Step back + keeping fixes from unified download + shared mem refactor

.forgejo/workflows/translations.yml

@@ -1,9 +1,9 @@
 name: tx-pull
 
 on:
-  #  monday at 4pm
+  #  tuesday, saturday at 2pm
   schedule:
-    - cron: '0 16 * * 1'
+    - cron: '0 14 * * 2,6'
   workflow_dispatch:
 
 jobs:

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -3,7 +3,7 @@ description: File a bug report
 body:
   - type: markdown
     attributes:
-      value: Tech support does not belong here. You should only file an issue here if you think you have experienced an actual bug with Eden.
+      value: Tech support does not belong here. You should only file an issue here if you think you have experienced an actual bug with yuzu.
   - type: checkboxes
     attributes:
       label: Is there an existing issue for this?
@@ -43,7 +43,7 @@ body:
     id: log
     attributes:
       label: Log File
-      description: A log file will help our developers to better diagnose and fix the issue. Instructions can be found [here](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/user/HowToAccessLogs.md).
+      description: A log file will help our developers to better diagnose and fix the issue. Instructions can be found [here](https://yuzu-emu.org/help/reference/log-files).
     validations:
       required: true
   - type: textarea

.github/ISSUE_TEMPLATE/feature_request.yml

@@ -4,7 +4,7 @@ labels: "request"
 body:
   - type: markdown
     attributes:
-      value: Tech support does not belong here. You should only file an issue here if you are requesting a feature you believe would make Eden better.
+      value: Tech support does not belong here. You should only file an issue here if you are requesting a feature you believe would make yuzu better.
   - type: checkboxes
     attributes:
       label: Is there an existing issue for this?
@@ -23,6 +23,6 @@ body:
     id: why-feature
     attributes:
       label: Why would this feature be useful?
-      description: A brief description of why this feature would make Eden better.
+      description: A brief description of why this feature would make yuzu better.
     validations:
       required: true

CMakeLists.txt

@@ -28,23 +28,23 @@ if (NOT DEFINED ARCHITECTURE)
 endif()
 
 # Needed for FFmpeg w/ VAAPI and DRM
-if (OPENBSD)
+if (PLATFORM_OPENBSD)
     # OpenBSD 7.8 broke libcxx when upgrading, so we must define the PSTL backend manually
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -I${CMAKE_SYSROOT}/usr/X11R6/include -D_LIBCPP_PSTL_BACKEND_SERIAL=1")
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -I${CMAKE_SYSROOT}/usr/X11R6/include -D_LIBCPP_PSTL_BACKEND_SERIAL=1")
     set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -L${CMAKE_SYSROOT}/usr/X11R6/lib")
-elseif (NETBSD)
+elseif (PLATFORM_NETBSD)
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -I${CMAKE_SYSROOT}/usr/X11R7/include -I${CMAKE_SYSROOT}/usr/pkg/include/c++/v1")
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -I${CMAKE_SYSROOT}/usr/X11R7/include -I${CMAKE_SYSROOT}/usr/pkg/include/c++/v1")
     set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -L${CMAKE_SYSROOT}/usr/X11R7/lib")
 endif()
 
 # NetBSD: Fun for the whole family!
-if (NETBSD)
+if (PLATFORM_NETBSD)
     set(ENV{PKG_CONFIG_PATH} "${PKG_CONFIG_PATH}:${CMAKE_SYSROOT}/usr/pkg/lib/ffmpeg7/pkgconfig")
 endif()
 
-cmake_dependent_option(YUZU_STATIC_ROOM "Build a static room executable only (CI only)" OFF "LINUX" OFF)
+cmake_dependent_option(YUZU_STATIC_ROOM "Build a static room executable only (CI only)" OFF "PLATFORM_LINUX" OFF)
 if (YUZU_STATIC_ROOM)
     set(YUZU_ROOM ON)
     set(YUZU_ROOM_STANDALONE ON)
@@ -79,7 +79,6 @@ set(YUZU_QT_MIRROR "" CACHE STRING "What mirror to use for downloading the bundl
 cmake_dependent_option(YUZU_USE_BUNDLED_QT "Download bundled Qt binaries" "${MSVC}" "ENABLE_QT" OFF)
 
 option(ENABLE_DEBUG_TOOLS "Enable debugging tools (maxwell disassembler, SPIRV translator, etc)" OFF)
-option(ENABLE_WERROR "Enable -Werror diagnostics" ON)
 
 # non-linux bundled qt are static
 if (YUZU_USE_BUNDLED_QT AND (APPLE OR NOT UNIX))
@@ -158,19 +157,6 @@ if (CXX_CLANG_CL)
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-reserved-identifier>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-deprecated-declarations>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-cast-function-type-mismatch>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c99-extensions>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++17-compat>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++11-compat-reserved-user-defined-literal>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++11-compat-deprecated-writable-strings>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++11-compat-pedantic>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++11-compat>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++0x-compat>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++98-c++11-compat-binary-literal>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++98-compat-pedantic>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++98-compat>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c99-compat>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c98-compat>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c99-extensions>
         $<$<COMPILE_LANGUAGE:C,CXX>:/EHsc>)
     # REQUIRED CPU features IN Windows-amd64
     if (ARCHITECTURE_x86_64)
@@ -224,7 +210,7 @@ endif()
 
 # ffmpeg
 option(YUZU_USE_BUNDLED_FFMPEG "Download bundled FFmpeg" ${EXT_DEFAULT})
-cmake_dependent_option(YUZU_USE_EXTERNAL_FFMPEG "Build FFmpeg from external source" "${SOLARIS}" "NOT WIN32 AND NOT ANDROID" OFF)
+cmake_dependent_option(YUZU_USE_EXTERNAL_FFMPEG "Build FFmpeg from external source" "${PLATFORM_SUN}" "NOT WIN32 AND NOT ANDROID" OFF)
 
 # sirit
 set(BUNDLED_SIRIT_DEFAULT OFF)
@@ -235,7 +221,7 @@ endif()
 option(YUZU_USE_BUNDLED_SIRIT "Download bundled sirit" ${BUNDLED_SIRIT_DEFAULT})
 
 # FreeBSD 15+ has libusb, versions below should disable it
-cmake_dependent_option(ENABLE_LIBUSB "Enable the use of LibUSB" ON "WIN32 OR LINUX OR FREEBSD OR APPLE" OFF)
+cmake_dependent_option(ENABLE_LIBUSB "Enable the use of LibUSB" ON "WIN32 OR PLATFORM_LINUX OR PLATFORM_FREEBSD OR APPLE" OFF)
 
 cmake_dependent_option(ENABLE_OPENGL "Enable OpenGL" ON "NOT (WIN32 AND ARCHITECTURE_arm64) AND NOT APPLE" OFF)
 mark_as_advanced(FORCE ENABLE_OPENGL)
@@ -249,7 +235,7 @@ option(YUZU_TESTS "Compile tests" "${BUILD_TESTING}")
 
 # Install udev rules on Linux (mainly for gyros)
 # Only acts on joysticks and nothing else.
-cmake_dependent_option(YUZU_INSTALL_UDEV_RULES "Install udev rules for gyro access" OFF "LINUX" OFF)
+cmake_dependent_option(YUZU_INSTALL_UDEV_RULES "Install udev rules for gyro access" OFF "PLATFORM_LINUX" OFF)
 
 option(YUZU_DOWNLOAD_ANDROID_VVL "Download validation layer binary for android" ON)
 
@@ -262,7 +248,7 @@ cmake_dependent_option(YUZU_ROOM_STANDALONE "Enable standalone room executable"
 
 cmake_dependent_option(YUZU_CMD "Compile the eden-cli executable" ON "NOT ANDROID" OFF)
 
-cmake_dependent_option(YUZU_CRASH_DUMPS "Compile crash dump (Minidump) support" OFF "WIN32 OR LINUX" OFF)
+cmake_dependent_option(YUZU_CRASH_DUMPS "Compile crash dump (Minidump) support" OFF "WIN32 OR PLATFORM_LINUX" OFF)
 
 option(YUZU_DOWNLOAD_TIME_ZONE_DATA "Always download time zone binaries" ON)
 set(YUZU_TZDB_PATH "" CACHE STRING "Path to a pre-downloaded timezone database")
@@ -272,7 +258,7 @@ cmake_dependent_option(YUZU_USE_BUNDLED_MOLTENVK "Download bundled MoltenVK lib"
 option(YUZU_DISABLE_LLVM "Disable LLVM (useful for CI)" OFF)
 
 set(DEFAULT_YUZU_USE_BUNDLED_OPENSSL OFF)
-if (EXT_DEFAULT OR SOLARIS OR OPENBSD)
+if (EXT_DEFAULT OR PLATFORM_SUN OR PLATFORM_OPENBSD)
     set(DEFAULT_YUZU_USE_BUNDLED_OPENSSL ON)
 endif()
 
@@ -295,7 +281,7 @@ if(EXISTS ${PROJECT_SOURCE_DIR}/hooks/pre-commit AND NOT EXISTS ${PROJECT_SOURCE
     endif()
 endif()
 
-if (ARCHITECTURE_arm64 AND (ANDROID OR LINUX))
+if (ARCHITECTURE_arm64 AND (ANDROID OR PLATFORM_LINUX))
     set(HAS_NCE 1)
     add_compile_definitions(HAS_NCE=1)
 endif()
@@ -304,7 +290,7 @@ if (YUZU_ROOM)
     add_compile_definitions(YUZU_ROOM)
 endif()
 
-if (UNIX AND NOT (LINUX OR WIN32))
+if (UNIX AND NOT (PLATFORM_LINUX OR WIN32))
     if(CXX_APPLE OR CXX_CLANG)
         # libc++ has stop_token and jthread as experimental
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fexperimental-library")
@@ -418,7 +404,7 @@ if (Boost_ADDED)
 
     if (NOT MSVC OR CXX_CLANG)
         # boost sucks
-        if (SOLARIS)
+        if (PLATFORM_SUN)
             add_compile_options($<$<COMPILE_LANGUAGE:C,CXX>:-pthreads>)
         endif()
 
@@ -522,9 +508,9 @@ elseif (WIN32)
         # PSAPI is the Process Status API
         set(PLATFORM_LIBRARIES ${PLATFORM_LIBRARIES} psapi imm32 version crypt32 rpcrt4 gdi32 wldap32 mswsock)
     endif()
-elseif (MANAGARM)
+elseif (PLATFORM_MANAGARM)
     set(PLATFORM_LIBRARIES iconv intl)
-elseif (HAIKUOS)
+elseif (PLATFORM_HAIKU)
     # Haiku is so special :)
     set(PLATFORM_LIBRARIES bsd /boot/system/lib/libnetwork.so)
 elseif (CMAKE_SYSTEM_NAME MATCHES "^(Linux|kFreeBSD|GNU|SunOS)$")
@@ -618,7 +604,7 @@ if (ENABLE_QT)
 
     # Best practice is to ask for all components at once, so they are from the same version
     set(YUZU_QT_COMPONENTS Core Widgets Charts Concurrent Gui)
-    if (LINUX OR FREEBSD)
+    if (PLATFORM_LINUX OR PLATFORM_FREEBSD)
         list(APPEND YUZU_QT_COMPONENTS DBus)
         # yes Qt, we get it
         set(QT_NO_PRIVATE_MODULE_WARNING ON)

CMakeModules/find/FindOpus.cmake

@@ -13,7 +13,7 @@ find_package_handle_standard_args(Opus
     VERSION_VAR OPUS_VERSION
 )
 
-if (MSYS2)
+if (PLATFORM_MSYS)
     FixMsysPath(PkgConfig::OPUS)
 endif()
 

CMakeModules/find/Findenet.cmake

@@ -14,7 +14,7 @@ find_package_handle_standard_args(enet
     VERSION_VAR ENET_VERSION
 )
 
-if (MSYS2)
+if (PLATFORM_MSYS)
     FixMsysPath(PkgConfig::ENET)
 endif()
 

CMakeModules/find/Findlibusb.cmake

@@ -14,7 +14,7 @@ find_package_handle_standard_args(libusb
     VERSION_VAR LIBUSB_VERSION
 )
 
-if (MSYS2)
+if (PLATFORM_MSYS)
     FixMsysPath(PkgConfig::LIBUSB)
 endif()
 

CMakeModules/find/Findlz4.cmake

@@ -13,7 +13,7 @@ else()
     find_package(PkgConfig QUIET)
     pkg_search_module(LZ4 QUIET IMPORTED_TARGET liblz4)
 
-    if (MSYS2)
+    if (PLATFORM_MSYS)
         FixMsysPath(PkgConfig::LZ4)
     endif()
 

CONTRIBUTING.md

@@ -22,13 +22,9 @@ Eden is free, open-source, copyleft software, licensed under the terms of the [G
 
 - No LLM or AI usage, *period*, for patches, pull requests, issues, comments, debugging, brainstorming, etc.
   - For details on why, see the [detailed AI policy](docs/policies/AI.md).
-- Usage of any form of profanity or otherwise unsavory language is generally discouraged.
-  - This is primarily because it rarely helps to actually understand what's going on.
-  - Remember that your comments should be focused and actually address what's happening. With very few exceptions, expletives are actively detrimental at best.
 - New code must follow the same general style as the surrounding codebase. Exceptions may be granted in certain cases.
 - Maintainers reserve the right to change your patches and pull requests at will. We will try to avoid this.
-  - You should generally respect all decisions made by the [code owners](docs/CODEOWNERS) in your particular subsystem.
-  - However, if you feel they are overstepping or are incorrect, don't be afraid to stand your ground! Maintainers are not always correct.
+  - You should respect all decisions made by the [code owners](docs/CODEOWNERS) in your particular subsystem. If you feel they are overstepping or are incorrect, don't be afraid to stand your ground!
 - While we do *not* adhere to the terms of a formal code of conduct, you will generally be expected to respect other developers, contributors, and community members.
 - You **must** have basic knowledge of [Git](https://git-scm.com/learn). Knowing how to manage your branches and follow proper fork policies is a necessity.
 

docs/Options.md

@@ -46,8 +46,6 @@ These options control dependencies.
 - `YUZU_INSTALL_UDEV_RULES` (OFF) Install udev rules to enable hidraw access
   - Needed for gyroscopes
   - Only available on Linux
-- `ENABLE_DEBUG_TOOLS` (OFF) Enables debugging and development tools, see [tools](../tools/README.md).
-- `ENABLE_WERROR` (ON) Enables warnings as errors (-Werror).
 
 ### Flavors
 

docs/policies/Coding.md

@@ -34,7 +34,7 @@ Everyone has their own way of viewing good/bad C++ practices, my general outline
   - The reason is because the project has `-fno-rtti` disabled by default, due to the costs of dynamic polymorphism.
 - Always copy-on-value for objects with `sizeof(void *) >= sizeof(T) * 2`, i.e objects sized as 2 pointers or less, for bigger objects you can use ref/pointer as usual.
 - Try using move semantics instead of references, whenever possible.
-- Function parameters are cheap. Don't be afraid to use as many as needed (API usability permitting).
+- Remember function parameters are extremelly cheap as fuck, don't be afraid to place upto 8 parameters on a given function.
 - Don't save a reference in structures of a parent object, i.e:
 
     ```c++

externals/CMakeLists.txt

@@ -164,7 +164,7 @@ if (NOT ANDROID)
         AddJsonPackage(sdl3)
     else()
         message(STATUS "Using bundled SDL3")
-        if (FREEBSD)
+        if (PLATFORM_FREEBSD)
             set(BUILD_SHARED_LIBS ON)
         endif()
         AddJsonPackage(sdl3-ci)

externals/cmake-modules/DetectPlatform.cmake

@@ -1,46 +1,45 @@
-# SPDX-FileCopyrightText: Copyright 2026 crueter
+# SPDX-FileCopyrightText: Copyright 2025 crueter
 # SPDX-License-Identifier: LGPL-3.0-or-later
 
 ## DetectPlatform ##
 
-# This is a small helper that sets platform variables for various
+# This is a small helper that sets PLATFORM_<platform> variables for various
 # operating systems and distributions. Note that Apple, Windows, Android, etc.
 # are not covered, as CMake already does that for us.
 
 # It also sets CXX_<compiler> for the C++ compiler.
 
+# Furthermore, some platforms have really silly requirements/quirks, so this
+# also does a few of those.
+
 # This module contains contributions from the Eden Emulator Project,
 # notably from crueter and Lizzie.
 
 if (${CMAKE_SYSTEM_NAME} STREQUAL "SunOS")
-    set(SOLARIS ON)
+    set(PLATFORM_SUN ON)
 elseif (${CMAKE_SYSTEM_NAME} STREQUAL "OpenOrbis")
-    set(OPENORBIS ON)
+    set(PLATFORM_PS4 ON)
 elseif (${CMAKE_SYSTEM_NAME} STREQUAL "managarm")
-    set(MANAGARM ON)
+    set(PLATFORM_MANAGARM ON)
+elseif (${CMAKE_SYSTEM_NAME} STREQUAL "FreeBSD")
+    set(PLATFORM_FREEBSD ON)
+elseif (${CMAKE_SYSTEM_NAME} STREQUAL "OpenBSD")
+    set(PLATFORM_OPENBSD ON)
+elseif (${CMAKE_SYSTEM_NAME} STREQUAL "NetBSD")
+    set(PLATFORM_NETBSD ON)
+elseif (${CMAKE_SYSTEM_NAME} STREQUAL "DragonFly")
+    set(PLATFORM_DRAGONFLYBSD ON)
 elseif (${CMAKE_SYSTEM_NAME} STREQUAL "Haiku")
-    set(HAIKUOS ON)
-endif()
-
-# BSD
-if (DEFINED BSD)
-    if ("${BSD}" STREQUAL "DragonFlyBSD")
-        set(DRAGONFLYBSD ON)
-    elseif ("${BSD}" STREQUAL "FreeBSD")
-        set(FREEBSD ON)
-    elseif ("${BSD}" STREQUAL "OpenBSD")
-        set(OPENBSD ON)
-    elseif ("${BSD}" STREQUAL "NetBSD")
-        set(NETBSD ON)
-    endif()
+    set(PLATFORM_HAIKU ON)
+elseif (${CMAKE_SYSTEM_NAME} STREQUAL "Linux")
+    set(PLATFORM_LINUX ON)
 endif()
 
 # dumb heuristic to detect msys2
 if (CMAKE_COMMAND MATCHES "msys64")
-    set(MSYS2 ON)
+    set(PLATFORM_MSYS ON)
 endif()
 
-# compiler checks
 if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
     set(CXX_CLANG ON)
     if (MSVC)
@@ -48,6 +47,8 @@ if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
     endif()
 elseif (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
     set(CXX_GCC ON)
+elseif (CMAKE_CXX_COMPILER_ID STREQUAL "MSVC")
+    set(CXX_CL ON)
 elseif (CMAKE_CXX_COMPILER_ID STREQUAL "IntelLLVM")
     set(CXX_ICC ON)
 elseif (CMAKE_CXX_COMPILER_ID STREQUAL "AppleClang")
@@ -60,13 +61,37 @@ if(MINGW AND CXX_CLANG)
     set(CMAKE_SYSTEM_VERSION 10.0.0)
 endif()
 
+# NB: this does not account for SPARC
+if (PLATFORM_SUN)
+    # Terrific OpenIndiana pkg shenanigans
+    list(APPEND CMAKE_PREFIX_PATH
+        "${CMAKE_SYSROOT}/usr/lib/qt/6.6/lib/amd64/cmake")
+    list(APPEND CMAKE_MODULE_PATH
+        "${CMAKE_SYSROOT}/usr/lib/qt/6.6/lib/amd64/cmake")
+
+    # Amazing - absolutely incredible
+    list(APPEND CMAKE_PREFIX_PATH "${CMAKE_SYSROOT}/usr/lib/amd64/cmake")
+    list(APPEND CMAKE_MODULE_PATH "${CMAKE_SYSROOT}/usr/lib/amd64/cmake")
+
+    # For some mighty reason, doing a normal release build sometimes
+    # may not trigger the proper -O3 switch to materialize
+    if (CMAKE_BUILD_TYPE MATCHES "Release")
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -O3")
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3")
+    endif()
+    if (CMAKE_BUILD_TYPE MATCHES "RelWithDebInfo")
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -O2")
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O2")
+    endif()
+endif()
+
 # MSYS2 utilities
 
 # Sometimes, PkgConfig modules will incorrectly reference / when CMake
 # wants you to reference it as C:/msys64/. This function corrects that.
 # Example in a Find module:
 #[[
-    if (MSYS2)
+    if (PLATFORM_MSYS)
         FixMsysPath(PkgConfig::OPUS)
     endif()
 ]]
@@ -91,9 +116,41 @@ function(FixMsysPath target)
         INTERFACE_INCLUDE_DIRECTORIES ${include_dir})
 endfunction()
 
-# Saves linking time
+# MSYSTEM handling + program_path
+if (PLATFORM_MSYS)
+    # really, really dumb heuristic to detect what environment we are in
+    macro(system var)
+        if (CMAKE_COMMAND MATCHES ${var})
+            set(MSYSTEM ${var})
+        endif()
+    endmacro()
+
+    system(mingw64)
+    system(clang64)
+    system(clangarm64)
+    system(ucrt64)
+
+    if (NOT DEFINED MSYSTEM)
+        set(MSYSTEM msys2)
+    endif()
+
+    # We generally want to prioritize environment-specific binaries if possible
+    # some, like autoconf, are not present on environments besides msys2 though
+    set(CMAKE_PROGRAM_PATH C:/msys64/${MSYSTEM}/bin C:/msys64/usr/bin)
+    set(ENV{PKG_CONFIG_PATH} C:/msys64/${MSYSTEM}/lib/pkgconfig)
+endif()
+
+# This saves a truly ridiculous amount of time during linking
+# In my tests, without this, Eden takes 2 mins, with this, it takes 3-5 seconds
+# or on GitHub Actions, 10 minutes -> 3 seconds
 if (MINGW)
     set(MINGW_FLAGS "-Wl,--strip-all -Wl,--gc-sections")
     set(CMAKE_EXE_LINKER_FLAGS_RELEASE
         "${CMAKE_EXE_LINKER_FLAGS_RELEASE} ${MINGW_FLAGS}")
 endif()
+
+# awesome
+if (PLATFORM_FREEBSD OR PLATFORM_DRAGONFLYBSD)
+    set(CMAKE_EXE_LINKER_FLAGS
+        "${CMAKE_EXE_LINKER_FLAGS} -L${CMAKE_SYSROOT}/usr/local/lib")
+endif()

externals/ffmpeg/CMakeLists.txt

@@ -56,7 +56,7 @@ if (NOT YUZU_USE_BUNDLED_FFMPEG)
     endif()
 endif()
 
-if (OPENORBIS OR MANAGARM)
+if (PLATFORM_PS4 OR PLATFORM_MANAGARM)
     # Doesn't support VA-API, don't go thru the embarrassment of trying to enable it
     list(APPEND FFmpeg_HWACCEL_FLAGS --disable-vaapi)
 elseif (ANDROID)
@@ -75,7 +75,7 @@ elseif (UNIX AND NOT DEFINED FFmpeg_IS_CROSS_COMPILING AND NOT ANDROID)
     if(X11_FOUND)
         if (NOT APPLE)
             # In Solaris needs explicit linking for ffmpeg which links to /lib/amd64/libX11.so
-            if(SOLARIS)
+            if(PLATFORM_SUN)
                 list(APPEND FFmpeg_HWACCEL_LIBRARIES
                     X11
                     "${CMAKE_SYSROOT}/usr/lib/xorg/amd64/libdrm.so")
@@ -158,7 +158,7 @@ elseif (UNIX AND NOT DEFINED FFmpeg_IS_CROSS_COMPILING AND NOT ANDROID)
     endif()
 endif()
 
-if (OPENORBIS)
+if (PLATFORM_PS4)
     list(APPEND FFmpeg_CROSS_COMPILE_LIBS
         -lkernel
         -lSceUserService
@@ -172,7 +172,7 @@ if (OPENORBIS)
         --extra-cxxflags=${CMAKE_SYSROOT}/usr/include
         --extra-libs="${FFmpeg_CROSS_COMPILE_LIBS}"
     )
-elseif (MANAGARM)
+elseif (PLATFORM_MANAGARM)
     # Required for proper stuff
     list(APPEND FFmpeg_CROSS_COMPILE_FLAGS
         --disable-pthreads
@@ -299,7 +299,7 @@ else()
     set(FFmpeg_BUILD_LIBRARIES ${FFmpeg_LIBRARIES})
 
     # BSD make or Solaris make don't support ffmpeg make-j8
-    if (LINUX OR ANDROID OR APPLE OR WIN32 OR FREEBSD)
+    if (PLATFORM_LINUX OR ANDROID OR APPLE OR WIN32 OR PLATFORM_FREEBSD)
         set(FFmpeg_MAKE_ARGS -j${SYSTEM_THREADS})
     else()
         set(FFmpeg_MAKE_ARGS "")

externals/libusb/CMakeLists.txt

@@ -11,7 +11,7 @@ if (NOT libusb_ADDED)
 endif()
 
 # TODO: *BSD fails to compile--may need different configs/symbols
-if (MINGW OR LINUX OR APPLE)
+if (MINGW OR PLATFORM_LINUX OR APPLE)
     set(LIBUSB_FOUND ON CACHE BOOL "libusb is present" FORCE)
     set(LIBUSB_VERSION "1.0.24" CACHE STRING "libusb version string" FORCE)
 

externals/stb/stb_image.h

@@ -1177,7 +1177,7 @@ static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int re
    #endif
 
    #ifndef STBI_NO_TGA
-   // test tga last because it's a bad test!
+   // test tga last because it's a crappy test!
    if (stbi__tga_test(s))
       return stbi__tga_load(s,x,y,comp,req_comp, ri);
    #endif
@@ -7662,7 +7662,7 @@ static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp)
    if (stbi__hdr_info(s, x, y, comp))  return 1;
    #endif
 
-   // test tga last because it's a bad test!
+   // test tga last because it's a crappy test!
    #ifndef STBI_NO_TGA
    if (stbi__tga_info(s, x, y, comp))
        return 1;

src/CMakeLists.txt

@@ -155,24 +155,12 @@ else()
             $<$<COMPILE_LANGUAGE:CXX>:-fno-rtti>)
     endif()
 
-    if (ENABLE_WERROR)
-        add_compile_options(
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=all>
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=extra>
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=missing-declarations>
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=shadow>
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=unused>)
-    else()
-        add_compile_options(
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Wall>
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Wextra>
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Wmissing-declarations>
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Wshadow>
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Wunused>
-            # Some compilers could particularly misbehave :)
-            $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-error-all>)
-    endif()
     add_compile_options(
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=all>
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=extra>
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=missing-declarations>
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=shadow>
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=unused>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-attributes>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-invalid-offsetof>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-unused-parameter>
@@ -181,9 +169,9 @@ else()
     if (CXX_CLANG OR CXX_ICC OR CXX_APPLE) # Clang, AppleClang, or Intel C++
         if (NOT MSVC)
             add_compile_options(
-                $<$<COMPILE_LANGUAGE:C,CXX>:-Wshadow-uncaptured-local>
-                $<$<COMPILE_LANGUAGE:C,CXX>:-Wimplicit-fallthrough>
-                $<$<COMPILE_LANGUAGE:C,CXX>:-Wtype-limits>)
+                $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=shadow-uncaptured-local>
+                $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=implicit-fallthrough>
+                $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=type-limits>)
         endif()
         add_compile_options(
             $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-braced-scalar-init>
@@ -193,7 +181,7 @@ else()
 
     if (ARCHITECTURE_x86_64)
         add_compile_options($<$<COMPILE_LANGUAGE:C,CXX>:-mcx16>)
-        if (LINUX OR FREEBSD)
+        if (PLATFORM_LINUX OR PLATFORM_FREEBSD)
             add_compile_options($<$<COMPILE_LANGUAGE:C,CXX>:-mtls-dialect=gnu2>)
         endif()
     endif()

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/BooleanSetting.kt

@@ -27,6 +27,7 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
     RENDERER_ASYNCHRONOUS_GPU_EMULATION("use_asynchronous_gpu_emulation"),
     RENDERER_ASYNC_PRESENTATION("async_presentation"),
     RENDERER_ASYNCHRONOUS_SHADERS("use_asynchronous_shaders"),
+    RENDERER_UNIFIED_MEMORY("use_unified_memory"),
     RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"),
     ENABLE_BUFFER_HISTORY("enable_buffer_history"),
     USE_OPTIMIZED_VERTEX_BUFFERS("use_optimized_vertex_buffers"),

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt

@@ -586,7 +586,6 @@ abstract class SettingsItem(
                     IntSetting.FSR_SHARPENING_SLIDER,
                     titleId = R.string.fsr_sharpness,
                     descriptionId = R.string.fsr_sharpness_description,
-                    max = 200,
                     units = "%"
                 )
             )
@@ -686,6 +685,13 @@ abstract class SettingsItem(
                     descriptionId = R.string.renderer_asynchronous_shaders_description
                 )
             )
+            put(
+                SwitchSetting(
+                    BooleanSetting.RENDERER_UNIFIED_MEMORY,
+                    titleId = R.string.renderer_unified_memory,
+                    descriptionId = R.string.renderer_unified_memory_description
+                )
+            )
             put(
                 SingleChoiceSetting(
                     IntSetting.FAST_GPU_TIME,

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/ui/SettingsFragmentPresenter.kt

@@ -304,6 +304,7 @@ class SettingsFragmentPresenter(
             add(BooleanSetting.EMULATE_BGR565.key)
             add(BooleanSetting.RESCALE_HACK.key)
             add(BooleanSetting.RENDERER_ASYNCHRONOUS_SHADERS.key)
+            add(BooleanSetting.RENDERER_UNIFIED_MEMORY.key)
             add(IntSetting.ANDROID_PIPELINE_WORKERS.key)
             add(BooleanSetting.RENDERER_ASYNCHRONOUS_GPU_EMULATION.key)
             add(BooleanSetting.RENDERER_ASYNC_PRESENTATION.key)

src/android/app/src/main/java/org/yuzu/yuzu_emu/fragments/EmulationFragment.kt

@@ -1182,7 +1182,7 @@ class EmulationFragment : Fragment(), SurfaceHolder.Callback {
                     container,
                     IntSetting.FSR_SHARPENING_SLIDER,
                     minValue = 0,
-                    maxValue = 200,
+                    maxValue = 100,
                     units = "%"
                 )
             }

src/android/app/src/main/res/values/strings.xml

@@ -524,6 +524,8 @@
     <string name="rescale_hack_description">Enables a legacy handling for the rescale configuration pass for games by using a quick rescale path</string>
     <string name="renderer_asynchronous_shaders">Use asynchronous shaders</string>
     <string name="renderer_asynchronous_shaders_description">Compiles shaders asynchronously. This may reduce stutters but may also introduce glitches.</string>
+    <string name="renderer_unified_memory">Unified memory access</string>
+    <string name="renderer_unified_memory_description">Allows GPU write buffer readbacks directly into guest memory, skipping the CPU staging copy.</string>
     <string name="gpu_unswizzle_settings">GPU Unswizzle Settings</string>
     <string name="gpu_unswizzle_settings_description">Configure GPU-based texture unswizzling parameters or disable it entirely. Adjust these settings to balance performance and texture loading quality.</string>
     <string name="gpu_unswizzle_enable">Enable GPU Unswizzle</string>

src/audio_core/CMakeLists.txt

@@ -222,7 +222,7 @@ if (MSVC)
     )
 else()
     target_compile_options(audio_core PRIVATE
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-conversion>)
 endif()
 

src/common/CMakeLists.txt

@@ -220,7 +220,7 @@ endif()
 if(CXX_CLANG)
     target_compile_options(common PRIVATE
         $<$<COMPILE_LANGUAGE:C,CXX>:-fsized-deallocation>
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wunreachable-code-aggressive>)
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=unreachable-code-aggressive>)
     target_compile_definitions(
         common
         PRIVATE

src/common/heap_tracker.cpp

@@ -31,7 +31,9 @@ s64 GetMaxPermissibleResidentMapCount() {
 } // namespace
 
 HeapTracker::HeapTracker(Common::HostMemory& buffer)
-    : m_buffer(buffer), m_max_resident_map_count(GetMaxPermissibleResidentMapCount()) {}
+    : m_buffer(buffer),
+      m_has_hardware_buffer_backing(!buffer.BackingHardwareBuffers().empty()),
+      m_max_resident_map_count(GetMaxPermissibleResidentMapCount()) {}
 HeapTracker::~HeapTracker() = default;
 
 void HeapTracker::Map(size_t virtual_offset, size_t host_offset, size_t length,
@@ -85,7 +87,8 @@ void HeapTracker::Unmap(size_t virtual_offset, size_t size, bool is_separate_hea
 
             // If resident, erase from resident map.
             if (item->is_resident) {
-                ASSERT(--m_resident_map_count >= 0);
+                m_resident_map_count -= this->HostMapCount(item->paddr, item->size);
+                ASSERT(m_resident_map_count >= 0);
                 m_resident_mappings.erase(m_resident_mappings.iterator_to(*item));
             }
 
@@ -191,7 +194,7 @@ bool HeapTracker::DeferredMapSeparateHeap(size_t virtual_offset) {
 
         // This map is now resident.
         it->is_resident = true;
-        m_resident_map_count++;
+        m_resident_map_count += this->HostMapCount(it->paddr, it->size);
         m_resident_mappings.insert(*it);
     }
 
@@ -213,17 +216,17 @@ void HeapTracker::RebuildSeparateHeapAddressSpace() {
     // Despite being worse in theory, this has proven to be better in practice than more
     // regularly dumping a smaller amount, because it significantly reduces average case
     // lock contention.
-    std::size_t const desired_count = (std::min)(m_resident_map_count, m_max_resident_map_count) / 2;
-    std::size_t const evict_count = m_resident_map_count - desired_count;
+    s64 const desired_count = (std::min)(m_resident_map_count, m_max_resident_map_count) / 2;
     auto it = m_resident_mappings.begin();
 
-    for (size_t i = 0; i < evict_count && it != m_resident_mappings.end(); i++) {
+    while (m_resident_map_count > desired_count && it != m_resident_mappings.end()) {
         // Unmark and unmap.
         it->is_resident = false;
         m_buffer.Unmap(it->vaddr, it->size, false);
 
