[host_shaders] Store the value of results instead of discarding it

the function generated the value but we never stored it effectively discarding the result.
[vk, cmake] Bump minimum VulkanTools to 1.4.317 and add UnifiedImageLayouts (#3318 )
2026-08-14 21:04:41 +00:00 · 2026-02-04 22:13:11 +01:00 · 2026-02-04 19:15:10 +01:00 · 2026-02-04 04:16:39 +01:00 · 2026-02-04 00:44:03 +01:00 · 2026-02-04 00:37:58 +01:00
33 changed files with 815 additions and 643 deletions
@@ -178,7 +178,9 @@ endif()

 # Disable Warnings as Errors for MSVC
 if (MSVC AND NOT CXX_CLANG)
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /W3 /WX-")
+    # This was dripping into spirv, being overriden, and causing cl flag override warning
+    # set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /W3 /WX-")
+    set(CMAKE_CXX_FLAGS_INIT "${CMAKE_CXX_FLAGS_INIT} /W3 /WX-")
 endif()

 # Set bundled sdl2/qt as dependent options.
@@ -587,6 +589,7 @@ if (NOT YUZU_STATIC_ROOM)
    find_package(sirit)
    find_package(gamemode)
    find_package(mcl)
+    find_package(frozen)

    if (ARCHITECTURE_riscv64)
        find_package(biscuit)
@@ -698,9 +701,6 @@ if (ENABLE_QT)

    # QuaZip
    AddJsonPackage(quazip)
-
-    # frozen
-    AddJsonPackage(frozen)
 endif()

 if (NOT YUZU_STATIC_ROOM AND NOT (YUZU_USE_BUNDLED_FFMPEG OR YUZU_USE_EXTERNAL_FFMPEG))
@@ -41,6 +41,11 @@ function(cpm_utils_message level name message)
    message(${level} "[CPMUtil] ${name}: ${message}")
 endfunction()

+# propagate a variable to parent scope
+macro(Propagate var)
+    set(${var} ${${var}} PARENT_SCOPE)
+endmacro()
+
 function(array_to_list array length out)
    math(EXPR range "${length} - 1")

@@ -72,6 +77,159 @@ function(get_json_element object out member default)
    set("${out}" "${outvar}" PARENT_SCOPE)
 endfunction()

+# Determine whether or not a package has a viable system candidate.
+function(SystemPackageViable JSON_NAME)
+    string(JSON object GET "${CPMFILE_CONTENT}" "${JSON_NAME}")
+
+    parse_object(${object})
+
+    string(REPLACE " " ";" find_args "${find_args}")
+    find_package(${package} ${version} ${find_args} QUIET NO_POLICY_SCOPE)
+
+    set(${pkg}_VIABLE ${${package}_FOUND} PARENT_SCOPE)
+    set(${pkg}_PACKAGE ${package} PARENT_SCOPE)
+endfunction()
+
+# Add several packages such that if one is bundled,
+# all the rest must also be bundled.
+function(AddDependentPackages)
+    set(_some_system OFF)
+    set(_some_bundled OFF)
+
+    foreach(pkg ${ARGN})
+        SystemPackageViable(${pkg})
+
+        if (${pkg}_VIABLE)
+            set(_some_system ON)
+            list(APPEND _system_pkgs ${${pkg}_PACKAGE})
+        else()
+            set(_some_bundled ON)
+            list(APPEND _bundled_pkgs ${${pkg}_PACKAGE})
+        endif()
+    endforeach()
+
+    if (_some_system AND _some_bundled)
+        foreach(pkg ${ARGN})
+            list(APPEND package_names ${${pkg}_PACKAGE})
+        endforeach()
+
+        string(REPLACE ";" ", " package_names "${package_names}")
+        string(REPLACE ";" ", " bundled_names "${_bundled_pkgs}")
+        foreach(sys ${_system_pkgs})
+            list(APPEND system_names ${sys}_FORCE_BUNDLED)
+        endforeach()
+
+        string(REPLACE ";" ", " system_names "${system_names}")
+
+        message(FATAL_ERROR "Partial dependency installation detected "
+            "for the following packages:\n${package_names}\n"
+            "You can solve this in one of two ways:\n"
+            "1. Install the following packages to your system if available:"
+            "\n\t${bundled_names}\n"
+            "2. Set the following variables to ON:"
+            "\n\t${system_names}\n"
+            "This may also be caused by a version mismatch, "
+            "such as one package being newer than the other.")
+    endif()
+
+    foreach(pkg ${ARGN})
+        AddJsonPackage(${pkg})
+    endforeach()
+endfunction()
+
+# json util
+macro(parse_object object)
+    get_json_element("${object}" package package ${JSON_NAME})
+    get_json_element("${object}" repo repo "")
+    get_json_element("${object}" ci ci OFF)
+    get_json_element("${object}" version version "")
+
+    if(ci)
+        get_json_element("${object}" name name "${JSON_NAME}")
+        get_json_element("${object}" extension extension "tar.zst")
+        get_json_element("${object}" min_version min_version "")
+        get_json_element("${object}" raw_disabled disabled_platforms "")
+
+        if(raw_disabled)
+            array_to_list("${raw_disabled}"
+                ${raw_disabled_LENGTH} disabled_platforms)
+        else()
+            set(disabled_platforms "")
+        endif()
+    else()
+        get_json_element("${object}" hash hash "")
+        get_json_element("${object}" hash_suffix hash_suffix "")
+        get_json_element("${object}" sha sha "")
+        get_json_element("${object}" url url "")
+        get_json_element("${object}" key key "")
+        get_json_element("${object}" tag tag "")
+        get_json_element("${object}" artifact artifact "")
+        get_json_element("${object}" git_version git_version "")
+        get_json_element("${object}" git_host git_host "")
+        get_json_element("${object}" source_subdir source_subdir "")
+        get_json_element("${object}" bundled bundled "unset")
+        get_json_element("${object}" find_args find_args "")
+        get_json_element("${object}" raw_patches patches "")
+
+        # okay here comes the fun part: REPLACEMENTS!
+        # first: tag gets %VERSION% replaced if applicable,
+        #   with either git_version (preferred) or version
+        # second: artifact gets %VERSION% and %TAG% replaced
+        #   accordingly (same rules for VERSION)
+
+        if(git_version)
+            set(version_replace ${git_version})
+        else()
+            set(version_replace ${version})
+        endif()
+
+        # TODO(crueter): fmt module for cmake
+        if(tag)
+            string(REPLACE "%VERSION%" "${version_replace}" tag ${tag})
+        endif()
+
+        if(artifact)
+            string(REPLACE "%VERSION%" "${version_replace}"
+                artifact ${artifact})
+            string(REPLACE "%TAG%" "${tag}" artifact ${artifact})
+        endif()
+
+        # format patchdir
+        if(raw_patches)
+            math(EXPR range "${raw_patches_LENGTH} - 1")
+
+            foreach(IDX RANGE ${range})
+                string(JSON _patch GET "${raw_patches}" "${IDX}")
+
+                set(full_patch
+                    "${PROJECT_SOURCE_DIR}/.patch/${JSON_NAME}/${_patch}")
+                if(NOT EXISTS ${full_patch})
+                    cpm_utils_message(FATAL_ERROR ${JSON_NAME}
+                        "specifies patch ${full_patch} which does not exist")
+                endif()
+
+                list(APPEND patches "${full_patch}")
+            endforeach()
+        endif()
+        # end format patchdir
+
+        # options
+        get_json_element("${object}" raw_options options "")
+
+        if(raw_options)
+            array_to_list("${raw_options}" ${raw_options_LENGTH} options)
+        endif()
+
+        set(options ${options} ${JSON_OPTIONS})
+        # end options
+
+        # system/bundled
+        if(bundled STREQUAL "unset" AND DEFINED JSON_BUNDLED_PACKAGE)
+            set(bundled ${JSON_BUNDLED_PACKAGE})
+        endif()
+    endif()
+endmacro()
+
 # The preferred usage
 function(AddJsonPackage)
    set(oneValueArgs
@@ -80,7 +238,8 @@ function(AddJsonPackage)
        # these are overrides that can be generated at runtime,
        # so can be defined separately from the json
        DOWNLOAD_ONLY
-        BUNDLED_PACKAGE)
+        BUNDLED_PACKAGE
+        FORCE_BUNDLED_PACKAGE)

    set(multiValueArgs OPTIONS)

@@ -111,24 +270,9 @@ function(AddJsonPackage)
        cpm_utils_message(FATAL_ERROR ${JSON_NAME} "Not found in cpmfile")
    endif()

-    get_json_element("${object}" package package ${JSON_NAME})
-    get_json_element("${object}" repo repo "")
-    get_json_element("${object}" ci ci OFF)
-    get_json_element("${object}" version version "")
+    parse_object(${object})

    if(ci)
-        get_json_element("${object}" name name "${JSON_NAME}")
-        get_json_element("${object}" extension extension "tar.zst")
-        get_json_element("${object}" min_version min_version "")
-        get_json_element("${object}" raw_disabled disabled_platforms "")
-
-        if(raw_disabled)
-            array_to_list("${raw_disabled}"
-                ${raw_disabled_LENGTH} disabled_platforms)
-        else()
-            set(disabled_platforms "")
-        endif()
-
        AddCIPackage(
            VERSION ${version}
            NAME ${name}
@@ -138,116 +282,38 @@ function(AddJsonPackage)
            MIN_VERSION ${min_version}
            DISABLED_PLATFORMS ${disabled_platforms})

-        # pass stuff to parent scope
-        set(${package}_ADDED "${${package}_ADDED}"
-            PARENT_SCOPE)
-        set(${package}_SOURCE_DIR "${${package}_SOURCE_DIR}"
-            PARENT_SCOPE)
-        set(${package}_BINARY_DIR "${${package}_BINARY_DIR}"
-            PARENT_SCOPE)
-
-        return()
-    endif()
-
-    get_json_element("${object}" hash hash "")
-    get_json_element("${object}" hash_suffix hash_suffix "")
-    get_json_element("${object}" sha sha "")
-    get_json_element("${object}" url url "")
-    get_json_element("${object}" key key "")
-    get_json_element("${object}" tag tag "")
-    get_json_element("${object}" artifact artifact "")
-    get_json_element("${object}" git_version git_version "")
-    get_json_element("${object}" git_host git_host "")
-    get_json_element("${object}" source_subdir source_subdir "")
-    get_json_element("${object}" bundled bundled "unset")
-    get_json_element("${object}" find_args find_args "")
-    get_json_element("${object}" raw_patches patches "")
-
-    # okay here comes the fun part: REPLACEMENTS!
-    # first: tag gets %VERSION% replaced if applicable,
-    #   with either git_version (preferred) or version
-    # second: artifact gets %VERSION% and %TAG% replaced
-    #   accordingly (same rules for VERSION)
-
-    if(git_version)
-        set(version_replace ${git_version})
    else()
-        set(version_replace ${version})
+        if (NOT DEFINED JSON_FORCE_BUNDLED_PACKAGE)
+            set(JSON_FORCE_BUNDLED_PACKAGE OFF)
+        endif()
+
+        AddPackage(
+            NAME "${package}"
+            VERSION "${version}"
+            URL "${url}"
+            HASH "${hash}"
+            HASH_SUFFIX "${hash_suffix}"
+            SHA "${sha}"
+            REPO "${repo}"
+            KEY "${key}"
+            PATCHES "${patches}"
+            OPTIONS "${options}"
+            FIND_PACKAGE_ARGUMENTS "${find_args}"
+            BUNDLED_PACKAGE "${bundled}"
+            FORCE_BUNDLED_PACKAGE "${JSON_FORCE_BUNDLED_PACKAGE}"
+            SOURCE_SUBDIR "${source_subdir}"
+
+            GIT_VERSION ${git_version}
+            GIT_HOST ${git_host}
+
+            ARTIFACT ${artifact}
+            TAG ${tag})
    endif()

