-uint64_t SysVirtualReserve(uint64_t address, uint64_t size) {
-	return SysVirtualReserveAligned(address, size, 1);
+uint64_t Reserve(uint64_t address, uint64_t size) {
+	return ReserveAligned(address, size, 1);
 }

-uint64_t SysVirtualReserveAligned(uint64_t address, uint64_t size, uint64_t alignment) {
+uint64_t ReserveAligned(uint64_t address, uint64_t size, uint64_t alignment) {
 	if (alignment == 0) {
 		return 0;
 	}
@@ -395,7 +393,7 @@ uint64_t SysVirtualReserveAligned(uint64_t address, uint64_t size, uint64_t alig
 	}

 	if (ptr == MAP_FAILED) {
-		return SysVirtualReserveAligned(address, size, alignment << 1u);
+		return ReserveAligned(address, size, alignment << 1u);
 	}

 	pthread_mutex_lock(&g_virtual_mutex);
@@ -405,7 +403,7 @@ uint64_t SysVirtualReserveAligned(uint64_t address, uint64_t size, uint64_t alig
 	return ret_addr;
 }

-bool SysVirtualReserveFixed(uint64_t address, uint64_t size) {
+bool ReserveFixed(uint64_t address, uint64_t size) {
 	EXIT_IF(g_allocs == nullptr);

 	auto addr = static_cast<uintptr_t>(address);
@@ -441,9 +439,9 @@ bool SysVirtualReserveFixed(uint64_t address, uint64_t size) {
 	return false;
 }

-bool SysVirtualDecommit(uint64_t address, uint64_t size) {
+bool Decommit(uint64_t address, uint64_t size) {
 	// Drop physical pages while preserving the reservation.
-	if (!SysVirtualProtect(address, size, VirtualMemory::Mode::NoAccess)) {
+	if (!Protect(address, size, Mode::NoAccess)) {
 		return false;
 	}

@@ -467,7 +465,7 @@ bool SysVirtualDecommit(uint64_t address, uint64_t size) {
 	return true;
 }

-bool SysVirtualFree(uint64_t address) {
+bool Free(uint64_t address) {
 	EXIT_IF(g_allocs == nullptr);
 	size_t size = 0;

@@ -487,7 +485,7 @@ bool SysVirtualFree(uint64_t address) {
 	return munmap(reinterpret_cast<void*>(addr), size) == 0;
 }

-bool SysVirtualFreeRange(uint64_t address, uint64_t size) {
+bool FreeRange(uint64_t address, uint64_t size) {
 	EXIT_IF(g_allocs == nullptr);
 	if (size == 0 || (address & 0xfffu) != 0 || (size & 0xfffu) != 0) {
 		return false;
@@ -543,7 +541,7 @@ bool SysVirtualFreeRange(uint64_t address, uint64_t size) {
 	return true;
 }

-bool SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
+bool Protect(uint64_t address, uint64_t size, Mode mode) {
 	const auto addr       = static_cast<uintptr_t>(address);
 	const auto page_start = addr >> 12u;
 	const auto page_end   = (addr + size - 1) >> 12u;
@@ -551,10 +549,10 @@ bool SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode
 	                get_protection_flag(mode)) == 0;
 }

-bool SysVirtualFlushInstructionCache(uint64_t /*address*/, uint64_t /*size*/) {
+bool FlushInstructionCache(uint64_t /*address*/, uint64_t /*size*/) {
 	return true;
 }

-} // namespace Common
+} // namespace Common::VirtualMemory

 #endif
@@ -1,53 +0,0 @@
-#ifndef KYTY_COMMON_PLATFORM_SYSSWAPBYTEORDER_H_
-#define KYTY_COMMON_PLATFORM_SYSSWAPBYTEORDER_H_
-
-#include "common/common.h"
-
-#include <type_traits>
-
-inline uint16_t SwapByteOrder16(uint16_t value) {
-	uint16_t hi = value << 8u;
-	uint16_t lo = value >> 8u;
-	return hi | lo;
-}
-
-inline uint32_t SwapByteOrder32(uint32_t value) {
-	return __builtin_bswap32(value);
-}
-
-inline uint64_t SwapByteOrder64(uint64_t value) {
-	return __builtin_bswap64(value);
-}
-
-template <typename T>
-inline void SwapByteOrder(T& x) {
-	if (sizeof(x) == 2) {
-		if (std::is_signed_v<T>) {
-			x = static_cast<std::make_signed_t<T>>(
-			    SwapByteOrder16(static_cast<std::make_unsigned_t<uint16_t>>(x)));
-		} else {
-			x = SwapByteOrder16(x);
-		}
-	}
-	if (sizeof(x) == 4) {
-		if (std::is_signed_v<T>) {
-			x = static_cast<std::make_signed_t<T>>(
-			    SwapByteOrder32(static_cast<std::make_unsigned_t<uint32_t>>(x)));
-		} else {
-			x = SwapByteOrder32(x);
-		}
-	}
-	if (sizeof(x) == 8) {
-		if (std::is_signed_v<T>) {
-			x = static_cast<std::make_signed_t<T>>(
-			    SwapByteOrder64(static_cast<std::make_unsigned_t<uint64_t>>(x)));
-		} else {
-			x = SwapByteOrder64(x);
-		}
-	}
-}
-
-template <typename T>
-inline void NoSwapByteOrder(T& x) {}
-
-#endif /* KYTY_COMMON_PLATFORM_SYSSWAPBYTEORDER_H_ */
@@ -151,10 +151,6 @@ inline void SysTimeTToSystem(time_t t, SysTimeStruct& s) {
 	SysFileToSystemTimeUtc(ft, s);
 }

-inline time_t SysTimegm(struct tm* tm) {
-	return timegm(tm);
-}
-
 inline void SysSystemToFileTimeUtc(const SysTimeStruct& f, SysFileTimeStruct& t) {
 	struct tm i {};

@@ -165,7 +161,7 @@ inline void SysSystemToFileTimeUtc(const SysTimeStruct& f, SysFileTimeStruct& t)
 	i.tm_min  = f.Minute;
 	i.tm_sec  = f.Second;

-	t.is_invalid = (f.is_invalid || (t.time = SysTimegm(&i)) == static_cast<time_t>(-1));
+	t.is_invalid = (f.is_invalid || (t.time = ::timegm(&i)) == static_cast<time_t>(-1));
 }

 // Retrieves the current local date and time.
@@ -1,27 +0,0 @@
-#ifndef KYTY_COMMON_PLATFORM_SYSVIRTUAL_H_
-#define KYTY_COMMON_PLATFORM_SYSVIRTUAL_H_
-
-#include "common/common.h"
-#include "common/virtualMemory.h"
-
-namespace Common {
-
-void SysVirtualInit();
-
-uint64_t SysVirtualAlloc(uint64_t address, uint64_t size, VirtualMemory::Mode mode);
-uint64_t SysVirtualAllocAligned(uint64_t address, uint64_t size, VirtualMemory::Mode mode,
-                                uint64_t alignment);
-bool     SysVirtualAllocFixed(uint64_t address, uint64_t size, VirtualMemory::Mode mode);
-bool     SysVirtualCommit(uint64_t address, uint64_t size, VirtualMemory::Mode mode);
-uint64_t SysVirtualReserve(uint64_t address, uint64_t size);
-uint64_t SysVirtualReserveAligned(uint64_t address, uint64_t size, uint64_t alignment);
-bool     SysVirtualReserveFixed(uint64_t address, uint64_t size);
-bool     SysVirtualDecommit(uint64_t address, uint64_t size);
-bool     SysVirtualFree(uint64_t address);
-bool     SysVirtualFreeRange(uint64_t address, uint64_t size);
-bool     SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode);
-bool     SysVirtualFlushInstructionCache(uint64_t address, uint64_t size);
-
-} // namespace Common
-
-#endif /* KYTY_COMMON_PLATFORM_SYSVIRTUAL_H_ */
@@ -14,6 +14,7 @@
 #include "common/stringUtils.h"

 #include <cstdlib>
+#include <cstring>
 #include <vector>

 // NOLINTNEXTLINE(readability-identifier-naming)
@@ -56,10 +57,10 @@ static DWORD GetCacheAccessType(sys_file_cache_type_t t) {
 	}

 	if (t == SYS_FILE_CACHE_SEQUENTIAL_SCAN) {
-		return SYS_FILE_CACHE_SEQUENTIAL_SCAN;
+		return FILE_FLAG_SEQUENTIAL_SCAN;
 	}

-	return SYS_FILE_CACHE_AUTO;
+	return FILE_ATTRIBUTE_NORMAL;
 }

 void SysFileRead(void* data, uint32_t size, sys_file_t& f, uint32_t* bytes_read) {
@@ -135,10 +136,6 @@ void SysFileWrite(const void* data, uint32_t size, sys_file_t& f, uint32_t* byte
 	}
 }

-void SysFileWrite(uint32_t n, sys_file_t& f) {
-	SysFileWrite(&n, 4, f);
-}
-
 sys_file_t* SysFileCreate(const std::filesystem::path& file_name) {
 	auto* ret = new sys_file_t;

@@ -479,66 +476,11 @@ bool SysFileSetLastAccessAndWriteTimeUtc(const std::filesystem::path& name,
 	return ok;
 }

-void SysFileFindFiles(const std::filesystem::path& path, std::vector<sys_file_find_t>& out) {
-	std::string real_path = Common::ReplaceChar(Common::PathToGenericString(path), '\\', '/');
-	if (!Common::EndsWith(real_path, "/")) {
-		real_path += "/";
-	}
-
-	std::string pattern = real_path + "*";
-
-	HANDLE           h = nullptr;
-	WIN32_FIND_DATAW data;
-
-	auto wide_pattern = std::filesystem::path(pattern).wstring();
-	h                 = FindFirstFileW(wide_pattern.c_str(), &data);
-
-	if (h == INVALID_HANDLE_VALUE) {
-		return;
-	}
-
-	do {
-		std::filesystem::path file_name(data.cFileName);
-		auto                  file_name_str = Common::PathToString(file_name);
-
-		if (file_name_str == "." || file_name_str == "..") {
-			continue;
-		}
-
-		if ((data.dwFileAttributes & static_cast<DWORD>(FILE_ATTRIBUTE_DIRECTORY)) != 0u) {
-			SysFileFindFiles(std::filesystem::path(real_path) / file_name, out);
-		} else {
-			sys_file_find_t r {};
-
-			r.path_with_name              = std::filesystem::path(real_path) / file_name;
-			r.size                        = (static_cast<uint64_t>(data.nFileSizeHigh) << 32u) +
-			                                static_cast<uint64_t>(data.nFileSizeLow);
-			r.last_access_time.is_invalid = false;
-			r.last_access_time.time       = data.ftLastAccessTime;
-			r.last_write_time.is_invalid  = false;
-			r.last_write_time.time        = data.ftLastWriteTime;
-
-			out.push_back(r);
-		}
-
-	} while (FindNextFileW(h, &data) != 0);
-
-	FindClose(h);
-}
-
 void SysFileGetDents(const std::filesystem::path& path, std::vector<sys_dir_entry_t>& out) {
-	std::string real_path = Common::ReplaceChar(Common::PathToGenericString(path), '\\', '/');
-	if (!Common::EndsWith(real_path, "/")) {
-		real_path += "/";
-	}
+	const auto pattern = path / L"*";

-	std::string pattern = real_path + "*";
-
-	HANDLE           h = nullptr;
-	WIN32_FIND_DATAW data;
-
-	auto wide_pattern = std::filesystem::path(pattern).wstring();
-	h                 = FindFirstFileW(wide_pattern.c_str(), &data);
+	WIN32_FIND_DATAW data {};
+	HANDLE h = FindFirstFileW(pattern.c_str(), &data);

 	if (h == INVALID_HANDLE_VALUE) {
 		return;
@@ -552,7 +494,7 @@ void SysFileGetDents(const std::filesystem::path& path, std::vector<sys_dir_entr
 		r.is_file = ((data.dwFileAttributes & static_cast<DWORD>(FILE_ATTRIBUTE_DIRECTORY)) == 0u);
 		r.name    = Common::PathToString(file_name);

-		out.push_back(r);
+		out.push_back(std::move(r));

 	} while (FindNextFileW(h, &data) != 0);

@@ -10,7 +10,6 @@
 #include <windows.h> // IWYU pragma: keep

 #include "common/assert.h"
-#include "common/platform/sysVirtual.h"
 #include "common/virtualMemory.h"

 // IWYU pragma: no_include <basetsd.h>
@@ -22,33 +21,33 @@
 // IWYU pragma: no_include <winerror.h>
 // IWYU pragma: no_include <wtypes.h>

-namespace Common {
+namespace Common::VirtualMemory {

-static DWORD GetProtectionFlag(VirtualMemory::Mode mode) {
+static DWORD GetProtectionFlag(Mode mode) {
 	DWORD protect = PAGE_NOACCESS;
 	switch (mode) {
-		case VirtualMemory::Mode::Read: protect = PAGE_READONLY; break;
+		case Mode::Read: protect = PAGE_READONLY; break;

-		case VirtualMemory::Mode::Write:
-		case VirtualMemory::Mode::ReadWrite: protect = PAGE_READWRITE; break;
+		case Mode::Write:
+		case Mode::ReadWrite: protect = PAGE_READWRITE; break;

-		case VirtualMemory::Mode::Execute: protect = PAGE_EXECUTE; break;
+		case Mode::Execute: protect = PAGE_EXECUTE; break;

-		case VirtualMemory::Mode::ExecuteRead: protect = PAGE_EXECUTE_READ; break;
+		case Mode::ExecuteRead: protect = PAGE_EXECUTE_READ; break;

-		case VirtualMemory::Mode::ExecuteWrite:
-		case VirtualMemory::Mode::ExecuteReadWrite: protect = PAGE_EXECUTE_READWRITE; break;
+		case Mode::ExecuteWrite:
+		case Mode::ExecuteReadWrite: protect = PAGE_EXECUTE_READWRITE; break;

-		case VirtualMemory::Mode::NoAccess:
+		case Mode::NoAccess:
 		default: protect = PAGE_NOACCESS; break;
 	}
 	return protect;
 }

-void SysVirtualInit() {}
+void Init() {}

-uint64_t SysVirtualAlloc(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
-	auto ptr = (address == 0 ? SysVirtualAllocAligned(address, size, mode, 1)
+uint64_t Alloc(uint64_t address, uint64_t size, Mode mode) {
+	auto ptr = (address == 0 ? AllocAligned(address, size, mode, 1)
 	                         : reinterpret_cast<uintptr_t>(VirtualAlloc(
 	                               reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)), size,
 	                               static_cast<DWORD>(MEM_COMMIT) | static_cast<DWORD>(MEM_RESERVE),
@@ -59,7 +58,7 @@ uint64_t SysVirtualAlloc(uint64_t address, uint64_t size, VirtualMemory::Mode mo
 		if (err != ERROR_INVALID_ADDRESS) {
 			printf("VirtualAlloc() failed: 0x%08" PRIx32 "\n", err);
 		} else {
-			return SysVirtualAllocAligned(address, size, mode, 1);
+			return AllocAligned(address, size, mode, 1);
 		}
 	}
 	return ptr;
@@ -80,8 +79,7 @@ static uint64_t AlignUp(uint64_t addr, uint64_t alignment) {
 	return (addr + alignment - 1) & ~(alignment - 1);
 }

-uint64_t SysVirtualAllocAligned(uint64_t address, uint64_t size, VirtualMemory::Mode mode,
-                                uint64_t alignment) {
+uint64_t AllocAligned(uint64_t address, uint64_t size, Mode mode, uint64_t alignment) {
 	if (alignment == 0) {
 		printf("VirtualAlloc2 failed: 0x%08" PRIx32 "\n", static_cast<uint32_t>(GetLastError()));
 		return 0;
@@ -135,13 +133,13 @@ uint64_t SysVirtualAllocAligned(uint64_t address, uint64_t size, VirtualMemory::
 			printf("VirtualAlloc2(alignment = 0x%016" PRIx64 ") failed: 0x%08" PRIx32 "\n",
 			       alignment, err);
 		} else {
-			return SysVirtualAllocAligned(address, size, mode, alignment << 1u);
+			return AllocAligned(address, size, mode, alignment << 1u);
 		}
 	}
 	return ptr;
 }

-bool SysVirtualAllocFixed(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
+bool AllocFixed(uint64_t address, uint64_t size, Mode mode) {
 	auto ptr = reinterpret_cast<uintptr_t>(VirtualAlloc(
 	    reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)), size,
 	    static_cast<DWORD>(MEM_COMMIT) | static_cast<DWORD>(MEM_RESERVE), GetProtectionFlag(mode)));
@@ -161,7 +159,7 @@ bool SysVirtualAllocFixed(uint64_t address, uint64_t size, VirtualMemory::Mode m
 	return true;
 }

-bool SysVirtualCommit(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
+bool Commit(uint64_t address, uint64_t size, Mode mode) {
 	auto ptr = reinterpret_cast<uintptr_t>(
 	    VirtualAlloc(reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)), size,
 	                 static_cast<DWORD>(MEM_COMMIT), GetProtectionFlag(mode)));
@@ -179,8 +177,8 @@ bool SysVirtualCommit(uint64_t address, uint64_t size, VirtualMemory::Mode mode)
 	return true;
 }

-uint64_t SysVirtualReserve(uint64_t address, uint64_t size) {
-	auto ptr = (address == 0 ? SysVirtualReserveAligned(address, size, 1)
+uint64_t Reserve(uint64_t address, uint64_t size) {
+	auto ptr = (address == 0 ? ReserveAligned(address, size, 1)
 	                         : reinterpret_cast<uintptr_t>(VirtualAlloc(
 	                               reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)), size,
 	                               static_cast<DWORD>(MEM_RESERVE), PAGE_NOACCESS)));
@@ -190,13 +188,13 @@ uint64_t SysVirtualReserve(uint64_t address, uint64_t size) {
 		if (err != ERROR_INVALID_ADDRESS) {
 			printf("VirtualAlloc(MEM_RESERVE) failed: 0x%08" PRIx32 "\n", err);
 		} else {
-			return SysVirtualReserveAligned(address, size, 1);
+			return ReserveAligned(address, size, 1);
 		}
 	}
 	return ptr;
 }

-uint64_t SysVirtualReserveAligned(uint64_t address, uint64_t size, uint64_t alignment) {
+uint64_t ReserveAligned(uint64_t address, uint64_t size, uint64_t alignment) {
 	if (alignment == 0) {
 		printf("VirtualAlloc2(MEM_RESERVE) failed: 0x%08" PRIx32 "\n",
 		       static_cast<uint32_t>(GetLastError()));
@@ -250,13 +248,13 @@ uint64_t SysVirtualReserveAligned(uint64_t address, uint64_t size, uint64_t alig
 			       "\n",
 			       alignment, err);
 		} else {
-			return SysVirtualReserveAligned(address, size, alignment << 1u);
+			return ReserveAligned(address, size, alignment << 1u);
 		}
 	}
 	return ptr;
 }

-bool SysVirtualReserveFixed(uint64_t address, uint64_t size) {
+bool ReserveFixed(uint64_t address, uint64_t size) {
 	auto ptr = reinterpret_cast<uintptr_t>(
 	    VirtualAlloc(reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)), size,
 	                 static_cast<DWORD>(MEM_RESERVE), PAGE_NOACCESS));
@@ -276,7 +274,7 @@ bool SysVirtualReserveFixed(uint64_t address, uint64_t size) {
 	return true;
 }

-bool SysVirtualDecommit(uint64_t address, uint64_t size) {
+bool Decommit(uint64_t address, uint64_t size) {
 	if (VirtualFree(reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)), size,
 	                MEM_DECOMMIT) == 0) {
 		printf("VirtualFree(MEM_DECOMMIT) failed: 0x%08" PRIx32 "\n",
@@ -286,7 +284,7 @@ bool SysVirtualDecommit(uint64_t address, uint64_t size) {
 	return true;
 }

-bool SysVirtualFree(uint64_t address) {
+bool Free(uint64_t address) {
 	if (VirtualFree(reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)), 0, MEM_RELEASE) ==
 	    0) {
 		printf("VirtualFree() failed: 0x%08" PRIx32 "\n", static_cast<uint32_t>(GetLastError()));
@@ -295,7 +293,7 @@ bool SysVirtualFree(uint64_t address) {
 	return true;
 }

-bool SysVirtualFreeRange(uint64_t address, uint64_t size) {
+bool FreeRange(uint64_t address, uint64_t size) {
 	if (address == 0 || size == 0) {
 		return false;
 	}
@@ -313,10 +311,10 @@ bool SysVirtualFreeRange(uint64_t address, uint64_t size) {
 		}
 		current = next;
 	}
-	return current - address == size && SysVirtualFree(address);
+	return current - address == size && Free(address);
 }

-bool SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode) {
+bool Protect(uint64_t address, uint64_t size, Mode mode) {
 	DWORD old_protect = 0;
 	if (VirtualProtect(reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)), size,
 	                   GetProtectionFlag(mode), &old_protect) == 0) {
@@ -326,7 +324,7 @@ bool SysVirtualProtect(uint64_t address, uint64_t size, VirtualMemory::Mode mode
 	return true;
 }

-bool SysVirtualFlushInstructionCache(uint64_t address, uint64_t size) {
+bool FlushInstructionCache(uint64_t address, uint64_t size) {
 	if (::FlushInstructionCache(GetCurrentProcess(),
 	                            reinterpret_cast<LPVOID>(static_cast<uintptr_t>(address)),
 	                            size) == 0) {
@@ -337,6 +335,6 @@ bool SysVirtualFlushInstructionCache(uint64_t address, uint64_t size) {
 	return true;
 }

-} // namespace Common
+} // namespace Common::VirtualMemory

 #endif
@@ -1,7 +1,6 @@
 #ifndef KYTY_COMMON_STRING_UTILS_H_
 #define KYTY_COMMON_STRING_UTILS_H_

-#include "common/byteBuffer.h"
 #include "common/common.h"

 #include <algorithm>
@@ -9,7 +8,6 @@
 #include <charconv>
 #include <codecvt>
 #include <filesystem>
-#include <fmt/format.h>
 #include <locale>
 #include <string>
 #include <string_view>
@@ -18,8 +16,6 @@

 namespace Common {

-constexpr uint32_t FIND_INVALID_INDEX = static_cast<uint32_t>(-1);
-
 inline std::string PathToString(const std::filesystem::path& path) {
 #if defined(__cpp_char8_t)
 	auto u8 = path.u8string();
@@ -38,8 +34,12 @@ inline std::string PathToGenericString(const std::filesystem::path& path) {
 #endif
 }

-inline bool IsSpace(char c) {
-	return std::isspace(static_cast<unsigned char>(c)) != 0;
+inline std::filesystem::path PathFromUtf8(std::string_view text) {
+#if defined(__cpp_char8_t)
+	return std::filesystem::path(std::u8string(text.begin(), text.end()));
+#else
+	return std::filesystem::u8path(text.begin(), text.end());
+#endif
 }

 inline bool EqualNoCase(std::string_view lhs, std::string_view rhs) {
@@ -63,65 +63,6 @@ inline std::string ToLower(std::string_view text) {
 	return ret;
 }

-inline bool ContainsStr(std::string_view text, std::string_view needle) {
-	return text.find(needle) != std::string_view::npos;
-}
-
-inline bool StartsWith(std::string_view text, std::string_view prefix) {
-	return text.starts_with(prefix);
-}
-
-inline bool StartsWith(std::string_view text, char prefix) {
-	return !text.empty() && text.front() == prefix;
-}
-
-inline bool EndsWith(std::string_view text, std::string_view suffix) {
-	return text.ends_with(suffix);
-}
-
-inline bool EndsWith(std::string_view text, char suffix) {
-	return !text.empty() && text.back() == suffix;
-}
-
-inline bool ContainsChar(std::string_view text, char needle) {
-	return text.find(needle) != std::string_view::npos;
-}
-
-inline bool IndexValid(std::string_view text, uint32_t index) {
-	return index < text.size();
-}
-
-inline uint32_t FindIndex(std::string_view text, std::string_view needle, size_t from = 0) {
-	const auto pos = text.find(needle, from);
-	return pos == std::string_view::npos ? FIND_INVALID_INDEX : static_cast<uint32_t>(pos);
-}
-
-inline uint32_t FindIndex(std::string_view text, char needle, size_t from = 0) {
-	const auto pos = text.find(needle, from);
-	return pos == std::string_view::npos ? FIND_INVALID_INDEX : static_cast<uint32_t>(pos);
-}
-
-inline uint32_t FindLastIndex(std::string_view text, std::string_view needle,
-                              size_t from = FIND_INVALID_INDEX) {
-	if (needle.empty()) {
-		return 0;
-	}
-	if (from == FIND_INVALID_INDEX || from > text.size()) {
-		from = text.size();
-	}
-	const auto pos = text.rfind(needle, from);
-	return pos == std::string_view::npos ? FIND_INVALID_INDEX : static_cast<uint32_t>(pos);
-}
-
-inline uint32_t FindLastIndex(std::string_view text, char needle,
-                              size_t from = FIND_INVALID_INDEX) {
-	if (from == FIND_INVALID_INDEX || from > text.size()) {
-		from = text.size();
-	}
-	const auto pos = text.rfind(needle, from);
-	return pos == std::string_view::npos ? FIND_INVALID_INDEX : static_cast<uint32_t>(pos);
-}
-
 inline std::string Mid(std::string_view text, size_t first, size_t count = std::string_view::npos) {
 	if (first >= text.size()) {
 		return {};
@@ -129,49 +70,18 @@ inline std::string Mid(std::string_view text, size_t first, size_t count = std::
 	return std::string(text.substr(first, count));
 }

-inline std::string Left(std::string_view text, size_t count) {
-	return Mid(text, 0, count);
-}
-
-inline std::string Right(std::string_view text, size_t count) {
-	if (count >= text.size()) {
-		return std::string(text);
-	}
-	return Mid(text, text.size() - count, count);
-}
-
 inline std::string RemoveLast(std::string_view text, size_t count) {
 	if (count >= text.size()) {
 		return {};
 	}
-	return Left(text, text.size() - count);
+	return std::string(text.substr(0, text.size() - count));
 }

 inline std::string RemoveFirst(std::string_view text, size_t count) {
 	if (count >= text.size()) {
 		return {};
 	}
-	return Right(text, text.size() - count);
-}
-
-inline std::string TrimRight(std::string_view text) {
-	size_t end = text.size();
-	while (end > 0 && IsSpace(text[end - 1])) {
-		end--;
-	}
-	return std::string(text.substr(0, end));
-}
-
-inline std::string TrimLeft(std::string_view text) {
-	size_t first = 0;
-	while (first < text.size() && IsSpace(text[first])) {
-		first++;
-	}
-	return std::string(text.substr(first));
-}
-
-inline std::string Trim(std::string_view text) {
-	return TrimRight(TrimLeft(text));
+	return std::string(text.substr(count));
 }

 inline std::string ReplaceChar(std::string_view text, char old_char, char new_char) {
@@ -204,16 +114,6 @@ inline std::string FilenameWithoutDirectory(std::string_view text) {
 	return pos == std::string_view::npos ? std::string(text) : std::string(text.substr(pos + 1));
 }

-inline std::string FilenameWithoutExtension(std::string_view text) {
-	const auto pos = text.find_last_of('.');
-	return pos == std::string_view::npos ? std::string(text) : std::string(text.substr(0, pos));
-}
-
-inline std::string ExtensionWithoutFilename(std::string_view text) {
-	const auto pos = text.find_last_of('.');
-	return pos == std::string_view::npos ? std::string() : std::string(text.substr(pos));
-}
-
 inline std::string FixFilenameSlash(std::string_view text) {
 	return ReplaceChar(text, '\\', '/');
 }
@@ -257,65 +157,12 @@ inline std::vector<std::string> Split(std::string_view text, char sep, bool keep
 	return Split(text, std::string_view(&sep, 1), keep_empty);
 }

-inline std::string Concat(const std::vector<std::string>& list, std::string_view sep) {
-	std::string ret;
-	for (size_t i = 0; i < list.size(); i++) {
-		if (i != 0) {
-			ret += sep;
-		}
-		ret += list[i];
-	}
-	return ret;
-}
-
-inline std::string Concat(const std::vector<std::string>& list, char sep) {
-	return Concat(list, std::string_view(&sep, 1));
-}
-
-inline uint32_t ToUint32(std::string_view text, int base = 10) {
-	uint32_t value = 0;
-	std::from_chars(text.data(), text.data() + text.size(), value, base);
-	return value;
-}
-
-inline uint64_t ToUint64(std::string_view text, int base = 10) {
-	uint64_t value = 0;
-	std::from_chars(text.data(), text.data() + text.size(), value, base);
-	return value;
-}
-
 inline int32_t ToInt32(std::string_view text, int base = 10) {
 	int32_t value = 0;
 	std::from_chars(text.data(), text.data() + text.size(), value, base);
 	return value;
 }

-inline int64_t ToInt64(std::string_view text, int base = 10) {
-	int64_t value = 0;
-	std::from_chars(text.data(), text.data() + text.size(), value, base);
-	return value;
-}
-
-inline double ToDouble(std::string_view text) {
-	return std::stod(std::string(text));
-}
-
-inline float ToFloat(std::string_view text) {
-	return std::stof(std::string(text));
-}
-
-inline std::string SafeCsv(std::string_view text) {
-	const bool add_space =
-	    !text.empty() && (text.front() == '+' || text.front() == '=' || text.front() == '-');
-	if (text.find('"') != std::string_view::npos || text.find(';') != std::string_view::npos ||
-	    text.find('+') != std::string_view::npos || text.find('=') != std::string_view::npos ||
-	    text.find('-') != std::string_view::npos) {
-		return "\"" + std::string(add_space ? " " : "") +
-		       ReplaceStr(std::string(text), "\"", "\"\"") + "\"";
-	}
-	return std::string(text);
-}
-
 inline std::string Utf16ToUtf8(std::u16string_view utf16) {
 	if (utf16.empty()) {
 		return {};
@@ -324,23 +171,6 @@ inline std::string Utf16ToUtf8(std::u16string_view utf16) {
 	return convert.to_bytes(utf16.data(), utf16.data() + utf16.size());
 }

-inline ByteBuffer HexToBin(std::string_view text) {
-	ByteBuffer ret;
-	for (size_t i = 0; i + 1 < text.size(); i += 2) {
-		ret.Add(static_cast<Byte>((ToUint32(text.substr(i, 2), 16) & 0xffu)));
-	}
-	return ret;
-}
-
-inline std::string HexFromBin(const ByteBuffer& bin) {
-	std::string ret;
-	ret.reserve(bin.Size() * 2);
-	for (uint32_t i = 0; i < bin.Size(); i++) {
-		ret += fmt::format("{:02X}", std::to_integer<uint8_t>(bin.At(i)));
-	}
-	return ret;
-}
-
 } // namespace Common

 #endif /* KYTY_COMMON_STRING_UTILS_H_ */
@@ -253,10 +253,6 @@ void Thread::Detach() {
 	m_thread->m_thread.detach();
 }

-void Thread::Sleep(uint32_t millis) {
-	std::this_thread::sleep_for(std::chrono::milliseconds(millis));
-}
-
 void Thread::SleepMicro(uint32_t micros) {
 #ifdef KYTY_WIN_CS
 	SleepHighResolution100ns(static_cast<uint64_t>(micros) * 10);
@@ -31,7 +31,6 @@ public:
 	// The id is unique and can't be reused by another thread.
 	[[nodiscard]] int GetUniqueId() const;

-	static void Sleep(uint32_t millis);
 	static void SleepMicro(uint32_t micros);
 	static void SleepNano(uint64_t nanos);
 	static bool IsMainThread();
@@ -38,16 +38,7 @@ bool Timer::IsPaused() const {

 // return time in milliseconds
 double Timer::GetTimeMs() const {
-	if (m_is_paused) {
-		return 1000.0 * (static_cast<double>(m_PauseTime - m_StartTime)) /
-		       static_cast<double>(m_Frequency);
-	}
-
-	uint64_t current_time = 0;
-	SysQueryPerformanceCounter(&current_time);
-
-	return 1000.0 * (static_cast<double>(current_time - m_StartTime)) /
-	       static_cast<double>(m_Frequency);
+	return GetTimeS() * 1000.0;
 }

 // return time in seconds
@@ -62,23 +53,6 @@ double Timer::GetTimeS() const {
 	return (static_cast<double>(current_time - m_StartTime)) / static_cast<double>(m_Frequency);
 }

-// return time in ticks
-uint64_t Timer::GetTicks() const {
-	if (m_is_paused) {
-		return (m_PauseTime - m_StartTime);
-	}
-
-	uint64_t current_time = 0;
-	SysQueryPerformanceCounter(&current_time);
-
-	return (current_time - m_StartTime);
-}
-
-// return ticks frequency
-uint64_t Timer::GetFrequency() const {
-	return m_Frequency;
-}
-
 uint64_t Timer::QueryPerformanceFrequency() {
 	uint64_t ret = 0;
 	SysQueryPerformanceFrequency(&ret);
@@ -24,12 +24,6 @@ public:
 	// return time in seconds
 	[[nodiscard]] double GetTimeS() const;

-	// return time in ticks
-	[[nodiscard]] uint64_t GetTicks() const;
-
-	// return ticks frequency
-	[[nodiscard]] uint64_t GetFrequency() const;
-
 	KYTY_CLASS_NO_COPY(Timer);

 	[[nodiscard]] static uint64_t QueryPerformanceFrequency();
@@ -1,63 +0,0 @@
-#include "common/virtualMemory.h"
-
-#include "common/platform/sysVirtual.h"
-
-namespace Common {
-
-namespace VirtualMemory {
-
-void Init() {
-	SysVirtualInit();
-}
-
-uint64_t Alloc(uint64_t address, uint64_t size, Mode mode) {
-	return SysVirtualAlloc(address, size, mode);
-}
-
-uint64_t AllocAligned(uint64_t address, uint64_t size, Mode mode, uint64_t alignment) {
-	return SysVirtualAllocAligned(address, size, mode, alignment);
-}
-
-bool AllocFixed(uint64_t address, uint64_t size, Mode mode) {
-	return SysVirtualAllocFixed(address, size, mode);
-}
-
-bool Commit(uint64_t address, uint64_t size, Mode mode) {
-	return SysVirtualCommit(address, size, mode);
-}
-
-uint64_t Reserve(uint64_t address, uint64_t size) {
-	return SysVirtualReserve(address, size);
-}
-
-uint64_t ReserveAligned(uint64_t address, uint64_t size, uint64_t alignment) {
-	return SysVirtualReserveAligned(address, size, alignment);
-}
-
-bool ReserveFixed(uint64_t address, uint64_t size) {
-	return SysVirtualReserveFixed(address, size);
-}
-
-bool Decommit(uint64_t address, uint64_t size) {
-	return SysVirtualDecommit(address, size);
-}
-
-bool Free(uint64_t address) {
-	return SysVirtualFree(address);
-}
-
-bool FreeRange(uint64_t address, uint64_t size) {
-	return SysVirtualFreeRange(address, size);
-}
-
-bool Protect(uint64_t address, uint64_t size, Mode mode) {
-	return SysVirtualProtect(address, size, mode);
-}
-
-bool FlushInstructionCache(uint64_t address, uint64_t size) {
-	return SysVirtualFlushInstructionCache(address, size);
-}
-
-} // namespace VirtualMemory
-
-} // namespace Common
@@ -73,6 +73,9 @@ public:
 	void SetIndexBufferSize(uint32_t index_buffer_size);
 	void SetDrawIndirectArgsBaseAddress(uint64_t draw_indirect_args_base_addr);
 	void SetDispatchIndirectArgsBaseAddress(uint64_t dispatch_indirect_args_base_addr);
+	[[nodiscard]] uint64_t GetDispatchIndirectArgsBaseAddress() const {
+		return m_dispatch_indirect_args_base_addr;
+	}
 	void SetNumInstances(uint32_t num_instances);
 	void DrawIndex(DrawIndexArgs args);
 	void DrawIndexOffset(uint32_t index_offset, uint32_t index_count);
@@ -99,7 +102,7 @@ public:
 	void TriggerEopEventAtEndOfPipe(uint32_t interrupt_context_id);
 	void DispatchDirect(uint32_t thread_group_x, uint32_t thread_group_y, uint32_t thread_group_z,
 	                    uint32_t mode);
-	void DispatchIndirect(uint32_t data_offset, uint32_t mode);
+	void DispatchIndirect(uint64_t args_addr, uint32_t mode);
 	void WaitFlipDone(uint32_t video_out_handle, uint32_t display_buffer_index);
 	void TriggerEvent(uint32_t event_type, uint32_t event_index, uint64_t event_address = 0);

@@ -148,7 +151,6 @@ private:
 	void SuspendPm4();
 	CommandScheduler&   GetScheduler() const { return m_renderer.GetCommandScheduler(); }
 	CommandBuffer&      CurrentBuffer() { return GetScheduler().Current(); }
-	void                CheckBuffer() const { GetScheduler().CheckActive(); }

 	RenderContext&   m_renderer;
 	HW::Context      m_ctx;
@@ -1331,26 +1331,13 @@ KYTY_CP_OP_PARSER(CpOpDispatchIndirect) {
 	EXIT_NOT_IMPLEMENTED(cmd_id != 0xc0011600 && cmd_id != 0xc0021600);

 	if (cmd_id == 0xc0021600) {
-		struct DispatchIndirectArgs {
-			uint32_t thread_group_x;
-			uint32_t thread_group_y;
-			uint32_t thread_group_z;
-		};
-
-		auto* args = reinterpret_cast<const DispatchIndirectArgs*>(
-		    buffer[0] | (static_cast<uint64_t>(buffer[1]) << 32u));
-		uint32_t mode = buffer[2];
-
-		EXIT_NOT_IMPLEMENTED(args == nullptr);
-		cp.DispatchDirect(args->thread_group_x, args->thread_group_y, args->thread_group_z, mode);
-
+		cp.DispatchIndirect(buffer[0] | (static_cast<uint64_t>(buffer[1]) << 32u), buffer[2]);
 		return 3;
 	}

-	uint32_t data_offset = buffer[0];
-	uint32_t mode        = buffer[1];
-
-	cp.DispatchIndirect(data_offset, mode);
+	const auto base_addr = cp.GetDispatchIndirectArgsBaseAddress();
+	EXIT_NOT_IMPLEMENTED(base_addr == 0);
+	cp.DispatchIndirect(base_addr + buffer[0], buffer[1]);

 	return 2;
 }
@@ -2017,6 +2004,14 @@ KYTY_CP_OP_PARSER(CpOpIndirectCxRegs) {
 		auto pfunc = g_hw_ctx_indirect_func[cmd_offset & (Pm4::CX_NUM - 1)];

 		if (pfunc == nullptr) {
+			if (raw_cmd_offset == 0x24au && value == 0u) {
+				static std::atomic_flag logged = ATOMIC_FLAG_INIT;
+				if (!logged.test_and_set(std::memory_order_relaxed)) {
+					Log::WriteToConsoleAndLog(
+					    "\t diagnostic: ignoring indirect CX {0x24a, 0}; hardware effect unresolved\n");
+				}
+				continue;
+			}
 			EXIT("unknown cx reg at %05" PRIx32 ": 0x%" PRIx32 "\n", num_dw - dw, cmd_offset);
 		}

@@ -26,6 +26,10 @@ enum class PrimitiveType : uint32_t {
 	kPolygon            = 21,
 };

+constexpr bool IsRectList(PrimitiveType type) {
+	return type == PrimitiveType::kRectList || type == PrimitiveType::kRectListLegacy;
+}
+
 enum class IndexType : uint32_t {
 	kIndex16 = 0,
 	kIndex32 = 1,
@@ -874,13 +874,11 @@ void CommandProcessor::SetPredication(uint32_t condition, uint32_t op, uint32_t
 }

 void CommandProcessor::DrawIndex(DrawIndexArgs args) {
-	CheckBuffer();
-
 	args.index_type_and_size = m_index_type_and_size;
 	if (args.instance_count == 0) {
 		args.instance_count = m_num_instances;
 	}
-	if (args.base_vertex != 0 || args.first_instance != 0) {
+	if (GraphicsRunDebugDumpEnabled() && (args.base_vertex != 0 || args.first_instance != 0)) {
 		LOGF("\t draw indexed offsets: base_vertex = %" PRId32 ", first_instance = %" PRIu32 "\n",
 		     args.base_vertex, args.first_instance);
 	}
@@ -912,17 +910,6 @@ void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiato
 	if (!indexed) {
 		DrawIndirectArgs args {};
 		std::memcpy(&args, args_addr, sizeof(args));
-		if (args.instance_count != 1u || args.start_vertex_location != 0u ||
-		    args.start_instance_location != 0u) {
-			static std::atomic<uint32_t> log_count {0};
-			if (log_count.fetch_add(1) < 64) {
-				LOGF("\t warning: partial DrawIndirect args: vertex_count=%" PRIu32
-				     ", instance_count=%" PRIu32 ", start_vertex=%" PRIu32
-				     ", start_instance=%" PRIu32 "\n",
-				     args.vertex_count_per_instance, args.instance_count,
-				     args.start_vertex_location, args.start_instance_location);
-			}
-		}
 		m_num_instances = args.instance_count;
 		DrawIndexAuto({.vertex_count   = args.vertex_count_per_instance,
 		               .instance_count = args.instance_count,
@@ -934,16 +921,6 @@ void CommandProcessor::DrawIndirect(uint32_t data_offset, uint32_t draw_initiato

 	DrawIndexedIndirectArgs args {};
 	std::memcpy(&args, args_addr, sizeof(args));
-	if (args.base_vertex_location != 0u || args.start_instance_location != 0u) {
-		static std::atomic<uint32_t> log_count {0};
-		if (log_count.fetch_add(1) < 64) {
-			LOGF("\t warning: partial DrawIndexIndirect args: index_count=%" PRIu32
-			     ", instance_count=%" PRIu32 ", start_index=%" PRIu32 ", base_vertex=%" PRIu32
-			     ", start_instance=%" PRIu32 "\n",
-			     args.index_count_per_instance, args.instance_count, args.start_index_location,
-			     args.base_vertex_location, args.start_instance_location);
-		}
-	}

 	uint64_t index_size = 0;
 	switch (m_index_type_and_size) {
@@ -999,23 +976,22 @@ void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_coun
 	const auto args_size = indexed ? sizeof(DrawIndexedIndirectArgs) : sizeof(DrawIndirectArgs);
 	EXIT_NOT_IMPLEMENTED(stride_in_bytes < args_size);

+	uint64_t index_size = 0;
+	if (indexed) {
+		switch (m_index_type_and_size) {
+			case 0: index_size = 2; break;
+			case 1: index_size = 4; break;
+			case 2: index_size = 1; break;
+			default: EXIT("unknown index_type_and_size: %u\n", m_index_type_and_size);
+		}
+	}
+
 	for (uint32_t i = 0; i < draw_count; i++) {
 		const auto args_addr = m_draw_indirect_args_base_addr + data_offset +
 		                       static_cast<uint64_t>(i) * stride_in_bytes;

 		if (!indexed) {
 			auto* args = reinterpret_cast<const DrawIndirectArgs*>(args_addr);
-			if (args->instance_count != 1u || args->start_vertex_location != 0u ||
-			    args->start_instance_location != 0u) {
-				static std::atomic<uint32_t> log_count {0};
-				if (log_count.fetch_add(1) < 64) {
-					LOGF("\t warning: partial DrawIndirectMulti args[%u]: vertex_count=%" PRIu32
-					     ", instance_count=%" PRIu32 ", start_vertex=%" PRIu32
-					     ", start_instance=%" PRIu32 "\n",
-					     i, args->vertex_count_per_instance, args->instance_count,
-					     args->start_vertex_location, args->start_instance_location);
-				}
-			}
 			m_num_instances = args->instance_count;
 			DrawIndexAuto({.vertex_count   = args->vertex_count_per_instance,
 			               .instance_count = args->instance_count,
@@ -1026,25 +1002,6 @@ void CommandProcessor::DrawIndirectMulti(uint32_t data_offset, uint32_t max_coun
 		}

 		auto* args = reinterpret_cast<const DrawIndexedIndirectArgs*>(args_addr);
-		if (args->base_vertex_location != 0u || args->start_instance_location != 0u) {
-			static std::atomic<uint32_t> log_count {0};
-			if (log_count.fetch_add(1) < 64) {
-				LOGF("\t warning: partial DrawIndexIndirectMulti args[%u]: index_count=%" PRIu32
-				     ", instance_count=%" PRIu32 ", start_index=%" PRIu32 ", base_vertex=%" PRIu32
-				     ", start_instance=%" PRIu32 "\n",
-				     i, args->index_count_per_instance, args->instance_count,
-				     args->start_index_location, args->base_vertex_location,
-				     args->start_instance_location);
-			}
-		}
-
-		uint64_t index_size = 0;
-		switch (m_index_type_and_size) {
-			case 0: index_size = 2; break;
-			case 1: index_size = 4; break;
-			case 2: index_size = 1; break;
-			default: EXIT("unknown index_type_and_size: %u\n", m_index_type_and_size);
-		}

 		auto* index_addr = reinterpret_cast<const void*>(
 		    m_index_base_addr + static_cast<uint64_t>(args->start_index_location) * index_size);
@@ -1082,7 +1039,6 @@ void CommandProcessor::DispatchDirect(uint32_t thread_group_x, uint32_t thread_g
 	// uint32_t local_z   = 1;

 	{
-		CheckBuffer();
 		frame_num = m_renderer.GetGpu().GetFrameNum();
 		if (GraphicsRunDebugDumpEnabled()) {
 			static std::atomic<uint32_t> log_count {0};
@@ -1133,24 +1089,18 @@ void CommandProcessor::DispatchDirect(uint32_t thread_group_x, uint32_t thread_g
 	}*/
 }

-void CommandProcessor::DispatchIndirect(uint32_t data_offset, uint32_t mode) {
-	struct DispatchIndirectArgs {
-		uint32_t thread_group_x;
-		uint32_t thread_group_y;
-		uint32_t thread_group_z;
-	};
-
-	EXIT_NOT_IMPLEMENTED(m_dispatch_indirect_args_base_addr == 0);
-
-	const auto args_addr = m_dispatch_indirect_args_base_addr + data_offset;
-	auto*      args      = reinterpret_cast<const DispatchIndirectArgs*>(args_addr);
-
-	DispatchDirect(args->thread_group_x, args->thread_group_y, args->thread_group_z, mode);
+void CommandProcessor::DispatchIndirect(uint64_t args_addr, uint32_t mode) {
+	EXIT_NOT_IMPLEMENTED(args_addr == 0 || (args_addr & 3u) != 0);
+	if ((mode & Pm4::COMPUTE_DISPATCH_INITIATOR_USE_THREAD_DIMENSIONS) != 0) {
+		const auto* args = reinterpret_cast<const vk::DispatchIndirectCommand*>(args_addr);
+		DispatchDirect(args->x, args->y, args->z, mode);
+		return;
+	}
+	m_sh_ctx.SetCsWaveSize(Pm4::ComputeWaveSize(mode));
+	m_renderer.GetRenderExecutor().DispatchIndirect(m_submit_id, CurrentBuffer(), args_addr, mode);
 }

 void CommandProcessor::DrawIndexAuto(DrawAutoArgs args) {
-	CheckBuffer();
-
 	if (args.instance_count == 0) {
 		args.instance_count = m_num_instances;
 	}
@@ -1172,7 +1122,7 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
                                        uint32_t interrupt_context_id) {
 	static_assert(sizeof(T) == sizeof(uint32_t) || sizeof(T) == sizeof(uint64_t));

-	CheckBuffer();
+	auto& command = CurrentBuffer();

 	if (GraphicsRunDebugDumpEnabled()) {
 		const auto bits      = static_cast<unsigned>(sizeof(T) * 8u);
@@ -1203,7 +1153,7 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 		case 0x03: with_interrupt = false; break;
 		case 0x01:
 			if (!IsAsyncComputeQueue()) {
-				Sync::TriggerEopEventAtEndOfPipe(CurrentBuffer(), m_interrupt_event_id,
+				Sync::TriggerEopEventAtEndOfPipe(command, m_interrupt_event_id,
 				                                 interrupt_context_id);
 				return;
 			}
@@ -1220,17 +1170,17 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr

 		if (with_interrupt) {
 			if (with_writeback) {
-				Sync::WriteAtEndOfPipeWithInterruptWriteBack32(m_submit_id, CurrentBuffer(), dst,
+				Sync::WriteAtEndOfPipeWithInterruptWriteBack32(m_submit_id, command, dst,
 				                                               data, m_interrupt_event_id,
 				                                               interrupt_context_id);
 			} else {
-				Sync::WriteAtEndOfPipeWithInterrupt32(m_submit_id, CurrentBuffer(), dst, data,
+				Sync::WriteAtEndOfPipeWithInterrupt32(m_submit_id, command, dst, data,
 				                                      m_interrupt_event_id, interrupt_context_id);
 			}
 		} else if (with_writeback) {
-			Sync::WriteAtEndOfPipeWithWriteBack32(m_submit_id, CurrentBuffer(), dst, data);
+			Sync::WriteAtEndOfPipeWithWriteBack32(m_submit_id, command, dst, data);
 		} else {
-			Sync::WriteAtEndOfPipe32(m_submit_id, CurrentBuffer(), dst, data);
+			Sync::WriteAtEndOfPipe32(m_submit_id, command, dst, data);
 		}
 	};