         // Advance.
-        ASSERT(--m_resident_map_count >= 0);
+        m_resident_map_count -= this->HostMapCount(it->paddr, it->size);
+        ASSERT(m_resident_map_count >= 0);
         it = m_resident_mappings.erase(it);
     }
 }
@@ -245,6 +248,7 @@ void HeapTracker::SplitHeapMapLocked(VAddr offset) {
     // Cache the original values.
     auto* const left = std::addressof(*it);
     const size_t orig_size = left->size;
+    const s64 orig_host_map_count = this->HostMapCount(left->paddr, orig_size);
 
     // Adjust the left map.
     const size_t left_size = offset - left->vaddr;
@@ -266,11 +270,20 @@ void HeapTracker::SplitHeapMapLocked(VAddr offset) {
 
     // If resident, also insert into resident map.
     if (right->is_resident) {
-        m_resident_map_count++;
+        m_resident_map_count += this->HostMapCount(left->paddr, left->size) +
+                                this->HostMapCount(right->paddr, right->size) -
+                                orig_host_map_count;
         m_resident_mappings.insert(*right);
     }
 }
 
+s64 HeapTracker::HostMapCount(PAddr paddr, size_t size) const {
+    if (!m_has_hardware_buffer_backing) {
+        return size != 0 ? 1 : 0;
+    }
+    return static_cast<s64>(m_buffer.BackingMapCount(paddr, size));
+}
+
 HeapTracker::AddrTree::iterator HeapTracker::GetNearestHeapMapLocked(VAddr offset) {
     const SeparateHeapMap key{
         .vaddr = offset,

src/common/heap_tracker.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -82,10 +85,13 @@ private:
 
     AddrTree::iterator GetNearestHeapMapLocked(VAddr offset);
 
+    s64 HostMapCount(PAddr paddr, size_t size) const;
+
     void RebuildSeparateHeapAddressSpace();
 
 private:
     Common::HostMemory& m_buffer;
+    const bool m_has_hardware_buffer_backing;
     const s64 m_max_resident_map_count;
 
     std::shared_mutex m_rebuild_lock{};

src/common/host_memory.cpp

@@ -51,14 +51,45 @@
 
 #endif // ^^^ POSIX ^^^
 
+#include <atomic>
 #include <mutex>
 #include <random>
+#include <vector>
 
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/free_region_manager.h"
 #include "common/host_memory.h"
 #include "common/logging.h"
+#include "common/memory_detect.h"
+#include "common/settings.h"
+
+#ifdef __ANDROID__
+#include <dlfcn.h>
+#include <android/hardware_buffer.h>
+
+namespace {
+
+struct NativeHandle {
+    int version;
+    int numFds;
+    int numInts;
+    int data[1];
+};
+
+using PFN_AHardwareBuffer_getNativeHandle = const NativeHandle* (*)(const AHardwareBuffer*);
+
+PFN_AHardwareBuffer_getNativeHandle ResolveGetNativeHandle() {
+    void* const lib = dlopen("libnativewindow.so", RTLD_NOW);
+    if (lib == nullptr) {
+        return nullptr;
+    }
+    return reinterpret_cast<PFN_AHardwareBuffer_getNativeHandle>(
+        dlsym(lib, "AHardwareBuffer_getNativeHandle"));
+}
+
+} // namespace
+#endif
 
 #if defined(__ANDROID__) && __ANDROID_API__ < 30
 #include <sys/syscall.h>
@@ -75,6 +106,12 @@ namespace Common {
 [[maybe_unused]] constexpr size_t PageAlignment = 0x1000;
 [[maybe_unused]] constexpr size_t HugePageSize = 0x200000;
 
+static std::atomic<u64> committed_backing_size{};
+
+u64 GetCommittedBackingSize() noexcept {
+    return committed_backing_size.load(std::memory_order_relaxed);
+}
+
 #ifdef _WIN32
 
 // Manually imported for MinGW compatibility
@@ -123,7 +160,7 @@ static void GetFuncAddress(Common::DynamicLibrary& dll, const char* name, T& pfn
 
 class HostMemory::Impl {
 public:
-    explicit Impl(size_t backing_size_, size_t virtual_size_)
+    explicit Impl(size_t backing_size_, size_t virtual_size_, size_t)
         : backing_size{backing_size_}
         , virtual_size{virtual_size_}
         , process{GetCurrentProcess()}
@@ -229,6 +266,10 @@ public:
         UNREACHABLE();
     }
 
+    bool IsBackingShared() const noexcept {
+        return true;
+    }
+
     const size_t backing_size; ///< Size of the backing memory in bytes
     const size_t virtual_size; ///< Size of the virtual address placeholder in bytes
 
@@ -501,9 +542,10 @@ static int shm_open_anon(int flags, mode_t mode) {
 
 class HostMemory::Impl {
 public:
-    explicit Impl(size_t backing_size_, size_t virtual_size_)
+    explicit Impl(size_t backing_size_, size_t virtual_size_, size_t preferred_offset_)
         : backing_size{backing_size_}
         , virtual_size{virtual_size_}
+        , preferred_offset{preferred_offset_}
     {}
 
     bool Init() {
@@ -543,10 +585,15 @@ public:
             LOG_WARNING(Common_Memory, "Using private mappings instead of shared ones");
             backing_base = static_cast<u8*>(mmap(nullptr, backing_size, PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0));
             if (fd > 0) {
-                fd = -1;
                 close(fd);
             }
+            fd = -1;
         } else {
+#ifdef __ANDROID__
+            if (InitAhbBacking()) {
+                return InitVirtual();
+            }
+#endif
             backing_base = static_cast<u8*>(mmap(nullptr, backing_size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0));
         }
         if (backing_base == MAP_FAILED) {
@@ -554,7 +601,10 @@ public:
             return false;
         }
 
-        // Virtual memory initialization
+        return InitVirtual();
+    }
+
+    bool InitVirtual() {
         virtual_base = virtual_map_base = static_cast<u8*>(ChooseVirtualBase(virtual_size));
         if (virtual_base == MAP_FAILED) {
             LOG_CRITICAL(HW_Memory, "mmap failed: {}", strerror(errno));
@@ -567,6 +617,242 @@ public:
         return true;
     }
 
+#ifdef __ANDROID__
+    static AHardwareBuffer_Desc MakeBlobDesc(size_t len) {
+        return AHardwareBuffer_Desc{
+            .width = static_cast<u32>(len),
+            .height = 1,
+            .layers = 1,
+            .format = AHARDWAREBUFFER_FORMAT_BLOB,
+            .usage = AHARDWAREBUFFER_USAGE_CPU_READ_OFTEN |
+                     AHARDWAREBUFFER_USAGE_CPU_WRITE_OFTEN |
+                     AHARDWAREBUFFER_USAGE_GPU_DATA_BUFFER,
+            .stride = 0,
+            .rfu0 = 0,
+            .rfu1 = 0,
+        };
+    }
+
+    static bool ProbeAhbBacking(PFN_AHardwareBuffer_getNativeHandle get_native_handle) {
+        const AHardwareBuffer_Desc desc = MakeBlobDesc(PageAlignment * 2);
+        AHardwareBuffer* buffer{};
+        if (AHardwareBuffer_allocate(&desc, &buffer) != 0 || buffer == nullptr) {
+            return false;
+        }
+        const NativeHandle* const handle = get_native_handle(buffer);
+        if (handle == nullptr || handle->numFds < 1) {
+            AHardwareBuffer_release(buffer);
+            return false;
+        }
+        const int probe_fd = handle->data[0];
+        bool ok = true;
+        const auto try_map = [&](int prot, off_t offset, const char* what) {
+            if (!ok) {
+                return;
+            }
+            void* const ptr = mmap(nullptr, PageAlignment, prot, MAP_SHARED, probe_fd, offset);
+            if (ptr == MAP_FAILED) {
+                ok = false;
+                return;
+            }
+            munmap(ptr, PageAlignment);
+        };
+        try_map(PROT_READ | PROT_WRITE, 0, "shared mappings");
+        try_map(PROT_READ | PROT_WRITE, static_cast<off_t>(PageAlignment), "mappings at an offset");
+#ifdef ARCHITECTURE_arm64
+        try_map(PROT_READ | PROT_EXEC, 0, "executable mappings");
+#endif
+        AHardwareBuffer_release(buffer);
+        return ok;
+    }
+
+    size_t ComputeAhbBudget(size_t window_size) const {
+        const u64 total_physical = Common::GetMemInfo().TotalPhysicalMemory;
+        constexpr u64 BaselineFootprint = 6ULL << 30;
+        if (total_physical <= BaselineFootprint) {
+            return 0;
+        }
+        const u64 max_map_count = Common::GetMaxMapCount();
+        constexpr u64 ReservedMaps = 24576;
+        if (max_map_count == 0 || max_map_count <= ReservedMaps) {
+            return 0;
+        }
+        u64 budget = (total_physical - BaselineFootprint) / 2;
+        constexpr u64 MapSlotsPerWindow = 4096;
+        const u64 affordable_windows = (max_map_count - ReservedMaps) / MapSlotsPerWindow;
+        budget = (std::min)(budget, affordable_windows * window_size);
+        const u64 available = Common::GetAvailablePhysicalMemory();
+        if (available != 0) {
+            budget = (std::min)(budget, available / 2);
+        }
+        budget = (std::min)(budget, static_cast<u64>(backing_size));
+        budget = Common::AlignDown(budget, window_size);
+        constexpr u64 MinimumBudget = 256ULL << 20;
+        if (budget < MinimumBudget) {
+            return 0;
+        }
+        return static_cast<size_t>(budget);
+    }
+
+    bool InitAhbBacking() {
+        if (!Settings::values.use_unified_memory.GetValue()) {
+            return false;
+        }
+        static const PFN_AHardwareBuffer_getNativeHandle get_native_handle =
+            ResolveGetNativeHandle();
+        if (get_native_handle == nullptr) {
+            return false;
+        }
+        constexpr size_t window_size = 256ULL << 20;
+        const size_t budget = ComputeAhbBudget(window_size);
+        if (budget == 0) {
+            return false;
+        }
+        if (!ProbeAhbBacking(get_native_handle)) {
+            return false;
+        }
+        const size_t aligned_backing = Common::AlignDown(backing_size, window_size);
+        const size_t max_windows = (std::min)(budget, aligned_backing) / window_size;
+
+        std::vector<AHardwareBuffer*> buffers;
+        std::vector<int> buffer_fds;
+        const auto cleanup = [&] {
+            for (AHardwareBuffer* buffer : buffers) {
+                AHardwareBuffer_release(buffer);
+            }
+            buffers.clear();
+            buffer_fds.clear();
+        };
+        for (size_t i = 0; i < max_windows; ++i) {
+            const AHardwareBuffer_Desc desc = MakeBlobDesc(window_size);
+            AHardwareBuffer* buffer{};
+            if (AHardwareBuffer_allocate(&desc, &buffer) != 0 || buffer == nullptr) {
+                break;
+            }
+            const NativeHandle* const handle = get_native_handle(buffer);
+            if (handle == nullptr || handle->numFds < 1) {
+                AHardwareBuffer_release(buffer);
+                break;
+            }
+            const int buffer_fd = handle->data[0];
+            const off_t buffer_len = lseek(buffer_fd, 0, SEEK_END);
+            if (buffer_len < static_cast<off_t>(window_size)) {
+                AHardwareBuffer_release(buffer);
+                break;
+            }
+            buffers.push_back(buffer);
+            buffer_fds.push_back(buffer_fd);
+        }
+        const size_t num_windows = buffers.size();
+        if (num_windows == 0) {
+            return false;
+        }
+        const size_t region_size = num_windows * window_size;
+        const size_t region_base = Common::AlignDown(
+            (std::min)(preferred_offset, aligned_backing - region_size), window_size);
+        u8* const base = static_cast<u8*>(mmap(nullptr, backing_size, PROT_NONE,
+                                               MAP_PRIVATE | MAP_ANONYMOUS | MAP_NORESERVE, -1, 0));
+        if (base == MAP_FAILED) {
+            cleanup();
+            return false;
+        }
+        const auto map_over_reservation = [&](size_t offset, size_t len, int map_fd,
+                                              off_t map_offset) {
+            if (len == 0) {
+                return true;
+            }
+            if (mmap(base + offset, len, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, map_fd,
+                     map_offset) == MAP_FAILED) {
+                munmap(base, backing_size);
+                cleanup();
+                return false;
+            }
+            return true;
+        };
+        if (!map_over_reservation(0, region_base, fd, 0)) {
+            return false;
+        }
+        for (size_t i = 0; i < num_windows; ++i) {
+            if (!map_over_reservation(region_base + i * window_size, window_size, buffer_fds[i],
+                                      0)) {
+                return false;
+            }
+        }
+        const size_t tail_offset = region_base + region_size;
+        if (!map_over_reservation(tail_offset, backing_size - tail_offset, fd,
+                                  static_cast<off_t>(tail_offset))) {
+            return false;
+        }
+        backing_base = base;
+        ahb_windows = std::move(buffers);
+        ahb_fds = std::move(buffer_fds);
+        ahb_window_size = window_size;
+        ahb_base = region_base;
+        ahb_bytes = region_size;
+        committed_backing_size.store(region_size, std::memory_order_relaxed);
+        return true;
+    }
+
+    void MapBackingRange(size_t virtual_offset, size_t host_offset, size_t length, int prot_flags) {
+        while (length > 0) {
+            int map_fd = fd;
+            off_t map_offset = static_cast<off_t>(host_offset);
+            size_t chunk = length;
+            if (host_offset < ahb_base) {
+                chunk = (std::min)(chunk, ahb_base - host_offset);
+            } else if (host_offset < ahb_base + ahb_bytes) {
+                const size_t relative = host_offset - ahb_base;
+                const size_t window = relative / ahb_window_size;
+                const size_t local = relative % ahb_window_size;
+                map_fd = ahb_fds[window];
+                map_offset = static_cast<off_t>(local);
+                chunk = (std::min)(chunk, ahb_window_size - local);
+            }
+            void* const ret = mmap(virtual_base + virtual_offset, chunk, prot_flags,
+                                   MAP_SHARED | MAP_FIXED, map_fd, map_offset);
+            ASSERT_MSG(ret != MAP_FAILED, "mmap: {}", strerror(errno));
+            virtual_offset += chunk;
+            host_offset += chunk;
+            length -= chunk;
+        }
+    }
+
+    size_t BackingMapCount(size_t host_offset, size_t length) const noexcept {
+        if (length == 0) {
+            return 0;
+        }
+        if (ahb_bytes == 0) {
+            return 1;
+        }
+        size_t count = 0;
+        while (length > 0) {
+            size_t chunk = length;
+            if (host_offset < ahb_base) {
+                chunk = (std::min)(chunk, ahb_base - host_offset);
+            } else if (host_offset < ahb_base + ahb_bytes) {
+                const size_t local = (host_offset - ahb_base) % ahb_window_size;
+                chunk = (std::min)(chunk, ahb_window_size - local);
+            }
+            host_offset += chunk;
+            length -= chunk;
+            ++count;
+        }
+        return count;
+    }
+
+    std::span<AHardwareBuffer* const> AhbWindows() const noexcept {
+        return ahb_windows;
+    }
+
+    size_t AhbWindowSize() const noexcept {
+        return ahb_bytes != 0 ? ahb_window_size : 0;
+    }
+
+    size_t AhbBase() const noexcept {
+        return ahb_base;
+    }
+#endif
+
     ~Impl() {
         Release();
     }
@@ -587,6 +873,12 @@ public:
 #ifdef ARCHITECTURE_arm64
         if (True(perms & MemoryPermission::Execute))
             prot_flags |= PROT_EXEC;
+#endif
+#ifdef __ANDROID__
+        if (ahb_bytes != 0) {
+            MapBackingRange(virtual_offset, host_offset, length, prot_flags);
+            return;
+        }
 #endif
         int flags = (fd >= 0 ? MAP_SHARED : MAP_PRIVATE) | MAP_FIXED;
         void* ret = mmap(virtual_base + virtual_offset, length, prot_flags, flags, fd, host_offset);
@@ -632,8 +924,18 @@ public:
         virtual_base = nullptr;
     }
 
+    bool IsBackingShared() const noexcept {
+#ifdef __ANDROID__
+        if (ahb_bytes != 0) {
+            return true;
+        }
+#endif
+        return fd >= 0;
+    }
+
     const size_t backing_size; ///< Size of the backing memory in bytes
     const size_t virtual_size; ///< Size of the virtual address placeholder in bytes
+    const size_t preferred_offset;
 
     u8* backing_base{reinterpret_cast<u8*>(MAP_FAILED)};
     u8* virtual_base{reinterpret_cast<u8*>(MAP_FAILED)};
@@ -656,6 +958,18 @@ private:
             int ret = close(fd);
             ASSERT_MSG(ret == 0, "close failed: {}", strerror(errno));
         }
+
+#ifdef __ANDROID__
+        for (AHardwareBuffer* buffer : ahb_windows) {
+            AHardwareBuffer_release(buffer);
+        }
+        ahb_windows.clear();
+        ahb_fds.clear();
+        if (ahb_bytes != 0) {
+            committed_backing_size.store(0, std::memory_order_relaxed);
+            ahb_bytes = 0;
+        }
+#endif
     }
 
     void AdjustMap(size_t* virtual_offset, size_t* length) {
@@ -681,11 +995,19 @@ private:
 
     int fd{-1}; // memfd file descriptor, -1 is the error value of memfd_create
     FreeRegionManager free_manager{};
+
+#ifdef __ANDROID__
+    std::vector<AHardwareBuffer*> ahb_windows;
+    std::vector<int> ahb_fds;
+    size_t ahb_window_size{};
+    size_t ahb_base{};
+    size_t ahb_bytes{};
+#endif
 };
 
 #endif // ^^^ POSIX ^^^
 
-HostMemory::HostMemory(size_t backing_size_, size_t virtual_size_)
+HostMemory::HostMemory(size_t backing_size_, size_t virtual_size_, size_t preferred_offset_)
     : backing_size(backing_size_)
     , virtual_size(virtual_size_)
 {
@@ -697,7 +1019,7 @@ HostMemory::HostMemory(size_t backing_size_, size_t virtual_size_)
 #else
     // Try to allocate a fastmem arena.
     // The implementation will fail with std::bad_alloc on errors.
-    impl = std::make_unique<HostMemory::Impl>(AlignUp(backing_size, PageAlignment), AlignUp(virtual_size, PageAlignment) + HugePageSize);
+    impl = std::make_unique<HostMemory::Impl>(AlignUp(backing_size, PageAlignment), AlignUp(virtual_size, PageAlignment) + HugePageSize, preferred_offset_);
     if (impl->Init()) {
         backing_base = impl->backing_base;
         virtual_base = impl->virtual_base;
@@ -767,6 +1089,46 @@ void HostMemory::ClearBackingRegion(size_t physical_offset, size_t length, u32 f
     std::memset(backing_base + physical_offset, fill_value, length);
 }
 
+std::span<AHardwareBuffer* const> HostMemory::BackingHardwareBuffers() const noexcept {
+#ifdef __ANDROID__
+    return impl ? impl->AhbWindows() : std::span<AHardwareBuffer* const>{};
+#else
+    return {};
+#endif
+}
+
+size_t HostMemory::BackingMapCount(size_t host_offset, size_t length) const noexcept {
+#ifdef __ANDROID__
+    return impl ? impl->BackingMapCount(host_offset, length) : (length != 0 ? 1 : 0);
+#else
+    return length != 0 ? 1 : 0;
+#endif
+}
+
+size_t HostMemory::BackingHardwareBufferWindowSize() const noexcept {
+#ifdef __ANDROID__
+    return impl ? impl->AhbWindowSize() : 0;
+#else
+    return 0;
+#endif
+}
+
+bool HostMemory::IsBackingShared() const noexcept {
+#if defined(__OPENORBIS__) || defined(__managarm__)
+    return false;
+#else
+    return impl && impl->IsBackingShared();
+#endif
+}
+
+size_t HostMemory::BackingHardwareBufferBase() const noexcept {
+#ifdef __ANDROID__
+    return impl ? impl->AhbBase() : 0;
+#else
+    return 0;
+#endif
+}
+
 void HostMemory::EnableDirectMappedAddress() {
 #if !(defined(__OPENORBIS__) || defined(__managarm__))
     if (impl) {

src/common/host_memory.h

@@ -8,12 +8,17 @@
 
 #include <memory>
 #include <optional>
+#include <span>
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/virtual_buffer.h"
 
+struct AHardwareBuffer;
+
 namespace Common {
 
+[[nodiscard]] u64 GetCommittedBackingSize() noexcept;
+
 enum class MemoryPermission : u32 {
     Read = 1 << 0,
     Write = 1 << 1,
@@ -28,7 +33,7 @@ DECLARE_ENUM_FLAG_OPERATORS(MemoryPermission)
  */
 class HostMemory {
 public:
-    explicit HostMemory(size_t backing_size_, size_t virtual_size_);
+    explicit HostMemory(size_t backing_size_, size_t virtual_size_, size_t preferred_offset_ = 0);
     ~HostMemory();
 
     /**
@@ -62,6 +67,20 @@ public:
         return backing_base;
     }
 
+    [[nodiscard]] size_t BackingSize() const noexcept {
+        return backing_size;
+    }
+
+    [[nodiscard]] size_t BackingMapCount(size_t host_offset, size_t length) const noexcept;
+
+    [[nodiscard]] std::span<AHardwareBuffer* const> BackingHardwareBuffers() const noexcept;
+
+    [[nodiscard]] size_t BackingHardwareBufferWindowSize() const noexcept;
+
+    [[nodiscard]] size_t BackingHardwareBufferBase() const noexcept;
+
+    [[nodiscard]] bool IsBackingShared() const noexcept;
+
     [[nodiscard]] u8* VirtualBasePointer() noexcept {
         return virtual_base;
     }

src/common/memory_detect.cpp

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -17,6 +20,10 @@
 #endif
 #endif
 
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+
 #include "common/memory_detect.h"
 
 namespace Common {
@@ -69,4 +76,55 @@ const MemoryInfo& GetMemInfo() {
     return mem_info;
 }
 
+u64 GetAvailablePhysicalMemory() {
+#ifdef _WIN32
+    MEMORYSTATUSEX memorystatus;
+    memorystatus.dwLength = sizeof(memorystatus);
+    if (GlobalMemoryStatusEx(&memorystatus)) {
+        return memorystatus.ullAvailPhys;
+    }
+    return 0;
+#elif defined(__linux__)
+    if (std::FILE* const file = std::fopen("/proc/meminfo", "re")) {
+        char line[256];
+        u64 available = 0;
+        while (std::fgets(line, sizeof(line), file) != nullptr) {
+            if (std::strncmp(line, "MemAvailable:", 13) == 0) {
+                available = std::strtoull(line + 13, nullptr, 10) * 1024ULL;
+                break;
+            }
+        }
+        std::fclose(file);
+        if (available != 0) {
+            return available;
+        }
+    }
+    struct sysinfo info;
+    if (sysinfo(&info) == 0) {
+        const u64 unit = info.mem_unit != 0 ? info.mem_unit : 1ULL;
+        return (static_cast<u64>(info.freeram) + static_cast<u64>(info.bufferram)) * unit;
+    }
+    return 0;
+#else
+    return 0;
+#endif
+}
+
+u64 GetMaxMapCount() {
+#ifdef __linux__
+    if (std::FILE* const file = std::fopen("/proc/sys/vm/max_map_count", "re")) {
+        char line[32];
+        u64 count = 0;
+        if (std::fgets(line, sizeof(line), file) != nullptr) {
+            count = std::strtoull(line, nullptr, 10);
+        }
+        std::fclose(file);
+        return count;
+    }
+    return 0;
+#else
+    return 0;
+#endif
+}
+
 } // namespace Common

src/common/memory_detect.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -18,4 +21,8 @@ struct MemoryInfo {
  */
 [[nodiscard]] const MemoryInfo& GetMemInfo();
 
+[[nodiscard]] u64 GetAvailablePhysicalMemory();
+
+[[nodiscard]] u64 GetMaxMapCount();
+
 } // namespace Common

src/common/settings.h

@@ -587,6 +587,9 @@ struct Values {
     SwitchableSetting<bool> use_asynchronous_shaders{linkage, false, "use_asynchronous_shaders",
                                                      Category::RendererHacks};
 
+    SwitchableSetting<bool> use_unified_memory{linkage, false, "use_unified_memory",
+                                               Category::RendererHacks};
+
     SwitchableSetting<GpuUnswizzleSize> gpu_unswizzle_texture_size{linkage,
                                                   GpuUnswizzleSize::Large,
                                                   "gpu_unswizzle_texture_size",
@@ -702,15 +705,7 @@ struct Values {
 
     // Controls
     InputSetting<std::array<PlayerInput, 10>> players;
-    Setting<bool> disable_wgi_xinput{
-        linkage, false, "disable_wgi_xinput", Category::Controls, Specialization::Default,
-// Only read/write disable_wgi_xinput on Windows platforms
-#ifdef _WIN32
-        true
-#else
-        false
-#endif
-    };
+
     Setting<bool> enable_raw_input{
         linkage, false, "enable_raw_input", Category::Controls, Specialization::Default,
 // Only read/write enable_raw_input on Windows platforms

src/core/CMakeLists.txt

@@ -1173,7 +1173,7 @@ add_library(core STATIC
 
 if (ENABLE_WIFI_SCAN)
     target_sources(core PRIVATE internal_network/wifi_scanner.cpp)
-    if (LINUX)
+    if (PLATFORM_LINUX)
         target_link_libraries(core PRIVATE iw)
     endif()
 else()
@@ -1199,7 +1199,7 @@ if (MSVC)
     )
 else()
     target_compile_options(core PRIVATE
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-conversion>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-cast-function-type>
         $<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>)

src/core/core.cpp

@@ -119,6 +119,7 @@ struct System::Impl {
 
         is_multicore = Settings::values.use_multi_core.GetValue();
         extended_memory_layout = Settings::values.memory_layout_mode.GetValue() != Settings::MemoryLayout::Memory_4Gb;
+        unified_memory = Settings::values.use_unified_memory.GetValue();
 
         core_timing.SetMulticore(is_multicore);
         core_timing.Initialize([&system]() { system.RegisterHostThread(); });
@@ -146,7 +147,8 @@ struct System::Impl {
             !device_memory.has_value() ||
             is_multicore != Settings::values.use_multi_core.GetValue() ||
             extended_memory_layout != (Settings::values.memory_layout_mode.GetValue() !=
-                                       Settings::MemoryLayout::Memory_4Gb);
+                                       Settings::MemoryLayout::Memory_4Gb) ||
+            unified_memory != Settings::values.use_unified_memory.GetValue();
 
         if (!must_reinitialize) {
             return;
@@ -157,6 +159,7 @@ struct System::Impl {
         is_multicore = Settings::values.use_multi_core.GetValue();
         extended_memory_layout =
             Settings::values.memory_layout_mode.GetValue() != Settings::MemoryLayout::Memory_4Gb;
+        unified_memory = Settings::values.use_unified_memory.GetValue();
 
         Initialize(system);
     }
@@ -503,6 +506,7 @@ struct System::Impl {
     std::atomic_bool is_powered_on{};
     bool is_multicore : 1 = false;
     bool extended_memory_layout : 1 = false;
+    bool unified_memory : 1 = false;
     bool exit_locked : 1 = false;
     bool exit_requested : 1 = false;
     bool nvdec_active : 1 = false;

src/core/device_memory.cpp

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -12,9 +15,18 @@ constexpr size_t VirtualReserveSize = 1ULL << 38;
 constexpr size_t VirtualReserveSize = 1ULL << 39;
 #endif
 
+namespace {
+size_t ApplicationPoolOffset() {
+    using Init = Kernel::Board::Nintendo::Nx::KSystemControl::Init;
+    const size_t dram_size = Init::GetIntendedMemorySize();
+    const size_t application_pool_size = Init::GetApplicationPoolSize();
+    return dram_size > application_pool_size ? dram_size - application_pool_size : 0;
+}
+}
+
 DeviceMemory::DeviceMemory()
     : buffer{Kernel::Board::Nintendo::Nx::KSystemControl::Init::GetIntendedMemorySize(),
-             VirtualReserveSize} {}
+             VirtualReserveSize, ApplicationPoolOffset()} {}
 
 DeviceMemory::~DeviceMemory() = default;
 

src/core/device_memory_manager.h

@@ -20,6 +20,8 @@
 #include "common/scratch_buffer.h"
 #include "common/virtual_buffer.h"
 
+struct AHardwareBuffer;
+
 namespace Core {
 
 constexpr size_t DEVICE_PAGEBITS = 12ULL;
@@ -95,6 +97,34 @@ public:
         ApplyOpOnPAddr(address, buffer, operation);
     }
 
+    u8* GetPhysicalBase() noexcept {
+        return reinterpret_cast<u8*>(physical_base);
+    }
+
+    const u8* GetPhysicalBase() const noexcept {
+        return reinterpret_cast<const u8*>(physical_base);
+    }
+
+    size_t GetPhysicalSize() const noexcept {
+        return physical_size;
+    }
+
+    std::span<AHardwareBuffer* const> GetBackingHardwareBuffers() const noexcept {
+        return ahb_windows;
+    }
+
+    size_t GetBackingHardwareBufferWindowSize() const noexcept {
+        return ahb_window_size;
+    }
+
+    size_t GetBackingHardwareBufferBase() const noexcept {
+        return ahb_base;
+    }
+
+    bool IsBackingShared() const noexcept {
+        return backing_is_shared;
+    }
+
     PAddr GetPhysicalRawAddressFromDAddr(DAddr address) const {
         PAddr subbits = PAddr(address & page_mask);
         auto paddr = tracked_entries[(address >> page_bits)].compressed_physical_ptr;
@@ -171,6 +201,11 @@ private:
     std::unique_ptr<DeviceMemoryManagerAllocator<Traits>> impl;
 
     const uintptr_t physical_base;
+    const size_t physical_size;
+    const std::span<AHardwareBuffer* const> ahb_windows;
+    const size_t ahb_window_size;
+    const size_t ahb_base;
+    const bool backing_is_shared;
     DeviceInterface* device_inter;
 
     struct TrackedEntry {

src/core/device_memory_manager.inc

@@ -171,6 +171,11 @@ struct DeviceMemoryManagerAllocator {
 template <typename Traits>
 DeviceMemoryManager<Traits>::DeviceMemoryManager(const DeviceMemory& device_memory_)
     : physical_base{uintptr_t(device_memory_.buffer.BackingBasePointer())}
+    , physical_size{device_memory_.buffer.BackingSize()}
+    , ahb_windows{device_memory_.buffer.BackingHardwareBuffers()}
+    , ahb_window_size{device_memory_.buffer.BackingHardwareBufferWindowSize()}
+    , ahb_base{device_memory_.buffer.BackingHardwareBufferBase()}
+    , backing_is_shared{device_memory_.buffer.IsBackingShared()}
     , device_inter{nullptr}
     , compressed_device_addr(1ULL << ((Settings::values.memory_layout_mode.GetValue() == Settings::MemoryLayout::Memory_4Gb ? physical_min_bits : physical_max_bits) - Memory::YUZU_PAGEBITS))
     , tracked_entries(device_as_size >> Memory::YUZU_PAGEBITS)

src/core/hle/kernel/svc/svc_synchronization.cpp

@@ -28,7 +28,7 @@ Result CloseHandle(Core::System& system, Handle handle) {
 
 /// Clears the signaled state of an event or process.
 Result ResetSignal(Core::System& system, Handle handle) {
-    LOG_TRACE(Kernel_SVC, "called handle {:#08x}", handle);
+    LOG_DEBUG(Kernel_SVC, "called handle {:#08x}", handle);
 
     // Get the current handle table.
     const auto& handle_table = GetCurrentProcess(system.Kernel()).GetHandleTable();

src/core/hle/service/pctl/parental_control_service.h

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator
@@ -84,7 +84,7 @@ private:
     RestrictionSettings restriction_settings{};
     std::array<char, 8> pin_code{};
     Capability capability{};
-    // TODO: this is raw
+    // TODO: this is RAW as fuck
     PlayTimerSettings raw_play_timer_settings{};
 
     KernelHelpers::ServiceContext service_context;

src/core/hle/service/set/system_settings_server.cpp

@@ -1020,7 +1020,7 @@ Result ISystemSettingsServer::GetBatteryLot(Out<BatteryLot> out_battery_lot) {
         c.lot_number[1] = 'H';
         c.lot_number[2] = 'A';
         c.lot_number[3] = 'C';
-        // TODO: what do the letters mean?
+        // TODO: I have no fucking idea what the letters mean
         c.lot_number[4] = 'H';
         c.lot_number[5] = 'Z';
         c.lot_number[6] = 'Z';

src/core/hle/service/vi/application_display_service.cpp

@@ -88,10 +88,25 @@ Result IApplicationDisplayService::GetIndirectDisplayTransactionService(
 }
 
 Result IApplicationDisplayService::OpenDisplay(Out<u64> out_display_id, DisplayName display_name) {
+    LOG_DEBUG(Service_VI, "called with display_name={}", display_name.data());
+
     // Ensure the display name is null-terminated
     display_name[display_name.size() - 1] = '\0';
 
-    LOG_DEBUG(Service_VI, "called with display_name={}", display_name.data());
+    // According to switchbrew, only "Default", "External", "Edid", "Internal" and "Null" are valid
+    const std::array<std::string_view, 5> valid_names = {
+        "Default", "External", "Edid", "Internal", "Null"
+    };
+
+    bool valid_name = false;
+    for (const auto& name : valid_names) {
+        if (name == display_name.data()) {
+            valid_name = true;
+            break;
+        }
+    }
+
+    R_UNLESS(valid_name, ResultOperationFailed);
 