-    # TODO(crueter): fmt module for cmake
-    if(tag)
-        string(REPLACE "%VERSION%" "${version_replace}" tag ${tag})
-    endif()
-
-    if(artifact)
-        string(REPLACE "%VERSION%" "${version_replace}" artifact ${artifact})
-        string(REPLACE "%TAG%" "${tag}" artifact ${artifact})
-    endif()
-
-    # format patchdir
-    if(raw_patches)
-        math(EXPR range "${raw_patches_LENGTH} - 1")
-
-        foreach(IDX RANGE ${range})
-            string(JSON _patch GET "${raw_patches}" "${IDX}")
-
-            set(full_patch
-                "${PROJECT_SOURCE_DIR}/.patch/${JSON_NAME}/${_patch}")
-            if(NOT EXISTS ${full_patch})
-                cpm_utils_message(FATAL_ERROR ${JSON_NAME}
-                    "specifies patch ${full_patch} which does not exist")
-            endif()
-
-            list(APPEND patches "${full_patch}")
-        endforeach()
-    endif()
-    # end format patchdir
-
-    # options
-    get_json_element("${object}" raw_options options "")
-
-    if(raw_options)
-        array_to_list("${raw_options}" ${raw_options_LENGTH} options)
-    endif()
-
-    set(options ${options} ${JSON_OPTIONS})
-    # end options
-
-    # system/bundled
-    if(bundled STREQUAL "unset" AND DEFINED JSON_BUNDLED_PACKAGE)
-        set(bundled ${JSON_BUNDLED_PACKAGE})
-    endif()
-
-    AddPackage(
-        NAME "${package}"
-        VERSION "${version}"
-        URL "${url}"
-        HASH "${hash}"
-        HASH_SUFFIX "${hash_suffix}"
-        SHA "${sha}"
-        REPO "${repo}"
-        KEY "${key}"
-        PATCHES "${patches}"
-        OPTIONS "${options}"
-        FIND_PACKAGE_ARGUMENTS "${find_args}"
-        BUNDLED_PACKAGE "${bundled}"
-        SOURCE_SUBDIR "${source_subdir}"
-
-        GIT_VERSION ${git_version}
-        GIT_HOST ${git_host}
-
-        ARTIFACT ${artifact}
-        TAG ${tag})
-
    # pass stuff to parent scope
-    set(${package}_ADDED "${${package}_ADDED}"
-        PARENT_SCOPE)
-    set(${package}_SOURCE_DIR "${${package}_SOURCE_DIR}"
-        PARENT_SCOPE)
-    set(${package}_BINARY_DIR "${${package}_BINARY_DIR}"
-        PARENT_SCOPE)
-
+    Propagate(${package}_ADDED)
+    Propagate(${package}_SOURCE_DIR)
+    Propagate(${package}_BINARY_DIR)
 endfunction()

 function(AddPackage)
@@ -343,7 +409,7 @@ function(AddPackage)

            if(DEFINED PKG_ARGS_ARTIFACT)
                set(pkg_url
-                    ${pkg_git_url}/releases/download/${PKG_ARGS_TAG}/${PKG_ARGS_ARTIFACT})
+                    "${pkg_git_url}/releases/download/${PKG_ARGS_TAG}/${PKG_ARGS_ARTIFACT}")
            else()
                set(pkg_url
                    ${pkg_git_url}/archive/refs/tags/${PKG_ARGS_TAG}.tar.gz)
@@ -625,7 +691,8 @@ function(AddCIPackage)
    endif()

    if (DEFINED pkgname AND NOT "${pkgname}" IN_LIST DISABLED_PLATFORMS)
-        set(ARTIFACT "${ARTIFACT_NAME}-${pkgname}-${ARTIFACT_VERSION}.${ARTIFACT_EXT}")
+        set(ARTIFACT
+            "${ARTIFACT_NAME}-${pkgname}-${ARTIFACT_VERSION}.${ARTIFACT_EXT}")

        AddPackage(
            NAME ${ARTIFACT_PACKAGE}
@@ -114,11 +114,5 @@
            "QUAZIP_INSTALL OFF",
            "QUAZIP_ENABLE_QTEXTCODEC OFF"
        ]
-    },
-    "frozen": {
-        "package": "frozen",
-        "repo": "serge-sans-paille/frozen",
-        "sha": "61dce5ae18",
-        "hash": "b8dfe741c82bc178dfc9749d4ab5a130cee718d9ee7b71d9b547cf5f7f23027ed0152ad250012a8546399fcc1e12187efc68d89d6731256c4d2df7d04eef8d5c"
    }
 }
@@ -18,7 +18,7 @@
 src/web_service      @AleksandrPopovich
 src/dynarmic         @Lizzie
 src/core             @Lizzie @Maufeat @PavelBARABANOV @MrPurple666 @JPikachu
-src/core/hle         @Maufeat @PavelBARABANOV @SDK-Chan
+src/core/hle         @Maufeat @PavelBARABANOV
 src/core/arm         @Lizzie @MrPurple666
 src/*_room           @AleksandrPopovich
 src/video_core       @CamilleLaVey @MaranBr @Wildcard @weakboson
@@ -0,0 +1,41 @@
+# AddDependentPackage
+
+Use `AddDependentPackage` when you have multiple packages that are required to all be from the system, OR bundled. This is useful in cases where e.g. versions must absolutely match.
+
+## Versioning
+
+Versioning must be handled by the package itself.
+
+## Examples
+
+### Vulkan
+
+`cpmfile.json`
+
+```json
+{
+    "vulkan-headers": {
+        "repo": "KhronosGroup/Vulkan-Headers",
+        "package": "VulkanHeaders",
+        "version": "1.4.317",
+        "hash": "26e0ad8fa34ab65a91ca62ddc54cc4410d209a94f64f2817dcdb8061dc621539a4262eab6387e9b9aa421db3dbf2cf8e2a4b041b696d0d03746bae1f25191272",
+        "git_version": "1.4.342",
+        "tag": "v%VERSION%"
+    },
+    "vulkan-utility-libraries": {
+        "repo": "KhronosGroup/Vulkan-Utility-Libraries",
+        "package": "VulkanUtilityLibraries",
+        "hash": "8147370f964fd82c315d6bb89adeda30186098427bf3efaa641d36282d42a263f31e96e4586bfd7ae0410ff015379c19aa4512ba160630444d3d8553afd1ec14",
+        "git_version": "1.4.342",
+        "tag": "v%VERSION%"
+    }
+}
+```
+
+`CMakeLists.txt`:
+
+```cmake
+AddDependentPackages(vulkan-headers vulkan-utility-libraries)
+```
+
+If Vulkan Headers are installed, but NOT Vulkan Utility Libraries, then CPMUtil will throw an error.
@@ -31,6 +31,10 @@ The core of CPMUtil is the [`AddPackage`](./AddPackage.md) function. [`AddPackag

 [`AddJsonPackage`](./AddJsonPackage.md) is the recommended method of usage for CPMUtil.

+## AddDependentPackage
+
+[`AddDependentPackage`](./AddDependentPackage.md) allows you to add multiple packages such that all of them must be from the system OR bundled.
+
 ## AddQt

 [`AddQt`](./AddQt.md) adds a specific version of Qt to your project.
@@ -84,13 +84,10 @@ endif()
 AddJsonPackage(mcl)

 # Vulkan stuff
-AddJsonPackage(vulkan-headers)
-AddJsonPackage(vulkan-utility-libraries)
+AddDependentPackages(vulkan-headers vulkan-utility-libraries)

-# small hack
-if (NOT VulkanUtilityLibraries_ADDED)
-    find_package(VulkanHeaders 1.3.274 REQUIRED)
-endif()
+# frozen
+AddJsonPackage(frozen)

 # DiscordRPC
 if (USE_DISCORD_PRESENCE)
@@ -276,7 +276,7 @@
    "vulkan-headers": {
        "repo": "KhronosGroup/Vulkan-Headers",
        "package": "VulkanHeaders",
-        "version": "1.3.274",
+        "version": "1.4.317",
        "hash": "26e0ad8fa34ab65a91ca62ddc54cc4410d209a94f64f2817dcdb8061dc621539a4262eab6387e9b9aa421db3dbf2cf8e2a4b041b696d0d03746bae1f25191272",
        "git_version": "1.4.342",
        "tag": "v%VERSION%"
@@ -287,5 +287,11 @@
        "hash": "8147370f964fd82c315d6bb89adeda30186098427bf3efaa641d36282d42a263f31e96e4586bfd7ae0410ff015379c19aa4512ba160630444d3d8553afd1ec14",
        "git_version": "1.4.342",
        "tag": "v%VERSION%"
+    },
+    "frozen": {
+        "package": "frozen",
+        "repo": "serge-sans-paille/frozen",
+        "sha": "61dce5ae18",
+        "hash": "b8dfe741c82bc178dfc9749d4ab5a130cee718d9ee7b71d9b547cf5f7f23027ed0152ad250012a8546399fcc1e12187efc68d89d6731256c4d2df7d04eef8d5c"
    }
 }
@@ -24,6 +24,7 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
    RENDERER_FORCE_MAX_CLOCK("force_max_clock"),
    RENDERER_ASYNCHRONOUS_SHADERS("use_asynchronous_shaders"),
    RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"),
+    ENABLE_BUFFER_HISTORY("enable_buffer_history"),
    SYNC_MEMORY_OPERATIONS("sync_memory_operations"),
    BUFFER_REORDER_DISABLE("disable_buffer_reorder"),
    RENDERER_DEBUG("debug"),
@@ -745,6 +745,13 @@ abstract class SettingsItem(
                    descriptionId = R.string.renderer_reactive_flushing_description
                )
            )
+            put(
+                SwitchSetting(
+                    BooleanSetting.ENABLE_BUFFER_HISTORY,
+                    titleId = R.string.enable_buffer_history,
+                    descriptionId = R.string.enable_buffer_history_description
+                )
+            )
            put(
                SwitchSetting(
                    BooleanSetting.SYNC_MEMORY_OPERATIONS,
@@ -275,6 +275,7 @@ class SettingsFragmentPresenter(
            add(BooleanSetting.RENDERER_USE_DISK_SHADER_CACHE.key)
            add(BooleanSetting.RENDERER_FORCE_MAX_CLOCK.key)
            add(BooleanSetting.RENDERER_REACTIVE_FLUSHING.key)
+            add(BooleanSetting.ENABLE_BUFFER_HISTORY.key)

            add(HeaderSetting(R.string.hacks))