@@ -1242,7 +1192,7 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 					SynchronizeGpu();
 					Sync::ReadGds(*m_renderer.GetBufferCache().GetGdsBuffer(), dst, value & 0xffffu,
 					              value >> 16u);
-					Sync::WriteAtEndOfPipeGds32(m_submit_id, CurrentBuffer(), dst, value & 0xffffu,
+					Sync::WriteAtEndOfPipeGds32(m_submit_id, command, dst, value & 0xffffu,
 					                            value >> 16u);
 					if (with_interrupt) {
 						m_renderer.TriggerInterrupt(m_interrupt_event_id, interrupt_context_id);
@@ -1255,8 +1205,9 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 			}
 			break;
 		case 0x02:
+		case 0x04:
 			if constexpr (sizeof(T) == sizeof(uint32_t)) {
-				if (eop_event_type == 0x2f && event_index == 0x06) {
+				if (event_write_source == 0x02 && eop_event_type == 0x2f && event_index == 0x06) {
 					switch (cache_action) {
 						case 0x00: write32(false); return;
 						case 0x38: write32(true); return;
@@ -1264,6 +1215,9 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 					}
 				}
 			} else {
+				if (event_write_source == 0x04) {
+					value = Sync::ReadReferenceClock();
+				}
 				auto write64 = [&](bool with_writeback) {
 					auto* dst = static_cast<uint64_t*>(dst_gpu_addr);
 					std::memcpy(dst, &value, sizeof(value));
@@ -1271,18 +1225,18 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 					if (with_interrupt) {
 						if (with_writeback) {
 							Sync::WriteAtEndOfPipeWithInterruptWriteBack64(
-							    m_submit_id, CurrentBuffer(), dst, value, m_interrupt_event_id,
+							    m_submit_id, command, dst, value, m_interrupt_event_id,
 							    interrupt_context_id);
 						} else {
-							Sync::WriteAtEndOfPipeWithInterrupt64(m_submit_id, CurrentBuffer(), dst,
+							Sync::WriteAtEndOfPipeWithInterrupt64(m_submit_id, command, dst,
 							                                      value, m_interrupt_event_id,
 							                                      interrupt_context_id);
 						}
 					} else if (with_writeback) {
-						Sync::WriteAtEndOfPipeWithWriteBack64(m_submit_id, CurrentBuffer(), dst,
+						Sync::WriteAtEndOfPipeWithWriteBack64(m_submit_id, command, dst,
 						                                      value);
 					} else {
-						Sync::WriteAtEndOfPipe64(m_submit_id, CurrentBuffer(), dst, value);
+						Sync::WriteAtEndOfPipe64(m_submit_id, command, dst, value);
 					}
 				};

@@ -1320,7 +1274,7 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 							case 0x14:
 							case 0x28:
 								if (((eop_event_type == 0x04 || eop_event_type == 0x28) &&
-								     event_index == 0x05 && !with_interrupt) ||
+								     event_index == 0x05) ||
 								    (event_index == 0x00)) {
 									write64(true);
 									return;
@@ -1352,45 +1306,6 @@ void CommandProcessor::WriteAtEndOfPipe(uint32_t cache_policy, uint32_t event_wr
 				}
 			}
 			break;
-		case 0x04:
-			if constexpr (sizeof(T) == sizeof(uint64_t)) {
-				const auto clock = Sync::ReadReferenceClock();
-				auto*      dst   = static_cast<uint64_t*>(dst_gpu_addr);
-				std::memcpy(dst, &clock, sizeof(clock));
-				switch (cache_action) {
-					case 0x00:
-						if ((eop_event_type == 0x04 && event_index == 0x05) ||
-						    (eop_event_type == 0x28 && event_index == 0x00)) {
-							if (with_interrupt) {
-								Sync::WriteAtEndOfPipeWithInterrupt64(
-								    m_submit_id, CurrentBuffer(), dst, clock, m_interrupt_event_id,
-								    interrupt_context_id);
-							} else {
-								Sync::WriteAtEndOfPipeClockCounter(m_submit_id, CurrentBuffer(),
-								                                   dst, clock);
-							}
-							return;
-						}
-						break;
-					case 0x38:
-						if ((eop_event_type == 0x04 &&
-						     (event_index == 0x00 || event_index == 0x05)) ||
-						    (eop_event_type == 0x28 && event_index == 0x00)) {
-							if (with_interrupt) {
-								Sync::WriteAtEndOfPipeWithInterruptWriteBack64(
-								    m_submit_id, CurrentBuffer(), dst, clock, m_interrupt_event_id,
-								    interrupt_context_id);
-							} else {
-								Sync::WriteAtEndOfPipeClockCounterWithWriteBack(
-								    m_submit_id, CurrentBuffer(), dst, clock);
-							}
-							return;
-						}
-						break;
-					default: break;
-				}
-			}
-			break;
 		default: break;
 	}

@@ -1420,8 +1335,6 @@ void CommandProcessor::WriteAtEndOfPipe64(uint32_t cache_policy, uint32_t event_
 }

 void CommandProcessor::EmitGlobalBarrier() {
-	CheckBuffer();
-
 	Common::LockGuard lock(m_renderer.GetMutex());

 	vk::MemoryBarrier2 barrier {};
@@ -1438,8 +1351,6 @@ void CommandProcessor::EmitGlobalBarrier() {
 }

 void CommandProcessor::TriggerEopEventAtEndOfPipe(uint32_t interrupt_context_id) {
-	CheckBuffer();
-
 	Sync::TriggerEopEventAtEndOfPipe(CurrentBuffer(), m_interrupt_event_id, interrupt_context_id);
 }

@@ -1523,8 +1434,6 @@ void CommandProcessor::TriggerEvent(uint32_t event_type, uint32_t event_index,
 }

 void CommandProcessor::Flip() {
-	CheckBuffer();
-
 	if (GraphicsRunDebugDumpEnabled()) {
 		LOGF("CommandProcessor::Flip()\n");
 	}
@@ -1538,7 +1447,7 @@ void CommandProcessor::Flip() {
 }

 void CommandProcessor::Flip(void* dst_gpu_addr, uint32_t value) {
-	CheckBuffer();
+	auto& command = CurrentBuffer();

 	if (GraphicsRunDebugDumpEnabled()) {
 		LOGF("CommandProcessor::Flip()\n"
@@ -1548,7 +1457,6 @@ void CommandProcessor::Flip(void* dst_gpu_addr, uint32_t value) {
 	}

 	std::memcpy(dst_gpu_addr, &value, sizeof(value));
-	auto& command = CurrentBuffer();
 	auto request = Sync::PrepareVideoOutFlip(command, m_flip.handle, m_flip.index, m_flip.flip_mode,
 	                                         m_flip.flip_arg);
 	Sync::WriteAtEndOfPipeWithFlip32(m_submit_id, command, static_cast<uint32_t*>(dst_gpu_addr),
@@ -1559,7 +1467,7 @@ void CommandProcessor::Flip(void* dst_gpu_addr, uint32_t value) {

 void CommandProcessor::FlipWithInterrupt(uint32_t eop_event_type, uint32_t cache_action,
                                         void* dst_gpu_addr, uint32_t value) {
-	CheckBuffer();
+	auto& command = CurrentBuffer();

 	if (GraphicsRunDebugDumpEnabled()) {
 		LOGF("CommandProcessor::FlipWithInterrupt()\n"
@@ -1574,7 +1482,6 @@ void CommandProcessor::FlipWithInterrupt(uint32_t eop_event_type, uint32_t cache
 		EXIT("unknown event type\n");
 	}
 	std::memcpy(dst_gpu_addr, &value, sizeof(value));
-	auto& command = CurrentBuffer();
 	auto request = Sync::PrepareVideoOutFlip(command, m_flip.handle, m_flip.index, m_flip.flip_mode,
 	                                         m_flip.flip_arg);
 	Sync::WriteAtEndOfPipeWithInterruptWriteBackFlip32(
@@ -1584,7 +1491,7 @@ void CommandProcessor::FlipWithInterrupt(uint32_t eop_event_type, uint32_t cache
 }

 void CommandProcessor::PrepareCpuFlip(uint64_t request_id) {
-	CheckBuffer();
+	auto& command = CurrentBuffer();
 	if (g_current_processor != nullptr) {
 		EXIT("invalid graphics-thread CPU flip preparation\n");
 	}
@@ -1594,7 +1501,7 @@ void CommandProcessor::PrepareCpuFlip(uint64_t request_id) {
 	};
 	ProcessorScope processor_scope(*this);

-	m_renderer.GetVideoOut().PrepareFlip(request_id, CurrentBuffer());
+	m_renderer.GetVideoOut().PrepareFlip(request_id, command);
 	GetScheduler().Flush();
 	m_renderer.GetVideoOut().CompleteFlip(request_id);
 }
@@ -910,6 +910,7 @@ constexpr uint32_t COMPUTE_PGM_RSRC1_WGP_MODE_MASK    = 0x1;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FWD_PROGRESS_SHIFT = 31;
 constexpr uint32_t COMPUTE_PGM_RSRC1_FWD_PROGRESS_MASK  = 0x1;

+constexpr uint32_t COMPUTE_DISPATCH_INITIATOR_USE_THREAD_DIMENSIONS = 1u << 5u;
 constexpr uint32_t COMPUTE_DISPATCH_INITIATOR_CS_W32_EN_SHIFT = 15;
 constexpr uint32_t COMPUTE_DISPATCH_INITIATOR_CS_W32_EN_MASK  = 0x1;

@@ -4,15 +4,12 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/profiler.h"
-#include "common/stringUtils.h"
 #include "graphics/guest_gpu/gpu_defs.h"
 #include "graphics/guest_gpu/gpu_format.h"

 #include <algorithm>
 #include <array>
 #include <bit>
-#include <fmt/format.h>
-#include <vector>

 namespace Libs::Graphics {

@@ -1322,13 +1319,8 @@ void TileGetTextureSize(Prospero::BufferFormat format, uint32_t width, uint32_t
 		return;
 	}
 	if (total_size != nullptr && total_size->size == 0) {
-		std::vector<std::string> list;
-		list.push_back(fmt::format("format = {}", static_cast<uint32_t>(format)));
-		list.push_back(fmt::format("width  = {}", width));
-		list.push_back(fmt::format("height = {}", height));
-		list.push_back(fmt::format("levels = {}", levels));
-		list.push_back(fmt::format("tile   = {}", static_cast<uint32_t>(tile)));
-		EXIT("unknown format:\n%s\n", Common::Concat(list, '\n').c_str());
+		EXIT("unknown format:\nformat = %u\nwidth  = %u\nheight = %u\nlevels = %u\ntile   = %u\n",
+		     static_cast<uint32_t>(format), width, height, levels, static_cast<uint32_t>(tile));
 	}
 }

@@ -16,6 +16,7 @@
 #include <windows.h>
 #undef min
 #undef max
+#undef MemoryBarrier
 #elif defined(__APPLE__)
 #include <pthread.h>
 #elif defined(__linux__)
@@ -129,7 +129,7 @@ bool BufferCache::DownloadBufferMemory(Buffer& buffer, uint64_t vaddr, uint64_t

 	const auto [mapped, offset] = m_download_buffer.Map(total_size, 64);
 	if (mapped == nullptr) {
-		EXIT("BufferCache: download exceeds 32 MiB staging buffer capacity\n");
+		EXIT("BufferCache: download exceeds 64 MiB staging buffer capacity\n");
 	}
 	m_download_buffer.Commit();
 	for (auto& copy: copies) {
@@ -183,7 +183,7 @@ BufferCache::BufferCache(GraphicContext& graphics, CommandScheduler& scheduler,
      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_download_buffer(graphics, scheduler, MemoryUsage::Download, 64 * MiB),
      m_device_buffer(graphics, scheduler, MemoryUsage::DeviceLocal, 128 * MiB),
      m_texture_cache(texture_cache) {
 	std::memset(m_gds_buffer.Mapped().data(), 0, static_cast<size_t>(m_gds_buffer.Size()));
@@ -107,45 +107,16 @@ constexpr uint64_t NumFramesBeforeRemoval = 32;
 	return true;
 }

-[[nodiscard]] const char* BindingTypeName(TextureCache::BindingType type) {
-	switch (type) {
-		case TextureCache::BindingType::Texture: return "Texture";
-		case TextureCache::BindingType::Storage: return "StorageTexture";
-		case TextureCache::BindingType::RenderTarget: return "ColorTarget";
-		case TextureCache::BindingType::DepthTarget: return "DepthTarget";
-		case TextureCache::BindingType::VideoOut: return "VideoOut";
-	}
-	return "Image";
-}
-
-void NameImageBinding(GraphicContext& graphics, Image& image, vk::ImageView view,
-                      TextureCache::BindingType type, const ImageViewInfo& view_info) {
-	const auto* role = BindingTypeName(type);
-	SetVulkanObjectNameF(
-	    graphics.device, image.backing.image,
-	    "Kyty.{}.Image[guest=0x{:016x} size=0x{:x} extent={}x{}x{} format={} mips={} layers={} "
-	    "samples={}]",
-	    role, image.info.data.address, image.info.data.size, image.info.extent.width,
-	    image.info.extent.height, image.info.extent.depth,
-	    static_cast<uint32_t>(image.info.pixel_format), image.info.resources.levels,
-	    image.info.resources.layers, image.info.samples);
-	SetVulkanObjectNameF(
-	    graphics.device, view,
-	    "Kyty.{}.View[guest=0x{:016x} format={} aspect=0x{:x} mip={}+{} layer={}+{}]", role,
-	    image.info.data.address, static_cast<uint32_t>(view_info.format),
-	    static_cast<vk::ImageAspectFlags::MaskType>(view_info.aspect), view_info.base_level,
-	    view_info.level_count, view_info.base_layer, view_info.layer_count);
-}
-
 [[nodiscard]] std::vector<vk::BufferImageCopy> BuildDepthCopies(const ImageInfo& info,
-                                                                uint64_t         slice_stride) {
+                                                              uint64_t slice_stride,
+                                                              vk::ImageAspectFlags aspect) {
 	std::vector<vk::BufferImageCopy> copies(info.resources.layers);
 	for (uint32_t layer = 0; layer < info.resources.layers; ++layer) {
 		auto& copy             = copies[layer];
 		copy.bufferOffset      = slice_stride * layer;
 		copy.bufferRowLength   = info.pitch;
 		copy.bufferImageHeight = info.extent.height;
-		copy.imageSubresource  = {vk::ImageAspectFlagBits::eDepth, 0, layer, 1};
+		copy.imageSubresource  = {aspect, 0, layer, 1};
 		copy.imageExtent       = {info.extent.width, info.extent.height, 1};
 	}
 	return copies;
@@ -1031,16 +1002,17 @@ TextureCache::ImageDownload TextureCache::BuildDownload(const Image& image) cons
 }

 void TextureCache::UploadImage(Image& image, Buffer& source, uint64_t source_offset) {
-	const auto& info    = image.info;
-	const auto  binding = UploadBinding(image);
+	auto& destination = image.depth_id ? m_slot_images[image.depth_id] : image;
+	const auto binding = image.depth_id ? BindingType::DepthTarget : UploadBinding(image);
 	const auto  upload  = [&](std::vector<vk::BufferImageCopy>& copies, TileManager::Result linear) {
 		for (auto& copy: copies) {
 			copy.bufferOffset += linear.offset;
 		}
-		image.Upload(copies, linear.buffer, linear.offset, linear.size);
+		destination.Upload(copies, linear.buffer, linear.offset, linear.size);
 	};

 	if (binding != BindingType::DepthTarget) {
+		const auto& info = image.info;
 		auto transfer = BuildTextureTransfer(image, binding, TransferDirection::Upload);
 		if (!transfer.valid) {
 			EXIT("TextureCache: invalid texture upload: binding=%u addr=0x%016" PRIx64
@@ -1064,21 +1036,32 @@ void TextureCache::UploadImage(Image& image, Buffer& source, uint64_t source_off
 		return;
 	}

-	if (info.samples != 1 || image.backing.samples != 1 ||
+	// The stencil plane has its own row pitch and shares the native image
+	// with the depth plane.
+	auto info = destination.info;
+	if (image.depth_id) {
+		info.data            = image.info.data;
+		info.guest_format    = Prospero::BufferFormat::k8UInt;
+		info.bytes_per_block = 1;
+		if (info.IsTiled()) info.pitch = TileGetDepthPitch(info.extent.width, 1, 0);
+	}
+	if (info.samples != 1 || destination.backing.samples != 1 ||
 	    info.resources.layers == 0 || info.data.size % info.resources.layers != 0 ||
 	    Prospero::NumBytesPerElement(info.guest_format) != info.bytes_per_block) {
 		EXIT("TextureCache: invalid depth upload\n");
 	}
 	const auto          layers          = info.resources.layers;
 	const auto          full_slice_size = info.data.size / layers;
-	auto                copies          = BuildDepthCopies(info, full_slice_size);
+	auto copies = BuildDepthCopies(info, full_slice_size, image.depth_id
+	                                                        ? vk::ImageAspectFlagBits::eStencil
+	                                                        : vk::ImageAspectFlagBits::eDepth);
 	TileManager::Result linear {source.Handle(), source_offset, source.Size() - source_offset};
 	if (info.IsTiled()) {
 		const auto tiles = BuildDepthTiles(info);
 		linear =
 		    m_tiler.Detile(source.Handle(), source_offset, info.data.size, info.data.size, tiles);
 	}
-	const auto transfer_bytes = DepthAspectTransferBytes(info.pixel_format);
+	const auto transfer_bytes = image.depth_id ? 1u : DepthAspectTransferBytes(info.pixel_format);
 	if (transfer_bytes != info.bytes_per_block) {
 		const uint64_t texels_per_slice = static_cast<uint64_t>(info.pitch) * info.extent.height;
 		EXIT_NOT_IMPLEMENTED(info.bytes_per_block != sizeof(uint16_t) ||
@@ -1197,13 +1180,20 @@ void TextureCache::MaterializeDccClear(ImageId id, const ImageDesc& desc,
 		}
 		// Native expanded keys own consumption. Existing buffer tracking publishes this CPU
 		// write to future GPU readers; FillBuffer can fault and must run outside the texture lock.
-		m_buffer_cache.FillBuffer(address, slice_size, UINT32_MAX, false);
+		if (desc.type != BindingType::VideoOut) {
+			m_buffer_cache.FillBuffer(address, slice_size, UINT32_MAX, false);
+		}
 	}
 }

 void TextureCache::RefreshImage(ImageId id) {
-	TrackImage(id);
 	auto& image = m_slot_images[id];
+	if (image.depth_id &&
+	    (m_slot_images[image.depth_id].info.metadata.stencil_compressed ||
+	     m_slot_images[image.depth_id].info.samples != 1)) {
+		return;
+	}
+	TrackImage(id);
 	if (image.IsMaybeCpuDirty()) {
 		const auto hash = image.HashGuestEdges();
 		if (image.NeedsMaybeCpuHash()) {
@@ -1225,7 +1215,7 @@ void TextureCache::RefreshImage(ImageId id) {
 	InitializeImage(id);
 }

-void TextureCache::AssociateStencil(ImageId depth_id, GuestRange stencil) {
+ImageId TextureCache::AssociateStencil(ImageId depth_id, GuestRange stencil) {
 	if (!stencil.Valid()) {
 		EXIT("TextureCache: invalid stencil association range\n");
 	}
@@ -1251,6 +1241,7 @@ void TextureCache::AssociateStencil(ImageId depth_id, GuestRange stencil) {
 	auto& record = m_slot_images[association];
 	TouchImage(record);
 	record.depth_id = depth_id;
+	return association;
 }

 ImageId TextureCache::FindImage(ImageDesc& desc, bool exact_format) {
@@ -1312,16 +1303,6 @@ ImageId TextureCache::FindImage(ImageDesc& desc, bool exact_format) {
 			}
 		}
 		auto& image = m_slot_images[result];
-		if (desc.type == BindingType::VideoOut &&
-		    desc.info.metadata.compression != VideoOutCompression::Uncompressed) {
-			const bool guest_dirty = image.IsBufferModified() || image.IsCpuDirty();
-			const bool native_current =
-			    (image.usage.render_target || image.IsGpuModified()) && !guest_dirty;
-			if (!native_current) {
-				EXIT("TextureCache: compressed video-out read requires clean native GPU "
-				     "contents\n");
-			}
-		}
 		if (view_mip >= 0) {
 			desc.view_info.base_level = static_cast<uint32_t>(view_mip);
 		}
@@ -1332,6 +1313,18 @@ ImageId TextureCache::FindImage(ImageDesc& desc, bool exact_format) {
 		TouchImage(image);
 	}
 	MaterializeDccClear(result, desc, metadata_base_layer);
+	if (desc.type == BindingType::VideoOut &&
+	    desc.info.metadata.compression != VideoOutCompression::Uncompressed) {
+		std::scoped_lock lock {m_lock};
+		const auto& image = m_slot_images[result];
+		const bool guest_dirty = image.IsBufferModified() || image.IsCpuDirty();
+		const bool native_current =
+		    (image.usage.render_target || image.IsGpuModified()) && !guest_dirty;
+		if (!native_current) {
+			EXIT("TextureCache: compressed video-out read requires clean native GPU "
+			     "contents\n");
+		}
+	}
 	return result;
 }

@@ -1396,6 +1389,19 @@ vk::ImageView TextureCache::FindTexture(ImageId id, const ImageDesc& desc) {
 	}
 	if (!image.info.data.Empty()) {
 		RefreshImage(id);
+		if (image.info.HasStencil() &&
+		    desc.info.data.address >= image.info.stencil.address &&
+		    desc.info.data.End() <= image.info.stencil.End()) {
+			for (const auto stencil_id:
+			     FindImagesInRegion(image.info.stencil.address, image.info.stencil.size, false)) {
+				const auto* stencil = m_slot_images.try_get(stencil_id);
+				if (stencil != nullptr && stencil->depth_id == id &&
+				    stencil->info.data == image.info.stencil) {
+					RefreshImage(stencil_id);
+					break;
+				}
+			}
+		}
 	}
 	switch (desc.type) {
 		case BindingType::Texture: break;
@@ -1407,9 +1413,7 @@ vk::ImageView TextureCache::FindTexture(ImageId id, const ImageDesc& desc) {
 			break;
 		default: EXIT("TextureCache: invalid texture binding\n");
 	}
-	const auto view = image.FindView(desc.view_info);
-	NameImageBinding(m_graphics, image, view, desc.type, desc.view_info);
-	return view;
+	return image.FindView(desc.view_info);
 }

 vk::ImageView TextureCache::FindRenderTarget(ImageId id, const ImageDesc& desc) {
@@ -1427,9 +1431,7 @@ vk::ImageView TextureCache::FindRenderTarget(ImageId id, const ImageDesc& desc)
 	RefreshImage(id);
 	CommitGpuWrite(image);
 	TrackImageDownload(id, image);
-	const auto view = image.FindView(desc.view_info);
-	NameImageBinding(m_graphics, image, view, desc.type, desc.view_info);
-	return view;
+	return image.FindView(desc.view_info);
 }

 vk::ImageView TextureCache::FindDepthTarget(ImageId id, const ImageDesc& desc) {
@@ -1444,20 +1446,19 @@ vk::ImageView TextureCache::FindDepthTarget(ImageId id, const ImageDesc& desc) {
 	TouchImage(image);
 	image.MarkGpuModified();
 	image.usage.depth_target = true;
-	RefreshImage(id);
+	image.info.stencil = desc.info.stencil;
+	image.info.metadata = desc.info.metadata;
 	if (desc.info.HasMetadata()) {
-		image.info.metadata = desc.info.metadata;
 		m_surface_metas.emplace(desc.info.metadata.range.address,
 		                        MetaDataInfo {.type       = MetaDataInfo::Type::HTile,
 		                                      .clear_mask = image.info.htile_clear_mask});
 	}
+	RefreshImage(id);
 	CommitGpuWrite(image);
 	if (desc.info.HasStencil()) {
-		AssociateStencil(id, desc.info.stencil);
+		RefreshImage(AssociateStencil(id, desc.info.stencil));
 	}
-	const auto view = image.FindView(desc.view_info);
-	NameImageBinding(m_graphics, image, view, desc.type, desc.view_info);
-	return view;
+	return image.FindView(desc.view_info);
 }

 void TextureCache::MarkGpuWritten(ImageId id) {
@@ -1468,11 +1469,16 @@ void TextureCache::MarkGpuWritten(ImageId id) {
 	}
 	TrackImage(id);
 	CommitGpuWrite(image);
+	if (image.info.HasStencil()) {
+		const auto stencil_id = AssociateStencil(id, image.info.stencil);
+		TrackImage(stencil_id);
+		CommitGpuWrite(m_slot_images[stencil_id]);
+	}
 }

 void TextureCache::CommitGpuWrite(Image& image) {
-	if (image.depth_id || image.backing.image == nullptr) {
-		EXIT("TextureCache: stencil association cannot own image contents\n");
+	if (!image.depth_id && image.backing.image == nullptr) {
+		EXIT("TextureCache: GPU writes require a native image or stencil association\n");
 	}
 	image.ClearBufferModified();
 	if (image.IsCpuDirty()) {
@@ -1556,9 +1562,10 @@ void TextureCache::ClearImage(CommandBuffer& command, ImageId id, vk::Format for
 	        range.layerCount == 0 || range.baseArrayLayer >= layers ||
 	        range.layerCount > layers - range.baseArrayLayer ||
 	        (range.aspectMask & aspects) != range.aspectMask);
-	const bool full_image = range.aspectMask == aspects && range.baseMipLevel == 0 &&
-	                        range.levelCount == image.info.resources.levels &&
-	                        range.baseArrayLayer == 0 && range.layerCount == layers;
+	const bool full_subresources = range.baseMipLevel == 0 &&
+	                               range.levelCount == image.info.resources.levels &&
+	                               range.baseArrayLayer == 0 && range.layerCount == layers;
+	const bool full_image = range.aspectMask == aspects && full_subresources;
 	TrackImage(id);
 	if (!full_image && (image.IsBufferModified() || image.IsCpuDirty())) {
 		InitializeImage(id);
@@ -1566,6 +1573,15 @@ void TextureCache::ClearImage(CommandBuffer& command, ImageId id, vk::Format for
 			EXIT("TextureCache: image clear retained guest ownership\n");
 		}
 	}
+	if (image.info.HasStencil() && (range.aspectMask & vk::ImageAspectFlagBits::eStencil)) {
+		const auto stencil_id = AssociateStencil(id, image.info.stencil);
+		if (!full_subresources) {
+			RefreshImage(stencil_id);
+		} else {
+			TrackImage(stencil_id);
+			CommitGpuWrite(m_slot_images[stencil_id]);
+		}
+	}
 	command.EndRendering();
 	// Transfer clears use the backing format; aliased clears must encode through their view.
 	if (format != image.backing.format || (image.info.IsVolume() && !full_image)) {
@@ -1638,7 +1654,7 @@ void TextureCache::DownloadDepth(Image& image, Buffer& destination, uint64_t des
 	EXIT_NOT_IMPLEMENTED(transfer_slice > UINT64_MAX / layers);
 	const uint64_t transfer_size = transfer_slice * layers;
 	EXIT_NOT_IMPLEMENTED(guest_slice > full_slice_size);
-	auto copies = BuildDepthCopies(info, full_slice_size);
+	auto copies = BuildDepthCopies(info, full_slice_size, vk::ImageAspectFlagBits::eDepth);
 	if (transfer_bytes == info.bytes_per_block) {
 		if (!info.IsTiled()) {
 			for (auto& copy: copies) {
@@ -1831,7 +1847,7 @@ void TextureCache::InvalidateMemoryFromGPU(uint64_t address, uint64_t size) {
 	std::scoped_lock lock {m_lock};
 	for (const auto id: FindImagesInRegion(address, size, true)) {
 		auto& image = m_slot_images[id];
-		if (image.depth_id || !image.Overlaps(address, size)) {
+		if (!image.Overlaps(address, size)) {
 			continue;
 		}
 		if (image.IsGpuModified()) {
@@ -157,7 +157,7 @@ private:
 	void RefreshCopySource(ImageId id);
 	[[nodiscard]] bool CopyD16(Image& destination, Image& source);
 	void               CopyImage(ImageId destination, ImageId source);
-	void               AssociateStencil(ImageId depth, GuestRange stencil);
+	[[nodiscard]] ImageId AssociateStencil(ImageId depth, GuestRange stencil);
 	void CopyImageMip(ImageId destination, ImageId source, uint32_t mip, uint32_t layer);
 	void ValidateImageDesc(const ImageDesc& desc) const;

@@ -209,17 +209,16 @@ static void RtCheck(const HW::RenderTarget& rt) {
 		}
 		if (rt.info.fmask_compression_enable) {
 			EXIT_NOT_IMPLEMENTED(rt.attrib.num_samples == 0 && rt.attrib.num_fragments == 0);
+			// Native MSAA stores expanded samples, independent of FMASK metadata compression.
 			static bool logged = false;
 			if (!logged) {
-				LOGF("RenderTarget: using native Vulkan MSAA without guest FMASK metadata, "
+				LOGF("RenderTarget: using expanded native Vulkan MSAA samples, "
 				     "fmask=0x%016" PRIx64 "\n",
 				     rt.fmask.addr);
 				logged = true;
 			}
 		}

-		EXIT_NOT_IMPLEMENTED(rt.info.fmask_data_compression_disable != false);
-
 		if (rt.info.cmask_fast_clear_enable || rt.info.dcc_compression_enable) {
 			static bool logged = false;
 			if (!logged) {
@@ -41,13 +41,42 @@ struct RenderDepthInfo {
 	[[nodiscard]] vk::ImageAspectFlags AttachmentWriteAspects() const;
 };

+inline vk::ImageAspectFlags DepthFeedbackAspects(vk::ImageAspectFlags draw_writes,
+                                                 const ImageViewInfo& target,
+                                                 const ImageViewInfo& sampled) {
+	if (!ImageRangeOverlaps(target.base_level, target.level_count, sampled.base_level,
+	                        sampled.level_count) ||
+	    !ImageRangeOverlaps(target.base_layer, target.layer_count, sampled.base_layer,
+	                        sampled.layer_count)) {
+		return {};
+	}
+	return draw_writes & sampled.aspect;
+}
+
+inline vk::ImageAspectFlags DepthReadableAspects(vk::ImageLayout layout) {
+	switch (layout) {
+		case vk::ImageLayout::eDepthReadOnlyOptimal:
+		case vk::ImageLayout::eDepthReadOnlyStencilAttachmentOptimal:
+			return vk::ImageAspectFlagBits::eDepth;
+		case vk::ImageLayout::eStencilReadOnlyOptimal:
+		case vk::ImageLayout::eDepthAttachmentStencilReadOnlyOptimal:
+			return vk::ImageAspectFlagBits::eStencil;
+		case vk::ImageLayout::eDepthStencilReadOnlyOptimal:
+		case vk::ImageLayout::eAttachmentFeedbackLoopOptimalEXT:
+		case vk::ImageLayout::eGeneral:
+			return vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil;
+		default:
+			return {};
+	}
+}
+
 inline vk::ImageLayout depth_attachment_layout(const RenderDepthInfo& depth) {
 	const auto available     = ImageViewOps::DepthAspectMask(depth.desc.view_info.format);
 	const auto writes        = depth.AttachmentWriteAspects();
 	const bool has_depth     = static_cast<bool>(available & vk::ImageAspectFlagBits::eDepth);
 	const bool has_stencil   = static_cast<bool>(available & vk::ImageAspectFlagBits::eStencil);
-	// LoadOp clears are separate from guest depth writes.
-	const bool depth_write   = depth.depth_write_enable;
+	// The attachment layout must permit load clears as well as draw writes.
+	const bool depth_write   = static_cast<bool>(writes & vk::ImageAspectFlagBits::eDepth);
 	const bool stencil_write = static_cast<bool>(writes & vk::ImageAspectFlagBits::eStencil);
 	if (!has_stencil) {
 		return depth_write ? vk::ImageLayout::eDepthAttachmentOptimal
@@ -690,6 +690,12 @@ Image::Image(GraphicContext& graphics, CommandScheduler& scheduler, const ImageI
 		     create.extent.width, create.extent.height, create.extent.depth,
 		     static_cast<int>(create.format), create.arrayLayers, create.mipLevels);
 	}
+	SetVulkanObjectNameF(
+	    graphics.device, backing.image,
+	    "Kyty.Image[guest=0x{:016x} size=0x{:x} extent={}x{}x{} format={} mips={} layers={} samples={}]",
+	    info.data.address, info.data.size, info.extent.width, info.extent.height, info.extent.depth,
+	    static_cast<uint32_t>(info.pixel_format), info.resources.levels, info.resources.layers,
+	    info.samples);
 }

 uint64_t Image::HashGuestEdges() const {
@@ -329,7 +329,9 @@ struct VideoOutFormatPolicy {
 	VideoOutPixelFormatInfo info;
 };

-inline constexpr std::array<VideoOutFormatPolicy, 6> VIDEO_OUT_FORMAT_POLICIES {{
+inline constexpr uint64_t VIDEO_OUT_PIXEL_FORMAT_R10_G10_B10_A2_BT2100_PQ = 0x8100070422000000ull;
+
+inline constexpr std::array<VideoOutFormatPolicy, 7> VIDEO_OUT_FORMAT_POLICIES {{
    {0x8000000022000000ull,
     {vk::Format::eR8G8B8A8Srgb, Prospero::BufferFormat::k8_8_8_8Srgb, 4, false}},
    {0x8000000000000000ull,
@@ -338,6 +340,8 @@ inline constexpr std::array<VideoOutFormatPolicy, 6> VIDEO_OUT_FORMAT_POLICIES {
     {vk::Format::eA2B10G10R10UnormPack32, Prospero::BufferFormat::k10_10_10_2UNorm, 4, false}},
    {0x8100000000000000ull,
     {vk::Format::eA2R10G10B10UnormPack32, Prospero::BufferFormat::k10_10_10_2UNorm, 4, false}},
+    {VIDEO_OUT_PIXEL_FORMAT_R10_G10_B10_A2_BT2100_PQ,
+     {vk::Format::eA2B10G10R10UnormPack32, Prospero::BufferFormat::k10_10_10_2UNorm, 4, false}},
    {0xc001000622000000ull,
     {vk::Format::eR16G16B16A16Sfloat, Prospero::BufferFormat::k16_16_16_16Float, 8, false}},
    {0xc001000600000000ull,
@@ -391,6 +391,12 @@ vk::ImageView Image::FindView(const ImageViewInfo& view_info) {
 		     view_info.level_count, view_info.base_layer, view_info.layer_count,
 		     static_cast<vk::ImageUsageFlags::MaskType>(view_info.usage));
 	}
+	SetVulkanObjectNameF(
+	    m_graphics.device, view,
+	    "Kyty.ImageView[guest=0x{:016x} format={} aspect=0x{:x} mip={}+{} layer={}+{}]",
+	    info.data.address, static_cast<uint32_t>(normalized.format),
+	    static_cast<vk::ImageAspectFlags::MaskType>(normalized.aspect), normalized.base_level,
+	    normalized.level_count, normalized.base_layer, normalized.layer_count);
 	views.push_back({normalized, view});
 	return view;
 }
@@ -17,6 +17,7 @@
 #include "graphics/host_gpu/hostMemory.h"
 #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/image/imageView.h"
 #include "graphics/host_gpu/renderer/image/textureCommon.h"
 #include "graphics/host_gpu/renderer/pipeline/shaderResourceBarrier.h"
@@ -184,17 +185,18 @@ bool IsSupportedDepthTextureEncoding(const ShaderTextureResource& descriptor, bo
 	}
 	const bool full = common && descriptor.fields[4] == field4_expected &&
 	                  descriptor.fields[5] == field5_expected;
-	if (!full || (descriptor.fields[6] == 0 && descriptor.fields[7] != 0) ||
+	if (!full ||
 	    (descriptor.MsaaDepth() && !IsMultisampledTexture(descriptor.Type()))) {
 		return false;
 	}
-	if (descriptor.fields[6] == 0) {
+	const auto metadata_control = descriptor.fields[6] & 0x00ffffffu;
+	if (metadata_control == 0) {
 		return true;
 	}
 	constexpr uint32_t htile_control = 0x00280000u;
 	const uint32_t expected_control  = htile_control | (descriptor.MsaaDepth() ? (1u << 10u) : 0u);
 	const auto     metadata_addr     = descriptor.MetaAddr() << 8u;
-	return (descriptor.fields[6] & 0x00ffffffu) == expected_control && metadata_addr != 0 &&
+	return metadata_control == expected_control && metadata_addr != 0 &&
 	       metadata_addr < TRACKER_ADDRESS_SIZE && (metadata_addr & 0x7fffu) == 0 &&
 	       descriptor.TileMode() == Prospero::TileMode::kDepth;
 }
@@ -252,10 +254,10 @@ static bool IsSupportedStorageTextureDescriptor(const ShaderRecompiler::IR::Imag
 	    (is_color_2d_array && descriptor.BaseArray5() <= descriptor.Depth());
 	const bool is_2d =
 	    resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2D && valid_2d_slice;
+	// Storage cube coordinates address individual faces, including partial cube views.
 	const bool is_cube = resource.cube && descriptor.Type() == Prospero::ImageType::kCube &&
 	                     descriptor.Width5() == descriptor.Height5() &&
-	                     descriptor.BaseArray5() <= descriptor.Depth() &&
-	                     (descriptor.Depth() - descriptor.BaseArray5() + 1u) % 6u == 0;
+	                     descriptor.BaseArray5() <= descriptor.Depth();
 	const bool is_2d_array =
 	    resource.dimension == ShaderRecompiler::Decoder::ImageDimension::Dim2DArray &&
 	    ((!resource.cube && is_color_2d_array && descriptor.BaseArray5() <= descriptor.Depth()) ||
@@ -550,8 +552,14 @@ TextureBinding RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageR
 	const bool multisampled    = IsMultisampledTexture(type);
 	const auto max_mip         = resource.r128 ? last_level : descriptor.MaxMip();
 	const auto physical_levels = multisampled ? 1u : static_cast<uint32_t>(max_mip) + 1u;
+	// IMAGE_STORE addresses BASE_LEVEL; only IMAGE_STORE_MIP selects other view mips.
+	const bool single_storage_mip =
+	    storage && resource.mip_mode != ShaderRecompiler::IR::ImageMipMode::DynamicStorage;
+	const auto view_levels = multisampled || single_storage_mip
+	                             ? 1u
+	                             : static_cast<uint32_t>(last_level - base_level) + 1u;
 	const auto levels =
-	    multisampled ? 1u : std::max(physical_levels, static_cast<uint32_t>(last_level) + 1u);
+	    multisampled ? 1u : std::max(physical_levels, base_level + view_levels);
 	const auto tile       = descriptor.TileMode();
 	const bool depth_tile = tile == Prospero::TileMode::kDepth;
 	const bool msaa_tile  = depth_tile || tile == Prospero::TileMode::kRenderTarget;
@@ -574,8 +582,6 @@ TextureBinding RenderExecutor::ResolveTexture(const ShaderRecompiler::IR::ImageR
 		     descriptor.fields[6], descriptor.fields[7]);
 	}
 	const auto samples = multisampled ? 1u << last_level : 1u;
-	const auto view_levels =
-	    multisampled ? 1u : static_cast<uint32_t>(last_level - base_level) + 1u;
 	const auto depth          = static_cast<uint32_t>(descriptor.Depth()) + 1u;
 	const auto format         = descriptor.Format();
 	const auto surface_format = TextureGetSurfaceFormatInfo(format);
@@ -744,18 +750,27 @@ void RenderExecutor::ResetBindings() {
 	m_bound_images.clear();
 }

-PreparedBindings RenderExecutor::PrepareBindings(const ShaderStageRuntime& runtime) {
+void RenderExecutor::PrepareBindings(const ShaderStageRuntime& runtime,
+                                     PreparedBindings& prepared) {
 	KYTY_PROFILER_FUNCTION();
 	EXIT_IF(!runtime);
 	const auto& program  = *runtime.program;
-	const auto& snapshot = runtime.resources;
-	PreparedBindings prepared;
+	const auto& snapshot = *runtime.resources;
 	prepared.runtime = &runtime;
-	prepared.images.reserve(program.info.images.size());
+	prepared.gds = {nullptr, 0, VK_WHOLE_SIZE};
+	prepared.flattened_srt = {};
+	prepared.shader_data_buffer = {};
+	prepared.buffer_sources.clear();
+	prepared.buffers.clear();
+	prepared.images.resize(program.info.images.size());
+	prepared.samplers.clear();
+	prepared.shader_data.clear();
 	for (uint32_t i = 0; i < program.info.images.size(); i++) {
 		auto binding = ResolveTexture(program.info.images[i], snapshot.images[i]);
 		BindImage(binding.image_id, binding.desc.type == TextureCache::BindingType::Storage);
-		prepared.images.push_back(std::move(binding));
+		binding.mip_views.swap(prepared.images[i].mip_views);
+		binding.mip_views.clear();
+		prepared.images[i] = std::move(binding);
 	}
 	prepared.samplers.reserve(program.info.samplers.size());
 	for (uint32_t i = 0; i < program.info.samplers.size(); i++) {
@@ -770,14 +785,13 @@ PreparedBindings RenderExecutor::PrepareBindings(const ShaderStageRuntime& runti
 	        program.bindings, ShaderRecompiler::IR::DescriptorBindingKind::Gds) != nullptr) {
 		prepared.gds.buffer = m_context.GetBufferCache().GetGdsBuffer()->Handle();
 	}
-	return prepared;
 }

 void RenderExecutor::FindBuffers(PreparedBindings& prepared) {
 	KYTY_PROFILER_FUNCTION();
 	EXIT_IF(prepared.runtime == nullptr || !*prepared.runtime);
 	const auto& program  = *prepared.runtime->program;
-	const auto& snapshot = prepared.runtime->resources;
+	const auto& snapshot = *prepared.runtime->resources;
 	auto&       cache    = m_context.GetBufferCache();

 	prepared.buffer_sources.clear();
@@ -799,7 +813,7 @@ void RenderExecutor::RebindBuffers(PreparedBindings& prepared) {
 	KYTY_PROFILER_FUNCTION();
 	EXIT_IF(prepared.runtime == nullptr || !*prepared.runtime);
 	const auto& program   = *prepared.runtime->program;
-	const auto& snapshot  = prepared.runtime->resources;
+	const auto& snapshot  = *prepared.runtime->resources;
 	const auto& layout    = program.bindings;
 	EXIT_IF(prepared.buffer_sources.size() != program.info.buffers.size());

@@ -834,7 +848,7 @@ void RenderExecutor::RebindImages(PreparedBindings& prepared) {
 	KYTY_PROFILER_FUNCTION();
 	EXIT_IF(prepared.runtime == nullptr || !*prepared.runtime);
 	const auto& program  = *prepared.runtime->program;
-	const auto& snapshot = prepared.runtime->resources;
+	const auto& snapshot = *prepared.runtime->resources;
 	auto&       images   = prepared.images;
 	EXIT_IF(images.size() != program.info.images.size());
 	auto& texture_cache = m_context.GetTextureCache();
@@ -866,11 +880,7 @@ void RenderExecutor::RebindImages(PreparedBindings& prepared) {
 			}
 			binding.image_view = binding.mip_views.front();
 		} else {
-			auto desc = binding.desc;
-			if (desc.type == TextureCache::BindingType::Storage) {
-				desc.view_info.level_count = 1;
-			}
-			binding.image_view = texture_cache.FindTexture(binding.image_id, desc);
+			binding.image_view = texture_cache.FindTexture(binding.image_id, binding.desc);
 		}
 		auto&      image   = texture_cache.GetImage(binding.image_id);
 		const bool storage = binding.desc.type == TextureCache::BindingType::Storage;
@@ -988,19 +998,7 @@ void RenderExecutor::CommitBindings(CommandBuffer&                     buffer,
 					    std::ranges::find(image.views, binding.image_view, &CachedImageView::view);
 					EXIT_IF(storage || host_view == image.views.end());
 					const auto aspect = host_view->info.aspect;
-					const bool depth_feedback =
-					    layout == vk::ImageLayout::eAttachmentFeedbackLoopOptimalEXT &&
-					    program.stage == ShaderType::Pixel;
-					const bool depth_read =
-					    depth_feedback || layout == vk::ImageLayout::eDepthReadOnlyOptimal ||
-					    layout == vk::ImageLayout::eDepthStencilReadOnlyOptimal ||
-					    layout == vk::ImageLayout::eDepthReadOnlyStencilAttachmentOptimal;
-					const bool stencil_read =
-					    layout == vk::ImageLayout::eStencilReadOnlyOptimal ||
-					    layout == vk::ImageLayout::eDepthStencilReadOnlyOptimal ||
-					    layout == vk::ImageLayout::eDepthAttachmentStencilReadOnlyOptimal;
-					if ((aspect & vk::ImageAspectFlagBits::eDepth && !depth_read) ||
-					    (aspect & vk::ImageAspectFlagBits::eStencil && !stencil_read)) {
+					if (aspect & ~DepthReadableAspects(layout)) {
 						EXIT("sampling a writable depth/stencil attachment aspect\n");
 					}
 				}
@@ -92,9 +92,9 @@ void PipelineCacheLog(fmt::format_string<Args...> format, Args&&... args) {
 	Log::WriteToConsoleAndLog(message);
 }

-bool ReadShaderGuestMemory(void*, uint64_t address, uint32_t* value) {
-	return value != nullptr &&
-	       Libs::LibKernel::Memory::TryReadGpuCleanBacking(address, value, sizeof(*value));
+bool ReadShaderGuestMemory(void*, uint64_t address, std::span<uint32_t> values) {
+	return !values.empty() &&
+	       Libs::LibKernel::Memory::TryReadGpuCleanBacking(address, values.data(), values.size_bytes());
 }

 void DumpShaderSpirv(const char* stage_name, uint64_t shader_hash,
@@ -200,8 +200,11 @@ struct PipelineCache::ProgramCache {
 			permutations.reserve(8);
 		}

-		ShaderRecompiler::IR::ResourcePlan resource_plan;
-		std::vector<Permutation>           permutations;
+		ShaderRecompiler::IR::ResourcePlan           resource_plan;
+		ShaderRecompiler::IR::ResourceSnapshot       resources;
+		ShaderRecompiler::IR::ResourceSpecialization specialization;
+		std::vector<Permutation>                    permutations;
+		bool                                        skip_dispatch = false;
 	};

 	struct ProgramKeyHash {
@@ -274,33 +277,36 @@ struct PipelineCache::ProgramCache {
 			stage = ShaderType::Compute;
 		}

+		const auto user_data = std::span(params.user_data).first(params.user_data_count);
 		lookup_key.stage           = stage;
 		lookup_key.hash            = params.hash;
-		lookup_key.user_data_count = static_cast<uint32_t>(params.user_data.size());
+		lookup_key.user_data_count = params.user_data_count;
 		lookup_key.code_size       = static_cast<uint32_t>(params.code.size());
 		BuildStageStaticKey(input_info, lookup_key.static_state);
 		auto                                         entry = programs.find(lookup_key);
-		ShaderRecompiler::IR::ResourceSnapshot       resources;
-		ShaderRecompiler::IR::ResourceSpecialization specialization;
+		if (entry != programs.end() && entry->second.skip_dispatch) {
+			return {};
+		}
 		const ShaderRecompiler::IR::SrtRuntime       runtime {
-		    .user_data                  = params.user_data,
+		    .user_data                  = user_data,
 		    .shader_base                = params.Base(),
 		    .read_specialization_memory = ReadShaderGuestMemory,
 		};
 		if (entry != programs.end()) {
 			EXIT_IF(!ShaderRecompiler::IR::MaterializeResources(
-			    entry->second.resource_plan, runtime, resources, specialization));
+			    entry->second.resource_plan, runtime, entry->second.resources,
+			    entry->second.specialization));
 			if (const auto permutation = std::ranges::find_if(
 			        entry->second.permutations, [&](const Permutation& candidate) {
 				        const auto& layout = candidate.program.bindings;
 				        return layout.push_data_start_dword ==
 				                   ShaderRecompiler::IR::PushData::StartFor(
 				                       push_data_cursor, layout.ShaderDataDwords()) &&
-				               candidate.specialization == specialization;
+				               candidate.specialization == entry->second.specialization;
 			        });
 			    permutation != entry->second.permutations.end()) {
 				input_info.stage = {.program   = &permutation->program,
-				                    .resources = std::move(resources)};
+				                    .resources = &entry->second.resources};
 				permutation->program.bindings.AdvancePushData(push_data_cursor);
 				return permutation->handle;
 			}
@@ -328,7 +334,7 @@ struct PipelineCache::ProgramCache {
 		ShaderRecompiler::CompileOptions options;
 		options.stage       = stage;
 		options.shader_hash = params.hash;
-		options.user_data   = params.user_data;
+		options.user_data   = user_data;
 		options.back_code      = params.back_code;
 		options.dump_ir     = Config::GetShaderLogDirection() != Config::LogDirection::Silent;
 		options.early_dump  = options.dump_ir;
@@ -337,6 +343,7 @@ struct PipelineCache::ProgramCache {

 		if constexpr (std::is_same_v<InputInfo, ShaderVertexInputInfo>) {
 			options.user_data_base = 8;
+			options.wave_size = input_info.wave_size;
 			if (stage == ShaderType::Mesh || stage == ShaderType::TessellationControl) {
 				options.user_data_base = 0;
 				options.wave_size = stage == ShaderType::Mesh ? input_info.mesh.wave_size : 64u;
@@ -345,16 +352,22 @@ struct PipelineCache::ProgramCache {
 			options.wave_size = input_info.wave_size;
 		}
 		auto translated = ShaderRecompiler::TranslateProgram(params.code, options);
+		if (translated.skip_dispatch) {
+			entry = programs.try_emplace(lookup_key, ShaderRecompiler::IR::ResourcePlan {}).first;
+			entry->second.skip_dispatch = true;
+			return {};
+		}
 		if (entry == programs.end()) {
-			auto resource_plan = ShaderRecompiler::IR::ExtractResourcePlan(translated.program);
-			EXIT_IF(!ShaderRecompiler::IR::MaterializeResources(resource_plan, runtime, resources,
-			                                                    specialization));
-			entry = programs.try_emplace(lookup_key, std::move(resource_plan)).first;
+			entry = programs.try_emplace(lookup_key,
+			    ShaderRecompiler::IR::ExtractResourcePlan(translated.program)).first;
+			EXIT_IF(!ShaderRecompiler::IR::MaterializeResources(
+			    entry->second.resource_plan, runtime, entry->second.resources,
+			    entry->second.specialization));
 		}
 		entry->second.permutations.push_back(CompilePermutation(
-		    params, options, std::move(translated), std::move(specialization), push_data_cursor));
+		    params, options, std::move(translated), entry->second.specialization, push_data_cursor));
 		const auto& permutation = entry->second.permutations.back();
-		input_info.stage = {.program = &permutation.program, .resources = std::move(resources)};
+		input_info.stage = {.program = &permutation.program, .resources = &entry->second.resources};
 		permutation.program.bindings.AdvancePushData(push_data_cursor);

 		std::array<size_t, static_cast<size_t>(ShaderType::TessellationEvaluation) + 1> counts {};
@@ -593,6 +606,21 @@ PipelineCache::GraphicsPrograms PipelineCache::GetGraphicsPrograms(
 	ShaderParams pixel_params;
 	if (pixel_active) {
 		pixel_params = PrepareProgram(pixel_regs, sh, target_export_mapping, pixel_info);
+		const auto& blend          = context.GetBlendControl(0);
+		const auto  is_dual_source = [](uint8_t factor) {
+			return factor >= static_cast<uint8_t>(Prospero::BlendFactor::kSrc1Color) &&
+			       factor <= static_cast<uint8_t>(Prospero::BlendFactor::kOneMinusSrc1Alpha);
+		};
+		pixel_info.dual_source_blending =
+		    blend.enable && !context.GetRenderTarget(0).info.blend_bypass &&
+		    (is_dual_source(blend.color_srcblend) || is_dual_source(blend.color_destblend) ||
+		     (blend.separate_alpha_blend &&
+		      (is_dual_source(blend.alpha_srcblend) || is_dual_source(blend.alpha_destblend))));
+		if (pixel_info.dual_source_blending) {
+			// MRT1 supplies a second blend source for the same render target as MRT0.
+			pixel_info.target_output_mode[1]    = pixel_info.target_output_mode[0];
+			pixel_info.target_export_mapping[1] = pixel_info.target_export_mapping[0];
+		}
 	}
 	if (context.GetClipControl().clip_disable) {
 		const auto& viewport = context.GetScreenViewport().viewports[0];
@@ -740,8 +768,7 @@ PipelineCache::Pipeline& PipelineCache::GetGraphicsPipeline(
 	static_params.depth_bounds_test_enable = depth.depth_bounds_test_enable;
 	static_params.depth_min_bounds         = depth.depth_min_bounds;
 	static_params.depth_max_bounds         = depth.depth_max_bounds;
-	const bool rect_list =
-	    command.GetUserConfig().GetPrimType() == Prospero::PrimitiveType::kRectList;
+	const bool rect_list = Prospero::IsRectList(command.GetUserConfig().GetPrimType());
 	static_params.cull_back  = !rect_list && mc.cull_back;
 	static_params.cull_front = !rect_list && mc.cull_front;
 	static_params.face       = mc.face;
@@ -46,8 +46,8 @@ vk::PipelineStageFlags ShaderPipelineStages(vk::ShaderStageFlags stages) {
 	return result;
 }