     R_RETURN(m_container->OpenDisplay(out_display_id, display_name));
 }

src/core/internal_network/network.cpp

@@ -531,8 +531,7 @@ int TranslateTypeToNative(Type type) {
     NETWORK_PROTOCOL_TRANSLATE_ELEM(MPLS) \
     NETWORK_PROTOCOL_TRANSLATE_ELEM(PFSYNC)
 #elif defined(__linux__)
-// Other platforms may not support some niche protocols.
-// This is usually not an issue
+// Other platforms get fucked
 #define NETWORK_PROTOCOL_TRANSLATE_LIST \
     NETWORK_PROTOCOL_TRANSLATE_ELEM(IP) \
     /*NETWORK_PROTOCOL_TRANSLATE_ELEM(HOPOPTS)*/ \

src/dynarmic/CMakeLists.txt

@@ -25,7 +25,7 @@ endif()
 # Dynarmic project options
 option(DYNARMIC_ENABLE_CPU_FEATURE_DETECTION "Turning this off causes dynarmic to assume the host CPU doesn't support anything later than SSE3" ON)
 
-if (OPENBSD OR DRAGONFLY OR NETBSD)
+if (PLATFORM_OPENBSD OR PLATFORM_DRAGONFLY OR PLATFORM_NETBSD)
     set(REQUIRE_WX ON)
 else()
     set(REQUIRE_WX OFF)

src/dynarmic/src/dynarmic/CMakeLists.txt

@@ -363,7 +363,7 @@ elseif (APPLE)
             backend/exception_handler_macos_mig.c
         )
     endif()
-elseif (UNIX AND NOT HAIKUOS)
+elseif (UNIX AND NOT PLATFORM_HAIKU)
     # Haiku lacks <ucontext.h>
     if (CMAKE_SYSTEM_NAME STREQUAL "Linux")
         target_link_libraries(dynarmic PRIVATE rt)

src/dynarmic/src/dynarmic/backend/x64/a32_interface.cpp

@@ -234,7 +234,7 @@ private:
     BlockOfCode block_of_code;
     A32EmitX64 emitter;
     Optimization::PolyfillOptions polyfill_options;
-    // Keep it here in order to not mess with initializer lists
+    // Keep it here, you don't wanna mess with the fuckery that's initializer lists
     const A32::UserConfig conf;
     Jit* jit_interface;
 

src/dynarmic/src/dynarmic/backend/x64/emit_x64_vector_floating_point.cpp

@@ -2160,6 +2160,8 @@ void EmitFPVectorToFixed(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst) {
     ROUNDING_MODE_CASE(CASE, 0x3e) \
     ROUNDING_MODE_CASE(CASE, 0x3f)
 
+        // FUCK YOU MSVC, FUCKING DEPTH CANT EVEN HANDLE 8+16+32+64 DEPTH OF A ELSE STATMENT YOU FUCKING STUPID
+        // BURN MSVC BURN IT STUPID COMPILER CAN'T EVEN COMPILE THE MOST BASIC C++
         ROUNDING_MODE_SWITCH(ToNearest_TieEven)
         ROUNDING_MODE_SWITCH(TowardsPlusInfinity)
         ROUNDING_MODE_SWITCH(TowardsMinusInfinity)

src/dynarmic/src/dynarmic/ir/opt_passes.cpp

@@ -787,7 +787,7 @@ static void FoldCountLeadingZeros(IR::Inst& inst, bool is_32_bit) {
 /// Folds division operations based on the following:
 ///
 /// 1. x / 0 -> 0 (NOTE: This is an ARM-specific behavior defined in the architecture reference manual)
-/// 2a. 0x8000_0000 / 0xFFFF_FFFF -> 0x8000_0000 (NOTE: More ARM errata)
+/// 2a. 0x8000_0000 / 0xFFFF_FFFF -> 0x8000_0000 (NOTE: More ARM bullshit)
 /// 2b. 0x8000_0000_0000_0000 / 0xFFFF_FFFF_FFFF_FFFF -> 0x8000_0000_0000_0000
 /// 3. imm_x / imm_y -> result
 /// 4. x / 1 -> x

src/hid_core/CMakeLists.txt

@@ -151,7 +151,7 @@ if (MSVC)
     )
 else()
     target_compile_options(hid_core PRIVATE
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-conversion>
         $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-cast-function-type>
         $<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>)

src/input_common/CMakeLists.txt

@@ -44,7 +44,7 @@ if (MSVC)
         /we4800 # Implicit conversion from 'type' to bool. Possible information loss
     )
 else()
-    target_compile_options(input_common PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>)
+    target_compile_options(input_common PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>)
 endif()
 
 if (ANDROID)

src/input_common/drivers/android.cpp

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
@@ -160,8 +157,8 @@ Common::ParamPackage Android::BuildButtonParamPackageForButton(PadIdentifier ide
 
 bool Android::MatchVID(Common::UUID device, const std::vector<std::string>& vids) const {
     for (size_t i = 0; i < vids.size(); ++i) {
-        auto dev_str = device.RawString();
-        if (dev_str.find(vids[i]) != std::string::npos) {
+        auto fucker = device.RawString();
+        if (fucker.find(vids[i]) != std::string::npos) {
             return true;
         }
     }

src/input_common/drivers/sdl_driver.cpp

@@ -648,13 +648,6 @@ SDLDriver::SDLDriver(std::string input_engine_) : InputEngine(std::move(input_en
     // Disable raw input. When enabled this setting causes SDL to die when a web applet opens
     SDL_SetHint(SDL_HINT_JOYSTICK_RAWINPUT, Settings::values.enable_raw_input ? "1" : "0");
 
-#ifdef _WIN32
-    if (Settings::values.disable_wgi_xinput) {
-        SDL_SetHintWithPriority(SDL_HINT_JOYSTICK_RAWINPUT_CORRELATE_XINPUT, "0", SDL_HINT_OVERRIDE);
-        SDL_SetHintWithPriority(SDL_HINT_JOYSTICK_WGI, "0", SDL_HINT_OVERRIDE);
-    }
-#endif
-
     // SDL3 defaults Steam Controller Bluetooth HIDAPI support to off, which can disable gyro.
     SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_STEAM, "1");
     SDL_SetHint(SDL_HINT_GAMECONTROLLER_SENSOR_FUSION, "1");

src/network/CMakeLists.txt

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+# SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 # SPDX-License-Identifier: GPL-3.0-or-later
 
 # SPDX-FileCopyrightText: 2022 yuzu Emulator Project
@@ -28,6 +28,6 @@ if (ENABLE_WEB_SERVICE)
 endif()
 
 # Solaris uses /lib/amd64/libsocket.so and /lib/amd64/libnsl.so
-if (SOLARIS)
+if (PLATFORM_SUN)
     target_link_libraries(network PRIVATE socket nsl)
 endif()

src/qt_common/config/shared_translation.cpp

@@ -227,6 +227,8 @@ std::unique_ptr<TranslationMap> InitializeTranslations(QObject* parent) {
            tr("Preserves GPU-modified data by reading it back before uploading.\nSome games require this to render certain effects properly."));
     INSERT(Settings, use_asynchronous_shaders, tr("Enable asynchronous shader compilation"),
            tr("May reduce shader stutter."));
+    INSERT(Settings, use_unified_memory, tr("Enable unified memory access"),
+           tr("Lets the GPU write buffer readbacks directly into guest memory."));
     INSERT(Settings, gpu_clock, tr("GPU Clocks"),
            tr("Makes the game believe GPU work finishes faster than it does, so it stops lowering "
               "resolution and render distance to fit the Switch's clocks."));

src/shader_recompiler/CMakeLists.txt

@@ -253,7 +253,7 @@ if (MSVC)
     )
 else()
     target_compile_options(shader_recompiler PRIVATE
-        $<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>
+        $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>
         # Bracket depth determines maximum size of a fold expression in Clang since 9c9974c3ccb6.
         # And this in turns limits the size of a std::array.
         $<$<CXX_COMPILER_ID:Clang>:-fbracket-depth=1024>

src/shader_recompiler/backend/spirv/emit_spirv_atomic.cpp

@@ -13,6 +13,7 @@
 namespace Shader::Backend::SPIRV {
 namespace {
 Id SharedPointer(EmitContext& ctx, Id offset, u32 index_offset = 0) {
+    offset = ctx.BoundSharedOffset(offset, 4 + index_offset * 4);
     const Id shift_id{ctx.Const(2U)};
     Id index{ctx.OpShiftRightArithmetic(ctx.U32[1], offset, shift_id)};
     if (index_offset > 0) {
@@ -160,7 +161,8 @@ Id EmitSharedAtomicExchange32(EmitContext& ctx, Id offset, Id value) {
 Id EmitSharedAtomicExchange64(EmitContext& ctx, Id offset, Id value) {
     if (ctx.profile.support_shared_int64_atomics && ctx.uses_explicit_workgroup_layout) {
         const Id shift_id{ctx.Const(3U)};
-        const Id index{ctx.OpShiftRightArithmetic(ctx.U32[1], offset, shift_id)};
+        const Id index{
+            ctx.OpShiftRightArithmetic(ctx.U32[1], ctx.BoundSharedOffset(offset, 8), shift_id)};
         const Id pointer{
             ctx.OpAccessChain(ctx.shared_u64, ctx.shared_memory_u64, ctx.u32_zero_value, index)};
         const auto [scope, semantics]{AtomicArgs(ctx)};

src/shader_recompiler/backend/spirv/emit_spirv_shared_memory.cpp

@@ -31,6 +31,7 @@ std::pair<Id, Id> ExtractArgs(EmitContext& ctx, Id offset, u32 mask, u32 count)
 } // Anonymous namespace
 
 Id EmitLoadSharedU8(EmitContext& ctx, Id offset) {
+    offset = ctx.BoundSharedOffset(offset, 1);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{
             ctx.OpAccessChain(ctx.shared_u8, ctx.shared_memory_u8, ctx.u32_zero_value, offset)};
@@ -42,6 +43,7 @@ Id EmitLoadSharedU8(EmitContext& ctx, Id offset) {
 }
 
 Id EmitLoadSharedS8(EmitContext& ctx, Id offset) {
+    offset = ctx.BoundSharedOffset(offset, 1);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{
             ctx.OpAccessChain(ctx.shared_u8, ctx.shared_memory_u8, ctx.u32_zero_value, offset)};
@@ -53,6 +55,7 @@ Id EmitLoadSharedS8(EmitContext& ctx, Id offset) {
 }
 
 Id EmitLoadSharedU16(EmitContext& ctx, Id offset) {
+    offset = ctx.BoundSharedOffset(offset, 2);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{Pointer(ctx, ctx.shared_u16, ctx.shared_memory_u16, offset, 1)};
         return ctx.OpUConvert(ctx.U32[1], ctx.OpLoad(ctx.U16, pointer));
@@ -63,6 +66,7 @@ Id EmitLoadSharedU16(EmitContext& ctx, Id offset) {
 }
 
 Id EmitLoadSharedS16(EmitContext& ctx, Id offset) {
+    offset = ctx.BoundSharedOffset(offset, 2);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{Pointer(ctx, ctx.shared_u16, ctx.shared_memory_u16, offset, 1)};
         return ctx.OpSConvert(ctx.U32[1], ctx.OpLoad(ctx.U16, pointer));
@@ -73,6 +77,7 @@ Id EmitLoadSharedS16(EmitContext& ctx, Id offset) {
 }
 
 Id EmitLoadSharedU32(EmitContext& ctx, Id offset) {
+    offset = ctx.BoundSharedOffset(offset, 4);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{Pointer(ctx, ctx.shared_u32, ctx.shared_memory_u32, offset, 2)};
         return ctx.OpLoad(ctx.U32[1], pointer);
@@ -82,6 +87,7 @@ Id EmitLoadSharedU32(EmitContext& ctx, Id offset) {
 }
 
 Id EmitLoadSharedU64(EmitContext& ctx, Id offset) {
+    offset = ctx.BoundSharedOffset(offset, 8);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{Pointer(ctx, ctx.shared_u32x2, ctx.shared_memory_u32x2, offset, 3)};
         return ctx.OpLoad(ctx.U32[2], pointer);
@@ -97,6 +103,7 @@ Id EmitLoadSharedU64(EmitContext& ctx, Id offset) {
 }
 
 Id EmitLoadSharedU128(EmitContext& ctx, Id offset) {
+    offset = ctx.BoundSharedOffset(offset, 16);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{Pointer(ctx, ctx.shared_u32x4, ctx.shared_memory_u32x4, offset, 4)};
         return ctx.OpLoad(ctx.U32[4], pointer);
@@ -113,6 +120,7 @@ Id EmitLoadSharedU128(EmitContext& ctx, Id offset) {
 }
 
 void EmitWriteSharedU8(EmitContext& ctx, Id offset, Id value) {
+    offset = ctx.BoundSharedOffset(offset, 1);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{
             ctx.OpAccessChain(ctx.shared_u8, ctx.shared_memory_u8, ctx.u32_zero_value, offset)};
@@ -123,6 +131,7 @@ void EmitWriteSharedU8(EmitContext& ctx, Id offset, Id value) {
 }
 
 void EmitWriteSharedU16(EmitContext& ctx, Id offset, Id value) {
+    offset = ctx.BoundSharedOffset(offset, 2);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{Pointer(ctx, ctx.shared_u16, ctx.shared_memory_u16, offset, 1)};
         ctx.OpStore(pointer, ctx.OpUConvert(ctx.U16, value));
@@ -132,6 +141,7 @@ void EmitWriteSharedU16(EmitContext& ctx, Id offset, Id value) {
 }
 
 void EmitWriteSharedU32(EmitContext& ctx, Id offset, Id value) {
+    offset = ctx.BoundSharedOffset(offset, 4);
     Id pointer{};
     if (ctx.uses_explicit_workgroup_layout) {
         pointer = Pointer(ctx, ctx.shared_u32, ctx.shared_memory_u32, offset, 2);
@@ -144,6 +154,7 @@ void EmitWriteSharedU32(EmitContext& ctx, Id offset, Id value) {
 }
 
 void EmitWriteSharedU64(EmitContext& ctx, Id offset, Id value) {
+    offset = ctx.BoundSharedOffset(offset, 8);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{Pointer(ctx, ctx.shared_u32x2, ctx.shared_memory_u32x2, offset, 3)};
         ctx.OpStore(pointer, value);
@@ -159,6 +170,7 @@ void EmitWriteSharedU64(EmitContext& ctx, Id offset, Id value) {
 }
 
 void EmitWriteSharedU128(EmitContext& ctx, Id offset, Id value) {
+    offset = ctx.BoundSharedOffset(offset, 16);
     if (ctx.uses_explicit_workgroup_layout) {
         const Id pointer{Pointer(ctx, ctx.shared_u32x4, ctx.shared_memory_u32x4, offset, 4)};
         ctx.OpStore(pointer, value);

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -600,6 +600,16 @@ void EmitContext::DefineLocalMemory(const IR::Program& program) {
     }
 }
 
+Id EmitContext::BoundSharedOffset(Id offset, u32 access_bytes) {
+    if (shared_memory_declared_bytes == 0) {
+        return offset;
+    }
+    const u32 last_valid{shared_memory_declared_bytes > access_bytes
+                             ? shared_memory_declared_bytes - access_bytes
+                             : 0U};
+    return OpUMin(U32[1], offset, Const(last_valid));
+}
+
 void EmitContext::DefineSharedMemory(const IR::Program& program) {
     uses_explicit_workgroup_layout =
         profile.support_explicit_workgroup_layout &&
@@ -608,8 +618,15 @@ void EmitContext::DefineSharedMemory(const IR::Program& program) {
     if (program.shared_memory_size == 0) {
         return;
     }
+    const u32 device_limit{profile.max_shared_memory_size};
+    const u32 shared_memory_size{device_limit != 0 && program.shared_memory_size > device_limit
+                                     ? device_limit
+                                     : program.shared_memory_size};
+    if (shared_memory_size != program.shared_memory_size) {
+        shared_memory_declared_bytes = shared_memory_size;
+    }
     const auto make{[&](Id element_type, u32 element_size) {
-        const u32 num_elements{Common::DivCeil(program.shared_memory_size, element_size)};
+        const u32 num_elements{Common::DivCeil(shared_memory_size, element_size)};
         const Id array_type{TypeArray(element_type, Const(num_elements))};
         Decorate(array_type, spv::Decoration::ArrayStride, element_size);
 
@@ -644,7 +661,7 @@ void EmitContext::DefineSharedMemory(const IR::Program& program) {
         std::tie(shared_memory_u32x4, shared_u32x4, std::ignore) = make(U32[4], 16);
         return;
     }
-    const u32 num_elements{Common::DivCeil(program.shared_memory_size, 4U)};
+    const u32 num_elements{Common::DivCeil(shared_memory_size, 4U)};
     const Id type{TypeArray(U32[1], Const(num_elements))};
     shared_memory_u32_type = TypePointer(spv::StorageClass::Workgroup, type);
 

src/shader_recompiler/backend/spirv/spirv_emit_context.h

@@ -312,6 +312,8 @@ public:
     Id local_memory{};
 
     bool uses_explicit_workgroup_layout{};
+    u32 shared_memory_declared_bytes{};
+    [[nodiscard]] Id BoundSharedOffset(Id offset, u32 access_bytes);
     Id shared_memory_u8{};
     Id shared_memory_u16{};
     Id shared_memory_u32{};

src/shader_recompiler/frontend/maxwell/translate/impl/texture_fetch_swizzled.cpp

@@ -50,6 +50,12 @@ constexpr std::array RGBA_LUT{
     R | G | B | A, //
 };
 
+void CheckAlignment(IR::Reg reg, size_t alignment) {
+    if (!IR::IsAligned(reg, alignment)) {
+        throw NotImplementedException("Unaligned source register {}", reg);
+    }
+}
+
 template <typename... Args>
 IR::Value Composite(TranslatorVisitor& v, Args... regs) {
     return v.ir.CompositeConstruct(v.F(regs)...);
@@ -80,53 +86,67 @@ IR::Value Sample(TranslatorVisitor& v, u64 insn) {
         info.type.Assign(TextureType::Color2D);
         return v.ir.ImageSampleExplicitLod(handle, Composite(v, reg_a, reg_b), zero, {}, info);
     case 3: // 2D.LL
+        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color2D);
         return v.ir.ImageSampleExplicitLod(handle, Composite(v, reg_a, reg_a + 1), v.F(reg_b), {},
                                            info);
     case 4: // 2D.DC
+        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color2D);
         info.is_depth.Assign(1);
         return v.ir.ImageSampleDrefImplicitLod(handle, Composite(v, reg_a, reg_a + 1), v.F(reg_b),
                                                {}, {}, {}, info);
     case 5: // 2D.LL.DC
+        CheckAlignment(reg_a, 2);
+        CheckAlignment(reg_b, 2);
         info.type.Assign(TextureType::Color2D);
         info.is_depth.Assign(1);
         return v.ir.ImageSampleDrefExplicitLod(handle, Composite(v, reg_a, reg_a + 1),
                                                v.F(reg_b + 1), v.F(reg_b), {}, info);
     case 6: // 2D.LZ.DC
+        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color2D);
         info.is_depth.Assign(1);
         return v.ir.ImageSampleDrefExplicitLod(handle, Composite(v, reg_a, reg_a + 1), v.F(reg_b),
                                                zero, {}, info);
     case 7: // ARRAY_2D
+        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::ColorArray2D);
         return v.ir.ImageSampleImplicitLod(
             handle, v.ir.CompositeConstruct(v.F(reg_a + 1), v.F(reg_b), ReadArray(v, v.X(reg_a))),
             {}, {}, {}, info);
     case 8: // ARRAY_2D.LZ
+        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::ColorArray2D);
         return v.ir.ImageSampleExplicitLod(
             handle, v.ir.CompositeConstruct(v.F(reg_a + 1), v.F(reg_b), ReadArray(v, v.X(reg_a))),
             zero, {}, info);
     case 9: // ARRAY_2D.LZ.DC
+        CheckAlignment(reg_a, 2);
+        CheckAlignment(reg_b, 2);
         info.type.Assign(TextureType::ColorArray2D);
         info.is_depth.Assign(1);
         return v.ir.ImageSampleDrefExplicitLod(
             handle, v.ir.CompositeConstruct(v.F(reg_a + 1), v.F(reg_b), ReadArray(v, v.X(reg_a))),
             v.F(reg_b + 1), zero, {}, info);
     case 10: // 3D
+        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color3D);
         return v.ir.ImageSampleImplicitLod(handle, Composite(v, reg_a, reg_a + 1, reg_b), {}, {},
                                            {}, info);
     case 11: // 3D.LZ
+        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::Color3D);
         return v.ir.ImageSampleExplicitLod(handle, Composite(v, reg_a, reg_a + 1, reg_b), zero, {},
                                            info);
     case 12: // CUBE
+        CheckAlignment(reg_a, 2);
         info.type.Assign(TextureType::ColorCube);
         return v.ir.ImageSampleImplicitLod(handle, Composite(v, reg_a, reg_a + 1, reg_b), {}, {},
                                            {}, info);
     case 13: // CUBE.LL
+        CheckAlignment(reg_a, 2);
+        CheckAlignment(reg_b, 2);
         info.type.Assign(TextureType::ColorCube);
         return v.ir.ImageSampleExplicitLod(handle, Composite(v, reg_a, reg_a + 1, reg_b),
                                            v.F(reg_b + 1), {}, info);
@@ -167,10 +187,12 @@ IR::Reg RegStoreComponent32(u64 insn, unsigned index) {
     case 0:
         return texs.dest_reg_a;
     case 1:
+        CheckAlignment(texs.dest_reg_a, 2);
         return texs.dest_reg_a + 1;
     case 2:
         return texs.dest_reg_b;
     case 3:
+        CheckAlignment(texs.dest_reg_b, 2);
         return texs.dest_reg_b + 1;
     }
     throw LogicError("Invalid store index {}", index);

src/shader_recompiler/frontend/maxwell/translate/impl/texture_gather_swizzled.cpp

@@ -34,6 +34,12 @@ union Encoding {
     BitField<36, 13, u64> cbuf_offset;
 };
 
+void CheckAlignment(IR::Reg reg, size_t alignment) {
+    if (!IR::IsAligned(reg, alignment)) {
+        throw NotImplementedException("Unaligned source register {}", reg);
+    }
+}
+
 IR::Value MakeOffset(TranslatorVisitor& v, IR::Reg reg) {
     const IR::U32 value{v.X(reg)};
     return v.ir.CompositeConstruct(v.ir.BitFieldExtract(value, v.ir.Imm32(0), v.ir.Imm32(6), true),
@@ -54,15 +60,18 @@ IR::Value Sample(TranslatorVisitor& v, u64 insn) {
     info.is_depth.Assign(tld4s.dc != 0 ? 1 : 0);
     IR::Value coords;
     if (tld4s.aoffi != 0) {
+        CheckAlignment(reg_a, 2);
         coords = v.ir.CompositeConstruct(v.F(reg_a), v.F(reg_a + 1));
         IR::Value offset = MakeOffset(v, reg_b);
         if (tld4s.dc != 0) {
+            CheckAlignment(reg_b, 2);
             IR::F32 dref = v.F(reg_b + 1);
             return v.ir.ImageGatherDref(handle, coords, offset, {}, dref, info);
         }
         return v.ir.ImageGather(handle, coords, offset, {}, info);
     }
     if (tld4s.dc != 0) {
+        CheckAlignment(reg_a, 2);
         coords = v.ir.CompositeConstruct(v.F(reg_a), v.F(reg_a + 1));
         IR::F32 dref = v.F(reg_b);
         return v.ir.ImageGatherDref(handle, coords, {}, {}, dref, info);
@@ -77,10 +86,12 @@ IR::Reg RegStoreComponent32(u64 insn, size_t index) {
     case 0:
         return tlds4.dest_reg_a;
     case 1:
+        CheckAlignment(tlds4.dest_reg_a, 2);
         return tlds4.dest_reg_a + 1;
     case 2:
         return tlds4.dest_reg_b;
     case 3:
+        CheckAlignment(tlds4.dest_reg_b, 2);
         return tlds4.dest_reg_b + 1;
     }
     throw LogicError("Invalid store index {}", index);

src/shader_recompiler/frontend/maxwell/translate/impl/texture_load_swizzled.cpp

@@ -55,6 +55,12 @@ union Encoding {
     BitField<53, 4, u64> encoding;
 };
 
+void CheckAlignment(IR::Reg reg, size_t alignment) {
+    if (!IR::IsAligned(reg, alignment)) {
+        throw NotImplementedException("Unaligned source register {}", reg);
+    }
+}
+
 IR::Value MakeOffset(TranslatorVisitor& v, IR::Reg reg) {
     const IR::U32 value{v.X(reg)};
     return v.ir.CompositeConstruct(v.ir.BitFieldExtract(value, v.ir.Imm32(0), v.ir.Imm32(4), true),
@@ -86,31 +92,38 @@ IR::Value Sample(TranslatorVisitor& v, u64 insn) {
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_b));
         break;
     case 4:
+        CheckAlignment(reg_a, 2);
         texture_type = Shader::TextureType::Color2D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1));
         offsets = MakeOffset(v, reg_b);
         break;
     case 5:
+        CheckAlignment(reg_a, 2);
         texture_type = Shader::TextureType::Color2D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1));
         lod = v.X(reg_b);
         break;
     case 6:
+        CheckAlignment(reg_a, 2);
         texture_type = Shader::TextureType::Color2D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1));
         multisample = v.X(reg_b);
         break;
     case 7:
+        CheckAlignment(reg_a, 2);
         texture_type = Shader::TextureType::Color3D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1), v.X(reg_b));
         break;
     case 8: {
+        CheckAlignment(reg_b, 2);
         const IR::U32 array{v.ir.BitFieldExtract(v.X(reg_a), v.ir.Imm32(0), v.ir.Imm32(16))};
         texture_type = Shader::TextureType::ColorArray2D;
         coords = v.ir.CompositeConstruct(v.X(reg_b), v.X(reg_b + 1), array);
         break;
     }
     case 12:
+        CheckAlignment(reg_a, 2);
+        CheckAlignment(reg_b, 2);
         texture_type = Shader::TextureType::Color2D;
         coords = v.ir.CompositeConstruct(v.X(reg_a), v.X(reg_a + 1));
         lod = v.X(reg_b);
@@ -153,10 +166,12 @@ IR::Reg RegStoreComponent32(u64 insn, unsigned index) {
     case 0:
         return tlds.dest_reg_a;
     case 1:
+        CheckAlignment(tlds.dest_reg_a, 2);
         return tlds.dest_reg_a + 1;
     case 2:
         return tlds.dest_reg_b;
     case 3:
+        CheckAlignment(tlds.dest_reg_b, 2);
         return tlds.dest_reg_b + 1;
     }
     throw LogicError("Invalid store index {}", index);

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -169,36 +169,11 @@ std::map<IR::Attribute, IR::Attribute> GenerateLegacyToGenericMappings(
     return mapping;
 }
 
-struct PassthroughVertices {
-    u32 count;
-    u32 first;
-    u32 stride;
-};
-
-PassthroughVertices GetPassthroughVertices(InputTopology input_topology) {
-    switch (input_topology) {
-    case InputTopology::Points:
-        return {1, 0, 1};
-    case InputTopology::Lines:
-        return {2, 0, 1};
-    case InputTopology::LinesAdjacency:
-        return {2, 1, 1};
-    case InputTopology::Triangles:
-        return {3, 0, 1};
-    case InputTopology::TrianglesAdjacency:
-        return {3, 0, 2};
-    }
-    return {3, 0, 1};
-}
-
 void EmitGeometryPassthrough(IR::IREmitter& ir, const IR::Program& program,
                              const Shader::VaryingState& passthrough_mask,
                              bool passthrough_position,
-                             std::optional<IR::Attribute> passthrough_layer_attr,
-                             InputTopology input_topology) {
-    const PassthroughVertices vertices{GetPassthroughVertices(input_topology)};
-    for (u32 vertex = 0; vertex < vertices.count; vertex++) {
-        const u32 i = vertices.first + vertex * vertices.stride;
+                             std::optional<IR::Attribute> passthrough_layer_attr) {
+    for (u32 i = 0; i < program.output_vertices; i++) {
         // Assign generics from input
         for (u32 j = 0; j < 32; j++) {
             if (!passthrough_mask.Generic(j)) {
@@ -233,16 +208,25 @@ void EmitGeometryPassthrough(IR::IREmitter& ir, const IR::Program& program,
     ir.EndPrimitive(ir.Imm32(0));
 }
 
-void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInfo& host_info,
-                              InputTopology input_topology) {
+u32 GetOutputTopologyVertices(OutputTopology output_topology) {
+    switch (output_topology) {
+    case OutputTopology::PointList:
+        return 1;
+    case OutputTopology::LineStrip:
+        return 2;
+    default:
+        return 3;
+    }
+}
+
+void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInfo& host_info) {
     for (IR::Block* const block : program.blocks) {
         for (IR::Inst& inst : block->Instructions()) {
             if (inst.GetOpcode() == IR::Opcode::Epilogue) {
                 IR::IREmitter ir{*block, IR::Block::InstructionList::s_iterator_to(inst)};
                 EmitGeometryPassthrough(
                     ir, program, program.info.passthrough,
-                    program.info.passthrough.AnyComponent(IR::Attribute::PositionX), {},
-                    input_topology);
+                    program.info.passthrough.AnyComponent(IR::Attribute::PositionX), {});
             }
         }
     }
@@ -251,8 +235,7 @@ void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInf
 } // Anonymous namespace
 
 IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
-                             Environment& env, Flow::CFG& cfg, const HostTranslateInfo& host_info,
-                             InputTopology input_topology) {
+                             Environment& env, Flow::CFG& cfg, const HostTranslateInfo& host_info) {
     HostTranslateInfo normalized_host_info{host_info};
     normalized_host_info.ApplyDescriptorLimitPolicy();
 
@@ -281,9 +264,8 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
             }
 
             if (!normalized_host_info.support_geometry_shader_passthrough) {
-                program.output_vertices = GetPassthroughVertices(input_topology).count;
-                LowerGeometryPassthrough(program, normalized_host_info, input_topology);
-                program.is_geometry_passthrough = false;
+                program.output_vertices = GetOutputTopologyVertices(program.output_topology);
+                LowerGeometryPassthrough(program, normalized_host_info);
             }
         }
         break;
@@ -432,12 +414,11 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
                                         ObjectPool<IR::Block>& block_pool,
                                         const HostTranslateInfo& host_info,
                                         IR::Program& source_program,
-                                        Shader::OutputTopology output_topology,
-                                        InputTopology input_topology) {
+                                        Shader::OutputTopology output_topology) {
     IR::Program program;
     program.stage = Stage::Geometry;
     program.output_topology = output_topology;
-    program.output_vertices = GetPassthroughVertices(input_topology).count;
+    program.output_vertices = GetOutputTopologyVertices(output_topology);
 
     program.is_geometry_passthrough = false;
     program.info.loads.mask = source_program.info.stores.mask;
@@ -452,7 +433,7 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
 
     IR::IREmitter ir{*current_block};
     EmitGeometryPassthrough(ir, program, program.info.stores, true,
-                            source_program.info.emulated_layer, input_topology);
+                            source_program.info.emulated_layer);
 
     IR::Block* return_block{block_pool.Create(inst_pool)};
     IR::IREmitter{*return_block}.Epilogue();

src/shader_recompiler/frontend/maxwell/translate_program.h

@@ -18,8 +18,7 @@ namespace Shader::Maxwell {
 
 [[nodiscard]] IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool,
                                            ObjectPool<IR::Block>& block_pool, Environment& env,
-                                           Flow::CFG& cfg, const HostTranslateInfo& host_info,
-                                           InputTopology input_topology);
+                                           Flow::CFG& cfg, const HostTranslateInfo& host_info);
 
 [[nodiscard]] IR::Program MergeDualVertexPrograms(IR::Program& vertex_a, IR::Program& vertex_b,
                                                   Environment& env_vertex_b);
@@ -33,7 +32,6 @@ void ConvertLegacyToGeneric(IR::Program& program, const RuntimeInfo& runtime_inf
                                                       ObjectPool<IR::Block>& block_pool,
                                                       const HostTranslateInfo& host_info,
                                                       IR::Program& source_program,
-                                                      Shader::OutputTopology output_topology,
-                                                      InputTopology input_topology);
+                                                      Shader::OutputTopology output_topology);
 
 } // namespace Shader::Maxwell

src/shader_recompiler/profile.h

@@ -101,6 +101,9 @@ struct Profile {
 
     u32 gl_max_compute_smem_size{};
 