@@ -1,18 +1,19 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later

 #include <string>

 #include <jni.h>
-#include <common/fs/path_util.h>

 #include "android_config.h"
 #include "android_settings.h"
 #include "common/android/android_common.h"
 #include "common/android/id_cache.h"
+#include "common/fs/path_util.h"
 #include "common/logging/log.h"
 #include "common/settings.h"
 #include "frontend_common/config.h"
+#include "frontend_common/settings_generator.h"
 #include "native.h"

 std::unique_ptr<AndroidConfig> global_config;
@@ -37,6 +38,7 @@ extern "C" {

 void Java_org_yuzu_yuzu_1emu_utils_NativeConfig_initializeGlobalConfig(JNIEnv* env, jobject obj) {
    global_config = std::make_unique<AndroidConfig>();
+    FrontendCommon::GenerateSettings();
 }

 void Java_org_yuzu_yuzu_1emu_utils_NativeConfig_unloadGlobalConfig(JNIEnv* env, jobject obj) {
@@ -493,6 +493,8 @@
    <string name="renderer_force_max_clock_description">Forces the GPU to run at the maximum possible clocks (thermal constraints will still be applied).</string>
    <string name="renderer_reactive_flushing">Use reactive flushing</string>
    <string name="renderer_reactive_flushing_description">Improves rendering accuracy in some games at the cost of performance.</string>
+    <string name="enable_buffer_history">Enable buffer history</string>
+    <string name="enable_buffer_history_description">Enables access to previous buffer states. This option may improve rendering quality and performance consistency in some games.</string>


    <string name="hacks">Hacks</string>
@@ -481,6 +481,14 @@ struct Values {
    SwitchableSetting<bool> barrier_feedback_loops{linkage, true, "barrier_feedback_loops",
                                                   Category::RendererAdvanced};

+    SwitchableSetting<bool> enable_buffer_history{linkage,
+                                                  false,
+                                                  "enable_buffer_history",
+                                                  Category::RendererAdvanced,
+                                                  Specialization::Default,
+                                                  true,
+                                                  true};
+
    // Renderer Hacks //
    SwitchableSetting<GpuOverclock> fast_gpu_time{linkage,
                                                  GpuOverclock::Medium,
@@ -778,7 +786,7 @@ struct Values {
                                     Category::WebService};
    Setting<std::string> eden_username{linkage, "Eden", "eden_username",
                                       Category::WebService};
-    Setting<std::string> eden_token{linkage, "njausoolxygtpvraofqunuufhmupriifnpfggjxefntlyglr",
+    Setting<std::string> eden_token{linkage, "",
                                    "eden_token", Category::WebService};

    // Add-Ons
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later

 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
@@ -99,11 +99,6 @@ Status BufferQueueConsumer::AcquireBuffer(BufferItem* out_buffer,
        slots[slot].acquire_called = true;
        slots[slot].needs_cleanup_on_release = false;
        slots[slot].buffer_state = BufferState::Acquired;
-
-        // TODO: for now, avoid resetting the fence, so that when we next return this
-        // slot to the producer, it will wait for the fence to pass. We should fix this
-        // by properly waiting for the fence in the BufferItemConsumer.
-        // slots[slot].fence = Fence::NoFence();
    }

    // If the buffer has previously been acquired by the consumer, set graphic_buffer to nullptr to
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later

 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
@@ -17,6 +17,39 @@ BufferQueueCore::BufferQueueCore() = default;

 BufferQueueCore::~BufferQueueCore() = default;

+void BufferQueueCore::PushHistory(u64 frame_number, s64 queue_time, s64 presentation_time, BufferState state) {
+    std::lock_guard lk(buffer_history_mutex);
+
+    auto it = buffer_history_map.find(frame_number);
+    if (it != buffer_history_map.end()) {
+        it->second.state = state;
+        return;
+    }
+
+    buffer_history_map.emplace(frame_number, BufferHistoryInfo{
+        frame_number,
+        queue_time,
+        presentation_time,
+        state
+    });
+    buffer_history_order.push_back(frame_number);
+
+    if (buffer_history_order.size() > BUFFER_HISTORY_SIZE) {
+        u64 oldest_frame = buffer_history_order.front();
+        buffer_history_order.pop_front();
+        buffer_history_map.erase(oldest_frame);
+    }
+}
+
+void BufferQueueCore::UpdateHistory(u64 frame_number, BufferState state) {
+    std::lock_guard lk(buffer_history_mutex);
+
+    auto it = buffer_history_map.find(frame_number);
+    if (it != buffer_history_map.end()) {
+        it->second.state = state;
+    }
+}
+
 void BufferQueueCore::SignalDequeueCondition() {
    dequeue_possible.store(true);
    dequeue_condition.notify_all();
@@ -30,7 +63,6 @@ bool BufferQueueCore::WaitForDequeueCondition(std::unique_lock<std::mutex>& lk)
 }

 s32 BufferQueueCore::GetMinUndequeuedBufferCountLocked(bool async) const {
-    // If DequeueBuffer is allowed to error out, we don't have to add an extra buffer.
    if (!use_async_buffer) {
        return 0;
    }
@@ -55,8 +87,6 @@ s32 BufferQueueCore::GetMaxBufferCountLocked(bool async) const {
        return override_max_buffer_count;
    }

-    // Any buffers that are dequeued by the producer or sitting in the queue waiting to be consumed
-    // need to have their slots preserved.
    for (s32 slot = max_buffer_count; slot < BufferQueueDefs::NUM_BUFFER_SLOTS; ++slot) {
        const auto state = slots[slot].buffer_state;
        if (state == BufferState::Queued || state == BufferState::Dequeued) {
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later

 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
@@ -10,20 +10,31 @@
 #pragma once

 #include <condition_variable>
+#include <deque>
 #include <list>
 #include <memory>
 #include <mutex>
 #include <set>
 #include <vector>
+#include <unordered_map>
+#include <algorithm>

 #include "core/hle/service/nvnflinger/buffer_item.h"
 #include "core/hle/service/nvnflinger/buffer_queue_defs.h"
+#include "core/hle/service/nvnflinger/buffer_slot.h"
 #include "core/hle/service/nvnflinger/pixel_format.h"
 #include "core/hle/service/nvnflinger/status.h"
 #include "core/hle/service/nvnflinger/window.h"

 namespace Service::android {

+struct BufferHistoryInfo {
+    u64 frame_number{};
+    s64 queue_time{};
+    s64 presentation_time{};
+    BufferState state{};
+};
+
 class IConsumerListener;
 class IProducerListener;

@@ -33,10 +44,14 @@ class BufferQueueCore final {

 public:
    static constexpr s32 INVALID_BUFFER_SLOT = BufferItem::INVALID_BUFFER_SLOT;
+    static constexpr u32 BUFFER_HISTORY_SIZE = 8;

    BufferQueueCore();
    ~BufferQueueCore();

+    void PushHistory(u64 frame_number, s64 queue_time, s64 presentation_time, BufferState state);
+    void UpdateHistory(u64 frame_number, BufferState state);
+
 private:
    void SignalDequeueCondition();
    bool WaitForDequeueCondition(std::unique_lock<std::mutex>& lk);
@@ -72,6 +87,11 @@ private:
    const s32 max_acquired_buffer_count{}; // This is always zero on HOS
    bool buffer_has_been_queued{};
    u64 frame_counter{};
+
+    std::unordered_map<u64, BufferHistoryInfo> buffer_history_map{};
+    mutable std::mutex buffer_history_mutex{};
+    std::deque<u64> buffer_history_order;
+
    u32 transform_hint{};
    bool is_allocating{};
    mutable std::condition_variable_any is_allocating_condition;
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later

 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
@@ -9,6 +9,7 @@

 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "common/settings.h"
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/kernel/k_readable_event.h"
 #include "core/hle/kernel/kernel.h"
@@ -26,7 +27,7 @@ BufferQueueProducer::BufferQueueProducer(Service::KernelHelpers::ServiceContext&
                                         std::shared_ptr<BufferQueueCore> buffer_queue_core_,
                                         Service::Nvidia::NvCore::NvMap& nvmap_)
    : service_context{service_context_}, core{std::move(buffer_queue_core_)}, slots(core->slots),
-      nvmap(nvmap_) {
+      clock{Common::CreateOptimalClock()}, nvmap(nvmap_) {
    buffer_wait_event = service_context.CreateEvent("BufferQueue:WaitEvent");
 }

@@ -428,8 +429,7 @@ Status BufferQueueProducer::AttachBuffer(s32* out_slot,
    return return_flags;
 }

-Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
-                                        QueueBufferOutput* output) {
+Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input, QueueBufferOutput* output) {
    s64 timestamp{};
    bool is_auto_timestamp{};
    Common::Rectangle<s32> crop;
@@ -440,8 +440,7 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
    s32 swap_interval{};
    Fence fence{};

-    input.Deflate(&timestamp, &is_auto_timestamp, &crop, &scaling_mode, &transform,
-                  &sticky_transform_, &async, &swap_interval, &fence);
+    input.Deflate(&timestamp, &is_auto_timestamp, &crop, &scaling_mode, &transform, &sticky_transform_, &async, &swap_interval, &fence);

    switch (scaling_mode) {
    case NativeWindowScalingMode::Freeze:
@@ -455,10 +454,9 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
        return Status::BadValue;
    }

-    std::shared_ptr<IConsumerListener> frame_available_listener;
-    std::shared_ptr<IConsumerListener> frame_replaced_listener;
-    s32 callback_ticket{};
    BufferItem item;
+    std::shared_ptr<IConsumerListener> listener_available;
+    std::shared_ptr<IConsumerListener> listener_replaced;