-VulkanMemoryBarrier MakeShaderWriteDependency() {
-	VulkanMemoryBarrier barrier {};
+vk::MemoryBarrier MakeShaderWriteDependency() {
+	vk::MemoryBarrier barrier {};
 	barrier.srcAccessMask = vk::AccessFlagBits::eShaderWrite;
 	barrier.dstAccessMask = vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite |
 	                        vk::AccessFlagBits::eVertexAttributeRead |
@@ -58,15 +58,15 @@ VulkanMemoryBarrier MakeShaderWriteDependency() {
 	return barrier;
 }

-VulkanMemoryBarrier MakeShaderAccessDependency() {
-	VulkanMemoryBarrier barrier {};
+vk::MemoryBarrier MakeShaderAccessDependency() {
+	vk::MemoryBarrier barrier {};
 	barrier.srcAccessMask = vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite;
 	barrier.dstAccessMask = vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite;
 	return barrier;
 }

-VulkanMemoryBarrier MakeShaderWriteHazardDependency() {
-	VulkanMemoryBarrier barrier {};
+vk::MemoryBarrier MakeShaderWriteHazardDependency() {
+	vk::MemoryBarrier barrier {};
 	barrier.srcAccessMask = vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite;
 	barrier.dstAccessMask = vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eShaderWrite;
 	return barrier;
@@ -90,7 +90,7 @@ vk::BufferMemoryBarrier MakeGdsDependency(vk::Buffer buffer) {
 bool HasShaderBufferWrites(const ShaderStageRuntime& runtime) {
 	EXIT_IF(!runtime);
 	const auto& program   = *runtime.program;
-	const auto& resources = runtime.resources;
+	const auto& resources = *runtime.resources;
 	EXIT_IF(resources.buffers.size() != program.info.buffers.size());
 	bool has_writes = false;
 	for (uint32_t i = 0; i < program.info.buffers.size(); i++) {
@@ -10,9 +10,9 @@ struct ShaderStageRuntime;

 vk::ShaderStageFlagBits NativeShaderStage(ShaderType stage);
 vk::PipelineStageFlags  ShaderPipelineStages(vk::ShaderStageFlags stages);
-VulkanMemoryBarrier     MakeShaderAccessDependency();
-VulkanMemoryBarrier     MakeShaderWriteHazardDependency();
-VulkanMemoryBarrier     MakeShaderWriteDependency();
+vk::MemoryBarrier       MakeShaderAccessDependency();
+vk::MemoryBarrier       MakeShaderWriteHazardDependency();
+vk::MemoryBarrier       MakeShaderWriteDependency();
 vk::BufferMemoryBarrier MakeGdsDependency(vk::Buffer buffer);
 bool HasShaderBufferWrites(const ShaderStageRuntime& runtime);
 void ShaderAccessBarrier(vk::CommandBuffer vk_buffer, vk::PipelineStageFlags source_stages);
@@ -368,6 +368,8 @@ void CreatePipelineInternal(GraphicContext& graphics, PipelineCache::Pipeline& p
 	// 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__)
+	// The DB clamps depth to the viewport range after polygon offset is applied.
+	rasterizer.depthClampEnable = VK_TRUE;
 	rasterizer.pNext = &clip_ext;
 #endif
 	vk::PipelineRasterizationProvokingVertexStateCreateInfoEXT provoking_vertex {};
@@ -45,6 +45,7 @@ enum class CommandBufferDebugOp : uint32_t {
 	EopFlip,
 	EopWriteBackFlip,
 	EopOnlyFlip,
+	DispatchIndirect,
 	Unknown,
 };

@@ -156,8 +157,10 @@ public:

 	void DispatchDirect(uint64_t submit_id, CommandBuffer& buffer, uint32_t thread_group_x,
 	                    uint32_t thread_group_y, uint32_t thread_group_z, uint32_t mode);
+	void DispatchIndirect(uint64_t submit_id, CommandBuffer& buffer, uint64_t args_addr,
+	                      uint32_t mode);

-	[[nodiscard]] PreparedBindings PrepareBindings(const ShaderStageRuntime& runtime);
+	void PrepareBindings(const ShaderStageRuntime& runtime, PreparedBindings& prepared);
 	void                           FindBuffers(PreparedBindings& bindings);
 	void                           RebindBuffers(PreparedBindings& bindings);
 	void                           RebindImages(PreparedBindings& bindings);
@@ -193,7 +196,8 @@ private:
 	                         bool primitive_restart_enable);
 	[[nodiscard]] RenderState AcquireRenderTargets(CommandBuffer& buffer, RenderColorInfo* colors,
 	                                               uint32_t color_count, RenderDepthInfo& depth,
-	                                               const std::optional<PreparedBindings>& pixel = std::nullopt);
+	                                               vk::ImageAspectFlags& feedback_aspects,
+	                                               std::span<PreparedBindings* const> stages = {});
 	[[nodiscard]] bool        ResolveColorTargets(CommandBuffer& buffer,
 	                                              uint32_t render_target_slice_offset);
 	void                      BindImage(ImageId id, bool storage);
@@ -206,6 +210,8 @@ private:
 	                                              uint32_t group_y, uint32_t group_z, uint32_t mode);

 	RenderContext&                        m_context;
+	GraphicsBindings                     m_graphics_bindings;
+	PreparedBindings                     m_compute_bindings;
 	std::vector<ImageId>                  m_bound_images;
 	std::vector<vk::DescriptorBufferInfo> m_descriptor_buffers;
 	std::vector<vk::DescriptorImageInfo>  m_descriptor_images;
@@ -9,6 +9,7 @@
 #include "graphics/guest_gpu/gpu_defs.h"
 #include "graphics/guest_gpu/graphicsRun.h"
 #include "graphics/guest_gpu/hardwareContext.h"
+#include "graphics/guest_gpu/pm4.h"
 #include "graphics/host_gpu/graphicContext.h"
 #include "graphics/host_gpu/renderer/image/imageInfo.h"
 #include "graphics/host_gpu/renderer/pipeline/descriptors.h"
@@ -51,7 +52,7 @@ static bool FillSourcesDisjoint(std::span<const ShaderRecompiler::IR::Descriptor
 bool RenderExecutor::TryConsumeComputeMetaClear(const ShaderComputeInputInfo& input,
                                                const CommandBuffer&          buffer) {
 	const auto& program   = *input.stage.program;
-	const auto& resources = input.stage.resources;
+	const auto& resources = *input.stage.resources;
 	if (resources.buffers.size() != program.info.buffers.size()) {
 		EXIT("compute runtime buffer count does not match shader metadata\n");
 	}
@@ -85,7 +86,7 @@ bool ResolveComputeBufferFill(const ShaderComputeInputInfo& input, uint32_t grou
                              uint32_t group_y, uint32_t group_z, uint32_t mode,
                              ShaderBufferResource& resolved_descriptor, uint32_t& resolved_clear,
                              uint64_t& resolved_size) {
-	const auto& resources = input.stage.resources;
+	const auto& resources = *input.stage.resources;
 	const auto& fill      = resources.uniform_fill;
 	if (fill.kind != ShaderRecompiler::IR::UniformFillKind::Buffer) {
 		return false;
@@ -128,7 +129,7 @@ bool RenderExecutor::TryConsumeComputeImageClear(const ShaderComputeInputInfo& i
                                                CommandBuffer& command, uint32_t group_x,
                                                uint32_t group_y, uint32_t group_z, uint32_t mode) {
 	const auto& program   = *input.stage.program;
-	const auto& resources = input.stage.resources;
+	const auto& resources = *input.stage.resources;
 	const auto& fill      = resources.uniform_fill;
 	auto&       cache     = command.GetContext().GetTextureCache();
 	if (fill.kind == ShaderRecompiler::IR::UniformFillKind::Image) {
@@ -225,7 +226,7 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 		return;
 	}

-	if (!ShaderAddressValid(sh_ctx.GetCs().cs_regs.data_addr)) {
+	if (sh_ctx.GetCs().cs_regs.data_addr == 0) {
 		return;
 	}

@@ -255,6 +256,10 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 	input_info.dispatch_thread_dimensions = use_thread_dimensions;
 	const auto compute_program =
 	    m_context.GetPipelineCache().GetComputeProgram(cs_regs, sh_regs, input_info);
+	if (!compute_program) {
+		// Temporary until RT is implemented.
+		return;
+	}
 	if (use_thread_dimensions) {
 		input_info.dispatch_threads_num[0]    = thread_group_x;
 		input_info.dispatch_threads_num[1]    = thread_group_y;
@@ -262,7 +267,7 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 	}

 	const auto& program   = *input_info.stage.program;
-	const auto& resources = input_info.stage.resources;
+	const auto& resources = *input_info.stage.resources;
 	if (TryConsumeComputeMetaClear(input_info, buffer)) {
 		ResetBindings();
 		return;
@@ -361,7 +366,8 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 	buffer.EndRendering();
 	auto& pipeline =
 	    m_context.GetPipelineCache().GetComputePipeline(input_info, compute_program);
-	auto bindings = PrepareBindings(input_info.stage);
+	auto& bindings = m_compute_bindings;
+	PrepareBindings(input_info.stage, bindings);
 	FindBuffers(bindings);
 	if (program.info.uses_dma) {
 		m_context.PrepareBda();
@@ -395,4 +401,65 @@ void RenderExecutor::DispatchDirect(uint64_t submit_id, CommandBuffer& buffer,
 	ResetBindings();
 }

+void RenderExecutor::DispatchIndirect(uint64_t submit_id, CommandBuffer& buffer,
+                                      uint64_t args_addr, uint32_t mode) {
+	EXIT_IF(buffer.IsInvalid() || args_addr == 0 || (args_addr & 3u) != 0 ||
+	        (mode & Pm4::COMPUTE_DISPATCH_INITIATOR_USE_THREAD_DIMENSIONS) != 0);
+	m_context.GetCommandScheduler().PopPendingOperations();
+	buffer.SetDebugInfo(static_cast<uint32_t>(CommandBufferDebugOp::DispatchIndirect), submit_id,
+	                    static_cast<uint32_t>(args_addr), static_cast<uint32_t>(args_addr >> 32u),
+	                    0, mode, buffer.GetShaders().GetCs().cs_regs.data_addr);
+	Common::LockGuard lock(m_context.GetMutex());
+	const auto& cs_regs = buffer.GetShaders().GetCs();
+	if (cs_regs.cs_regs.data_addr == 0) {
+		return;
+	}
+	ShaderComputeInputInfo input_info {};
+	const auto compute_program = m_context.GetPipelineCache().GetComputeProgram(
+	    cs_regs, buffer.GetRegisters().GetShaderRegisters(), input_info);
+	if (!compute_program) {
+		// Temporary until RT is implemented.
+		return;
+	}
+	buffer.EndRendering();
+	auto& pipeline = m_context.GetPipelineCache().GetComputePipeline(input_info, compute_program);
+	auto& bindings = m_compute_bindings;
+	PrepareBindings(input_info.stage, bindings);
+	FindBuffers(bindings);
+	const auto& program = *input_info.stage.program;
+	if (program.info.uses_dma) {
+		m_context.PrepareBda();
+	}
+	RebindImages(bindings);
+	// Acquiring arguments can merge cache buffers; finalize shader bindings afterward.
+	const auto [args_buffer, args_offset] = m_context.GetBufferCache().ObtainBuffer(
+	    args_addr, sizeof(vk::DispatchIndirectCommand), false);
+	EXIT_IF(args_buffer == nullptr || (args_offset & 3u) != 0);
+	RebindBuffers(bindings);
+	PreparedBindings* descriptor_stage = &bindings;
+	CommitBindings(buffer, vk::PipelineBindPoint::eCompute, pipeline,
+	               std::span {&descriptor_stage, 1u});
+	const auto vk_buffer = buffer.Handle();
+	const bool has_storage_writes = HasShaderBufferWrites(input_info.stage) ||
+	    std::any_of(program.info.images.begin(), program.info.images.end(), [](const auto& image) {
+		    return image.written && image.resource_class ==
+		                                ShaderRecompiler::IR::ImageResourceClass::Storage;
+	    });
+	if (has_storage_writes) {
+		ShaderWriteHazardBarrier(vk_buffer, vk::PipelineStageFlagBits::eComputeShader);
+	}
+	vk::MemoryBarrier barrier {};
+	barrier.srcAccessMask = vk::AccessFlagBits::eShaderWrite | vk::AccessFlagBits::eTransferWrite;
+	barrier.dstAccessMask = vk::AccessFlagBits::eIndirectCommandRead;
+	vk_buffer.pipelineBarrier(vk::PipelineStageFlagBits::eAllGraphics |
+	                              vk::PipelineStageFlagBits::eComputeShader |
+	                              vk::PipelineStageFlagBits::eTransfer,
+	                          vk::PipelineStageFlagBits::eDrawIndirect, {},
+	                          1, &barrier, 0, nullptr, 0, nullptr);
+	vk_buffer.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline.pipeline);
+	vk_buffer.dispatchIndirect(args_buffer->Handle(), args_offset);
+	ShaderAccessBarrier(vk_buffer, vk::PipelineStageFlagBits::eComputeShader);
+	ResetBindings();
+}
+
 } // namespace Libs::Graphics
@@ -2,6 +2,7 @@

 #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"
@@ -57,7 +58,7 @@ std::pair<int32_t, uint32_t> ResolveDrawOffsets(uint32_t index_offset,

 	EXIT_IF(!vs_input_info.stage);
 	const auto& program   = *vs_input_info.stage.program;
-	const auto& resources = vs_input_info.stage.resources;
+	const auto& resources = *vs_input_info.stage.resources;
 	if (index_offset == 0 &&
 	    program.info.vertex_offset_sgpr >= static_cast<int32_t>(program.user_data_base)) {
 		const auto index =
@@ -421,7 +422,7 @@ static void SetGraphicsDynamicParams(const CommandBuffer& buffer, vk::CommandBuf
 static bool DrawHasValidVertexShader(const HW::Shader& sh_ctx) {

 	const auto& vs = sh_ctx.GetVs();
-	return ShaderAddressValid(vs.es_regs.data_addr);
+	return vs.es_regs.data_addr != 0;
 }

 static bool PixelShaderHasDepthOrCoverageSideEffects(const HW::ShaderRegisters& sh_regs) {
@@ -452,15 +453,16 @@ struct DrawCallInfo {

 RenderState RenderExecutor::AcquireRenderTargets(CommandBuffer& buffer, RenderColorInfo* colors,
                                                 uint32_t color_count, RenderDepthInfo& depth,
-                                                 const std::optional<PreparedBindings>& pixel) {
+                                                 vk::ImageAspectFlags& feedback_aspects,
+                                                 std::span<PreparedBindings* const> stages) {
 	EXIT_IF(colors == nullptr || color_count > RENDER_COLOR_ATTACHMENTS_MAX);
+	feedback_aspects = {};
 	auto&       cache = m_context.GetTextureCache();
 	RenderState state {};
 	state.width                 = std::numeric_limits<uint32_t>::max();
 	state.height                = std::numeric_limits<uint32_t>::max();
 	state.num_layers            = std::numeric_limits<uint32_t>::max();
 	state.num_color_attachments = 0;
-	uint32_t attachment_samples = 0;
 	for (uint32_t i = 0; i < color_count; i++) {
 		auto& target = colors[i];
 		EXIT_IF(!target.image_id);
@@ -471,22 +473,7 @@ RenderState RenderExecutor::AcquireRenderTargets(CommandBuffer& buffer, RenderCo
 		}
 		const auto image_view = cache.FindRenderTarget(target.image_id, target.desc);
 		auto&      image      = cache.GetImage(target.image_id);
-		SetVulkanObjectNameF(m_context.GetGraphics().device, image.backing.image,
-		                     "Kyty.MRT{}.Image[guest=0x{:016x} size=0x{:x} format={}]",
-		                     target.target_slot, image.info.data.address, image.info.data.size,
-		                     static_cast<uint32_t>(image.info.pixel_format));
-		SetVulkanObjectNameF(m_context.GetGraphics().device, image_view,
-		                     "Kyty.MRT{}.View[guest=0x{:016x} mip={} layer={}+{}]",
-		                     target.target_slot, image.info.data.address,
-		                     target.desc.view_info.base_level, target.desc.view_info.base_layer,
-		                     target.desc.view_info.layer_count);
 		EXIT_IF(image.backing.samples != target.desc.info.samples || image_view == nullptr);
-		if (attachment_samples == 0) {
-			attachment_samples = target.desc.info.samples;
-		} else if (attachment_samples != target.desc.info.samples) {
-			EXIT("mixed color attachment sample counts are unsupported: %u and %u\n",
-			     attachment_samples, target.desc.info.samples);
-		}
 		const auto& view   = target.desc.view_info;
 		const auto  layout = image.binding.is_bound ? vk::ImageLayout::eGeneral
 		                                            : vk::ImageLayout::eColorAttachmentOptimal;
@@ -526,43 +513,30 @@ RenderState RenderExecutor::AcquireRenderTargets(CommandBuffer& buffer, RenderCo
 			EXIT("failed to consume HTile clear state\n");
 		}
 		auto& image = cache.GetImage(depth.image_id);
-		SetVulkanObjectNameF(m_context.GetGraphics().device, image.backing.image,
-		                     "Kyty.DepthTarget.Image[guest=0x{:016x} size=0x{:x} format={}]",
-		                     image.info.data.address, image.info.data.size,
-		                     static_cast<uint32_t>(image.info.pixel_format));
-		SetVulkanObjectNameF(m_context.GetGraphics().device, image_view,
-		                     "Kyty.DepthTarget.View[guest=0x{:016x} layer={}+{}]",
-		                     image.info.data.address, depth.desc.view_info.base_layer,
-		                     depth.desc.view_info.layer_count);
 		EXIT_IF(image_view == nullptr || image.backing.samples != depth.desc.info.samples);
-		if (attachment_samples == 0) {
-			attachment_samples = depth.desc.info.samples;
-		} else if (attachment_samples != depth.desc.info.samples) {
-			EXIT("mixed color/depth sample counts are unsupported: %u and %u\n", attachment_samples,
-			     depth.desc.info.samples);
+		const auto draw_writes = depth.AttachmentWriteAspects();
+		vk::ImageAspectFlags sampled_aspects;
+		for (const auto* stage: stages) {
+			for (const auto& binding: stage->images) {
+				if (binding.image_id != depth.image_id ||
+				    binding.desc.type != TextureCache::BindingType::Texture) continue;
+				const auto native =
+				    std::ranges::find(image.views, binding.image_view, &CachedImageView::view);
+				EXIT_IF(native == image.views.end());
+				sampled_aspects |= native->info.aspect;
+				feedback_aspects |= DepthFeedbackAspects(draw_writes, depth.desc.view_info,
+				                                         native->info);
+			}
 		}
-		const bool feedback = depth.depth_write_enable && pixel &&
-		    std::ranges::any_of(pixel->images, [&](const TextureBinding& binding) {
-			    if (binding.image_id != depth.image_id ||
-			        binding.desc.type != TextureCache::BindingType::Texture) {
-				    return false;
-			    }
-			    const auto native =
-			        std::ranges::find(image.views, binding.image_view, &CachedImageView::view);
-			    EXIT_IF(native == image.views.end());
-			    const auto& sampled = native->info;
-			    const auto& target = depth.desc.view_info;
-			    return (sampled.aspect & vk::ImageAspectFlagBits::eDepth) &&
-			           ImageRangeOverlaps(sampled.base_level, sampled.level_count,
-			                              target.base_level, target.level_count) &&
-			           ImageRangeOverlaps(sampled.base_layer, sampled.layer_count,
-			                              target.base_layer, target.layer_count);
-		    });
-		if (feedback && !m_context.GetGraphics().attachment_feedback_loop_enabled) {
+		if (feedback_aspects && !m_context.GetGraphics().attachment_feedback_loop_enabled) {
 			EXIT("depth attachment feedback loop is not supported by the host\n");
 		}
-		const auto layout = feedback ? vk::ImageLayout::eAttachmentFeedbackLoopOptimalEXT
-		                             : depth_attachment_layout(depth);
+		auto layout = depth_attachment_layout(depth);
+		if (sampled_aspects & ~DepthReadableAspects(layout)) {
+			layout = m_context.GetGraphics().attachment_feedback_loop_enabled
+			             ? vk::ImageLayout::eAttachmentFeedbackLoopOptimalEXT
+			             : vk::ImageLayout::eGeneral;
+		}
 		// The attachment store writes even when guest depth/stencil tests do not.
 		const auto access = vk::AccessFlagBits2::eDepthStencilAttachmentRead |
 		                    vk::AccessFlagBits2::eDepthStencilAttachmentWrite;
@@ -591,9 +565,6 @@ RenderState RenderExecutor::AcquireRenderTargets(CommandBuffer& buffer, RenderCo
 		const auto& limits = buffer.GetGraphics().GetPhysicalDeviceProperties().limits;
 		state.width        = limits.maxFramebufferWidth;
 		state.height       = limits.maxFramebufferHeight;
-	} else if (attachment_samples == 0 ||
-	           vulkan_sample_count(attachment_samples) == vk::SampleCountFlagBits {}) {
-		EXIT("render state has no valid attachments\n");
 	}
 	if (state.num_layers == std::numeric_limits<uint32_t>::max()) {
 		state.num_layers = 1;
@@ -608,7 +579,7 @@ static bool DrawHasActivePixelShader(const CommandBuffer& buffer) {
 	const auto& ctx              = buffer.GetRegisters();
 	const auto& sh_regs          = ctx.GetShaderRegisters();
 	const bool  has_color_output = (ctx.GetRenderTargetMask() & sh_regs.m_cbShaderMask) != 0;
-	return ShaderAddressValid(buffer.GetShaders().GetPs().ps_regs.data_addr) &&
+	return buffer.GetShaders().GetPs().ps_regs.data_addr != 0 &&
 	       (has_color_output || PixelShaderHasDepthOrCoverageSideEffects(sh_regs));
 }

@@ -698,11 +669,13 @@ static PreparedVertexBuffers AcquireVertexBuffers(CommandBuffer&               b
 	        vs_input_info.buffers_num > ShaderVertexInputInfo::RES_MAX);

 	// Collect the non-empty guest vertex ranges.
+	std::array<uint64_t, ShaderVertexInputInfo::RES_MAX>          sizes {};
 	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, vs_input_info);
+		sizes[i]           = size;
 		if (size == 0) {
 			continue;
 		}
@@ -751,7 +724,7 @@ static PreparedVertexBuffers AcquireVertexBuffers(CommandBuffer&               b
 	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, vs_input_info);
+		const auto  size   = sizes[i];
 		if (size == 0) {
 			if (null_buffer == nullptr) {
 				null_buffer = cache.GetBuffer(NULL_BUFFER_ID).Handle();
@@ -789,8 +762,7 @@ static void SetDrawDebugPhase(CommandBuffer& buffer, uint64_t submit_id, const D
 	                    draw.instance_count, draw.first_instance);
 }

-static bool GetDrawTopology(const HW::UserConfig& ucfg, bool auto_draw,
-                            vk::PrimitiveTopology& topology) {
+static bool GetDrawTopology(const HW::UserConfig& ucfg, vk::PrimitiveTopology& topology) {

 	topology = vk::PrimitiveTopology::ePointList;

@@ -813,14 +785,11 @@ static bool GetDrawTopology(const HW::UserConfig& ucfg, bool auto_draw,
 			topology = vk::PrimitiveTopology::eTriangleStrip;
 			break;
 		case Prospero::PrimitiveType::kPatch:
+			if (!Config::TessellationEnabled()) return false;
+			[[fallthrough]];
 		case Prospero::PrimitiveType::kRectList:
-			topology = vk::PrimitiveTopology::ePatchList;
-			break;
 		case Prospero::PrimitiveType::kRectListLegacy:
-			if (!auto_draw) {
-				EXIT("unknown primitive type: %u\n", static_cast<uint32_t>(ucfg.GetPrimType()));
-			}
-			topology = vk::PrimitiveTopology::eTriangleStrip;
+			topology = vk::PrimitiveTopology::ePatchList;
 			break;
 		case Prospero::PrimitiveType::kQuadListLegacy:
 			topology = vk::PrimitiveTopology::eTriangleFan;
@@ -999,8 +968,7 @@ static void LogDrawStateIfNeeded(const CommandBuffer& buffer, const DrawCallInfo
 		return;
 	}

-	if (!draw.IsIndexed() &&
-	    buffer.GetUserConfig().GetPrimType() != Prospero::PrimitiveType::kRectListLegacy) {
+	if (!draw.IsIndexed() && !Prospero::IsRectList(buffer.GetUserConfig().GetPrimType())) {
 		return;
 	}

@@ -1013,8 +981,7 @@ static void LogDrawStateIfNeeded(const CommandBuffer& buffer, const DrawCallInfo
 }

 static void EmitDrawPrimitives(const HW::UserConfig& ucfg, vk::CommandBuffer vk_buffer,
-                               const ShaderVertexInputInfo& vs_input_info, const DrawCallInfo& draw,
-                               const DrawEmitInfo& emit) {
+                               const DrawCallInfo& draw, const DrawEmitInfo& emit) {
 	switch (ucfg.GetPrimType()) {
 		case Prospero::PrimitiveType::kPointList:
 		case Prospero::PrimitiveType::kLineList:
@@ -1023,6 +990,7 @@ static void EmitDrawPrimitives(const HW::UserConfig& ucfg, vk::CommandBuffer vk_
 		case Prospero::PrimitiveType::kTriFan:
 		case Prospero::PrimitiveType::kTriStrip:
 		case Prospero::PrimitiveType::kRectList:
+		case Prospero::PrimitiveType::kRectListLegacy:
 		case Prospero::PrimitiveType::kPatch:
 			if (draw.IsIndexed()) {
 				vk_buffer.drawIndexed(draw.index_count, draw.instance_count, 0, emit.vertex_offset,
@@ -1032,14 +1000,6 @@ static void EmitDrawPrimitives(const HW::UserConfig& ucfg, vk::CommandBuffer vk_
 				               emit.first_instance);
 			}
 			break;
-		case Prospero::PrimitiveType::kRectListLegacy:
-			if (draw.IsIndexed()) {
-				EXIT("unknown primitive type: %u\n", static_cast<uint32_t>(ucfg.GetPrimType()));
-			}
-			// Sarah
-			EXIT_NOT_IMPLEMENTED(!(draw.index_count == 3 && vs_input_info.buffers_num == 0));
-			vk_buffer.draw(4, draw.instance_count, emit.first_vertex, emit.first_instance);
-			break;
 		case Prospero::PrimitiveType::kQuadListLegacy:
 			EXIT_NOT_IMPLEMENTED((draw.index_count & 0x3u) != 0);
 			for (uint32_t i = 0; i < draw.index_count; i += 4) {
@@ -1101,15 +1061,16 @@ void RenderExecutor::ExecutePreparedDraw(uint64_t submit_id, CommandBuffer& buff
 		                              index_source.guest_element_size);
 	}
 	LogDrawPhase(draw.Name(), "PrepareBindings");
-	GraphicsBindings                 bindings;
+	auto&                            bindings = m_graphics_bindings;
 	std::array<PreparedBindings*, 4> descriptor_stages {};
 	uint32_t                         stage_count = 0;
 	for (uint32_t i = 0; i < vertex_stages.size(); i++) {
-		bindings.vertex[i]               = PrepareBindings(state.vertex_info[i].stage);
+		PrepareBindings(state.vertex_info[i].stage, bindings.vertex[i]);
 		descriptor_stages[stage_count++] = &bindings.vertex[i];
 	}
 	if (state.ps_active) {
-		bindings.pixel.emplace(PrepareBindings(state.ps_input_info.stage));
+		if (!bindings.pixel) bindings.pixel.emplace();
+		PrepareBindings(state.ps_input_info.stage, *bindings.pixel);
 		descriptor_stages[stage_count++] = &*bindings.pixel;
 	}
 	const auto stages = std::span {descriptor_stages.data(), stage_count};
@@ -1121,10 +1082,6 @@ void RenderExecutor::ExecutePreparedDraw(uint64_t submit_id, CommandBuffer& buff
 		vertex_bindings = AcquireVertexBuffers(buffer, state.vertex_info[0]);
 		index_binding   = PrepareIndexBuffer(buffer, index_source);
 	}
-	const auto rendering =
-	    AcquireRenderTargets(buffer, state.color_info, state.color_count, state.depth_info,
-	                         bindings.pixel);
-
 	if (draw.IsIndexed()) {
 		LogDrawPhase(draw.Name(), "CreatePipeline");
 	}
@@ -1132,6 +1089,10 @@ void RenderExecutor::ExecutePreparedDraw(uint64_t submit_id, CommandBuffer& buff
 	    std::span {state.color_info, state.color_count}, state.depth_info, vertex_stages, buffer,
 	    state.ps_active ? &state.ps_input_info : nullptr, topology, primitive_restart_enable,
 	    state.programs);
+	vk::ImageAspectFlags feedback_aspects;
+	const auto rendering =
+	    AcquireRenderTargets(buffer, state.color_info, state.color_count, state.depth_info,
+	                         feedback_aspects, stages);

 	// Resource preparation above may synchronously finish and restart the scheduler. From this
 	// point onward, every operation targets the current command buffer and cannot touch guest
@@ -1141,7 +1102,7 @@ void RenderExecutor::ExecutePreparedDraw(uint64_t submit_id, CommandBuffer& buff
 	if (!mesh_active) {
 		CommitVertexBuffers(vk_buffer, vertex_bindings);
 	}
-	if (bindings.pixel && !draw.IsIndexed()) {
+	if (state.ps_active && !draw.IsIndexed()) {
 		SetDrawDebugPhase(buffer, submit_id, draw, 0x300u);
 	}
 	CommitBindings(buffer, vk::PipelineBindPoint::eGraphics, pipeline, stages);
@@ -1163,11 +1124,7 @@ void RenderExecutor::ExecutePreparedDraw(uint64_t submit_id, CommandBuffer& buff

 	SetGraphicsDynamicParams(buffer, vk_buffer, vertex_stages.back(), state.depth_info, rendering);
 	if (m_context.GetGraphics().attachment_feedback_loop_enabled) {
-		vk_buffer.setAttachmentFeedbackLoopEnableEXT(
-		    rendering.depth_stencil_attachment.image_layout ==
-		            vk::ImageLayout::eAttachmentFeedbackLoopOptimalEXT
-		        ? vk::ImageAspectFlags {vk::ImageAspectFlagBits::eDepth}
-		        : vk::ImageAspectFlags {});
+		vk_buffer.setAttachmentFeedbackLoopEnableEXT(feedback_aspects);
 	}

 	LogDrawPhase(draw.Name(), "BeginRendering");
@@ -1182,7 +1139,7 @@ void RenderExecutor::ExecutePreparedDraw(uint64_t submit_id, CommandBuffer& buff
 	if (mesh_active) {
 		vk_buffer.drawMeshTasksEXT(mesh_groups, draw.instance_count, 1);
 	} else {
-		EmitDrawPrimitives(ucfg, vk_buffer, state.vertex_info[0], draw, emit);
+		EmitDrawPrimitives(ucfg, vk_buffer, draw, emit);
 	}

 	if (!draw.IsIndexed()) {
@@ -1258,7 +1215,7 @@ void RenderExecutor::DrawIndex(uint64_t submit_id, CommandBuffer& buffer,
 	hw_check(buffer);

 	vk::PrimitiveTopology topology = vk::PrimitiveTopology::ePointList;
-	if (!GetDrawTopology(ucfg, false, topology)) {
+	if (!GetDrawTopology(ucfg, topology)) {
 		return;
 	}

@@ -1368,7 +1325,7 @@ void RenderExecutor::DrawAuto(uint64_t submit_id, CommandBuffer& buffer, const D
 	                         args.vertex_count, args.instance_count, args.first_instance};

 	vk::PrimitiveTopology topology = vk::PrimitiveTopology::ePointList;
-	if (!GetDrawTopology(ucfg, true, topology)) {
+	if (!GetDrawTopology(ucfg, topology)) {
 		ResetBindings();
 		return;
 	}
@@ -1378,7 +1335,7 @@ void RenderExecutor::DrawAuto(uint64_t submit_id, CommandBuffer& buffer, const D
 		return;
 	}

-	const bool rect_list = ucfg.GetPrimType() == Prospero::PrimitiveType::kRectList;
+	const bool rect_list = Prospero::IsRectList(ucfg.GetPrimType());
 	if (rect_list && state.vertex_info[0].buffers_num == 0 &&
 	    state.vertex_info[0].stage.program->param_export_mask == 0 &&
 	    state.ps_input_info.input_num != 0) {
@@ -112,26 +112,6 @@ void WriteAtEndOfPipe64(uint64_t submit_id, CommandBuffer& buffer, uint64_t* dst
 	                     EndOfPipeWriteSize::Qword, EndOfPipeWriteAction::Write);
 }

-void WriteAtEndOfPipeClockCounter(uint64_t submit_id, CommandBuffer& buffer, uint64_t* dst_gpu_addr,
-                                  uint64_t value) {
-	RecordEndOfPipeWrite(submit_id, buffer, reinterpret_cast<uint64_t>(dst_gpu_addr), 0,
-	                     EndOfPipeWriteSize::Qword, EndOfPipeWriteAction::Write);
-
-	LOGF_COLOR(Log::Color::BrightGreen,
-	           "EndOfPipe Signal!!! [0x%016" PRIx64 "] <- Clock: 0x%016" PRIx64 "\n",
-	           reinterpret_cast<uint64_t>(dst_gpu_addr), value);
-}
-
-void WriteAtEndOfPipeClockCounterWithWriteBack(uint64_t submit_id, CommandBuffer& buffer,
-                                               uint64_t* dst_gpu_addr, uint64_t value) {
-	RecordEndOfPipeWrite(submit_id, buffer, reinterpret_cast<uint64_t>(dst_gpu_addr), 0,
-	                     EndOfPipeWriteSize::Qword, EndOfPipeWriteAction::WriteBack);
-
-	LOGF_COLOR(Log::Color::BrightGreen,
-	           "EndOfPipe Signal!!! [0x%016" PRIx64 "] <- Clock: 0x%016" PRIx64 "\n",
-	           reinterpret_cast<uint64_t>(dst_gpu_addr), value);
-}
-
 void WriteAtEndOfPipeWithWriteBack64(uint64_t submit_id, CommandBuffer& buffer,
                                     uint64_t* dst_gpu_addr, uint64_t value) {
 	RecordEndOfPipeWrite(submit_id, buffer, reinterpret_cast<uint64_t>(dst_gpu_addr), value,
@@ -25,10 +25,6 @@ void WriteAtEndOfPipe64(uint64_t submit_id, CommandBuffer& buffer, uint64_t* dst
                        uint64_t value);
 void WriteAtEndOfPipeGds32(uint64_t submit_id, CommandBuffer& buffer, uint32_t* dst_gpu_addr,
                           uint32_t dw_offset, uint32_t dw_num);
-void WriteAtEndOfPipeClockCounter(uint64_t submit_id, CommandBuffer& buffer, uint64_t* dst_gpu_addr,
-                                  uint64_t value);
-void WriteAtEndOfPipeClockCounterWithWriteBack(uint64_t submit_id, CommandBuffer& buffer,
-                                               uint64_t* dst_gpu_addr, uint64_t value);
 void WriteAtEndOfPipeWithWriteBack32(uint64_t submit_id, CommandBuffer& buffer,
                                     uint32_t* dst_gpu_addr, uint32_t value);
 void WriteAtEndOfPipeWithWriteBack64(uint64_t submit_id, CommandBuffer& buffer,
@@ -24,8 +24,6 @@

 namespace Libs::Graphics {

-using VulkanMemoryBarrier = vk::MemoryBarrier;
-
 vk::Format  VulkanFormat(Prospero::BufferFormat guest_format);
 void        RequireVulkanSuccess(vk::Result result, const char* operation);
 vk::ShaderModule CompileSPV(std::span<const uint32_t> code, vk::Device device);
@@ -1,6 +1,7 @@
 #include "graphics/presentation/systemOverlay.h"

-#include "SDL.h"
+#include <SDL3/SDL.h>
+
 #include "common/assert.h"
 #include "common/stringUtils.h"
 #include "graphics/host_gpu/graphicContext.h"
@@ -176,6 +177,7 @@ bool                         g_input_multiline            = false;
 bool                         g_input_lifecycle_active     = false;
 bool                         g_controller_captured        = false;
 OverlaySession               g_session;
+SDL_Window*                  g_input_window               = nullptr;

 void ClearInputEvents() {
 	std::scoped_lock lock(g_input_mutex);
@@ -212,7 +214,7 @@ void RetryVisibilityWakeup() {
 	}
 	SDL_Event event {};
 	event.type = type;
-	if (SDL_PushEvent(&event) > 0) {
+	if (SDL_PushEvent(&event)) {
 		g_missing_visibility_wakeups--;
 	}
 }
@@ -252,7 +254,7 @@ void RefreshVisibility() {
 		std::scoped_lock input_lock(g_input_mutex);
 		g_visibility_updates.push_back({snapshot.session, visible, capture_controller,
 		                                capture_keyboard, text_input, multiline});
-		if (g_missing_visibility_wakeups != 0 || SDL_PushEvent(&event) <= 0) {
+		if (g_missing_visibility_wakeups != 0 || !SDL_PushEvent(&event)) {
 			g_missing_visibility_wakeups++;
 		}
 	}
@@ -268,16 +270,16 @@ void OnDialogVisibilityChanged(bool, uint64_t) {

 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;
+	if ((modifiers & SDL_KMOD_LCTRL) != 0) status |= 0x00000100;
+	if ((modifiers & SDL_KMOD_LSHIFT) != 0) status |= 0x00000200;
+	if ((modifiers & SDL_KMOD_LALT) != 0) status |= 0x00000400;
+	if ((modifiers & SDL_KMOD_LGUI) != 0) status |= 0x00000800;
+	if ((modifiers & SDL_KMOD_RCTRL) != 0) status |= 0x00001000;
+	if ((modifiers & SDL_KMOD_RSHIFT) != 0) status |= 0x00002000;
+	if ((modifiers & SDL_KMOD_RALT) != 0) status |= 0x00004000;
+	if ((modifiers & SDL_KMOD_RGUI) != 0) status |= 0x00008000;
+	if ((modifiers & SDL_KMOD_NUM) != 0) status |= 0x00010000;
+	if ((modifiers & SDL_KMOD_CAPS) != 0) status |= 0x00020000;
 	return status;
 }

@@ -392,14 +394,14 @@ float AlignmentPivot(Ime::Alignment alignment) {

 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;
+		case SDL_GAMEPAD_BUTTON_SOUTH: return ImGuiKey_GamepadFaceDown;
+		case SDL_GAMEPAD_BUTTON_EAST: return ImGuiKey_GamepadFaceRight;
+		case SDL_GAMEPAD_BUTTON_WEST: return ImGuiKey_GamepadFaceLeft;
+		case SDL_GAMEPAD_BUTTON_NORTH: return ImGuiKey_GamepadFaceUp;
+		case SDL_GAMEPAD_BUTTON_DPAD_LEFT: return ImGuiKey_GamepadDpadLeft;
+		case SDL_GAMEPAD_BUTTON_DPAD_RIGHT: return ImGuiKey_GamepadDpadRight;
+		case SDL_GAMEPAD_BUTTON_DPAD_UP: return ImGuiKey_GamepadDpadUp;
+		case SDL_GAMEPAD_BUTTON_DPAD_DOWN: return ImGuiKey_GamepadDpadDown;
 		default: return ImGuiKey_None;
 	}
 }
@@ -415,9 +417,11 @@ void CheckVulkanResult(VkResult result) {

 } // namespace

-void InitializeSystemOverlayInput() {
+void InitializeSystemOverlayInput(SDL_Window* window) {
+	EXIT_IF(window == nullptr);
 	const Uint32 type = SDL_RegisterEvents(1);
 	EXIT_IF(type == static_cast<Uint32>(-1));
+	g_input_window = window;
 	{
 		std::scoped_lock lock(g_visibility_mutex);
 		g_input_lifecycle_active = true;
@@ -453,9 +457,10 @@ void ShutdownSystemOverlayInput() {
 	g_input_controller = false;
 	g_input_keyboard   = false;
 	g_input_multiline  = false;
-	if (SDL_IsTextInputActive() == SDL_TRUE) {
-		SDL_StopTextInput();
+	if (g_input_window != nullptr && SDL_TextInputActive(g_input_window)) {
+		SDL_StopTextInput(g_input_window);
 	}
+	g_input_window = nullptr;
 }

 SystemOverlayVisualState GetSystemOverlayVisualState() noexcept {
@@ -487,13 +492,13 @@ bool ProcessSystemOverlayInput(const SDL_Event& event) {
 		g_input_keyboard        = update.capture_keyboard;
 		g_input_multiline       = update.multiline;
 		if (update.text_input) {
-			SDL_StartTextInput();
-		} else if (SDL_IsTextInputActive() == SDL_TRUE) {
-			SDL_StopTextInput();
+			SDL_StartTextInput(g_input_window);
+		} else if (SDL_TextInputActive(g_input_window)) {
+			SDL_StopTextInput(g_input_window);
 		}
 		return true;
 	}
-	if (event.type == SDL_CONTROLLERDEVICEREMOVED) {
+	if (event.type == SDL_EVENT_GAMEPAD_REMOVED) {
 		if (g_input_active && g_input_controller) {
 			QueueInput({InputKind::ResetController, g_input_session, 0, 0.0f, 0.0f});
 		}
@@ -505,26 +510,26 @@ bool ProcessSystemOverlayInput(const SDL_Event& event) {

 	const auto     session          = g_input_session;
 	const uint64_t generation       = session.generation;
-	const bool     controller_event = event.type == SDL_CONTROLLERBUTTONDOWN ||
-	                                  event.type == SDL_CONTROLLERBUTTONUP ||
-	                                  event.type == SDL_CONTROLLERAXISMOTION;
+	const bool     controller_event = event.type == SDL_EVENT_GAMEPAD_BUTTON_DOWN ||
+	                                  event.type == SDL_EVENT_GAMEPAD_BUTTON_UP ||
+	                                  event.type == SDL_EVENT_GAMEPAD_AXIS_MOTION;
 	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;
+	const bool keyboard_event = event.type == SDL_EVENT_TEXT_INPUT || event.type == SDL_EVENT_TEXT_EDITING ||
+	                            event.type == SDL_EVENT_KEY_DOWN || event.type == SDL_EVENT_KEY_UP;
 	if (keyboard_event && !g_input_keyboard) {
 		return false;
 	}
 	if (keyboard_event && session.kind == OverlayKind::Error) {
-		if (event.type == SDL_KEYDOWN && event.key.repeat == 0 &&
-		    (event.key.keysym.sym == SDLK_RETURN || event.key.keysym.sym == SDLK_KP_ENTER)) {
+		if (event.type == SDL_EVENT_KEY_DOWN && !event.key.repeat &&
+		    (event.key.key == SDLK_RETURN || event.key.key == SDLK_KP_ENTER)) {
 			ErrorDialog::HostAccept(generation);
 		}
 		return true;
 	}
 	switch (event.type) {
-		case SDL_TEXTINPUT: {
+		case SDL_EVENT_TEXT_INPUT: {
 			const auto text = Utf8ToUtf16(event.text.text);
 			if (!text.empty()) {
 				auto input = MakeExternalInput(Ime::ExternalAction::Text, g_last_external_keycode,
@@ -535,26 +540,24 @@ bool ProcessSystemOverlayInput(const SDL_Event& event) {
 			}
 			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);
+		case SDL_EVENT_TEXT_EDITING: return true;
+		case SDL_EVENT_KEY_DOWN: {
+			g_last_external_keycode = static_cast<uint16_t>(event.key.scancode);
+			g_last_external_status  = ExternalKeyStatus(event.key.mod, false);
 			auto action = Ime::ExternalAction::Text;
 			bool queue  = true;
-			if (event.key.keysym.sym == SDLK_BACKSPACE) {
+			if (event.key.key == SDLK_BACKSPACE) {
 				action = Ime::ExternalAction::Backspace;
-			} else if (event.key.keysym.sym == SDLK_LEFT) {
+			} else if (event.key.key == SDLK_LEFT) {
 				action = Ime::ExternalAction::MoveLeft;
-			} else if (event.key.keysym.sym == SDLK_RIGHT) {
+			} else if (event.key.key == SDLK_RIGHT) {
 				action = Ime::ExternalAction::MoveRight;
-			} else if (event.key.keysym.sym == SDLK_ESCAPE) {
+			} else if (event.key.key == SDLK_ESCAPE) {
 				action = Ime::ExternalAction::Cancel;
-			} else if (event.key.keysym.sym == SDLK_RETURN ||
-			           event.key.keysym.sym == SDLK_KP_ENTER) {
+			} else if (event.key.key == SDLK_RETURN || event.key.key == SDLK_KP_ENTER) {
 				action =
 				    g_input_multiline ? Ime::ExternalAction::Newline : Ime::ExternalAction::Accept;
-			} else if (event.key.keysym.sym == SDLK_TAB) {
+			} else if (event.key.key == SDLK_TAB) {
 				action = Ime::ExternalAction::None;
 			} else {
 				queue = false;
@@ -566,28 +569,28 @@ bool ProcessSystemOverlayInput(const SDL_Event& event) {
 			}
 			return true;
 		}
-		case SDL_KEYUP: return true;
-		case SDL_CONTROLLERBUTTONDOWN:
-		case SDL_CONTROLLERBUTTONUP:
-			QueueInput({InputKind::Button, session, event.cbutton.button,
-			            event.type == SDL_CONTROLLERBUTTONDOWN ? 1.0f : 0.0f, 0.0f});
+		case SDL_EVENT_KEY_UP: return true;
+		case SDL_EVENT_GAMEPAD_BUTTON_DOWN:
+		case SDL_EVENT_GAMEPAD_BUTTON_UP:
+			QueueInput({InputKind::Button, session, event.gbutton.button,
+			            event.type == SDL_EVENT_GAMEPAD_BUTTON_DOWN ? 1.0f : 0.0f, 0.0f});
 			return true;
-		case SDL_CONTROLLERAXISMOTION:
-			QueueInput({InputKind::Axis, session, event.caxis.axis,
-			            static_cast<float>(event.caxis.value), 0.0f});
+		case SDL_EVENT_GAMEPAD_AXIS_MOTION:
+			QueueInput({InputKind::Axis, session, event.gaxis.axis,
+			            static_cast<float>(event.gaxis.value), 0.0f});
 			return true;
-		case SDL_MOUSEMOTION:
+		case SDL_EVENT_MOUSE_MOTION:
 			QueueInput({InputKind::MousePosition, session, 0, static_cast<float>(event.motion.x),
 			            static_cast<float>(event.motion.y)});
 			return true;
-		case SDL_MOUSEBUTTONDOWN:
-		case SDL_MOUSEBUTTONUP:
+		case SDL_EVENT_MOUSE_BUTTON_DOWN:
+		case SDL_EVENT_MOUSE_BUTTON_UP:
 			QueueInput({InputKind::MousePosition, session, 0, static_cast<float>(event.button.x),
 			            static_cast<float>(event.button.y)});
 			QueueInput({InputKind::MouseButton, session, event.button.button,
-			            event.type == SDL_MOUSEBUTTONDOWN ? 1.0f : 0.0f, 0.0f});
+			            event.type == SDL_EVENT_MOUSE_BUTTON_DOWN ? 1.0f : 0.0f, 0.0f});
 			return true;
-		case SDL_MOUSEWHEEL:
+		case SDL_EVENT_MOUSE_WHEEL:
 			QueueInput({InputKind::MouseWheel, session, 0, static_cast<float>(event.wheel.x),
 			            static_cast<float>(event.wheel.y)});
 			return true;
@@ -681,9 +684,9 @@ struct SystemOverlay::Impl {
 				case InputKind::Button: {
 					const bool down = event.x != 0.0f;
 					if (session.kind == OverlayKind::Ime && down) {
-						if (event.id == SDL_CONTROLLER_BUTTON_B) {
+						if (event.id == SDL_GAMEPAD_BUTTON_EAST) {
 							Ime::HostCancel(session.generation);
-						} else if (event.id == SDL_CONTROLLER_BUTTON_Y) {
+						} else if (event.id == SDL_GAMEPAD_BUTTON_NORTH) {
 							Ime::HostBackspace(session.generation);
 						}
 					}
@@ -695,19 +698,19 @@ struct SystemOverlay::Impl {
 				}
 				case InputKind::Axis: {
 					const float value = event.x / (event.x < 0.0f ? 32768.0f : 32767.0f);
-					if (event.id == SDL_CONTROLLER_AXIS_LEFTX) {
+					if (event.id == SDL_GAMEPAD_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) {
+					} else if (event.id == SDL_GAMEPAD_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) {
+					} else if (event.id == SDL_GAMEPAD_AXIS_RIGHTX) {
 						right_stick.x = std::abs(value) > 0.2f ? value : 0.0f;
-					} else if (event.id == SDL_CONTROLLER_AXIS_RIGHTY) {
+					} else if (event.id == SDL_GAMEPAD_AXIS_RIGHTY) {
 						right_stick.y = std::abs(value) > 0.2f ? value : 0.0f;
 					}
 					break;
@@ -7,6 +7,7 @@
 #include <memory>

 union SDL_Event;
+struct SDL_Window;

 namespace Libs::Graphics {

@@ -17,7 +18,7 @@ struct SystemOverlayVisualState {
 	uint64_t revision;
 };