+    /// Largest workgroup shared memory allocation the device accepts, 0 when unconstrained
+    u32 max_shared_memory_size{};
+
     /// Maxwell and earlier nVidia architectures have broken robust support
     bool has_broken_robust{};
 

src/video_core/CMakeLists.txt

@@ -373,7 +373,7 @@ else()
             $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-shadow>
             $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-unused-local-typedef>)
     else()
-        target_compile_options(video_core PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>)
+        target_compile_options(video_core PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>)
     endif()
 
     target_compile_options(video_core PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-conversion>)

src/video_core/buffer_cache/buffer_cache.h

@@ -7,7 +7,6 @@
 #pragma once
 
 #include <algorithm>
-#include <bit>
 #include <memory>
 #include <numeric>
 
@@ -572,7 +571,11 @@ void BufferCache<P>::AccumulateFlushes() {
 
 template <class P>
 bool BufferCache<P>::ShouldWaitAsyncFlushes() const noexcept {
-    return (!async_buffers.empty() && async_buffers.front().has_value());
+    if (async_buffers.empty()) {
+        return false;
+    }
+    return async_buffers.front().has_value() ||
+           !pending_downloads.front().unified_copies.empty();
 }
 
 template <class P>
@@ -580,6 +583,7 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
     AccumulateFlushes();
 
     if (committed_gpu_modified_ranges.empty()) {
+        pending_downloads.emplace_back();
         async_buffers.emplace_back(std::optional<Async_Buffer>{});
         return;
     }
@@ -639,27 +643,84 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
     }
     committed_gpu_modified_ranges.clear();
     if (downloads.empty()) {
+        pending_downloads.emplace_back();
         async_buffers.emplace_back(std::optional<Async_Buffer>{});
         return;
     }
-    auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes, true);
-    boost::container::small_vector<BufferCopy, 4> normalized_copies;
-    runtime.PreCopyBarrier();
+
+    struct QueuedUnifiedCopy {
+        u64 window;
+        BufferId buffer_id;
+        boost::container::small_vector<BufferCopy, 16> copies;
+    };
+
+    AsyncDownloadBatch batch;
+    boost::container::small_vector<std::pair<BufferCopy, BufferId>, 16> staging_downloads;
+    boost::container::small_vector<QueuedUnifiedCopy, 4> unified_copy_queue;
+    boost::container::small_vector<u64, 4> window_ids;
+    UnifiedWindowGroups groups;
+    u64 staging_size_bytes = 0;
     for (auto& [copy, buffer_id] : downloads) {
-        copy.dst_offset += download_staging.offset;
-        const std::array copies{copy};
-        BufferCopy second_copy{copy};
         Buffer& buffer = slot_buffers[buffer_id];
-        second_copy.src_offset = static_cast<size_t>(buffer.CpuAddr()) + copy.src_offset;
-        const DAddr orig_device_addr = static_cast<DAddr>(second_copy.src_offset);
+        const DAddr orig_device_addr = buffer.CpuAddr() + copy.src_offset;
+        bool unified = false;
+        if constexpr (USE_UNIFIED_MEMORY) {
+            if (runtime.HasUnifiedMemory()) {
+                window_ids.clear();
+                groups.clear();
+                unified = ResolveUnifiedWindows(orig_device_addr, copy.src_offset, copy.size,
+                                                window_ids, groups);
+            }
+        }
+        BufferCopy record{copy};
+        record.src_offset = static_cast<size_t>(orig_device_addr);
+        if (unified) {
+            async_downloads.Add(orig_device_addr, copy.size);
+            buffer.MarkUsage(copy.src_offset, copy.size);
+            for (size_t i = 0; i < window_ids.size(); ++i) {
+                unified_copy_queue.push_back(
+                    QueuedUnifiedCopy{window_ids[i], buffer_id, std::move(groups[i])});
+            }
+            batch.unified_copies.push_back(record);
+            continue;
+        }
+        copy.dst_offset = staging_size_bytes;
+        constexpr u64 align = 64ULL;
+        staging_size_bytes += (copy.size + align - 1) & ~(align - 1ULL);
+        staging_downloads.push_back({copy, buffer_id});
+    }
+
+    std::optional<Async_Buffer> download_staging;
+    if (!staging_downloads.empty()) {
+        download_staging = runtime.DownloadStagingBuffer(staging_size_bytes, true);
+    }
+    runtime.PreCopyBarrier();
+    for (auto& [copy, buffer_id] : staging_downloads) {
+        copy.dst_offset += download_staging->offset;
+        const std::array copies{copy};
+        Buffer& buffer = slot_buffers[buffer_id];
+        BufferCopy record{copy};
+        record.src_offset = static_cast<size_t>(buffer.CpuAddr()) + copy.src_offset;
+        const DAddr orig_device_addr = static_cast<DAddr>(record.src_offset);
         async_downloads.Add(orig_device_addr, copy.size);
         buffer.MarkUsage(copy.src_offset, copy.size);
-        runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
-        normalized_copies.push_back(second_copy);
+        runtime.CopyBuffer(download_staging->buffer, buffer, copies, false);
+        batch.staging_copies.push_back(record);
+    }
+    if constexpr (USE_UNIFIED_MEMORY) {
+        for (const auto& queued : unified_copy_queue) {
+            const std::span<const BufferCopy> group_span(queued.copies.data(),
+                                                         queued.copies.size());
+            runtime.CopyToUnifiedMemory(queued.window, slot_buffers[queued.buffer_id], group_span);
+        }
+        if (!unified_copy_queue.empty()) {
+            runtime.FlushUnifiedMemoryCopies();
+            runtime.UnifiedMemoryHostBarrier();
+        }
     }
     runtime.PostCopyBarrier();
-    pending_downloads.emplace_back(std::move(normalized_copies));
-    async_buffers.emplace_back(download_staging);
+    pending_downloads.emplace_back(std::move(batch));
+    async_buffers.emplace_back(std::move(download_staging));
 }
 
 template <class P>
@@ -674,32 +735,49 @@ void BufferCache<P>::PopAsyncFlushes() {
 
 template <class P>
 void BufferCache<P>::PopAsyncBuffers() {
-    if (async_buffers.empty()) {
-        return;
-    }
-    if (!async_buffers.front().has_value()) {
+    struct Writeback {
+        DAddr addr;
+        const u8* src;
+        u64 size;
+    };
+    boost::container::small_vector<Writeback, 8> writebacks;
+    {
+        std::scoped_lock lock{mutex};
+        if (async_buffers.empty()) {
+            return;
+        }
+        auto& batch = pending_downloads.front();
+        auto& async_buffer = async_buffers.front();
+        if (async_buffer.has_value()) {
+            const u8* base = async_buffer->mapped_span.data();
+            const size_t base_offset = async_buffer->offset;
+            for (const auto& copy : batch.staging_copies) {
+                const DAddr device_addr = static_cast<DAddr>(copy.src_offset);
+                const u64 dst_offset = copy.dst_offset - base_offset;
+                const u8* read_mapped_memory = base + dst_offset;
+                async_downloads.ForEachInRange(
+                    device_addr, copy.size, [&](DAddr start, DAddr end, s32) {
+                        writebacks.push_back(
+                            {start, &read_mapped_memory[start - device_addr], end - start});
+                    });
+                async_downloads.Subtract(device_addr, copy.size, [&](DAddr start, DAddr end) {
+                    gpu_modified_ranges.Subtract(start, end - start);
+                });
+            }
+            async_buffers_death_ring.emplace_back(*async_buffer);
+        }
+        for (const auto& copy : batch.unified_copies) {
+            const DAddr device_addr = static_cast<DAddr>(copy.src_offset);
+            async_downloads.Subtract(device_addr, copy.size, [&](DAddr start, DAddr end) {
+                gpu_modified_ranges.Subtract(start, end - start);
+            });
+        }
         async_buffers.pop_front();
-        return;
+        pending_downloads.pop_front();
     }
-    auto& downloads = pending_downloads.front();
-    auto& async_buffer = async_buffers.front();
-    u8* base = async_buffer->mapped_span.data();
-    const size_t base_offset = async_buffer->offset;
-    for (const auto& copy : downloads) {
-        const DAddr device_addr = static_cast<DAddr>(copy.src_offset);
-        const u64 dst_offset = copy.dst_offset - base_offset;
-        const u8* read_mapped_memory = base + dst_offset;
-        async_downloads.ForEachInRange(device_addr, copy.size, [&](DAddr start, DAddr end, s32) {
-            device_memory.WriteBlockUnsafe(start, &read_mapped_memory[start - device_addr],
-                                           end - start);
-        });
-        async_downloads.Subtract(device_addr, copy.size, [&](DAddr start, DAddr end) {
-            gpu_modified_ranges.Subtract(start, end - start);
-        });
+    for (const auto& wb : writebacks) {
+        device_memory.WriteBlockUnsafe(wb.addr, wb.src, wb.size);
     }
-    async_buffers_death_ring.emplace_back(*async_buffer);
-    async_buffers.pop_front();
-    pending_downloads.pop_front();
 }
 
 template <class P>
@@ -810,46 +888,46 @@ void BufferCache<P>::BindHostVertexBuffers() {
 
     if (use_optimized_vertex_buffers) {
         auto& flags = maxwell3d->dirty.flags;
-        const u32 enabled_mask = enabled_vertex_buffers_mask;
-        bool any_dirty = false;
-        u32 pending_mask = enabled_mask;
-        while (pending_mask != 0) {
-            const u32 index = std::countr_zero(pending_mask);
-            pending_mask &= (pending_mask - 1);
+        u32 enabled_mask = enabled_vertex_buffers_mask;
+        HostBindings<Buffer> bindings{};
+        u32 last_index = (std::numeric_limits<u32>::max)();
+        const auto flush_bindings = [&]() {
+            if (bindings.buffers.empty()) {
+                return;
+            }
+            bindings.max_index = bindings.min_index + static_cast<u32>(bindings.buffers.size());
+            runtime.BindVertexBuffers(bindings);
+            bindings = HostBindings<Buffer>{};
+            last_index = (std::numeric_limits<u32>::max)();
+        };
+        while (enabled_mask != 0) {
+            const u32 index = std::countr_zero(enabled_mask);
+            enabled_mask &= (enabled_mask - 1);
             const Binding& binding = VertexBufferSlot(index);
             Buffer& buffer = slot_buffers[binding.buffer_id];
             TouchBuffer(buffer, binding.buffer_id);
             SynchronizeBuffer(buffer, binding.device_addr, binding.size);
-            any_dirty |= flags[Dirty::VertexBuffer0 + index];
-        }
-        if (enabled_mask == 0 || !any_dirty) {
-            return;
-        }
-        const u32 min_index = static_cast<u32>(std::countr_zero(enabled_mask));
-        const u32 max_index = 32u - static_cast<u32>(std::countl_zero(enabled_mask));
-        HostBindings<Buffer> bindings{};
-        bindings.min_index = min_index;
-        bindings.max_index = max_index;
-        for (u32 index = min_index; index < max_index; ++index) {
-            flags[Dirty::VertexBuffer0 + index] = false;
-            const u32 stride = maxwell3d->regs.vertex_streams[index].stride;
-            if ((enabled_mask & (1u << index)) == 0) {
-                bindings.buffers.push_back(&slot_buffers[NULL_BUFFER_ID]);
-                bindings.offsets.push_back(0);
-                bindings.sizes.push_back(0);
-                bindings.strides.push_back(stride);
+            if (!flags[Dirty::VertexBuffer0 + index]) {
+                flush_bindings();
                 continue;
             }
-            const Binding& binding = VertexBufferSlot(index);
-            Buffer& buffer = slot_buffers[binding.buffer_id];
+            flags[Dirty::VertexBuffer0 + index] = false;
+            const u32 stride = maxwell3d->regs.vertex_streams[index].stride;
             const u32 offset = buffer.Offset(binding.device_addr);
             buffer.MarkUsage(offset, binding.size);
+            if (!bindings.buffers.empty() && index != last_index + 1) {
+                flush_bindings();
+            }
+            if (bindings.buffers.empty()) {
+                bindings.min_index = index;
+            }
             bindings.buffers.push_back(&buffer);
             bindings.offsets.push_back(offset);
             bindings.sizes.push_back(binding.size);
             bindings.strides.push_back(stride);
+            last_index = index;
         }
-        runtime.BindVertexBuffers(bindings);
+        flush_bindings();
     } else {
         HostBindings<typename P::Buffer> host_bindings;
         bool any_valid{false};
@@ -1700,6 +1778,99 @@ void BufferCache<P>::ImmediateUploadMemory([[maybe_unused]] Buffer& buffer,
     }
 }
 
+template <class P>
+bool BufferCache<P>::ResolveUnifiedWindows(
+    [[maybe_unused]] DAddr device_addr, [[maybe_unused]] u64 buffer_offset,
+    [[maybe_unused]] u64 size, [[maybe_unused]] boost::container::small_vector<u64, 4>& window_ids,
+    [[maybe_unused]] UnifiedWindowGroups& groups) {
+    if constexpr (USE_UNIFIED_MEMORY) {
+        const u8* const physical_base = device_memory.GetPhysicalBase();
+        const u64 unified_base = runtime.UnifiedMemoryBase();
+        const u64 unified_size = runtime.UnifiedMemorySize();
+        const u64 window_size = runtime.UnifiedMemoryWindowSize();
+        if (window_size == 0) {
+            return false;
+        }
+        const auto group_for = [&](u64 window) -> boost::container::small_vector<BufferCopy, 16>& {
+            for (size_t i = 0; i < window_ids.size(); ++i) {
+                if (window_ids[i] == window) {
+                    return groups[i];
+                }
+            }
+            window_ids.push_back(window);
+            groups.emplace_back();
+            return groups.back();
+        };
+        u64 downloaded = 0;
+        while (downloaded < size) {
+            const DAddr page_addr = device_addr + downloaded;
+            const u8* const ptr = device_memory.GetPointer<u8>(page_addr);
+            if (ptr == nullptr) {
+                return false;
+            }
+            const u64 page_offset = page_addr & Core::DEVICE_PAGEMASK;
+            u64 chunk = (std::min)(size - downloaded,
+                                   static_cast<u64>(Core::DEVICE_PAGESIZE) - page_offset);
+            const u64 phys_offset = static_cast<u64>(ptr - physical_base);
+            if (phys_offset < unified_base || phys_offset - unified_base + chunk > unified_size) {
+                return false;
+            }
+            const u64 relative = phys_offset - unified_base;
+            const u64 window = relative / window_size;
+            const u64 local_offset = relative % window_size;
+            chunk = (std::min)(chunk, window_size - local_offset);
+            auto& group = group_for(window);
+            if (!group.empty()) {
+                BufferCopy& last = group.back();
+                if (last.src_offset + last.size == buffer_offset + downloaded &&
+                    last.dst_offset + last.size == local_offset) {
+                    last.size += chunk;
+                    downloaded += chunk;
+                    continue;
+                }
+            }
+            group.push_back(BufferCopy{
+                .src_offset = buffer_offset + downloaded,
+                .dst_offset = local_offset,
+                .size = chunk,
+            });
+            downloaded += chunk;
+        }
+        return true;
+    } else {
+        return false;
+    }
+}
+
+template <class P>
+bool BufferCache<P>::TryUnifiedDownloadMemory([[maybe_unused]] Buffer& buffer,
+                                              [[maybe_unused]] std::span<BufferCopy> copies) {
+    if constexpr (USE_UNIFIED_MEMORY) {
+        boost::container::small_vector<u64, 4> window_ids;
+        UnifiedWindowGroups groups;
+        for (const BufferCopy& copy : copies) {
+            if (!ResolveUnifiedWindows(buffer.CpuAddr() + copy.src_offset, copy.src_offset,
+                                       copy.size, window_ids, groups)) {
+                return false;
+            }
+        }
+        for (const BufferCopy& copy : copies) {
+            buffer.MarkUsage(copy.src_offset, copy.size);
+        }
+        runtime.PreCopyBarrier();
+        for (size_t i = 0; i < window_ids.size(); ++i) {
+            const std::span<const BufferCopy> group_span(groups[i].data(), groups[i].size());
+            runtime.CopyToUnifiedMemory(window_ids[i], buffer, group_span);
+        }
+        runtime.FlushUnifiedMemoryCopies();
+        runtime.UnifiedMemoryHostBarrier();
+        runtime.Finish();
+        return true;
+    } else {
+        return false;
+    }
+}
+
 template <class P>
 void BufferCache<P>::MappedUploadMemory([[maybe_unused]] Buffer& buffer,
                                         [[maybe_unused]] u64 total_size_bytes,
@@ -1803,6 +1974,12 @@ void BufferCache<P>::DownloadBufferMemory(Buffer& buffer, DAddr device_addr, u64
     }
 
     if constexpr (USE_MEMORY_MAPS) {
+        if constexpr (USE_UNIFIED_MEMORY) {
+            if (runtime.HasUnifiedMemory() &&
+                TryUnifiedDownloadMemory(buffer, std::span(copies.data(), copies.size()))) {
+                return;
+            }
+        }
         auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes);
         const u8* const mapped_memory = download_staging.mapped_span.data();
         const std::span<BufferCopy> copies_span(copies.data(), copies.data() + copies.size());

src/video_core/buffer_cache/buffer_cache_base.h

@@ -180,6 +180,7 @@ class BufferCache : public VideoCommon::ChannelSetupCaches<BufferCacheChannelInf
     static constexpr bool USE_MEMORY_MAPS = P::USE_MEMORY_MAPS;
     static constexpr bool SEPARATE_IMAGE_BUFFERS_BINDINGS = P::SEPARATE_IMAGE_BUFFER_BINDINGS;
     static constexpr bool USE_MEMORY_MAPS_FOR_UPLOADS = P::USE_MEMORY_MAPS_FOR_UPLOADS;
+    static constexpr bool USE_UNIFIED_MEMORY = P::USE_UNIFIED_MEMORY;
 
 #ifdef YUZU_LEGACY
     static constexpr s64 TARGET_THRESHOLD = 3_GiB;
@@ -443,6 +444,15 @@ private:
 
     void MappedUploadMemory(Buffer& buffer, u64 total_size_bytes, std::span<BufferCopy> copies);
 
+    bool TryUnifiedDownloadMemory(Buffer& buffer, std::span<BufferCopy> copies);
+
+    using UnifiedWindowGroups =
+        boost::container::small_vector<boost::container::small_vector<BufferCopy, 16>, 4>;
+
+    bool ResolveUnifiedWindows(DAddr device_addr, u64 buffer_offset, u64 size,
+                               boost::container::small_vector<u64, 4>& window_ids,
+                               UnifiedWindowGroups& groups);
+
     void DownloadBufferMemory(Buffer& buffer_id);
 
     void DownloadBufferMemory(Buffer& buffer_id, DAddr device_addr, u64 size);
@@ -498,9 +508,14 @@ private:
     std::deque<Common::RangeSet<DAddr>> committed_gpu_modified_ranges;
 
     // Async Buffers
+    struct AsyncDownloadBatch {
+        boost::container::small_vector<BufferCopy, 4> staging_copies;
+        boost::container::small_vector<BufferCopy, 4> unified_copies;
+    };
+
     Common::OverlapRangeSet<DAddr> async_downloads;
     std::deque<std::optional<Async_Buffer>> async_buffers;
-    std::deque<boost::container::small_vector<BufferCopy, 4>> pending_downloads;
+    std::deque<AsyncDownloadBatch> pending_downloads;
     std::optional<Async_Buffer> current_buffer;
 
     std::deque<Async_Buffer> async_buffers_death_ring;

src/video_core/engines/maxwell_3d.h

@@ -2257,7 +2257,7 @@ public:
             /// Returns whether the vertex array specified by index is supposed to be
             /// accessed per instance or not.
             bool IsInstancingEnabled(std::size_t index) const {
-                return bool(is_instanced[index]);
+                return bool(is_instanced[index]); //FUCK YOU MSVC
             }
         };
 

src/video_core/engines/nv01_timer.h

@@ -40,6 +40,7 @@ public:
     }
 
     struct Regs {
+        // No fucking idea
         INSERT_PADDING_BYTES_NOINIT(0x48);
     } regs{};
 private:

src/video_core/host1x/vic.cpp

@@ -963,7 +963,7 @@ void Vic::WriteABGR(const OutputSurfaceConfig& output_surface_config, VideoPixel
                     auto pixel1213 = _mm_load_si128((__m128i*)&inp[src + x + 12]);
                     auto pixel1415 = _mm_load_si128((__m128i*)&inp[src + x + 14]);
 
-                    // Right-shift the channels by 16 to un-do the left shift on read and bring the range
+                    // Right-shift the channels by 16 to un-do the left shit on read and bring the range
                     // back to 8-bit.
                     pixel01 = _mm_srli_epi16(pixel01, 2);
                     pixel23 = _mm_srli_epi16(pixel23, 2);

src/video_core/host_shaders/CMakeLists.txt

@@ -17,13 +17,11 @@ set(SHADER_FILES
     ${CMAKE_CURRENT_SOURCE_DIR}/astc_decoder.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_color_float.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_2d.comp
-    ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_2d_buffer.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_color_msaa.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_depth_msaa.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_depth_stencil_msaa.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_3d.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_3d_bcn.comp
-    ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_3d_buffer.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_abgr8_to_d24s8.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_abgr8_to_d32f.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_d32f_to_abgr8.frag
@@ -34,8 +32,6 @@ set(SHADER_FILES
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa.frag
-    ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa_depth.frag
-    ${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa_depth_stencil.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/convert_s8d24_to_abgr8.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/full_screen_triangle.vert
     ${CMAKE_CURRENT_SOURCE_DIR}/fxaa.frag
@@ -92,7 +88,7 @@ set(SHADER_FILES
     ${CMAKE_CURRENT_SOURCE_DIR}/sgsr1_shader_mobile_edge_direction.frag
 )
 
-if (HAIKUOS)
+if (PLATFORM_HAIKU)
     # glslangValidator WILL crash, glslang will not
     set(GLSLANGVALIDATOR "glslang")
 else()

src/video_core/host_shaders/block_linear_unswizzle_2d_buffer.comp

@@ -1,104 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-#version 430
-
-#extension GL_EXT_shader_16bit_storage : require
-#extension GL_EXT_shader_8bit_storage : require
-
-#define BINDING_INPUT_BUFFER 0
-#define BINDING_OUTPUT_BUFFER 1
-
-layout(push_constant) uniform PushConstants {
-    uvec3 dim;
-    uint bytes_per_block_log2;
-
-    uvec3 origin;
-    uint layer_stride;
-
-    uint block_size;
-    uint x_shift;
-    uint block_height;
-    uint block_height_mask;
-} pc;
-
-layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU32 { uint u32data[]; };
-layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU64 { uvec2 u64data[]; };
-layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU128 { uvec4 u128data[]; };
-
-layout(binding = BINDING_OUTPUT_BUFFER, std430) writeonly buffer OutputBuffer {
-    uint out_u32[];
-};
-
-layout(local_size_x = 16, local_size_y = 8, local_size_z = 1) in;
-
-const uint GOB_SIZE_X = 64;
-const uint GOB_SIZE_Y = 8;
-
-const uint GOB_SIZE_X_SHIFT = 6;
-const uint GOB_SIZE_Y_SHIFT = 3;
-const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT;
-
-const uvec2 SWIZZLE_MASK = uvec2(GOB_SIZE_X - 1u, GOB_SIZE_Y - 1u);
-
-uint SwizzleTable(uint pos) {
-    const uint t[8] = uint[](
-        0x12100200, 0x13110301, 0x16140604, 0x17150705,
-        0x1a180a08, 0x1b190b09, 0x1e1c0e0c, 0x1f1d0f0d
-    );
-    const uint i = pos >> 4;
-    const uint h = (t[i / 4] >> ((i % 4) * 8)) & 0xff;
-    return (h << 4) | (pos & 0xf);
-}
-
-uint SwizzleOffset(uvec2 pos) {
-    pos = pos & SWIZZLE_MASK;
-    return SwizzleTable(pos.y * 64u + pos.x);
-}
-
-uvec4 ReadTexel(uint offset) {
-    switch (pc.bytes_per_block_log2) {
-    case 2u:
-        return uvec4(u32data[offset / 4u], 0u, 0u, 0u);
-    case 3u:
-        return uvec4(u64data[offset / 8u], 0u, 0u);
-    case 4u:
-        return u128data[offset / 16u];
-    }
-    return uvec4(0u);
-}
-
-void main() {
-    uvec3 coord = gl_GlobalInvocationID;
-    if (coord.x >= pc.dim.x || coord.y >= pc.dim.y || coord.z >= pc.dim.z) {
-        return;
-    }
-
-    uvec3 pos = coord + pc.origin;
-    pos.x <<= pc.bytes_per_block_log2;
-
-    uint swizzle = SwizzleOffset(pos.xy);
-    uint block_y = pos.y >> GOB_SIZE_Y_SHIFT;
-
-    uint offset = 0u;
-    offset += pos.z * pc.layer_stride;
-    offset += (block_y >> pc.block_height) * pc.block_size;
-    offset += (block_y & pc.block_height_mask) << GOB_SIZE_SHIFT;
-    offset += (pos.x >> GOB_SIZE_X_SHIFT) << pc.x_shift;
-    offset += swizzle;
-
-    uvec4 texel = ReadTexel(offset);
-
-    uint words = 1u << (pc.bytes_per_block_log2 - 2u);
-    uint linear_index = coord.x + coord.y * pc.dim.x + coord.z * pc.dim.x * pc.dim.y;
-    uint out_idx = linear_index * words;
-
-    out_u32[out_idx] = texel.x;
-    if (words > 1u) {
-        out_u32[out_idx + 1u] = texel.y;
-    }
-    if (words > 2u) {
-        out_u32[out_idx + 2u] = texel.z;
-        out_u32[out_idx + 3u] = texel.w;
-    }
-}

src/video_core/host_shaders/block_linear_unswizzle_3d_buffer.comp

@@ -1,105 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-#version 430
-
-#define BINDING_INPUT_BUFFER 0
-#define BINDING_OUTPUT_BUFFER 1
-
-layout(push_constant) uniform PushConstants {
-    uvec3 dim;
-    uint bytes_per_block_log2;
-
-    uvec3 origin;
-    uint slice_size;
-
-    uint block_size;
-    uint x_shift;
-    uint block_height;
-    uint block_height_mask;
-
-    uint block_depth;
-    uint block_depth_mask;
-} pc;
-
-layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU32 { uint u32data[]; };
-layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU64 { uvec2 u64data[]; };
-layout(binding = BINDING_INPUT_BUFFER, std430) buffer InputBufferU128 { uvec4 u128data[]; };
-
-layout(binding = BINDING_OUTPUT_BUFFER, std430) writeonly buffer OutputBuffer {
-    uint out_u32[];
-};
-
-layout(local_size_x = 8, local_size_y = 8, local_size_z = 4) in;
-
-const uint GOB_SIZE_X = 64;
-const uint GOB_SIZE_Y = 8;
-
-const uint GOB_SIZE_X_SHIFT = 6;
-const uint GOB_SIZE_Y_SHIFT = 3;
-const uint GOB_SIZE_SHIFT = GOB_SIZE_X_SHIFT + GOB_SIZE_Y_SHIFT;
-
-const uvec2 SWIZZLE_MASK = uvec2(GOB_SIZE_X - 1u, GOB_SIZE_Y - 1u);
-
-uint SwizzleTable(uint pos) {
-    const uint t[8] = uint[](
-        0x12100200, 0x13110301, 0x16140604, 0x17150705,
-        0x1a180a08, 0x1b190b09, 0x1e1c0e0c, 0x1f1d0f0d
-    );
-    const uint i = pos >> 4;
-    const uint h = (t[i / 4] >> ((i % 4) * 8)) & 0xff;
-    return (h << 4) | (pos & 0xf);
-}
-
-uint SwizzleOffset(uvec2 pos) {
-    pos = pos & SWIZZLE_MASK;
-    return SwizzleTable(pos.y * 64u + pos.x);
-}
-
-uvec4 ReadTexel(uint offset) {
-    switch (pc.bytes_per_block_log2) {
-    case 2u:
-        return uvec4(u32data[offset / 4u], 0u, 0u, 0u);
-    case 3u:
-        return uvec4(u64data[offset / 8u], 0u, 0u);
-    case 4u:
-        return u128data[offset / 16u];
-    }
-    return uvec4(0u);
-}
-
-void main() {
-    uvec3 coord = gl_GlobalInvocationID;
-    if (coord.x >= pc.dim.x || coord.y >= pc.dim.y || coord.z >= pc.dim.z) {
-        return;
-    }
-
-    uvec3 pos = coord + pc.origin;
-    pos.x <<= pc.bytes_per_block_log2;
-
-    uint swizzle = SwizzleOffset(pos.xy);
-    uint block_y = pos.y >> GOB_SIZE_Y_SHIFT;
-
-    uint offset = 0u;
-    offset += (pos.z >> pc.block_depth) * pc.slice_size;
-    offset += (pos.z & pc.block_depth_mask) << (GOB_SIZE_SHIFT + pc.block_height);
-    offset += (block_y >> pc.block_height) * pc.block_size;
-    offset += (block_y & pc.block_height_mask) << GOB_SIZE_SHIFT;
-    offset += (pos.x >> GOB_SIZE_X_SHIFT) << pc.x_shift;
-    offset += swizzle;
-
-    uvec4 texel = ReadTexel(offset);
-
-    uint words = 1u << (pc.bytes_per_block_log2 - 2u);
-    uint linear_index = coord.x + coord.y * pc.dim.x + coord.z * pc.dim.x * pc.dim.y;
-    uint out_idx = linear_index * words;
-
-    out_u32[out_idx] = texel.x;
-    if (words > 1u) {
-        out_u32[out_idx + 1u] = texel.y;
-    }
-    if (words > 2u) {
-        out_u32[out_idx + 2u] = texel.z;
-        out_u32[out_idx + 3u] = texel.w;
-    }
-}

src/video_core/host_shaders/convert_non_msaa_to_msaa_depth.frag

@@ -1,19 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-#version 450 core
-
-layout(binding = 0) uniform sampler2D img_in;
-
-layout(push_constant) uniform PushConstants {
-    ivec2 dst_offset;
-    ivec2 src_offset;
-    ivec2 scale;
-};
-
-void main() {
-    const ivec2 msaa_coord = ivec2(gl_FragCoord.xy) - dst_offset;
-    const ivec2 sample_offset = ivec2(gl_SampleID % scale.x, gl_SampleID / scale.x);
-    const ivec2 coord = msaa_coord * scale + sample_offset + src_offset;
-    gl_FragDepth = texelFetch(img_in, coord, 0).r;
-}

src/video_core/host_shaders/convert_non_msaa_to_msaa_depth_stencil.frag

@@ -1,22 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-#version 450 core
-#extension GL_ARB_shader_stencil_export : require
-
-layout(binding = 0) uniform sampler2D depth_tex;
-layout(binding = 1) uniform usampler2D stencil_tex;
-
-layout(push_constant) uniform PushConstants {
-    ivec2 dst_offset;
-    ivec2 src_offset;
-    ivec2 scale;
-};
-
-void main() {
-    const ivec2 msaa_coord = ivec2(gl_FragCoord.xy) - dst_offset;
-    const ivec2 sample_offset = ivec2(gl_SampleID % scale.x, gl_SampleID / scale.x);
-    const ivec2 coord = msaa_coord * scale + sample_offset + src_offset;
-    gl_FragDepth = texelFetch(depth_tex, coord, 0).r;
-    gl_FragStencilRefARB = int(texelFetch(stencil_tex, coord, 0).r);
-}

src/video_core/host_shaders/sgsr1_shader.vert

@@ -7,7 +7,6 @@ layout(push_constant) uniform constants {
     vec2 scale;
     vec2 size;
     vec2 resize_factor;
-    vec2 crop_offset;
     float edge_sharpness;
 };
 layout(location = 0) out highp vec2 texcoord;
@@ -16,5 +15,5 @@ void main() {
     float x = float((gl_VertexIndex & 1) << 2);
     float y = float((gl_VertexIndex & 2) << 1);
     gl_Position = vec4(x - 1.0f, y - 1.0f, 0.0, 1.0f) * vec4(sign(resize_factor), 1.f, 1.f);
-    texcoord = crop_offset + vec2(x, y) * abs(resize_factor) * 0.5;
+    texcoord = vec2(x, y) * abs(resize_factor) * 0.5;
 }

src/video_core/host_shaders/sgsr1_shader_mobile.frag

@@ -14,7 +14,6 @@ layout(push_constant) uniform constants {
     vec2 scale;
     vec2 size;
     vec2 resize_factor;
-    vec2 crop_offset;
     float edge_sharpness;
 };
 layout(set = 0, binding = 0) uniform sampler2D sampler0;

src/video_core/host_shaders/sgsr1_shader_mobile_edge_direction.frag

@@ -13,7 +13,6 @@
 layout( push_constant ) uniform constants {
     vec4 ViewportInfo[1];
     vec2 ResizeFactor;
-    vec2 CropOffset;
     float EdgeSharpness;
 };
 layout(set = 0, binding = 0) uniform sampler2D ps0;

src/video_core/macro.cpp

@@ -441,7 +441,7 @@ void MacroInterpreterImpl::Reset() {
     pc = 0;
     delayed_pc = {};
     method_address.raw = 0;
-    // Vector must hold its last indices otherwise chaos will ensue
+    // Vector must hold its last indices otherwise wonky shit will happen
     // The next parameter index starts at 1, because $r1 already has the value of the first
     // parameter.
     next_parameter_index = 1;

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -261,6 +261,7 @@ struct BufferCacheParams {
 
     // TODO: Investigate why OpenGL seems to perform worse with persistently mapped buffer uploads
     static constexpr bool USE_MEMORY_MAPS_FOR_UPLOADS = false;
+    static constexpr bool USE_UNIFIED_MEMORY = false;
 };
 
 using BufferCache = VideoCommon::BufferCache<BufferCacheParams>;

src/video_core/renderer_opengl/gl_device.cpp

@@ -59,7 +59,7 @@ std::vector<std::string> GetExtensions() {
     std::vector<std::string> extensions;
     for (GLint index = 0; index < num_extensions; ++index) {
         auto const* p = reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, GLuint(index)));
-        if (p != nullptr)
+        if (p != nullptr) // Fuck you? - sincerely, buggy mesa drivers
             extensions.push_back(std::string{p});
     }
     return extensions;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -72,25 +72,6 @@ Shader::OutputTopology MaxwellToOutputTopology(Maxwell::PrimitiveTopology topolo
     }
 }
 