    {
        std::scoped_lock lock{core->mutex};
@@ -469,127 +467,82 @@ Status BufferQueueProducer::QueueBuffer(s32 slot, const QueueBufferInput& input,
        }

        const s32 max_buffer_count = core->GetMaxBufferCountLocked(async);
-        if (async && core->override_max_buffer_count) {
-            if (core->override_max_buffer_count < max_buffer_count) {
-                LOG_ERROR(Service_Nvnflinger, "async mode is invalid with "
-                                              "buffer count override");
-                return Status::BadValue;
-            }
-        }

        if (slot < 0 || slot >= max_buffer_count) {
-            LOG_ERROR(Service_Nvnflinger, "slot index {} out of range [0, {})", slot,
-                      max_buffer_count);
+            LOG_ERROR(Service_Nvnflinger, "slot {} out of range [0, {})", slot, max_buffer_count);
            return Status::BadValue;
-        } else if (slots[slot].buffer_state != BufferState::Dequeued) {
-            LOG_ERROR(Service_Nvnflinger,
-                      "slot {} is not owned by the producer "
-                      "(state = {})",
-                      slot, slots[slot].buffer_state);
+        }
+        if (slots[slot].buffer_state != BufferState::Dequeued) {
+            LOG_ERROR(Service_Nvnflinger, "slot {} is not owned by producer", slot);
            return Status::BadValue;
-        } else if (!slots[slot].request_buffer_called) {
-            LOG_ERROR(Service_Nvnflinger,
-                      "slot {} was queued without requesting "
-                      "a buffer",
-                      slot);
+        }
+        if (!slots[slot].request_buffer_called) {
+            LOG_ERROR(Service_Nvnflinger, "slot {} was queued without request", slot);
            return Status::BadValue;
        }

-        LOG_DEBUG(Service_Nvnflinger,
-                  "slot={} frame={} time={} crop=[{},{},{},{}] transform={} scale={}", slot,
-                  core->frame_counter + 1, timestamp, crop.Left(), crop.Top(), crop.Right(),
-                  crop.Bottom(), transform, scaling_mode);
-
-        const std::shared_ptr<GraphicBuffer>& graphic_buffer(slots[slot].graphic_buffer);
-        Common::Rectangle<s32> buffer_rect(graphic_buffer->Width(), graphic_buffer->Height());
-        Common::Rectangle<s32> cropped_rect;
-        [[maybe_unused]] const bool unused = crop.Intersect(buffer_rect, &cropped_rect);
-
-        if (cropped_rect != crop) {
-            LOG_ERROR(Service_Nvnflinger, "crop rect is not contained within the buffer in slot {}",
-                      slot);
-            return Status::BadValue;
-        }
-
-        slots[slot].fence = fence;
-        slots[slot].buffer_state = BufferState::Queued;
        ++core->frame_counter;
+        slots[slot].buffer_state = BufferState::Queued;
        slots[slot].frame_number = core->frame_counter;
+        slots[slot].queue_time = timestamp;
+        slots[slot].presentation_time = clock->GetTimeNS().count();
+        slots[slot].fence = fence;

-        item.acquire_called = slots[slot].acquire_called;
+        item.slot = slot;
        item.graphic_buffer = slots[slot].graphic_buffer;
-        item.crop = crop;
-        item.transform = transform & ~NativeWindowTransform::InverseDisplay;
-        item.transform_to_display_inverse =
-            (transform & NativeWindowTransform::InverseDisplay) != NativeWindowTransform::None;
-        item.scaling_mode = static_cast<u32>(scaling_mode);
+        item.frame_number = core->frame_counter;
        item.timestamp = timestamp;
        item.is_auto_timestamp = is_auto_timestamp;
-        item.frame_number = core->frame_counter;
-        item.slot = slot;
+        item.crop = crop;
+        item.transform = transform & ~NativeWindowTransform::InverseDisplay;
+        item.transform_to_display_inverse = (transform & NativeWindowTransform::InverseDisplay) != NativeWindowTransform::None;
+        item.scaling_mode = static_cast<u32>(scaling_mode);
        item.fence = fence;
        item.is_droppable = core->dequeue_buffer_cannot_block || async;
        item.swap_interval = swap_interval;
+        item.acquire_called = slots[slot].acquire_called;

        sticky_transform = sticky_transform_;

        if (core->queue.empty()) {
-            // When the queue is empty, we can simply queue this buffer
            core->queue.push_back(item);
-            frame_available_listener = core->consumer_listener;
+            listener_available = core->consumer_listener;
        } else {
-            // When the queue is not empty, we need to look at the front buffer
-            // state to see if we need to replace it
-            auto front(core->queue.begin());
+            auto front = core->queue.begin();
+            if (front->is_droppable && core->StillTracking(*front)) {
+                slots[front->slot].buffer_state = BufferState::Free;
+                if (Settings::values.enable_buffer_history.GetValue()) {
+                    core->UpdateHistory(front->frame_number, BufferState::Free);
+                }
+                slots[front->slot].frame_number = 0;
+            }

            if (front->is_droppable) {
-                // If the front queued buffer is still being tracked, we first
-                // mark it as freed
-                if (core->StillTracking(*front)) {
-                    slots[front->slot].buffer_state = BufferState::Free;
-                    // Reset the frame number of the freed buffer so that it is the first in line to
-                    // be dequeued again
-                    slots[front->slot].frame_number = 0;
-                }
-                // Overwrite the droppable buffer with the incoming one
                *front = item;
-                frame_replaced_listener = core->consumer_listener;
+                listener_replaced = core->consumer_listener;
            } else {
                core->queue.push_back(item);
-                frame_available_listener = core->consumer_listener;
+                listener_available = core->consumer_listener;
            }
        }

+        if (Settings::values.enable_buffer_history.GetValue()) {
+            core->PushHistory(core->frame_counter, slots[slot].queue_time, slots[slot].presentation_time, BufferState::Queued);
+        }
+
        core->buffer_has_been_queued = true;
        core->SignalDequeueCondition();
-        output->Inflate(core->default_width, core->default_height, core->transform_hint,
-                        static_cast<u32>(core->queue.size()));

-        // Take a ticket for the callback functions
-        callback_ticket = next_callback_ticket++;
+        output->Inflate(core->default_width, core->default_height, core->transform_hint, static_cast<u32>(core->queue.size()));
    }

-    // Don't send the GraphicBuffer through the callback, and don't send the slot number, since the
-    // consumer shouldn't need it
    item.graphic_buffer.reset();
    item.slot = BufferItem::INVALID_BUFFER_SLOT;

-    // Call back without the main BufferQueue lock held, but with the callback lock held so we can
-    // ensure that callbacks occur in order
-    {
-        std::scoped_lock lock{callback_mutex};
-        while (callback_ticket != current_callback_ticket) {
-            callback_condition.wait(callback_mutex);
-        }
-
-        if (frame_available_listener != nullptr) {
-            frame_available_listener->OnFrameAvailable(item);
-        } else if (frame_replaced_listener != nullptr) {
-            frame_replaced_listener->OnFrameReplaced(item);
-        }
-
-        ++current_callback_ticket;
-        callback_condition.notify_all();
+    if (listener_available) {
+        listener_available->OnFrameAvailable(item);
+    } else if (listener_replaced) {
+        listener_replaced->OnFrameReplaced(item);
    }

    return Status::NoError;
@@ -810,6 +763,10 @@ Status BufferQueueProducer::SetPreallocatedBuffer(s32 slot,
    return Status::NoError;
 }

+Kernel::KReadableEvent* BufferQueueProducer::GetNativeHandle(u32 type_id) {
+    return &buffer_wait_event->GetReadableEvent();
+}
+
 void BufferQueueProducer::Transact(u32 code, std::span<const u8> parcel_data,
                                   std::span<u8> parcel_reply, u32 flags) {
    // Values used by BnGraphicBufferProducer onTransact
@@ -929,9 +886,42 @@ void BufferQueueProducer::Transact(u32 code, std::span<const u8> parcel_data,
        status = SetBufferCount(buffer_count);
        break;
    }
-    case TransactionId::GetBufferHistory:
-        LOG_DEBUG(Service_Nvnflinger, "(STUBBED) called, transaction=GetBufferHistory");
+    case TransactionId::GetBufferHistory: {
+        if (!Settings::values.enable_buffer_history.GetValue()) {
+            LOG_DEBUG(Service_Nvnflinger, "(STUBBED) called");
+            break;
+        }
+
+        LOG_DEBUG(Service_Nvnflinger, "called, transaction=GetBufferHistory");
+
+        const s32 request = parcel_in.Read<s32>();
+        if (request <= 0) {
+            parcel_out.Write(Status::BadValue);
+            parcel_out.Write<s32>(0);
+            break;
+        }
+
+        std::vector<BufferHistoryInfo> snapshot;
+
+        {
+            std::scoped_lock lk(core->buffer_history_mutex);
+            for (auto& [frame, info] : core->buffer_history_map) {
+                snapshot.push_back(info);
+            }
+        }
+
+        std::sort(snapshot.begin(), snapshot.end(), [](auto& a, auto& b){
+            return a.frame_number > b.frame_number;
+        });
+
+        const s32 limit = std::min(request, (s32)snapshot.size());
+        parcel_out.Write(Status::NoError);
+        parcel_out.Write<s32>(limit);
+        for (s32 i = 0; i < limit; ++i) {
+            parcel_out.Write(snapshot[i]);
+        }
        break;
+    }
    default:
        ASSERT_MSG(false, "Unimplemented TransactionId {}", code);
        break;
@@ -944,8 +934,4 @@ void BufferQueueProducer::Transact(u32 code, std::span<const u8> parcel_data,
                (std::min)(parcel_reply.size(), serialized.size()));
 }

-Kernel::KReadableEvent* BufferQueueProducer::GetNativeHandle(u32 type_id) {
-    return &buffer_wait_event->GetReadableEvent();
-}
-
 } // namespace Service::android
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later

 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
@@ -14,6 +14,7 @@
 #include <mutex>

 #include "common/common_funcs.h"
+#include "common/wall_clock.h"
 #include "core/hle/service/nvdrv/nvdata.h"
 #include "core/hle/service/nvnflinger/binder.h"
 #include "core/hle/service/nvnflinger/buffer_queue_defs.h"
@@ -88,6 +89,7 @@ private:
    s32 next_callback_ticket{};
    s32 current_callback_ticket{};
    std::condition_variable_any callback_condition;
+    std::unique_ptr<Common::WallClock> clock;

    Service::Nvidia::NvCore::NvMap& nvmap;
 };
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later

 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
@@ -37,6 +37,7 @@ struct BufferSlot final {
    bool needs_cleanup_on_release{};
    bool attached_by_consumer{};
    bool is_preallocated{};
+    s64 queue_time{}, presentation_time{};
 };

 } // namespace Service::android
@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+# SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
 # SPDX-License-Identifier: GPL-3.0-or-later