-void                     InitializeSystemOverlayInput();
+void                     InitializeSystemOverlayInput(SDL_Window* window);
 void                     ShutdownSystemOverlayInput();
 bool                     ProcessSystemOverlayInput(const SDL_Event& event);
 SystemOverlayVisualState GetSystemOverlayVisualState() noexcept;
@@ -20,9 +20,12 @@
 #include "kernel/pthread.h"
 #include "libs/errno.h"
 #include "libs/libs.h"
+#include "loader/systemContent.h"

 #include <algorithm>
 #include <array>
+#include <atomic>
+#include <fmt/format.h>
 #include <list>
 #include <thread>
 #include <vector>
@@ -540,6 +543,14 @@ Graphics::ImageInfo BufferAttributeGroup::ImageInfo(const VideoOutBuffer& buffer
 	if (!Graphics::DecodeVideoOutPixelFormat(attribute.pixel_format, pixel_format)) {
 		EXIT("unsupported video-out pixel format: 0x%016" PRIx64 "\n", attribute.pixel_format);
 	}
+	if (attribute.pixel_format == Graphics::VIDEO_OUT_PIXEL_FORMAT_R10_G10_B10_A2_BT2100_PQ) {
+		static std::atomic_flag warned = ATOMIC_FLAG_INIT;
+		if (!warned.test_and_set(std::memory_order_relaxed)) {
+			Log::WriteToConsoleAndLog(fmt::format(
+			    "Warning: HDR-to-SDR conversion is not implemented; displaying PQ video-out "
+			    "pixels unchanged (format=0x{:016x}).\n", attribute.pixel_format));
+		}
+	}
 	const auto tile_mode = Graphics::Prospero::TileMode::kRenderTarget;
 	const auto pitch =
 	    Graphics::TileGetTexturePitch(pixel_format.guest_format, attribute.width, tile_mode);
@@ -564,7 +575,12 @@ Graphics::ImageInfo BufferAttributeGroup::ImageInfo(const VideoOutBuffer& buffer
 	info.bgra16          = pixel_format.bgra16;
 	info.mip_layout[0]   = {0, total.size, pitch, attribute.height};
 	if (compression != Graphics::VideoOutCompression::Uncompressed) {
-		info.metadata.range       = {buffer.metadata_address, 0};
+		Graphics::TileSizeAlign dcc_size {};
+		if (!Graphics::TileGetDccSize(attribute.width, attribute.height, 1,
+		                              pixel_format.bytes_per_element, 1, tile_mode, dcc_size)) {
+			EXIT("invalid video-out DCC footprint\n");
+		}
+		info.metadata.range       = {buffer.metadata_address, dcc_size.size};
 		info.metadata.kind        = Graphics::ImageMetadataKind::Dcc;
 		info.metadata.control     = attribute.dcc_control;
 		info.metadata.compression = compression;
@@ -1540,7 +1556,9 @@ KYTY_SYSV_ABI int VideoOutIsFlipPending(int handle) {
 	VideoOutFlipStatus status {};
 	DriverState().GetFlipQueue().GetFlipStatus(*ctx, status);

-	LOGF("\t flipPendingNum = %d\n", status.flipPendingNum);
+	if (Config::GraphicsDebugDumpEnabled()) {
+		LOGF("\t flipPendingNum = %d\n", status.flipPendingNum);
+	}

 	return status.flipPendingNum;
 }
@@ -1654,8 +1672,12 @@ KYTY_SYSV_ABI int VideoOutGetOutputStatus(int handle, VideoOutOutputStatus* stat
 		return VIDEO_OUT_ERROR_INVALID_HANDLE;
 	}

+	int32_t attribute3 = 0;
+	Loader::SystemContentParamSfoGetInt("ATTRIBUTE3", &attribute3);
 	ctx->mutex.Lock();
-	status->resolution   = (ctx->width >= 3840 || ctx->height >= 2160 ? 2u : 1u);
+	// Primary output reports 4K unless param.json Video-out Info enables resolution detection.
+	status->resolution =
+	    ((attribute3 & 4) != 0 && ctx->width < 3840 && ctx->height < 2160 ? 1u : 2u);
 	status->dynamicRange = 1;
 	status->refreshRate =
 	    (ctx->output_mode == VIDEO_OUT_OUTPUT_MODE_119_88HZ || Config::GetVblankFrequency() >= 119
@@ -1,12 +1,7 @@
 #include "graphics/presentation/window/hostInput.h"

-#include "SDL_error.h"
-#include "SDL_events.h"
-#include "SDL_keyboard.h"
-#include "SDL_keycode.h"
-#include "SDL_mouse.h"
-#include "SDL_timer.h"
-#include "SDL_video.h"
+#include <SDL3/SDL.h>
+
 #include "common/assert.h"
 #include "common/emulatorConfig.h"
 #include "common/logging/log.h"
@@ -80,6 +75,7 @@ struct MouseJoystickState {
 };

 MouseJoystickState g_mouse;
+SDL_Window*        g_mouse_window = nullptr;

 std::size_t ControlFromName(std::string_view name) {
 	const auto info = std::find_if(CONTROL_INFO.begin(), CONTROL_INFO.end(),
@@ -141,9 +137,8 @@ public:
 				}
 			}

-			const bool reserved = binding.key == SDLK_ESCAPE || binding.key == SDLK_SPACE ||
-			                      binding.key == SDLK_F1 || binding.key == SDLK_F7 ||
-			                      binding.key == SDLK_F11;
+			const bool reserved = binding.key == SDLK_ESCAPE || binding.key == SDLK_F1 ||
+			                      binding.key == SDLK_F7 || binding.key == SDLK_F11;
 			if (binding.control == INVALID_CONTROL || reserved ||
 			    (binding.key == SDLK_UNKNOWN && binding.mouse_button == 0)) {
 				EXIT("Invalid input mapping: %s\n", value.c_str());
@@ -214,14 +209,14 @@ uint32_t DefaultKeyboardButton(int key_code) {
 		case SDLK_LEFT: return Controller::PAD_BUTTON_LEFT;
 		case SDLK_DOWN: return Controller::PAD_BUTTON_DOWN;
 		case SDLK_RIGHT: return Controller::PAD_BUTTON_RIGHT;
-		case SDLK_j: return Controller::PAD_BUTTON_CROSS;
-		case SDLK_i: return Controller::PAD_BUTTON_TRIANGLE;
-		case SDLK_k: return Controller::PAD_BUTTON_SQUARE;
-		case SDLK_l: return Controller::PAD_BUTTON_CIRCLE;
-		case SDLK_q: return Controller::PAD_BUTTON_L1;
-		case SDLK_e: return Controller::PAD_BUTTON_R1;
-		case SDLK_z: return Controller::PAD_BUTTON_L2;
-		case SDLK_c: return Controller::PAD_BUTTON_R2;
+		case SDLK_J: return Controller::PAD_BUTTON_CROSS;
+		case SDLK_I: return Controller::PAD_BUTTON_TRIANGLE;
+		case SDLK_K: return Controller::PAD_BUTTON_SQUARE;
+		case SDLK_L: return Controller::PAD_BUTTON_CIRCLE;
+		case SDLK_Q: return Controller::PAD_BUTTON_L1;
+		case SDLK_E: return Controller::PAD_BUTTON_R1;
+		case SDLK_Z: return Controller::PAD_BUTTON_L2;
+		case SDLK_C: return Controller::PAD_BUTTON_R2;
 		case SDLK_LSHIFT: return Controller::PAD_BUTTON_L3;
 		case SDLK_LCTRL: return Controller::PAD_BUTTON_R3;
 		case SDLK_RETURN:
@@ -285,35 +280,35 @@ void DefaultKeyboardInput(int key_code, bool down) {
 	switch (NormalizeKey(static_cast<SDL_Keycode>(key_code))) {
 		case SDLK_BACKSPACE: SetTouchPad(0.25f, down); return;
 		case SDLK_TAB: SetTouchPad(0.75f, down); return;
-		case SDLK_a:
+		case SDLK_A:
 			left.left = down;
 			SetStickAxis(Controller::Axis::LeftX, left.left, left.right);
 			return;
-		case SDLK_d:
+		case SDLK_D:
 			left.right = down;
 			SetStickAxis(Controller::Axis::LeftX, left.left, left.right);
 			return;
-		case SDLK_w:
+		case SDLK_W:
 			left.up = down;
 			SetStickAxis(Controller::Axis::LeftY, left.up, left.down);
 			return;
-		case SDLK_s:
+		case SDLK_S:
 			left.down = down;
 			SetStickAxis(Controller::Axis::LeftY, left.up, left.down);
 			return;
-		case SDLK_f:
+		case SDLK_F:
 			right.left = down;
 			SetStickAxis(Controller::Axis::RightX, right.left, right.right);
 			return;
-		case SDLK_h:
+		case SDLK_H:
 			right.right = down;
 			SetStickAxis(Controller::Axis::RightX, right.left, right.right);
 			return;
-		case SDLK_t:
+		case SDLK_T:
 			right.up = down;
 			SetStickAxis(Controller::Axis::RightY, right.up, right.down);
 			return;
-		case SDLK_g:
+		case SDLK_G:
 			right.down = down;
 			SetStickAxis(Controller::Axis::RightY, right.up, right.down);
 			return;
@@ -321,11 +316,11 @@ void DefaultKeyboardInput(int key_code, bool down) {
 	}
 }

-void MouseToJoystick(int delta_x, int delta_y) {
+void MouseToJoystick(float delta_x, float delta_y) {
 	const double distance = std::hypot(delta_x, delta_y);
 	const double scale =
 	    std::clamp(distance * GetInputMap().MouseSensitivity() + 16.0, 64.0, 128.0) / distance;
-	const auto map_axis = [scale](int delta) {
+	const auto map_axis = [scale](float delta) {
 		return std::clamp(128 + static_cast<int>(std::lround(delta * scale)), 0, 255);
 	};
 	Controller::SetRightStick(Controller::HOST_INPUT_CONTROLLER_ID, map_axis(delta_x),
@@ -341,16 +336,45 @@ void CenterMouseStick() {
 }

 bool SetRelativeMouseMode(bool enabled) {
-	if (SDL_SetRelativeMouseMode(enabled ? SDL_TRUE : SDL_FALSE) == 0) {
+	if (SDL_SetWindowRelativeMouseMode(g_mouse_window, enabled)) {
 		return true;
 	}
 	LOGF("Mouse-to-joystick relative mode failed: %s\n", SDL_GetError());
 	return false;
 }
+
+int PollMouse(uint64_t now_ms) {
+	if (now_ms < g_mouse.next_poll) {
+		return static_cast<int>(g_mouse.next_poll - now_ms);
+	}
+	g_mouse.next_poll = now_ms + MOUSE_POLL_INTERVAL_MS;
+
+	float delta_x = 0;
+	float delta_y = 0;
+	SDL_GetRelativeMouseState(&delta_x, &delta_y);
+	if (delta_x == 0 && delta_y == 0) {
+		CenterMouseStick();
+	} else {
+		MouseToJoystick(delta_x, delta_y);
+		g_mouse.output = true;
+	}
+	return MOUSE_POLL_INTERVAL_MS;
+}
+
 } // namespace

-void HostInputInit() {
+void HostInputInit(SDL_Window* window) {
 	GetInputMap();
+	g_mouse_window = window;
+}
+
+void HostInputShutdown() {
+	if (g_mouse.enabled) {
+		SetRelativeMouseMode(false);
+		CenterMouseStick();
+		g_mouse = {};
+	}
+	g_mouse_window = nullptr;
 }

 void HostInputKey(int key_code, bool down) {
@@ -381,51 +405,33 @@ void HostInputToggleMouseToJoystick() {
 	if (!SetRelativeMouseMode(true)) {
 		return;
 	}
-	int ignored_x = 0;
-	int ignored_y = 0;
-	SDL_GetRelativeMouseState(&ignored_x, &ignored_y);
-	g_mouse.enabled   = true;
-	g_mouse.next_poll = SDL_GetTicks64() + MOUSE_POLL_INTERVAL_MS;
+	g_mouse.enabled = true;
 	LOGF("Mouse to right stick: enabled (F7 to release)\n");
 }

-int PollMouse(uint64_t now_ms) {
-	if (now_ms < g_mouse.next_poll) {
-		return static_cast<int>(g_mouse.next_poll - now_ms);
-	}
-	g_mouse.next_poll = now_ms + MOUSE_POLL_INTERVAL_MS;
-
-	int delta_x = 0;
-	int delta_y = 0;
-	SDL_GetRelativeMouseState(&delta_x, &delta_y);
-	if (delta_x == 0 && delta_y == 0) {
-		CenterMouseStick();
-		return MOUSE_POLL_INTERVAL_MS;
-	}
-
-	MouseToJoystick(delta_x, delta_y);
-	g_mouse.output = true;
-	return MOUSE_POLL_INTERVAL_MS;
-}
-
 bool HostInputWaitEvent(SDL_Event* event) {
-	if (!g_mouse.enabled || SDL_GetKeyboardFocus() == nullptr) {
+	bool has_event;
+	if (!g_mouse.enabled || SDL_GetKeyboardFocus() != g_mouse_window) {
+		g_mouse.next_poll = 0;
 		CenterMouseStick();
-		if (SDL_WaitEvent(event) == 0) {
+		has_event = SDL_WaitEvent(event);
+		if (!has_event) {
 			EXIT("%s\n", SDL_GetError());
 		}
-		return true;
+	} else {
+		if (g_mouse.next_poll == 0) {
+			SDL_GetRelativeMouseState(nullptr, nullptr);
+			g_mouse.next_poll = SDL_GetTicks() + MOUSE_POLL_INTERVAL_MS;
+		}
+		has_event = SDL_WaitEventTimeout(event, PollMouse(SDL_GetTicks()));
 	}

-	const int timeout_ms = PollMouse(SDL_GetTicks64());
-	SDL_ClearError();
-	if (SDL_WaitEventTimeout(event, timeout_ms) != 0) {
-		return true;
+	if (has_event && event->type == SDL_EVENT_WINDOW_FOCUS_LOST &&
+	    event->window.windowID == SDL_GetWindowID(g_mouse_window)) {
+		g_mouse.next_poll = 0;
+		CenterMouseStick();
 	}
-	if (SDL_GetError()[0] != '\0') {
-		EXIT("%s\n", SDL_GetError());
-	}
-	return false;
+	return has_event;
 }

 } // namespace Libs::Graphics
@@ -3,11 +3,13 @@

 #include <cstdint>

+struct SDL_Window;
 union SDL_Event;

 namespace Libs::Graphics {

-void               HostInputInit();
+void               HostInputInit(SDL_Window* window);
+void               HostInputShutdown();
 void               HostInputKey(int key_code, bool down);
 void               HostInputMouseButton(uint8_t mouse_button, bool down);
 void               HostInputToggleMouseToJoystick();
@@ -22,9 +22,6 @@

 // IWYU pragma: no_include <intrin.h>

-#define KYTY_ENABLE_DEBUG_PRINTF
-#define KYTY_DBG_INPUT
-
 namespace Libs::Graphics {

 struct Presenter::Frame {
@@ -174,6 +171,13 @@ void Presenter::Frame::Configure(GraphicContext& graphics, vk::Extent2D extent,
 		EXIT("unsupported prepared frame, extent=%ux%u format=%d\n", extent.width, extent.height,
 		     static_cast<int>(format));
 	}
+	auto&      dst        = image;
+	const bool compatible = dst.image != nullptr && dst.extent.width == extent.width &&
+	                        dst.extent.height == extent.height && dst.format == format;
+	if (compatible) {
+		return;
+	}
+
 	const auto features = graphics.GetFormatProperties(format).optimalTilingFeatures;
 	const auto required =
 	    vk::FormatFeatureFlagBits::eBlitSrc | vk::FormatFeatureFlagBits::eSampledImageFilterLinear |
@@ -183,12 +187,6 @@ void Presenter::Frame::Configure(GraphicContext& graphics, vk::Extent2D extent,
 		     static_cast<int>(format), static_cast<vk::FormatFeatureFlags::MaskType>(features));
 	}

-	auto&      dst        = image;
-	const bool compatible = dst.image != nullptr && dst.extent.width == extent.width &&
-	                        dst.extent.height == extent.height && dst.format == format;
-	if (compatible) {
-		return;
-	}
 	if (dst.image != nullptr) {
 		graphics.DeleteImage(dst);
 	}
@@ -282,6 +280,7 @@ public:

 	void                 Create();
 	void                 Recreate(bool surface_lost = false);
+	[[nodiscard]] bool   NeedsResize() const;
 	[[nodiscard]] Status AcquireNextImage();
 	[[nodiscard]] bool   PrepareSystemOverlay();
 	void                 RecordPresentCommands(CommandBuffer& command, VulkanImage& source,
@@ -301,6 +300,8 @@ private:
 	vk::SwapchainKHR            m_handle = nullptr;
 	vk::Format                  m_format = vk::Format::eUndefined;
 	vk::Extent2D                m_extent {};
+	// Drawable pixel size observed when this swapchain was created.
+	vk::Extent2D                m_window_extent {};
 	std::vector<vk::Image>      m_images;
 	std::vector<vk::ImageView>  m_image_views;
 	std::vector<vk::Semaphore>  m_image_acquired;
@@ -360,19 +361,25 @@ void Swapchain::Create() {
 	EXIT_IF(graphics.device == nullptr);
 	EXIT_IF(m_window.surface == nullptr);

-	Common::LockGuard lock(m_window.mutex);
-	EXIT_IF(graphics.screen_width == 0);
-	EXIT_IF(graphics.screen_height == 0);
+	// Capture the size before querying surface capabilities. If the window changes
+	// during creation, the next presentation will detect the newer size.
+	{
+		Common::LockGuard lock(m_window.mutex);
+		m_window_extent = {graphics.screen_width, graphics.screen_height};
+	}
+	EXIT_IF(m_window_extent.width == 0);
+	EXIT_IF(m_window_extent.height == 0);
+	m_window.RefreshSurfaceCapabilities();
 	const auto&       surface = m_window.surface_capabilities;
 	EXIT_NOT_IMPLEMENTED(surface.formats.empty());

 	m_extent = surface.capabilities.currentExtent;
 	if (m_extent.width == std::numeric_limits<uint32_t>::max()) {
 		m_extent.width =
-		    std::clamp(graphics.screen_width, surface.capabilities.minImageExtent.width,
+		    std::clamp(m_window_extent.width, surface.capabilities.minImageExtent.width,
 		               surface.capabilities.maxImageExtent.width);
 		m_extent.height =
-		    std::clamp(graphics.screen_height, surface.capabilities.minImageExtent.height,
+		    std::clamp(m_window_extent.height, surface.capabilities.minImageExtent.height,
 		               surface.capabilities.maxImageExtent.height);
 	}
 	uint32_t image_count = surface.capabilities.minImageCount + 1;
@@ -476,8 +483,8 @@ void Swapchain::Create() {
 		    graphics.device.createSemaphore(&semaphore_info, nullptr, &m_render_complete[i]),
 		    "create swapchain render-complete semaphore");
 	}
-	m_image_index = static_cast<uint32_t>(-1);
-	m_frame_index = 0;
+	m_image_index   = static_cast<uint32_t>(-1);
+	m_frame_index   = 0;
 }

 Swapchain::~Swapchain() {
@@ -518,11 +525,12 @@ void Swapchain::Destroy() {
 		graphics.device.destroySwapchainKHR(m_handle, nullptr);
 	}

-	m_handle      = nullptr;
-	m_format      = vk::Format::eUndefined;
-	m_extent      = {};
-	m_image_index = static_cast<uint32_t>(-1);
-	m_frame_index = 0;
+	m_handle        = nullptr;
+	m_format        = vk::Format::eUndefined;
+	m_extent        = {};
+	m_window_extent = {};
+	m_image_index   = static_cast<uint32_t>(-1);
+	m_frame_index   = 0;
 	m_images.clear();
 	m_image_views.clear();
 	m_image_acquired.clear();
@@ -535,15 +543,22 @@ void Swapchain::Recreate(bool surface_lost) {
 #if defined(__APPLE__)
 		// Surface recreation goes through SDL_Vulkan_CreateSurface, which touches the
 		// window's view/layer and must run on the main thread on macOS.
-		m_window.RunOnMainThread([this] { m_window.RecreateSurface(); });
+		EXIT_IF(!SDL_RunOnMainThread(
+		    [](void* window) { static_cast<WindowContext*>(window)->RecreateSurface(); },
+		    &m_window, true));
 #else
 		m_window.RecreateSurface();
 #endif
 	}
-	m_window.RefreshSurfaceCapabilities();
 	Create();
 }

+bool Swapchain::NeedsResize() const {
+	Common::LockGuard lock(m_window.mutex);
+	return m_window_extent.width != m_window.graphic_ctx.screen_width ||
+	       m_window_extent.height != m_window.graphic_ctx.screen_height;
+}
+
 Swapchain::Status Swapchain::AcquireNextImage() {
 	EXIT_IF(m_handle == nullptr || m_frame_index >= m_image_acquired.size());
 	m_image_index     = static_cast<uint32_t>(-1);
@@ -769,6 +784,10 @@ void Presenter::Present(Frame& frame, bool reuse) {

 	const auto overlay_visual = GetSystemOverlayVisualState();
 	auto&      swapchain  = m_impl->swapchain;
+	// Some window systems keep presenting an old swapchain after a resize.
+	if (swapchain.NeedsResize()) {
+		m_impl->RecoverSwapchain(Swapchain::Status::Recreate);
+	}
 	for (uint32_t attempt = 0; attempt < 2; attempt++) {
 		auto status = swapchain.AcquireNextImage();
 		if (status != Swapchain::Status::Success) {
@@ -1,20 +1,6 @@
-#include "SDL.h"
-#include "SDL_error.h"
-#include "SDL_events.h"
-#include "SDL_gamecontroller.h"
-#include "SDL_hints.h"
-#include "SDL_joystick.h"
-#include "SDL_keyboard.h"
-#include "SDL_keycode.h"
-#include "SDL_mouse.h"
-#include "SDL_pixels.h"
-#include "SDL_rwops.h"
-#include "SDL_stdinc.h"
-#include "SDL_surface.h"
-#include "SDL_thread.h"
-#include "SDL_touch.h"
-#include "SDL_video.h"
-#include "SDL_vulkan.h"
+#include <SDL3/SDL.h>
+#include <SDL3/SDL_vulkan.h>
+
 #include "common/assert.h"
 #include "common/common.h"
 #include "common/emulatorConfig.h"
@@ -49,7 +35,6 @@
 // IWYU pragma: no_include <intrin.h>

 #define KYTY_ENABLE_DEBUG_PRINTF
-#define KYTY_DBG_INPUT

 namespace Libs::Graphics {

@@ -253,6 +238,10 @@ static void VulkanFindPhysicalDevice(vk::Instance instance, vk::SurfaceKHR surfa
 #endif
 		}
 #if !defined(__APPLE__)
+		if (device_features2.features.depthClamp != VK_TRUE) {
+			LOGF("depthClamp is not supported\n");
+			skip_device = true;
+		}
 		if (fragment_barycentric.fragmentShaderBarycentric != VK_TRUE) {
 			LOGF("fragmentShaderBarycentric is not supported\n");
 			skip_device = true;
@@ -611,20 +600,23 @@ static vk::Device VulkanCreateDevice(GraphicContext& graphics,
 		     graphics.mesh_shader_properties.maxMeshOutputPrimitives,
 		     graphics.mesh_shader_properties.maxMeshSharedMemorySize);
 	}
-	// VulkanFindPhysicalDevice already checked the required creation features. These two
+	// VulkanFindPhysicalDevice already checked the required creation features. These
 	// requirements are specific to this creation path and are not part of device selection.
 	EXIT_NOT_IMPLEMENTED(supported_features2.features.shaderInt64 != VK_TRUE);
 	EXIT_NOT_IMPLEMENTED(supported_features2.features.vertexPipelineStoresAndAtomics != VK_TRUE);
+	EXIT_NOT_IMPLEMENTED(supported_features2.features.dualSrcBlend != VK_TRUE);
 	vk::PhysicalDeviceFeatures device_features {};
 	device_features.fragmentStoresAndAtomics = VK_TRUE;
 	device_features.samplerAnisotropy        = VK_TRUE;
 	device_features.robustBufferAccess       = VK_TRUE;
 #if !defined(__APPLE__)
 	device_features.depthBounds = VK_TRUE; // unsupported by MoltenVK
+	device_features.depthClamp  = VK_TRUE;
 #endif
 	device_features.shaderStorageImageWriteWithoutFormat = VK_TRUE;
 	device_features.shaderImageGatherExtended            = VK_TRUE;
 	device_features.independentBlend                     = VK_TRUE;
+	device_features.dualSrcBlend                         = VK_TRUE;
 	device_features.tessellationShader                   = VK_TRUE;
 	device_features.sampleRateShading                    = VK_TRUE;
 	device_features.depthBiasClamp                       = VK_TRUE;
@@ -699,25 +691,13 @@ static vk::Device VulkanCreateDevice(GraphicContext& graphics,
 	return device;
 }

-static void VulkanGetExtensions(SDL_Window* window, VulkanExtensions& r) {
-	EXIT_IF(window == nullptr);
-
+static void VulkanGetExtensions(VulkanExtensions& r) {
 	uint32_t required_extensions_count = 0;

-	auto sdl_result = SDL_Vulkan_GetInstanceExtensions(window, &required_extensions_count, nullptr);
-
-	EXIT_NOT_IMPLEMENTED(sdl_result == SDL_FALSE);
+	const char* const* extensions = SDL_Vulkan_GetInstanceExtensions(&required_extensions_count);
+	EXIT_NOT_IMPLEMENTED(extensions == nullptr);
 	EXIT_NOT_IMPLEMENTED(required_extensions_count == 0);
-
-	r.required_extensions =
-	    std::vector<const char*>(required_extensions_count); // @suppress("Ambiguous problem")
-
-	sdl_result = SDL_Vulkan_GetInstanceExtensions(window, &required_extensions_count,
-	                                              r.required_extensions.data());
-
-	EXIT_NOT_IMPLEMENTED(sdl_result == SDL_FALSE);
-	EXIT_NOT_IMPLEMENTED(required_extensions_count == 0);
-	EXIT_NOT_IMPLEMENTED(required_extensions_count != r.required_extensions.size());
+	r.required_extensions.assign(extensions, extensions + required_extensions_count);

 	r.available_extensions =
 	    EnumerateVulkan<vk::ExtensionProperties>( // @suppress("Ambiguous problem")
@@ -898,7 +878,7 @@ void WindowContext::CreateVulkan() {
 	VULKAN_HPP_DEFAULT_DISPATCHER.init(get_instance_proc_addr);

 	VulkanExtensions r;
-	VulkanGetExtensions(window, r);
+	VulkanGetExtensions(r);
 	VulkanCheckInstanceVersion();

 	vk::ApplicationInfo app_info {};
@@ -985,8 +965,8 @@ void WindowContext::CreateVulkan() {
 	}

 	vk::SurfaceKHR::CType native_surface = VK_NULL_HANDLE;
-	if (SDL_Vulkan_CreateSurface(window, static_cast<vk::Instance::CType>(graphic_ctx.instance),
-	                             &native_surface) == SDL_FALSE) {
+	if (!SDL_Vulkan_CreateSurface(window, static_cast<vk::Instance::CType>(graphic_ctx.instance),
+	                              nullptr, &native_surface)) {
 		EXIT("Could not create a Vulkan surface");
 	}
 	surface = native_surface;
@@ -1099,8 +1079,8 @@ void WindowContext::RecreateSurface() {
 		surface = nullptr;
 	}
 	vk::SurfaceKHR::CType native_surface = VK_NULL_HANDLE;
-	if (SDL_Vulkan_CreateSurface(window, static_cast<vk::Instance::CType>(graphic_ctx.instance),
-	                             &native_surface) == SDL_FALSE) {
+	if (!SDL_Vulkan_CreateSurface(window, static_cast<vk::Instance::CType>(graphic_ctx.instance),
+	                              nullptr, &native_surface)) {
 		EXIT("Could not recreate the Vulkan surface: %s\n", SDL_GetError());
 	}
 	surface = native_surface;
@@ -1136,7 +1116,7 @@ WindowContext::~WindowContext() {
 		SDL_DestroyWindow(window);
 		window = nullptr;
 	}
-	SDL_QuitSubSystem(SDL_INIT_VIDEO | SDL_INIT_GAMECONTROLLER);
+	SDL_QuitSubSystem(SDL_INIT_VIDEO | SDL_INIT_GAMEPAD);
 }

 } // namespace Libs::Graphics
@@ -1,23 +1,7 @@
 #include "graphics/presentation/window.h"

-#include "SDL.h"
-#include "SDL_error.h"
-#include "SDL_events.h"
-#include "SDL_filesystem.h"
-#include "SDL_gamecontroller.h"
-#include "SDL_hints.h"
-#include "SDL_joystick.h"
-#include "SDL_keyboard.h"
-#include "SDL_keycode.h"
-#include "SDL_mouse.h"
-#include "SDL_pixels.h"
-#include "SDL_rwops.h"
-#include "SDL_stdinc.h"
-#include "SDL_surface.h"
-#include "SDL_thread.h"
-#include "SDL_touch.h"
-#include "SDL_video.h"
-#include "SDL_vulkan.h"
+#include <SDL3/SDL.h>
+
 #include "common/assert.h"
 #include "common/common.h"
 #include "common/emulatorConfig.h"
@@ -27,6 +11,7 @@
 #include "common/systemInfo.h"
 #include "common/threads.h"
 #include "common/timer.h"
+#include "common/stringUtils.h"
 #include "graphics/host_gpu/graphicContext.h"
 #include "graphics/host_gpu/renderer/render.h"
 #include "graphics/host_gpu/renderer/renderContext.h"
@@ -45,6 +30,7 @@
 #include <string>
 #include <vector>
 #include <vulkan/vk_platform.h>
+#include <filesystem>

 #define STB_IMAGE_IMPLEMENTATION
 #define STBI_NO_SIMD
@@ -52,7 +38,6 @@

 // IWYU pragma: no_include <intrin.h>

-#define KYTY_ENABLE_DEBUG_PRINTF
 #define KYTY_DBG_INPUT

 namespace Libs::Graphics {
@@ -71,39 +56,39 @@ struct EventKeyboard {

 static uint32_t ControllerButtonToPadButton(int button) {
 	switch (button) {
-		case SDL_CONTROLLER_BUTTON_A: return Controller::PAD_BUTTON_CROSS;
-		case SDL_CONTROLLER_BUTTON_B: return Controller::PAD_BUTTON_CIRCLE;
-		case SDL_CONTROLLER_BUTTON_X: return Controller::PAD_BUTTON_SQUARE;
-		case SDL_CONTROLLER_BUTTON_Y: return Controller::PAD_BUTTON_TRIANGLE;
-		case SDL_CONTROLLER_BUTTON_START: return Controller::PAD_BUTTON_OPTIONS;
-		case SDL_CONTROLLER_BUTTON_LEFTSTICK: return Controller::PAD_BUTTON_L3;
-		case SDL_CONTROLLER_BUTTON_RIGHTSTICK: return Controller::PAD_BUTTON_R3;
-		case SDL_CONTROLLER_BUTTON_LEFTSHOULDER: return Controller::PAD_BUTTON_L1;
-		case SDL_CONTROLLER_BUTTON_RIGHTSHOULDER: return Controller::PAD_BUTTON_R1;
-		case SDL_CONTROLLER_BUTTON_DPAD_UP: return Controller::PAD_BUTTON_UP;
-		case SDL_CONTROLLER_BUTTON_DPAD_DOWN: return Controller::PAD_BUTTON_DOWN;
-		case SDL_CONTROLLER_BUTTON_DPAD_LEFT: return Controller::PAD_BUTTON_LEFT;
-		case SDL_CONTROLLER_BUTTON_DPAD_RIGHT: return Controller::PAD_BUTTON_RIGHT;
-		case SDL_CONTROLLER_BUTTON_TOUCHPAD: return Controller::PAD_BUTTON_TOUCH_PAD;
+		case SDL_GAMEPAD_BUTTON_SOUTH: return Controller::PAD_BUTTON_CROSS;
+		case SDL_GAMEPAD_BUTTON_EAST: return Controller::PAD_BUTTON_CIRCLE;
+		case SDL_GAMEPAD_BUTTON_WEST: return Controller::PAD_BUTTON_SQUARE;
+		case SDL_GAMEPAD_BUTTON_NORTH: return Controller::PAD_BUTTON_TRIANGLE;
+		case SDL_GAMEPAD_BUTTON_START: return Controller::PAD_BUTTON_OPTIONS;
+		case SDL_GAMEPAD_BUTTON_LEFT_STICK: return Controller::PAD_BUTTON_L3;
+		case SDL_GAMEPAD_BUTTON_RIGHT_STICK: return Controller::PAD_BUTTON_R3;
+		case SDL_GAMEPAD_BUTTON_LEFT_SHOULDER: return Controller::PAD_BUTTON_L1;
+		case SDL_GAMEPAD_BUTTON_RIGHT_SHOULDER: return Controller::PAD_BUTTON_R1;
+		case SDL_GAMEPAD_BUTTON_DPAD_UP: return Controller::PAD_BUTTON_UP;
+		case SDL_GAMEPAD_BUTTON_DPAD_DOWN: return Controller::PAD_BUTTON_DOWN;
+		case SDL_GAMEPAD_BUTTON_DPAD_LEFT: return Controller::PAD_BUTTON_LEFT;
+		case SDL_GAMEPAD_BUTTON_DPAD_RIGHT: return Controller::PAD_BUTTON_RIGHT;
+		case SDL_GAMEPAD_BUTTON_TOUCHPAD: return Controller::PAD_BUTTON_TOUCH_PAD;
 		default: return 0;
 	}
 }

 static Controller::Axis ControllerAxisFromSdl(int axis_id) {
 	switch (axis_id) {
-		case SDL_CONTROLLER_AXIS_LEFTX: return Controller::Axis::LeftX;
-		case SDL_CONTROLLER_AXIS_LEFTY: return Controller::Axis::LeftY;
-		case SDL_CONTROLLER_AXIS_RIGHTX: return Controller::Axis::RightX;
-		case SDL_CONTROLLER_AXIS_RIGHTY: return Controller::Axis::RightY;
-		case SDL_CONTROLLER_AXIS_TRIGGERLEFT: return Controller::Axis::TriggerLeft;
-		case SDL_CONTROLLER_AXIS_TRIGGERRIGHT: return Controller::Axis::TriggerRight;
+		case SDL_GAMEPAD_AXIS_LEFTX: return Controller::Axis::LeftX;
+		case SDL_GAMEPAD_AXIS_LEFTY: return Controller::Axis::LeftY;
+		case SDL_GAMEPAD_AXIS_RIGHTX: return Controller::Axis::RightX;
+		case SDL_GAMEPAD_AXIS_RIGHTY: return Controller::Axis::RightY;
+		case SDL_GAMEPAD_AXIS_LEFT_TRIGGER: return Controller::Axis::TriggerLeft;
+		case SDL_GAMEPAD_AXIS_RIGHT_TRIGGER: return Controller::Axis::TriggerRight;
 		default: return Controller::Axis::AxisMax;
 	}
 }

 static bool ControllerAxisIsTrigger(int axis_id) {
-	return axis_id == SDL_CONTROLLER_AXIS_TRIGGERLEFT ||
-	       axis_id == SDL_CONTROLLER_AXIS_TRIGGERRIGHT;
+	return axis_id == SDL_GAMEPAD_AXIS_LEFT_TRIGGER ||
+	       axis_id == SDL_GAMEPAD_AXIS_RIGHT_TRIGGER;
 }

 static int ControllerAxisValueFromSdl(int axis_id, int axis_value) {
@@ -164,28 +149,6 @@ struct EventController {
 	double timestamp_seconds;
 };

-enum class DisplayOrientation {
-	Unknown,   /* The display orientation can't be determined */
-	Landscape, /* The display is in landscape mode, with the right side up, relative to portrait
-	              mode */
-	LandscapeFlipped, /* The display is in landscape mode, with the left side up, relative to
-	                     portrait mode */
-	Portrait,         /* The display is in portrait mode */
-	PortraitFlipped,  /* The display is in portrait mode, upside down */
-
-	DisplayEventOrientation = 0xF0
-};
-
-struct EventDisplay {
-	DisplayOrientation orientation;
-};
-
-constexpr uint32_t KYTY_SDL_BUTTON_LMASK  = SDL_BUTTON_LMASK;  // NOLINT(hicpp-signed-bitwise)
-constexpr uint32_t KYTY_SDL_BUTTON_MMASK  = SDL_BUTTON_MMASK;  // NOLINT(hicpp-signed-bitwise)
-constexpr uint32_t KYTY_SDL_BUTTON_RMASK  = SDL_BUTTON_RMASK;  // NOLINT(hicpp-signed-bitwise)
-constexpr uint32_t KYTY_SDL_BUTTON_X1MASK = SDL_BUTTON_X1MASK; // NOLINT(hicpp-signed-bitwise)
-constexpr uint32_t KYTY_SDL_BUTTON_X2MASK = SDL_BUTTON_X2MASK; // NOLINT(hicpp-signed-bitwise)
-
 namespace {

 std::unique_ptr<WindowContext> g_window;
@@ -193,7 +156,6 @@ std::unique_ptr<WindowContext> g_window;
 } // namespace

 constexpr const char* KYTY_SDL_WINDOW_CAPTION = "Game";
-constexpr int KYTY_SDL_WINDOWPOS_CENTERED = SDL_WINDOWPOS_CENTERED; /*NOLINT(hicpp-signed-bitwise)*/

 static void SetPause(WindowLoopState& game, bool flag) {
 	LOGF("Pause: %s\n", flag ? "true" : "false");
@@ -219,16 +181,13 @@ static void ToggleDesktopFullscreen() {
 	}

 	const auto flags = static_cast<uint32_t>(SDL_GetWindowFlags(g_window->window));
-	const bool fullscreen =
-	    (flags & static_cast<uint32_t>(SDL_WINDOW_FULLSCREEN_DESKTOP)) != 0u;
-	const auto mode =
-	    fullscreen ? 0u : static_cast<uint32_t>(SDL_WINDOW_FULLSCREEN_DESKTOP);
-	if (SDL_SetWindowFullscreen(g_window->window, mode) != 0) {
+	const bool fullscreen = (flags & SDL_WINDOW_FULLSCREEN) != 0u;
+	if (!SDL_SetWindowFullscreen(g_window->window, !fullscreen)) {
 		LOGF("Toggle fullscreen failed: %s\n", SDL_GetError());
 	}
 }

-static void GameEventKeyboard(WindowLoopState& game, const EventKeyboard& key) {
+static void GameEventKeyboard(const EventKeyboard& key) {
 	static SDL_Keycode fullscreen_key = SDLK_UNKNOWN;

 #ifdef KYTY_DBG_INPUT
@@ -241,7 +200,6 @@ static void GameEventKeyboard(WindowLoopState& game, const EventKeyboard& key) {
 #if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS || KYTY_PLATFORM == KYTY_PLATFORM_LINUX
 	if (key.down) {
 		switch (key.key_code) {
-			case SDLK_SPACE: SetPause(game, !game.paused.load(std::memory_order_acquire)); break;
 			case SDLK_F1:
 				if (!key.repeat) {
 					RenderDocRequestCapture();
@@ -254,7 +212,7 @@ static void GameEventKeyboard(WindowLoopState& game, const EventKeyboard& key) {
 				break;
 			case SDLK_RETURN:
 			case SDLK_KP_ENTER:
-				if (!key.repeat && (key.mod & KMOD_ALT) != 0) {
+				if (!key.repeat && (key.mod & SDL_KMOD_ALT) != 0) {
 					ToggleDesktopFullscreen();
 					fullscreen_key = key.key_code;
 				}
@@ -347,15 +305,15 @@ static void GameEventController([[maybe_unused]] const EventController& f) {
 #endif

 	if (f.added) {
-		auto* pad = SDL_GameControllerOpen(f.id);
+		auto* pad = SDL_OpenGamepad(f.id);
 		EXIT_NOT_IMPLEMENTED(pad == nullptr);
-		int id = SDL_JoystickInstanceID(SDL_GameControllerGetJoystick(pad));
+		int id = SDL_GetJoystickID(SDL_GetGamepadJoystick(pad));
 		Controller::Connect(id);
 	}

 	if (f.removed) {
 		Controller::Disconnect(f.id);
-		SDL_GameControllerClose(SDL_GameControllerFromInstanceID(f.id));
+		SDL_CloseGamepad(SDL_GetGamepadFromID(f.id));
 	}

 	if (f.down || f.up) {
@@ -407,81 +365,73 @@ void WindowContext::Resize(uint32_t new_width, uint32_t new_height) {

 void WindowContext::ProcessWindowEvent(const SDL_WindowEvent& event) {
 	const auto& window_event = event;
-	switch (window_event.event) {
-		case SDL_WINDOWEVENT_SHOWN:
+	switch (window_event.type) {
+		case SDL_EVENT_WINDOW_SHOWN:
 			LOGF("Window %" PRIu32 " shown\n", window_event.windowID);
 			break;

-		case SDL_WINDOWEVENT_HIDDEN:
+		case SDL_EVENT_WINDOW_HIDDEN:
 			LOGF("Window %" PRIu32 " hidden\n", window_event.windowID);
 			break;

-		case SDL_WINDOWEVENT_EXPOSED:
+		case SDL_EVENT_WINDOW_EXPOSED:
 			LOGF("Window %" PRIu32 " exposed\n", window_event.windowID);
 			break;

-		case SDL_WINDOWEVENT_MOVED:
+		case SDL_EVENT_WINDOW_MOVED:
 			LOGF("Window %" PRIu32 " moved to %" PRId32 ",%" PRId32 "\n", window_event.windowID,
 			     window_event.data1, window_event.data2);
 			break;

-		case SDL_WINDOWEVENT_RESIZED:
+		case SDL_EVENT_WINDOW_RESIZED:
 			LOGF("Window %" PRIu32 " resized to %" PRId32 "x%" PRId32 "\n", window_event.windowID,
 			     window_event.data1, window_event.data2);
-
-			LOGF("m: %d\n", static_cast<int>(SDL_ThreadID()));
-			Resize(window_event.data1, window_event.data2);
-
 			break;

-		case SDL_WINDOWEVENT_SIZE_CHANGED:
+		case SDL_EVENT_WINDOW_PIXEL_SIZE_CHANGED:
 			LOGF("Window %" PRIu32 " size changed to %" PRId32 "x%" PRId32 "\n",
 			     window_event.windowID, window_event.data1, window_event.data2);

-			LOGF("m: %d\n", static_cast<int>(SDL_ThreadID()));
 			Resize(window_event.data1, window_event.data2);

 			break;

-		case SDL_WINDOWEVENT_MINIMIZED:
+		case SDL_EVENT_WINDOW_MINIMIZED:
 			LOGF("Window %" PRIu32 " minimized\n", window_event.windowID);
 			break;
-		case SDL_WINDOWEVENT_MAXIMIZED:
+		case SDL_EVENT_WINDOW_MAXIMIZED:
 			LOGF("Window %" PRIu32 " maximized\n", window_event.windowID);
 			break;
-		case SDL_WINDOWEVENT_RESTORED:
+		case SDL_EVENT_WINDOW_RESTORED:
 			LOGF("Window %" PRIu32 " restored\n", window_event.windowID);
 			break;
-		case SDL_WINDOWEVENT_ENTER:
+		case SDL_EVENT_WINDOW_MOUSE_ENTER:
 			LOGF("Mouse entered window %" PRIu32 "\n", window_event.windowID);
 			break;
-		case SDL_WINDOWEVENT_LEAVE:
+		case SDL_EVENT_WINDOW_MOUSE_LEAVE:
 			LOGF("Mouse left window %" PRIu32 "\n", window_event.windowID);
 			break;
-		case SDL_WINDOWEVENT_FOCUS_GAINED:
+		case SDL_EVENT_WINDOW_FOCUS_GAINED:
 			LOGF("Window %" PRIu32 " gained keyboard focus\n", window_event.windowID);
 			break;
-		case SDL_WINDOWEVENT_FOCUS_LOST:
+		case SDL_EVENT_WINDOW_FOCUS_LOST:
 			LOGF("Window %" PRIu32 " lost keyboard focus\n", window_event.windowID);
 			break;
-		case SDL_WINDOWEVENT_CLOSE:
+		case SDL_EVENT_WINDOW_CLOSE_REQUESTED:
 			LOGF("Window %" PRIu32 " closed\n", window_event.windowID);
 			break;
 		default:
-			LOGF("Window %" PRIu32 " got unknown event %" PRIu8 "\n", window_event.windowID,
-			     window_event.event);
+			LOGF("Window %" PRIu32 " got unknown event %" PRIu32 "\n", window_event.windowID,
+			     static_cast<uint32_t>(window_event.type));
 			break;
 	}
 }

 void WindowContext::ProcessDisplayEvent(const SDL_DisplayEvent& display) {
-	bool sdl = false;
-
-	switch (display.event) {
-		case SDL_DISPLAYEVENT_ORIENTATION: sdl = true; [[fallthrough]];
-		case static_cast<Uint8>(DisplayOrientation::DisplayEventOrientation): {
-			LOGF("Display %" PRIu32 "[%s] changed orientation to %d - ", display.display,
-			     sdl ? "SDL" : "Kyty", static_cast<int>(display.data1));
+	switch (display.type) {
+		case SDL_EVENT_DISPLAY_ORIENTATION: {
+			LOGF("Display %" PRIu32 " changed orientation to %d - ", display.displayID,
+			     static_cast<int>(display.data1));

 			switch (display.data1) {
 				case SDL_ORIENTATION_UNKNOWN: LOGF("UNKNOWN\n"); break;
@@ -495,8 +445,8 @@ void WindowContext::ProcessDisplayEvent(const SDL_DisplayEvent& display) {
 			break;
 		}
 		default:
-			LOGF("Display %" PRIu32 " got unknown event 0x%" PRIx8 "\n", display.display,
-			     display.event);
+			LOGF("Display %" PRIu32 " got unknown event 0x%" PRIx32 "\n", display.displayID,
+			     static_cast<uint32_t>(display.type));
 			break;
 	}
 }
@@ -504,10 +454,9 @@ void WindowContext::ProcessDisplayEvent(const SDL_DisplayEvent& display) {
 void WindowContext::ProcessEvent(double time_s) {
 	auto& game  = loop;
 	auto* event = &game.event;
-	EXIT_IF(SDL_GetEventState(SDL_DISPLAYEVENT) != SDL_ENABLE);
-	if ((event->type == SDL_KEYDOWN || event->type == SDL_KEYUP) &&
-	    event->key.keysym.sym == SDLK_F7) {
-		if (event->type == SDL_KEYDOWN && event->key.repeat == 0) {
+	if ((event->type == SDL_EVENT_KEY_DOWN || event->type == SDL_EVENT_KEY_UP) &&
+	    event->key.key == SDLK_F7) {
+		if (event->type == SDL_EVENT_KEY_DOWN && event->key.repeat == 0) {
 			HostInputToggleMouseToJoystick();
 		}
 		return;
@@ -515,59 +464,63 @@ void WindowContext::ProcessEvent(double time_s) {
 	if (ProcessSystemOverlayInput(*event)) {
 		return;
 	}
+	if (event->type >= SDL_EVENT_WINDOW_FIRST && event->type <= SDL_EVENT_WINDOW_LAST) {
+		ProcessWindowEvent(event->window);
+		return;
+	}
+	if (event->type >= SDL_EVENT_DISPLAY_FIRST && event->type <= SDL_EVENT_DISPLAY_LAST) {
+		ProcessDisplayEvent(event->display);
+		return;
+	}

 	switch (event->type) {
-		case SDL_QUIT: GameEventQuit(game); break;
+		case SDL_EVENT_QUIT: GameEventQuit(game); break;