-Shader::InputTopology MaxwellToInputTopology(Maxwell::PrimitiveTopology topology) {
-    switch (topology) {
-    case Maxwell::PrimitiveTopology::Points:
-        return Shader::InputTopology::Points;
-    case Maxwell::PrimitiveTopology::Lines:
-    case Maxwell::PrimitiveTopology::LineLoop:
-    case Maxwell::PrimitiveTopology::LineStrip:
-        return Shader::InputTopology::Lines;
-    case Maxwell::PrimitiveTopology::LinesAdjacency:
-    case Maxwell::PrimitiveTopology::LineStripAdjacency:
-        return Shader::InputTopology::LinesAdjacency;
-    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
-    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
-        return Shader::InputTopology::TrianglesAdjacency;
-    default:
-        return Shader::InputTopology::Triangles;
-    }
-}
-
 Shader::RuntimeInfo MakeRuntimeInfo(const GraphicsPipelineKey& key,
                                     const Shader::IR::Program& program,
                                     const Shader::IR::Program* previous_program,
@@ -146,7 +127,33 @@ Shader::RuntimeInfo MakeRuntimeInfo(const GraphicsPipelineKey& key,
     default:
         break;
     }
-    info.input_topology = MaxwellToInputTopology(key.gs_input_topology);
+    switch (key.gs_input_topology) {
+    case Maxwell::PrimitiveTopology::Points:
+        info.input_topology = Shader::InputTopology::Points;
+        break;
+    case Maxwell::PrimitiveTopology::Lines:
+    case Maxwell::PrimitiveTopology::LineLoop:
+    case Maxwell::PrimitiveTopology::LineStrip:
+        info.input_topology = Shader::InputTopology::Lines;
+        break;
+    case Maxwell::PrimitiveTopology::Triangles:
+    case Maxwell::PrimitiveTopology::TriangleStrip:
+    case Maxwell::PrimitiveTopology::TriangleFan:
+    case Maxwell::PrimitiveTopology::Quads:
+    case Maxwell::PrimitiveTopology::QuadStrip:
+    case Maxwell::PrimitiveTopology::Polygon:
+    case Maxwell::PrimitiveTopology::Patches:
+        info.input_topology = Shader::InputTopology::Triangles;
+        break;
+    case Maxwell::PrimitiveTopology::LinesAdjacency:
+    case Maxwell::PrimitiveTopology::LineStripAdjacency:
+        info.input_topology = Shader::InputTopology::LinesAdjacency;
+        break;
+    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
+    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
+        info.input_topology = Shader::InputTopology::TrianglesAdjacency;
+        break;
+    }
     info.glasm_use_storage_buffers = glasm_use_storage_buffers;
     return info;
 }
@@ -230,6 +237,7 @@ ShaderCache::ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
           .has_gl_bool_ref_bug = device.HasBoolRefBug(),
           .ignore_nan_fp_comparisons = true,
           .gl_max_compute_smem_size = device.GetMaxComputeSharedMemorySize(),
+          .max_shared_memory_size = device.GetMaxComputeSharedMemorySize(),
           .min_ssbo_alignment = device.GetShaderStorageBufferAlignment(),
           // Use the host limit, but never more than the guest can produce. Maxwell exposes 8 clip
           // distances and the SPIR-V output array is sized for at most 8, so clamping here keeps a
@@ -476,10 +484,8 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
             && index == u32(Maxwell::ShaderType::Geometry);
         if (key.unique_hashes[index] == 0 && is_emulated_stage) {
             auto topology = MaxwellToOutputTopology(key.gs_input_topology);
-            programs[index] =
-                GenerateGeometryPassthrough(pools.inst, pools.block, host_info,
-                                            *layer_source_program, topology,
-                                            MaxwellToInputTopology(key.gs_input_topology));
+            programs[index] = GenerateGeometryPassthrough(pools.inst, pools.block, host_info,
+                                                          *layer_source_program, topology);
             continue;
         }
         if (key.unique_hashes[index] == 0) {
@@ -497,15 +503,13 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
 
         if (!uses_vertex_a || index != 1) {
             // Normal path
-            programs[index] = TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
-                                              MaxwellToInputTopology(key.gs_input_topology));
+            programs[index] = TranslateProgram(pools.inst, pools.block, env, cfg, host_info);
 
             total_storage_buffers += Shader::NumDescriptors(programs[index].info.storage_buffers_descriptors);
         } else {
             // VertexB path when VertexA is present.
             auto& program_va{programs[0]};
-            auto program_vb{TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
-                                             MaxwellToInputTopology(key.gs_input_topology))};
+            auto program_vb{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
             total_storage_buffers += Shader::NumDescriptors(program_vb.info.storage_buffers_descriptors);
             programs[index] = MergeDualVertexPrograms(program_va, program_vb, env);
         }
@@ -594,8 +598,7 @@ std::unique_ptr<ComputePipeline> ShaderCache::CreateComputePipeline(
         env.Dump(hash, key.unique_hash);
     }
 
-    auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
-                                  Shader::InputTopology::Points)};
+    auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
     const u32 num_storage_buffers{Shader::NumDescriptors(program.info.storage_buffers_descriptors)};
     Shader::RuntimeInfo info;
     info.glasm_use_storage_buffers = num_storage_buffers <= device.GetMaxGLASMStorageBufferBlocks();

src/video_core/renderer_vulkan/blit_image.cpp

@@ -22,8 +22,6 @@
 #include "video_core/host_shaders/convert_float_to_depth_frag_spv.h"
 #include "video_core/host_shaders/convert_msaa_to_non_msaa_frag_spv.h"
 #include "video_core/host_shaders/convert_non_msaa_to_msaa_frag_spv.h"
-#include "video_core/host_shaders/convert_non_msaa_to_msaa_depth_frag_spv.h"
-#include "video_core/host_shaders/convert_non_msaa_to_msaa_depth_stencil_frag_spv.h"
 #include "video_core/host_shaders/convert_s8d24_to_abgr8_frag_spv.h"
 #include "video_core/host_shaders/full_screen_triangle_vert_spv.h"
 #include "video_core/host_shaders/vulkan_blit_depth_stencil_frag_spv.h"
@@ -521,8 +519,7 @@ void RecordShaderReadBarrier(Scheduler& scheduler, const ImageView& image_view)
 }
 
 [[nodiscard]] vk::ImageView MakeMSAACopyView(const vk::Device& device, VkImage image,
-                                             VkFormat format, u32 base_level,
-                                             VkImageAspectFlags aspect_mask) {
+                                             VkFormat format, u32 base_level) {
     return device.CreateImageView(VkImageViewCreateInfo{
         .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
         .pNext = nullptr,
@@ -537,7 +534,7 @@ void RecordShaderReadBarrier(Scheduler& scheduler, const ImageView& image_view)
             .a = VK_COMPONENT_SWIZZLE_IDENTITY,
         },
         .subresourceRange{
-            .aspectMask = aspect_mask,
+            .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
             .baseMipLevel = base_level,
             .levelCount = 1,
             .baseArrayLayer = 0,
@@ -589,10 +586,6 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
       msaa_copy_pipeline_layout(device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo(
           one_texture_set_layout.address(),
           PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(MSAACopyPushConstants)>))),
-      msaa_copy_depth_stencil_pipeline_layout(
-          device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo(
-              two_textures_set_layout.address(),
-              PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(MSAACopyPushConstants)>))),
       full_screen_vert(BuildShader(device, FULL_SCREEN_TRIANGLE_VERT_SPV)),
       blit_color_to_color_frag(BuildShader(device, BLIT_COLOR_FLOAT_FRAG_SPV)),
       blit_color_msaa_frag(BuildShader(device, BLIT_COLOR_MSAA_FRAG_SPV)),
@@ -617,12 +610,6 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
       convert_s8d24_to_abgr8_frag(BuildShader(device, CONVERT_S8D24_TO_ABGR8_FRAG_SPV)),
       convert_msaa_to_non_msaa_frag(BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_FRAG_SPV)),
       convert_non_msaa_to_msaa_frag(BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_FRAG_SPV)),
-      convert_non_msaa_to_msaa_depth_frag(
-          BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_DEPTH_FRAG_SPV)),
-      convert_non_msaa_to_msaa_depth_stencil_frag(
-          device.IsExtShaderStencilExportSupported()
-              ? BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_DEPTH_STENCIL_FRAG_SPV)
-              : vk::ShaderModule{}),
       linear_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_LINEAR>)),
       nearest_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_NEAREST>)) {}
 
@@ -933,12 +920,10 @@ void BlitImageHelper::CopyMSAA(RenderPassCache& render_pass_cache, VkImage dst_i
         ASSERT(copy.dst_subresource.num_layers == 1);
         vk::ImageView src_view =
             MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format,
-                             static_cast<u32>(copy.src_subresource.base_level),
-                             VK_IMAGE_ASPECT_COLOR_BIT);
+                             static_cast<u32>(copy.src_subresource.base_level));
         vk::ImageView dst_view =
             MakeMSAACopyView(device.GetLogical(), dst_image, dst_vk_format,
-                             static_cast<u32>(copy.dst_subresource.base_level),
-                             VK_IMAGE_ASPECT_COLOR_BIT);
+                             static_cast<u32>(copy.dst_subresource.base_level));
         const VkOffset2D dst_offset{copy.dst_offset.x, copy.dst_offset.y};
         const VkExtent2D dst_extent{copy.extent.width, copy.extent.height};
         const VkRect2D render_area{
@@ -1206,7 +1191,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceColorPipeline(const BlitImagePipelineKe
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache()));
+    }));
     return *blit_color_pipelines.back();
 }
 
@@ -1238,7 +1223,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceDepthStencilPipeline(const BlitImagePip
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache()));
+    }));
     return *blit_depth_stencil_pipelines.back();
 }
 
@@ -1291,7 +1276,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceClearColorPipeline(const BlitImagePipel
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache()));
+    }));
     return *clear_color_pipelines.back();
 }
 
@@ -1347,7 +1332,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceClearStencilPipeline(
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache()));
+    }));
     return *clear_stencil_pipelines.back();
 }
 
@@ -1390,7 +1375,7 @@ VkPipeline BlitImageHelper::FindOrEmplaceBlitColorMSAAPipeline(const BlitMSAAPip
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache()));
+    }));
     return *blit_msaa_color_pipelines.back();
 }
 
@@ -1428,202 +1413,10 @@ VkPipeline BlitImageHelper::FindOrEmplaceResolveDepthStencilPipeline(VkRenderPas
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache()));
+    }));
     return *pipelines.back();
 }
 
-void BlitImageHelper::CopyMSAADepth(RenderPassCache& render_pass_cache, VkImage dst_image,
-                                    VideoCore::Surface::PixelFormat dst_format, VkImage src_image,
-                                    VideoCore::Surface::PixelFormat src_format, u32 num_samples,
-                                    std::span<const VideoCommon::ImageCopy> copies,
-                                    bool copy_stencil) {
-    while (!msaa_copy_resources.empty() && scheduler.IsFree(msaa_copy_resources.front().tick)) {
-        msaa_copy_resources.pop_front();
-    }
-    const auto [samples_x, samples_y] = VideoCommon::SamplesLog2(static_cast<int>(num_samples));
-    const s32 scale_x = 1 << samples_x;
-    const s32 scale_y = 1 << samples_y;
-    const VkSampleCountFlagBits samples = SampleCountFlag(num_samples);
-    RenderPassKey renderpass_key{};
-    renderpass_key.color_formats.fill(VideoCore::Surface::PixelFormat::Invalid);
-    renderpass_key.depth_format = dst_format;
-    renderpass_key.samples = samples;
-    const VkRenderPass renderpass = render_pass_cache.Get(renderpass_key);
-    const MSAACopyPipelineKey key{
-        .renderpass = renderpass,
-        .samples = samples,
-        .msaa_to_non_msaa = false,
-    };
-    const VkPipeline pipeline = FindOrEmplaceMSAACopyDepthPipeline(key, copy_stencil);
-    const VkPipelineLayout layout = copy_stencil ? *msaa_copy_depth_stencil_pipeline_layout
-                                                 : *msaa_copy_pipeline_layout;
-    const VkSampler sampler = *nearest_sampler;
-    const VkFormat src_vk_format =
-        MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, true, src_format).format;
-    const VkFormat dst_vk_format =
-        MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, true, dst_format).format;
-    const VkImageAspectFlags attachment_aspect =
-        VideoCore::Surface::GetFormatType(dst_format) ==
-                VideoCore::Surface::SurfaceType::DepthStencil
-            ? VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT
-            : VK_IMAGE_ASPECT_DEPTH_BIT;
-    for (const VideoCommon::ImageCopy& copy : copies) {
-        ASSERT(copy.src_subresource.base_layer == 0);
-        ASSERT(copy.src_subresource.num_layers == 1);
-        ASSERT(copy.dst_subresource.base_layer == 0);
-        ASSERT(copy.dst_subresource.num_layers == 1);
-        vk::ImageView src_view =
-            MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format,
-                             static_cast<u32>(copy.src_subresource.base_level),
-                             VK_IMAGE_ASPECT_DEPTH_BIT);
-        vk::ImageView src_stencil_view =
-            copy_stencil ? MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format,
-                                            static_cast<u32>(copy.src_subresource.base_level),
-                                            VK_IMAGE_ASPECT_STENCIL_BIT)
-                         : vk::ImageView{};
-        vk::ImageView dst_view =
-            MakeMSAACopyView(device.GetLogical(), dst_image, dst_vk_format,
-                             static_cast<u32>(copy.dst_subresource.base_level),
-                             attachment_aspect);
-        const VkOffset2D dst_offset{copy.dst_offset.x, copy.dst_offset.y};
-        const VkExtent2D dst_extent{copy.extent.width, copy.extent.height};
-        const VkRect2D render_area{
-            .offset = dst_offset,
-            .extent = dst_extent,
-        };
-        vk::Framebuffer framebuffer = device.GetLogical().CreateFramebuffer(VkFramebufferCreateInfo{
-            .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
-            .pNext = nullptr,
-            .flags = 0,
-            .renderPass = renderpass,
-            .attachmentCount = 1,
-            .pAttachments = dst_view.address(),
-            .width = static_cast<u32>(dst_offset.x) + dst_extent.width,
-            .height = static_cast<u32>(dst_offset.y) + dst_extent.height,
-            .layers = 1,
-        });
-        const MSAACopyPushConstants push_constants{
-            .dst_offset = {dst_offset.x, dst_offset.y},
-            .src_offset = {copy.src_offset.x, copy.src_offset.y},
-            .scale = {scale_x, scale_y},
-        };
-        scheduler.RequestOutsideRenderPassOperationContext();
-        const VkImageView src_stencil_handle = copy_stencil ? *src_stencil_view : VK_NULL_HANDLE;
-        scheduler.Record([this, pipeline, layout, sampler, renderpass,
-                          framebuffer_handle = *framebuffer, src_view_handle = *src_view,
-                          src_stencil_handle, src = src_image, dst = dst_image, render_area,
-                          attachment_aspect, push_constants](vk::CommandBuffer cmdbuf) {
-            const VkImageSubresourceRange src_range{
-                .aspectMask = attachment_aspect,
-                .baseMipLevel = 0,
-                .levelCount = VK_REMAINING_MIP_LEVELS,
-                .baseArrayLayer = 0,
-                .layerCount = VK_REMAINING_ARRAY_LAYERS,
-            };
-            const std::array pre_barriers{
-                VkImageMemoryBarrier{
-                    .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-                    .pNext = nullptr,
-                    .srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
-                                     VK_ACCESS_TRANSFER_WRITE_BIT,
-                    .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
-                    .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
-                    .newLayout = VK_IMAGE_LAYOUT_GENERAL,
-                    .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                    .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                    .image = src,
-                    .subresourceRange = src_range,
-                },
-                VkImageMemoryBarrier{
-                    .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-                    .pNext = nullptr,
-                    .srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
-                                     VK_ACCESS_TRANSFER_WRITE_BIT,
-                    .dstAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
-                                     VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
-                    .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
-                    .newLayout = VK_IMAGE_LAYOUT_GENERAL,
-                    .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                    .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                    .image = dst,
-                    .subresourceRange = src_range,
-                },
-            };
-            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
-                                       VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
-                                       VK_PIPELINE_STAGE_TRANSFER_BIT,
-                                   VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
-                                       VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT,
-                                   0, nullptr, nullptr, pre_barriers);
-            const VkRenderPassBeginInfo renderpass_bi{
-                .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
-                .pNext = nullptr,
-                .renderPass = renderpass,
-                .framebuffer = framebuffer_handle,
-                .renderArea = render_area,
-                .clearValueCount = 0,
-                .pClearValues = nullptr,
-            };
-            cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
-            const VkDescriptorSet descriptor_set =
-                src_stencil_handle != VK_NULL_HANDLE
-                    ? two_textures_descriptor_allocator.Commit()
-                    : one_texture_descriptor_allocator.Commit();
-            if (src_stencil_handle != VK_NULL_HANDLE) {
-                UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_view_handle,
-                                               src_stencil_handle);
-            } else {
-                UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view_handle);
-            }
-            cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
-            cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
-                                      nullptr);
-            const VkViewport viewport{
-                .x = static_cast<float>(render_area.offset.x),
-                .y = static_cast<float>(render_area.offset.y),
-                .width = static_cast<float>(render_area.extent.width),
-                .height = static_cast<float>(render_area.extent.height),
-                .minDepth = 0.0f,
-                .maxDepth = 1.0f,
-            };
-            cmdbuf.SetViewport(0, viewport);
-            cmdbuf.SetScissor(0, render_area);
-            cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, push_constants);
-            cmdbuf.Draw(3, 1, 0, 0);
-            cmdbuf.EndRenderPass();
-            const VkImageMemoryBarrier post_barrier{
-                .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-                .pNext = nullptr,
-                .srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
-                .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT |
-                                 VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT,
-                .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
-                .newLayout = VK_IMAGE_LAYOUT_GENERAL,
-                .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-                .image = dst,
-                .subresourceRange = src_range,
-            };
-            cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
-                                   vk::PIPELINE_STAGE_GRAPHICS_COMPUTE_TRANSFER, 0, post_barrier);
-        });
-        msaa_copy_resources.push_back(MSAACopyResources{
-            .tick = scheduler.CurrentTick(),
-            .src_view = std::move(src_view),
-            .dst_view = std::move(dst_view),
-            .framebuffer = std::move(framebuffer),
-        });
-        if (copy_stencil) {
-            msaa_copy_resources.push_back(MSAACopyResources{
-                .tick = scheduler.CurrentTick(),
-                .src_view = std::move(src_stencil_view),
-                .dst_view = vk::ImageView{},
-                .framebuffer = vk::Framebuffer{},
-            });
-        }
-    }
-}
-
 VkPipeline BlitImageHelper::FindOrEmplaceMSAACopyPipeline(const MSAACopyPipelineKey& key) {
     const auto it = std::ranges::find(msaa_copy_keys, key);
     if (it != msaa_copy_keys.end()) {
@@ -1665,83 +1458,10 @@ VkPipeline BlitImageHelper::FindOrEmplaceMSAACopyPipeline(const MSAACopyPipeline
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache()));
+    }));
     return *msaa_copy_pipelines.back();
 }
 
-VkPipeline BlitImageHelper::FindOrEmplaceMSAACopyDepthPipeline(const MSAACopyPipelineKey& key,
-                                                               bool copy_stencil) {
-    auto& keys = copy_stencil ? msaa_copy_depth_stencil_keys : msaa_copy_depth_keys;
-    auto& pipelines = copy_stencil ? msaa_copy_depth_stencil_pipelines : msaa_copy_depth_pipelines;
-    const auto it = std::ranges::find(keys, key);
-    if (it != keys.end()) {
-        return *pipelines[std::distance(keys.begin(), it)];
-    }
-    keys.push_back(key);
-    const std::array stages =
-        MakeStages(*clear_color_vert, copy_stencil ? *convert_non_msaa_to_msaa_depth_stencil_frag
-                                                   : *convert_non_msaa_to_msaa_depth_frag);
-    const VkPipelineMultisampleStateCreateInfo multisample_ci{
-        .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
-        .pNext = nullptr,
-        .flags = 0,
-        .rasterizationSamples = key.samples,
-        .sampleShadingEnable = VK_TRUE,
-        .minSampleShading = 1.0f,
-        .pSampleMask = nullptr,
-        .alphaToCoverageEnable = VK_FALSE,
-        .alphaToOneEnable = VK_FALSE,
-    };
-    static constexpr VkStencilOpState REPLACE_STENCIL_OP{
-        .failOp = VK_STENCIL_OP_REPLACE,
-        .passOp = VK_STENCIL_OP_REPLACE,
-        .depthFailOp = VK_STENCIL_OP_REPLACE,
-        .compareOp = VK_COMPARE_OP_ALWAYS,
-        .compareMask = 0xFF,
-        .writeMask = 0xFF,
-        .reference = 0,
-    };
-    const VkPipelineDepthStencilStateCreateInfo depth_stencil_ci{
-        .sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
-        .pNext = nullptr,
-        .flags = 0,
-        .depthTestEnable = VK_TRUE,
-        .depthWriteEnable = VK_TRUE,
-        .depthCompareOp = VK_COMPARE_OP_ALWAYS,
-        .depthBoundsTestEnable = VK_FALSE,
-        .stencilTestEnable = copy_stencil ? VK_TRUE : VK_FALSE,
-        .front = copy_stencil ? REPLACE_STENCIL_OP : VkStencilOpState{},
-        .back = copy_stencil ? REPLACE_STENCIL_OP : VkStencilOpState{},
-        .minDepthBounds = 0.0f,
-        .maxDepthBounds = 0.0f,
-    };
-    const VkPipelineInputAssemblyStateCreateInfo input_assembly_ci =
-        GetPipelineInputAssemblyStateCreateInfo(device);
-    pipelines.push_back(device.GetLogical().CreateGraphicsPipeline({
-        .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
-        .pNext = nullptr,
-        .flags = 0,
-        .stageCount = static_cast<u32>(stages.size()),
-        .pStages = stages.data(),
-        .pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
-        .pInputAssemblyState = &input_assembly_ci,
-        .pTessellationState = nullptr,
-        .pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
-        .pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
-        .pMultisampleState = &multisample_ci,
-        .pDepthStencilState = &depth_stencil_ci,
-        .pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
-        .pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
-        .layout = copy_stencil ? *msaa_copy_depth_stencil_pipeline_layout
-                               : *msaa_copy_pipeline_layout,
-        .renderPass = key.renderpass,
-        .subpass = 0,
-        .basePipelineHandle = VK_NULL_HANDLE,
-        .basePipelineIndex = 0,
-    }, device.StaticPipelineCache()));
-    return *pipelines.back();
-}
-
 void BlitImageHelper::ConvertDepthToColorPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass) {
     ConvertPipeline(pipeline, renderpass, false);
 }
@@ -1779,7 +1499,7 @@ void BlitImageHelper::ConvertPipelineEx(vk::Pipeline& pipeline, VkRenderPass ren
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache());
+    });
 }
 
 void BlitImageHelper::ConvertPipelineColorTargetEx(vk::Pipeline& pipeline, VkRenderPass renderpass,
@@ -1822,7 +1542,7 @@ void BlitImageHelper::ConvertPipeline(vk::Pipeline& pipeline, VkRenderPass rende
         .subpass = 0,
         .basePipelineHandle = VK_NULL_HANDLE,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache());
+    });
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/blit_image.h

@@ -116,11 +116,6 @@ public:
                   VideoCore::Surface::PixelFormat src_format, u32 num_samples,
                   std::span<const VideoCommon::ImageCopy> copies, bool msaa_to_non_msaa);
 
-    void CopyMSAADepth(RenderPassCache& render_pass_cache, VkImage dst_image,
-                       VideoCore::Surface::PixelFormat dst_format, VkImage src_image,
-                       VideoCore::Surface::PixelFormat src_format, u32 num_samples,
-                       std::span<const VideoCommon::ImageCopy> copies, bool copy_stencil);
-
 private:
     void Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
                  const ImageView& src_image_view);
@@ -136,9 +131,6 @@ private:
     [[nodiscard]] VkPipeline FindOrEmplaceClearStencilPipeline(
         const BlitDepthStencilPipelineKey& key);
     [[nodiscard]] VkPipeline FindOrEmplaceMSAACopyPipeline(const MSAACopyPipelineKey& key);
-
-    [[nodiscard]] VkPipeline FindOrEmplaceMSAACopyDepthPipeline(const MSAACopyPipelineKey& key,
-                                                                bool copy_stencil);
     [[nodiscard]] VkPipeline FindOrEmplaceBlitColorMSAAPipeline(const BlitMSAAPipelineKey& key);
     [[nodiscard]] VkPipeline FindOrEmplaceResolveDepthStencilPipeline(VkRenderPass renderpass,
                                                                       bool resolve_stencil);
@@ -170,7 +162,6 @@ private:
     vk::PipelineLayout two_textures_pipeline_layout;
     vk::PipelineLayout clear_color_pipeline_layout;
     vk::PipelineLayout msaa_copy_pipeline_layout;
-    vk::PipelineLayout msaa_copy_depth_stencil_pipeline_layout;
     vk::ShaderModule full_screen_vert;
     vk::ShaderModule blit_color_to_color_frag;
     vk::ShaderModule blit_color_msaa_frag;
@@ -189,8 +180,6 @@ private:
     vk::ShaderModule convert_s8d24_to_abgr8_frag;
     vk::ShaderModule convert_msaa_to_non_msaa_frag;
     vk::ShaderModule convert_non_msaa_to_msaa_frag;
-    vk::ShaderModule convert_non_msaa_to_msaa_depth_frag;
-    vk::ShaderModule convert_non_msaa_to_msaa_depth_stencil_frag;
     vk::Sampler linear_sampler;
     vk::Sampler nearest_sampler;
 
@@ -204,10 +193,6 @@ private:
     std::vector<vk::Pipeline> clear_stencil_pipelines;
     std::vector<MSAACopyPipelineKey> msaa_copy_keys;
     std::vector<vk::Pipeline> msaa_copy_pipelines;
-    std::vector<MSAACopyPipelineKey> msaa_copy_depth_keys;
-    std::vector<vk::Pipeline> msaa_copy_depth_pipelines;
-    std::vector<MSAACopyPipelineKey> msaa_copy_depth_stencil_keys;
-    std::vector<vk::Pipeline> msaa_copy_depth_stencil_pipelines;
     std::vector<BlitMSAAPipelineKey> blit_msaa_color_keys;
     std::vector<vk::Pipeline> blit_msaa_color_pipelines;
     std::vector<VkRenderPass> resolve_depth_keys;

src/video_core/renderer_vulkan/present/sgsr.cpp

@@ -1,8 +1,6 @@
 // SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
-#include <algorithm>
-
 #include "common/common_types.h"
 #include "common/div_ceil.h"
 #include "common/settings.h"
@@ -19,7 +17,7 @@
 
 namespace Vulkan {
 
-using PushConstants = std::array<u32, 4 + 2 + 2 + 1>;
+using PushConstants = std::array<u32, 4 + 2 + 1>;
 
 SGSR::SGSR(const Device& device, MemoryAllocator& memory_allocator, size_t image_count, VkExtent2D extent, bool edge_dir)
     : m_memory_allocator{memory_allocator}
@@ -102,28 +100,26 @@ VkImageView SGSR::Draw(const Device& device, Scheduler& scheduler, size_t image_
 
     const f32 input_image_width = f32(input_image_extent.width);
     const f32 input_image_height = f32(input_image_extent.height);
-    const f32 crop_width = (crop_rect.right - crop_rect.left) * input_image_width;
-    const f32 crop_height = (crop_rect.bottom - crop_rect.top) * input_image_height;
-    static constexpr f32 EDGE_SHARPNESS_MAX = 2.0f;
-    const f32 edge_sharpness =
-        EDGE_SHARPNESS_MAX - f32(Settings::values.fsr_sharpening_slider.GetValue()) / 200.0f;
+    const f32 viewport_width = (crop_rect.right - crop_rect.left) * input_image_width;
+    const f32 viewport_height = (crop_rect.bottom - crop_rect.top) * input_image_height;
+    // expected [0, 2]
+    const f32 sharpening = f32(Settings::values.fsr_sharpening_slider.GetValue()) / 100.0f;
 
+    // p = (tex * viewport) / input = [0,n] (normalized texcoords)
+    // p * input = [0,1024], [0,768]
     // layout( push_constant ) uniform constants {
     //     highp vec4 ViewportInfo[1];
     //     highp vec2 ResizeFactor;
-    //     highp vec2 CropOffset;
     //     highp float EdgeSharpness;
     // };
     PushConstants viewport_con{};
-    viewport_con[0] = std::bit_cast<u32>(1.f / input_image_width);
-    viewport_con[1] = std::bit_cast<u32>(1.f / input_image_height);
-    viewport_con[2] = std::bit_cast<u32>(input_image_width);
-    viewport_con[3] = std::bit_cast<u32>(input_image_height);
-    viewport_con[4] = std::bit_cast<u32>(crop_width / input_image_width);
-    viewport_con[5] = std::bit_cast<u32>(crop_height / input_image_height);
-    viewport_con[6] = std::bit_cast<u32>((std::min)(crop_rect.left, crop_rect.right));
-    viewport_con[7] = std::bit_cast<u32>((std::min)(crop_rect.top, crop_rect.bottom));
-    viewport_con[8] = std::bit_cast<u32>(edge_sharpness);
+    viewport_con[0] = std::bit_cast<u32>(std::abs(1.f / viewport_width));
+    viewport_con[1] = std::bit_cast<u32>(std::abs(1.f / viewport_height));
+    viewport_con[2] = std::bit_cast<u32>(std::abs(viewport_width));
+    viewport_con[3] = std::bit_cast<u32>(std::abs(viewport_height));
+    viewport_con[4] = std::bit_cast<u32>(viewport_width / input_image_width);
+    viewport_con[5] = std::bit_cast<u32>(viewport_height / input_image_height);
+    viewport_con[6] = std::bit_cast<u32>(sharpening);
 
     UploadImages(device, scheduler);
     UpdateDescriptorSets(device, source_image_view, image_index);

src/video_core/renderer_vulkan/present/util.cpp

@@ -491,7 +491,7 @@ static vk::Pipeline CreateWrappedPipelineImpl(
         .subpass = 0,
         .basePipelineHandle = 0,
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache());
+    });
 }
 
 vk::Pipeline CreateWrappedPipeline(const Device& device, vk::RenderPass& renderpass,

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -8,6 +8,7 @@
 #include <array>
 #include <cstring>
 #include <span>
+#include <utility>
 #include <vector>
 
 #include "video_core/buffer_cache/buffer_cache_base.h"
@@ -32,6 +33,32 @@ VkBufferCopy MakeBufferCopy(const VideoCommon::BufferCopy& copy) {
     };
 }
 
+constexpr size_t MAX_WINDOW_BARRIER_RANGES = 8;
+
+using WindowRange = std::pair<VkDeviceSize, VkDeviceSize>;
+using WindowRanges = boost::container::small_vector<WindowRange, MAX_WINDOW_BARRIER_RANGES>;
+
+void CoalesceWindowRanges(WindowRanges& ranges) {
+    if (ranges.size() < 2) {
+        return;
+    }
+    std::sort(ranges.begin(), ranges.end());
+    size_t merged = 0;
+    for (size_t index = 1; index < ranges.size(); ++index) {
+        if (ranges[index].first <= ranges[merged].second) {
+            ranges[merged].second = (std::max)(ranges[merged].second, ranges[index].second);
+        } else {
+            ranges[++merged] = ranges[index];
+        }
+    }
+    ranges.resize(merged + 1);
+    if (ranges.size() > MAX_WINDOW_BARRIER_RANGES) {
+        const WindowRange bounding{ranges.front().first, ranges.back().second};
+        ranges.clear();
+        ranges.push_back(bounding);
+    }
+}
+
 VkIndexType IndexTypeFromNumElements(const Device& device, u32 num_elements) {
     if (num_elements <= 0xff && device.IsExtIndexTypeUint8Supported()) {
         return VK_INDEX_TYPE_UINT8_EXT;
@@ -365,6 +392,141 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& m
                                                                      scheduler_, staging_pool_);
 }
 
+void BufferCacheRuntime::TryEnableUnifiedMemory(void* base, size_t size,
+                                                std::span<AHardwareBuffer* const> hardware_buffers,
+                                                size_t hardware_buffer_window,
+                                                size_t hardware_buffer_base) {
+    unified_memory = std::make_unique<HostMemoryImport>(
+        device, base, size, hardware_buffers, hardware_buffer_window, hardware_buffer_base);
+    if (!unified_memory->IsValid()) {
+        unified_memory.reset();
+    }
+}
+
+void BufferCacheRuntime::CopyToUnifiedMemory(
+    size_t window_index, VkBuffer src_buffer,
+    std::span<const VideoCommon::BufferCopy> copies) {
+    if (!unified_memory || src_buffer == VK_NULL_HANDLE || copies.empty() ||
+        window_index >= unified_memory->GetWindowCount() ||
+        unified_memory->GetWindowBuffer(window_index) == VK_NULL_HANDLE) {
+        return;
+    }
+    PendingUnifiedCopy& pending = pending_unified_copies.emplace_back();
+    pending.window = window_index;
+    pending.buffer = src_buffer;
+    pending.copies.resize(copies.size());
+    std::ranges::transform(copies, pending.copies.begin(), MakeBufferCopy);
+}
+
+void BufferCacheRuntime::FlushUnifiedMemoryCopies() {
+    if (pending_unified_copies.empty()) {
+        return;
+    }
+    struct UnifiedCopyCommand {
+        VkBuffer buffer;
+        boost::container::small_vector<VkBufferCopy, 8> copies;
+    };
+
+    std::stable_sort(pending_unified_copies.begin(), pending_unified_copies.end(),
+                     [](const PendingUnifiedCopy& lhs, const PendingUnifiedCopy& rhs) {
+                         return lhs.window < rhs.window;
+                     });
+
+    const bool foreign = unified_memory->NeedsForeignOwnershipTransfer();
+    const u32 queue_family = device.GetGraphicsFamily();
+
+    size_t group_begin = 0;
+    while (group_begin < pending_unified_copies.size()) {
+        const size_t window = pending_unified_copies[group_begin].window;
+        size_t group_end = group_begin;
+        while (group_end < pending_unified_copies.size() &&
+               pending_unified_copies[group_end].window == window) {
+            ++group_end;
+        }
+        const VkBuffer window_buffer = unified_memory->GetWindowBuffer(window);
+
+        WindowRanges ranges;
+        for (size_t index = group_begin; index < group_end; ++index) {
+            for (const VkBufferCopy& copy : pending_unified_copies[index].copies) {
+                ranges.emplace_back(copy.dstOffset, copy.dstOffset + copy.size);
+            }
+        }
+        CoalesceWindowRanges(ranges);
+
+        boost::container::small_vector<VkBufferMemoryBarrier, MAX_WINDOW_BARRIER_RANGES> acquire;
+        boost::container::small_vector<VkBufferMemoryBarrier, MAX_WINDOW_BARRIER_RANGES> release;
+        if (foreign) {
+            for (const WindowRange& range : ranges) {
+                acquire.push_back(VkBufferMemoryBarrier{
+                    .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
+                    .pNext = nullptr,
+                    .srcAccessMask = 0,
+                    .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+                    .srcQueueFamilyIndex = VK_QUEUE_FAMILY_FOREIGN_EXT,
+                    .dstQueueFamilyIndex = queue_family,
+                    .buffer = window_buffer,
+                    .offset = range.first,
+                    .size = range.second - range.first,
+                });
+                release.push_back(VkBufferMemoryBarrier{
+                    .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
+                    .pNext = nullptr,
+                    .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+                    .dstAccessMask = 0,
+                    .srcQueueFamilyIndex = queue_family,
+                    .dstQueueFamilyIndex = VK_QUEUE_FAMILY_FOREIGN_EXT,
+                    .buffer = window_buffer,
+                    .offset = range.first,
+                    .size = range.second - range.first,
+                });
+            }
+        }
+
+        boost::container::small_vector<UnifiedCopyCommand, 4> commands;
+        commands.reserve(group_end - group_begin);
+        for (size_t index = group_begin; index < group_end; ++index) {
+            PendingUnifiedCopy& pending = pending_unified_copies[index];
+            commands.push_back(UnifiedCopyCommand{pending.buffer, std::move(pending.copies)});
+        }
+
+        scheduler.RequestOutsideRenderPassOperationContext();
+        scheduler.Record([window_buffer, acquire = std::move(acquire), release = std::move(release),
+                          commands = std::move(commands)](vk::CommandBuffer cmdbuf) {
+            if (!acquire.empty()) {
+                cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
+                                       VK_PIPELINE_STAGE_TRANSFER_BIT, 0, {},
+                                       VideoCommon::FixSmallVectorADL(acquire), {});
+            }
+            for (const UnifiedCopyCommand& command : commands) {
+                cmdbuf.CopyBuffer(command.buffer, window_buffer,
+                                  VideoCommon::FixSmallVectorADL(command.copies));
+            }
+            if (!release.empty()) {
+                cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
+                                       VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, 0, {},
+                                       VideoCommon::FixSmallVectorADL(release), {});
+            }
+        });
+
+        group_begin = group_end;
+    }
+    pending_unified_copies.clear();
+}
+
+void BufferCacheRuntime::UnifiedMemoryHostBarrier() {
+    static constexpr VkMemoryBarrier HOST_BARRIER{
+        .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
+        .pNext = nullptr,
+        .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+        .dstAccessMask = VK_ACCESS_HOST_READ_BIT,
+    };
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([](vk::CommandBuffer cmdbuf) {
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_HOST_BIT, 0,
+                               HOST_BARRIER);
+    });
+}
+
 StagingBufferRef BufferCacheRuntime::UploadStagingBuffer(size_t size) {
     return staging_pool.Request(size, MemoryUsage::Upload);
 }
@@ -402,6 +564,7 @@ u32 BufferCacheRuntime::GetStorageBufferAlignment() const {
 }
 
 void BufferCacheRuntime::TickFrame(Common::SlotVector<Buffer>& slot_buffers) noexcept {
+    FlushUnifiedMemoryCopies();
     for (auto it = slot_buffers.begin(); it != slot_buffers.end(); it++) {
         if (scheduler.IsFree(it->LastUsageTick())) {
             it->ResetUsageTracking();