 # SPDX-FileCopyrightText: 2023 yuzu Emulator Project
@@ -12,7 +12,8 @@ add_library(frontend_common STATIC
    firmware_manager.cpp
    data_manager.h data_manager.cpp
    play_time_manager.cpp
-    play_time_manager.h)
+    play_time_manager.h
+    settings_generator.h settings_generator.cpp)

 if (ENABLE_UPDATE_CHECKER)
    target_link_libraries(frontend_common PRIVATE httplib::httplib)
@@ -29,4 +30,6 @@ if (ENABLE_UPDATE_CHECKER)
 endif()

 create_target_directory_groups(frontend_common)
-target_link_libraries(frontend_common PUBLIC core SimpleIni::SimpleIni PRIVATE common Boost::headers)
+target_link_libraries(frontend_common
+    PUBLIC core SimpleIni::SimpleIni frozen::frozen-headers
+    PRIVATE common Boost::headers)
@@ -0,0 +1,31 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include <random>
+#include <frozen/string.h>
+#include "common/settings.h"
+#include "settings_generator.h"
+
+namespace FrontendCommon {
+
+void GenerateSettings() {
+    static std::random_device rd;
+
+    // Web Token //
+    auto &token_setting = Settings::values.eden_token;
+    if (token_setting.GetValue().empty()) {
+        static constexpr const size_t token_length = 48;
+        static constexpr const frozen::string token_set = "abcdefghijklmnopqrstuvwxyz";
+        static std::uniform_int_distribution<int> token_dist(0, token_set.size() - 1);
+        std::string result;
+
+        for (size_t i = 0; i < token_length; ++i) {
+            size_t idx = token_dist(rd);
+            result += token_set[idx];
+        }
+
+        token_setting.SetValue(result);
+    }
+}
+
+}
@@ -0,0 +1,14 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+namespace FrontendCommon {
+
+/**
+ * @brief GenerateSettings Generate platform-specific or randomized settings.
+ * Run this function at initialization time for your frontend.
+ */
+void GenerateSettings();
+
+}
@@ -78,12 +78,9 @@ target_compile_definitions(qt_common PUBLIC
    QT_NO_URL_CAST_FROM_STRING
 )

-# pass targets
-find_package(frozen)
-
 target_link_libraries(qt_common PRIVATE core Qt6::Core Qt6::Concurrent SimpleIni::SimpleIni QuaZip::QuaZip)
 target_link_libraries(qt_common PUBLIC frozen::frozen-headers)
-target_link_libraries(qt_common PRIVATE gamemode::headers)
+target_link_libraries(qt_common PRIVATE gamemode::headers frontend_common)

 if (NOT APPLE AND ENABLE_OPENGL)
    target_compile_definitions(qt_common PUBLIC HAS_OPENGL)
@@ -329,6 +329,10 @@ std::unique_ptr<TranslationMap> InitializeTranslations(QObject* parent)
           barrier_feedback_loops,
           tr("Barrier feedback loops"),
           tr("Improves rendering of transparency effects in specific games."));
+    INSERT(Settings,
+           enable_buffer_history,
+           tr("Enable buffer history"),
+           tr("Enables access to previous buffer states.\nThis option may improve rendering quality and performance consistency in some games."));
    INSERT(Settings,
           fix_bloom_effects,
           tr("Fix bloom effects"),
@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later

@@ -148,10 +151,13 @@ Id EmitConvertU32U64(EmitContext& ctx, Id value) {
 }

 Id EmitConvertF16F32(EmitContext& ctx, Id value) {
+#ifdef ANDROID
+    return ctx.OpFConvert(ctx.F16[1], value);
+#else
    const auto result = ctx.OpFConvert(ctx.F16[1], value);
    const auto isOverflowing = ctx.OpIsNan(ctx.U1, result);
    return ctx.OpSelect(ctx.F16[1], isOverflowing, ctx.Constant(ctx.F16[1], 0), result);
-    //return ctx.OpFConvert(ctx.F16[1], value);
+#endif
 }

 Id EmitConvertF32F16(EmitContext& ctx, Id value) {
@@ -36,12 +36,6 @@ struct EncodingData {
    uint data;
 };

-struct PartitionTable {
-    uint s1, s2, s3, s4, s5, s6, s7, s8;
-    uint rnum;
-    bool small_block;
-};
-
 layout(binding = BINDING_INPUT_BUFFER, std430) readonly restrict buffer InputBufferU32 {
    uvec4 astc_data[];
 };
@@ -68,40 +62,26 @@ const uint encoding_values[22] = uint[](
    (QUINT | (4u << 8u)), (TRIT | (5u << 8u)), (JUST_BITS | (7u << 8u)), (QUINT | (5u << 8u)),
    (TRIT | (6u << 8u)), (JUST_BITS | (8u << 8u)));

-// Shared memory for workgroup processing
-shared uvec4 local_buff;
-shared int total_bitsread;
+// Input ASTC texture globals
+int total_bitsread = 0;
+uvec4 local_buff;

 // Color data globals
-shared uvec4 color_endpoint_data;
-shared int color_bitsread;
+uvec4 color_endpoint_data;
+int color_bitsread = 0;

 // Global "vector" to be pushed into when decoding
+// At most will require BLOCK_WIDTH x BLOCK_HEIGHT in single plane mode
+// At most will require BLOCK_WIDTH x BLOCK_HEIGHT x 2 in dual plane mode
+// So the maximum would be 144 (12 x 12) elements, x 2 for two planes
 #define DIVCEIL(number, divisor) (number + divisor - 1) / divisor
 #define ARRAY_NUM_ELEMENTS 144
 #define VECTOR_ARRAY_SIZE DIVCEIL(ARRAY_NUM_ELEMENTS * 2, 4)
-shared uint result_vector[ARRAY_NUM_ELEMENTS * 2];
+uint result_vector[ARRAY_NUM_ELEMENTS * 2];

-shared int result_index;
-shared uint result_vector_max_index;
-shared bool result_limit_reached;
-
-// avoid intermediate result_vector storage during color decode phase
-shared bool write_color_values;
-shared uint color_values_direct[32];
-shared uint color_out_index;
-shared uint color_num_values;
-
-// Shared variables for DecompressBlock interthread communication
-shared uvec4 endpoints0[4];
-shared uvec4 endpoints1[4];
-shared PartitionTable pt;
-shared uvec2 size_params;
-shared uint num_partitions;
-shared uint partition_index;
-shared uint plane_index;
-shared bool dual_plane;
-shared vec4 fill_color;
+int result_index = 0;
+uint result_vector_max_index;
+bool result_limit_reached = false;

 // EncodingData helpers
 uint Encoding(EncodingData val) {
@@ -134,110 +114,9 @@ EncodingData CreateEncodingData(uint encoding, uint num_bits, uint bit_val, uint
    return EncodingData(((encoding) << 0u) | ((num_bits) << 8u) |
                        ((bit_val) << 16u) | ((quint_trit_val) << 24u));
 }
-uint ReplicateBitTo9(uint bit);
-uint FastReplicateTo8(uint value, uint num_bits);
-
-void EmitColorValue(EncodingData val) {
-    // write directly to color_values_direct[]
-    const uint encoding = Encoding(val);
-    const uint bitlen = NumBits(val);
-    const uint bitval = BitValue(val);
-
-    if (encoding == JUST_BITS) {
-        color_values_direct[++color_out_index] = FastReplicateTo8(bitval, bitlen);
-        return;
-    }
-
-    uint A = ReplicateBitTo9((bitval & 1));
-    uint B = 0, C = 0, D = QuintTritValue(val);
-
-    if (encoding == TRIT) {
-        switch (bitlen) {
-            case 1:
-            C = 204;
-            break;
-            case 2: {
-                C = 93;
-                const uint b = (bitval >> 1) & 1;
-                B = (b << 8) | (b << 4) | (b << 2) | (b << 1);
-                break;
-            }
-            case 3: {
-                C = 44;
-                const uint cb = (bitval >> 1) & 3;
-                B = (cb << 7) | (cb << 2) | cb;
-                break;
-            }
-            case 4: {
-                C = 22;
-                const uint dcb = (bitval >> 1) & 7;
-                B = (dcb << 6) | dcb;
-                break;
-            }
-            case 5: {
-                C = 11;
-                const uint edcb = (bitval >> 1) & 0xF;
-                B = (edcb << 5) | (edcb >> 2);
-                break;
-            }
-            case 6: {
-                C = 5;
-                const uint fedcb = (bitval >> 1) & 0x1F;
-                B = (fedcb << 4) | (fedcb >> 4);
-                break;
-            }
-        }
-    } else { // QUINT
-        switch (bitlen) {
-            case 1:
-            C = 113;
-            break;
-            case 2: {
-                C = 54;
-                const uint b = (bitval >> 1) & 1;
-                B = (b << 8) | (b << 3) | (b << 2);
-                break;
-            }
-            case 3: {
-                C = 26;
-                const uint cb = (bitval >> 1) & 3;
-                B = (cb << 7) | (cb << 1) | (cb >> 1);
-                break;
-            }
-            case 4: {
-                C = 13;
-                const uint dcb = (bitval >> 1) & 7;
-                B = (dcb << 6) | (dcb >> 1);
-                break;
-            }
-            case 5: {
-                C = 6;
-                const uint edcb = (bitval >> 1) & 0xF;
-                B = (edcb << 5) | (edcb >> 3);
-                break;
-            }
-        }
-    }
-
-    uint T = (D * C) + B;
-    T ^= A;
-    T = (A & 0x80) | (T >> 2);
-    color_values_direct[++color_out_index] = T;
-}
-


 void ResultEmplaceBack(EncodingData val) {
-    if (write_color_values) {
-        if (color_out_index >= color_num_values) {
-            // avoid decoding more than needed by this phase
-            result_limit_reached = true;
-            return;
-        }
-        EmitColorValue(val);
-        return;
-    }
-
    if (result_index >= result_vector_max_index) {
        // Alert callers to avoid decoding more than needed by this phase
        result_limit_reached = true;
@@ -318,31 +197,32 @@ uint Hash52(uint p) {
    return p;
 }

-
-PartitionTable GetPartitionTable(uint seed, uint partition_count) {
-    PartitionTable pt;
-    pt.small_block = (block_dims.y * block_dims.x) < 32;
+uint Select2DPartition(uint seed, uint x, uint y, uint partition_count) {
+    if ((block_dims.y * block_dims.x) < 32) {
+        x <<= 1;
+        y <<= 1;
+    }

    seed += (partition_count - 1) * 1024;
-    uint rnum = Hash52(uint(seed));
-    pt.rnum = rnum;