-		case SDL_APP_TERMINATING: GameEventTerminate(game); break;
+		case SDL_EVENT_TERMINATING: GameEventTerminate(game); break;

-		case SDL_APP_LOWMEMORY: GameEventLowMemory(); break;
+		case SDL_EVENT_LOW_MEMORY: GameEventLowMemory(); break;

-		case SDL_APP_WILLENTERBACKGROUND: GameEventWillEnterBackground(game); break;
+		case SDL_EVENT_WILL_ENTER_BACKGROUND: GameEventWillEnterBackground(game); break;

-		case SDL_APP_DIDENTERBACKGROUND: GameEventDidEnterBackground(); break;
+		case SDL_EVENT_DID_ENTER_BACKGROUND: GameEventDidEnterBackground(); break;

-		case SDL_APP_WILLENTERFOREGROUND: GameEventWillEnterForeground(); break;
+		case SDL_EVENT_WILL_ENTER_FOREGROUND: GameEventWillEnterForeground(); break;

-		case SDL_APP_DIDENTERFOREGROUND: GameEventDidEnterForeground(game); break;
+		case SDL_EVENT_DID_ENTER_FOREGROUND: GameEventDidEnterForeground(game); break;

-		case SDL_KEYDOWN:
-		case SDL_KEYUP: {
+		case SDL_EVENT_KEY_DOWN:
+		case SDL_EVENT_KEY_UP: {
 			EventKeyboard key {};

-			key.down              = (event->type == SDL_KEYDOWN);
-			key.up                = (event->type == SDL_KEYUP);
-			key.pressed           = (event->key.state == SDL_PRESSED);
-			key.released          = (event->key.state == SDL_RELEASED);
+			key.down              = (event->type == SDL_EVENT_KEY_DOWN);
+			key.up                = (event->type == SDL_EVENT_KEY_UP);
+			key.pressed           = event->key.down;
+			key.released          = !event->key.down;
 			key.repeat            = (event->key.repeat != 0u);
-			key.scan_code         = event->key.keysym.scancode;
-			key.key_code          = event->key.keysym.sym;
-			key.mod               = event->key.keysym.mod;
+			key.scan_code         = event->key.scancode;
+			key.key_code          = event->key.key;
+			key.mod               = event->key.mod;
 			key.timestamp_seconds = time_s;

-			GameEventKeyboard(game, key);
+			GameEventKeyboard(key);

 			break;
 		}

-		case SDL_WINDOWEVENT: ProcessWindowEvent(event->window); break;
-
-		case SDL_DISPLAYEVENT: ProcessDisplayEvent(event->display); break;
-
-		case SDL_MOUSEBUTTONDOWN:
-		case SDL_MOUSEBUTTONUP: {
+		case SDL_EVENT_MOUSE_BUTTON_DOWN:
+		case SDL_EVENT_MOUSE_BUTTON_UP: {
 			EventMouse mb {};

-			mb.down              = (event->button.type == SDL_MOUSEBUTTONDOWN);
-			mb.up                = (event->button.type == SDL_MOUSEBUTTONUP);
+			mb.down              = (event->button.type == SDL_EVENT_MOUSE_BUTTON_DOWN);
+			mb.up                = (event->button.type == SDL_EVENT_MOUSE_BUTTON_UP);
 			mb.left              = (event->button.button == SDL_BUTTON_LEFT);
 			mb.middle            = (event->button.button == SDL_BUTTON_MIDDLE);
 			mb.right             = (event->button.button == SDL_BUTTON_RIGHT);
 			mb.x1                = (event->button.button == SDL_BUTTON_X1);
 			mb.x2                = (event->button.button == SDL_BUTTON_X2);
 			mb.touch             = (event->button.which == SDL_TOUCH_MOUSEID);
-			mb.pressed           = (event->button.state == SDL_PRESSED);
-			mb.released          = (event->button.state == SDL_RELEASED);
+			mb.pressed           = event->button.down;
+			mb.released          = !event->button.down;
 			mb.num_of_clicks     = event->button.clicks;
 			mb.wheel             = false;
 			mb.x                 = event->button.x;
@@ -582,7 +535,7 @@ void WindowContext::ProcessEvent(double time_s) {
 			break;
 		}

-		case SDL_MOUSEWHEEL: {
+		case SDL_EVENT_MOUSE_WHEEL: {
 			EventMouse mb {};

 			mb.down              = false;
@@ -609,16 +562,16 @@ void WindowContext::ProcessEvent(double time_s) {
 			break;
 		}

-		case SDL_MOUSEMOTION: {
+		case SDL_EVENT_MOUSE_MOTION: {
 			EventMouse mb {};

 			mb.down              = false;
 			mb.up                = false;
-			mb.left              = ((event->motion.state & KYTY_SDL_BUTTON_LMASK) != 0u);
-			mb.middle            = ((event->motion.state & KYTY_SDL_BUTTON_MMASK) != 0u);
-			mb.right             = ((event->motion.state & KYTY_SDL_BUTTON_RMASK) != 0u);
-			mb.x1                = ((event->motion.state & KYTY_SDL_BUTTON_X1MASK) != 0u);
-			mb.x2                = ((event->motion.state & KYTY_SDL_BUTTON_X2MASK) != 0u);
+			mb.left              = ((event->motion.state & SDL_BUTTON_LMASK) != 0u);
+			mb.middle            = ((event->motion.state & SDL_BUTTON_MMASK) != 0u);
+			mb.right             = ((event->motion.state & SDL_BUTTON_RMASK) != 0u);
+			mb.x1                = ((event->motion.state & SDL_BUTTON_X1MASK) != 0u);
+			mb.x2                = ((event->motion.state & SDL_BUTTON_X2MASK) != 0u);
 			mb.touch             = (event->motion.which == SDL_TOUCH_MOUSEID);
 			mb.pressed           = false;
 			mb.released          = false;
@@ -636,16 +589,16 @@ void WindowContext::ProcessEvent(double time_s) {
 			break;
 		}

-		case SDL_FINGERMOTION:
-		case SDL_FINGERDOWN:
-		case SDL_FINGERUP: {
+		case SDL_EVENT_FINGER_MOTION:
+		case SDL_EVENT_FINGER_DOWN:
+		case SDL_EVENT_FINGER_UP: {
 			EventFinger f {};

-			f.down              = (event->tfinger.type == SDL_FINGERDOWN);
-			f.up                = (event->tfinger.type == SDL_FINGERUP);
-			f.motion            = (event->tfinger.type == SDL_FINGERMOTION);
-			f.finger_id         = static_cast<int>(event->tfinger.fingerId);
-			f.touch_id          = static_cast<int>(event->tfinger.touchId);
+			f.down              = (event->tfinger.type == SDL_EVENT_FINGER_DOWN);
+			f.up                = (event->tfinger.type == SDL_EVENT_FINGER_UP);
+			f.motion            = (event->tfinger.type == SDL_EVENT_FINGER_MOTION);
+			f.finger_id         = static_cast<int>(event->tfinger.fingerID);
+			f.touch_id          = static_cast<int>(event->tfinger.touchID);
 			f.x                 = event->tfinger.x;
 			f.y                 = event->tfinger.y;
 			f.dx                = event->tfinger.dx;
@@ -658,13 +611,13 @@ void WindowContext::ProcessEvent(double time_s) {
 			break;
 		}

-		case SDL_CONTROLLERAXISMOTION: {
+		case SDL_EVENT_GAMEPAD_AXIS_MOTION: {
 			EventController c {};

-			c.id                = event->caxis.which;
-			c.button            = SDL_CONTROLLER_BUTTON_INVALID;
-			c.axis_id           = event->caxis.axis;
-			c.axis_value        = event->caxis.value;
+			c.id                = event->gaxis.which;
+			c.button            = SDL_GAMEPAD_BUTTON_INVALID;
+			c.axis_id           = event->gaxis.axis;
+			c.axis_value        = event->gaxis.value;
 			c.down              = false;
 			c.up                = false;
 			c.added             = false;
@@ -680,22 +633,22 @@ void WindowContext::ProcessEvent(double time_s) {
 			break;
 		}

-		case SDL_CONTROLLERBUTTONDOWN:
-		case SDL_CONTROLLERBUTTONUP: {
+		case SDL_EVENT_GAMEPAD_BUTTON_DOWN:
+		case SDL_EVENT_GAMEPAD_BUTTON_UP: {
 			EventController c {};

-			c.id                = event->cbutton.which;
-			c.button            = event->cbutton.button;
-			c.axis_id           = SDL_CONTROLLER_AXIS_INVALID;
+			c.id                = event->gbutton.which;
+			c.button            = event->gbutton.button;
+			c.axis_id           = SDL_GAMEPAD_AXIS_INVALID;
 			c.axis_value        = 0;
-			c.down              = (event->cbutton.type == SDL_CONTROLLERBUTTONDOWN);
-			c.up                = (event->cbutton.type == SDL_CONTROLLERBUTTONUP);
+			c.down              = event->gbutton.down;
+			c.up                = !event->gbutton.down;
 			c.added             = false;
 			c.removed           = false;
 			c.remapped          = false;
 			c.axis              = false;
-			c.pressed           = (event->cbutton.state == SDL_PRESSED);
-			c.released          = (event->cbutton.state == SDL_RELEASED);
+			c.pressed           = event->gbutton.down;
+			c.released          = !event->gbutton.down;
 			c.timestamp_seconds = time_s;

 			GameEventController(c);
@@ -703,30 +656,44 @@ void WindowContext::ProcessEvent(double time_s) {
 			break;
 		}

-		case SDL_CONTROLLERTOUCHPADDOWN:
-		case SDL_CONTROLLERTOUCHPADMOTION:
-		case SDL_CONTROLLERTOUCHPADUP:
-			if (event->ctouchpad.touchpad == 0) {
-				Controller::SetTouchPad(event->ctouchpad.which, event->ctouchpad.finger,
-				                        event->ctouchpad.type != SDL_CONTROLLERTOUCHPADUP,
-				                        event->ctouchpad.x, event->ctouchpad.y);
+		case SDL_EVENT_GAMEPAD_TOUCHPAD_DOWN:
+		case SDL_EVENT_GAMEPAD_TOUCHPAD_MOTION:
+		case SDL_EVENT_GAMEPAD_TOUCHPAD_UP:
+			if (event->gtouchpad.touchpad == 0) {
+				Controller::SetTouchPad(event->gtouchpad.which, event->gtouchpad.finger,
+				                        event->gtouchpad.type != SDL_EVENT_GAMEPAD_TOUCHPAD_UP,
+				                        event->gtouchpad.x, event->gtouchpad.y);
 			}
 			break;

-		case SDL_CONTROLLERDEVICEADDED:
-		case SDL_CONTROLLERDEVICEREMOVED:
-		case SDL_CONTROLLERDEVICEREMAPPED: {
+		case SDL_EVENT_GAMEPAD_SENSOR_UPDATE: {
+			const auto& sensor = event->gsensor;
+			if (sensor.sensor == SDL_SENSOR_ACCEL || sensor.sensor == SDL_SENSOR_GYRO) {
+				Controller::SetSensor(sensor.which,
+				                      sensor.sensor == SDL_SENSOR_ACCEL ? Controller::Sensor::Accel
+				                                                        : Controller::Sensor::Gyro,
+				                      sensor.data,
+				                      sensor.sensor_timestamp != 0
+				                          ? sensor.sensor_timestamp / 1000
+				                          : sensor.timestamp / 1000);
+			}
+			break;
+		}
+
+		case SDL_EVENT_GAMEPAD_ADDED:
+		case SDL_EVENT_GAMEPAD_REMOVED:
+		case SDL_EVENT_GAMEPAD_REMAPPED: {
 			EventController c {};

-			c.id                = event->cdevice.which;
-			c.button            = SDL_CONTROLLER_BUTTON_INVALID;
-			c.axis_id           = SDL_CONTROLLER_AXIS_INVALID;
+			c.id                = event->gdevice.which;
+			c.button            = SDL_GAMEPAD_BUTTON_INVALID;
+			c.axis_id           = SDL_GAMEPAD_AXIS_INVALID;
 			c.axis_value        = 0;
 			c.down              = false;
 			c.up                = false;
-			c.added             = (event->cdevice.type == SDL_CONTROLLERDEVICEADDED);
-			c.removed           = (event->cdevice.type == SDL_CONTROLLERDEVICEREMOVED);
-			c.remapped          = (event->cdevice.type == SDL_CONTROLLERDEVICEREMAPPED);
+			c.added             = (event->gdevice.type == SDL_EVENT_GAMEPAD_ADDED);
+			c.removed           = (event->gdevice.type == SDL_EVENT_GAMEPAD_REMOVED);
+			c.remapped          = (event->gdevice.type == SDL_EVENT_GAMEPAD_REMAPPED);
 			c.axis              = false;
 			c.pressed           = false;
 			c.released          = false;
@@ -739,47 +706,6 @@ void WindowContext::ProcessEvent(double time_s) {
 	}
 }

-void WindowContext::RunOnMainThread(std::function<void()> task) {
-	if (Common::Thread::IsMainThread()) {
-		task();
-		return;
-	}
-
-	uint64_t ticket = 0;
-	{
-		Common::LockGuard lock(main_task_mutex);
-		main_tasks.push_back(std::move(task));
-		ticket = ++main_tasks_queued;
-	}
-
-	// Wake the main loop in case it is blocked in SDL_WaitEvent.
-	SDL_Event event {};
-	event.type = SDL_USEREVENT;
-	SDL_PushEvent(&event);
-
-	Common::LockGuard lock(main_task_mutex);
-	while (main_tasks_run < ticket) {
-		main_task_done.Wait(&main_task_mutex);
-	}
-}
-
-void WindowContext::DrainMainThreadTasks() {
-	std::vector<std::function<void()>> tasks;
-	{
-		Common::LockGuard lock(main_task_mutex);
-		tasks.swap(main_tasks);
-	}
-	if (tasks.empty()) {
-		return;
-	}
-	for (auto& task: tasks) {
-		task();
-	}
-	Common::LockGuard lock(main_task_mutex);
-	main_tasks_run += tasks.size();
-	main_task_done.SignalAll();
-}
-
 void WindowContext::Run() {
 	Common::Timer timer;
 	timer.Start();
@@ -789,7 +715,6 @@ void WindowContext::Run() {
 	loop.paused.store(false, std::memory_order_release);

 	while (!loop.need_exit) {
-		DrainMainThreadTasks();
 		if (loop.paused.load(std::memory_order_acquire)) {
 			if (!timer.IsPaused()) {
 				timer.Pause();
@@ -813,17 +738,15 @@ static void WindowCreate(WindowContext& context) {
 	int width  = static_cast<int>(context.graphic_ctx.screen_width);
 	int height = static_cast<int>(context.graphic_ctx.screen_height);

-	SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_PS4_RUMBLE, "1");
-	SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_PS5_RUMBLE, "1");
-#if KYTY_PLATFORM == KYTY_PLATFORM_WINDOWS
-	SDL_SetHint(SDL_HINT_WINDOWS_DPI_SCALING, "0");
+#if KYTY_PLATFORM == KYTY_PLATFORM_LINUX
+	// RenderDoc can hide Wayland Vulkan surface support; prefer X11 for captures.
+	if (Config::RenderDocEnabled()) {
+		SDL_SetHint(SDL_HINT_VIDEO_DRIVER, "x11,wayland");
+	}
 #endif
-
-	if (SDL_InitSubSystem(SDL_INIT_VIDEO | SDL_INIT_GAMECONTROLLER) < 0) {
+	if (!SDL_InitSubSystem(SDL_INIT_VIDEO | SDL_INIT_GAMEPAD)) {
 		EXIT("%s\n", SDL_GetError());
 	}
-	HostInputInit();
-	InitializeSystemOverlayInput();

 	LOGF("WindowCreate(): width = %d, height = %d\n", width, height);

@@ -833,15 +756,14 @@ static void WindowCreate(WindowContext& context) {
 	// MoltenVK loader before calling SDL_CreateWindow. Keep an explicit user
 	// override, and fall back to SDL's normal loader search when no bundle is present.
 	if (std::getenv("SDL_VULKAN_LIBRARY") == nullptr) {
-		if (char* base_path = SDL_GetBasePath(); base_path != nullptr) {
+		if (const char* base_path = SDL_GetBasePath(); base_path != nullptr) {
 			const std::string base_path_str = base_path;
-			SDL_free(base_path);
 			std::string moltenvk_path = base_path_str + "libMoltenVK.dylib";
 			if (!Common::File::IsFileExisting(moltenvk_path)) {
 				moltenvk_path = base_path_str + "../Frameworks/libMoltenVK.dylib";
 			}
 			if (Common::File::IsFileExisting(moltenvk_path) &&
-			    SDL_setenv("SDL_VULKAN_LIBRARY", moltenvk_path.c_str(), 0) == 0) {
+			    SDL_setenv_unsafe("SDL_VULKAN_LIBRARY", moltenvk_path.c_str(), 0) == 0) {
 				LOGF("Vulkan loader: %s\n", moltenvk_path.c_str());
 			}
 		}
@@ -854,21 +776,22 @@ static void WindowCreate(WindowContext& context) {
 		window_flags |= static_cast<uint32_t>(SDL_WINDOW_BORDERLESS);
 	}
 #endif
-	context.window = SDL_CreateWindow(KYTY_SDL_WINDOW_CAPTION, KYTY_SDL_WINDOWPOS_CENTERED,
-	                                  KYTY_SDL_WINDOWPOS_CENTERED, width, height, window_flags);
+	context.window = SDL_CreateWindow(KYTY_SDL_WINDOW_CAPTION, width, height, window_flags);

 	if (context.window == nullptr) {
 		EXIT("%s\n", SDL_GetError());
 	}

-	SDL_SetWindowResizable(context.window, SDL_FALSE);
+	SDL_SetWindowPosition(context.window, SDL_WINDOWPOS_CENTERED, SDL_WINDOWPOS_CENTERED);
+	HostInputInit(context.window);
+	InitializeSystemOverlayInput(context.window);
 	context.UpdateIcon();
 }

 uint32_t WindowContext::InitialWindowFlags(bool fullscreen) noexcept {
-	auto flags = static_cast<uint32_t>(SDL_WINDOW_VULKAN);
+	auto flags = static_cast<uint32_t>(SDL_WINDOW_VULKAN | SDL_WINDOW_RESIZABLE);
 	if (fullscreen) {
-		flags |= static_cast<uint32_t>(SDL_WINDOW_FULLSCREEN_DESKTOP);
+		flags |= static_cast<uint32_t>(SDL_WINDOW_FULLSCREEN);
 	}
 	return flags;
 }
@@ -899,6 +822,7 @@ void WindowRun() {

 void WindowShutdown() {
 	if (g_window != nullptr) {
+		HostInputShutdown();
 		Controller::EmergencyShutdown();
 		g_window.reset();
 	}
@@ -934,15 +858,15 @@ struct WindowIcon {
 	void*        pixels  = nullptr;

 	~WindowIcon() {
-		SDL_FreeSurface(surface);
+		SDL_DestroySurface(surface);
 		stbi_image_free(pixels);
 	}
 };

-static void WindowLoadPngIcon(const std::string& path, WindowIcon* icon) {
+static void WindowLoadPngIcon(const std::filesystem::path& path, WindowIcon* icon) {
 	Common::File f;
 	if (!f.Open(path, Common::File::Mode::Read)) {
-		EXIT("Can't open icon file %s\n", path.c_str());
+		EXIT("Can't open icon file %s\n", Common::PathToString(path).c_str());
 	}

 	int width  = 0;
@@ -958,8 +882,8 @@ static void WindowLoadPngIcon(const std::string& path, WindowIcon* icon) {

 	EXIT_IF(icon->pixels == nullptr);

-	icon->surface = SDL_CreateRGBSurfaceWithFormatFrom(icon->pixels, width, height, 32, width * 4,
-	                                                   SDL_PIXELFORMAT_RGBA32);
+	icon->surface = SDL_CreateSurfaceFrom(width, height, SDL_PIXELFORMAT_RGBA32, icon->pixels,
+	                                      width * 4);
 	EXIT_NOT_IMPLEMENTED(icon->surface == nullptr);
 }

@@ -968,7 +892,7 @@ void WindowContext::UpdateIcon() {
 	static bool       icon_loaded = false;

 	if (!icon_loaded) {
-		std::string icon_path;
+		std::filesystem::path icon_path;
 		if (Loader::SystemContentGetIconPath(&icon_path)) {
 			WindowLoadPngIcon(icon_path, &icon);
 		}
@@ -1021,7 +945,16 @@ void WindowContext::UpdateTitle() {
 	    (has_title_id ? ", " : ""), (has_app_ver ? app_ver : ""), (has_app_ver ? " " : ""),
 	    device_name, processor_name, frame_num, current_fps);

-	RunOnMainThread([this, text = std::move(text)] { SDL_SetWindowTitle(window, text.c_str()); });
+	struct TitleUpdate {
+		SDL_Window*  window;
+		std::string* text;
+	} update {window, &text};
+	EXIT_IF(!SDL_RunOnMainThread(
+	    [](void* data) {
+		    auto& title = *static_cast<TitleUpdate*>(data);
+		    SDL_SetWindowTitle(title.window, title.text->c_str());
+	    },
+	    &update, true));
 }

 } // namespace Libs::Graphics
@@ -1,15 +1,14 @@
 #ifndef EMULATOR_SRC_GRAPHICS_PRESENTATION_WINDOW_WINDOWINTERNAL_H_
 #define EMULATOR_SRC_GRAPHICS_PRESENTATION_WINDOW_WINDOWINTERNAL_H_

-#include "SDL_events.h"
-#include "SDL_video.h"
+#include <SDL3/SDL.h>
+
 #include "common/threads.h"
 #include "graphics/host_gpu/graphicContext.h"
 #include "graphics/host_gpu/vulkanCommon.h"

 #include <atomic>
 #include <cstdint>
-#include <functional>
 #include <memory>
 #include <vector>

@@ -48,9 +47,6 @@ struct WindowContext {
 	void ProcessDisplayEvent(const SDL_DisplayEvent& event);
 	void ProcessEvent(double time_seconds);
 	void Run();
-	// SDL window operations must complete on the main thread.
-	void RunOnMainThread(std::function<void()> task);
-	void DrainMainThreadTasks();

 	GraphicContext                 graphic_ctx;
 	SDL_Window*                    window        = nullptr;
@@ -61,12 +57,6 @@ struct WindowContext {
 	WindowLoopState                loop;

 	Common::Mutex mutex;
-
-	Common::Mutex                      main_task_mutex;
-	Common::CondVar                    main_task_done;
-	std::vector<std::function<void()>> main_tasks;            // guarded by main_task_mutex
-	uint64_t                           main_tasks_queued = 0; // guarded by main_task_mutex
-	uint64_t                           main_tasks_run    = 0; // guarded by main_task_mutex
 };

 } // namespace Libs::Graphics
@@ -20,6 +20,7 @@

 #include <algorithm>
 #include <array>
+#include <atomic>
 #include <chrono>
 #include <fmt/format.h>
 #include <map>
@@ -105,8 +106,6 @@ void ClearEmbeddedFetchVectorLanes(EmbeddedFetchVectorLanes* lanes, uint32_t reg
 	lanes->erase(first, last);
 }

-using EmbeddedFetchData = Frontend::EmbeddedFetchPlan;
-
 bool IsDecodedSgpr(const Decoder::Operand& op) {
 	return op.kind == Decoder::OperandKind::Sgpr || op.kind == Decoder::OperandKind::VccLo ||
 	       op.kind == Decoder::OperandKind::VccHi;
@@ -210,13 +209,12 @@ int BufferTableAttribFromOffset(uint32_t raw_offset, int dword) {
 	return static_cast<int>((raw_offset + static_cast<uint32_t>(dword) * 4u) / 16u);
 }

-EmbeddedFetchData DetectEmbeddedVertexFetch(const Decoder::Program&      decoded,
-                                            const ShaderVertexInputInfo* input_info,
-                                            uint32_t user_data_base, uint32_t user_data_count,
-                                            uint32_t wave_size) {
+Frontend::EmbeddedFetchPlan DetectEmbeddedVertexFetch(
+    const Decoder::Program& decoded, const ShaderVertexInputInfo* input_info,
+    uint32_t user_data_base, uint32_t user_data_count, uint32_t wave_size) {
 	const uint32_t    vertex_index_reg   = input_info->logical_stage == ShaderType::Local ? 2u : 5u;
 	const uint32_t    instance_index_reg = input_info->logical_stage == ShaderType::Local ? 5u : 8u;
-	EmbeddedFetchData data;
+	Frontend::EmbeddedFetchPlan data;
 	data.loads.reserve(input_info->resources_num);
 	int32_t vertex_offset_candidate   = -1;
 	int32_t instance_offset_candidate = -1;
@@ -526,6 +524,20 @@ TranslateResult TranslateProgram(std::span<const uint32_t> code, const CompileOp
 	     GetDumpLabel(options), StageName(options.stage), options.shader_hash,
 	     static_cast<uint64_t>(decoded.instructions.size()), phase_ms());

+	// Temporary workaround for games that compile ray-tracing shaders before
+	// the player can select a mode without ray tracing.
+	if (options.stage == ShaderType::Compute && decoded.has_bvh) {
+		static std::atomic_flag warned = ATOMIC_FLAG_INIT;
+		if (!warned.test_and_set(std::memory_order_relaxed)) {
+			const auto& bvh = decoded.instructions.back();
+			Log::WriteToConsoleAndLog(fmt::format(
+			    "Warning: ray tracing is not implemented; skipping compute dispatches containing "
+			    "BVH intersection instructions (shader=0x{:016x}, pc=0x{:08x}, opcode=0x{:02x}).\n",
+			    options.shader_hash, bvh.pc, bvh.opcode_id));
+		}
+		return {.skip_dispatch = true};
+	}
+
 	std::string decoded_dump;
 	if (options.dump_ir) {
 		decoded_dump = Decoder::ProgramToString(decoded);
@@ -560,7 +572,7 @@ TranslateResult TranslateProgram(std::span<const uint32_t> code, const CompileOp
 		     static_cast<uint64_t>(cfg.natural_loops.size()), phase_ms());
 	}

-	EmbeddedFetchData embedded_fetch;
+	Frontend::EmbeddedFetchPlan embedded_fetch;
 	if ((options.stage == ShaderType::Vertex || options.stage == ShaderType::Local) &&
 	    options.input_info.vertex != nullptr && options.input_info.vertex->fetch_embedded) {
 		embedded_fetch = DetectEmbeddedVertexFetch(
@@ -618,6 +630,7 @@ TranslateResult TranslateProgram(std::span<const uint32_t> code, const CompileOp
 CompileResult CompileProgram(TranslateResult translated, const CompileOptions& options,
                             const IR::ResourceSpecialization& specialization,
                             uint32_t push_data_start_dword) {
+	EXIT_IF(translated.skip_dispatch);
 	const auto emit_begin = std::chrono::steady_clock::now();
 	auto& ir = translated.program;
 	IR::ApplyResourceSpecialization(ir, specialization);
@@ -28,6 +28,7 @@ struct TranslateResult {
 	IR::Program program;
 	std::string decoded_dump;
 	std::string cfg_dump;
+	bool        skip_dispatch = false;
 };

 struct CompileResult {
@@ -132,12 +132,23 @@ void ValidateNativeProgram(const IR::Program& program) {
 			              program.memory_info[index].planning_only;
 		       });
 	};
+	const auto indirect_buffer_handle = [&](const IR::Inst& handle) {
+		return program.info.uses_dma && handle.NumArgs() == 4u && !handle.Uses().empty() &&
+		       std::ranges::all_of(handle.Uses(), [&](const IR::Use& use) {
+			       if (IR::BufferAccessOf(use.user->GetOpcode()) != IR::BufferAccess::Read) {
+				       return false;
+			       }
+			       const auto index = use.user->Flags<IR::MemoryFlags>().index;
+			       return index < program.memory_info.size() &&
+			              program.memory_info[index].kind == IR::ResourceKind::IndirectBuffer;
+		       });
+	};
 	for (const auto* block: program.blocks) {
 		for (const auto& inst: *block) {
 			const auto dense = inst.Flags<uint32_t>();
 			switch (inst.GetOpcode()) {
 				case IR::ValueOpcode::GetBufferResource:
-					if (planning_only_handle(inst)) {
+					if (planning_only_handle(inst) || indirect_buffer_handle(inst)) {
 						break;
 					}
 					if (dense >= program.info.buffers.size()) {
@@ -212,7 +223,11 @@ Emitter::SpirvRequirements Emitter::AnalyzeProgramRequirements(const IR::Program
 					Fail(program, "buffer operation has invalid memory metadata");
 				}
 				const auto& memory = program.memory_info[memory_index];
+				if (memory.kind == IR::ResourceKind::IndirectBuffer) {
+					requirements.subgroup_local_invocation_id = true;
+				}
 				if (memory.kind == IR::ResourceKind::Buffer) {
+					requirements.coherent_buffers |= memory.coherent;
 					if (memory.resource >= program.info.buffers.size()) {
 						Fail(program, "buffer operation has invalid resource metadata");
 					}
@@ -138,7 +138,7 @@ uint32_t EmitDppWriteCondition(ValueEmitContext& ctx, const IR::DppMoveFlags& fl
 	state.builder.AddFunction(spv::OpLogicalAnd, TypeBool(state), masks_ok, bank_ok, row_ok);
 	uint32_t writable = masks_ok;
 	if (!flags.bound_control) {
-		const auto target  = EmitDppTargetLane(state, flags.control);
+		const auto target  = EmitDppTargetLane(state, flags);
 		const auto bounded = state.builder.AllocateId();
 		state.builder.AddFunction(spv::OpLogicalAnd, TypeBool(state), bounded, writable,
 		                          target.valid);
@@ -559,6 +559,10 @@ void EmitVoid(ValueEmitContext&) {}

 void EmitBarrier(EmitterState& state) {
 	const auto tessellation = state.program.stage == ShaderType::TessellationControl;
+	if (!tessellation && ShaderWorkgroupInput(state.program.stage, state.input_info) == nullptr) {
+		// Independent graphics invocations have no native workgroup left to synchronize.
+		return;
+	}
 	const auto memory_scope = tessellation ? spv::ScopeInvocation : spv::ScopeWorkgroup;
 	const auto semantics    = tessellation ? spv::MemorySemanticsMaskNone
 	                                       : spv::MemorySemanticsAcquireReleaseMask |
@@ -611,7 +615,7 @@ uint32_t EmitUndefU1(EmitterState& state, const IR::Inst& inst) {
 uint32_t EmitDppMoveU32(ValueEmitContext& ctx, const IR::Inst& inst) {
 	auto&      state    = ctx.state;
 	const auto flags    = inst.Flags<IR::DppMoveFlags>();
-	const auto target   = EmitDppTargetLane(state, flags.control);
+	const auto target   = EmitDppTargetLane(state, flags);
 	const auto shuffled = ctx.Shuffle(inst, 0, target.lane);
 	if (flags.fetch_inactive) {
 		return shuffled;
@@ -2,24 +2,26 @@

 namespace Libs::Graphics::ShaderRecompiler::Spirv::Emitter {

-DppTargetLane EmitDppQuadPermTargetLane(EmitterState& state, uint32_t subid, uint32_t control) {
-	const auto quad_base = state.builder.AllocateId();
+DppTargetLane EmitDppPermTargetLane(EmitterState& state, uint32_t subid, uint32_t control,
+                                    uint32_t lane_bits) {
+	const auto lane_mask  = (1u << lane_bits) - 1u;
+	const auto group_base = state.builder.AllocateId();
 	const auto lane      = state.builder.AllocateId();
 	const auto shift     = state.builder.AllocateId();
 	const auto selected0 = state.builder.AllocateId();
 	const auto selected  = state.builder.AllocateId();
 	const auto target    = state.builder.AllocateId();
-	state.builder.AddFunction(spv::OpBitwiseAnd, TypeU32(state), quad_base, subid,
-	                          ConstantU32(state, 0xfffffffcu));
+	state.builder.AddFunction(spv::OpBitwiseAnd, TypeU32(state), group_base, subid,
+	                          ConstantU32(state, ~lane_mask));
 	state.builder.AddFunction(spv::OpBitwiseAnd, TypeU32(state), lane, subid,
-	                          ConstantU32(state, 3));
-	state.builder.AddFunction(spv::OpShiftLeftLogical, TypeU32(state), shift, lane,
-	                          ConstantU32(state, 1));
+	                          ConstantU32(state, lane_mask));
+	state.builder.AddFunction(spv::OpIMul, TypeU32(state), shift, lane,
+	                          ConstantU32(state, lane_bits));
 	state.builder.AddFunction(spv::OpShiftRightLogical, TypeU32(state), selected0,
 	                          ConstantU32(state, control), shift);
 	state.builder.AddFunction(spv::OpBitwiseAnd, TypeU32(state), selected, selected0,
-	                          ConstantU32(state, 3));
-	state.builder.AddFunction(spv::OpBitwiseOr, TypeU32(state), target, quad_base, selected);
+	                          ConstantU32(state, lane_mask));
+	state.builder.AddFunction(spv::OpBitwiseOr, TypeU32(state), target, group_base, selected);
 	return {target, ConstantBool(state, true)};
 }

@@ -89,10 +91,14 @@ DppTargetLane EmitDppMirrorTargetLane(EmitterState& state, uint32_t subid, bool
 	return {target, ConstantBool(state, true)};
 }

-DppTargetLane EmitDppTargetLane(EmitterState& state, uint32_t control) {
+DppTargetLane EmitDppTargetLane(EmitterState& state, const IR::DppMoveFlags& flags) {
 	const auto subid = EmitSubgroupLocalInvocationId(state);
+	const auto control = flags.control;
+	if (flags.dpp8) {
+		return EmitDppPermTargetLane(state, subid, control, 3u);
+	}
 	if (control <= 0xffu) {
-		return EmitDppQuadPermTargetLane(state, subid, control);
+		return EmitDppPermTargetLane(state, subid, control, 2u);
 	}
 	if (control >= 0x101u && control <= 0x10fu) {
 		return EmitDppRowShiftTargetLane(state, subid, control & 0xfu, true);
@@ -63,10 +63,10 @@ uint32_t AddressF32(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::
 	           : Unary(ctx.state, spv::OpBitcast, TypeF32(ctx.state), value);
 }

-ImageSampleLayout Layout(const IR::MemoryInfo& mem, ImageDimension dimension) {
+ImageSampleLayout Layout(const IR::MemoryInfo& mem) {
 	ImageSampleLayout layout;
 	uint32_t          cursor = 0;
-	const auto&       info   = ImageDimensionInfoFor(dimension);
+	const auto&       info   = ImageDimensionInfoFor(mem.image_dimension);
 	if (HasFlag(mem, Decoder::ImageSampleFlagOffset)) layout.offset = cursor++;
 	if (HasFlag(mem, Decoder::ImageSampleFlagBias)) layout.bias = cursor++;
 	if (HasFlag(mem, Decoder::ImageSampleFlagCompare)) layout.dref = cursor++;
@@ -104,8 +104,7 @@ uint32_t CubeLayer(EmitterState& state, uint32_t value) {
 }

 uint32_t CoordF32(ValueEmitContext& ctx, const IR::MemoryInfo& mem, const IR::Inst& address,
-                  uint32_t first, uint32_t components) {
-	const bool cube = ctx.state.program.info.images.at(mem.resource).cube;
+                  uint32_t first, uint32_t components, bool cube = false) {
 	auto x = AddressF32(ctx, mem, address, first);
 	if (components == 1u) return x;
 	auto y = mem.image_address_components > first + 1u ? AddressF32(ctx, mem, address, first + 1u)
@@ -576,7 +575,8 @@ void EmitImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 		const auto lod       = state.builder.AllocateId();
 		state.builder.AddFunction(
 		    spv::OpImageQueryLod, TypeF32Vector(state, 2), lod, sampled,
-		    CoordF32(ctx, mem, *address, 0, ImageDimensionInfoFor(dimension).spatial_components));
+		    CoordF32(ctx, mem, *address, 0, ImageDimensionInfoFor(dimension).spatial_components,
+		             image.cube));
 		uint32_t values[4] = {ConstantU32(state, 0), ConstantU32(state, 0), ConstantU32(state, 0),
 		                      ConstantU32(state, 0)};
 		for (uint32_t index = 0; index < 2u; index++) {
@@ -637,7 +637,7 @@ void EmitImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 	if (op == IR::ValueOpcode::ImageSampleRaw || op == IR::ValueOpcode::ImageGatherRaw) {
 		const auto  dimension      = image.dimension;
 		const auto& dimension_info = ImageDimensionInfoFor(dimension);
-		const auto  layout         = Layout(mem, dimension);
+		const auto  layout         = Layout(mem);
 		const auto  numeric_class  = image.numeric_class;
 		const bool  dref           = HasFlag(mem, Decoder::ImageSampleFlagCompare);
 		if (dref && state.program.info.images[mem.resource].conversion_format !=
@@ -645,9 +645,9 @@ void EmitImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 			ctx.Fail(inst, "uses depth comparison with a packed integer image");
 			return;
 		}
-		const auto coord =
-		    CoordF32(ctx, mem, *address, layout.coord, dimension_info.coordinate_components);
 		if (op == IR::ValueOpcode::ImageGatherRaw) {
+			const auto coord = CoordF32(ctx, mem, *address, layout.coord,
+			                            dimension_info.coordinate_components, image.cube);
 			if (HasFlag(mem, Decoder::ImageSampleFlagLod)) {
 				static std::atomic_flag warned = ATOMIC_FLAG_INIT;
 				if (!warned.test_and_set(std::memory_order_relaxed)) {
@@ -752,6 +752,11 @@ void EmitImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 			operands.push_back(AddressF32(ctx, mem, *address, layout.bias));
 		}
 		const auto EmitSample = [&](uint32_t resource) {
+			const auto& candidate = state.program.info.images[resource];
+			const auto coord =
+			    CoordF32(ctx, mem, *address, layout.coord,
+			             ImageDimensionInfoFor(candidate.dimension).coordinate_components,
+			             candidate.cube);
 			const auto            sampled = MakeSampledImage(state, resource, mem.sampler);
 			const auto            sample  = state.builder.AllocateId();
 			std::vector<uint32_t> sample_operands {result_type, sample, sampled, coord};
@@ -779,11 +784,11 @@ void EmitImage(ValueEmitContext& ctx, const IR::Inst& inst) {
 		                         ? &state.program.descriptor_sources[image.source]
 		                         : nullptr;
 		if (handle == nullptr || source == nullptr || !source->indirect_image.has_value() ||
-		    source->indirect_image->key_arg >= handle->NumArgs()) {
+		    handle->NumArgs() == 0u) {
 			ctx.Fail(inst, "has invalid indirect image key provenance");
 			return;
 		}
-		const auto key = ctx.Def(handle->Arg(source->indirect_image->key_arg));
+		const auto key = ctx.Def(handle->Arg(0));
 		if (state.flattened_srt_variable == 0 || image.indirect_search_iterations == 0u ||
 		    image.indirect_resources.size() < 2u) {
 			ctx.Fail(inst, "has no indirect image runtime mapping");
@@ -67,6 +67,7 @@ struct SpirvRequirements {
 	bool function_scratch             = false;
 	bool pixel_valid_mask             = false;
 	bool buffer_int64_atomics         = false;
+	bool coherent_buffers             = false;
 };

 SpirvRequirements AnalyzeProgramRequirements(const IR::Program& program);
@@ -308,8 +309,6 @@ uint32_t ConstantI32(EmitterState& state, int32_t value);

 uint32_t ConstantF32(EmitterState& state, uint32_t bits);

-uint32_t FloatBits(float value);
-
 uint32_t ConstantF32Value(EmitterState& state, float value);

 uint32_t ConstantBool(EmitterState& state, bool value);
@@ -329,7 +328,8 @@ void     EmitMeshAllocate(ValueEmitContext& ctx, const IR::Inst& inst);
 uint32_t MeshOutputPointer(EmitterState& state, IR::StageOutputKind kind, uint32_t index = 0);
 uint32_t MeshPrimitivePointer(EmitterState& state);

-DppTargetLane EmitDppQuadPermTargetLane(EmitterState& state, uint32_t subid, uint32_t control);
+DppTargetLane EmitDppPermTargetLane(EmitterState& state, uint32_t subid, uint32_t control,
+                                    uint32_t lane_bits);

 DppTargetLane EmitDppRowShiftTargetLane(EmitterState& state, uint32_t subid, uint32_t amount,
                                        bool left);
@@ -338,7 +338,7 @@ DppTargetLane EmitDppRowRotateRightTargetLane(EmitterState& state, uint32_t subi

 DppTargetLane EmitDppMirrorTargetLane(EmitterState& state, uint32_t subid, bool half_row);

-DppTargetLane EmitDppTargetLane(EmitterState& state, uint32_t control);
+DppTargetLane EmitDppTargetLane(EmitterState& state, const IR::DppMoveFlags& flags);

 uint32_t EmitSubgroupLocalInvocationId(EmitterState& state);

@@ -372,12 +372,13 @@ void EmitMemoryOffsets(EmitterState& state);
 uint32_t LdsDwordCount(const EmitterState& state);

 struct MemoryResourceAccess {
-	IR::ResourceKind kind             = IR::ResourceKind::None;
-	uint32_t         object_pointer   = 0;
-	uint32_t         length           = 0;
-	uint32_t         index_offset     = 0;
-	uint32_t         byte_offset      = 0;
-	bool             add_index_offset = false;
+	IR::ResourceKind      kind             = IR::ResourceKind::None;
+	uint32_t              object_pointer   = 0;
+	uint32_t              length           = 0;
+	uint32_t              index_offset     = 0;
+	uint32_t              byte_offset      = 0;
+	bool                  add_index_offset = false;
+	spv::MemoryAccessMask memory_access    = spv::MemoryAccessMaskNone;
 };

 MemoryResourceAccess PrepareMemoryResourceAccess(EmitterState& state, const IR::MemoryInfo& mem);
@@ -495,22 +496,18 @@ template <typename Fn>
 uint32_t EmitValueOrDefaultIfCondition(EmitterState& state, uint32_t condition, uint32_t type,
                                       uint32_t default_value, Fn&& fn) {
 	const auto then_label  = state.builder.AllocateId();
-	const auto then_exit   = state.builder.AllocateId();
-	const auto else_label  = state.builder.AllocateId();
+	const auto header     = state.current_label;
 	const auto merge_label = state.builder.AllocateId();
 	state.builder.AddFunction(spv::OpSelectionMerge, merge_label, spv::SelectionControlMaskNone);
-	state.builder.AddFunction(spv::OpBranchConditional, condition, then_label, else_label);
+	state.builder.AddFunction(spv::OpBranchConditional, condition, then_label, merge_label);
 	EmitLabel(state, then_label);
 	const auto then_value = fn();
-	state.builder.AddFunction(spv::OpBranch, then_exit);
-	EmitLabel(state, then_exit);
-	state.builder.AddFunction(spv::OpBranch, merge_label);
-	EmitLabel(state, else_label);
+	const auto then_exit  = state.current_label;
 	state.builder.AddFunction(spv::OpBranch, merge_label);
 	EmitLabel(state, merge_label);
 	const auto value = state.builder.AllocateId();
 	state.builder.AddFunction(spv::OpPhi, type, value, then_value, then_exit, default_value,
-	                          else_label);
+	                          header);
 	return value;
 }

@@ -27,61 +27,71 @@ uint32_t EmitDsMaskedLaneRead(EmitterState& state, uint32_t source, uint32_t tar
 	return Select(state, TypeU32(state), source_active, shuffled, ConstantU32(state, 0));
 }

-uint32_t BufferByteAddress(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem,
-                           uint32_t index, uint32_t offset, uint32_t soffset) {
-	auto&          state   = ctx.state;
-	const uint32_t packed  = StorageBufferPackedStride(state, mem);
-	const uint32_t stride  = packed & 0x3fffu;
-	const bool     swizzle = stride != 0u && ((packed >> 14u) & 1u) != 0u;
-	if (((packed >> 20u) & 1u) != 0u) {
-		const auto lane = Binary(state, spv::OpBitwiseAnd, TypeU32(state),
-		                         EmitSubgroupLocalInvocationId(state), ConstantU32(state, 63));
-		index           = Binary(state, spv::OpIAdd, TypeU32(state), index, lane);
-	}
-	if (mem.offset != 0u) {
-		offset = Binary(state, spv::OpIAdd, TypeU32(state), offset, ConstantU32(state, mem.offset));
-	}
+struct BufferAddress {
+	uint32_t offset;
+	uint32_t byte;
+};

-	uint32_t address = 0;
-	if (!swizzle) {
-		if (stride == 0u) {
-			address = offset;
-		} else {
-			const auto indexed = stride == 1u ? index
-			                                  : Binary(state, spv::OpIMul, TypeU32(state), index,
-			                                           ConstantU32(state, stride));
-			address            = Binary(state, spv::OpIAdd, TypeU32(state), indexed, offset);
-		}
-	} else {
-		const uint32_t stride_enum  = (packed >> 16u) & 3u;
-		const uint32_t index_stride = 8u << stride_enum;
-		const auto     index_msb    = Binary(state, spv::OpShiftRightLogical, TypeU32(state), index,
-		                                     ConstantU32(state, stride_enum + 3u));
-		const auto     index_lsb    = Binary(state, spv::OpBitwiseAnd, TypeU32(state), index,
-		                                     ConstantU32(state, index_stride - 1u));
-		const auto     offset_msb =
+BufferAddress CalculateBufferAddress(EmitterState& state, uint32_t index, uint32_t offset,
+                                     uint32_t soffset, uint32_t immediate, uint32_t stride,
+                                     uint32_t swizzle, uint32_t index_stride) {
+	const auto zero = ConstantU32(state, 0);
+	const auto one  = ConstantU32(state, 1);
+	const auto add = [&](uint32_t lhs, uint32_t rhs) {
+		return lhs == zero ? rhs : rhs == zero ? lhs
+		                                      : Binary(state, spv::OpIAdd, TypeU32(state), lhs, rhs);
+	};
+	const auto mul = [&](uint32_t lhs, uint32_t rhs) {
+		return lhs == zero || rhs == zero ? zero
+		       : lhs == one              ? rhs
+		       : rhs == one              ? lhs
+		                                 : Binary(state, spv::OpIMul, TypeU32(state), lhs, rhs);
+	};
+	if (immediate != 0u) {
+		offset = add(offset, ConstantU32(state, immediate));
+	}
+	auto address = add(mul(index, stride), offset);
+	if (swizzle != 0u) {
+		const auto index_shift =
+		    Binary(state, spv::OpIAdd, TypeU32(state), index_stride, ConstantU32(state, 3));
+		const auto indices = Binary(state, spv::OpShiftLeftLogical, TypeU32(state),
+		                            ConstantU32(state, 1), index_shift);
+		const auto index_msb =
+		    Binary(state, spv::OpShiftRightLogical, TypeU32(state), index, index_shift);
+		const auto index_lsb =
+		    Binary(state, spv::OpBitwiseAnd, TypeU32(state), index,
+		           Binary(state, spv::OpISub, TypeU32(state), indices, ConstantU32(state, 1)));
+		const auto offset_msb =
 		    Binary(state, spv::OpBitwiseAnd, TypeU32(state), offset, ConstantU32(state, ~3u));
 		const auto offset_lsb =
 		    Binary(state, spv::OpBitwiseAnd, TypeU32(state), offset, ConstantU32(state, 3u));
-		const auto indexed_msb = stride == 1u ? index_msb
-		                                      : Binary(state, spv::OpIMul, TypeU32(state),
-		                                               index_msb, ConstantU32(state, stride));
-		const auto msb = Binary(state, spv::OpIMul, TypeU32(state),
-		                        Binary(state, spv::OpIAdd, TypeU32(state), indexed_msb, offset_msb),
-		                        ConstantU32(state, index_stride));
-		const auto lsb = Binary(state, spv::OpIAdd, TypeU32(state),
-		                        Binary(state, spv::OpShiftLeftLogical, TypeU32(state), index_lsb,
-		                               ConstantU32(state, 2u)),
-		                        offset_lsb);
-		address        = Binary(state, spv::OpIAdd, TypeU32(state), msb, lsb);
+		const auto msb = mul(add(mul(index_msb, stride), offset_msb), indices);
+		const auto lsb = add(Binary(state, spv::OpShiftLeftLogical, TypeU32(state), index_lsb,
+		                            ConstantU32(state, 2u)), offset_lsb);
+		address = Select(state, TypeU32(state), swizzle, add(msb, lsb), address);
 	}
+	return {offset, add(address, soffset)};
+}