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -7,6 +7,10 @@
 #pragma once
 
 #include <limits>
+#include <memory>
+#include <span>
+
+#include <boost/container/small_vector.hpp>
 
 #include "video_core/buffer_cache/buffer_cache_base.h"
 #include "video_core/buffer_cache/memory_tracker_base.h"
@@ -97,6 +101,33 @@ public:
 
     void TickFrame(Common::SlotVector<Buffer>& slot_buffers) noexcept;
 
+    void TryEnableUnifiedMemory(void* base, size_t size,
+                                std::span<AHardwareBuffer* const> hardware_buffers,
+                                size_t hardware_buffer_window, size_t hardware_buffer_base);
+
+    [[nodiscard]] bool HasUnifiedMemory() const noexcept {
+        return unified_memory != nullptr && unified_memory->IsValid();
+    }
+
+    [[nodiscard]] u64 UnifiedMemorySize() const noexcept {
+        return unified_memory ? unified_memory->GetSize() : 0;
+    }
+
+    [[nodiscard]] u64 UnifiedMemoryBase() const noexcept {
+        return unified_memory ? unified_memory->GetBaseOffset() : 0;
+    }
+
+    [[nodiscard]] u64 UnifiedMemoryWindowSize() const noexcept {
+        return unified_memory ? unified_memory->GetWindowSize() : 0;
+    }
+
+    void CopyToUnifiedMemory(size_t window_index, VkBuffer src_buffer,
+                             std::span<const VideoCommon::BufferCopy> copies);
+
+    void FlushUnifiedMemoryCopies();
+
+    void UnifiedMemoryHostBarrier();
+
     u64 CurrentTick();
 
     u64 KnownGpuTick();
@@ -180,6 +211,12 @@ public:
     }
 
 private:
+    struct PendingUnifiedCopy {
+        size_t window;
+        VkBuffer buffer;
+        boost::container::small_vector<VkBufferCopy, 8> copies;
+    };
+
     void BindBuffer(const Buffer& buffer, u32 offset, u32 size) {
         const VkBuffer handle = buffer.Handle();
         if (handle == VK_NULL_HANDLE) {
@@ -204,6 +241,8 @@ private:
     std::shared_ptr<QuadStripIndexBuffer> quad_strip_index_buffer;
 
     vk::Buffer null_buffer;
+    std::unique_ptr<HostMemoryImport> unified_memory;
+    boost::container::small_vector<PendingUnifiedCopy, 8> pending_unified_copies;
 
     std::unique_ptr<Uint8Pass> uint8_pass;
     QuadIndexedPass quad_index_pass;
@@ -226,6 +265,7 @@ struct BufferCacheParams {
     static constexpr bool USE_MEMORY_MAPS = true;
     static constexpr bool SEPARATE_IMAGE_BUFFER_BINDINGS = false;
     static constexpr bool USE_MEMORY_MAPS_FOR_UPLOADS = true;
+    static constexpr bool USE_UNIFIED_MEMORY = true;
 };
 
 using BufferCache = VideoCommon::BufferCache<BufferCacheParams>;

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -22,9 +22,7 @@
 #include "video_core/host_shaders/resolve_conditional_render_comp_spv.h"
 #include "video_core/host_shaders/vulkan_quad_indexed_comp_spv.h"
 #include "video_core/host_shaders/vulkan_uint8_comp_spv.h"
-#include "video_core/host_shaders/block_linear_unswizzle_2d_buffer_comp_spv.h"
 #include "video_core/host_shaders/block_linear_unswizzle_3d_bcn_comp_spv.h"
-#include "video_core/host_shaders/block_linear_unswizzle_3d_buffer_comp_spv.h"
 #include "video_core/renderer_vulkan/vk_compute_pass.h"
 #include "video_core/surface.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
@@ -270,7 +268,7 @@ ComputePass::ComputePass(const Device& device_, Scheduler& scheduler, Descriptor
         .layout = *layout,
         .basePipelineHandle = {},
         .basePipelineIndex = 0,
-    }, device.StaticPipelineCache());
+    });
 }
 
 ComputePass::~ComputePass() = default;
@@ -874,479 +872,4 @@ void BlockLinearUnswizzle3DPass::UnswizzleChunk(
     });
 }
 
-namespace {
-constexpr u32 BL2D_BINDING_INPUT_BUFFER = 0;
-constexpr u32 BL2D_BINDING_OUTPUT_BUFFER = 1;
-
-struct alignas(16) BlockLinearUnswizzle2DPushConstants {
-    std::array<u32, 3> dim;
-    u32 bytes_per_block_log2;
-    std::array<u32, 3> origin;
-    u32 layer_stride;
-    u32 block_size;
-    u32 x_shift;
-    u32 block_height;
-    u32 block_height_mask;
-};
-static_assert(sizeof(BlockLinearUnswizzle2DPushConstants) <= 128);
-
-constexpr std::array<VkDescriptorSetLayoutBinding, 2> BL2D_BINDINGS{{
-    {
-        .binding = BL2D_BINDING_INPUT_BUFFER,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .descriptorCount = 1,
-        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
-        .pImmutableSamplers = nullptr,
-    },
-    {
-        .binding = BL2D_BINDING_OUTPUT_BUFFER,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .descriptorCount = 1,
-        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
-        .pImmutableSamplers = nullptr,
-    },
-}};
-
-constexpr std::array<VkDescriptorUpdateTemplateEntry, 2> BL2D_TEMPLATE{{
-    {
-        .dstBinding = BL2D_BINDING_INPUT_BUFFER,
-        .dstArrayElement = 0,
-        .descriptorCount = 1,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .offset = BL2D_BINDING_INPUT_BUFFER * sizeof(DescriptorUpdateEntry),
-        .stride = sizeof(DescriptorUpdateEntry),
-    },
-    {
-        .dstBinding = BL2D_BINDING_OUTPUT_BUFFER,
-        .dstArrayElement = 0,
-        .descriptorCount = 1,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .offset = BL2D_BINDING_OUTPUT_BUFFER * sizeof(DescriptorUpdateEntry),
-        .stride = sizeof(DescriptorUpdateEntry),
-    },
-}};
-
-constexpr DescriptorBankInfo BL2D_BANK_INFO{
-    .uniform_buffers = 0,
-    .storage_buffers = 2,
-    .texture_buffers = 0,
-    .image_buffers = 0,
-    .textures = 0,
-    .images = 0,
-    .score = 2,
-};
-} // Anonymous namespace
-
-BlockLinearUnswizzle2DPass::BlockLinearUnswizzle2DPass(
-    const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
-    StagingBufferPool& staging_buffer_pool_,
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
-    : ComputePass(device_, scheduler_, descriptor_pool_, BL2D_BINDINGS, BL2D_TEMPLATE,
-                  BL2D_BANK_INFO,
-                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(BlockLinearUnswizzle2DPushConstants)>,
-                  BLOCK_LINEAR_UNSWIZZLE_2D_BUFFER_COMP_SPV),
-      scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
-      compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
-
-BlockLinearUnswizzle2DPass::~BlockLinearUnswizzle2DPass() = default;
-
-bool BlockLinearUnswizzle2DPass::IsSupported(const VideoCommon::ImageInfo& info) {
-    if (info.type != VideoCommon::ImageType::e2D) {
-        return false;
-    }
-    if (info.resources.levels != 1 || info.resources.layers != 1) {
-        return false;
-    }
-    if (info.num_samples > 1) {
-        return false;
-    }
-    if (VideoCore::Surface::IsPixelFormatASTC(info.format) ||
-        VideoCore::Surface::IsPixelFormatBCn(info.format)) {
-        return false;
-    }
-    const u32 bytes_per_block = VideoCore::Surface::BytesPerBlock(info.format);
-    if (bytes_per_block != 4 && bytes_per_block != 8 && bytes_per_block != 16) {
-        return false;
-    }
-    return VideoCore::Surface::DefaultBlockWidth(info.format) == 1 &&
-           VideoCore::Surface::DefaultBlockHeight(info.format) == 1;
-}
-
-void BlockLinearUnswizzle2DPass::Unswizzle(
-    Image& image, const StagingBufferRef& swizzled,
-    std::span<const VideoCommon::SwizzleParameters> swizzles) {
-    if (swizzles.empty()) {
-        return;
-    }
-    const VideoCommon::SwizzleParameters& sw = swizzles.front();
-    const auto params = VideoCommon::Accelerated::MakeBlockLinearSwizzle2DParams(sw, image.info);
-
-    const u32 width = sw.num_tiles.width;
-    const u32 height = sw.num_tiles.height;
-    const u32 depth = image.info.resources.layers;
-    const u32 bytes_per_block = 1u << params.bytes_per_block_log2;
-    const VkDeviceSize output_size =
-        static_cast<VkDeviceSize>(width) * height * depth * bytes_per_block;
-
-    const StagingBufferRef output =
-        staging_buffer_pool.Request(static_cast<size_t>(output_size), MemoryUsage::DeviceLocal);
-
-    BlockLinearUnswizzle2DPushConstants pc{};
-    pc.dim = {width, height, depth};
-    pc.bytes_per_block_log2 = params.bytes_per_block_log2;
-    pc.origin = params.origin;
-    pc.layer_stride = params.layer_stride;
-    pc.block_size = params.block_size;
-    pc.x_shift = params.x_shift;
-    pc.block_height = params.block_height;
-    pc.block_height_mask = params.block_height_mask;
-
-    scheduler.RequestOutsideRenderPassOperationContext();
-
-    compute_pass_descriptor_queue.Acquire(scheduler, 2);
-    compute_pass_descriptor_queue.AddBuffer(swizzled.buffer, sw.buffer_offset + swizzled.offset,
-                                            image.guest_size_bytes - sw.buffer_offset);
-    compute_pass_descriptor_queue.AddBuffer(output.buffer, output.offset, output_size);
-
-    const void* descriptor_data = compute_pass_descriptor_queue.UpdateData();
-    const VkDescriptorSet set = descriptor_allocator.Commit();
-
-    const u32 gx = Common::DivCeil(width, 16u);
-    const u32 gy = Common::DivCeil(height, 8u);
-    const bool is_initialized = image.ExchangeInitialization();
-
-    const VkBuffer out_buffer = output.buffer;
-    const VkDeviceSize out_offset = output.offset;
-    const VkImage dst_image = image.Handle();
-    const VkImageAspectFlags aspect = image.AspectMask();
-
-    scheduler.Record([this, set, descriptor_data, pc, gx, gy, depth, output_size, out_buffer,
-                      out_offset, dst_image, aspect, width, height,
-                      is_initialized](vk::CommandBuffer cmdbuf) {
-        if (dst_image == VK_NULL_HANDLE || out_buffer == VK_NULL_HANDLE) {
-            return;
-        }
-        device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
-        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
-        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
-        cmdbuf.PushConstants(*layout, VK_SHADER_STAGE_COMPUTE_BIT, 0, sizeof(pc), &pc);
-        cmdbuf.Dispatch(gx, gy, depth);
-
-        const VkBufferMemoryBarrier buffer_barrier{
-            .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
-            .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
-            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .buffer = out_buffer,
-            .offset = out_offset,
-            .size = output_size,
-        };
-        const VkImageMemoryBarrier pre_copy{
-            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = static_cast<VkAccessFlags>(is_initialized ? VK_ACCESS_SHADER_READ_BIT
-                                                                      : VK_ACCESS_NONE),
-            .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
-            .oldLayout = is_initialized ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_UNDEFINED,
-            .newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
-            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .image = dst_image,
-            .subresourceRange{
-                .aspectMask = aspect,
-                .baseMipLevel = 0,
-                .levelCount = VK_REMAINING_MIP_LEVELS,
-                .baseArrayLayer = 0,
-                .layerCount = VK_REMAINING_ARRAY_LAYERS,
-            },
-        };
-        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
-                                   (is_initialized ? vk::PIPELINE_STAGE_GRAPHICS_COMPUTE
-                                                   : VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT),
-                               VK_PIPELINE_STAGE_TRANSFER_BIT, 0, {}, buffer_barrier, pre_copy);
-
-        const VkBufferImageCopy copy{
-            .bufferOffset = out_offset,
-            .bufferRowLength = 0,
-            .bufferImageHeight = 0,
-            .imageSubresource{
-                .aspectMask = aspect,
-                .mipLevel = 0,
-                .baseArrayLayer = 0,
-                .layerCount = depth,
-            },
-            .imageOffset = {0, 0, 0},
-            .imageExtent = {width, height, 1},
-        };
-        cmdbuf.CopyBufferToImage(out_buffer, dst_image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy);
-
-        const VkImageMemoryBarrier post_copy{
-            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
-            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
-            .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
-            .newLayout = VK_IMAGE_LAYOUT_GENERAL,
-            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .image = dst_image,
-            .subresourceRange{
-                .aspectMask = aspect,
-                .baseMipLevel = 0,
-                .levelCount = VK_REMAINING_MIP_LEVELS,
-                .baseArrayLayer = 0,
-                .layerCount = VK_REMAINING_ARRAY_LAYERS,
-            },
-        };
-        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE,
-                               0, {}, {}, post_copy);
-    });
-}
-
-namespace {
-constexpr u32 BL3DB_BINDING_INPUT_BUFFER = 0;
-constexpr u32 BL3DB_BINDING_OUTPUT_BUFFER = 1;
-
-struct alignas(16) BlockLinearUnswizzle3DBufferPushConstants {
-    std::array<u32, 3> dim;
-    u32 bytes_per_block_log2;
-    std::array<u32, 3> origin;
-    u32 slice_size;
-    u32 block_size;
-    u32 x_shift;
-    u32 block_height;
-    u32 block_height_mask;
-    u32 block_depth;
-    u32 block_depth_mask;
-};
-static_assert(sizeof(BlockLinearUnswizzle3DBufferPushConstants) <= 128);
-
-constexpr std::array<VkDescriptorSetLayoutBinding, 2> BL3DB_BINDINGS{{
-    {
-        .binding = BL3DB_BINDING_INPUT_BUFFER,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .descriptorCount = 1,
-        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
-        .pImmutableSamplers = nullptr,
-    },
-    {
-        .binding = BL3DB_BINDING_OUTPUT_BUFFER,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .descriptorCount = 1,
-        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
-        .pImmutableSamplers = nullptr,
-    },
-}};
-
-constexpr std::array<VkDescriptorUpdateTemplateEntry, 2> BL3DB_TEMPLATE{{
-    {
-        .dstBinding = BL3DB_BINDING_INPUT_BUFFER,
-        .dstArrayElement = 0,
-        .descriptorCount = 1,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .offset = BL3DB_BINDING_INPUT_BUFFER * sizeof(DescriptorUpdateEntry),
-        .stride = sizeof(DescriptorUpdateEntry),
-    },
-    {
-        .dstBinding = BL3DB_BINDING_OUTPUT_BUFFER,
-        .dstArrayElement = 0,
-        .descriptorCount = 1,
-        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
-        .offset = BL3DB_BINDING_OUTPUT_BUFFER * sizeof(DescriptorUpdateEntry),
-        .stride = sizeof(DescriptorUpdateEntry),
-    },
-}};
-
-constexpr DescriptorBankInfo BL3DB_BANK_INFO{
-    .uniform_buffers = 0,
-    .storage_buffers = 2,
-    .texture_buffers = 0,
-    .image_buffers = 0,
-    .textures = 0,
-    .images = 0,
-    .score = 2,
-};
-} // Anonymous namespace
-
-BlockLinearUnswizzle3DBufferPass::BlockLinearUnswizzle3DBufferPass(
-    const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
-    StagingBufferPool& staging_buffer_pool_,
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue_)
-    : ComputePass(device_, scheduler_, descriptor_pool_, BL3DB_BINDINGS, BL3DB_TEMPLATE,
-                  BL3DB_BANK_INFO,
-                  COMPUTE_PUSH_CONSTANT_RANGE<sizeof(BlockLinearUnswizzle3DBufferPushConstants)>,
-                  BLOCK_LINEAR_UNSWIZZLE_3D_BUFFER_COMP_SPV),
-      scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
-      compute_pass_descriptor_queue{compute_pass_descriptor_queue_} {}
-
-BlockLinearUnswizzle3DBufferPass::~BlockLinearUnswizzle3DBufferPass() = default;
-
-bool BlockLinearUnswizzle3DBufferPass::IsSupported(const Device& device,
-                                                   const VideoCommon::ImageInfo& info) {
-    if (info.type != VideoCommon::ImageType::e3D) {
-        return false;
-    }
-    if (info.resources.levels != 1 || info.resources.layers != 1) {
-        return false;
-    }
-    if (info.num_samples > 1) {
-        return false;
-    }
-    if (info.size.depth <= 1) {
-        return false;
-    }
-    if (VideoCore::Surface::IsPixelFormatASTC(info.format)) {
-        return false;
-    }
-    if (VideoCore::Surface::IsPixelFormatBCn(info.format) && !device.IsOptimalBcnSupported()) {
-        return false;
-    }
-    const u32 bytes_per_block = VideoCore::Surface::BytesPerBlock(info.format);
-    return bytes_per_block == 4 || bytes_per_block == 8 || bytes_per_block == 16;
-}
-
-void BlockLinearUnswizzle3DBufferPass::Unswizzle(
-    Image& image, const StagingBufferRef& swizzled,
-    std::span<const VideoCommon::SwizzleParameters> swizzles) {
-    if (swizzles.empty()) {
-        return;
-    }
-
-    const VideoCommon::SwizzleParameters& sw = swizzles.front();
-    const auto params = VideoCommon::Accelerated::MakeBlockLinearSwizzle3DParams(sw, image.info);
-
-    const u32 blocks_x = sw.num_tiles.width;
-    const u32 blocks_y = sw.num_tiles.height;
-    const u32 blocks_z = sw.num_tiles.depth;
-    const u32 bytes_per_block = 1u << params.bytes_per_block_log2;
-    const VkDeviceSize output_size =
-        static_cast<VkDeviceSize>(blocks_x) * blocks_y * blocks_z * bytes_per_block;
-
-    const StagingBufferRef output =
-        staging_buffer_pool.Request(static_cast<size_t>(output_size), MemoryUsage::DeviceLocal);
-
-    BlockLinearUnswizzle3DBufferPushConstants pc{};
-    pc.dim = {blocks_x, blocks_y, blocks_z};
-    pc.bytes_per_block_log2 = params.bytes_per_block_log2;
-    pc.origin = params.origin;
-    pc.slice_size = params.slice_size;
-    pc.block_size = params.block_size;
-    pc.x_shift = params.x_shift;
-    pc.block_height = params.block_height;
-    pc.block_height_mask = params.block_height_mask;
-    pc.block_depth = params.block_depth;
-    pc.block_depth_mask = params.block_depth_mask;
-
-    scheduler.RequestOutsideRenderPassOperationContext();
-
-    compute_pass_descriptor_queue.Acquire(scheduler, 2);
-    compute_pass_descriptor_queue.AddBuffer(swizzled.buffer, sw.buffer_offset + swizzled.offset,
-                                            image.guest_size_bytes - sw.buffer_offset);
-    compute_pass_descriptor_queue.AddBuffer(output.buffer, output.offset, output_size);
-
-    const void* descriptor_data = compute_pass_descriptor_queue.UpdateData();
-    const VkDescriptorSet set = descriptor_allocator.Commit();
-
-    const u32 gx = Common::DivCeil(blocks_x, 8u);
-    const u32 gy = Common::DivCeil(blocks_y, 8u);
-    const u32 gz = Common::DivCeil(blocks_z, 4u);
-    const bool is_initialized = image.ExchangeInitialization();
-
-    const VkBuffer out_buffer = output.buffer;
-    const VkDeviceSize out_offset = output.offset;
-    const VkImage dst_image = image.Handle();
-    const VkImageAspectFlags aspect = image.AspectMask();
-    const VkExtent3D extent{
-        .width = image.info.size.width,
-        .height = image.info.size.height,
-        .depth = image.info.size.depth,
-    };
-
-    scheduler.Record([this, set, descriptor_data, pc, gx, gy, gz, output_size, out_buffer,
-                      out_offset, dst_image, aspect, extent,
-                      is_initialized](vk::CommandBuffer cmdbuf) {
-        if (dst_image == VK_NULL_HANDLE || out_buffer == VK_NULL_HANDLE) {
-            return;
-        }
-        device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
-        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
-        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *layout, 0, set, {});
-        cmdbuf.PushConstants(*layout, VK_SHADER_STAGE_COMPUTE_BIT, 0, sizeof(pc), &pc);
-        cmdbuf.Dispatch(gx, gy, gz);
-
-        const VkBufferMemoryBarrier buffer_barrier{
-            .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
-            .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
-            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .buffer = out_buffer,
-            .offset = out_offset,
-            .size = output_size,
-        };
-        const VkImageMemoryBarrier pre_copy{
-            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = static_cast<VkAccessFlags>(is_initialized ? VK_ACCESS_SHADER_READ_BIT
-                                                                      : VK_ACCESS_NONE),
-            .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
-            .oldLayout = is_initialized ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_UNDEFINED,
-            .newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
-            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .image = dst_image,
-            .subresourceRange{
-                .aspectMask = aspect,
-                .baseMipLevel = 0,
-                .levelCount = VK_REMAINING_MIP_LEVELS,
-                .baseArrayLayer = 0,
-                .layerCount = VK_REMAINING_ARRAY_LAYERS,
-            },
-        };
-        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
-                                   (is_initialized ? vk::PIPELINE_STAGE_GRAPHICS_COMPUTE
-                                                   : VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT),
-                               VK_PIPELINE_STAGE_TRANSFER_BIT, 0, {}, buffer_barrier, pre_copy);
-
-        const VkBufferImageCopy copy{
-            .bufferOffset = out_offset,
-            .bufferRowLength = 0,
-            .bufferImageHeight = 0,
-            .imageSubresource{
-                .aspectMask = aspect,
-                .mipLevel = 0,
-                .baseArrayLayer = 0,
-                .layerCount = 1,
-            },
-            .imageOffset = {0, 0, 0},
-            .imageExtent = extent,
-        };
-        cmdbuf.CopyBufferToImage(out_buffer, dst_image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy);
-
-        const VkImageMemoryBarrier post_copy{
-            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
-            .pNext = nullptr,
-            .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
-            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
-            .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
-            .newLayout = VK_IMAGE_LAYOUT_GENERAL,
-            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-            .image = dst_image,
-            .subresourceRange{
-                .aspectMask = aspect,
-                .baseMipLevel = 0,
-                .levelCount = VK_REMAINING_MIP_LEVELS,
-                .baseArrayLayer = 0,
-                .layerCount = VK_REMAINING_ARRAY_LAYERS,
-            },
-        };
-        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE,
-                               0, {}, {}, post_copy);
-    });
-}
-
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pass.h

@@ -164,42 +164,4 @@ private:
     ComputePassDescriptorQueue& compute_pass_descriptor_queue;
 };
 
-class BlockLinearUnswizzle2DPass final : public ComputePass {
-public:
-    explicit BlockLinearUnswizzle2DPass(const Device& device_, Scheduler& scheduler_,
-                                        DescriptorPool& descriptor_pool_,
-                                        StagingBufferPool& staging_buffer_pool_,
-                                        ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
-    ~BlockLinearUnswizzle2DPass();
-
-    [[nodiscard]] static bool IsSupported(const VideoCommon::ImageInfo& info);
-
-    void Unswizzle(Image& image, const StagingBufferRef& swizzled,
-                   std::span<const VideoCommon::SwizzleParameters> swizzles);
-
-private:
-    Scheduler& scheduler;
-    StagingBufferPool& staging_buffer_pool;
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
-};
-
-class BlockLinearUnswizzle3DBufferPass final : public ComputePass {
-public:
-    explicit BlockLinearUnswizzle3DBufferPass(
-        const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
-        StagingBufferPool& staging_buffer_pool_,
-        ComputePassDescriptorQueue& compute_pass_descriptor_queue_);
-    ~BlockLinearUnswizzle3DBufferPass();
-
-    [[nodiscard]] static bool IsSupported(const Device& device, const VideoCommon::ImageInfo& info);
-
-    void Unswizzle(Image& image, const StagingBufferRef& swizzled,
-                   std::span<const VideoCommon::SwizzleParameters> swizzles);
-
-private:
-    Scheduler& scheduler;
-    StagingBufferPool& staging_buffer_pool;
-    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
-};
-
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -140,8 +140,6 @@ RenderPassKey MakeRenderPassKey(const FixedPipelineState& state, const Device& d
     });
     key.resolve_color =
         key.samples != VK_SAMPLE_COUNT_1_BIT && has_color && device.IsTiler();
-    key.resolve_depth_stencil = key.samples != VK_SAMPLE_COUNT_1_BIT && device.IsTiler() &&
-                                SupportsDepthStencilResolve(device, key.depth_format);
     return key;
 }
 

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -63,8 +63,6 @@ using VideoCommon::GenericEnvironment;
 using VideoCommon::GraphicsEnvironment;
 
 constexpr u32 CACHE_VERSION = 18;
-constexpr size_t VULKAN_CACHE_FLUSH_PIPELINES = 128;
-constexpr size_t VULKAN_CACHE_FLUSH_MIN_SECONDS = 30;
 constexpr std::array<char, 8> VULKAN_CACHE_MAGIC_NUMBER{'y', 'u', 'z', 'u', 'v', 'k', 'c', 'h'};
 
 template <typename Container>
@@ -152,25 +150,6 @@ Shader::AttributeType AttributeType(const FixedPipelineState& state, size_t inde
     return Shader::AttributeType::Disabled;
 }
 
-Shader::InputTopology MaxwellToInputTopology(Maxwell::PrimitiveTopology topology) {
-    switch (topology) {
-    case Maxwell::PrimitiveTopology::Points:
-        return Shader::InputTopology::Points;
-    case Maxwell::PrimitiveTopology::Lines:
-    case Maxwell::PrimitiveTopology::LineLoop:
-    case Maxwell::PrimitiveTopology::LineStrip:
-        return Shader::InputTopology::Lines;
-    case Maxwell::PrimitiveTopology::LinesAdjacency:
-    case Maxwell::PrimitiveTopology::LineStripAdjacency:
-        return Shader::InputTopology::LinesAdjacency;
-    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
-    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
-        return Shader::InputTopology::TrianglesAdjacency;
-    default:
-        return Shader::InputTopology::Triangles;
-    }
-}
-
 Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> programs,
                                     const GraphicsPipelineCacheKey& key,
                                     const Shader::IR::Program& program,
@@ -289,7 +268,33 @@ Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> program
     default:
         break;
     }
-    info.input_topology = MaxwellToInputTopology(key.state.topology);
+    switch (key.state.topology) {
+    case Maxwell::PrimitiveTopology::Points:
+        info.input_topology = Shader::InputTopology::Points;
+        break;
+    case Maxwell::PrimitiveTopology::Lines:
+    case Maxwell::PrimitiveTopology::LineLoop:
+    case Maxwell::PrimitiveTopology::LineStrip:
+        info.input_topology = Shader::InputTopology::Lines;
+        break;
+    case Maxwell::PrimitiveTopology::Triangles:
+    case Maxwell::PrimitiveTopology::TriangleStrip:
+    case Maxwell::PrimitiveTopology::TriangleFan:
+    case Maxwell::PrimitiveTopology::Quads:
+    case Maxwell::PrimitiveTopology::QuadStrip:
+    case Maxwell::PrimitiveTopology::Polygon:
+    case Maxwell::PrimitiveTopology::Patches:
+        info.input_topology = Shader::InputTopology::Triangles;
+        break;
+    case Maxwell::PrimitiveTopology::LinesAdjacency:
+    case Maxwell::PrimitiveTopology::LineStripAdjacency:
+        info.input_topology = Shader::InputTopology::LinesAdjacency;
+        break;
+    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
+    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
+        info.input_topology = Shader::InputTopology::TrianglesAdjacency;
+        break;
+    }
     info.force_early_z = key.state.early_z != 0;
     info.y_negate = key.state.y_negate != 0;
     return info;
@@ -441,6 +446,7 @@ PipelineCache::PipelineCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
         .has_broken_fp16_float_controls = driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY,
         .ignore_nan_fp_comparisons = false,
         .has_broken_spirv_subgroup_mask_vector_extract_dynamic = false,
+        .max_shared_memory_size = device.GetMaxComputeSharedMemorySize(),
         .has_broken_robust =
             device.IsNvidia() && device.GetNvidiaArch() <= NvidiaArchitecture::Arch_Pascal,
         .min_ssbo_alignment = device.GetStorageBufferAlignment(),
@@ -694,10 +700,6 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
     if (use_vulkan_pipeline_cache) {
         SerializeVulkanPipelineCache(vulkan_pipeline_cache_filename, vulkan_pipeline_cache,
                                      CACHE_VERSION);
-        size_t size = 0;
-        vulkan_pipeline_cache.Read(&size, nullptr);
-        last_cache_size.store(size, std::memory_order_relaxed);
-        last_flush = std::chrono::steady_clock::now();
     }
 
     if (state.statistics) {
@@ -705,35 +707,6 @@ void PipelineCache::LoadDiskResources(u64 title_id, std::stop_token stop_loading
     }
 }
 
-void PipelineCache::QueueVulkanPipelineCacheFlush() {
-    if (!use_vulkan_pipeline_cache || vulkan_pipeline_cache_filename.empty()) {
-        return;
-    }
-    if (++pipelines_since_flush < VULKAN_CACHE_FLUSH_PIPELINES) {
-        return;
-    }
-    const auto now = std::chrono::steady_clock::now();
-    const auto megabytes = last_cache_size.load(std::memory_order_relaxed) / (1024 * 1024);
-    const std::chrono::seconds interval{
-        std::max<size_t>(VULKAN_CACHE_FLUSH_MIN_SECONDS, megabytes)};
-    if (last_flush.time_since_epoch().count() != 0 && now - last_flush < interval) {
-        return;
-    }
-    if (flush_in_flight.exchange(true, std::memory_order_acq_rel)) {
-        return;
-    }
-    pipelines_since_flush = 0;
-    last_flush = now;
-    serialization_thread.QueueWork([this] {
-        SerializeVulkanPipelineCache(vulkan_pipeline_cache_filename, vulkan_pipeline_cache,
-                                     CACHE_VERSION);
-        size_t size = 0;
-        vulkan_pipeline_cache.Read(&size, nullptr);
-        last_cache_size.store(size, std::memory_order_relaxed);
-        flush_in_flight.store(false, std::memory_order_release);
-    });
-}
-
 GraphicsPipeline* PipelineCache::CurrentGraphicsPipelineSlowPath() {
     const auto [pair, is_new]{graphics_cache.try_emplace(graphics_key)};
     auto& pipeline{pair->second};
@@ -772,7 +745,7 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
     std::span<Shader::Environment* const> envs, PipelineStatistics* statistics,
     bool build_in_parallel) try {
     auto hash = key.Hash();
-    LOG_DEBUG(Render_Vulkan, "{:#016x}", hash);
+    LOG_INFO(Render_Vulkan, "{:#016x}", hash);
     size_t env_index{0};
     std::array<Shader::IR::Program, Maxwell::MaxShaderProgram> programs;
     const bool uses_vertex_a{key.unique_hashes[0] != 0};
@@ -786,10 +759,8 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
                                        index == static_cast<u32>(Maxwell::ShaderType::Geometry);
         if (key.unique_hashes[index] == 0 && is_emulated_stage) {
             auto topology = MaxwellToOutputTopology(key.state.topology);
-            programs[index] =
-                GenerateGeometryPassthrough(pools.inst, pools.block, host_info,
-                                            *layer_source_program, topology,
-                                            MaxwellToInputTopology(key.state.topology));
+            programs[index] = GenerateGeometryPassthrough(pools.inst, pools.block, host_info,
+                                                          *layer_source_program, topology);
             continue;
         }
         if (key.unique_hashes[index] == 0) {
@@ -802,13 +773,11 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline(
         Shader::Maxwell::Flow::CFG cfg(env, pools.flow_block, cfg_offset, index == 0);
         if (!uses_vertex_a || index != 1) {
             // Normal path
-            programs[index] = TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
-                                              MaxwellToInputTopology(key.state.topology));
+            programs[index] = TranslateProgram(pools.inst, pools.block, env, cfg, host_info);
         } else {
             // VertexB path when VertexA is present.
             auto& program_va{programs[0]};
-            auto program_vb{TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
-                                             MaxwellToInputTopology(key.state.topology))};
+            auto program_vb{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
             programs[index] = MergeDualVertexPrograms(program_va, program_vb, env);
         }
 