-    uint seed1 = (rnum & 0xF);
-    seed1 *= seed1;
-    uint seed2 = (rnum >> 4) & 0xF;
-    seed2 *= seed2;
-    uint seed3 = (rnum >> 8) & 0xF;
-    seed3 *= seed3;
-    uint seed4 = (rnum >> 12) & 0xF;
-    seed4 *= seed4;
-    uint seed5 = (rnum >> 16) & 0xF;
-    seed5 *= seed5;
-    uint seed6 = (rnum >> 20) & 0xF;
-    seed6 *= seed6;
-    uint seed7 = (rnum >> 24) & 0xF;
-    seed7 *= seed7;
-    uint seed8 = (rnum >> 28) & 0xF;
-    seed8 *= seed8;
+    const uint rnum = Hash52(uint(seed));
+    uint seed1 = uint(rnum & 0xF);
+    uint seed2 = uint((rnum >> 4) & 0xF);
+    uint seed3 = uint((rnum >> 8) & 0xF);
+    uint seed4 = uint((rnum >> 12) & 0xF);
+    uint seed5 = uint((rnum >> 16) & 0xF);
+    uint seed6 = uint((rnum >> 20) & 0xF);
+    uint seed7 = uint((rnum >> 24) & 0xF);
+    uint seed8 = uint((rnum >> 28) & 0xF);
+
+    seed1 = (seed1 * seed1);
+    seed2 = (seed2 * seed2);
+    seed3 = (seed3 * seed3);
+    seed4 = (seed4 * seed4);
+    seed5 = (seed5 * seed5);
+    seed6 = (seed6 * seed6);
+    seed7 = (seed7 * seed7);
+    seed8 = (seed8 * seed8);

    uint sh1, sh2;
    if ((seed & 1) > 0) {
@@ -352,37 +232,31 @@ PartitionTable GetPartitionTable(uint seed, uint partition_count) {
        sh1 = (partition_count == 3) ? 6 : 5;
        sh2 = (seed & 2) > 0 ? 4 : 5;
    }
+    seed1 >>= sh1;
+    seed2 >>= sh2;
+    seed3 >>= sh1;
+    seed4 >>= sh2;
+    seed5 >>= sh1;
+    seed6 >>= sh2;
+    seed7 >>= sh1;
+    seed8 >>= sh2;

-    pt.s1 = seed1 >> sh1;
-    pt.s2 = seed2 >> sh2;
-    pt.s3 = seed3 >> sh1;
-    pt.s4 = seed4 >> sh2;
-    pt.s5 = seed5 >> sh1;
-    pt.s6 = seed6 >> sh2;
-    pt.s7 = seed7 >> sh1;
-    pt.s8 = seed8 >> sh2;
-
-    return pt;
-    }
-
-uint SelectPartition(PartitionTable pt, uint x, uint y, uint partition_count) {
-    if (pt.small_block) {
-        x <<= 1;
-        y <<= 1;
-    }
-
-    uint a = pt.s1 * x + pt.s2 * y + (pt.rnum >> 14);
-    uint b = pt.s3 * x + pt.s4 * y + (pt.rnum >> 10);
-    uint c = pt.s5 * x + pt.s6 * y + (pt.rnum >> 6);
-    uint d = pt.s7 * x + pt.s8 * y + (pt.rnum >> 2);
+    uint a = seed1 * x + seed2 * y + (rnum >> 14);
+    uint b = seed3 * x + seed4 * y + (rnum >> 10);
+    uint c = seed5 * x + seed6 * y + (rnum >> 6);
+    uint d = seed7 * x + seed8 * y + (rnum >> 2);

    a &= 0x3F;
    b &= 0x3F;
    c &= 0x3F;
    d &= 0x3F;

-    if (partition_count < 4) d = 0;
-    if (partition_count < 3) c = 0;
+    if (partition_count < 4) {
+        d = 0;
+    }
+    if (partition_count < 3) {
+        c = 0;
+    }

    if (a >= b && a >= c && a >= d) {
        return 0;
@@ -583,7 +457,7 @@ void DecodeIntegerSequence(uint max_range, uint num_values) {
    }
 }

-void DecodeColorValues(uvec4 modes, uint num_partitions, uint color_data_bits) {
+void DecodeColorValues(uvec4 modes, uint num_partitions, uint color_data_bits, out uint color_values[32]) {
    uint num_values = 0;
    for (uint i = 0; i < num_partitions; i++) {
        num_values += ((modes[i] >> 2) + 1) << 1;
@@ -597,21 +471,104 @@ void DecodeColorValues(uvec4 modes, uint num_partitions, uint color_data_bits) {
            break;
        }
    }
-    // Decode directly into color_values_direct[]
-    write_color_values = true;
-    color_out_index = 0;
-    color_num_values = num_values;
-    for (uint i = 0; i < 32; ++i) {
-        color_values_direct[i] = 0;
-    }
-
    DecodeIntegerSequence(range - 1, num_values);
-
-    write_color_values = false;
+    uint out_index = 0;
+    for (int itr = 0; itr < result_index; ++itr) {
+        if (out_index >= num_values) {
+            break;
+        }
+        const EncodingData val = GetEncodingFromVector(itr);
+        const uint encoding = Encoding(val);
+        const uint bitlen = NumBits(val);
+        const uint bitval = BitValue(val);
+        uint A = 0, B = 0, C = 0, D = 0;
+        A = ReplicateBitTo9((bitval & 1));
+        switch (encoding) {
+        case JUST_BITS:
+            color_values[++out_index] = FastReplicateTo8(bitval, bitlen);
+            break;
+        case TRIT: {
+            D = QuintTritValue(val);
+            switch (bitlen) {
+            case 1:
+                C = 204;
+                break;
+            case 2: {
+                C = 93;
+                const uint b = (bitval >> 1) & 1;
+                B = (b << 8) | (b << 4) | (b << 2) | (b << 1);
+                break;
+            }
+            case 3: {
+                C = 44;
+                const uint cb = (bitval >> 1) & 3;
+                B = (cb << 7) | (cb << 2) | cb;
+                break;
+            }
+            case 4: {
+                C = 22;
+                const uint dcb = (bitval >> 1) & 7;
+                B = (dcb << 6) | dcb;
+                break;
+            }
+            case 5: {
+                C = 11;
+                const uint edcb = (bitval >> 1) & 0xF;
+                B = (edcb << 5) | (edcb >> 2);
+                break;
+            }
+            case 6: {
+                C = 5;
+                const uint fedcb = (bitval >> 1) & 0x1F;
+                B = (fedcb << 4) | (fedcb >> 4);
+                break;
+            }
+            }
+            break;
+        }
+        case QUINT: {
+            D = QuintTritValue(val);
+            switch (bitlen) {
+            case 1:
+                C = 113;
+                break;
+            case 2: {
+                C = 54;
+                const uint b = (bitval >> 1) & 1;
+                B = (b << 8) | (b << 3) | (b << 2);
+                break;
+            }
+            case 3: {
+                C = 26;
+                const uint cb = (bitval >> 1) & 3;
+                B = (cb << 7) | (cb << 1) | (cb >> 1);
+                break;
+            }
+            case 4: {
+                C = 13;
+                const uint dcb = (bitval >> 1) & 7;
+                B = (dcb << 6) | (dcb >> 1);
+                break;
+            }
+            case 5: {
+                C = 6;
+                const uint edcb = (bitval >> 1) & 0xF;
+                B = (edcb << 5) | (edcb >> 3);
+                break;
+            }
+            }
+            break;
+        }
+        }
+        if (encoding != JUST_BITS) {
+            uint T = (D * C) + B;
+            T ^= A;
+            T = (A & 0x80) | (T >> 2);
+            color_values[++out_index] = T;
+        }
+    }
 }

-
-
 ivec2 BitTransferSigned(int a, int b) {
    ivec2 transferred;
    transferred.y = b >> 1;
@@ -773,7 +730,7 @@ uint UnquantizeTexelWeight(EncodingData val) {
    uint encoding = Encoding(val), bitlen = NumBits(val), bitval = BitValue(val);
    if (encoding == JUST_BITS) {
        return (bitlen >= 1 && bitlen <= 5)
-            ? ((bitval * 64) + ((1 << bitlen) - 1) / 2) / ((1 << bitlen) - 1)
+            ? uint(floor(0.5f + float(bitval) * 64.0f / float((1 << bitlen) - 1)))
            : FastReplicateTo6(bitval, bitlen);
    } else if (encoding == TRIT || encoding == QUINT) {
        uint B = 0, C = 0, D = 0;
@@ -907,32 +864,27 @@ int FindLayout(uint mode) {


 void FillError(ivec3 coord) {
-    const uint total_texels = block_dims.x * block_dims.y;
-    for (uint tid = gl_LocalInvocationIndex; tid < total_texels; tid += gl_WorkGroupSize.x * gl_WorkGroupSize.y) {
-        uint x = tid % block_dims.x;
-        uint y = tid / block_dims.x;
-        imageStore(dest_image, coord + ivec3(x, y, 0), vec4(0.0, 0.0, 0.0, 0.0));
+    for (uint j = 0; j < block_dims.y; j++) {
+        for (uint i = 0; i < block_dims.x; i++) {
+            imageStore(dest_image, coord + ivec3(i, j, 0), vec4(0.0, 0.0, 0.0, 0.0));
+        }
    }
 }

 void FillVoidExtentLDR(ivec3 coord) {
-    // Thread 0 decodes color
-
-    if (gl_LocalInvocationIndex == 0) {
-        SkipBits(52);
-        const uint r_u = StreamBits(16);
-        const uint g_u = StreamBits(16);
-        const uint b_u = StreamBits(16);
-        const uint a_u = StreamBits(16);
-        fill_color = vec4(float(r_u) / 65535.0f, float(g_u) / 65535.0f, float(b_u) / 65535.0f, float(a_u) / 65535.0f);
-    }
-    barrier();
-
-    const uint total_texels = block_dims.x * block_dims.y;
-    for (uint tid = gl_LocalInvocationIndex; tid < total_texels; tid += gl_WorkGroupSize.x * gl_WorkGroupSize.y) {
-        uint x = tid % block_dims.x;
-        uint y = tid / block_dims.x;
-        imageStore(dest_image, coord + ivec3(x, y, 0), fill_color);
+    SkipBits(52);
+    const uint r_u = StreamBits(16);
+    const uint g_u = StreamBits(16);
+    const uint b_u = StreamBits(16);
+    const uint a_u = StreamBits(16);
+    const float a = float(a_u) / 65535.0f;
+    const float r = float(r_u) / 65535.0f;
+    const float g = float(g_u) / 65535.0f;
+    const float b = float(b_u) / 65535.0f;
+    for (uint j = 0; j < block_dims.y; j++) {
+        for (uint i = 0; i < block_dims.x; i++) {
+            imageStore(dest_image, coord + ivec3(i, j, 0), vec4(r, g, b, a));
+        }
    }
 }