-	const auto soffset_value = inst.Arg(3).Resolve();
-	if (soffset_value.IsImmediate() && soffset_value.GetType() == IR::Type::U32 &&
-	    soffset_value.U32() == 0u) {
-		return address;
+uint32_t BufferLane(EmitterState& state) {
+	return Binary(state, spv::OpBitwiseAnd, TypeU32(state), EmitSubgroupLocalInvocationId(state),
+	              ConstantU32(state, 63));
+}
+
+uint32_t BufferByteAddress(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem) {
+	auto&      state  = ctx.state;
+	const auto packed = StorageBufferPackedStride(state, mem);
+	const auto stride = packed & 0x3fffu;
+	auto       index  = ctx.Arg(inst, 1);
+	if ((packed & (1u << 20u)) != 0u) {
+		index = Binary(state, spv::OpIAdd, TypeU32(state), index, BufferLane(state));
 	}
-	return Binary(state, spv::OpIAdd, TypeU32(state), address, soffset);
+	const bool swizzle = stride != 0u && (packed & (1u << 14u)) != 0u;
+	return CalculateBufferAddress(state, index, ctx.Arg(inst, 2), ctx.Arg(inst, 3), mem.offset,
+	                              ConstantU32(state, stride),
+	                              swizzle ? ConstantBool(state, true) : 0u,
+	                              ConstantU32(state, (packed >> 16u) & 3u))
+	    .byte;
 }

 uint32_t AddU64Low(EmitterState& state, uint32_t low, uint32_t high, uint32_t add_low,
@@ -196,11 +206,8 @@ uint32_t LoadBdaDword(ValueEmitContext& ctx, uint32_t address) {
 	});
 }

-uint32_t LoadBda(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem,
-	             uint32_t bits) {
-	auto&      state   = ctx.state;
-	const auto address = GuestAddress(ctx, inst, mem);
-	const auto active  = ctx.Arg(inst, inst.NumArgs() - 1);
+uint32_t LoadBda(ValueEmitContext& ctx, uint32_t address, uint32_t active, uint32_t bits) {
+	auto& state = ctx.state;
 	return EmitValueOrZeroIfCondition(state, active, [&]() {
 		const auto aligned = Binary(state, spv::OpBitwiseAnd, TypeScalarU64(state), address,
 		                            ConstantDeviceAddress(state, ~uint64_t {3}));
@@ -236,8 +243,7 @@ uint32_t LoadBda(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryIn

 uint32_t ByteAddress(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem) {
 	if (mem.kind == IR::ResourceKind::Buffer) {
-		return BufferByteAddress(ctx, inst, mem, ctx.Arg(inst, 1), ctx.Arg(inst, 2),
-		                         ctx.Arg(inst, 3));
+		return BufferByteAddress(ctx, inst, mem);
 	}
 	if (mem.kind == IR::ResourceKind::Lds || mem.kind == IR::ResourceKind::Gds) {
 		if (mem.offset == 0u) {
@@ -307,7 +313,8 @@ uint32_t LoadWordInBounds(ValueEmitContext& ctx, const MemoryResourceAccess& res
                          uint32_t index) {
 	const auto value   = ctx.state.builder.AllocateId();
 	const auto pointer = EmitMemoryElementPointer(ctx.state, resource, index);
-	ctx.state.builder.AddFunction(spv::OpLoad, TypeU32(ctx.state), value, pointer);
+	ctx.state.builder.AddFunction(spv::OpLoad, TypeU32(ctx.state), value, pointer,
+	                              resource.memory_access);
 	return value;
 }

@@ -492,21 +499,21 @@ void StoreSubword(ValueEmitContext& ctx, const IR::Inst& inst, IR::MemoryInfo me
 	});
 }

+void StoreWordInBounds(ValueEmitContext& ctx, const MemoryResourceAccess& resource, uint32_t index,
+                       uint32_t data) {
+	ctx.state.builder.AddFunction(spv::OpStore,
+	                              EmitMemoryElementPointer(ctx.state, resource, index), data,
+	                              resource.memory_access);
+}
+
 void StoreWordPrepared(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem,
                       const MemoryResourceAccess& resource, uint32_t data) {
 	const auto index = EmitMemoryElementIndex(ctx.state, resource, DwordIndex(ctx, inst, mem));
 	EmitIfCondition(ctx.state, EmitMemoryElementInBounds(ctx.state, resource, index), [&]() {
-		ctx.state.builder.AddFunction(spv::OpStore,
-		                              EmitMemoryElementPointer(ctx.state, resource, index), data);
+		StoreWordInBounds(ctx, resource, index, data);
 	});
 }

-void StoreWordInBounds(ValueEmitContext& ctx, const MemoryResourceAccess& resource, uint32_t index,
-                       uint32_t data) {
-	ctx.state.builder.AddFunction(spv::OpStore,
-	                              EmitMemoryElementPointer(ctx.state, resource, index), data);
-}
-
 void StoreWord(ValueEmitContext& ctx, const IR::Inst& inst, IR::MemoryInfo mem) {
 	EmitIfCondition(ctx.state, ctx.Arg(inst, inst.NumArgs() - 1), [&]() {
 		const auto resource = PrepareMemoryResourceAccess(ctx.state, mem);
@@ -514,31 +521,6 @@ void StoreWord(ValueEmitContext& ctx, const IR::Inst& inst, IR::MemoryInfo mem)
 	});
 }

-void FormattedStorePrepared(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem,
-                            uint32_t component, const MemoryResourceAccess& resource,
-                            uint32_t data) {
-	const auto info = Format::GetFormatInfo(BufferFormat(ctx, mem));
-	if (info.type == Format::ComponentType::Unknown) {
-		StoreWordPrepared(ctx, inst, RebaseRawComponent(mem, component), resource, data);
-		return;
-	}
-	if (component >= info.component_count) return;
-	const auto bits          = info.component_bits[component];
-	const auto component_mem = RebaseFormattedComponent(mem, info, component);
-	if (bits == 8u || bits == 16u) {
-		StoreSubwordPrepared(ctx, inst, component_mem, resource, bits, data);
-	} else {
-		StoreWordPrepared(ctx, inst, component_mem, resource, data);
-	}
-}
-
-void FormattedStore(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem) {
-	EmitIfCondition(ctx.state, ctx.Arg(inst, inst.NumArgs() - 1), [&]() {
-		const auto resource = PrepareMemoryResourceAccess(ctx.state, mem);
-		FormattedStorePrepared(ctx, inst, mem, 0u, resource, ctx.Arg(inst, inst.NumArgs() - 2));
-	});
-}
-
 spv::Op SpirvAtomicOpcode(IR::ValueOpcode opcode) {
 	switch (opcode) {
 		case IR::ValueOpcode::BufferAtomicCmpSwap32: return spv::OpAtomicCompareExchange;
@@ -729,6 +711,15 @@ void StoreFormattedInBounds(ValueEmitContext& ctx, const IR::MemoryInfo& mem,
                            uint32_t data) {
 	if (component >= plan.info.component_count) return;
 	const auto bits = plan.info.component_bits[component];
+	if (bits == 16u && (plan.info.type == Format::ComponentType::Snorm ||
+	                    plan.info.type == Format::ComponentType::Float)) {
+		const auto value = EmitBitCastF32U32(ctx.state, data);
+		const auto pair = EmitCompositeConstructF32x2(ctx.state, value,
+		                                               ConstantF32Value(ctx.state, 0.0f));
+		data = plan.info.type == Format::ComponentType::Float
+		           ? EmitPackHalf2x16(ctx.state, pair)
+		           : EmitPackSnorm2x16(ctx.state, pair);
+	}
 	if (bits == 8u || bits == 16u) {
 		StoreSubwordInBounds(ctx, mem, plan.resource, plan.addresses[component],
 		                     plan.indices[component], bits, data);
@@ -737,12 +728,103 @@ void StoreFormattedInBounds(ValueEmitContext& ctx, const IR::MemoryInfo& mem,
 	}
 }

+void FormattedStore(ValueEmitContext& ctx, const IR::Inst& inst, const IR::MemoryInfo& mem) {
+	EmitIfCondition(ctx.state, ctx.Arg(inst, inst.NumArgs() - 1), [&]() {
+		const auto resource = PrepareMemoryResourceAccess(ctx.state, mem);
+		const auto data     = ctx.Arg(inst, inst.NumArgs() - 2);
+		const auto info     = Format::GetFormatInfo(BufferFormat(ctx, mem));
+		if (info.type == Format::ComponentType::Unknown) {
+			StoreWordPrepared(ctx, inst, RebaseRawComponent(mem, 0u), resource, data);
+			return;
+		}
+		const auto plan = PrepareFormattedMemory(ctx, inst, mem, resource, info, 1u,
+		                                         FormattedAccess::Store);
+		EmitIfCondition(ctx.state, plan.in_bounds, [&]() {
+			StoreFormattedInBounds(ctx, mem, plan, 0u, data);
+		});
+	});
+}
+
+uint32_t LoadIndirectBuffer(ValueEmitContext& ctx, const IR::Inst& inst, uint32_t components) {
+	auto&       state   = ctx.state;
+	const auto& handle  = *inst.Arg(0).ResolveInstruction();
+	const auto  word1   = ctx.Arg(handle, 1);
+	const auto  records = ctx.Arg(handle, 2);
+	const auto  word3   = ctx.Arg(handle, 3);
+	const auto  field   = [&](uint32_t word, uint32_t first, uint32_t count) {
+		return EmitBitFieldUExtract(state, word, ConstantU32(state, first),
+		                            ConstantU32(state, count));
+	};
+	const auto nonzero = [&](uint32_t value) {
+		return Binary(state, spv::OpINotEqual, TypeBool(state), value, ConstantU32(state, 0));
+	};
+	const auto has_dword = [&](uint32_t offset, uint32_t size) {
+		return AndCondition(
+		    state,
+		    Binary(state, spv::OpUGreaterThanEqual, TypeBool(state), size, ConstantU32(state, 4)),
+		    Binary(state, spv::OpULessThanEqual, TypeBool(state), offset,
+		           Binary(state, spv::OpISub, TypeU32(state), size, ConstantU32(state, 4))));
+	};
+	const auto stride       = field(word1, 16, 14);
+	const auto swizzle      = AndCondition(state, nonzero(stride), nonzero(field(word1, 31, 1)));
+	const auto index_stride = field(word3, 21, 2);
+	const auto add_tid      = nonzero(field(word3, 23, 1));
+	const auto index =
+	    Binary(state, spv::OpIAdd, TypeU32(state), ctx.Arg(inst, 1),
+	           Select(state, TypeU32(state), add_tid, BufferLane(state), ConstantU32(state, 0)));
+	const auto soffset = ctx.Arg(inst, 3);
+	const auto base    = DeviceAddressFromWords(state, ctx.Arg(handle, 0), field(word1, 0, 16));
+	const auto valid_format    = nonzero(field(word3, 12, 7));
+	const auto mode            = field(word3, 28, 2);
+	const auto index_in_bounds = Binary(state, spv::OpULessThan, TypeBool(state), index, records);
+	const auto scalar_in_bounds =
+	    Binary(state, spv::OpULessThanEqual, TypeBool(state), soffset, records);
+	const auto raw_records = Binary(state, spv::OpISub, TypeU32(state), records, soffset);
+	const auto raw_index_in_bounds =
+	    Binary(state, spv::OpULessThan, TypeBool(state), index, raw_records);
+	std::array<uint32_t, 4> values {};
+	for (uint32_t component = 0; component < components; component++) {
+		const auto address = CalculateBufferAddress(state, index, ctx.Arg(inst, 2), soffset,
+		                                            ctx.Memory(inst).offset + component * 4u,
+		                                            stride, swizzle, index_stride);
+		// RDNA2 raw DWORD vectors check each component, unlike formatted accesses.
+		const auto structured_bounds =
+		    AndCondition(state, index_in_bounds,
+		                 Binary(state, spv::OpULessThan, TypeBool(state), address.offset, stride));
+		// OOB_SELECT=3 reduces NUM_RECORDS by SOFFSET before the offset/index check.
+		const auto raw_bounds = AndCondition(
+		    state, scalar_in_bounds,
+		    Select(state, TypeBool(state), swizzle,
+		           AndCondition(state, raw_index_in_bounds, has_dword(address.offset, stride)),
+		           has_dword(address.offset, raw_records)));
+		auto in_bounds =
+		    Select(state, TypeBool(state),
+		           Binary(state, spv::OpIEqual, TypeBool(state), mode, ConstantU32(state, 0)),
+		           structured_bounds, index_in_bounds);
+		in_bounds = Select(
+		    state, TypeBool(state),
+		    Binary(state, spv::OpULessThan, TypeBool(state), mode, ConstantU32(state, 2)),
+		    in_bounds,
+		    Select(state, TypeBool(state),
+		           Binary(state, spv::OpIEqual, TypeBool(state), mode, ConstantU32(state, 2)),
+		           nonzero(records), raw_bounds));
+		const auto guest =
+		    Binary(state, spv::OpIAdd, TypeScalarU64(state), base,
+		           Unary(state, spv::OpUConvert, TypeScalarU64(state), address.byte));
+		values[component] = LoadBda(ctx, guest, AndCondition(state, valid_format, in_bounds), 32u);
+	}
+	return ConstructU32Composite(state, components, values);
+}
+
 uint32_t LoadWideBuffer(ValueEmitContext& ctx, const IR::Inst& inst, uint32_t components) {
 	auto& state = ctx.state;
 	return EmitValueOrDefaultIfCondition(
 	    state, ctx.Arg(inst, inst.NumArgs() - 1), TypeU32Composite(state, components),
 	    ConstantU32CompositeZero(state, components), [&]() {
 		    const auto mem      = ctx.Memory(inst);
+		    if (mem.kind == IR::ResourceKind::IndirectBuffer) {
+			    return LoadIndirectBuffer(ctx, inst, components);
+		    }
 		    const auto resource = PrepareMemoryResourceAccess(state, mem);
 		    const auto info = Format::GetFormatInfo(
 		        mem.formatted ? BufferFormat(ctx, mem) : Prospero::BufferFormat::kInvalid);
@@ -1091,7 +1173,8 @@ void EmitLoadMemory(ValueEmitContext& ctx, const IR::Inst& inst) {
 		value = LoadWideShared(ctx, inst, shared_components);
 	else if (address_info.access == IR::AddressAccess::Read &&
 	         mem.kind != IR::ResourceKind::Scratch)
-		value = LoadBda(ctx, inst, mem, address_info.data_bits);
+		value = LoadBda(ctx, GuestAddress(ctx, inst, mem), ctx.Arg(inst, inst.NumArgs() - 1),
+		                address_info.data_bits);
 	else if (op == IR::ValueOpcode::LoadBufferU32 && mem.formatted)
 		value = FormattedLoad(ctx, inst, mem);
 	else if (inst.GetType() == IR::Type::U8)
@@ -86,7 +86,9 @@ MemoryResourceAccess PrepareStorageBufferResourceAccess(EmitterState& state,
 	}
 	const auto array_index =
 	    ResourceForDescriptor(state, IR::DescriptorBindingKind::Buffers, mem.resource);
-	MemoryResourceAccess access {.kind = mem.kind};
+	MemoryResourceAccess access {
+	    .kind = mem.kind,
+	    .memory_access = mem.coherent ? spv::MemoryAccessVolatileMask : spv::MemoryAccessMaskNone};
 	access.object_pointer = state.builder.AllocateId();
 	state.builder.AddFunction(spv::OpAccessChain, pointer_type, access.object_pointer, variable,
 	                          ConstantU32(state, array_index));
@@ -1,4 +1,4 @@
-#include "graphics/shader/recompiler/backend/spirv/spirvEmitterInternal.h"
+#include "graphics/shader/recompiler/backend/spirv/spirvEmitterInstructions.h"

 namespace Libs::Graphics::ShaderRecompiler::Spirv::Emitter {
 namespace {
@@ -127,10 +127,7 @@ void EmitMeshEntryPoint(EmitterState& state) {
 	state.builder.AddFunction(spv::OpFunctionCall, TypeVoid(state), state.builder.AllocateId(),
 	                          state.mesh_guest_func);
 	// All guest waves finish before the uniform Vulkan allocation and output stores.
-	state.builder.AddFunction(spv::OpControlBarrier, ConstantU32(state, spv::ScopeWorkgroup),
-	                          ConstantU32(state, spv::ScopeWorkgroup),
-	                          ConstantU32(state, spv::MemorySemanticsAcquireReleaseMask |
-	                                                 spv::MemorySemanticsWorkgroupMemoryMask));
+	EmitBarrier(state);
 	const auto vertices   = MeshLoad(state, state.mesh_allocation, spv::StorageClassWorkgroup,
 	                                 TypeU32(state), ConstantU32(state, 0));
 	const auto primitives = MeshLoad(state, state.mesh_allocation, spv::StorageClassWorkgroup,
@@ -1,6 +1,7 @@
 #include "graphics/shader/recompiler/backend/spirv/spirvEmitterInternal.h"

 #include <algorithm>
+#include <bit>

 namespace Libs::Graphics::ShaderRecompiler::Spirv::Emitter {

@@ -200,6 +201,16 @@ void DefineDescriptors(EmitterState& state) {
 					state.builder.AddAnnotation(spv::OpDecorate, state.storage_buffer_u64_variable,
 					                            spv::DecorationAliased);
 				}
+				if (state.requirements.coherent_buffers) {
+					// RDNA2 stores publish to L2 even without GLC; every alias of the buffer
+					// must participate in visibility for cache-bypassing polling loads.
+					state.builder.AddAnnotation(spv::OpDecorate, state.storage_buffer_variable,
+					                            spv::DecorationCoherent);
+					if (state.storage_buffer_u64_variable != 0) {
+						state.builder.AddAnnotation(spv::OpDecorate, state.storage_buffer_u64_variable,
+						                            spv::DecorationCoherent);
+					}
+				}
 				break;
 			case IR::DescriptorBindingKind::BdaPagetable:
 				state.bda_pagetable_variable = Define(StorageBufferU64Type(state), "bda_pagetable");
@@ -256,14 +267,8 @@ uint32_t ConstantF32(EmitterState& state, uint32_t bits) {
 	return state.builder.Constant(spv::OpConstant, TypeF32(state), bits);
 }

-uint32_t FloatBits(float value) {
-	uint32_t bits = 0;
-	std::memcpy(&bits, &value, sizeof(bits));
-	return bits;
-}
-
 uint32_t ConstantF32Value(EmitterState& state, float value) {
-	return ConstantF32(state, FloatBits(value));
+	return ConstantF32(state, std::bit_cast<uint32_t>(value));
 }

 uint32_t ConstantBool(EmitterState& state, bool value) {
@@ -558,8 +563,17 @@ void DefineOutputs(EmitterState& state) {
 				const auto type = uint_output ? TypeU32Vector(state, 4) : TypeF32Vector(state, 4);
 				binding.variable_id = DefineInterfaceVariable(state, type, spv::StorageClassOutput,
 				                                              binding.debug_name.c_str());
+				const bool dual_source = binding.kind == IR::StageOutputKind::Mrt &&
+				                         state.program.stage == ShaderType::Pixel &&
+				                         state.input_info.pixel->dual_source_blending;
+				EXIT_NOT_IMPLEMENTED(dual_source && binding.index > 1);
 				state.builder.AddAnnotation(spv::OpDecorate, binding.variable_id,
-				                            spv::DecorationLocation, binding.location);
+				                            spv::DecorationLocation,
+				                            dual_source ? 0u : binding.location);
+				if (dual_source) {
+					state.builder.AddAnnotation(spv::OpDecorate, binding.variable_id,
+					                            spv::DecorationIndex, binding.index);
+				}
 				break;
 			}
 		}
@@ -1389,6 +1389,15 @@ bool SplitSharedMergeBlock(Graph& graph, uint32_t merge,
 	if (construct_predecessors.empty() || (!force_split && !has_external_predecessor)) {
 		return false;
 	}
+	if (construct_predecessors.size() == 1u) {
+		const auto* predecessor = graph.FindBlock(construct_predecessors.front());
+		if (predecessor != nullptr && predecessor->inst_begin == predecessor->inst_end &&
+		    predecessor->terminator.kind == TerminatorKind::Branch &&
+		    predecessor->terminator.true_block == merge &&
+		    predecessor->start_pc == merge_block->start_pc) {
+			return false;
+		}
+	}
 	const auto synthetic_merge = AppendSyntheticBranchBlock(graph, merge);
 	auto*      synthetic_block = graph.FindBlock(synthetic_merge);
 	if (synthetic_block != nullptr) {
@@ -3,6 +3,7 @@
 #include "graphics/shader/recompiler/frontend/decode/OpcodeTable.h"

 #include <algorithm>
+#include <bit>

 namespace Libs::Graphics::ShaderRecompiler::Decoder {
 namespace {
@@ -203,18 +204,6 @@ constexpr auto MIMG_SAMPLE_OPS = Detail::MakeOpcodeTable<0x100>(MIMG_SAMPLE_OPCO
 constexpr auto MIMG_GATHER_OPS = Detail::MakeOpcodeTable<0x100>(MIMG_GATHER_OPCODE_LIST);
 constexpr auto MIMG_ATOMIC_OPS = Detail::MakeOpcodeTable<0x100>(MIMG_ATOMIC_OPCODE_LIST);

-const MimgSampleInfo* LookupSample(uint32_t opcode) {
-	return Detail::FindOpcode(MIMG_SAMPLE_OPS, opcode);
-}
-
-const MimgGatherInfo* LookupGather(uint32_t opcode) {
-	return Detail::FindOpcode(MIMG_GATHER_OPS, opcode);
-}
-
-const Detail::OpcodeMap* LookupAtomic(uint32_t opcode) {
-	return Detail::FindOpcode(MIMG_ATOMIC_OPS, opcode);
-}
-
 Opcode DecodeMimgOpcode(uint32_t opcode, const MimgSampleInfo* sample, const MimgGatherInfo* gather,
                        const Detail::OpcodeMap* atomic) {
 	if (sample != nullptr) {
@@ -280,10 +269,6 @@ uint32_t CountDmaskComponents(uint32_t dmask) {
 	return count != 0 ? count : 1u;
 }

-bool IsSingleDmaskBit(uint32_t dmask) {
-	return dmask != 0u && (dmask & (dmask - 1u)) == 0u;
-}
-
 } // namespace

 ImageAddressComponent ImageAddressComponentLayout(uint32_t flags, uint32_t component) {
@@ -332,9 +317,9 @@ void DecodeMimg(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 	const bool     r128   = ((word0 >> 15u) & 0x1u) != 0u;
 	const bool     a16    = ((word1 >> 30u) & 0x1u) != 0u;
 	const bool     d16    = ((word1 >> 31u) & 0x1u) != 0u;
-	const auto*    sample = LookupSample(opcode);
-	const auto*    gather = LookupGather(opcode);
-	const auto*    atomic = LookupAtomic(opcode);
+	const auto*    sample = Detail::FindOpcode(MIMG_SAMPLE_OPS, opcode);
+	const auto*    gather = Detail::FindOpcode(MIMG_GATHER_OPS, opcode);
+	const auto*    atomic = Detail::FindOpcode(MIMG_ATOMIC_OPS, opcode);

 	inst.pc                 = pc;
 	inst.word_count         = word_count;
@@ -364,7 +349,7 @@ void DecodeMimg(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 	if (inst.opcode == Opcode::UNSUPPORTED) {
 		SetUnsupported(inst, Family::MIMG, opcode, "MIMG opcode is not implemented");
 	}
-	if (gather != nullptr && !IsSingleDmaskBit(inst.dmask)) {
+	if (gather != nullptr && !std::has_single_bit(inst.dmask)) {
 		SetUnsupported(inst, Family::MIMG, opcode,
 		               "MIMG image gather requires exactly one dmask bit");
 	}
@@ -382,7 +367,7 @@ void DecodeMimg(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 }

 const char* MimgSampleOpcodeName(uint32_t opcode) {
-	const auto* sample = LookupSample(opcode);
+	const auto* sample = Detail::FindOpcode(MIMG_SAMPLE_OPS, opcode);
 	return sample != nullptr ? sample->name : nullptr;
 }

@@ -21,6 +21,7 @@ constexpr MemoryOpcodeInfo SMEM_OPCODE_LIST[] = {
    {0x04u, Opcode::S_LOAD_DWORDX16, 16, 32},      {0x08u, Opcode::S_BUFFER_LOAD_DWORD, 1, 32},
    {0x09u, Opcode::S_BUFFER_LOAD_DWORDX2, 2, 32}, {0x0au, Opcode::S_BUFFER_LOAD_DWORDX4, 4, 32},
    {0x0bu, Opcode::S_BUFFER_LOAD_DWORDX8, 8, 32}, {0x0cu, Opcode::S_BUFFER_LOAD_DWORDX16, 16, 32},
+    {0x25u, Opcode::S_MEMREALTIME, 2, 32},
 };

 constexpr MemoryOpcodeInfo MUBUF_OPCODE_LIST[] = {
@@ -86,6 +87,7 @@ constexpr MemoryOpcodeInfo FLAT_OPCODE_LIST[] = {

 constexpr MemoryOpcodeInfo DS_OPCODE_LIST[] = {
    {0x00u, Opcode::DS_ADD_U32, 1, 32},          {0x01u, Opcode::DS_SUB_U32, 1, 32},
+    {0x03u, Opcode::DS_INC_U32, 1, 32},          {0x04u, Opcode::DS_DEC_U32, 1, 32},
    {0x05u, Opcode::DS_MIN_I32, 1, 32},          {0x06u, Opcode::DS_MAX_I32, 1, 32},
    {0x07u, Opcode::DS_MIN_U32, 1, 32},          {0x08u, Opcode::DS_MAX_U32, 1, 32},
    {0x09u, Opcode::DS_AND_B32, 1, 32},          {0x0au, Opcode::DS_OR_B32, 1, 32},
@@ -107,6 +109,7 @@ constexpr MemoryOpcodeInfo DS_OPCODE_LIST[] = {
    {0x4du, Opcode::DS_WRITE_B64, 2, 32},        {0x4eu, Opcode::DS_WRITE2_B64, 4, 32},
    {0x4fu, Opcode::DS_WRITE2ST64_B64, 4, 32},   {0x76u, Opcode::DS_READ_B64, 2, 32},
    {0x77u, Opcode::DS_READ2_B64, 4, 32},        {0x78u, Opcode::DS_READ2ST64_B64, 4, 32},
+    {0xa0u, Opcode::DS_WRITE_B8_D16_HI, 1, 8},
    {0xa1u, Opcode::DS_WRITE_B16_D16_HI, 1, 16},
    {0xa6u, Opcode::DS_READ_U16_D16, 1, 16},
    {0xa7u, Opcode::DS_READ_U16_D16_HI, 1, 16},
@@ -147,6 +150,7 @@ bool IsDsWriteOpcode(Opcode opcode) {
 	switch (opcode) {
 		case Opcode::DS_WRITE_B8:
 		case Opcode::DS_WRITE_B16:
+		case Opcode::DS_WRITE_B8_D16_HI:
 		case Opcode::DS_WRITE_B16_D16_HI:
 		case Opcode::DS_WRITE2_B32:
 		case Opcode::DS_WRITE2ST64_B32:
@@ -166,7 +170,9 @@ bool IsDsAtomicOpcode(Opcode opcode) {
 		case Opcode::DS_ADD_RTN_U32:
 		case Opcode::DS_SUB_U32:
 		case Opcode::DS_SUB_RTN_U32:
+		case Opcode::DS_INC_U32:
 		case Opcode::DS_INC_RTN_U32:
+		case Opcode::DS_DEC_U32:
 		case Opcode::DS_DEC_RTN_U32:
 		case Opcode::DS_MIN_I32:
 		case Opcode::DS_MIN_RTN_I32:
@@ -240,6 +246,10 @@ void DecodeSmem(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 	}

 	DecodeScalarDestination(sdst, pc, inst.dst);
+	if (inst.opcode == Opcode::S_MEMREALTIME) {
+		inst.src_count = 0;
+		return;
+	}
 	// SMEM encodes SBASE in SGPR pairs. Scalar-buffer loads still use the same
 	// pair index; their descriptor operand consumes four SGPRs from that base.
 	DecodeScalarSource(sbase * 2u, pc, inst.src0);
@@ -263,6 +273,7 @@ void DecodeMubuf(uint32_t pc, std::span<const uint32_t> code, uint32_t word_inde
 	inst.idxen       = ((word0 >> 13u) & 1u) != 0;
 	inst.offen       = ((word0 >> 12u) & 1u) != 0;
 	inst.glc         = ((word0 >> 14u) & 1u) != 0;
+	inst.dlc         = ((word0 >> 15u) & 1u) != 0;
 	inst.slc         = ((word1 >> 22u) & 1u) != 0;
 	inst.family      = Family::MUBUF;
 	inst.opcode_id   = opcode;
@@ -298,6 +309,7 @@ void DecodeMtbuf(uint32_t pc, std::span<const uint32_t> code, uint32_t word_inde
 	inst.idxen         = ((word0 >> 13u) & 1u) != 0;
 	inst.offen         = ((word0 >> 12u) & 1u) != 0;
 	inst.glc           = ((word0 >> 14u) & 1u) != 0;
+	inst.dlc           = ((word0 >> 15u) & 1u) != 0;
 	inst.slc           = ((word1 >> 22u) & 1u) != 0;
 	inst.family        = Family::MTBUF;
 	inst.opcode_id     = opcode;
@@ -429,7 +441,9 @@ void DecodeDs(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index,
 	}
 	DecodeVectorGpr(addr, inst.src0);
 	DecodeVectorGpr(data0, inst.src1);
-	if (inst.opcode == Opcode::DS_WRITE_B16_D16_HI) {
+	if (inst.opcode == Opcode::DS_WRITE_B8_D16_HI) {
+		inst.src1.sdwa_sel = 2u;
+	} else if (inst.opcode == Opcode::DS_WRITE_B16_D16_HI) {
 		inst.src1.sdwa_sel = 5u;
 	}
 	DecodeVectorGpr(data1, inst.src2);
@@ -24,7 +24,8 @@ constexpr OpcodeMap SOP2_OPCODE_LIST[] = {
    {0x1cu, Opcode::S_XNOR_B32},        {0x1du, Opcode::S_XNOR_B64},
    {0x1eu, Opcode::S_LSHL_B32},        {0x1fu, Opcode::S_LSHL_B64},
    {0x20u, Opcode::S_LSHR_B32},        {0x21u, Opcode::S_LSHR_B64},
-    {0x22u, Opcode::S_ASHR_I32},        {0x24u, Opcode::S_BFM_B32},
+    {0x22u, Opcode::S_ASHR_I32},        {0x23u, Opcode::S_ASHR_I64},
+    {0x24u, Opcode::S_BFM_B32},
    {0x25u, Opcode::S_BFM_B64},         {0x26u, Opcode::S_MUL_I32},
    {0x27u, Opcode::S_BFE_U32},         {0x28u, Opcode::S_BFE_I32},
    {0x29u, Opcode::S_BFE_U64},         {0x2cu, Opcode::S_ABSDIFF_I32},
@@ -73,7 +74,8 @@ constexpr OpcodeMap SOPC_OPCODE_LIST[] = {
    {0x03u, Opcode::S_CMP_GE_I32},  {0x04u, Opcode::S_CMP_LT_I32},  {0x05u, Opcode::S_CMP_LE_I32},
    {0x06u, Opcode::S_CMP_EQ_U32},  {0x07u, Opcode::S_CMP_LG_U32},  {0x08u, Opcode::S_CMP_GT_U32},
    {0x09u, Opcode::S_CMP_GE_U32},  {0x0au, Opcode::S_CMP_LT_U32},  {0x0bu, Opcode::S_CMP_LE_U32},
-    {0x0cu, Opcode::S_BITCMP0_B32}, {0x0du, Opcode::S_BITCMP1_B32}, {0x12u, Opcode::S_CMP_EQ_U64},
+    {0x0cu, Opcode::S_BITCMP0_B32}, {0x0du, Opcode::S_BITCMP1_B32},
+    {0x0eu, Opcode::S_BITCMP0_B64}, {0x0fu, Opcode::S_BITCMP1_B64}, {0x12u, Opcode::S_CMP_EQ_U64},
    {0x13u, Opcode::S_CMP_LG_U64},
 };

@@ -14,10 +14,6 @@
 namespace Libs::Graphics::ShaderRecompiler::Decoder {
 namespace {

-uint32_t FloatBits(float value) {
-	return std::bit_cast<uint32_t>(value);
-}
-
 bool HasLiteral(const Instruction& inst) {
 	return inst.src0.kind == OperandKind::LiteralConstant ||
 	       inst.src1.kind == OperandKind::LiteralConstant ||
@@ -77,11 +73,11 @@ std::string FormatMemory(const Instruction& inst) {
 		text += OperandToString(*sources[i]);
 	}
 	text += fmt::format(" ; offset={} offset2={} dwords={} bits={} dfmt={} nfmt={} signed={} "
-	                    "typed={} formatted={} segment={} glc={} slc={} idxen={} offen={}",
+	                    "typed={} formatted={} segment={} glc={} dlc={} slc={} idxen={} offen={}",
 	                    inst.offset, inst.secondary_offset, inst.data_dwords, inst.data_bits,
 	                    inst.data_format, inst.number_format, inst.data_signed ? 1u : 0u,
 	                    inst.typed ? 1u : 0u, inst.formatted ? 1u : 0u, inst.memory_segment,
-	                    inst.glc ? 1u : 0u, inst.slc ? 1u : 0u, inst.idxen ? 1u : 0u,
+	                    inst.glc ? 1u : 0u, inst.dlc ? 1u : 0u, inst.slc ? 1u : 0u, inst.idxen ? 1u : 0u,
 	                    inst.offen ? 1u : 0u);
 	return text;
 }
@@ -242,7 +238,7 @@ void DecodeScalarSource(uint32_t code, uint32_t pc, Operand& operand) {
 	if (code >= 240u && code <= 247u) {
 		constexpr float values[] = {0.5f, -0.5f, 1.0f, -1.0f, 2.0f, -2.0f, 4.0f, -4.0f};
 		operand.kind             = OperandKind::FloatInlineConstant;
-		operand.value            = FloatBits(values[code - 240u]);
+		operand.value            = std::bit_cast<uint32_t>(values[code - 240u]);
 		return;
 	}
 	if (code >= 256u && code <= 511u) {
@@ -260,7 +256,7 @@ void DecodeScalarSource(uint32_t code, uint32_t pc, Operand& operand) {
 		case 239u: operand.kind = OperandKind::PopsExitingWaveId; return;
 		case 248u:
 			operand.kind      = OperandKind::FloatInlineConstant;
-			operand.value = FloatBits(0.15915494309189535f);
+			operand.value = std::bit_cast<uint32_t>(0.15915494309189535f);
 			return;
 		case 251u: operand.kind = OperandKind::VccZ; return;
 		case 252u: operand.kind = OperandKind::ExecZ; return;
@@ -407,6 +403,7 @@ void DecodeProgram(std::span<const uint32_t> code, Program& program) {
 	program.instructions.clear();
 	program.instructions.reserve(code.size());
 	program.code = code;
+	program.has_bvh = false;

 	std::vector<bool> branch_targets;
 	for (uint32_t word_index = 0; word_index < code.size();) {
@@ -415,6 +412,10 @@ void DecodeProgram(std::span<const uint32_t> code, Program& program) {

 		const auto& inst = program.instructions.back();
 		word_index += inst.word_count;
+		if (inst.family == Family::MIMG && (inst.opcode_id == 0xe6u || inst.opcode_id == 0xe7u)) {
+			program.has_bvh = true;
+			return;
+		}

 		if (IsDirectBranch(inst.opcode)) {
 			const auto target_index = inst.branch_target / sizeof(uint32_t);
@@ -476,8 +477,9 @@ std::string OperandToString(const Operand& operand) {
 		text += ".clamp";
 	}
 	if (operand.dpp) {
-		text += fmt::format(".dpp(ctrl=0x{:x},fi={},bc={})", operand.dpp_ctrl,
-		                    operand.dpp_fetch_inactive ? 1u : 0u, operand.dpp_bound_ctrl ? 1u : 0u);
+		text += fmt::format(".{}(ctrl=0x{:x},fi={},bc={})", operand.dpp8 ? "dpp8" : "dpp",
+		                    operand.dpp_ctrl, operand.dpp_fetch_inactive ? 1u : 0u,
+		                    operand.dpp_bound_ctrl ? 1u : 0u);
 	}
 	return text;
 }
@@ -662,7 +664,9 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::DS_ADD_RTN_U32:
 		case Opcode::DS_SUB_U32:
 		case Opcode::DS_SUB_RTN_U32:
+		case Opcode::DS_INC_U32:
 		case Opcode::DS_INC_RTN_U32:
+		case Opcode::DS_DEC_U32:
 		case Opcode::DS_DEC_RTN_U32:
 		case Opcode::DS_MIN_I32:
 		case Opcode::DS_MIN_RTN_I32:
@@ -697,6 +701,7 @@ std::string InstructionToString(const Instruction& inst) {
 		case Opcode::DS_READ_B128:
 		case Opcode::DS_WRITE_B8:
 		case Opcode::DS_WRITE_B16:
+		case Opcode::DS_WRITE_B8_D16_HI:
 		case Opcode::DS_WRITE_B16_D16_HI:
 		case Opcode::DS_WRITE2_B32:
 		case Opcode::DS_WRITE2ST64_B32:
@@ -2,8 +2,8 @@
 #define EMULATOR_INCLUDE_EMULATOR_GRAPHICS_SHADER_RECOMPILER_SHADERDECODER_H_

 #include "common/common.h"
-#include "common/magicEnum.h"

+#include <magic_enum.hpp>
 #include <span>
 #include <string_view>
 #include <vector>
@@ -73,6 +73,8 @@ enum class Opcode {
 	S_SUB_I32,
 	S_BITCMP0_B32,
 	S_BITCMP1_B32,
+	S_BITCMP0_B64,
+	S_BITCMP1_B64,
 	S_BITSET0_B32,
 	S_BITSET1_B32,
 	S_BITSET0_B64,
@@ -106,6 +108,7 @@ enum class Opcode {
 	S_LSHR_B32,
 	S_LSHR_B64,
 	S_ASHR_I32,
+	S_ASHR_I64,
 	S_MUL_I32,
 	S_MUL_HI_U32,
 	S_MUL_HI_I32,
@@ -227,6 +230,7 @@ enum class Opcode {
 	V_CVT_PKRTZ_F16_F32,
 	V_CVT_PK_U8_F32,
 	V_MAD_F32,
+	V_MAD_I16,
 	V_MAD_I32_I24,
 	V_MAD_U32_U24,
 	V_MAD_U64_U32,
@@ -299,6 +303,7 @@ enum class Opcode {
 	V_SUBREV_NC_U32,
 	V_ADD_NC_U16,
 	V_SUB_NC_U16,
+	V_MUL_LO_U16,
 	V_MAX_U16,
 	V_MAX_I16,
 	V_MIN_U16,
@@ -349,6 +354,7 @@ enum class Opcode {
 	V_CMPX_GT_F32,
 	V_CMPX_LG_F32,
 	V_CMPX_GE_F32,
+	V_CMPX_O_F32,
 	V_CMPX_NGE_F32,
 	V_CMPX_NLG_F32,
 	V_CMPX_NGT_F32,
@@ -377,7 +383,9 @@ enum class Opcode {
 	V_CMP_GT_F16,
 	V_CMP_LG_F16,
 	V_CMP_GE_F16,
+	V_CMP_NGT_F16,
 	V_CMP_NEQ_F16,
+	V_CMP_NLT_F16,
 	V_CMPX_LT_F16,
 	V_CMPX_EQ_F16,
 	V_CMPX_LE_F16,
@@ -397,6 +405,7 @@ enum class Opcode {
 	V_CMP_LE_U16,
 	V_CMP_GT_U16,
 	V_CMPX_LT_U16,
+	V_CMPX_EQ_U16,
 	V_CMPX_GT_U16,
 	V_CMP_NE_U16,
 	V_CMP_GE_U16,
@@ -432,6 +441,7 @@ enum class Opcode {
 	S_BUFFER_LOAD_DWORDX4,
 	S_BUFFER_LOAD_DWORDX8,
 	S_BUFFER_LOAD_DWORDX16,
+	S_MEMREALTIME,
 	BUFFER_LOAD_FORMAT_X,
 	BUFFER_LOAD_FORMAT_XY,
 	BUFFER_LOAD_FORMAT_XYZ,
@@ -495,7 +505,9 @@ enum class Opcode {
 	DS_ADD_RTN_U32,
 	DS_SUB_U32,
 	DS_SUB_RTN_U32,
+	DS_INC_U32,
 	DS_INC_RTN_U32,
+	DS_DEC_U32,
 	DS_DEC_RTN_U32,
 	DS_MIN_I32,
 	DS_MIN_RTN_I32,
@@ -534,6 +546,7 @@ enum class Opcode {
 	DS_READ_B128,
 	DS_WRITE_B8,
 	DS_WRITE_B16,
+	DS_WRITE_B8_D16_HI,
 	DS_WRITE_B16_D16_HI,
 	DS_WRITE2_B32,
 	DS_WRITE2ST64_B32,
@@ -665,6 +678,7 @@ struct Operand {
 	bool     absolute           = false;
 	bool     clamp              = false;
 	bool     dpp                = false;
+	bool     dpp8               = false;
 };

 struct Instruction {
@@ -700,6 +714,7 @@ struct Instruction {
 	bool           formatted                                    = false;
 	bool           gds                                          = false;
 	bool           glc                                          = false;
+	bool           dlc                                          = false;
 	bool           slc                                          = false;
 	bool           idxen                                        = false;
 	bool           offen                                        = false;
@@ -718,6 +733,7 @@ struct Instruction {
 struct Program {
 	std::span<const uint32_t> code;
 	std::vector<Instruction>  instructions;
+	bool                     has_bvh = false;
 };

 // Code spans are trusted to contain complete instructions, valid branch targets, and 32-bit PCs.
@@ -40,7 +40,7 @@ constexpr Vop2OpcodeInfo VOP2_OPCODE_LIST[] = {
    {0x0fu, Opcode::V_MIN_F32},
    {0x10u, Opcode::V_MAX_F32},
    {0x11u, Opcode::V_MIN_I32},
-    {0x12u, Opcode::V_MAX_I32},
+    {0x12u, Opcode::V_MAX_I32, Vop2SdwaProfile::IntegerFullDestination},
    {0x13u, Opcode::V_MIN_U32, Vop2SdwaProfile::IntegerPartialDestination},
    {0x14u, Opcode::V_MAX_U32, Vop2SdwaProfile::IntegerFullDestination},
    {0x15u, Opcode::V_LSHR_B32},
@@ -83,10 +83,6 @@ constexpr Vop2OpcodeInfo VOP2_OPCODE_LIST[] = {

 constexpr auto VOP2_OPS = Detail::MakeOpcodeTable<0x40>(VOP2_OPCODE_LIST);

-constexpr Opcode LookupVop2Opcode(uint32_t encoding) {
-	return Detail::LookupOpcode(VOP2_OPS, encoding);
-}
-
 constexpr OpcodeMap VOP1_OPCODE_LIST[] = {
    {0x00u, Opcode::V_NOP},
    {0x01u, Opcode::V_MOV_B32},
@@ -218,6 +214,7 @@ constexpr VopcOpcodeInfo VOPC_OPCODE_LIST[] = {
    {0x11u, Opcode::V_CMPX_LT_F32},        {0x12u, Opcode::V_CMPX_EQ_F32},
    {0x13u, Opcode::V_CMPX_LE_F32},        {0x14u, Opcode::V_CMPX_GT_F32},
    {0x15u, Opcode::V_CMPX_LG_F32},        {0x16u, Opcode::V_CMPX_GE_F32},
+    {0x17u, Opcode::V_CMPX_O_F32},
    {0x19u, Opcode::V_CMPX_NGE_F32},       {0x1au, Opcode::V_CMPX_NLG_F32},
    {0x1bu, Opcode::V_CMPX_NGT_F32},       {0x1cu, Opcode::V_CMPX_NLE_F32},
    {0x1du, Opcode::V_CMPX_NEQ_F32},       {0x1eu, Opcode::V_CMPX_NLT_F32},
@@ -235,7 +232,7 @@ constexpr VopcOpcodeInfo VOPC_OPCODE_LIST[] = {
    {0xa9u, Opcode::V_CMP_LT_U16},         {0xaau, Opcode::V_CMP_EQ_U16},
    {0xabu, Opcode::V_CMP_LE_U16},         {0xacu, Opcode::V_CMP_GT_U16},
    {0xadu, Opcode::V_CMP_NE_U16},         {0xaeu, Opcode::V_CMP_GE_U16},
-    {0xb9u, Opcode::V_CMPX_LT_U16, false},
+    {0xb9u, Opcode::V_CMPX_LT_U16, false}, {0xbau, Opcode::V_CMPX_EQ_U16, false},
    {0xbcu, Opcode::V_CMPX_GT_U16},        {0xc0u, Opcode::V_CMP_F_U32},
    {0xc1u, Opcode::V_CMP_LT_U32},         {0xc2u, Opcode::V_CMP_EQ_U32},
    {0xc3u, Opcode::V_CMP_LE_U32},         {0xc4u, Opcode::V_CMP_GT_U32},
@@ -250,7 +247,8 @@ constexpr VopcOpcodeInfo VOPC_OPCODE_LIST[] = {
    {0xf5u, Opcode::V_CMPX_NE_U64, false}, {0xc9u, Opcode::V_CMP_LT_F16},
    {0xcau, Opcode::V_CMP_EQ_F16},         {0xcbu, Opcode::V_CMP_LE_F16},
    {0xccu, Opcode::V_CMP_GT_F16},         {0xcdu, Opcode::V_CMP_LG_F16},
-    {0xceu, Opcode::V_CMP_GE_F16},         {0xedu, Opcode::V_CMP_NEQ_F16},
+    {0xceu, Opcode::V_CMP_GE_F16},         {0xebu, Opcode::V_CMP_NGT_F16},
+    {0xedu, Opcode::V_CMP_NEQ_F16},        {0xeeu, Opcode::V_CMP_NLT_F16},
    {0xd9u, Opcode::V_CMPX_LT_F16},        {0xdau, Opcode::V_CMPX_EQ_F16},
    {0xdbu, Opcode::V_CMPX_LE_F16},        {0xdcu, Opcode::V_CMPX_GT_F16},
    {0xdeu, Opcode::V_CMPX_GE_F16},        {0xfbu, Opcode::V_CMPX_NGT_F16},
@@ -298,6 +296,7 @@ constexpr OpcodeMap VOP3_OPCODE_LIST[] = {
    {0x300u, Opcode::V_LSHRREV_B64},
    {0x303u, Opcode::V_ADD_NC_U16},
    {0x304u, Opcode::V_SUB_NC_U16},
+    {0x305u, Opcode::V_MUL_LO_U16},
    {0x307u, Opcode::V_LSHRREV_B16},
    {0x308u, Opcode::V_ASHRREV_I16},
    {0x309u, Opcode::V_MAX_U16},
@@ -314,6 +313,7 @@ constexpr OpcodeMap VOP3_OPCODE_LIST[] = {
    {0x345u, Opcode::V_XAD_U32},
    {0x346u, Opcode::V_LSHL_ADD_U32},
    {0x347u, Opcode::V_ADD_LSHL_U32},
+    {0x35eu, Opcode::V_MAD_I16},
    {0x360u, Opcode::V_READLANE_B32},
    {0x361u, Opcode::V_WRITELANE_B32},
    {0x362u, Opcode::V_LDEXP_F32},
@@ -374,7 +374,7 @@ Opcode LookupVop3Opcode(uint32_t opcode) {
 		if (IsUnsupportedVop3EncodedVop2Alias(opcode - 0x100u)) {
 			return Opcode::UNSUPPORTED;
 		}
-		return LookupVop2Opcode(opcode - 0x100u);
+		return Detail::LookupOpcode(VOP2_OPS, opcode - 0x100u);
 	}
 	if (opcode >= 0x180u && opcode <= 0x1ffu) {
 		return Detail::LookupOpcode(VOP3_ENCODED_VOP1_OPS, opcode - 0x180u);
@@ -426,13 +426,14 @@ bool IsNativeVop3F16TernaryOpcode(Opcode opcode) {
 }

 bool IsNativeVop3I16TernaryOpcode(Opcode opcode) {
-	return opcode == Opcode::V_MED3_I16;
+	return opcode == Opcode::V_MED3_I16 || opcode == Opcode::V_MAD_I16;
 }

 bool IsNativeVop3B16BinaryOpcode(Opcode opcode) {
 	switch (opcode) {
 		case Opcode::V_ADD_NC_U16:
 		case Opcode::V_SUB_NC_U16:
+		case Opcode::V_MUL_LO_U16:
 		case Opcode::V_MAX_U16:
 		case Opcode::V_MAX_I16:
 		case Opcode::V_MIN_U16:
@@ -521,13 +522,16 @@ constexpr Vop1SdwaRule VOP1_SDWA_RULES[] = {
    {Opcode::V_CVT_F32_U32, SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), 0, 0, false},
    {Opcode::V_CVT_F32_I32, SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), 0, 0, false},
    {Opcode::V_CVT_F32_UBYTE0, SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), 0, 0, false},
-    {Opcode::V_NOT_B32, SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), 0, 0, false},
+    // The source field extract and the destination insert are independent (Translate.cpp), so a
+    // partial destination may combine with any source selector.
+    {Opcode::V_NOT_B32, SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), SdwaSelBytes() | SdwaSelWords(),
+     SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), false},
    {Opcode::V_FFBL_B32, SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), 0, 0, false},
    {Opcode::V_CVT_F32_F16, SdwaSelWords() | SdwaSelFull(), 0, 0, true},
    {Opcode::V_CVT_F16_F32, SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), SdwaSelWords(),
     SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), true},
-    {Opcode::V_CVT_F16_U16, SdwaSelWords() | SdwaSelFull(), SdwaSelWords(),
-     SdwaSelWords() | SdwaSelFull(), false},
+    {Opcode::V_CVT_F16_U16, SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), SdwaSelWords(),
+     SdwaSelBytes() | SdwaSelWords() | SdwaSelFull(), false},
    {Opcode::V_CVT_U16_F16, SdwaSelWords() | SdwaSelFull(), SdwaSelWords(),
     SdwaSelWords() | SdwaSelFull(), true},
    {Opcode::V_CVT_F16_I16, SdwaSelWords() | SdwaSelFull(), SdwaSelWords(),
@@ -617,13 +621,14 @@ bool SupportsVop1Clamp(Opcode opcode) {
 }

 bool ValidateVop1Sdwa(Instruction& inst, uint32_t opcode, uint32_t modifier) {
-	const auto dst_sel  = (modifier >> 8u) & 0x7u;
-	const auto dst_u    = (modifier >> 11u) & 0x3u;
-	const auto clamp    = (modifier >> 13u) & 0x1u;
-	const auto omod     = (modifier >> 14u) & 0x3u;
-	const auto src0_sel = (modifier >> 16u) & 0x7u;
-	const auto src0_neg = (modifier >> 20u) & 0x1u;
-	const auto src0_abs = (modifier >> 21u) & 0x1u;
+	const auto dst_sel   = (modifier >> 8u) & 0x7u;
+	const auto dst_u     = (modifier >> 11u) & 0x3u;
+	const auto clamp     = (modifier >> 13u) & 0x1u;
+	const auto omod      = (modifier >> 14u) & 0x3u;
+	const auto src0_sel  = (modifier >> 16u) & 0x7u;
+	const auto src0_sext = (modifier >> 19u) & 0x1u;
+	const auto src0_neg  = (modifier >> 20u) & 0x1u;
+	const auto src0_abs  = (modifier >> 21u) & 0x1u;

 	if (src0_sel > 6u || dst_sel > 6u) {
 		SetUnsupported(inst, Family::VOP1, opcode, "VOP1 SDWA selector is invalid");
@@ -639,7 +644,8 @@ bool ValidateVop1Sdwa(Instruction& inst, uint32_t opcode, uint32_t modifier) {
 		               "VOP1 SDWA destination selector is not supported");
 		return false;
 	}
-	if (!IsVop1SdwaSourceSupported(inst.opcode, src0_sel, src0_neg != 0u, src0_abs != 0u)) {
+	if (!IsVop1SdwaSourceSupported(inst.opcode, src0_sel, src0_neg != 0u, src0_abs != 0u) ||
+	    (inst.opcode == Opcode::V_CVT_F16_U16 && src0_sel <= 3u && src0_sext != 0u)) {
 		SetUnsupported(inst, Family::VOP1, opcode, "VOP1 SDWA source selector is not supported");
 		return false;
 	}
@@ -685,10 +691,15 @@ void DecodeVop1Sdwa(uint32_t pc, std::span<const uint32_t> code, uint32_t word_i
 	ReadLiteralOperands(code, word_index, inst);
 }