@@ -912,7 +881,6 @@ std::unique_ptr<GraphicsPipeline> PipelineCache::CreateGraphicsPipeline() {
         }
         SerializePipeline(key, env_ptrs, pipeline_cache_filename, CACHE_VERSION);
     });
-    QueueVulkanPipelineCacheFlush();
     return pipeline;
 }
 
@@ -932,7 +900,6 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
         SerializePipeline(key, std::array<const GenericEnvironment*, 1>{&env_},
                           pipeline_cache_filename, CACHE_VERSION);
     });
-    QueueVulkanPipelineCacheFlush();
     return pipeline;
 }
 
@@ -945,7 +912,7 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
         return nullptr;
     }
 
-    LOG_DEBUG(Render_Vulkan, "{:#016x}", hash);
+    LOG_INFO(Render_Vulkan, "{:#016x}", hash);
 
     Shader::Maxwell::Flow::CFG cfg{env, pools.flow_block, env.StartAddress()};
 
@@ -954,21 +921,7 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
         env.Dump(hash, key.unique_hash);
     }
 
-    auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info,
-                                  Shader::InputTopology::Points)};
-    const VkDriverIdKHR driver_id = device.GetDriverID();
-    const bool needs_shared_mem_clamp =
-        driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY ||
-        driver_id == VK_DRIVER_ID_ARM_PROPRIETARY;
-    const u32 max_shared_memory = device.GetMaxComputeSharedMemorySize();
-    if (needs_shared_mem_clamp && program.shared_memory_size > max_shared_memory) {
-        LOG_WARNING(Render_Vulkan,
-                    "Compute shader {:#016x} requests {}KB shared memory but device max is {}KB - clamping",
-                    key.unique_hash,
-                    program.shared_memory_size / 1024,
-                    max_shared_memory / 1024);
-        program.shared_memory_size = max_shared_memory;
-    }
+    auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
     const std::vector<u32> code{EmitSPIRV(profile, program)};
     device.SaveShader(code);
     vk::ShaderModule spv_module{BuildShader(device, code)};

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -7,8 +7,6 @@
 #pragma once
 
 #include <array>
-#include <atomic>
-#include <chrono>
 #include <cstddef>
 #include <filesystem>
 #include <memory>
@@ -146,8 +144,6 @@ private:
     vk::PipelineCache LoadVulkanPipelineCache(const std::filesystem::path& filename,
                                               u32 expected_cache_version);
 
-    void QueueVulkanPipelineCacheFlush();
-
     const Device& device;
     Scheduler& scheduler;
     DescriptorPool& descriptor_pool;
@@ -175,10 +171,6 @@ private:
 
     std::filesystem::path vulkan_pipeline_cache_filename;
     vk::PipelineCache vulkan_pipeline_cache;
-    size_t pipelines_since_flush{};
-    std::chrono::steady_clock::time_point last_flush{};
-    std::atomic<size_t> last_cache_size{};
-    std::atomic_bool flush_in_flight{};
 
     Common::ThreadWorker workers;
     Common::ThreadWorker serialization_thread;

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -725,9 +725,6 @@ public:
         if (!device.IsExtTransformFeedbackSupported()) {
             return;
         }
-        if (!scheduler.IsRenderPassActive()) {
-            return;
-        }
         FlushBeginTFB();
         has_started = true;
     }
@@ -744,9 +741,6 @@ public:
         if (has_flushed_end_pending) {
             if (scheduler.IsRenderPassActive()) {
                 FlushEndTFB();
-            } else {
-                has_flushed_end_pending = false;
-                has_started = false;
             }
         }
         runtime.View3DRegs([this](Maxwell3D& maxwell3d) {

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -225,6 +225,13 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
       fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache, device, scheduler),
       wfi_event(device.GetLogical().CreateEvent()) {
     scheduler.SetQueryCache(query_cache);
+    if (Settings::values.use_unified_memory.GetValue() && device_memory.IsBackingShared()) {
+        buffer_cache_runtime.TryEnableUnifiedMemory(
+            device_memory.GetPhysicalBase(), device_memory.GetPhysicalSize(),
+            device_memory.GetBackingHardwareBuffers(),
+            device_memory.GetBackingHardwareBufferWindowSize(),
+            device_memory.GetBackingHardwareBufferBase());
+    }
 }
 
 RasterizerVulkan::~RasterizerVulkan() {
@@ -1286,42 +1293,33 @@ void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) {
                         regs.zeta.format == Tegra::DepthFormat::X8Z24_UNORM ||
                         regs.zeta.format == Tegra::DepthFormat::S8Z24_UNORM ||
                         regs.zeta.format == Tegra::DepthFormat::V8Z24_UNORM;
-    const bool forces_unorm_representation = device.IsExtDepthBiasControlSupported();
 
     if (is_d24 && !device.SupportsD24DepthBuffer()) {
-        static constexpr double GUEST_TO_HOST_UNORM_BITS =
-            static_cast<double>(1ULL << (32 - 24));
+        static constexpr const size_t length = sizeof(NEEDS_D24) / sizeof(NEEDS_D24[0]);
 
-        if (forces_unorm_representation) {
-            units = static_cast<float>(static_cast<double>(units) * GUEST_TO_HOST_UNORM_BITS);
-        } else {
-            static constexpr const size_t length = sizeof(NEEDS_D24) / sizeof(NEEDS_D24[0]);
+        static constexpr const u64* start = NEEDS_D24;
+        static constexpr const u64* end = NEEDS_D24 + length;
 
-            static constexpr const u64* start = NEEDS_D24;
-            static constexpr const u64* end = NEEDS_D24 + length;
+        const u64* it = std::find(start, end, program_id);
 
-            const u64* it = std::find(start, end, program_id);
-
-            if (it != end) {
-                // the base formulas can be obtained from here:
-                //   https://docs.microsoft.com/en-us/windows/win32/direct3d11/d3d10-graphics-programming-guide-output-merger-stage-depth-bias
-                const double rescale_factor =
-                    GUEST_TO_HOST_UNORM_BITS / (static_cast<double>(0x1.ep+127));
-                units = static_cast<float>(static_cast<double>(units) * rescale_factor);
-            }
+        if (it != end) {
+            // the base formulas can be obtained from here:
+            //   https://docs.microsoft.com/en-us/windows/win32/direct3d11/d3d10-graphics-programming-guide-output-merger-stage-depth-bias
+            const double rescale_factor =
+                static_cast<double>(1ULL << (32 - 24)) / (static_cast<double>(0x1.ep+127));
+            units = static_cast<float>(static_cast<double>(units) * rescale_factor);
         }
     }
 
     scheduler.Record([constant = units, clamp = regs.depth_bias_clamp,
-                      factor = regs.slope_scale_depth_bias,
-                      forces_unorm_representation, this](vk::CommandBuffer cmdbuf) {
-        if (forces_unorm_representation) {
-            const VkDepthBiasRepresentationInfoEXT bias_info{
+                      factor = regs.slope_scale_depth_bias, this](vk::CommandBuffer cmdbuf) {
+        if (device.IsExtDepthBiasControlSupported()) {
+            static VkDepthBiasRepresentationInfoEXT bias_info{
                 .sType = VK_STRUCTURE_TYPE_DEPTH_BIAS_REPRESENTATION_INFO_EXT,
                 .pNext = nullptr,
                 .depthBiasRepresentation =
                     VK_DEPTH_BIAS_REPRESENTATION_LEAST_REPRESENTABLE_VALUE_FORCE_UNORM_EXT,
-                .depthBiasExact = static_cast<VkBool32>(device.HasExactDepthBiasControl()),
+                .depthBiasExact = VK_FALSE,
             };
 
             cmdbuf.SetDepthBias(constant, clamp, factor, &bias_info);
@@ -1345,14 +1343,8 @@ void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs)
     if (!state_tracker.TouchDepthBounds()) {
         return;
     }
-    float min = regs.depth_bounds[0];
-    float max = regs.depth_bounds[1];
-    if (!device.IsExtDepthRangeUnrestrictedSupported()) {
-        min = std::clamp(min, 0.0f, 1.0f);
-        max = std::clamp(max, 0.0f, 1.0f);
-    }
-    scheduler.Record(
-        [min, max](vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBounds(min, max); });
+    scheduler.Record([min = regs.depth_bounds[0], max = regs.depth_bounds[1]](
+                         vk::CommandBuffer cmdbuf) { cmdbuf.SetDepthBounds(min, max); });
 }
 
 void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) {

src/video_core/renderer_vulkan/vk_render_pass_cache.cpp

@@ -65,61 +65,8 @@ using VideoCore::Surface::SurfaceType;
                 .finalLayout = VK_IMAGE_LAYOUT_GENERAL,
             };
         }
-
-        struct ResolveAspects {
-            bool depth;
-            bool stencil;
-        };
-
-        struct ResolveModes {
-            VkResolveModeFlagBits depth;
-            VkResolveModeFlagBits stencil;
-        };
-
-        constexpr ResolveAspects GetResolveAspects(PixelFormat format) {
-            const SurfaceType surface_type = GetSurfaceType(format);
-            return ResolveAspects{
-                .depth = surface_type == SurfaceType::Depth ||
-                         surface_type == SurfaceType::DepthStencil,
-                .stencil = surface_type == SurfaceType::Stencil ||
-                           surface_type == SurfaceType::DepthStencil,
-            };
-        }
-
-        ResolveModes PickResolveModes(const Device& device, PixelFormat format) {
-            constexpr VkResolveModeFlagBits mode = VK_RESOLVE_MODE_SAMPLE_ZERO_BIT;
-
-            const ResolveAspects aspects = GetResolveAspects(format);
-            ResolveModes modes{
-                .depth = VK_RESOLVE_MODE_NONE,
-                .stencil = VK_RESOLVE_MODE_NONE,
-            };
-            if (aspects.depth && (device.GetDepthResolveModes() & mode) != 0) {
-                modes.depth = mode;
-            }
-            if (aspects.stencil && (device.GetStencilResolveModes() & mode) != 0) {
-                modes.stencil = mode;
-            }
-            return modes;
-        }
     } // Anonymous namespace
 
-bool SupportsDepthStencilResolve(const Device& device, PixelFormat depth_format) {
-    if (depth_format == PixelFormat::Invalid || !device.IsKhrDepthStencilResolveSupported()) {
-        return false;
-    }
-    const ResolveAspects aspects = GetResolveAspects(depth_format);
-    if (!aspects.depth && !aspects.stencil) {
-        return false;
-    }
-    const ResolveModes modes = PickResolveModes(device, depth_format);
-    if ((aspects.depth && modes.depth == VK_RESOLVE_MODE_NONE) ||
-        (aspects.stencil && modes.stencil == VK_RESOLVE_MODE_NONE)) {
-        return false;
-    }
-    return modes.depth == modes.stencil || device.SupportsIndependentResolveNone();
-}
-
 RenderPassCache::RenderPassCache(const Device& device_) : device{&device_} {}
 
 VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
@@ -128,9 +75,7 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
     if (!is_new) {
         return *pair->second;
     }
-    static constexpr size_t MAX_ATTACHMENTS =
-        2 * std::tuple_size_v<decltype(RenderPassKey::color_formats)> + 2;
-    boost::container::static_vector<VkAttachmentDescription, MAX_ATTACHMENTS> descriptions;
+    boost::container::static_vector<VkAttachmentDescription, 9> descriptions;
     std::array<VkAttachmentReference, 8> references{};
     u32 num_attachments{};
     u32 num_colors{};
@@ -188,21 +133,6 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
             }
         }
     }
-    const bool do_resolve_depth_stencil = key.resolve_depth_stencil && has_depth &&
-                                          key.samples != VK_SAMPLE_COUNT_1_BIT &&
-                                          SupportsDepthStencilResolve(*device, key.depth_format);
-    VkAttachmentReference depth_resolve_reference{};
-    if (do_resolve_depth_stencil) {
-        depth_resolve_reference = VkAttachmentReference{
-            .attachment = static_cast<u32>(descriptions.size()),
-            .layout = VK_IMAGE_LAYOUT_GENERAL,
-        };
-        VkAttachmentDescription resolve_desc =
-            AttachmentDescription(*device, key.depth_format, VK_SAMPLE_COUNT_1_BIT,
-                                  VK_ATTACHMENT_LOAD_OP_DONT_CARE, VK_ATTACHMENT_STORE_OP_STORE);
-        resolve_desc.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
-        descriptions.push_back(resolve_desc);
-    }
     const VkSubpassDescription subpass{
         .flags = 0,
         .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
@@ -215,117 +145,18 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
         .preserveAttachmentCount = 0,
         .pPreserveAttachments = nullptr,
     };
-    const VkSubpassDependency counter_resume_dependency{
-        .srcSubpass = 0,
-        .dstSubpass = 0,
-        .srcStageMask = VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT,
-        .dstStageMask = VK_PIPELINE_STAGE_TRANSFORM_FEEDBACK_BIT_EXT,
-        .srcAccessMask = VK_ACCESS_TRANSFORM_FEEDBACK_COUNTER_WRITE_BIT_EXT,
-        .dstAccessMask = VK_ACCESS_TRANSFORM_FEEDBACK_COUNTER_READ_BIT_EXT,
-        .dependencyFlags = 0,
+    const VkSubpassDependency dependency{
+            .srcSubpass = 0,  // Current subpass
+            .dstSubpass = 0,  // Same subpass (self-dependency)
+            .srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
+                            VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
+                            VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
+            .dstStageMask = VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT,
+            .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+                             VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
+            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
+            .dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT
     };
-    const bool can_resume_transform_feedback = device->IsExtTransformFeedbackSupported();
-
-    if (device->IsKhrCreateRenderPass2Supported()) {
-        boost::container::static_vector<VkAttachmentDescription2, MAX_ATTACHMENTS> descriptions2;
-        for (const VkAttachmentDescription& description : descriptions) {
-            descriptions2.push_back(VkAttachmentDescription2{
-                .sType = VK_STRUCTURE_TYPE_ATTACHMENT_DESCRIPTION_2,
-                .pNext = nullptr,
-                .flags = description.flags,
-                .format = description.format,
-                .samples = description.samples,
-                .loadOp = description.loadOp,
-                .storeOp = description.storeOp,
-                .stencilLoadOp = description.stencilLoadOp,
-                .stencilStoreOp = description.stencilStoreOp,
-                .initialLayout = description.initialLayout,
-                .finalLayout = description.finalLayout,
-            });
-        }
-        const auto promote = [](const VkAttachmentReference& reference) {
-            return VkAttachmentReference2{
-                .sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2,
-                .pNext = nullptr,
-                .attachment = reference.attachment,
-                .layout = reference.layout,
-                .aspectMask = 0,
-            };
-        };
-        std::array<VkAttachmentReference2, 8> references2{};
-        std::array<VkAttachmentReference2, 8> resolve_references2{};
-        for (size_t index = 0; index < references.size(); ++index) {
-            references2[index] = promote(references[index]);
-            resolve_references2[index] = promote(resolve_references[index]);
-        }
-        const VkAttachmentReference2 depth_reference2 = promote(depth_reference);
-        const VkAttachmentReference2 depth_resolve_reference2 = promote(depth_resolve_reference);
-        const ResolveModes resolve_modes = PickResolveModes(*device, key.depth_format);
-        const VkSubpassDescriptionDepthStencilResolve depth_stencil_resolve{
-            .sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_DEPTH_STENCIL_RESOLVE,
-            .pNext = nullptr,
-            .depthResolveMode = resolve_modes.depth,
-            .stencilResolveMode = resolve_modes.stencil,
-            .pDepthStencilResolveAttachment = &depth_resolve_reference2,
-        };
-        const VkSubpassDescription2 subpass2{
-            .sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_2,
-            .pNext = do_resolve_depth_stencil ? &depth_stencil_resolve : nullptr,
-            .flags = 0,
-            .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
-            .viewMask = 0,
-            .inputAttachmentCount = 0,
-            .pInputAttachments = nullptr,
-            .colorAttachmentCount = num_attachments,
-            .pColorAttachments = references2.data(),
-            .pResolveAttachments = do_resolve_color ? resolve_references2.data() : nullptr,
-            .pDepthStencilAttachment = has_depth ? &depth_reference2 : nullptr,
-            .preserveAttachmentCount = 0,
-            .pPreserveAttachments = nullptr,
-        };
-        const VkMemoryBarrier2 counter_resume_barrier{
-            .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER_2,
-            .pNext = nullptr,
-            .srcStageMask = VK_PIPELINE_STAGE_2_TRANSFORM_FEEDBACK_BIT_EXT,
-            .srcAccessMask = VK_ACCESS_2_TRANSFORM_FEEDBACK_COUNTER_WRITE_BIT_EXT,
-            .dstStageMask = VK_PIPELINE_STAGE_2_TRANSFORM_FEEDBACK_BIT_EXT,
-            .dstAccessMask = VK_ACCESS_2_TRANSFORM_FEEDBACK_COUNTER_READ_BIT_EXT,
-        };
-        VkSubpassDependency2 counter_resume_dependency2{
-            .sType = VK_STRUCTURE_TYPE_SUBPASS_DEPENDENCY_2,
-            .pNext = nullptr,
-            .srcSubpass = counter_resume_dependency.srcSubpass,
-            .dstSubpass = counter_resume_dependency.dstSubpass,
-            .srcStageMask = counter_resume_dependency.srcStageMask,
-            .dstStageMask = counter_resume_dependency.dstStageMask,
-            .srcAccessMask = counter_resume_dependency.srcAccessMask,
-            .dstAccessMask = counter_resume_dependency.dstAccessMask,
-            .dependencyFlags = counter_resume_dependency.dependencyFlags,
-            .viewOffset = 0,
-        };
-        if (device->HasSynchronization2()) {
-            counter_resume_dependency2.pNext = &counter_resume_barrier;
-            counter_resume_dependency2.srcStageMask = 0;
-            counter_resume_dependency2.dstStageMask = 0;
-            counter_resume_dependency2.srcAccessMask = 0;
-            counter_resume_dependency2.dstAccessMask = 0;
-        }
-        pair->second = device->GetLogical().CreateRenderPass2({
-            .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO_2,
-            .pNext = nullptr,
-            .flags = 0,
-            .attachmentCount = static_cast<u32>(descriptions2.size()),
-            .pAttachments = descriptions2.empty() ? nullptr : descriptions2.data(),
-            .subpassCount = 1,
-            .pSubpasses = &subpass2,
-            .dependencyCount = can_resume_transform_feedback ? 1u : 0u,
-            .pDependencies = can_resume_transform_feedback ? &counter_resume_dependency2 : nullptr,
-            .correlatedViewMaskCount = 0,
-            .pCorrelatedViewMasks = nullptr,
-        });
-        return *pair->second;
-    }
-
     pair->second = device->GetLogical().CreateRenderPass({
         .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
         .pNext = nullptr,
@@ -334,8 +165,8 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
         .pAttachments = descriptions.empty() ? nullptr : descriptions.data(),
         .subpassCount = 1,
         .pSubpasses = &subpass,
-        .dependencyCount = can_resume_transform_feedback ? 1u : 0u,
-        .pDependencies = can_resume_transform_feedback ? &counter_resume_dependency : nullptr,
+        .dependencyCount = 1,
+        .pDependencies = &dependency,
     });
     return *pair->second;
 }

src/video_core/renderer_vulkan/vk_render_pass_cache.h

@@ -9,7 +9,6 @@
 #include <mutex>
 #include <ankerl/unordered_dense.h>
 
-#include "common/container_hash.h"
 #include "video_core/surface.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
@@ -22,7 +21,6 @@ struct RenderPassKey {
     VideoCore::Surface::PixelFormat depth_format;
     VkSampleCountFlagBits samples;
     bool resolve_color;
-    bool resolve_depth_stencil;
     u32 color_clear_mask;
     bool depth_stencil_clear;
     u32 color_discard_mask;
@@ -33,27 +31,17 @@ struct RenderPassKey {
 namespace std {
 template <>
 struct hash<Vulkan::RenderPassKey> {
-    static_assert(std::tuple_size_v<decltype(Vulkan::RenderPassKey::color_formats)> <= 8);
-    static_assert(static_cast<u32>(VideoCore::Surface::PixelFormat::Invalid) <= 0xFF);
-    static_assert(static_cast<u32>(VideoCore::Surface::PixelFormat::Max) <= 0xFF);
-    static_assert(VK_SAMPLE_COUNT_64_BIT <= 0xFF);
-
     [[nodiscard]] size_t operator()(const Vulkan::RenderPassKey& key) const noexcept {
-        u64 formats = 0;
-        for (size_t index = 0; index < key.color_formats.size(); ++index) {
-            formats |= static_cast<u64>(key.color_formats[index]) << (index * 8);
+        size_t value = static_cast<size_t>(key.depth_format) << 48;
+        value ^= static_cast<size_t>(key.samples) << 52;
+        value ^= static_cast<size_t>(key.resolve_color) << 63;
+        value ^= static_cast<size_t>(key.color_clear_mask) << 54;
+        value ^= static_cast<size_t>(key.depth_stencil_clear) << 62;
+        value ^= static_cast<size_t>(key.color_discard_mask) << 24;
+        for (size_t i = 0; i < key.color_formats.size(); ++i) {
+            value ^= static_cast<size_t>(key.color_formats[i]) << (i * 6);
         }
-        const u64 state = static_cast<u64>(key.depth_format) |
-                          (static_cast<u64>(key.samples) << 8) |
-                          (static_cast<u64>(key.color_clear_mask) << 16) |
-                          (static_cast<u64>(key.color_discard_mask) << 24) |
-                          (static_cast<u64>(key.resolve_color) << 32) |
-                          (static_cast<u64>(key.depth_stencil_clear) << 33) |
-                          (static_cast<u64>(key.resolve_depth_stencil) << 34);
-        size_t seed = 0;
-        Common::HashCombine(seed, formats);
-        Common::HashCombine(seed, state);
-        return seed;
+        return value;
     }
 };
 } // namespace std
@@ -62,9 +50,6 @@ namespace Vulkan {
 
 class Device;
 
-[[nodiscard]] bool SupportsDepthStencilResolve(const Device& device,
-                                               VideoCore::Surface::PixelFormat depth_format);
-
 class RenderPassCache {
 public:
     explicit RenderPassCache(const Device& device_);

src/video_core/renderer_vulkan/vk_scheduler.cpp

@@ -410,17 +410,8 @@ void Scheduler::EndRenderPass()
         Record([num_images = num_renderpass_images,
                        images = renderpass_images,
                        ranges = renderpass_image_ranges,
-                       consumer_stages = device.AttachmentConsumerStages(),
                        has_transform_feedback = device.IsExtTransformFeedbackSupported()](
                           vk::CommandBuffer cmdbuf) {
-            static constexpr VkAccessFlags SHADER_ACCESS =
-                VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT;
-            static constexpr VkAccessFlags COLOR_ACCESS =
-                VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
-            static constexpr VkAccessFlags DEPTH_STENCIL_ACCESS =
-                VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
-                VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
-
             std::array<VkImageMemoryBarrier, 9> barriers;
             for (size_t i = 0; i < num_images; ++i) {
                 const VkImageSubresourceRange& range = ranges[i];
@@ -430,25 +421,24 @@ void Scheduler::EndRenderPass()
                                                  | VK_IMAGE_ASPECT_STENCIL_BIT)) !=0;
 
                 VkAccessFlags src_access = 0;
-                VkAccessFlags dst_access = SHADER_ACCESS;
 
-                if (is_color) {
+                if (is_color)
                     src_access |= VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
-                    dst_access |= COLOR_ACCESS;
-                } else if (is_depth_stencil) {
+                else if (is_depth_stencil)
                     src_access |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
-                    dst_access |= DEPTH_STENCIL_ACCESS;
-                } else {
+                else
                     src_access |= VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT
                                   | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
-                    dst_access |= COLOR_ACCESS | DEPTH_STENCIL_ACCESS;
-                }
 
                 barriers[i] = VkImageMemoryBarrier{
                         .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
                         .pNext = nullptr,
                         .srcAccessMask = src_access,
-                        .dstAccessMask = dst_access,
+                        .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT
+                                         | VK_ACCESS_COLOR_ATTACHMENT_READ_BIT
+                                         | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT
+                                         | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT
+                                         | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
                         .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
                         .newLayout = VK_IMAGE_LAYOUT_GENERAL,
                         .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
@@ -459,7 +449,7 @@ void Scheduler::EndRenderPass()
             }
             cmdbuf.EndRenderPass();
             cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
-                                   VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, consumer_stages,
+                                   VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE,
                                    0, nullptr, nullptr, vk::Span(barriers.data(), num_images));
             if (has_transform_feedback) {
                 static constexpr VkMemoryBarrier XFB_OUTPUT_BARRIER{

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -628,12 +628,18 @@ void CopyBufferToImage(vk::CommandBuffer cmdbuf, VkBuffer src_buffer, VkImage im
             .subresourceRange = subresource_range,
     };
 
-    cmdbuf.PipelineBarrier(vk::PIPELINE_STAGE_GRAPHICS_COMPUTE, VK_PIPELINE_STAGE_TRANSFER_BIT, 0,
+    cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
+                           VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
+                           VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0,
                            read_barrier);
     cmdbuf.CopyBufferToImage(src_buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copies);
     // TODO: Move this to another API
-    cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE, 0,
-                           nullptr, nullptr, write_barrier);
+    cmdbuf.PipelineBarrier(
+            VK_PIPELINE_STAGE_TRANSFER_BIT,
+            VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
+            VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
+            VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
+            0, nullptr, nullptr, write_barrier);
 }
 
 [[nodiscard]] VkImageBlit MakeImageBlit(const Region2D& dst_region, const Region2D& src_region,
@@ -927,10 +933,6 @@ TextureCacheRuntime::TextureCacheRuntime(const Device& device_, Scheduler& sched
         bl3d_unswizzle_pass.emplace(device, scheduler, descriptor_pool,
                                    staging_buffer_pool, compute_pass_descriptor_queue);
     }
-    bl2d_unswizzle_pass.emplace(device, scheduler, descriptor_pool, staging_buffer_pool,
-                                compute_pass_descriptor_queue);
-    bl3db_unswizzle_pass.emplace(device, scheduler, descriptor_pool, staging_buffer_pool,
-                                 compute_pass_descriptor_queue);
 }
 
 void TextureCacheRuntime::Finish() {
@@ -993,25 +995,13 @@ VkBuffer TextureCacheRuntime::GetTemporaryBuffer(size_t needed_size) {
 }
 
 VkImageView TextureCacheRuntime::GetOrCreateResolveShadow(VkImage msaa_image, VkFormat format,
-                                                          VkExtent2D extent, u32 layers,
-                                                          VkImageAspectFlags aspect_mask) {
+                                                          VkExtent2D extent, u32 layers) {
     ResolveShadow& shadow = resolve_shadows[msaa_image];
     if (shadow.image && shadow.format == format && shadow.extent.width == extent.width &&
-        shadow.extent.height == extent.height && shadow.layers == layers &&
-        shadow.aspect_mask == aspect_mask) {
+        shadow.extent.height == extent.height && shadow.layers == layers) {
         shadow.up_to_date = true;
         return *shadow.view;
     }
-    VkImageUsageFlags shadow_usage =
-        VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
-    if ((aspect_mask & (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) != 0) {
-        shadow_usage |= VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
-    } else {
-        shadow_usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
-    }
-    if (shadow.image) {
-        pending_resolve_shadows.emplace_back(scheduler.CurrentTick(), std::move(shadow));
-    }
     shadow.image = memory_allocator.CreateImage(VkImageCreateInfo{
         .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
         .pNext = nullptr,
@@ -1023,7 +1013,8 @@ VkImageView TextureCacheRuntime::GetOrCreateResolveShadow(VkImage msaa_image, Vk
         .arrayLayers = layers,
         .samples = VK_SAMPLE_COUNT_1_BIT,
         .tiling = VK_IMAGE_TILING_OPTIMAL,
-        .usage = shadow_usage,
+        .usage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT | VK_IMAGE_USAGE_SAMPLED_BIT |
+                 VK_IMAGE_USAGE_TRANSFER_SRC_BIT,
         .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
         .queueFamilyIndexCount = 0,
         .pQueueFamilyIndices = nullptr,
@@ -1038,7 +1029,7 @@ VkImageView TextureCacheRuntime::GetOrCreateResolveShadow(VkImage msaa_image, Vk
         .format = format,
         .components{},
         .subresourceRange{
-            .aspectMask = aspect_mask,
+            .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
             .baseMipLevel = 0,
             .levelCount = 1,
             .baseArrayLayer = 0,
@@ -1048,7 +1039,6 @@ VkImageView TextureCacheRuntime::GetOrCreateResolveShadow(VkImage msaa_image, Vk
     shadow.format = format;
     shadow.extent = extent;
     shadow.layers = layers;
-    shadow.aspect_mask = aspect_mask;
     shadow.up_to_date = true;
     return *shadow.view;
 }
@@ -1070,14 +1060,7 @@ void TextureCacheRuntime::InvalidateResolveShadow(VkImage msaa_image) {
 }
 
 void TextureCacheRuntime::EraseResolveShadow(VkImage msaa_image) {
-    const auto it = resolve_shadows.find(msaa_image);
-    if (it == resolve_shadows.end()) {
-        return;
-    }
-    if (it->second.image) {
-        pending_resolve_shadows.emplace_back(scheduler.CurrentTick(), std::move(it->second));
-    }
-    resolve_shadows.erase(it);
+    resolve_shadows.erase(msaa_image);
 }
 
 void TextureCacheRuntime::BarrierFeedbackLoop() {
@@ -1212,8 +1195,7 @@ void TextureCacheRuntime::ReinterpretImage(Image& dst, Image& src,
 
         cmdbuf.PipelineBarrier(vk::PIPELINE_STAGE_GRAPHICS_COMPUTE_TRANSFER, VK_PIPELINE_STAGE_TRANSFER_BIT,
                        0, READ_BARRIER, {}, middle_out_barrier);
-        cmdbuf.CopyBufferToImage(copy_buffer, dst_image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
-                                 vk_out_copies);
+        cmdbuf.CopyBufferToImage(copy_buffer, dst_image, VK_IMAGE_LAYOUT_GENERAL, vk_out_copies);
         cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE,
                        0, {}, {}, post_barriers);
     });
@@ -1375,83 +1357,133 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
     }
 