@@ -1014,156 +966,160 @@ uint DecodeMaxWeight(uint mode) {
 }

 void DecompressBlock(ivec3 coord) {
-    if (gl_LocalInvocationIndex == 0) {
-        uint mode = StreamBits(11);
-        bool early_exit = false;
-        if (IsError(mode)) {
-            size_params = uvec2(0);
-            early_exit = true;
-        } else if ((mode & 0x1ff) == 0x1fc) {
-            size_params = uvec2(0xFFFFFFFF);
-            early_exit = true;
-        } else {
-            size_params = DecodeBlockSize(mode);
-            if ((size_params.x > block_dims.x) || (size_params.y > block_dims.y)) {
-               size_params = uvec2(0);
-               early_exit = true;
-            }
-        }
-
-        if (!early_exit) {
-            num_partitions = StreamBits(2) + 1;
-            uint mode_layout = FindLayout(mode);
-            dual_plane = (mode_layout != 9) && ((mode & 0x400) != 0);
-            if (num_partitions > 4 || (num_partitions == 4 && dual_plane)) {
-                size_params = uvec2(0);
-                early_exit = true;
-            }
-        }
-
-        if (!early_exit) {
-            uint partition_index_local = 1;
-            uvec4 color_endpoint_mode = uvec4(0);
-            uint ced_pointer = 0;
-            uint base_cem = 0;
-            if (num_partitions == 1) {
-                color_endpoint_mode.x = StreamBits(4);
-                partition_index_local = 0;
-            } else {
-                partition_index_local = StreamBits(10);
-                base_cem = StreamBits(6);
-            }
-            partition_index = partition_index_local; // Store to shared
-            const uint base_mode = base_cem & 3;
-            const uint max_weight = DecodeMaxWeight(mode);
-            const uint weight_bits = GetPackedBitSize(size_params, dual_plane, max_weight);
-            uint remaining_bits = 128 - weight_bits - total_bitsread;
-            uint extra_cem_bits = 0;
-            if (base_mode > 0) {
-                switch (num_partitions) {
-                case 2: extra_cem_bits += 2; break;
-                case 3: extra_cem_bits += 5; break;
-                case 4: extra_cem_bits += 8; break;
-                }
-            }
-            remaining_bits -= extra_cem_bits;
-            const uint plane_selector_bits = dual_plane ? 2 : 0;
-            remaining_bits -= plane_selector_bits;
-            if (remaining_bits > 128) {
-                size_params = uvec2(0); // Error
-            } else {
-                const uint color_data_bits = remaining_bits;
-                while (remaining_bits > 0) {
-                    const int nb = int(min(remaining_bits, 32U));
-                    const uint b = StreamBits(nb);
-                    color_endpoint_data[ced_pointer] = uint(bitfieldExtract(b, 0, nb));
-                    ++ced_pointer;
-                    remaining_bits -= nb;
-                }
-                plane_index = uint(StreamBits(plane_selector_bits));
-                if (base_mode > 0) {
-                    const uint extra_cem = StreamBits(extra_cem_bits);
-                    uint cem = (extra_cem << 6) | base_cem;
-                    cem >>= 2;
-                    uvec4 C = uvec4(0);
-                    for (uint i = 0; i < num_partitions; i++) {
-                        C[i] = (cem & 1); cem >>= 1;
-                    }
-                    uvec4 M = uvec4(0);
-                    for (uint i = 0; i < num_partitions; i++) {
-                        M[i] = cem & 3; cem >>= 2;
-                    }
-                    for (uint i = 0; i < num_partitions; i++) {
-                        color_endpoint_mode[i] = base_mode;
-                        if (C[i] == 0) --color_endpoint_mode[i];
-                        color_endpoint_mode[i] <<= 2;
-                        color_endpoint_mode[i] |= M[i];
-                    }
-                } else if (num_partitions > 1) {
-                    const uint cem = base_cem >> 2;
-                    for (uint i = 0; i < num_partitions; i++) {
-                        color_endpoint_mode[i] = cem;
-                    }
-                }
-
-                result_limit_reached = false;
-                uint colvals_index = 0;
-                DecodeColorValues(color_endpoint_mode, num_partitions, color_data_bits);
-                for (uint i = 0; i < num_partitions; i++) {
-                    ComputeEndpoints(endpoints0[i], endpoints1[i], color_endpoint_mode[i], color_values_direct, colvals_index);
-                }
-
-                color_endpoint_data = local_buff;
-                color_endpoint_data = bitfieldReverse(color_endpoint_data).wzyx;
-                const uint clear_byte_start = (weight_bits >> 3) + 1;
-                const uint byte_insert = ExtractBits(color_endpoint_data, int(clear_byte_start - 1) * 8, 8) & uint(((1 << (weight_bits % 8)) - 1));
-                const uint vec_index = (clear_byte_start - 1) >> 2;
-                color_endpoint_data[vec_index] = bitfieldInsert(color_endpoint_data[vec_index], byte_insert, int((clear_byte_start - 1) % 4) * 8, 8);
-                for (uint i = clear_byte_start; i < 16; ++i) {
-                    const uint idx = i >> 2;
-                    color_endpoint_data[idx] = bitfieldInsert(color_endpoint_data[idx], 0, int(i % 4) * 8, 8);
-                }
-
-                result_index = 0;
-                color_bitsread = 0;
-                result_limit_reached = false;
-                result_vector_max_index = size_params.x * size_params.y;
-                if (dual_plane) result_vector_max_index *= 2;
-                DecodeIntegerSequence(max_weight, GetNumWeightValues(size_params, dual_plane));
-                UnquantizeTexelWeights(size_params, dual_plane);
-
-                if (num_partitions > 1) {
-                    pt = GetPartitionTable(partition_index, num_partitions);
-                }
-            }
-        }
-    }
-    barrier();
-
-    if (size_params.x == 0) {
+    uint mode = StreamBits(11);
+    if (IsError(mode)) {
        FillError(coord);
        return;
    }
-    if (size_params.x == 0xFFFFFFFF) {
+    if ((mode & 0x1ff) == 0x1fc) {
+        // params.void_extent_ldr = true;
        FillVoidExtentLDR(coord);
        return;
    }
-
-    const uint total_texels = block_dims.x * block_dims.y;
-    for (uint tid = gl_LocalInvocationIndex; tid < total_texels; tid += gl_WorkGroupSize.x * gl_WorkGroupSize.y) {
-        uint x = tid % block_dims.x;
-        uint y = tid / block_dims.x;
-
-        uint local_partition = 0;
-        if (num_partitions > 1) {
-            local_partition = SelectPartition(pt, x, y, num_partitions);
+    const uvec2 size_params = DecodeBlockSize(mode);
+    if ((size_params.x > block_dims.x) || (size_params.y > block_dims.y)) {
+        FillError(coord);
+        return;
+    }
+    const uint num_partitions = StreamBits(2) + 1;
+    const uint mode_layout = FindLayout(mode);
+    const bool dual_plane = (mode_layout != 9) && ((mode & 0x400) != 0);
+    if (num_partitions > 4 || (num_partitions == 4 && dual_plane)) {
+        FillError(coord);
+        return;
+    }
+    uint partition_index = 1;
+    uvec4 color_endpoint_mode = uvec4(0);
+    uint ced_pointer = 0;
+    uint base_cem = 0;
+    if (num_partitions == 1) {
+        color_endpoint_mode.x = StreamBits(4);
+        partition_index = 0;
+    } else {
+        partition_index = StreamBits(10);
+        base_cem = StreamBits(6);
+    }
+    const uint base_mode = base_cem & 3;
+    const uint max_weight = DecodeMaxWeight(mode);
+    const uint weight_bits = GetPackedBitSize(size_params, dual_plane, max_weight);
+    uint remaining_bits = 128 - weight_bits - total_bitsread;
+    uint extra_cem_bits = 0;
+    if (base_mode > 0) {
+        switch (num_partitions) {
+        case 2:
+            extra_cem_bits += 2;
+            break;
+        case 3:
+            extra_cem_bits += 5;
+            break;
+        case 4:
+            extra_cem_bits += 8;
+            break;
+        default:
+            return;
+        }
+    }
+    remaining_bits -= extra_cem_bits;
+    const uint plane_selector_bits = dual_plane ? 2 : 0;
+    remaining_bits -= plane_selector_bits;
+    if (remaining_bits > 128) {
+        // Bad data, more remaining bits than 4 bytes
+        // return early
+        return;
+    }
+    // Read color data...
+    const uint color_data_bits = remaining_bits;
+    while (remaining_bits > 0) {
+        const int nb = int(min(remaining_bits, 32U));
+        const uint b = StreamBits(nb);
+        color_endpoint_data[ced_pointer] = uint(bitfieldExtract(b, 0, nb));
+        ++ced_pointer;
+        remaining_bits -= nb;
+    }
+    const uint plane_index = uint(StreamBits(plane_selector_bits));
+    if (base_mode > 0) {
+        const uint extra_cem = StreamBits(extra_cem_bits);
+        uint cem = (extra_cem << 6) | base_cem;
+        cem >>= 2;
+        uvec4 C = uvec4(0);
+        for (uint i = 0; i < num_partitions; i++) {
+            C[i] = (cem & 1);
+            cem >>= 1;
+        }
+        uvec4 M = uvec4(0);
+        for (uint i = 0; i < num_partitions; i++) {
+            M[i] = cem & 3;
+            cem >>= 2;
+        }
+        for (uint i = 0; i < num_partitions; i++) {
+            color_endpoint_mode[i] = base_mode;
+            if (C[i] == 0) {
+                --color_endpoint_mode[i];
+            }
+            color_endpoint_mode[i] <<= 2;
+            color_endpoint_mode[i] |= M[i];
+        }
+    } else if (num_partitions > 1) {
+        const uint cem = base_cem >> 2;
+        for (uint i = 0; i < num_partitions; i++) {
+            color_endpoint_mode[i] = cem;
+        }
+    }
+
+    uvec4 endpoints0[4];
+    uvec4 endpoints1[4];
+    {
+        // This decode phase should at most push 32 elements into the vector
+        result_vector_max_index = 32;
+        uint color_values[32];
+        uint colvals_index = 0;
+        DecodeColorValues(color_endpoint_mode, num_partitions, color_data_bits, color_values);
+        for (uint i = 0; i < num_partitions; i++) {
+            ComputeEndpoints(endpoints0[i], endpoints1[i], color_endpoint_mode[i], color_values,
+                             colvals_index);
+        }
+    }
+    color_endpoint_data = local_buff;
+    color_endpoint_data = bitfieldReverse(color_endpoint_data).wzyx;
+    const uint clear_byte_start = (weight_bits >> 3) + 1;
+
+    const uint byte_insert = ExtractBits(color_endpoint_data, int(clear_byte_start - 1) * 8, 8) &
+                             uint(((1 << (weight_bits % 8)) - 1));
+    const uint vec_index = (clear_byte_start - 1) >> 2;
+    color_endpoint_data[vec_index] = bitfieldInsert(color_endpoint_data[vec_index], byte_insert,
+                                                    int((clear_byte_start - 1) % 4) * 8, 8);
+    for (uint i = clear_byte_start; i < 16; ++i) {
+        const uint idx = i >> 2;
+        color_endpoint_data[idx] = bitfieldInsert(color_endpoint_data[idx], 0, int(i % 4) * 8, 8);
+    }
+
+    // Re-init vector variables for next decode phase
+    result_index = 0;
+    color_bitsread = 0;
+    result_limit_reached = false;
+
+    // The limit for the Unquantize phase, avoids decoding more data than needed.
+    result_vector_max_index = size_params.x * size_params.y;
+    if (dual_plane) {
+        result_vector_max_index *= 2;
+    }
+    DecodeIntegerSequence(max_weight, GetNumWeightValues(size_params, dual_plane));
+
+    UnquantizeTexelWeights(size_params, dual_plane);
+    for (uint j = 0; j < block_dims.y; j++) {
+        for (uint i = 0; i < block_dims.x; i++) {
+            uint local_partition = 0;
+            if (num_partitions > 1) {
+                local_partition = Select2DPartition(partition_index, i, j, num_partitions);
+            }
+            const uvec4 C0 = ReplicateByteTo16(endpoints0[local_partition]);
+            const uvec4 C1 = ReplicateByteTo16(endpoints1[local_partition]);
+            const uvec4 weight_vec = GetUnquantizedWeightVector(j, i, size_params, plane_index, dual_plane);
+            const vec4 Cf =
+                vec4((C0 * (uvec4(64) - weight_vec) + C1 * weight_vec + uvec4(32)) / 64);
+            const vec4 p = (Cf / 65535.0f);
+            imageStore(dest_image, coord + ivec3(i, j, 0), p.gbar);
        }
-        const uvec4 C0 = ReplicateByteTo16(endpoints0[local_partition]);
-        const uvec4 C1 = ReplicateByteTo16(endpoints1[local_partition]);
-        const uvec4 weight_vec = GetUnquantizedWeightVector(y, x, size_params, plane_index, dual_plane);
-        const vec4 Cf = vec4((C0 * (uvec4(64) - weight_vec) + C1 * weight_vec + uvec4(32)) / 64);
-        const vec4 p = (Cf / 65535.0f);
-        imageStore(dest_image, coord + ivec3(x, y, 0), p.gbar);
    }
 }