-void ApplyDppModifier(Operand& operand, uint32_t modifier) {
+void ApplyDppModifier(Operand& operand, uint32_t modifier, uint32_t encoding) {
+	operand.dpp = true;
+	if (encoding == 233u) {
+		operand.dpp8     = true;
+		operand.dpp_ctrl = modifier >> 8u;
+		return;
+	}
 	operand.negate             = ((modifier >> 20u) & 0x1u) != 0u;
 	operand.absolute           = ((modifier >> 21u) & 0x1u) != 0u;
-	operand.dpp                = true;
 	operand.dpp_ctrl           = (modifier >> 8u) & 0x1ffu;
 	operand.dpp_fetch_inactive = ((modifier >> 18u) & 0x1u) != 0u;
 	operand.dpp_bound_ctrl     = ((modifier >> 19u) & 0x1u) != 0u;
@@ -709,7 +720,7 @@ void DecodeVop1Dpp(uint32_t pc, std::span<const uint32_t> code, uint32_t word_in
 		DecodeVectorGpr(vdst, inst.dst);
 	}
 	DecodeScalarSource(src0 + 256u, pc, inst.src0);
-	ApplyDppModifier(inst.src0, modifier);
+	ApplyDppModifier(inst.src0, modifier, code[word_index] & 0x1ffu);
 	inst.src_count = 1;

 	if (!IsVop1FloatSourceOpcode(inst.opcode) && (inst.src0.negate || inst.src0.absolute)) {
@@ -812,6 +823,7 @@ bool IsVopcFloatCompareOpcode(Opcode opcode) {
 		case Opcode::V_CMPX_GT_F32:
 		case Opcode::V_CMPX_LG_F32:
 		case Opcode::V_CMPX_GE_F32:
+		case Opcode::V_CMPX_O_F32:
 		case Opcode::V_CMPX_NGE_F32:
 		case Opcode::V_CMPX_NLG_F32:
 		case Opcode::V_CMPX_NGT_F32:
@@ -824,7 +836,9 @@ bool IsVopcFloatCompareOpcode(Opcode opcode) {
 		case Opcode::V_CMP_GT_F16:
 		case Opcode::V_CMP_LG_F16:
 		case Opcode::V_CMP_GE_F16:
+		case Opcode::V_CMP_NGT_F16:
 		case Opcode::V_CMP_NEQ_F16:
+		case Opcode::V_CMP_NLT_F16:
 		case Opcode::V_CMPX_LT_F16:
 		case Opcode::V_CMPX_EQ_F16:
 		case Opcode::V_CMPX_LE_F16:
@@ -1046,7 +1060,7 @@ void DecodeVop2Dpp(uint32_t pc, std::span<const uint32_t> code, uint32_t word_in
 	DecodeVectorGpr(vdst, inst.dst);
 	DecodeVectorGpr(vsrc1, inst.src1);
 	DecodeScalarSource(src0 + 256u, pc, inst.src0);
-	ApplyDppModifier(inst.src0, modifier);
+	ApplyDppModifier(inst.src0, modifier, code[word_index] & 0x1ffu);
 	inst.src1.negate       = ((modifier >> 22u) & 0x1u) != 0u;
 	inst.src1.absolute     = ((modifier >> 23u) & 0x1u) != 0u;
 	const bool packed_fmac = inst.opcode == Opcode::V_PK_FMAC_F16;
@@ -1155,7 +1169,7 @@ void DecodeVopcDpp(uint32_t pc, std::span<const uint32_t> code, uint32_t word_in
 	DecodeVectorGpr(vsrc1, inst.src1);
 	DecodeScalarSource(src0 + 256u, pc, inst.src0);
 	inst.dst.kind = IsVopcCompareExec(inst.opcode) ? OperandKind::ExecLo : OperandKind::VccLo;
-	ApplyDppModifier(inst.src0, modifier);
+	ApplyDppModifier(inst.src0, modifier, code[word_index] & 0x1ffu);
 	inst.src1.negate   = ((modifier >> 22u) & 0x1u) != 0u;
 	inst.src1.absolute = ((modifier >> 23u) & 0x1u) != 0u;
 	inst.src_count     = 2;
@@ -1171,6 +1185,7 @@ uint32_t NativeVop3SourceCount(Opcode opcode) {
 		case Opcode::V_MUL_I32_I24:
 		case Opcode::V_ADD_NC_U16:
 		case Opcode::V_SUB_NC_U16:
+		case Opcode::V_MUL_LO_U16:
 		case Opcode::V_MAX_U16:
 		case Opcode::V_MAX_I16:
 		case Opcode::V_MIN_U16:
@@ -1355,7 +1370,8 @@ bool SupportsNativeVop3ResultModifiers(Opcode opcode) {
 }

 bool SupportsNativeVop3Clamp(Opcode opcode) {
-	return SupportsNativeVop3ResultModifiers(opcode) || UsesInexactClampControl(opcode);
+	return SupportsNativeVop3ResultModifiers(opcode) || UsesInexactClampControl(opcode) ||
+	       opcode == Opcode::V_MAD_I16;
 }

 bool HasUnsupportedNativeVop3Modifiers(Opcode opcode, bool permlane, bool mad_mix,
@@ -1376,7 +1392,7 @@ bool HasUnsupportedNativeVop3Modifiers(Opcode opcode, bool permlane, bool mad_mi
 		return opcode != Opcode::V_FMA_F16 && (clamp != 0u || omod != 0u);
 	}
 	if (IsNativeVop3I16TernaryOpcode(opcode)) {
-		return abs != 0u || clamp != 0u || omod != 0u || neg != 0u;
+		return abs != 0u || (clamp != 0u && !clamp_modifier) || omod != 0u || neg != 0u;
 	}
 	if (IsNativeVop3B16BinaryOpcode(opcode)) {
 		return abs != 0u || clamp != 0u || omod != 0u || neg != 0u;
@@ -1420,6 +1436,7 @@ bool IsVopcCompareExec(Opcode opcode) {
 		case Opcode::V_CMPX_GT_F32:
 		case Opcode::V_CMPX_LG_F32:
 		case Opcode::V_CMPX_GE_F32:
+		case Opcode::V_CMPX_O_F32:
 		case Opcode::V_CMPX_NGE_F32:
 		case Opcode::V_CMPX_NLG_F32:
 		case Opcode::V_CMPX_NGT_F32:
@@ -1442,6 +1459,7 @@ bool IsVopcCompareExec(Opcode opcode) {
 		case Opcode::V_CMPX_NE_I64:
 		case Opcode::V_CMPX_NE_U64:
 		case Opcode::V_CMPX_LT_U16:
+		case Opcode::V_CMPX_EQ_U16:
 		case Opcode::V_CMPX_GT_U16:
 		case Opcode::V_CMPX_LT_F16:
 		case Opcode::V_CMPX_EQ_F16:
@@ -1468,7 +1486,7 @@ void DecodeVop2(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 	inst.pc        = pc;
 	inst.family    = Family::VOP2;
 	inst.opcode_id = opcode;
-	inst.opcode    = LookupVop2Opcode(opcode);
+	inst.opcode    = Detail::LookupOpcode(VOP2_OPS, opcode);
 	SetRawWords(inst, code, word_index, 1);

 	if (inst.opcode == Opcode::UNSUPPORTED) {
@@ -1510,6 +1528,7 @@ void DecodeVop1(uint32_t pc, std::span<const uint32_t> code, uint32_t word_index
 		return;
 	}
 	switch (src0) {
+		case 233u: DecodeVop1Dpp(pc, code, word_index, opcode, vdst, inst); return;
 		case 249u: DecodeVop1Sdwa(pc, code, word_index, opcode, vdst, inst); return;
 		case 250u: DecodeVop1Dpp(pc, code, word_index, opcode, vdst, inst); return;
 		default: break;
@@ -106,12 +106,12 @@ void Translator::EXP(const Decoder::Instruction& inst) {
 	ir.Emit(IR::ValueOpcode::SetAttribute, {data, ir.GetExec()}, AddExportInfo(inst));
 }

-bool Translator::EmitInterpolation(const Decoder::Instruction& inst) {
+void Translator::EmitInterpolation(const Decoder::Instruction& inst) {
 	switch (inst.opcode) {
-		case Decoder::Opcode::V_INTERP_P1_F32: V_INTERP_P1_F32(); return true;
-		case Decoder::Opcode::V_INTERP_P2_F32: V_INTERP_P2_F32(inst); return true;
-		case Decoder::Opcode::V_INTERP_MOV_F32: V_INTERP_MOV_F32(inst); return true;
-		default: return false;
+		case Decoder::Opcode::V_INTERP_P1_F32: V_INTERP_P1_F32(); return;
+		case Decoder::Opcode::V_INTERP_P2_F32: V_INTERP_P2_F32(inst); return;
+		case Decoder::Opcode::V_INTERP_MOV_F32: V_INTERP_MOV_F32(inst); return;
+		default: return FailMissingTranslation(inst);
 	}
 }

@@ -47,12 +47,12 @@ void Translator::EmitFloatCompare(const Decoder::Instruction& inst, IR::ValueOpc
 	EmitCompareResult(inst, IR::U1(ir.Emit(opcode, {lhs, rhs})), false, cmpx);
 }

-void Translator::EmitFloatOrderedCompare(const Decoder::Instruction& inst, bool ordered) {
+void Translator::EmitFloatOrderedCompare(const Decoder::Instruction& inst, bool ordered, bool cmpx) {
 	const auto lhs       = IR::F32(ReadOperand(inst.src0, IR::Type::F32));
 	const auto rhs       = IR::F32(ReadOperand(inst.src1, IR::Type::F32));
 	const auto unordered = ir.LogicalOr(IR::U1(ir.Emit(IR::ValueOpcode::FPIsNan32, {lhs})),
 	                                    IR::U1(ir.Emit(IR::ValueOpcode::FPIsNan32, {rhs})));
-	EmitCompareResult(inst, ordered ? ir.LogicalNot(unordered) : unordered, false, false);
+	EmitCompareResult(inst, ordered ? ir.LogicalNot(unordered) : unordered, false, cmpx);
 }

 void Translator::EmitFloatClassCompare(const Decoder::Instruction& inst, bool cmpx) {
@@ -302,23 +302,95 @@ void Translator::S_MOV_B64(const Decoder::Instruction& inst) {
 }

 void Translator::S_WQM(const Decoder::Instruction& inst, bool wide) {
-	const auto source = wide ? ReadU64(inst.src0)
-	                         : ir.ConstructU64(ReadU32(inst.src0), IR::U32(IR::Value(0u)));
+	const bool source_is_mask = inst.src0.kind == Decoder::OperandKind::Sgpr ||
+	                            inst.src0.kind == Decoder::OperandKind::ExecLo ||
+	                            inst.src0.kind == Decoder::OperandKind::VccLo ||
+	                            (!wide && IsExecOrVcc(inst.src0));
+	const auto source_valid = ReadMaskValid(inst.src0);
+	auto source_live = IR::U1(IR::Value(false));
+	if (source_is_mask) {
+		if (inst.src0.kind == Decoder::OperandKind::VccLo ||
+		    inst.src0.kind == Decoder::OperandKind::VccHi) {
+			source_live = ir.GetVcc();
+		} else {
+			source_live = ReadMask(inst.src0);
+		}
+	}
+	const auto retained_live = ir.LogicalAnd(source_valid, source_live);
+
+	IR::U64 source;
+	if (wide) {
+		source = ReadU64(inst.src0);
+	} else {
+		source = ir.ConstructU64(ReadU32(inst.src0), IR::U32(IR::Value(0u)));
+	}
 	const auto result = IR::U64(ir.Emit(IR::ValueOpcode::WqmU64, {source}));
 	if (!wide) {
 		const auto low = ir.CompositeExtract(result, 0);
+		const bool destination_is_mask = IsExecOrVcc(inst.dst);
+		const bool destination_is_exec = inst.dst.kind == Decoder::OperandKind::ExecLo ||
+		                                 inst.dst.kind == Decoder::OperandKind::ExecHi;
+		auto old_live = IR::U1(IR::Value(false));
+		if (destination_is_mask) {
+			if (destination_is_exec) {
+				old_live = ir.GetExec();
+			} else {
+				old_live = ir.GetVcc();
+			}
+		}
+
 		// A 32-bit write preserves the other EXEC/VCC half and invalidates any
 		// overlapping SGPR-pair mask provenance through the ordinary scalar path.
 		WriteRawU32(inst.dst, low);
+		if (destination_is_mask) {
+			const auto is_high_half = [](const Decoder::Operand& operand) {
+				return operand.kind == Decoder::OperandKind::ExecHi ||
+				       operand.kind == Decoder::OperandKind::VccHi;
+			};
+			const bool destination_is_high = is_high_half(inst.dst);
+			auto affected_lanes = IR::U1(IR::Value(!destination_is_high));
+			if (program.wave_size == 64u) {
+				const auto lane = IR::U32(ir.Emit(IR::ValueOpcode::LaneId));
+				affected_lanes = ir.ULessThan(lane, IR::U32(IR::Value(32u)));
+				if (destination_is_high) {
+					affected_lanes = ir.LogicalNot(affected_lanes);
+				}
+			}
+
+			IR::U1 live;
+			if (destination_is_exec) {
+				live = ir.GetExec();
+			} else {
+				live = ir.GetVcc();
+			}
+			// A cross-half copy refers to different lanes, so only scalar bits apply.
+			if (is_high_half(inst.src0) == destination_is_high) {
+				live = ir.LogicalOr(retained_live, live);
+			}
+			live = IR::U1(ir.Emit(IR::ValueOpcode::SelectU1, {affected_lanes, live, old_live}));
+			if (destination_is_exec) {
+				ir.SetExec(live);
+			} else {
+				ir.SetVcc(live);
+			}
+		}
 		ir.SetScc(ir.INotEqual(low, IR::U32(IR::Value(0u))));
 		return;
 	}
-	const auto mask_valid = ReadMaskValid(inst.src0);
+
+	// WQM adds live lanes but never removes a source lane. Keep that fact separate
+	// from the scalar mask: reconstructing an entry-true predicate from a ballot
+	// otherwise introduces a spurious dependency into every vector write.
+	const auto live = ir.LogicalOr(retained_live, ThreadBit(ExtractU64(result)));
 	WriteOperand(DestinationOperand(inst), result);
-	if (inst.dst.kind == Decoder::OperandKind::Sgpr) {
+	if (inst.dst.kind == Decoder::OperandKind::ExecLo) {
+		ir.SetExec(live);
+	} else if (inst.dst.kind == Decoder::OperandKind::VccLo) {
+		ir.SetVcc(live);
+	} else if (inst.dst.kind == Decoder::OperandKind::Sgpr) {
 		const auto dst = static_cast<IR::ScalarReg>(inst.dst.reg);
-		ir.SetThreadBitScalarReg(dst, ThreadBit(ExtractU64(result)));
-		ir.SetScalarMaskTag(dst, mask_valid);
+		ir.SetThreadBitScalarReg(dst, live);
+		ir.SetScalarMaskTag(dst, source_valid);
 	}
 	ir.SetScc(IR::U1(ir.Emit(IR::ValueOpcode::INotEqual64, {result, IR::Value(uint64_t {0})})));
 }
@@ -166,9 +166,9 @@ void Translator::V_CVT_PK_U8_F32(const Decoder::Instruction& inst) {
 }

 void Translator::V_PACK_B32_F16(const Decoder::Instruction& inst) {
-	const auto low = ReadF16LaneBits(inst.src0, false);
+	const auto low = Read16LaneBits(inst.src0, false);
 	const auto high =
-	    ir.ShiftLeftLogical(ReadF16LaneBits(inst.src1, false), IR::U32(IR::Value(16u)));
+	    ir.ShiftLeftLogical(Read16LaneBits(inst.src1, false), IR::U32(IR::Value(16u)));
 	WriteOperand(DestinationOperand(inst), ir.BitwiseOr(low, high));
 }

@@ -3,6 +3,11 @@

 namespace Libs::Graphics::ShaderRecompiler::Frontend {

+void Translator::FailMissingTranslation(const Decoder::Instruction& inst) {
+	EXIT("opcode %s at pc 0x%08x has no IR translation",
+	     Decoder::InstructionToString(inst).c_str(), inst.pc);
+}
+
 void Translator::TranslateInstruction(const Decoder::Instruction& inst) {
 	current_opcode = inst.opcode;
 	current_pc     = inst.pc;
@@ -16,35 +21,26 @@ void Translator::TranslateInstruction(const Decoder::Instruction& inst) {
 		default: break;
 	}

-	bool translated = false;
 	switch (inst.family) {
 		case Decoder::Family::SOP1:
 		case Decoder::Family::SOP2:
 		case Decoder::Family::SOPK:
 		case Decoder::Family::SOPC:
-		case Decoder::Family::SOPP: translated = EmitScalar(inst); break;
+		case Decoder::Family::SOPP: return EmitScalar(inst);
 		case Decoder::Family::VOP1:
 		case Decoder::Family::VOP2:
 		case Decoder::Family::VOP3:
 		case Decoder::Family::VOP3P:
-		case Decoder::Family::VOPC: translated = EmitVector(inst); break;
+		case Decoder::Family::VOPC: return EmitVector(inst);
 		case Decoder::Family::SMEM:
 		case Decoder::Family::MUBUF:
 		case Decoder::Family::MTBUF:
 		case Decoder::Family::FLAT:
 		case Decoder::Family::DS:
-		case Decoder::Family::MIMG: translated = EmitMemory(inst); break;
-		case Decoder::Family::VINTRP: translated = EmitInterpolation(inst); break;
-		case Decoder::Family::EXP:
-			EXP(inst);
-			translated = true;
-			break;
-		default: break;
-	}
-
-	if (!translated) {
-		EXIT("opcode %s at pc 0x%08x has no IR translation",
-		     Decoder::InstructionToString(inst).c_str(), inst.pc);
+		case Decoder::Family::MIMG: return EmitMemory(inst);
+		case Decoder::Family::VINTRP: return EmitInterpolation(inst);
+		case Decoder::Family::EXP: return EXP(inst);
+		default: return FailMissingTranslation(inst);
 	}
 }

@@ -5,7 +5,7 @@

 namespace Libs::Graphics::ShaderRecompiler::Frontend {

-bool Translator::PackedFloat16(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::PackedFloat16(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                               bool accumulator, bool quiet_snan) {
 	const auto translate_lane = [&](bool high) {
 		const auto lhs = ReadF16LaneAsF32(inst.src0, high, true);
@@ -43,16 +43,15 @@ bool Translator::PackedFloat16(const Decoder::Instruction& inst, IR::ValueOpcode
 		result = PackU16Lanes(override_lane(false), override_lane(true));
 	}
 	WriteOperand(raw, result);
-	return true;
 }

-bool Translator::Float16Unary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::Float16Unary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                              bool invalid_negative) {
 	const auto argument = ReadF16AsF32(inst.src0);
 	auto       result   = IR::F32(ir.Emit(opcode, {argument}));
 	if (!invalid_negative) {
 		WriteF16(DestinationOperand(inst), result);
-		return true;
+		return;
 	}
 	const auto negative =
 	    IR::U1(ir.Emit(IR::ValueOpcode::FPOrdLessThan32, {argument, IR::Value::F32(0.0f)}));
@@ -60,10 +59,9 @@ bool Translator::Float16Unary(const Decoder::Instruction& inst, IR::ValueOpcode
 	const auto bits    = PackHalf2x16(result, IR::F32(IR::Value::F32(0.0f)));
 	const auto invalid = IR::U32(IR::Value(inst.dst.clamp ? 0u : 0xfe00u));
 	Write16Bits(DestinationOperand(inst), ir.Select(negative, invalid, bits));
-	return true;
 }

-bool Translator::Float16Trig(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
+void Translator::Float16Trig(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
 	const auto argument = ReadF16AsF32(inst.src0);
 	const auto magnitude =
 	    ir.BitwiseAnd(ir.BitCastU32(argument), IR::U32(IR::Value(0x7fffffffu)));
@@ -96,18 +94,16 @@ bool Translator::Float16Trig(const Decoder::Instruction& inst, IR::ValueOpcode o
 	const auto bits    = PackHalf2x16(result, IR::F32(IR::Value::F32(0.0f)));
 	const auto invalid = IR::U32(IR::Value(inst.dst.clamp ? 0u : 0xfe00u));
 	Write16Bits(DestinationOperand(inst), ir.Select(infinite, invalid, bits));
-	return true;
 }

-bool Translator::Float16Binary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::Float16Binary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                               bool reverse) {
 	const auto lhs = ReadF16AsF32(reverse ? inst.src1 : inst.src0);
 	const auto rhs = ReadF16AsF32(reverse ? inst.src0 : inst.src1);
 	WriteF16(DestinationOperand(inst), IR::F32(ir.Emit(opcode, {lhs, rhs})));
-	return true;
 }

-bool Translator::Float16Ternary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::Float16Ternary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                                bool accumulator, bool mix) {
 	std::array<IR::Value, 3> args;
 	for (uint32_t index = 0; index < args.size(); index++) {
@@ -115,16 +111,14 @@ bool Translator::Float16Ternary(const Decoder::Instruction& inst, IR::ValueOpcod
 		args[index] = mix ? IR::Value(ReadMixF32(operand)) : IR::Value(ReadF16AsF32(operand));
 	}
 	WriteF16(DestinationOperand(inst), IR::F32(ir.Emit(opcode, {args[0], args[1], args[2]})));
-	return true;
 }

-bool Translator::FloatUnary(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
+void Translator::FloatUnary(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
 	const auto type = IR::ArgTypeOf(opcode, 0);
 	WriteOperand(DestinationOperand(inst), ir.Emit(opcode, {ReadOperand(inst.src0, type)}));
-	return true;
 }

-bool Translator::FloatBinary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::FloatBinary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                             bool reverse) {
 	std::array<IR::Value, 2> args;
 	for (uint32_t index = 0; index < args.size(); index++) {
@@ -132,10 +126,9 @@ bool Translator::FloatBinary(const Decoder::Instruction& inst, IR::ValueOpcode o
 		args[index]        = ReadOperand(operand, IR::ArgTypeOf(opcode, index));
 	}
 	WriteOperand(DestinationOperand(inst), ir.Emit(opcode, {args[0], args[1]}));
-	return true;
 }

-bool Translator::FloatTernary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::FloatTernary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                              bool accumulator, bool mix) {
 	std::array<IR::Value, 3> args;
 	for (uint32_t index = 0; index < args.size(); index++) {
@@ -145,10 +138,9 @@ bool Translator::FloatTernary(const Decoder::Instruction& inst, IR::ValueOpcode
 		                                                  : ReadOperand(operand, type);
 	}
 	WriteOperand(DestinationOperand(inst), ir.Emit(opcode, {args[0], args[1], args[2]}));
-	return true;
 }

-bool Translator::V_FREXP_MANT_F32(const Decoder::Instruction& inst) {
+void Translator::V_FREXP_MANT_F32(const Decoder::Instruction& inst) {
 	const auto bits = ReadU32(inst.src0);
 	const auto exponent =
 	    IR::U32(ir.Emit(IR::ValueOpcode::BitFieldUExtract, {bits, IR::Value(23u), IR::Value(8u)}));
@@ -166,10 +158,9 @@ bool Translator::V_FREXP_MANT_F32(const Decoder::Instruction& inst) {
 	                                 ir.Select(zero, bits, subnormal));
 	const auto result    = ir.Select(ir.IEqual(exponent, IR::U32(IR::Value(0xffu))), bits, finite);
 	WriteOperand(DestinationOperand(inst), ir.BitCastF32(result));
-	return true;
 }

-bool Translator::V_DOT2C_F32_F16(const Decoder::Instruction& inst) {
+void Translator::V_DOT2C_F32_F16(const Decoder::Instruction& inst) {
 	auto a          = inst.src0;
 	a.op_sel        = false;
 	a.op_sel_hi     = true;
@@ -183,26 +174,25 @@ bool Translator::V_DOT2C_F32_F16(const Decoder::Instruction& inst) {
 	const auto acc  = ReadMixF32(inst.dst);
 	const auto lo   = IR::F32(ir.Emit(IR::ValueOpcode::FPFma32, {a_lo, b_lo, acc}));
 	WriteOperand(DestinationOperand(inst), ir.Emit(IR::ValueOpcode::FPFma32, {a_hi, b_hi, lo}));
-	return true;
 }

-bool Translator::V_CUBEID_F32(const Decoder::Instruction& inst) {
+void Translator::V_CUBEID_F32(const Decoder::Instruction& inst) {
 	return FloatCube(inst, 0u);
 }

-bool Translator::V_CUBESC_F32(const Decoder::Instruction& inst) {
+void Translator::V_CUBESC_F32(const Decoder::Instruction& inst) {
 	return FloatCube(inst, 1u);
 }

-bool Translator::V_CUBETC_F32(const Decoder::Instruction& inst) {
+void Translator::V_CUBETC_F32(const Decoder::Instruction& inst) {
 	return FloatCube(inst, 2u);
 }

-bool Translator::V_CUBEMA_F32(const Decoder::Instruction& inst) {
+void Translator::V_CUBEMA_F32(const Decoder::Instruction& inst) {
 	return FloatCube(inst, 3u);
 }

-bool Translator::FloatCube(const Decoder::Instruction& inst, uint32_t result_kind) {
+void Translator::FloatCube(const Decoder::Instruction& inst, uint32_t result_kind) {
 	const auto x  = ReadMixF32(inst.src0);
 	const auto y  = ReadMixF32(inst.src1);
 	const auto z  = ReadMixF32(inst.src2);
@@ -242,7 +232,6 @@ bool Translator::FloatCube(const Decoder::Instruction& inst, uint32_t result_kin
 		default: EXIT("invalid cube result kind");
 	}
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

 } // namespace Libs::Graphics::ShaderRecompiler::Frontend
@@ -4,33 +4,41 @@

 namespace Libs::Graphics::ShaderRecompiler::Frontend {

-bool Translator::Integer16Shift(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::Integer16Shift(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                                bool arithmetic) {
 	const auto value  = ReadU16AsU32(inst.src1, arithmetic);
 	const auto count  = ir.BitwiseAnd(ReadU16AsU32(inst.src0, false), IR::U32(IR::Value(15u)));
 	const auto result = IR::U32(ir.Emit(opcode, {value, count}));
 	Write16Bits(DestinationOperand(inst), ir.BitwiseAnd(result, IR::U32(IR::Value(0xffffu))));
-	return true;
 }

-bool Translator::Integer16Binary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::Integer16Binary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                                 bool sign) {
 	const auto lhs    = ReadU16AsU32(inst.src0, sign);
 	const auto rhs    = ReadU16AsU32(inst.src1, sign);
 	const auto result = IR::U32(ir.Emit(opcode, {lhs, rhs}));
 	Write16Bits(DestinationOperand(inst), ir.BitwiseAnd(result, IR::U32(IR::Value(0xffffu))));
-	return true;
 }

-bool Translator::V_MED3_I16(const Decoder::Instruction& inst) {
+void Translator::V_MAD_I16(const Decoder::Instruction& inst) {
+	const auto lhs    = ReadU16AsU32(inst.src0, true);
+	const auto rhs    = ReadU16AsU32(inst.src1, true);
+	auto       result = ir.IAdd(ir.IMul(lhs, rhs), ReadU16AsU32(inst.src2, true));
+	if (inst.dst.clamp) {
+		result = IR::U32(ir.Emit(IR::ValueOpcode::SMax32, {result, IR::Value(0xffff8000u)}));
+		result = IR::U32(ir.Emit(IR::ValueOpcode::SMin32, {result, IR::Value(0x7fffu)}));
+	}
+	Write16Bits(DestinationOperand(inst), ir.BitwiseAnd(result, IR::U32(IR::Value(0xffffu))));
+}
+
+void Translator::V_MED3_I16(const Decoder::Instruction& inst) {
 	const auto result = IR::U32(ir.Emit(
 	    IR::ValueOpcode::SMedTri32, {ReadU16AsU32(inst.src0, true), ReadU16AsU32(inst.src1, true),
 	                                 ReadU16AsU32(inst.src2, true)}));
 	Write16Bits(DestinationOperand(inst), ir.BitwiseAnd(result, IR::U32(IR::Value(0xffffu))));
-	return true;
 }

-bool Translator::PackedInteger16Shift(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::PackedInteger16Shift(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                                      bool arithmetic) {
 	const auto translate_lane = [&](bool high) {
 		const auto count =
@@ -40,10 +48,9 @@ bool Translator::PackedInteger16Shift(const Decoder::Instruction& inst, IR::Valu
 	};
 	WriteOperand(DestinationOperand(inst),
 	             PackU16Lanes(translate_lane(false), translate_lane(true)));
-	return true;
 }

-bool Translator::PackedInteger16Binary(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
+void Translator::PackedInteger16Binary(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
 	const auto translate_lane = [&](bool high) {
 		const auto lhs = ReadU16LaneAsU32(inst.src0, high, false);
 		const auto rhs = ReadU16LaneAsU32(inst.src1, high, false);
@@ -51,10 +58,9 @@ bool Translator::PackedInteger16Binary(const Decoder::Instruction& inst, IR::Val
 	};
 	WriteOperand(DestinationOperand(inst),
 	             PackU16Lanes(translate_lane(false), translate_lane(true)));
-	return true;
 }

-bool Translator::PackedInteger16Mad(const Decoder::Instruction& inst, bool sign) {
+void Translator::PackedInteger16Mad(const Decoder::Instruction& inst, bool sign) {
 	const auto translate_lane = [&](bool high) {
 		const auto lhs = ReadU16LaneAsU32(inst.src0, high, false);
 		const auto rhs = ReadU16LaneAsU32(inst.src1, high, false);
@@ -62,10 +68,9 @@ bool Translator::PackedInteger16Mad(const Decoder::Instruction& inst, bool sign)
 	};
 	WriteOperand(DestinationOperand(inst),
 	             PackU16Lanes(translate_lane(false), translate_lane(true)));
-	return true;
 }

-bool Translator::PackedInteger16MinMax(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::PackedInteger16MinMax(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                                       bool sign) {
 	const auto translate_lane = [&](bool high) {
 		const auto lhs = ReadU16LaneAsU32(inst.src0, high, sign);
@@ -74,7 +79,6 @@ bool Translator::PackedInteger16MinMax(const Decoder::Instruction& inst, IR::Val
 	};
 	WriteOperand(DestinationOperand(inst),
 	             PackU16Lanes(translate_lane(false), translate_lane(true)));
-	return true;
 }

 IR::U1 Translator::U64MaskBinary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
@@ -88,7 +92,7 @@ IR::U1 Translator::U64MaskBinary(const Decoder::Instruction& inst, IR::ValueOpco
 	return negate_result ? ir.LogicalNot(result) : result;
 }

-bool Translator::S_U64_MASK(const Decoder::Instruction& inst, IR::ValueOpcode logical_opcode,
+void Translator::S_U64_MASK(const Decoder::Instruction& inst, IR::ValueOpcode logical_opcode,
                            IR::ValueOpcode bit_opcode, bool negate_rhs, bool negate_result,
                            bool unary) {
 	const auto invocation_result =
@@ -107,7 +111,7 @@ bool Translator::S_U64_MASK(const Decoder::Instruction& inst, IR::ValueOpcode lo
 	    (inst.src_count > 1u && is_exec_or_vcc(inst.src1))) {
 		const auto mask = WriteMask(inst.dst, invocation_result, true);
 		ir.SetScc(ir.INotEqual(ir.BitwiseOr(mask[0], mask[1]), IR::U32(IR::Value(0u))));
-		return true;
+		return;
 	}

 	const auto lhs        = ReadU32Pair(inst.src0);
@@ -134,10 +138,9 @@ bool Translator::S_U64_MASK(const Decoder::Instruction& inst, IR::ValueOpcode lo
 		ir.SetScalarMaskTag(dst, mask_valid);
 	}
 	ir.SetScc(ir.INotEqual(ir.BitwiseOr(result[0], result[1]), IR::U32(IR::Value(0u))));
-	return true;
 }

-bool Translator::SimpleInteger(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::SimpleInteger(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                               IR::Type type, bool reverse, bool mask_shift_count,
                               bool update_scc) {
 	std::array<IR::Value, 3> args;
@@ -165,10 +168,23 @@ bool Translator::SimpleInteger(const Decoder::Instruction& inst, IR::ValueOpcode
 			ir.SetScc(ir.INotEqual(IR::U32(result), IR::U32(IR::Value(0u))));
 		}
 	}
-	return true;
 }

-bool Translator::ComposedIntegerBinary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::S_ASHR_I64(const Decoder::Instruction& inst) {
+	// Signed 64-bit sources sign-extend literals; generic B64 operands zero-extend them.
+	const auto source =
+	    inst.src0.kind == Decoder::OperandKind::LiteralConstant
+	        ? ir.ConstructU64(
+	              ReadU32(inst.src0),
+	              IR::U32(IR::Value((inst.src0.value & 0x80000000u) ? 0xffffffffu : 0u)))
+	        : ReadU64(inst.src0);
+	const auto result =
+	    ir.Emit(IR::ValueOpcode::ShiftRightArithmetic64, {source, ReadU32(inst.src1)});
+	WriteOperand(inst.dst, result);
+	ir.SetScc(IR::U1(ir.Emit(IR::ValueOpcode::INotEqual64, {result, IR::Value(uint64_t {0})})));
+}
+
+void Translator::ComposedIntegerBinary(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                                       bool negate_rhs, bool negate_result, bool update_scc) {
 	const auto lhs = ReadU32(inst.src0);
 	auto       rhs = ReadU32(inst.src1);
@@ -183,31 +199,27 @@ bool Translator::ComposedIntegerBinary(const Decoder::Instruction& inst, IR::Val
 	if (update_scc) {
 		ir.SetScc(ir.INotEqual(result, IR::U32(IR::Value(0u))));
 	}
-	return true;
 }

-bool Translator::V_AND_OR_B32(const Decoder::Instruction& inst) {
+void Translator::V_AND_OR_B32(const Decoder::Instruction& inst) {
 	const auto result =
 	    ir.BitwiseOr(ir.BitwiseAnd(ReadU32(inst.src0), ReadU32(inst.src1)), ReadU32(inst.src2));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::V_OR3_B32(const Decoder::Instruction& inst) {
+void Translator::V_OR3_B32(const Decoder::Instruction& inst) {
 	const auto result =
 	    ir.BitwiseOr(ir.BitwiseOr(ReadU32(inst.src0), ReadU32(inst.src1)), ReadU32(inst.src2));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::V_XOR3_B32(const Decoder::Instruction& inst) {
+void Translator::V_XOR3_B32(const Decoder::Instruction& inst) {
 	const auto result =
 	    ir.BitwiseXor(ir.BitwiseXor(ReadU32(inst.src0), ReadU32(inst.src1)), ReadU32(inst.src2));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::S_FF1_I32_B64(const Decoder::Instruction& inst) {
+void Translator::S_FF1_I32_B64(const Decoder::Instruction& inst) {
 	const auto source        = ExtractU64(ReadU64(inst.src0));
 	const auto low_lsb       = IR::U32(ir.Emit(IR::ValueOpcode::FindILsb32, {source[0]}));
 	const auto high_lsb      = IR::U32(ir.Emit(IR::ValueOpcode::FindILsb32, {source[1]}));
@@ -216,10 +228,9 @@ bool Translator::S_FF1_I32_B64(const Decoder::Instruction& inst) {
 	                                     ir.Select(ir.INotEqual(source[1], IR::U32(IR::Value(0u))),
 	                                               high_position, IR::U32(IR::Value(0xffffffffu))));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::V_FFBH_32(const Decoder::Instruction& inst, bool sign) {
+void Translator::V_FFBH_32(const Decoder::Instruction& inst, bool sign) {
 	const auto source = ReadU32(inst.src0);
 	const auto value  = sign ? ir.BitwiseXor(
 	                              source, ir.ShiftRightArithmetic(source, IR::U32(IR::Value(31u))))
@@ -229,10 +240,9 @@ bool Translator::V_FFBH_32(const Decoder::Instruction& inst, bool sign) {
 	const auto result   = ir.Select(ir.INotEqual(value, IR::U32(IR::Value(0u))), position,
 	                                IR::U32(IR::Value(0xffffffffu)));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::S_FLBIT_I32_B64(const Decoder::Instruction& inst) {
+void Translator::S_FLBIT_I32_B64(const Decoder::Instruction& inst) {
 	const auto source   = ReadU64(inst.src0);
 	const auto msb      = IR::U32(ir.Emit(IR::ValueOpcode::FindUMsb64, {source}));
 	const auto position = ir.ISub(IR::U32(IR::Value(63u)), msb);
@@ -241,10 +251,9 @@ bool Translator::S_FLBIT_I32_B64(const Decoder::Instruction& inst) {
 	                   {source, ir.ConstructU64(IR::U32(IR::Value(0u)), IR::U32(IR::Value(0u)))}));
 	const auto result = ir.Select(nonzero, position, IR::U32(IR::Value(0xffffffffu)));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::Integer24(const Decoder::Instruction& inst, bool sign, bool addend) {
+void Translator::Integer24(const Decoder::Instruction& inst, bool sign, bool addend) {
 	const auto extract24 = [&](IR::U32 value) {
 		return IR::U32(
 		    ir.Emit(sign ? IR::ValueOpcode::BitFieldSExtract : IR::ValueOpcode::BitFieldUExtract,
@@ -257,10 +266,9 @@ bool Translator::Integer24(const Decoder::Instruction& inst, bool sign, bool add
 		result = ir.IAdd(result, ReadU32(inst.src2));
 	}
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::V_MAD_U64_U32(const Decoder::Instruction& inst) {
+void Translator::V_MAD_U64_U32(const Decoder::Instruction& inst) {
 	const auto lhs       = ReadU32(inst.src0);
 	const auto rhs       = ReadU32(inst.src1);
 	const auto add       = ExtractU64(ReadU64(inst.src2));
@@ -278,36 +286,32 @@ bool Translator::V_MAD_U64_U32(const Decoder::Instruction& inst) {
 	    inst.dst2.kind != Decoder::OperandKind::Unknown) {
 		WriteMask(inst.dst2, ir.LogicalOr(carry0, carry1));
 	}
-	return true;
 }

-bool Translator::V_SAD_U32(const Decoder::Instruction& inst) {
+void Translator::V_SAD_U32(const Decoder::Instruction& inst) {
 	const auto lhs    = ReadU32(inst.src0);
 	const auto rhs    = ReadU32(inst.src1);
 	const auto lo     = IR::U32(ir.Emit(IR::ValueOpcode::UMin32, {lhs, rhs}));
 	const auto hi     = IR::U32(ir.Emit(IR::ValueOpcode::UMax32, {lhs, rhs}));
 	const auto result = ir.IAdd(ir.ISub(hi, lo), ReadU32(inst.src2));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::V_ADD3_U32(const Decoder::Instruction& inst) {
+void Translator::V_ADD3_U32(const Decoder::Instruction& inst) {
 	const auto result =
 	    ir.IAdd(ir.IAdd(ReadU32(inst.src0), ReadU32(inst.src1)), ReadU32(inst.src2));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::S_BITSET_B32(const Decoder::Instruction& inst, bool set) {
+void Translator::S_BITSET_B32(const Decoder::Instruction& inst, bool set) {
 	const auto offset = ir.BitwiseAnd(ReadU32(inst.src0), IR::U32(IR::Value(31u)));
 	const auto bit    = ir.ShiftLeftLogical(IR::U32(IR::Value(1u)), offset);
 	const auto old    = ReadU32(inst.dst);
 	const auto result = set ? ir.BitwiseOr(old, bit) : ir.BitwiseAnd(old, ir.BitwiseNot(bit));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::S_BITSET_B64(const Decoder::Instruction& inst, bool set) {
+void Translator::S_BITSET_B64(const Decoder::Instruction& inst, bool set) {
 	const auto offset    = ir.BitwiseAnd(ReadU32(inst.src0), IR::U32(IR::Value(63u)));
 	const auto word_bit  = ir.BitwiseAnd(offset, IR::U32(IR::Value(31u)));
 	const auto bit       = ir.ShiftLeftLogical(IR::U32(IR::Value(1u)), word_bit);
@@ -320,16 +324,14 @@ bool Translator::S_BITSET_B64(const Decoder::Instruction& inst, bool set) {
 	                                {offset, IR::Value(32u)}));
 	WriteU32Pair(inst.dst,
 	             {ir.Select(high, old[0], low_value), ir.Select(high, high_value, old[1])});
-	return true;
 }

-bool Translator::V_BCNT_U32_B32(const Decoder::Instruction& inst) {
+void Translator::V_BCNT_U32_B32(const Decoder::Instruction& inst) {
 	const auto count = IR::U32(ir.Emit(IR::ValueOpcode::BitCount32, {ReadU32(inst.src0)}));
 	WriteOperand(DestinationOperand(inst), ir.IAdd(count, ReadU32(inst.src1)));
-	return true;
 }

-bool Translator::V_MBCNT_U32_B32(const Decoder::Instruction& inst, bool low) {
+void Translator::V_MBCNT_U32_B32(const Decoder::Instruction& inst, bool low) {
 	const auto lane  = IR::U32(ir.Emit(IR::ValueOpcode::LaneId));
 	const auto local = ir.BitwiseAnd(lane, IR::U32(IR::Value(31u)));
 	const auto below =
@@ -341,10 +343,9 @@ bool Translator::V_MBCNT_U32_B32(const Decoder::Instruction& inst, bool low) {
 	const auto active      = ir.BitwiseAnd(ReadU32(inst.src0), thread_mask);
 	const auto count       = IR::U32(ir.Emit(IR::ValueOpcode::BitCount32, {active}));
 	WriteOperand(DestinationOperand(inst), ir.IAdd(count, ReadU32(inst.src1)));
-	return true;
 }

-bool Translator::S_BITREPLICATE_B64_B32(const Decoder::Instruction& inst) {
+void Translator::S_BITREPLICATE_B64_B32(const Decoder::Instruction& inst) {
 	const auto replicate = [&](IR::U32 value) {
 		auto bits = ir.BitwiseOr(value, ir.ShiftLeftLogical(value, IR::U32(IR::Value(8u))));
 		bits      = ir.BitwiseAnd(bits, IR::U32(IR::Value(0x00ff00ffu)));
@@ -360,10 +361,9 @@ bool Translator::S_BITREPLICATE_B64_B32(const Decoder::Instruction& inst) {
 	const auto low    = ir.BitwiseAnd(source, IR::U32(IR::Value(0xffffu)));
 	const auto high   = ir.ShiftRightLogical(source, IR::U32(IR::Value(16u)));
 	WriteOperand(DestinationOperand(inst), ir.ConstructU64(replicate(low), replicate(high)));
-	return true;
 }

-bool Translator::S_QUADMASK_B64(const Decoder::Instruction& inst) {
+void Translator::S_QUADMASK_B64(const Decoder::Instruction& inst) {
 	const auto compact = [&](IR::U32 value) {
 		auto bits = ir.BitwiseOr(value, ir.ShiftRightLogical(value, IR::U32(IR::Value(1u))));
 		bits      = ir.BitwiseOr(bits, ir.ShiftRightLogical(bits, IR::U32(IR::Value(2u))));
@@ -382,16 +382,14 @@ bool Translator::S_QUADMASK_B64(const Decoder::Instruction& inst) {
 	const auto result = ir.ConstructU64(quads, IR::U32(IR::Value(0u)));
 	WriteOperand(DestinationOperand(inst), result);
 	ir.SetScc(IR::U1(ir.Emit(IR::ValueOpcode::INotEqual64, {result, IR::Value(uint64_t {0})})));
-	return true;
 }

-bool Translator::BFM_B32(const Decoder::Instruction& inst) {
+void Translator::BFM_B32(const Decoder::Instruction& inst) {
 	const auto count  = ir.BitwiseAnd(ReadU32(inst.src0), IR::U32(IR::Value(31u)));
 	const auto offset = ir.BitwiseAnd(ReadU32(inst.src1), IR::U32(IR::Value(31u)));
 	const auto result = ir.Emit(IR::ValueOpcode::BitFieldInsert,
 	                            {IR::Value(0u), IR::Value(0xffffffffu), offset, count});
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

 IR::U32 Translator::RightMask32(IR::U32 count) {
@@ -408,15 +406,14 @@ IR::U64 Translator::RightMask64(IR::U32 count) {
 	return ir.ConstructU64(RightMask32(low_count), RightMask32(high_count));
 }

-bool Translator::S_BFM_B64(const Decoder::Instruction& inst) {
+void Translator::S_BFM_B64(const Decoder::Instruction& inst) {
 	const auto count  = ir.BitwiseAnd(ReadU32(inst.src0), IR::U32(IR::Value(63u)));
 	const auto offset = ir.BitwiseAnd(ReadU32(inst.src1), IR::U32(IR::Value(63u)));
 	const auto result = ir.Emit(IR::ValueOpcode::ShiftLeftLogical64, {RightMask64(count), offset});
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::S_BFE_U32(const Decoder::Instruction& inst, bool sign) {
+void Translator::S_BFE_U32(const Decoder::Instruction& inst, bool sign) {
 	const auto source = ReadU32(inst.src0);
 	const auto field  = ReadU32(inst.src1);
 	const auto offset =
@@ -430,10 +427,9 @@ bool Translator::S_BFE_U32(const Decoder::Instruction& inst, bool sign) {
 	const auto result = IR::U32(ir.Emit(opcode, {source, offset, count}));
 	WriteOperand(DestinationOperand(inst), result);
 	ir.SetScc(ir.INotEqual(result, IR::U32(IR::Value(0u))));
-	return true;
 }

-bool Translator::S_BFE_U64(const Decoder::Instruction& inst) {
+void Translator::S_BFE_U64(const Decoder::Instruction& inst) {
 	const auto source = ReadU64(inst.src0);
 	const auto field  = ReadU32(inst.src1);
 	const auto offset =
@@ -446,10 +442,9 @@ bool Translator::S_BFE_U64(const Decoder::Instruction& inst) {
 	const auto result    = ir.Emit(IR::ValueOpcode::BitwiseAnd64, {shifted, RightMask64(count)});
 	WriteOperand(DestinationOperand(inst), result);
 	ir.SetScc(IR::U1(ir.Emit(IR::ValueOpcode::INotEqual64, {result, IR::Value(uint64_t {0})})));
-	return true;
 }

-bool Translator::V_BFE_U32(const Decoder::Instruction& inst, bool sign) {
+void Translator::V_BFE_U32(const Decoder::Instruction& inst, bool sign) {
 	const auto source    = ReadU32(inst.src0);
 	const auto offset    = ir.BitwiseAnd(ReadU32(inst.src1), IR::U32(IR::Value(31u)));
 	const auto raw_count = ir.BitwiseAnd(ReadU32(inst.src2), IR::U32(IR::Value(31u)));
@@ -458,29 +453,37 @@ bool Translator::V_BFE_U32(const Decoder::Instruction& inst, bool sign) {
 	const auto opcode =
 	    sign ? IR::ValueOpcode::BitFieldSExtract : IR::ValueOpcode::BitFieldUExtract;
 	WriteOperand(DestinationOperand(inst), ir.Emit(opcode, {source, offset, count}));
-	return true;
 }