     switch (dst_view.format) {
-    case PixelFormat::R16_UNORM:
-        if (src_view.format == PixelFormat::D16_UNORM) {
-            return blit_image_helper.ConvertD16ToR16(dst, src_view);
-        }
-        break;
-    case PixelFormat::A8B8G8R8_SRGB:
-    case PixelFormat::B8G8R8A8_SRGB:
-    case PixelFormat::B8G8R8A8_UNORM:
-        if (src_view.format == PixelFormat::D32_FLOAT) {
-            return blit_image_helper.ConvertD32FToABGR8(dst, src_view);
+    case PixelFormat::D24_UNORM_S8_UINT:
+        if (src_view.format == PixelFormat::A8B8G8R8_UNORM
+        || src_view.format == PixelFormat::B8G8R8A8_UNORM
+        || src_view.format == PixelFormat::A8B8G8R8_SRGB
+        || src_view.format == PixelFormat::B8G8R8A8_SRGB) {
+            return blit_image_helper.ConvertABGR8ToD24S8(dst, src_view);
         }
         break;
     case PixelFormat::A8B8G8R8_UNORM:
-        if (src_view.format == PixelFormat::S8_UINT_D24_UNORM) {
-            return blit_image_helper.ConvertD24S8ToABGR8(dst, src_view);
-        }
-        if (src_view.format == PixelFormat::D24_UNORM_S8_UINT) {
-            return blit_image_helper.ConvertS8D24ToABGR8(dst, src_view);
-        }
-        if (src_view.format == PixelFormat::D32_FLOAT) {
-            return blit_image_helper.ConvertD32FToABGR8(dst, src_view);
-        }
-        break;
+    case PixelFormat::A8B8G8R8_SNORM:
+    case PixelFormat::A8B8G8R8_SINT:
+    case PixelFormat::A8B8G8R8_UINT:
+    case PixelFormat::R5G6B5_UNORM:
+    case PixelFormat::B5G6R5_UNORM:
+    case PixelFormat::A1R5G5B5_UNORM:
+    case PixelFormat::A2B10G10R10_UNORM:
+    case PixelFormat::A2B10G10R10_UINT:
+    case PixelFormat::A2R10G10B10_UNORM:
+    case PixelFormat::A1B5G5R5_UNORM:
+    case PixelFormat::A5B5G5R1_UNORM:
+    case PixelFormat::R8_UNORM:
+    case PixelFormat::R8_SNORM:
+    case PixelFormat::R8_SINT:
+    case PixelFormat::R8_UINT:
+    case PixelFormat::R16G16B16A16_FLOAT:
+    case PixelFormat::R16G16B16A16_UNORM:
+    case PixelFormat::R16G16B16A16_SNORM:
+    case PixelFormat::R16G16B16A16_SINT:
+    case PixelFormat::R16G16B16A16_UINT:
+    case PixelFormat::B10G11R11_FLOAT:
+    case PixelFormat::R32G32B32A32_UINT:
+    case PixelFormat::BC1_RGBA_UNORM:
+    case PixelFormat::BC2_UNORM:
+    case PixelFormat::BC3_UNORM:
+    case PixelFormat::BC4_UNORM:
+    case PixelFormat::BC4_SNORM:
+    case PixelFormat::BC5_UNORM:
+    case PixelFormat::BC5_SNORM:
+    case PixelFormat::BC7_UNORM:
+    case PixelFormat::BC6H_UFLOAT:
+    case PixelFormat::BC6H_SFLOAT:
+    case PixelFormat::ASTC_2D_4X4_UNORM:
+    case PixelFormat::B8G8R8A8_UNORM:
+    case PixelFormat::R32G32B32A32_FLOAT:
+    case PixelFormat::R32G32B32A32_SINT:
+    case PixelFormat::R32G32_FLOAT:
+    case PixelFormat::R32G32_SINT:
     case PixelFormat::R32_FLOAT:
-        if (src_view.format == PixelFormat::D32_FLOAT) {
-            return blit_image_helper.ConvertD32ToR32(dst, src_view);
-        }
-        break;
-    case PixelFormat::D16_UNORM:
-        if (src_view.format == PixelFormat::R16_UNORM) {
-            return blit_image_helper.ConvertR16ToD16(dst, src_view);
-        }
-        break;
-    case PixelFormat::S8_UINT_D24_UNORM:
-        if (src_view.format == PixelFormat::A8B8G8R8_UNORM ||
-            src_view.format == PixelFormat::B8G8R8A8_UNORM) {
-            return blit_image_helper.ConvertABGR8ToD24S8(dst, src_view);
+        if (src_view.format == PixelFormat::D32_FLOAT &&
+            (dst_view.format == PixelFormat::B5G6R5_UNORM ||
+             Settings::values.fix_bloom_effects.GetValue())) {
+            const Region2D region{
+                .start = {0, 0},
+                .end = {static_cast<s32>(dst->RenderArea().width),
+                        static_cast<s32>(dst->RenderArea().height)},
+            };
+            return blit_image_helper.BlitColor(dst, src_view, region, region,
+                                            Tegra::Engines::Fermi2D::Filter::Point,
+                                            Tegra::Engines::Fermi2D::Operation::SrcCopy);
         }
         break;
+    case PixelFormat::R16_FLOAT:
+    case PixelFormat::R16_UNORM:
+    case PixelFormat::R16_SNORM:
+    case PixelFormat::R16_UINT:
+    case PixelFormat::R16_SINT:
+    case PixelFormat::R16G16_UNORM:
+    case PixelFormat::R16G16_FLOAT:
+    case PixelFormat::R16G16_UINT:
+    case PixelFormat::R16G16_SINT:
+    case PixelFormat::R16G16_SNORM:
+    case PixelFormat::R32G32B32_FLOAT:
+    case PixelFormat::A8B8G8R8_SRGB:
+    case PixelFormat::R8G8_UNORM:
+    case PixelFormat::R8G8_SNORM:
+    case PixelFormat::R8G8_SINT:
+    case PixelFormat::R8G8_UINT:
+    case PixelFormat::R32G32_UINT:
+    case PixelFormat::R16G16B16X16_FLOAT:
+    case PixelFormat::R32_UINT:
+    case PixelFormat::R32_SINT:
+    case PixelFormat::ASTC_2D_8X8_UNORM:
+    case PixelFormat::ASTC_2D_8X5_UNORM:
+    case PixelFormat::ASTC_2D_5X4_UNORM:
+    case PixelFormat::B8G8R8A8_SRGB:
+    case PixelFormat::BC1_RGBA_SRGB:
+    case PixelFormat::BC2_SRGB:
+    case PixelFormat::BC3_SRGB:
+    case PixelFormat::BC7_SRGB:
+    case PixelFormat::A4B4G4R4_UNORM:
+    case PixelFormat::G4R4_UNORM:
+    case PixelFormat::ASTC_2D_4X4_SRGB:
+    case PixelFormat::ASTC_2D_8X8_SRGB:
+    case PixelFormat::ASTC_2D_8X5_SRGB:
+    case PixelFormat::ASTC_2D_5X4_SRGB:
+    case PixelFormat::ASTC_2D_5X5_UNORM:
+    case PixelFormat::ASTC_2D_5X5_SRGB:
+    case PixelFormat::ASTC_2D_10X8_UNORM:
+    case PixelFormat::ASTC_2D_10X8_SRGB:
+    case PixelFormat::ASTC_2D_6X6_UNORM:
+    case PixelFormat::ASTC_2D_6X6_SRGB:
+    case PixelFormat::ASTC_2D_10X6_UNORM:
+    case PixelFormat::ASTC_2D_10X6_SRGB:
+    case PixelFormat::ASTC_2D_10X5_UNORM:
+    case PixelFormat::ASTC_2D_10X5_SRGB:
+    case PixelFormat::ASTC_2D_10X10_UNORM:
+    case PixelFormat::ASTC_2D_10X10_SRGB:
+    case PixelFormat::ASTC_2D_12X10_UNORM:
+    case PixelFormat::ASTC_2D_12X10_SRGB:
+    case PixelFormat::ASTC_2D_12X12_UNORM:
+    case PixelFormat::ASTC_2D_12X12_SRGB:
+    case PixelFormat::ASTC_2D_8X6_UNORM:
+    case PixelFormat::ASTC_2D_8X6_SRGB:
+    case PixelFormat::ASTC_2D_6X5_UNORM:
+    case PixelFormat::ASTC_2D_6X5_SRGB:
+    case PixelFormat::E5B9G9R9_FLOAT:
     case PixelFormat::D32_FLOAT:
-        if (src_view.format == PixelFormat::A8B8G8R8_UNORM ||
-            src_view.format == PixelFormat::B8G8R8A8_UNORM ||
-            src_view.format == PixelFormat::A8B8G8R8_SRGB ||
-            src_view.format == PixelFormat::B8G8R8A8_SRGB) {
-            return blit_image_helper.ConvertABGR8ToD32F(dst, src_view);
-        }
-        if (src_view.format == PixelFormat::R32_FLOAT) {
-            return blit_image_helper.ConvertR32ToD32(dst, src_view);
-        }
-        break;
-    case PixelFormat::D24_UNORM_S8_UINT:
-        if (src_view.format == PixelFormat::A8B8G8R8_UNORM ||
-            src_view.format == PixelFormat::B8G8R8A8_UNORM ||
-            src_view.format == PixelFormat::A8B8G8R8_SRGB ||
-            src_view.format == PixelFormat::B8G8R8A8_SRGB) {
-            return blit_image_helper.ConvertABGR8ToD24S8(dst, src_view);
-        }
-        break;
+    case PixelFormat::D16_UNORM:
+    case PixelFormat::X8_D24_UNORM:
+    case PixelFormat::S8_UINT:
+    case PixelFormat::S8_UINT_D24_UNORM:
+    case PixelFormat::D32_FLOAT_S8_UINT:
+    case PixelFormat::Invalid:
     default:
+        LOG_DEBUG(Render_Vulkan, "Unimplemented texture conversion from {} to {} format type", src_view.format, dst_view.format);
         break;
     }
-
-    if (src_view.format == PixelFormat::D32_FLOAT &&
-        VideoCore::Surface::GetFormatType(dst_view.format) == SurfaceType::ColorTexture &&
-        (dst_view.format == PixelFormat::B5G6R5_UNORM ||
-         Settings::values.fix_bloom_effects.GetValue())) {
-        const Region2D region{
-            .start = {0, 0},
-            .end = {static_cast<s32>(dst->RenderArea().width),
-                    static_cast<s32>(dst->RenderArea().height)},
-        };
-        return blit_image_helper.BlitColor(dst, src_view, region, region,
-                                        Tegra::Engines::Fermi2D::Filter::Point,
-                                        Tegra::Engines::Fermi2D::Operation::SrcCopy);
-    }
-
-    LOG_DEBUG(Render_Vulkan, "Unimplemented texture conversion from {} to {} format type", src_view.format, dst_view.format);
 }
 
 VkFormat TextureCacheRuntime::GetSupportedFormat(VkFormat requested_format,
@@ -1608,41 +1640,31 @@ void TextureCacheRuntime::CopyImageMSAA(Image& dst, Image& src,
                                         std::span<const VideoCommon::ImageCopy> copies) {
     const bool msaa_to_non_msaa = src.info.num_samples > 1 && dst.info.num_samples == 1;
     const u32 num_samples = msaa_to_non_msaa ? src.info.num_samples : dst.info.num_samples;
+    if (dst.AspectMask() != VK_IMAGE_ASPECT_COLOR_BIT ||
+        VideoCore::Surface::IsPixelFormatInteger(dst.info.format)) {
+        UNIMPLEMENTED_MSG("Copying images with different samples is not supported.");
+        return;
+    }
     if (ENABLE_MSAA_RESOLVE_CONSUME && msaa_to_non_msaa && copies.size() == 1 &&
         src.info.format == dst.info.format) {
         const VideoCommon::ImageCopy& copy = copies.front();
         const ResolveShadow* const shadow = GetValidResolveShadow(src.Handle());
-        if (shadow != nullptr && shadow->aspect_mask == dst.AspectMask() &&
-            copy.src_offset.x == 0 && copy.src_offset.y == 0 &&
+        if (shadow != nullptr && copy.src_offset.x == 0 && copy.src_offset.y == 0 &&
             copy.src_subresource.base_level == 0 &&
             static_cast<u32>(copy.extent.width) <= shadow->extent.width &&
-            static_cast<u32>(copy.extent.height) <= shadow->extent.height &&
-            static_cast<u32>(copy.src_subresource.base_layer + copy.src_subresource.num_layers) <=
-                shadow->layers) {
-            const VkImageAspectFlags aspect_mask = shadow->aspect_mask;
-            VkPipelineStageFlags attachment_stage = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
-            VkAccessFlags attachment_write = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
-            VkAccessFlags attachment_read_write =
-                VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
-            if ((aspect_mask & (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) != 0) {
-                attachment_stage = VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
-                                   VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
-                attachment_write = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
-                attachment_read_write = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
-                                        VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
-            }
+            static_cast<u32>(copy.extent.height) <= shadow->extent.height) {
             const VkImage shadow_image = *shadow->image;
             const VkImage dst_image = dst.Handle();
             const VkImageCopy region{
                 .srcSubresource{
-                    .aspectMask = aspect_mask,
+                    .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
                     .mipLevel = 0,
                     .baseArrayLayer = static_cast<u32>(copy.src_subresource.base_layer),
                     .layerCount = static_cast<u32>(copy.src_subresource.num_layers),
                 },
                 .srcOffset = {0, 0, 0},
                 .dstSubresource{
-                    .aspectMask = aspect_mask,
+                    .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
                     .mipLevel = static_cast<u32>(copy.dst_subresource.base_level),
                     .baseArrayLayer = static_cast<u32>(copy.dst_subresource.base_layer),
                     .layerCount = static_cast<u32>(copy.dst_subresource.num_layers),
@@ -1651,27 +1673,26 @@ void TextureCacheRuntime::CopyImageMSAA(Image& dst, Image& src,
                 .extent = {copy.extent.width, copy.extent.height, 1},
             };
             scheduler.RequestOutsideRenderPassOperationContext();
-            scheduler.Record([shadow_image, dst_image, region, aspect_mask, attachment_stage,
-                              attachment_write,
-                              attachment_read_write](vk::CommandBuffer cmdbuf) {
+            scheduler.Record([shadow_image, dst_image, region](vk::CommandBuffer cmdbuf) {
                 const std::array pre_barriers{
                     VkImageMemoryBarrier{
                         .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
                         .pNext = nullptr,
-                        .srcAccessMask = attachment_write,
+                        .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
                         .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
                         .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
                         .newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
                         .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                         .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                         .image = shadow_image,
-                        .subresourceRange{aspect_mask, 0, VK_REMAINING_MIP_LEVELS, 0,
+                        .subresourceRange{VK_IMAGE_ASPECT_COLOR_BIT, 0, VK_REMAINING_MIP_LEVELS, 0,
                                           VK_REMAINING_ARRAY_LAYERS},
                     },
                     VkImageMemoryBarrier{
                         .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
                         .pNext = nullptr,
-                        .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT | attachment_write |
+                        .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT |
+                                         VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
                                          VK_ACCESS_TRANSFER_WRITE_BIT,
                         .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
                         .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
@@ -1679,7 +1700,7 @@ void TextureCacheRuntime::CopyImageMSAA(Image& dst, Image& src,
                         .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                         .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                         .image = dst_image,
-                        .subresourceRange{aspect_mask, 0, VK_REMAINING_MIP_LEVELS, 0,
+                        .subresourceRange{VK_IMAGE_ASPECT_COLOR_BIT, 0, VK_REMAINING_MIP_LEVELS, 0,
                                           VK_REMAINING_ARRAY_LAYERS},
                     },
                 };
@@ -1694,25 +1715,28 @@ void TextureCacheRuntime::CopyImageMSAA(Image& dst, Image& src,
                         .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                         .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                         .image = shadow_image,
-                        .subresourceRange{aspect_mask, 0, VK_REMAINING_MIP_LEVELS, 0,
+                        .subresourceRange{VK_IMAGE_ASPECT_COLOR_BIT, 0, VK_REMAINING_MIP_LEVELS, 0,
                                           VK_REMAINING_ARRAY_LAYERS},
                     },
                     VkImageMemoryBarrier{
                         .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
                         .pNext = nullptr,
                         .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
-                        .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | attachment_read_write |
+                        .dstAccessMask = VK_ACCESS_SHADER_READ_BIT |
+                                         VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
+                                         VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
                                          VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
                         .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
                         .newLayout = VK_IMAGE_LAYOUT_GENERAL,
                         .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                         .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
                         .image = dst_image,
-                        .subresourceRange{aspect_mask, 0, VK_REMAINING_MIP_LEVELS, 0,
+                        .subresourceRange{VK_IMAGE_ASPECT_COLOR_BIT, 0, VK_REMAINING_MIP_LEVELS, 0,
                                           VK_REMAINING_ARRAY_LAYERS},
                     },
                 };
-                cmdbuf.PipelineBarrier(attachment_stage | VK_PIPELINE_STAGE_TRANSFER_BIT,
+                cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
+                                           VK_PIPELINE_STAGE_TRANSFER_BIT,
                                        VK_PIPELINE_STAGE_TRANSFER_BIT, 0, nullptr, nullptr,
                                        pre_barriers);
                 cmdbuf.CopyImage(shadow_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_image,
@@ -1724,11 +1748,6 @@ void TextureCacheRuntime::CopyImageMSAA(Image& dst, Image& src,
             return;
         }
     }
-    if (dst.AspectMask() != VK_IMAGE_ASPECT_COLOR_BIT ||
-        VideoCore::Surface::IsPixelFormatInteger(dst.info.format)) {
-        UNIMPLEMENTED_MSG("Copying images with different samples is not supported.");
-        return;
-    }
     blit_image_helper.CopyMSAA(render_pass_cache, dst.Handle(), dst.info.format, src.Handle(),
                                src.info.format, num_samples, copies, msaa_to_non_msaa);
 }
@@ -1753,9 +1772,6 @@ void TextureCacheRuntime::TickFrame() {
     std::erase_if(pending_msaa_images, [this](const auto& pending) {
         return scheduler.IsFree(pending.first);
     });
-    std::erase_if(pending_resolve_shadows, [this](const auto& pending) {
-        return scheduler.IsFree(pending.first);
-    });
 }
 
 Image::Image(TextureCacheRuntime& runtime_, const ImageInfo& info_, GPUVAddr gpu_addr_,
@@ -1780,13 +1796,6 @@ Image::Image(TextureCacheRuntime& runtime_, const ImageInfo& info_, GPUVAddr gpu
         }
         flags |= VideoCommon::ImageFlagBits::Converted;
         flags |= VideoCommon::ImageFlagBits::CostlyLoad;
-    } else if (runtime->bl2d_unswizzle_pass && BlockLinearUnswizzle2DPass::IsSupported(info)) {
-        flags |= VideoCommon::ImageFlagBits::AcceleratedUpload;
-        flags |= VideoCommon::ImageFlagBits::CostlyLoad;
-    } else if (runtime->bl3db_unswizzle_pass &&
-               BlockLinearUnswizzle3DBufferPass::IsSupported(runtime->device, info)) {
-        flags |= VideoCommon::ImageFlagBits::AcceleratedUpload;
-        flags |= VideoCommon::ImageFlagBits::CostlyLoad;
     }
     if (IsPixelFormatBCn(info.format) && !runtime->device.IsOptimalBcnSupported()) {
         flags |= VideoCommon::ImageFlagBits::Converted;
@@ -1873,21 +1882,11 @@ void Image::UploadMemory(VkBuffer buffer, VkDeviceSize offset,
         ScaleDown(true);
     }
 
-    const bool msaa_upload_is_depth = (aspect_mask & VK_IMAGE_ASPECT_DEPTH_BIT) != 0;
     const bool wants_msaa_upload = info.num_samples > 1
-        && ((aspect_mask & VK_IMAGE_ASPECT_COLOR_BIT) != 0 || msaa_upload_is_depth)
+        && (aspect_mask & VK_IMAGE_ASPECT_COLOR_BIT) != 0
         && !VideoCore::Surface::IsPixelFormatInteger(info.format);
 
     if (wants_msaa_upload) {
-        const bool msaa_upload_copies_stencil =
-            msaa_upload_is_depth && (aspect_mask & VK_IMAGE_ASPECT_STENCIL_BIT) != 0 &&
-            runtime->device.IsExtShaderStencilExportSupported();
-        const VkImageAspectFlags upload_aspect_mask =
-            msaa_upload_is_depth
-                ? (msaa_upload_copies_stencil
-                       ? VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT
-                       : VK_IMAGE_ASPECT_DEPTH_BIT)
-                : aspect_mask;
         ImageInfo temp_info = info;
         temp_info.num_samples = 1;
 
@@ -1899,10 +1898,10 @@ void Image::UploadMemory(VkBuffer buffer, VkDeviceSize offset,
         vk::Image temp_image = runtime->memory_allocator.CreateImage(image_ci);
 
         scheduler->RequestOutsideRenderPassOperationContext();
-        auto vk_copies = TransformBufferImageCopies(copies, offset, upload_aspect_mask);
+        auto vk_copies = TransformBufferImageCopies(copies, offset, aspect_mask);
         const VkBuffer src_buffer = buffer;
         const VkImage temp_vk_image = *temp_image;
-        const VkImageAspectFlags vk_aspect_mask = upload_aspect_mask;
+        const VkImageAspectFlags vk_aspect_mask = aspect_mask;
 
         scheduler->Record([src_buffer, temp_vk_image, vk_aspect_mask,
                            vk_copies](vk::CommandBuffer cmdbuf) {
@@ -1924,16 +1923,9 @@ void Image::UploadMemory(VkBuffer buffer, VkDeviceSize offset,
             image_copies.push_back(image_copy);
         }
 
-        if (msaa_upload_is_depth) {
-            runtime->blit_image_helper.CopyMSAADepth(runtime->render_pass_cache, Handle(),
-                                                     info.format, temp_vk_image, info.format,
-                                                     info.num_samples, image_copies,
-                                                     msaa_upload_copies_stencil);
-        } else {
-            runtime->blit_image_helper.CopyMSAA(runtime->render_pass_cache, Handle(), info.format,
-                                                temp_vk_image, info.format, info.num_samples,
-                                                image_copies, false);
-        }
+        runtime->blit_image_helper.CopyMSAA(runtime->render_pass_cache, Handle(), info.format,
+                                            temp_vk_image, info.format, info.num_samples,
+                                            image_copies, false);
         initialized = true;
         runtime->pending_msaa_images.emplace_back(scheduler->CurrentTick(), std::move(temp_image));
 
@@ -2698,7 +2690,7 @@ Framebuffer::~Framebuffer() = default;
 void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime,
                                     std::span<ImageView*, NUM_RT> color_buffers,
                                     ImageView* depth_buffer, bool is_rescaled_) {
-    boost::container::small_vector<VkImageView, NUM_RT * 2 + 2> attachments;
+    boost::container::small_vector<VkImageView, NUM_RT + 1> attachments;
     RenderPassKey renderpass_key{};
     s32 num_layers = 1;
 
@@ -2727,8 +2719,6 @@ void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime,
         ++num_images;
     }
     const size_t num_colors = attachments.size();
-    VkImage depth_image = VK_NULL_HANDLE;
-    VkImageAspectFlags depth_aspect_mask = 0;
     if (depth_buffer) {
         width = (std::min)(width, is_rescaled ? resolution.ScaleUp(depth_buffer->size.width)
                                             : depth_buffer->size.width);
@@ -2744,8 +2734,6 @@ void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime,
         ++num_images;
         has_depth = (subresource_range.aspectMask & VK_IMAGE_ASPECT_DEPTH_BIT) != 0;
         has_stencil = (subresource_range.aspectMask & VK_IMAGE_ASPECT_STENCIL_BIT) != 0;
-        depth_image = depth_buffer->ImageHandle();
-        depth_aspect_mask = subresource_range.aspectMask;
     } else {
         renderpass_key.depth_format = PixelFormat::Invalid;
     }
@@ -2754,12 +2742,6 @@ void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime,
         samples != VK_SAMPLE_COUNT_1_BIT && num_colors > 0 && runtime.device.IsTiler();
     renderpass_key.resolve_color = do_resolve_color;
 
-    const bool do_resolve_depth_stencil =
-        samples != VK_SAMPLE_COUNT_1_BIT && depth_image != VK_NULL_HANDLE &&
-        runtime.device.IsTiler() &&
-        SupportsDepthStencilResolve(runtime.device, renderpass_key.depth_format);
-    renderpass_key.resolve_depth_stencil = do_resolve_depth_stencil;
-
     discard_msaa_color =
         ENABLE_MSAA_RESOLVE_CONSUME && ENABLE_MSAA_COLOR_DISCARD && do_resolve_color;
 
@@ -2780,8 +2762,8 @@ void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime,
                 MaxwellToVK::SurfaceFormat(runtime.device, FormatType::Optimal, true, format).format;
             if (ENABLE_MSAA_RESOLVE_CONSUME) {
                 const VkImage msaa_image = images[rt_map[index]];
-                attachments.push_back(runtime.GetOrCreateResolveShadow(
-                    msaa_image, vk_format, render_area, layers, VK_IMAGE_ASPECT_COLOR_BIT));
+                attachments.push_back(runtime.GetOrCreateResolveShadow(msaa_image, vk_format,
+                                                                       render_area, layers));
                 continue;
             }
             VkImageCreateInfo resolve_ci{
@@ -2826,16 +2808,6 @@ void Framebuffer::CreateFramebuffer(TextureCacheRuntime& runtime,
         }
     }
 
-    if (do_resolve_depth_stencil) {
-        const u32 layers = static_cast<u32>((std::max)(num_layers, 1));
-        const VkFormat vk_format =
-            MaxwellToVK::SurfaceFormat(runtime.device, FormatType::Optimal, true,
-                                       renderpass_key.depth_format)
-                .format;
-        attachments.push_back(runtime.GetOrCreateResolveShadow(depth_image, vk_format, render_area,
-                                                               layers, depth_aspect_mask));
-    }
-
     num_color_buffers = static_cast<u32>(num_colors);
     framebuffer = runtime.device.GetLogical().CreateFramebuffer({
         .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
@@ -2855,25 +2827,11 @@ VkRenderPass Framebuffer::RenderPassVariant(u32 color_clear_mask, bool depth_ste
     if (color_clear_mask == 0 && !depth_stencil_clear && color_discard_mask == 0) {
         return renderpass;
     }
-    static_assert(NUM_RT <= 8);
-    const u32 variant_key = color_clear_mask | (color_discard_mask << 8) |
-                            (static_cast<u32>(depth_stencil_clear) << 16);
-    for (u32 index = 0; index < num_memoized_variants; ++index) {
-        if (variant_keys[index] == variant_key) {
-            return variant_render_passes[index];
-        }
-    }
     RenderPassKey key = render_pass_key;
     key.color_clear_mask = color_clear_mask;
     key.depth_stencil_clear = depth_stencil_clear;
     key.color_discard_mask = color_discard_mask;
-    const VkRenderPass variant = render_pass_cache->Get(key);
-    if (num_memoized_variants < variant_keys.size()) {
-        variant_keys[num_memoized_variants] = variant_key;
-        variant_render_passes[num_memoized_variants] = variant;
-        ++num_memoized_variants;
-    }
-    return variant;
+    return render_pass_cache->Get(key);
 }
 
 void TextureCacheRuntime::AccelerateImageUpload(
@@ -2885,15 +2843,6 @@ void TextureCacheRuntime::AccelerateImageUpload(
         return astc_decoder_pass->Assemble(image, map, swizzles);
     }
 
-    if (bl2d_unswizzle_pass && BlockLinearUnswizzle2DPass::IsSupported(image.info)) {
-        return bl2d_unswizzle_pass->Unswizzle(image, map, swizzles);
-    }
-
-    if (bl3db_unswizzle_pass &&
-        BlockLinearUnswizzle3DBufferPass::IsSupported(device, image.info)) {
-        return bl3db_unswizzle_pass->Unswizzle(image, map, swizzles);
-    }
-
     if (!Settings::values.gpu_unswizzle_enabled.GetValue() || !bl3d_unswizzle_pass) {
         if (IsPixelFormatBCn(image.info.format) && image.info.type == ImageType::e3D) {
             ASSERT(false && "GPU unswizzle is disabled for BCn 3D texture");

src/video_core/renderer_vulkan/vk_texture_cache.h

@@ -117,13 +117,11 @@ public:
         VkFormat format = VK_FORMAT_UNDEFINED;
         VkExtent2D extent{};
         u32 layers = 0;
-        VkImageAspectFlags aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
         bool up_to_date = false;
     };
 
     [[nodiscard]] VkImageView GetOrCreateResolveShadow(VkImage msaa_image, VkFormat format,
-                                                       VkExtent2D extent, u32 layers,
-                                                       VkImageAspectFlags aspect_mask);
+                                                       VkExtent2D extent, u32 layers);
 
     [[nodiscard]] const ResolveShadow* GetValidResolveShadow(VkImage msaa_image) const;
 
@@ -151,8 +149,6 @@ public:
     std::optional<ASTCDecoderPass> astc_decoder_pass;
 
     std::optional<BlockLinearUnswizzle3DPass> bl3d_unswizzle_pass;
-    std::optional<BlockLinearUnswizzle2DPass> bl2d_unswizzle_pass;
-    std::optional<BlockLinearUnswizzle3DBufferPass> bl3db_unswizzle_pass;
     const Settings::ResolutionScalingInfo& resolution;
     std::array<std::vector<VkFormat>, VideoCore::Surface::MaxPixelFormat> view_formats;
 
@@ -160,7 +156,6 @@ public:
     std::array<vk::Buffer, indexing_slots> buffers{};
     std::vector<std::pair<u64, vk::Image>> pending_msaa_images;
     ankerl::unordered_dense::map<VkImage, ResolveShadow> resolve_shadows;
-    std::vector<std::pair<u64, ResolveShadow>> pending_resolve_shadows;
 };
 
 class Framebuffer {
@@ -251,8 +246,6 @@ public:
     }
 
 private:
-    static constexpr size_t NUM_MEMOIZED_RENDER_PASS_VARIANTS = 8;
-
     vk::Framebuffer framebuffer;
     VkRenderPass renderpass{};
     VkExtent2D render_area{};
@@ -270,9 +263,6 @@ private:
     RenderPassKey render_pass_key{};
     RenderPassCache* render_pass_cache{nullptr};
     bool discard_msaa_color{};
-    mutable std::array<u32, NUM_MEMOIZED_RENDER_PASS_VARIANTS> variant_keys{};
-    mutable std::array<VkRenderPass, NUM_MEMOIZED_RENDER_PASS_VARIANTS> variant_render_passes{};
-    mutable u32 num_memoized_variants{};
 };
 
 class Image : public VideoCommon::ImageBase {

src/video_core/textures/decoders.cpp

@@ -1,10 +1,9 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <algorithm>
 #include <array>
 #include <cmath>
 #include <cstring>
@@ -31,62 +30,12 @@ constexpr u32 pdep(u32 value) {
     return result;
 }
 
-constexpr u32 SWIZZLE_RUN_BYTES = 16;
-constexpr u32 SWIZZLE_RUN_SHIFT = 4;
-constexpr u32 SWIZZLE_RUN_MASK = SWIZZLE_RUN_BYTES - 1;
-constexpr u32 SWIZZLE_RUN_INDEX_MASK = GOB_SIZE_X / SWIZZLE_RUN_BYTES - 1;
-
-static_assert((SWIZZLE_X_BITS & SWIZZLE_RUN_MASK) == SWIZZLE_RUN_MASK);
-
-constexpr std::array<u32, GOB_SIZE_X / SWIZZLE_RUN_BYTES> SWIZZLE_X_RUN_TABLE = [] {
-    std::array<u32, GOB_SIZE_X / SWIZZLE_RUN_BYTES> table{};
-    for (u32 index = 0; index < static_cast<u32>(table.size()); ++index) {
-        table[index] = pdep<SWIZZLE_X_BITS>(index << SWIZZLE_RUN_SHIFT);
-    }
-    return table;
-}();
-
 template <u32 mask, u32 incr_amount>
 void incrpdep(u32& value) {
     static constexpr u32 swizzled_incr = pdep<mask>(incr_amount);
     value = ((value | ~mask) + swizzled_incr) & mask;
 }
 
-template <bool TO_LINEAR>
-void SwizzleRow(std::span<u8> output, std::span<const u8> input, u32 offset_zy, u32 swizzled_y,
-                u32 x_shift, u32 x, u32 num_bytes, u32 linear) {
-    const auto copy = [&](u32 swizzled_x, u32 count) {
-        const u32 swizzled =
-            offset_zy + ((x >> GOB_SIZE_X_SHIFT) << x_shift) + (swizzled_x | swizzled_y);
-        u8* const dst = &output[TO_LINEAR ? swizzled : linear];
-        const u8* const src = &input[TO_LINEAR ? linear : swizzled];
-        std::memcpy(dst, src, count);
-        x += count;
-        linear += count;
-    };
-
-    u32 swizzled_run = SWIZZLE_X_RUN_TABLE[(x >> SWIZZLE_RUN_SHIFT) & SWIZZLE_RUN_INDEX_MASK];
-    u32 remaining = num_bytes;
-
-    const u32 head =
-        (std::min)(SWIZZLE_RUN_BYTES - (x & SWIZZLE_RUN_MASK), remaining) & SWIZZLE_RUN_MASK;
-    if (head != 0) {
-        copy(swizzled_run | (x & SWIZZLE_RUN_MASK), head);
-        remaining -= head;
-        incrpdep<SWIZZLE_X_BITS, SWIZZLE_RUN_BYTES>(swizzled_run);
-    }
-
-    while (remaining >= SWIZZLE_RUN_BYTES) {
-        copy(swizzled_run, SWIZZLE_RUN_BYTES);
-        remaining -= SWIZZLE_RUN_BYTES;
-        incrpdep<SWIZZLE_X_BITS, SWIZZLE_RUN_BYTES>(swizzled_run);
-    }
-
-    if (remaining != 0) {
-        copy(swizzled_run, remaining);
-    }
-}
-
 template <bool TO_LINEAR, u32 BYTES_PER_PIXEL>
 void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32 height, u32 depth,
                  u32 block_height, u32 block_depth, u32 stride) {
@@ -121,9 +70,23 @@ void SwizzleImpl(std::span<u8> output, std::span<const u8> input, u32 width, u32
             const u32 offset_y = (block_y >> block_height) * block_size +
                                  ((block_y & block_height_mask) << GOB_SIZE_SHIFT);
 
-            SwizzleRow<TO_LINEAR>(output, input, offset_z + offset_y, swizzled_y, x_shift,
-                                  origin_x * BYTES_PER_PIXEL, width * BYTES_PER_PIXEL,
-                                  slice * pitch * height + line * pitch);
+            u32 swizzled_x = pdep<SWIZZLE_X_BITS>(origin_x * BYTES_PER_PIXEL);
+            for (u32 column = 0; column < width;
+                 ++column, incrpdep<SWIZZLE_X_BITS, BYTES_PER_PIXEL>(swizzled_x)) {
+                const u32 x = (column + origin_x) * BYTES_PER_PIXEL;
+                const u32 offset_x = (x >> GOB_SIZE_X_SHIFT) << x_shift;
+
+                const u32 base_swizzled_offset = offset_z + offset_y + offset_x;
+                const u32 swizzled_offset = base_swizzled_offset + (swizzled_x | swizzled_y);
+
+                const u32 unswizzled_offset =
+                    slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL;
+
+                u8* const dst = &output[TO_LINEAR ? swizzled_offset : unswizzled_offset];
+                const u8* const src = &input[TO_LINEAR ? unswizzled_offset : swizzled_offset];
+
+                std::memcpy(dst, src, BYTES_PER_PIXEL);
+            }
         }
     }
 }
@@ -166,9 +129,23 @@ void SwizzleSubrectImpl(std::span<u8> output, std::span<const u8> input, u32 wid
             const u32 offset_y = (block_y >> block_height) * block_size +
                                  ((block_y & block_height_mask) << GOB_SIZE_SHIFT);
 
-            SwizzleRow<TO_LINEAR>(output, input, offset_z + offset_y, swizzled_y, x_shift,
-                                  origin_x * BYTES_PER_PIXEL, extent_x * BYTES_PER_PIXEL,
-                                  slice * pitch * height + line * pitch);
+            u32 swizzled_x = pdep<SWIZZLE_X_BITS>(origin_x * BYTES_PER_PIXEL);
+            for (u32 column = 0; column < extent_x;
+                 ++column, incrpdep<SWIZZLE_X_BITS, BYTES_PER_PIXEL>(swizzled_x)) {
+                const u32 x = (column + origin_x) * BYTES_PER_PIXEL;
+                const u32 offset_x = (x >> GOB_SIZE_X_SHIFT) << x_shift;
+
+                const u32 base_swizzled_offset = offset_z + offset_y + offset_x;
+                const u32 swizzled_offset = base_swizzled_offset + (swizzled_x | swizzled_y);
+
+                const u32 unswizzled_offset =
+                    slice * pitch * height + line * pitch + column * BYTES_PER_PIXEL;
+
+                u8* const dst = &output[TO_LINEAR ? swizzled_offset : unswizzled_offset];
+                const u8* const src = &input[TO_LINEAR ? unswizzled_offset : swizzled_offset];
+
+                std::memcpy(dst, src, BYTES_PER_PIXEL);
+            }
         }
         unprocessed_lines -= lines_in_y;
         if (unprocessed_lines == 0) {