-bool Translator::V_BFI_B32(const Decoder::Instruction& inst) {
+void Translator::V_BFI_B32(const Decoder::Instruction& inst) {
 	const auto bits   = ReadU32(inst.src0);
 	const auto insert = ReadU32(inst.src1);
 	const auto base   = ReadU32(inst.src2);
 	const auto result =
 	    ir.BitwiseOr(ir.BitwiseAnd(bits, insert), ir.BitwiseAnd(ir.BitwiseNot(bits), base));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::S_BITCMP_B32(const Decoder::Instruction& inst, bool expected) {
+void Translator::S_BITCMP_B32(const Decoder::Instruction& inst, bool expected) {
 	const auto value  = ReadU32(inst.src0);
 	const auto offset = ir.BitwiseAnd(ReadU32(inst.src1), IR::U32(IR::Value(31u)));
 	const auto bit =
 	    IR::U32(ir.Emit(IR::ValueOpcode::BitFieldUExtract, {value, offset, IR::Value(1u)}));
 	WriteCompareResult(inst.dst, ir.IEqual(bit, IR::U32(IR::Value(expected ? 1u : 0u))));
-	return true;
 }

-bool Translator::V_ALIGNBIT_B32(const Decoder::Instruction& inst) {
+void Translator::S_BITCMP_B64(const Decoder::Instruction& inst, bool expected) {
+	const auto value    = ReadU32Pair(inst.src0);
+	const auto offset   = ir.BitwiseAnd(ReadU32(inst.src1), IR::U32(IR::Value(63u)));
+	const auto word_bit = ir.BitwiseAnd(offset, IR::U32(IR::Value(31u)));
+	const auto word =
+	    ir.Select(ir.ULessThan(offset, IR::U32(IR::Value(32u))), value[0], value[1]);
+	const auto bit =
+	    IR::U32(ir.Emit(IR::ValueOpcode::BitFieldUExtract, {word, word_bit, IR::Value(1u)}));
+	WriteCompareResult(inst.dst, ir.IEqual(bit, IR::U32(IR::Value(expected ? 1u : 0u))));
+}
+
+void Translator::V_ALIGNBIT_B32(const Decoder::Instruction& inst) {
 	const auto hi      = ReadU32(inst.src0);
 	const auto lo      = ReadU32(inst.src1);
 	const auto shift   = ir.BitwiseAnd(ReadU32(inst.src2), IR::U32(IR::Value(31u)));
@@ -491,10 +494,9 @@ bool Translator::V_ALIGNBIT_B32(const Decoder::Instruction& inst) {
 	const auto hi_part =
 	    ir.Select(ir.INotEqual(shift, IR::U32(IR::Value(0u))), hi_part_raw, IR::U32(IR::Value(0u)));
 	WriteOperand(DestinationOperand(inst), ir.BitwiseOr(lo_part, hi_part));
-	return true;
 }

-bool Translator::V_ALIGNBYTE_B32(const Decoder::Instruction& inst) {
+void Translator::V_ALIGNBYTE_B32(const Decoder::Instruction& inst) {
 	const auto hi           = ReadU32(inst.src0);
 	const auto lo           = ReadU32(inst.src1);
 	const auto byte_offset  = ir.BitwiseAnd(ReadU32(inst.src2), IR::U32(IR::Value(31u)));
@@ -507,39 +509,34 @@ bool Translator::V_ALIGNBYTE_B32(const Decoder::Instruction& inst) {
 	    IR::U1(ir.Emit(IR::ValueOpcode::ULessThan32, {byte_offset, IR::Value(8u)}));
 	WriteOperand(DestinationOperand(inst),
 	             ir.Select(in_range, ExtractU64(shifted)[0], IR::U32(IR::Value(0u))));
-	return true;
 }

-bool Translator::V_LSHL_ADD_U32(const Decoder::Instruction& inst) {
+void Translator::V_LSHL_ADD_U32(const Decoder::Instruction& inst) {
 	const auto shift  = ir.BitwiseAnd(ReadU32(inst.src1), IR::U32(IR::Value(31u)));
 	const auto result = ir.IAdd(ir.ShiftLeftLogical(ReadU32(inst.src0), shift), ReadU32(inst.src2));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::V_ADD_LSHL_U32(const Decoder::Instruction& inst) {
+void Translator::V_ADD_LSHL_U32(const Decoder::Instruction& inst) {
 	const auto shift  = ir.BitwiseAnd(ReadU32(inst.src2), IR::U32(IR::Value(31u)));
 	const auto result = ir.ShiftLeftLogical(ir.IAdd(ReadU32(inst.src0), ReadU32(inst.src1)), shift);
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::V_XAD_U32(const Decoder::Instruction& inst) {
+void Translator::V_XAD_U32(const Decoder::Instruction& inst) {
 	const auto result =
 	    ir.IAdd(ir.BitwiseXor(ReadU32(inst.src0), ReadU32(inst.src1)), ReadU32(inst.src2));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::V_LSHL_OR_B32(const Decoder::Instruction& inst) {
+void Translator::V_LSHL_OR_B32(const Decoder::Instruction& inst) {
 	const auto shift = ir.BitwiseAnd(ReadU32(inst.src1), IR::U32(IR::Value(31u)));
 	const auto result =
 	    ir.BitwiseOr(ir.ShiftLeftLogical(ReadU32(inst.src0), shift), ReadU32(inst.src2));
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::V_CNDMASK_B32(const Decoder::Instruction& inst) {
+void Translator::V_CNDMASK_B32(const Decoder::Instruction& inst) {
 	Decoder::Operand mask_operand;
 	mask_operand.kind = Decoder::OperandKind::VccLo;
 	if (inst.src_count >= 3u) {
@@ -555,17 +552,15 @@ bool Translator::V_CNDMASK_B32(const Decoder::Instruction& inst) {
 		result = ir.Select(condition, ReadU32(inst.src1), ReadU32(inst.src0));
 	}
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

-bool Translator::PackB16(const Decoder::Instruction& inst, bool high0, bool high1) {
+void Translator::PackB16(const Decoder::Instruction& inst, bool high0, bool high1) {
 	const auto lo        = high0 ? ir.ShiftRightLogical(ReadU32(inst.src0), IR::U32(IR::Value(16u)))
 	                             : ReadU32(inst.src0);
 	const auto hi        = high1 ? ir.ShiftRightLogical(ReadU32(inst.src1), IR::U32(IR::Value(16u)))
 	                             : ReadU32(inst.src1);
 	const auto result = PackU16Lanes(lo, hi);
 	WriteOperand(DestinationOperand(inst), result);
-	return true;
 }

 } // namespace Libs::Graphics::ShaderRecompiler::Frontend
@@ -1,8 +1,11 @@
+#include "common/assert.h"
+#include "common/logging/log.h"
 #include "graphics/shader/recompiler/frontend/translate/Translator.h"
 #include "graphics/shader/recompiler/frontend/decode/ImageOps.h"

 #include <algorithm>
 #include <array>
+#include <atomic>

 namespace Libs::Graphics::ShaderRecompiler::Frontend {

@@ -34,6 +37,7 @@ Decoder::Operand OffsetDecodedRegister(const Decoder::Operand& operand, uint32_t
 	result.negate_hi          = false;
 	result.absolute           = false;
 	result.dpp_ctrl           = 0;
+	result.dpp8               = false;
 	result.dpp_row_mask       = 0xf;
 	result.dpp_bank_mask      = 0xf;
 	result.explicit_sdwa_dst  = false;
@@ -122,6 +126,11 @@ IR::MemoryInfo MemoryInfoFromDecoded(const Decoder::Instruction& decoded) {
 	memory.image_r128    = decoded.image_r128;
 	memory.idxen         = decoded.idxen;
 	memory.offen         = decoded.offen;
+	// Vector loads use GLC/DLC to bypass L0/GL1; atomics use GLC only to return data.
+	const bool buffer_atomic = decoded.opcode >= Decoder::Opcode::BUFFER_ATOMIC_SWAP &&
+	                           decoded.opcode <= Decoder::Opcode::BUFFER_ATOMIC_FMAX;
+	memory.coherent = memory.kind == ResourceKind::Buffer && !buffer_atomic &&
+	                  (decoded.glc || decoded.dlc);
 	memory.resource      = ResourceIndexFromOperand(decoded.src1);
 	memory.sampler       = ResourceIndexFromOperand(decoded.src2);
 	if (memory.kind == ResourceKind::ScalarBuffer) {
@@ -234,6 +243,7 @@ Decoder::Operand MemorySourceAt(const Decoder::Instruction& decoded, uint32_t in
 				return index == 0u ? decoded.src1 : index == 1u ? decoded.src0 : decoded.src2;
 			case Decoder::Opcode::DS_WRITE_B8:
 			case Decoder::Opcode::DS_WRITE_B16:
+			case Decoder::Opcode::DS_WRITE_B8_D16_HI:
 			case Decoder::Opcode::DS_WRITE_B16_D16_HI:
 			case Decoder::Opcode::DS_WRITE_B32:
 			case Decoder::Opcode::DS_WRITE_B64:
@@ -395,7 +405,7 @@ IR::Value Translator::NarrowSubdword(IR::U32 value, uint32_t bits) {
 	                  : ir.Emit(IR::ValueOpcode::ConvertU16U32, {value});
 }

-bool Translator::S_LOAD(const Decoder::Instruction& inst, bool raw) {
+void Translator::S_LOAD(const Decoder::Instruction& inst, bool raw) {
 	const auto memory = MemoryInfoFromDecoded(inst);
 	const auto resource =
 	    raw ? GetScalarAddressResource(RawScalarLoadBase(inst.src0)) : GetBufferResource(memory);
@@ -418,10 +428,9 @@ bool Translator::S_LOAD(const Decoder::Instruction& inst, bool raw) {
 	for (uint32_t component = 0; component < memory.data_dwords; component++) {
 		WriteOperand(ScalarDestinationOperand(inst.dst, component), loaded[component]);
 	}
-	return true;
 }

-bool Translator::BUFFER_LOAD(const Decoder::Instruction& inst) {
+void Translator::BUFFER_LOAD(const Decoder::Instruction& inst) {
 	const auto      memory = MemoryInfoFromDecoded(inst);
 	IR::ValueOpcode opcode;
 	const auto      bits = memory.data_bits;
@@ -435,10 +444,15 @@ bool Translator::BUFFER_LOAD(const Decoder::Instruction& inst) {
 				case 2u: opcode = IR::ValueOpcode::LoadBufferU32x2; break;
 				case 3u: opcode = IR::ValueOpcode::LoadBufferU32x3; break;
 				case 4u: opcode = IR::ValueOpcode::LoadBufferU32x4; break;
-				default: return false;
+				default:
+					EXIT("opcode %s at pc 0x%08x has unsupported buffer load dword count %u",
+					     Decoder::InstructionToString(inst).c_str(), inst.pc,
+					     memory.data_dwords);
 			}
 			break;
-		default: return false;
+		default:
+			EXIT("opcode %s at pc 0x%08x has unsupported buffer load width %u",
+			     Decoder::InstructionToString(inst).c_str(), inst.pc, bits);
 	}
 	const auto resource = GetBufferResource(memory);
 	const auto address  = ReadBufferAddress(inst, 0);
@@ -455,10 +469,9 @@ bool Translator::BUFFER_LOAD(const Decoder::Instruction& inst) {
 			             ir.CompositeExtract(loaded, component));
 		}
 	}
-	return true;
 }

-bool Translator::BUFFER_STORE(const Decoder::Instruction& inst) {
+void Translator::BUFFER_STORE(const Decoder::Instruction& inst) {
 	const auto      memory   = MemoryInfoFromDecoded(inst);
 	const auto      resource = GetBufferResource(memory);
 	const auto      address  = ReadBufferAddress(inst, 1);
@@ -499,17 +512,21 @@ bool Translator::BUFFER_STORE(const Decoder::Instruction& inst) {
 					                  ReadU32(OffsetOperand(data_src, 2u)),
 					                  ReadU32(OffsetOperand(data_src, 3u))});
 					break;
-				default: return false;
+				default:
+					EXIT("opcode %s at pc 0x%08x has unsupported buffer store dword count %u",
+					     Decoder::InstructionToString(inst).c_str(), inst.pc,
+					     memory.data_dwords);
 			}
 			break;
-		default: return false;
+		default:
+			EXIT("opcode %s at pc 0x%08x has unsupported buffer store width %u",
+			     Decoder::InstructionToString(inst).c_str(), inst.pc, memory.data_bits);
 	}
 	ir.Emit(opcode, {resource, address.index, address.offset, address.soffset, value, ir.GetExec()},
 	        AddMemoryInfo(memory, inst.pc));
-	return true;
 }

-bool Translator::BUFFER_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
+void Translator::BUFFER_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
 	const auto memory   = MemoryInfoFromDecoded(inst);
 	const auto resource = GetBufferResource(memory);
 	const auto address  = ReadBufferAddress(inst, 1);
@@ -534,10 +551,9 @@ bool Translator::BUFFER_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode
 	if (inst.glc) {
 		WriteOperand(inst.dst, result);
 	}
-	return true;
 }

-bool Translator::IMAGE_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
+void Translator::IMAGE_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
 	const auto memory   = MemoryInfoFromDecoded(inst);
 	const auto resource = GetImageResource(memory);
 	const auto address  = MakeImageAddress(inst, MemorySourceAt(inst, 1));
@@ -547,10 +563,9 @@ bool Translator::IMAGE_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode
 	if (inst.glc) {
 		WriteOperand(inst.dst, result);
 	}
-	return true;
 }

-bool Translator::DS_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
+void Translator::DS_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode opcode,
                           bool returns_value) {
 	const auto memory  = MemoryInfoFromDecoded(inst);
 	const auto address = ReadU32(MemorySourceAt(inst, 1));
@@ -559,10 +574,9 @@ bool Translator::DS_ATOMIC(const Decoder::Instruction& inst, IR::ValueOpcode opc
 	if (returns_value) {
 		WriteOperand(inst.dst, result);
 	}
-	return true;
 }

-bool Translator::FLAT_LOAD(const Decoder::Instruction& inst) {
+void Translator::FLAT_LOAD(const Decoder::Instruction& inst) {
 	const auto      memory = MemoryInfoFromDecoded(inst);
 	IR::ValueOpcode opcode;
 	const auto      bits = memory.data_bits;
@@ -571,7 +585,9 @@ bool Translator::FLAT_LOAD(const Decoder::Instruction& inst) {
 		case 8u: opcode = IR::ValueOpcode::LoadAddressU8; break;
 		case 16u: opcode = IR::ValueOpcode::LoadAddressU16; break;
 		case 32u: opcode = IR::ValueOpcode::LoadAddressU32; break;
-		default: return false;
+		default:
+			EXIT("opcode %s at pc 0x%08x has unsupported flat load width %u",
+			     Decoder::InstructionToString(inst).c_str(), inst.pc, bits);
 	}
 	const auto address = ReadAddressOperands(inst, 0);
 	const auto active  = ir.GetExec();
@@ -586,10 +602,9 @@ bool Translator::FLAT_LOAD(const Decoder::Instruction& inst) {
 		WriteOperand(OffsetOperand(inst.dst, index),
 		             bits == 32u ? loaded : WidenSubdword(loaded, bits, sign));
 	}
-	return true;
 }

-bool Translator::FLAT_STORE(const Decoder::Instruction& inst) {
+void Translator::FLAT_STORE(const Decoder::Instruction& inst) {
 	const auto      memory  = MemoryInfoFromDecoded(inst);
 	const auto      data_op = MemorySourceAt(inst, 0);
 	const auto      address = ReadAddressOperands(inst, 1);
@@ -598,7 +613,9 @@ bool Translator::FLAT_STORE(const Decoder::Instruction& inst) {
 		case 8u: opcode = IR::ValueOpcode::StoreAddressU8; break;
 		case 16u: opcode = IR::ValueOpcode::StoreAddressU16; break;
 		case 32u: opcode = IR::ValueOpcode::StoreAddressU32; break;
-		default: return false;
+		default:
+			EXIT("opcode %s at pc 0x%08x has unsupported flat store width %u",
+			     Decoder::InstructionToString(inst).c_str(), inst.pc, memory.data_bits);
 	}
 	const auto count = memory.data_bits == 32u ? memory.data_dwords : 1u;
 	for (uint32_t index = 0; index < count; index++) {
@@ -613,20 +630,18 @@ bool Translator::FLAT_STORE(const Decoder::Instruction& inst) {
 		ir.Emit(opcode, {address.resource, address.low, address.high, value, ir.GetExec()},
 		        AddMemoryInfo(component, inst.pc));
 	}
-	return true;
 }

-bool Translator::IMAGE_GET_RESINFO(const Decoder::Instruction& inst) {
+void Translator::IMAGE_GET_RESINFO(const Decoder::Instruction& inst) {
 	const auto memory   = MemoryInfoFromDecoded(inst);
 	const auto resource = GetImageResource(memory);
 	const auto address  = MakeImageAddress(inst, MemorySourceAt(inst, 0));
 	const auto result   = ir.Emit(IR::ValueOpcode::ImageQueryDimensions, {resource, address},
 	                              AddMemoryInfo(memory, inst.pc));
 	WriteImageComponents(inst.dst, result, memory, 4u);
-	return true;
 }

-bool Translator::IMAGE_GET_LOD(const Decoder::Instruction& inst) {
+void Translator::IMAGE_GET_LOD(const Decoder::Instruction& inst) {
 	const auto memory   = MemoryInfoFromDecoded(inst);
 	const auto resource = GetImageResource(memory);
 	const auto sampler  = GetSamplerResource(memory);
@@ -634,30 +649,27 @@ bool Translator::IMAGE_GET_LOD(const Decoder::Instruction& inst) {
 	const auto result   = ir.Emit(IR::ValueOpcode::ImageQueryLod, {resource, sampler, address},
 	                              AddMemoryInfo(memory, inst.pc));
 	WriteImageComponents(inst.dst, result, memory, 2u);
-	return true;
 }

-bool Translator::IMAGE_LOAD(const Decoder::Instruction& inst) {
+void Translator::IMAGE_LOAD(const Decoder::Instruction& inst) {
 	const auto memory   = MemoryInfoFromDecoded(inst);
 	const auto resource = GetImageResource(memory);
 	const auto address  = MakeImageAddress(inst, MemorySourceAt(inst, 0));
 	const auto result   = ir.Emit(IR::ValueOpcode::ImageRead, {resource, address, ir.GetExec()},
 	                              AddMemoryInfo(memory, inst.pc));
 	WriteImageComponents(inst.dst, result, memory, 4u);
-	return true;
 }

-bool Translator::IMAGE_STORE(const Decoder::Instruction& inst) {
+void Translator::IMAGE_STORE(const Decoder::Instruction& inst) {
 	const auto memory   = MemoryInfoFromDecoded(inst);
 	const auto resource = GetImageResource(memory);
 	const auto address  = MakeImageAddress(inst, MemorySourceAt(inst, 1));
 	const auto data     = ConstructU32x4(MemorySourceAt(inst, 0), memory.data_dwords);
 	ir.Emit(IR::ValueOpcode::ImageWrite, {resource, address, data, ir.GetExec()},
 	        AddMemoryInfo(memory, inst.pc));
-	return true;
 }

-bool Translator::IMAGE_SAMPLE(const Decoder::Instruction& inst) {
+void Translator::IMAGE_SAMPLE(const Decoder::Instruction& inst) {
 	const auto memory   = MemoryInfoFromDecoded(inst);
 	const auto resource = GetImageResource(memory);
 	const auto sampler  = GetSamplerResource(memory);
@@ -674,10 +686,9 @@ bool Translator::IMAGE_SAMPLE(const Decoder::Instruction& inst) {
 	} else {
 		WriteImageComponents(inst.dst, result, memory, 4u);
 	}
-	return true;
 }

-bool Translator::IMAGE_GATHER(const Decoder::Instruction& inst) {
+void Translator::IMAGE_GATHER(const Decoder::Instruction& inst) {
 	const auto memory   = MemoryInfoFromDecoded(inst);
 	const auto resource = GetImageResource(memory);
 	const auto sampler  = GetSamplerResource(memory);
@@ -688,7 +699,6 @@ bool Translator::IMAGE_GATHER(const Decoder::Instruction& inst) {
 		WriteOperand(OffsetOperand(inst.dst, index),
 		             ir.Emit(IR::ValueOpcode::CompositeExtractU32x4, {result, IR::Value(index)}));
 	}
-	return true;
 }

 IR::Value Translator::LoadSharedU32(uint32_t width, IR::U32 address, const IR::MemoryInfo& memory,
@@ -744,7 +754,7 @@ void Translator::WriteSharedU32(uint32_t width, IR::U32 address,
 	}
 }

-bool Translator::DS_READ(const Decoder::Instruction& inst) {
+void Translator::DS_READ(const Decoder::Instruction& inst) {
 	const auto memory  = MemoryInfoFromDecoded(inst);
 	const auto address = ReadU32(MemorySourceAt(inst, 0));
 	if (memory.data_bits == 32u) {
@@ -753,16 +763,15 @@ bool Translator::DS_READ(const Decoder::Instruction& inst) {
 		for (uint32_t index = 0; index < width; index++) {
 			WriteOperand(OffsetOperand(inst.dst, index), ExtractSharedU32(loaded, width, index));
 		}
-		return true;
+		return;
 	}
 	const auto opcode =
 	    memory.data_bits == 8u ? IR::ValueOpcode::LoadSharedU8 : IR::ValueOpcode::LoadSharedU16;
 	const auto loaded = ir.Emit(opcode, {address, ir.GetExec()}, AddMemoryInfo(memory, inst.pc));
 	WriteOperand(inst.dst, WidenSubdword(loaded, memory.data_bits, memory.data_signed));
-	return true;
 }

-bool Translator::DS_READ2(const Decoder::Instruction& inst) {
+void Translator::DS_READ2(const Decoder::Instruction& inst) {
 	const auto memory       = MemoryInfoFromDecoded(inst);
 	const auto width        = memory.data_dwords / 2u;
 	const auto address      = ReadU32(MemorySourceAt(inst, 0));
@@ -781,10 +790,9 @@ bool Translator::DS_READ2(const Decoder::Instruction& inst) {
 		WriteOperand(OffsetOperand(inst.dst, width + index),
 		             ExtractSharedU32(second_value, width, index));
 	}
-	return true;
 }

-bool Translator::DS_WRITE(const Decoder::Instruction& inst) {
+void Translator::DS_WRITE(const Decoder::Instruction& inst) {
 	const auto               memory       = MemoryInfoFromDecoded(inst);
 	const auto               width        = memory.data_dwords;
 	const auto               data_operand = MemorySourceAt(inst, 0);
@@ -795,16 +803,15 @@ bool Translator::DS_WRITE(const Decoder::Instruction& inst) {
 	const auto address = ReadU32(MemorySourceAt(inst, 1));
 	if (memory.data_bits == 32u) {
 		WriteSharedU32(width, address, values, memory, inst.pc);
-		return true;
+		return;
 	}
 	const auto opcode =
 	    memory.data_bits == 8u ? IR::ValueOpcode::WriteSharedU8 : IR::ValueOpcode::WriteSharedU16;
 	ir.Emit(opcode, {address, NarrowSubdword(IR::U32(values[0]), memory.data_bits), ir.GetExec()},
 	        AddMemoryInfo(memory, inst.pc));
-	return true;
 }

-bool Translator::DS_WRITE2(const Decoder::Instruction& inst) {
+void Translator::DS_WRITE2(const Decoder::Instruction& inst) {
 	const auto               memory      = MemoryInfoFromDecoded(inst);
 	const auto               width       = memory.data_dwords / 2u;
 	const auto               address     = ReadU32(MemorySourceAt(inst, 1));
@@ -825,28 +832,25 @@ bool Translator::DS_WRITE2(const Decoder::Instruction& inst) {
 		second.offset = memory.secondary_offset;
 		WriteSharedU32(width, address, second_values, second, inst.pc);
 	}
-	return true;
 }

-bool Translator::DS_MINMAX_F32(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
+void Translator::DS_MINMAX_F32(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
 	const auto memory = MemoryInfoFromDecoded(inst);
 	ir.Emit(opcode,
 	        {ReadU32(MemorySourceAt(inst, 1)), ReadU32(MemorySourceAt(inst, 0)),
 	         ReadU32(MemorySourceAt(inst, 2)), ir.GetExec()},
 	        AddMemoryInfo(memory, inst.pc));
-	return true;
 }

-bool Translator::DS_APPEND_CONSUME(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
+void Translator::DS_APPEND_CONSUME(const Decoder::Instruction& inst, IR::ValueOpcode opcode) {
 	const auto memory = MemoryInfoFromDecoded(inst);
 	WriteOperand(inst.dst, ir.Emit(opcode,
 	                               {ReadU32(MemorySourceAt(inst, 0)), ir.GetExec(), ir.GetExecLo(),
 	                                ir.GetExecHi()},
 	                               AddMemoryInfo(memory, inst.pc)));
-	return true;
 }

-bool Translator::DS_ADDTID(const Decoder::Instruction& inst, bool write) {
+void Translator::DS_ADDTID(const Decoder::Instruction& inst, bool write) {
 	const auto memory = MemoryInfoFromDecoded(inst);
 	const auto base =
 	    ir.BitwiseAnd(ReadU32(MemorySourceAt(inst, write ? 1u : 0u)), IR::U32(IR::Value(0xffffu)));
@@ -860,24 +864,21 @@ bool Translator::DS_ADDTID(const Decoder::Instruction& inst, bool write) {
 		WriteOperand(inst.dst, ir.Emit(IR::ValueOpcode::LoadSharedU32, {address, ir.GetExec()},
 		                               AddMemoryInfo(memory, inst.pc)));
 	}
-	return true;
 }

-bool Translator::DS_SWIZZLE_B32(const Decoder::Instruction& inst) {
+void Translator::DS_SWIZZLE_B32(const Decoder::Instruction& inst) {
 	WriteOperand(inst.dst, ir.Emit(IR::ValueOpcode::SwizzleU32,
 	                               {ReadU32(MemorySourceAt(inst, 0)),
 	                                ReadU32(MemorySourceAt(inst, 1)), ir.GetExec()}));
-	return true;
 }

-bool Translator::DS_BPERMUTE_B32(const Decoder::Instruction& inst) {
+void Translator::DS_BPERMUTE_B32(const Decoder::Instruction& inst) {
 	const auto address = ir.IAdd(ReadU32(inst.src0), IR::U32(IR::Value(inst.offset)));
 	WriteOperand(inst.dst, ir.Emit(IR::ValueOpcode::BpermuteU32,
 	                               {ReadU32(inst.src1), address, ir.GetExec()}));
-	return true;
 }

-bool Translator::EmitMemory(const Decoder::Instruction& inst) {
+void Translator::EmitMemory(const Decoder::Instruction& inst) {
 	switch (inst.opcode) {
 		case Decoder::Opcode::S_LOAD_DWORD:
 		case Decoder::Opcode::S_LOAD_DWORDX2:
@@ -889,6 +890,17 @@ bool Translator::EmitMemory(const Decoder::Instruction& inst) {
 		case Decoder::Opcode::S_BUFFER_LOAD_DWORDX4:
 		case Decoder::Opcode::S_BUFFER_LOAD_DWORDX8:
 		case Decoder::Opcode::S_BUFFER_LOAD_DWORDX16: return S_LOAD(inst, false);
+		case Decoder::Opcode::S_MEMREALTIME: {
+			static std::atomic_flag warned = ATOMIC_FLAG_INIT;
+			if (!warned.test_and_set(std::memory_order_relaxed)) {
+				Log::WriteToConsoleAndLog(
+				    "Warning: S_MEMREALTIME uses placeholder UINT64_MAX; real-time clock not implemented.\n");
+			}
+			for (uint32_t component = 0; component < 2; component++) {
+				WriteOperand(ScalarDestinationOperand(inst.dst, component), IR::Value(UINT32_MAX));
+			}
+			return;
+		}

 		case Decoder::Opcode::BUFFER_LOAD_UBYTE:
 		case Decoder::Opcode::BUFFER_LOAD_SBYTE:
@@ -961,8 +973,12 @@ bool Translator::EmitMemory(const Decoder::Instruction& inst) {
 			return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicISub32, false);
 		case Decoder::Opcode::DS_SUB_RTN_U32:
 			return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicISub32, true);
+		case Decoder::Opcode::DS_INC_U32:
+			return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicInc32, false);
 		case Decoder::Opcode::DS_INC_RTN_U32:
 			return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicInc32, true);
+		case Decoder::Opcode::DS_DEC_U32:
+			return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicDec32, false);
 		case Decoder::Opcode::DS_DEC_RTN_U32:
 			return DS_ATOMIC(inst, IR::ValueOpcode::SharedAtomicDec32, true);
 		case Decoder::Opcode::DS_MIN_I32:
@@ -1075,12 +1091,13 @@ bool Translator::EmitMemory(const Decoder::Instruction& inst) {
 		case Decoder::Opcode::DS_WRITE2ST64_B64: return DS_WRITE2(inst);
 		case Decoder::Opcode::DS_WRITE_B8:
 		case Decoder::Opcode::DS_WRITE_B16:
+		case Decoder::Opcode::DS_WRITE_B8_D16_HI:
 		case Decoder::Opcode::DS_WRITE_B16_D16_HI:
 		case Decoder::Opcode::DS_WRITE_B32:
 		case Decoder::Opcode::DS_WRITE_B64:
 		case Decoder::Opcode::DS_WRITE_B96:
 		case Decoder::Opcode::DS_WRITE_B128: return DS_WRITE(inst);
-		default: return false;
+		default: return FailMissingTranslation(inst);
 	}
 }

@@ -2,106 +2,106 @@

 namespace Libs::Graphics::ShaderRecompiler::Frontend {

-bool Translator::EmitScalar(const Decoder::Instruction& inst) {
+void Translator::EmitScalar(const Decoder::Instruction& inst) {
 	using O = Decoder::Opcode;
 	switch (inst.opcode) {
 		case O::S_MOV_B32:
-		case O::S_MOVK_I32: MOV_B32(inst, false); return true;
-		case O::S_MOV_B64: S_MOV_B64(inst); return true;
-		case O::S_WQM_B32: S_WQM(inst, false); return true;
-		case O::S_WQM_B64: S_WQM(inst, true); return true;
-		case O::S_GETPC_B64: S_GETPC_B64(inst); return true;
-		case O::S_SETPC_B64: return true;
-		case O::S_SUBVECTOR_LOOP_BEGIN: S_SUBVECTOR_LOOP(inst, true); return true;
-		case O::S_SUBVECTOR_LOOP_END: S_SUBVECTOR_LOOP(inst, false); return true;
-		case O::S_CSELECT_B32: S_CSELECT_B32(inst); return true;
-		case O::S_CSELECT_B64: ScalarSelect64(inst, inst.src1); return true;
-		case O::S_CMOV_B64: ScalarSelect64(inst, inst.dst); return true;
-		case O::S_SETREG_B32: EmitControlNop(); return true;
-		case O::S_WAITCNT: EmitWaitcnt(); return true;
+		case O::S_MOVK_I32: MOV_B32(inst, false); return;
+		case O::S_MOV_B64: S_MOV_B64(inst); return;
+		case O::S_WQM_B32: S_WQM(inst, false); return;
+		case O::S_WQM_B64: S_WQM(inst, true); return;
+		case O::S_GETPC_B64: S_GETPC_B64(inst); return;
+		case O::S_SETPC_B64: return;
+		case O::S_SUBVECTOR_LOOP_BEGIN: S_SUBVECTOR_LOOP(inst, true); return;
+		case O::S_SUBVECTOR_LOOP_END: S_SUBVECTOR_LOOP(inst, false); return;
+		case O::S_CSELECT_B32: S_CSELECT_B32(inst); return;
+		case O::S_CSELECT_B64: ScalarSelect64(inst, inst.src1); return;
+		case O::S_CMOV_B64: ScalarSelect64(inst, inst.dst); return;
+		case O::S_SETREG_B32: EmitControlNop(); return;
+		case O::S_WAITCNT: EmitWaitcnt(); return;

 		case O::S_AND_SAVEEXEC_B32:
 			S_SAVEEXEC(inst, IR::ValueOpcode::LogicalAnd, false, false, false);
-			return true;
+			return;
 		case O::S_ANDN1_SAVEEXEC_B32:
 			S_SAVEEXEC(inst, IR::ValueOpcode::LogicalAnd, false, true, false);
-			return true;
+			return;
 		case O::S_ORN2_SAVEEXEC_B32:
 			S_SAVEEXEC(inst, IR::ValueOpcode::LogicalOr, true, false, false);
-			return true;
+			return;
 		case O::S_AND_SAVEEXEC_B64:
 			S_SAVEEXEC(inst, IR::ValueOpcode::LogicalAnd, false, false, true);
-			return true;
+			return;
 		case O::S_ANDN1_SAVEEXEC_B64:
 			S_SAVEEXEC(inst, IR::ValueOpcode::LogicalAnd, false, true, true);
-			return true;
+			return;
 		case O::S_ORN2_SAVEEXEC_B64:
 			S_SAVEEXEC(inst, IR::ValueOpcode::LogicalOr, true, false, true);
-			return true;
-		case O::S_ADD_U32: ADD_U32(inst, false, false); return true;
-		case O::S_ADDC_U32: ADD_U32(inst, false, true); return true;
-		case O::S_SUB_U32: SUB_U32(inst, false, false); return true;
-		case O::S_SUBB_U32: SUBB_U32(inst, false, false); return true;
-		case O::S_ABSDIFF_I32: S_ABSDIFF_I32(inst); return true;
-		case O::S_ADD_I32: S_ADD_SUB_I32(inst, false); return true;
-		case O::S_SUB_I32: S_ADD_SUB_I32(inst, true); return true;
-		case O::S_LSHL1_ADD_U32: S_LSHL_ADD_U32(inst, 1u); return true;
-		case O::S_LSHL2_ADD_U32: S_LSHL_ADD_U32(inst, 2u); return true;
-		case O::S_LSHL3_ADD_U32: S_LSHL_ADD_U32(inst, 3u); return true;
-		case O::S_LSHL4_ADD_U32: S_LSHL_ADD_U32(inst, 4u); return true;
+			return;
+		case O::S_ADD_U32: ADD_U32(inst, false, false); return;
+		case O::S_ADDC_U32: ADD_U32(inst, false, true); return;
+		case O::S_SUB_U32: SUB_U32(inst, false, false); return;
+		case O::S_SUBB_U32: SUBB_U32(inst, false, false); return;
+		case O::S_ABSDIFF_I32: S_ABSDIFF_I32(inst); return;
+		case O::S_ADD_I32: S_ADD_SUB_I32(inst, false); return;
+		case O::S_SUB_I32: S_ADD_SUB_I32(inst, true); return;
+		case O::S_LSHL1_ADD_U32: S_LSHL_ADD_U32(inst, 1u); return;
+		case O::S_LSHL2_ADD_U32: S_LSHL_ADD_U32(inst, 2u); return;
+		case O::S_LSHL3_ADD_U32: S_LSHL_ADD_U32(inst, 3u); return;
+		case O::S_LSHL4_ADD_U32: S_LSHL_ADD_U32(inst, 4u); return;
 		case O::S_MIN_I32:
 			ScalarMinMax32(inst, IR::ValueOpcode::SMin32, IR::ValueOpcode::SLessThan32);
-			return true;
+			return;
 		case O::S_MAX_I32:
 			ScalarMinMax32(inst, IR::ValueOpcode::SMax32, IR::ValueOpcode::SGreaterThan32);
-			return true;
+			return;
 		case O::S_MIN_U32:
 			ScalarMinMax32(inst, IR::ValueOpcode::UMin32, IR::ValueOpcode::ULessThan32);
-			return true;
+			return;
 		case O::S_MAX_U32:
 			ScalarMinMax32(inst, IR::ValueOpcode::UMax32, IR::ValueOpcode::UGreaterThan32);
-			return true;
+			return;

 		case O::S_CMP_EQ_U32:
 		case O::S_CMP_EQ_I32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::IEqual32, IR::Type::U32, true, false);
-			return true;
+			return;
 		case O::S_CMP_LG_U32:
 		case O::S_CMP_LG_I32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::INotEqual32, IR::Type::U32, true, false);
-			return true;
+			return;
 		case O::S_CMP_GT_U32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::UGreaterThan32, IR::Type::U32, true, false);
-			return true;
+			return;
 		case O::S_CMP_GE_U32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::UGreaterThanEqual32, IR::Type::U32, true,
 			                   false);
-			return true;
+			return;
 		case O::S_CMP_LT_U32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::ULessThan32, IR::Type::U32, true, false);
-			return true;
+			return;
 		case O::S_CMP_LE_U32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::ULessThanEqual32, IR::Type::U32, true, false);
-			return true;
+			return;
 		case O::S_CMP_GT_I32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::SGreaterThan32, IR::Type::U32, true, false);
-			return true;
+			return;
 		case O::S_CMP_GE_I32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::SGreaterThanEqual32, IR::Type::U32, true,
 			                   false);
-			return true;
+			return;
 		case O::S_CMP_LT_I32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::SLessThan32, IR::Type::U32, true, false);
-			return true;
+			return;
 		case O::S_CMP_LE_I32:
 			EmitIntegerCompare(inst, IR::ValueOpcode::SLessThanEqual32, IR::Type::U32, true, false);
-			return true;
+			return;
 		case O::S_CMP_EQ_U64:
 			EmitIntegerCompare(inst, IR::ValueOpcode::IEqual64, IR::Type::U64, true, false);
-			return true;
+			return;
 		case O::S_CMP_LG_U64:
 			EmitIntegerCompare(inst, IR::ValueOpcode::INotEqual64, IR::Type::U64, true, false);
-			return true;
+			return;

 		case O::S_AND_B64:
 			return S_U64_MASK(inst, IR::ValueOpcode::LogicalAnd, IR::ValueOpcode::BitwiseAnd32,
@@ -179,6 +179,7 @@ bool Translator::EmitScalar(const Decoder::Instruction& inst) {
 		case O::S_LSHR_B64:
 			return SimpleInteger(inst, IR::ValueOpcode::ShiftRightLogical64, IR::Type::U64, false,
 			                     false, true);
+		case O::S_ASHR_I64: return S_ASHR_I64(inst);

 		case O::S_ANDN2_B32:
 			return ComposedIntegerBinary(inst, IR::ValueOpcode::BitwiseAnd32, true, false, true);
@@ -207,6 +208,8 @@ bool Translator::EmitScalar(const Decoder::Instruction& inst) {
 		case O::S_BFE_U64: return S_BFE_U64(inst);
 		case O::S_BITCMP0_B32: return S_BITCMP_B32(inst, false);
 		case O::S_BITCMP1_B32: return S_BITCMP_B32(inst, true);
+		case O::S_BITCMP0_B64: return S_BITCMP_B64(inst, false);
+		case O::S_BITCMP1_B64: return S_BITCMP_B64(inst, true);
 		case O::S_PACK_LL_B32_B16: return PackB16(inst, false, false);
 		case O::S_PACK_LH_B32_B16: return PackB16(inst, false, true);
 		case O::S_PACK_HH_B32_B16: return PackB16(inst, true, true);
@@ -214,12 +217,12 @@ bool Translator::EmitScalar(const Decoder::Instruction& inst) {
 		case O::S_NOP:
 		case O::S_SLEEP:
 		case O::S_SETPRIO:
-		case O::S_TRAP: EmitControlNop(); return true;
-		case O::S_WAITCNT_DEPCTR: EmitWaitcnt(); return true;
-		case O::S_BARRIER: S_BARRIER(); return true;
-		case O::S_SENDMSG: S_SENDMSG(inst); return true;
-		case O::S_TTRACEDATA: S_TTRACEDATA(); return true;
-		case O::S_INST_PREFETCH: S_INST_PREFETCH(); return true;
+		case O::S_TRAP: EmitControlNop(); return;
+		case O::S_WAITCNT_DEPCTR: EmitWaitcnt(); return;
+		case O::S_BARRIER: S_BARRIER(); return;
+		case O::S_SENDMSG: S_SENDMSG(inst); return;
+		case O::S_TTRACEDATA: S_TTRACEDATA(); return;
+		case O::S_INST_PREFETCH: S_INST_PREFETCH(); return;
 		case O::S_BRANCH:
 		case O::S_CBRANCH_SCC0:
 		case O::S_CBRANCH_SCC1:
@@ -228,8 +231,8 @@ bool Translator::EmitScalar(const Decoder::Instruction& inst) {
 		case O::S_CBRANCH_EXECZ:
 		case O::S_CBRANCH_EXECNZ:
 		case O::S_CBRANCH_CDBGSYS:
-		case O::S_ENDPGM: return true;
-		default: return false;
+		case O::S_ENDPGM: return;
+		default: return FailMissingTranslation(inst);
 	}
 }

@@ -9,6 +9,17 @@

 namespace Libs::Graphics::ShaderRecompiler::Frontend {

+static IR::DppMoveFlags DppFlags(const Decoder::Operand& operand) {
+	return {
+	    .control        = operand.dpp_ctrl,
+	    .row_mask       = operand.dpp_row_mask,
+	    .bank_mask      = operand.dpp_bank_mask,
+	    .fetch_inactive = operand.dpp_fetch_inactive,
+	    .bound_control  = operand.dpp_bound_ctrl,
+	    .dpp8           = operand.dpp8,
+	};
+}
+
 const Decoder::Operand& Translator::SourceAt(const Decoder::Instruction& inst, uint32_t index) {
 	switch (index) {
 		case 0: return inst.src0;
@@ -30,6 +41,7 @@ Decoder::Operand Translator::DestinationOperand(const Decoder::Instruction& inst
 			continue;
 		}
 		destination.dpp                = true;
+		destination.dpp8               = source.dpp8;
 		destination.dpp_ctrl           = source.dpp_ctrl;
 		destination.dpp_row_mask       = source.dpp_row_mask;
 		destination.dpp_bank_mask      = source.dpp_bank_mask;
@@ -106,6 +118,7 @@ Decoder::Operand Translator::PlainOperand(const Decoder::Operand& operand) {
 	result.dpp_fetch_inactive = false;
 	result.dpp_bound_ctrl     = false;
 	result.dpp                = false;
+	result.dpp8               = false;
 	return result;
 }

@@ -167,14 +180,8 @@ IR::U32 Translator::ReadScalarCode(uint32_t code) {

 IR::U32 Translator::ApplyBitSourceModifiers(const Decoder::Operand& operand, IR::U32 value) {
 	if (operand.dpp) {
-		const IR::DppMoveFlags flags {
-		    .control        = static_cast<uint16_t>(operand.dpp_ctrl),
-		    .row_mask       = static_cast<uint8_t>(operand.dpp_row_mask),
-		    .bank_mask      = static_cast<uint8_t>(operand.dpp_bank_mask),
-		    .fetch_inactive = operand.dpp_fetch_inactive,
-		    .bound_control  = operand.dpp_bound_ctrl,
-		};
-		value = IR::U32(ir.Emit(IR::ValueOpcode::DppMoveU32, {value, ir.GetExec()}, flags));
+		value =
+		    IR::U32(ir.Emit(IR::ValueOpcode::DppMoveU32, {value, ir.GetExec()}, DppFlags(operand)));
 	}
 	if (operand.sdwa_sel != 6u) {
 		uint32_t offset = 0;
@@ -301,15 +308,8 @@ void Translator::WriteRawU32(const Decoder::Operand& operand, IR::U32 value) {
 			const auto reg = static_cast<IR::VectorReg>(operand.reg);
 			const auto old = ir.GetVectorReg(reg);
 			if (operand.dpp) {
-				const IR::DppMoveFlags flags {
-				    .control        = static_cast<uint16_t>(operand.dpp_ctrl),
-				    .row_mask       = static_cast<uint8_t>(operand.dpp_row_mask),
-				    .bank_mask      = static_cast<uint8_t>(operand.dpp_bank_mask),
-				    .fetch_inactive = operand.dpp_fetch_inactive,
-				    .bound_control  = operand.dpp_bound_ctrl,
-				};
-				value = IR::U32(
-				    ir.Emit(IR::ValueOpcode::DppUpdateU32, {value, old, ir.GetExec()}, flags));
+				value = IR::U32(ir.Emit(IR::ValueOpcode::DppUpdateU32, {value, old, ir.GetExec()},
+				                        DppFlags(operand)));
 			} else {
 				value = ir.Select(ir.GetExec(), value, old);
 			}
@@ -541,10 +541,7 @@ IR::U32 Translator::ReadU16LaneRaw(const Decoder::Operand& operand, bool high_la

 IR::U32 Translator::ReadU16LaneAsU32(const Decoder::Operand& operand, bool high_lane,
                                     bool sign_extend) {
-	auto value = ReadU16LaneRaw(operand, high_lane);
-	if (high_lane ? operand.negate_hi : operand.negate) {
-		value = ir.BitwiseAnd(ir.ISub(IR::U32(IR::Value(0u)), value), IR::U32(IR::Value(0xffffu)));
-	}
+	auto value = Read16LaneBits(operand, high_lane);
 	if (sign_extend || operand.sdwa_sext) {
 		value = IR::U32(
 		    ir.Emit(IR::ValueOpcode::BitFieldSExtract, {value, IR::Value(0u), IR::Value(16u)}));
@@ -556,12 +553,13 @@ IR::U32 Translator::ReadU16AsU32(const Decoder::Operand& operand, bool sign_exte
 	return ReadU16LaneAsU32(operand, false, sign_extend);
 }

-IR::U32 Translator::ReadF16LaneBits(const Decoder::Operand& operand, bool high_lane) {
+IR::U32 Translator::Read16LaneBits(const Decoder::Operand& operand, bool high_lane) {
 	auto value = ReadU16LaneRaw(operand, high_lane);
 	if (operand.absolute) {
 		value = ir.BitwiseAnd(value, IR::U32(IR::Value(0x7fffu)));
 	}
 	if (high_lane ? operand.negate_hi : operand.negate) {
+		// RDNA2 source NEG flips the sign bit, including packed integer operands.
 		value = ir.BitwiseXor(value, IR::U32(IR::Value(0x8000u)));
 	}
 	return value;
@@ -1065,8 +1063,8 @@ IR::Program TranslateProgram(const Decoder::Program& decoded, const CFG::Graph&
 			}
 		} else if (options.stage == ShaderType::Mesh) {
 			const auto& mesh = options.input_info.vertex->mesh;
-			EXIT_NOT_IMPLEMENTED(options.wave_size != 64u || mesh.primitives_per_group == 0u ||
-			                     mesh.vertices_per_group > 64u || total_threads > 15u * 64u);
+			EXIT_NOT_IMPLEMENTED(mesh.primitives_per_group == 0u || mesh.vertices_per_group > 64u ||
+			                     total_threads > 15u * options.wave_size);
 			const auto u32  = [](uint32_t value) { return IR::U32(IR::Value(value)); };
 			const auto draw = [&](uint32_t index) {
 				return IR::U32(
@@ -1091,12 +1089,15 @@ IR::Program TranslateProgram(const Decoder::Program& decoded, const CFG::Graph&
 			    entry_ir.IAdd(IR::U32(entry_ir.Emit(IR::ValueOpcode::UDiv32,
 			                                       {subtract_saturate(vertices, size), step})),
 			                  u32(1)));
-			const auto wave            = entry_ir.ShiftRightLogical(local, u32(6));
-			const auto wave_base       = entry_ir.BitwiseAnd(local, u32(~63u));
-			const auto vertex_count    = minimum(subtract_saturate(vertices, wave_base), u32(64));
-			const auto primitive_count = minimum(subtract_saturate(primitives, wave_base), u32(64));
+			const auto wave = entry_ir.ShiftRightLogical(local, u32(options.wave_size == 32u ? 5u : 6u));
+			const auto wave_base = entry_ir.BitwiseAnd(local, u32(~(options.wave_size - 1u)));
+			const auto vertex_count =
+			    minimum(subtract_saturate(vertices, wave_base), u32(options.wave_size));
+			const auto primitive_count =
+			    minimum(subtract_saturate(primitives, wave_base), u32(options.wave_size));
 			const auto wave_info = entry_ir.BitwiseOr(entry_ir.ShiftLeftLogical(wave, u32(24)),
-			                                          u32(((total_threads + 63u) / 64u) << 28u));
+			                                          u32(((total_threads + options.wave_size - 1u) /
+			                                               options.wave_size) << 28u));
 			entry_ir.SetScalarReg(
 			    static_cast<IR::ScalarReg>(3),
 			    entry_ir.BitwiseOr(wave_info, entry_ir.BitwiseOr(entry_ir.ShiftLeftLogical(
@@ -1227,6 +1228,12 @@ IR::Program TranslateProgram(const Decoder::Program& decoded, const CFG::Graph&
 				                      builtin(IR::StageInputKind::PackedAncillary));
 			}
 		} else if (options.stage == ShaderType::Vertex) {
+			// Vulkan owns primitive assembly; each vertex subgroup is one NGG wave.
+			// Keep its full lane extent: mbcnt(-1) uses lane ordinals, not active counts.
+			entry_ir.SetScalarReg(static_cast<IR::ScalarReg>(2),
+			                      IR::U32(IR::Value(options.wave_size << 12u)));
+			entry_ir.SetScalarReg(static_cast<IR::ScalarReg>(3),
+			                      IR::U32(IR::Value((1u << 28u) | options.wave_size)));
 			entry_ir.SetVectorReg(static_cast<IR::VectorReg>(5),
 			                      builtin(IR::StageInputKind::VertexIndex));
 			entry_ir.SetVectorReg(static_cast<IR::VectorReg>(8),