@@ -1176,8 +1132,7 @@ uint SwizzleOffset(uvec2 pos) {
 }

 void main() {
-    uvec3 block_id = gl_WorkGroupID;
-    uvec3 pos = block_id;
+    uvec3 pos = gl_GlobalInvocationID;
    pos.x <<= BYTES_PER_BLOCK_LOG2;
    const uint swizzle = SwizzleOffset(pos.xy);
    const uint block_y = pos.y >> GOB_SIZE_Y_SHIFT;
@@ -1189,21 +1144,10 @@ void main() {
    offset += (pos.x >> GOB_SIZE_X_SHIFT) << x_shift;
    offset += swizzle;

-    if (gl_LocalInvocationIndex == 0) {
-       total_bitsread = 0;
-       result_index = 0;
-       color_bitsread = 0;
-       write_color_values = false;
-       result_limit_reached = false;
-       color_out_index = 0;
-       color_num_values = 0;
-       local_buff = astc_data[offset / 16];
-    }
-    barrier();
-
-    ivec3 coord = ivec3(block_id * uvec3(block_dims, 1));
+    const ivec3 coord = ivec3(gl_GlobalInvocationID * uvec3(block_dims, 1));
    if (any(greaterThanEqual(coord, imageSize(dest_image)))) {
        return;
    }
+    local_buff = astc_data[offset / 16];
    DecompressBlock(coord);
 }
@@ -86,7 +86,7 @@ void WriteResults(uvec2 results[LOCAL_RESULTS]) {
    const uvec2 accum = accumulated_data;
    for (uint i = 0; i < LOCAL_RESULTS; i++) {
        uvec2 base_data = current_id * LOCAL_RESULTS + i < min_accumulation_base ? accum : uvec2(0, 0);
-        AddUint64(results[i], base_data);
+        results[i] = AddUint64(results[i], base_data);
    }
    for (uint i = 0; i < LOCAL_RESULTS; i++) {
        output_data[buffer_offset + current_id * LOCAL_RESULTS + i] = results[i];
@@ -586,8 +586,8 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
    });
    for (const VideoCommon::SwizzleParameters& swizzle : swizzles) {
        const size_t input_offset = swizzle.buffer_offset + map.offset;
-        const u32 num_dispatches_x = swizzle.num_tiles.width;
-        const u32 num_dispatches_y = swizzle.num_tiles.height;
+        const u32 num_dispatches_x = Common::DivCeil(swizzle.num_tiles.width, 8U);
+        const u32 num_dispatches_y = Common::DivCeil(swizzle.num_tiles.height, 8U);
        const u32 num_dispatches_z = image.info.resources.layers;

        compute_pass_descriptor_queue.Acquire();
@@ -640,6 +640,7 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
                               VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, 0, image_barrier);
    });
+    scheduler.Finish();
 }

 constexpr u32 BL3D_BINDING_SWIZZLE_TABLE = 0;
@@ -1230,6 +1230,10 @@ void Device::RemoveUnsuitableExtensions() {
    }
    RemoveExtensionFeatureIfUnsuitable(extensions.custom_border_color, features.custom_border_color,
                                       VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME);
+    // VK_KHR_unified_image_layouts
+    extensions.unified_image_layouts = features.unified_image_layouts.unifiedImageLayouts;
+    RemoveExtensionFeatureIfUnsuitable(extensions.unified_image_layouts, features.unified_image_layouts,
+                                       VK_KHR_UNIFIED_IMAGE_LAYOUTS_EXTENSION_NAME);

    // VK_EXT_depth_bias_control
    extensions.depth_bias_control =
@@ -70,7 +70,9 @@ VK_DEFINE_HANDLE(VmaAllocator)
    FEATURE(KHR, PipelineExecutableProperties, PIPELINE_EXECUTABLE_PROPERTIES,                     \
            pipeline_executable_properties)                                                        \
    FEATURE(KHR, WorkgroupMemoryExplicitLayout, WORKGROUP_MEMORY_EXPLICIT_LAYOUT,                  \
-            workgroup_memory_explicit_layout)
+            workgroup_memory_explicit_layout)                                                      \
+    FEATURE(KHR, UnifiedImageLayouts, UNIFIED_IMAGE_LAYOUTS, unified_image_layouts)
+

 // Define miscellaneous extensions which may be used by the implementation here.
 #define FOR_EACH_VK_EXTENSION(EXTENSION)                                                           \
@@ -3,6 +3,7 @@

 // Qt on macOS doesn't define VMA shit
 #include <boost/algorithm/string/split.hpp>
+#include "frontend_common/settings_generator.h"
 #include "qt_common/qt_string_lookup.h"
 #if defined(QT_STATICPLUGIN) && !defined(__APPLE__)
 #undef VMA_IMPLEMENTATION
@@ -432,6 +433,7 @@ MainWindow::MainWindow(bool has_broken_vulkan)
    Common::Log::Start();

    LoadTranslation();
+    FrontendCommon::GenerateSettings();

    setAcceptDrops(true);
    ui->setupUi(this);
@@ -23,7 +23,7 @@ MAXDEPTH=3
 # For your project you'll want to change this to define what dirs you have cpmfiles in
 # Remember to account for the MAXDEPTH variable!
 # Adding ./ before each will help to remove duplicates
-CPMFILES=$(find . src -maxdepth "$MAXDEPTH" -name cpmfile.json | sort | uniq)
+CPMFILES=$(find . -maxdepth "$MAXDEPTH" -name cpmfile.json | sort | uniq)

 # shellcheck disable=SC2016
 PACKAGES=$(echo "$CPMFILES" | xargs jq -s 'reduce .[] as $item ({}; . * $item)')
Author	SHA1	Message	Date
wildcard	213a9a7813	[host_shaders] Store the value of results instead of discarding it the function generated the value but we never stored it effectively discarding the result.	2026-02-04 22:13:11 +01:00
wildcard	48ba1f3f24	[vk, cmake] Bump minimum VulkanTools to 1.4.317 and add UnifiedImageLayouts (#3318 ) It seems too easy, the specification does not state anything more to be done for it towork. Requires performance testing on android. Co-authored-by: DraVee <dravee@eden-emu.dev> Co-authored-by: crueter <crueter@eden-emu.dev> Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3318 Reviewed-by: DraVee <dravee@eden-emu.dev> Reviewed-by: CamilleLaVey <camillelavey99@gmail.com> Co-authored-by: wildcard <wildcard@eden-emu.dev> Co-committed-by: wildcard <wildcard@eden-emu.dev>	2026-02-04 19:15:10 +01:00
crueter	d59fcf01bf	[frontend] Generate web token at runtime (#3462 ) Rather than having that hardcoded one like before. Also adds infrastructure which should make it easier to setup defaults at runtime (e.g. GPU stuff?) Signed-off-by: crueter <crueter@eden-emu.dev> Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3462 Reviewed-by: DraVee <dravee@eden-emu.dev> Reviewed-by: Lizzie <lizzie@eden-emu.dev>	2026-02-04 04:16:39 +01:00
PavelBARABANOV	b8456394f1	[spirv] Fix F16 conversion on Android (skip NaN check) (#3458 ) Fixes the yellow texture issue in Outer Wilds on Android. Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3458 Reviewed-by: Maufeat <sahyno1996@gmail.com> Reviewed-by: DraVee <dravee@eden-emu.dev> Co-authored-by: PavelBARABANOV <pavelbarabanov94@gmail.com> Co-committed-by: PavelBARABANOV <pavelbarabanov94@gmail.com>	2026-02-04 00:44:03 +01:00
wildcard	a467dd1ba6	[vk] fix flickering caused by recent swizzle changes (#3457 ) Co-authored-by: Maufeat <sahyno1996@gmail.com> Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3457 Reviewed-by: Lizzie <lizzie@eden-emu.dev> Reviewed-by: DraVee <dravee@eden-emu.dev> Reviewed-by: Maufeat <sahyno1996@gmail.com> Co-authored-by: wildcard <wildcard@eden-emu.dev> Co-committed-by: wildcard <wildcard@eden-emu.dev>	2026-02-04 00:37:58 +01:00
lizzie	c156f4760f	[meta] update codeowners (#3460 ) Signed-off-by: lizzie <lizzie@eden-emu.dev> Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3460 Reviewed-by: MaranBr <maranbr@eden-emu.dev> Reviewed-by: Maufeat <sahyno1996@gmail.com> Co-authored-by: lizzie <lizzie@eden-emu.dev> Co-committed-by: lizzie <lizzie@eden-emu.dev>	2026-02-03 23:42:53 +01:00
MaranBr	13f11ebf49	[nvnflinger] Reimplement GetBufferHistory (#3394 ) Reimplements GetBufferHistory, enabling tracking and retrieval of recent buffer states. This can improve rendering performance and stability in some games. Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3394 Reviewed-by: CamilleLaVey <camillelavey99@gmail.com> Co-authored-by: MaranBr <maranbr@outlook.com> Co-committed-by: MaranBr <maranbr@outlook.com>	2026-02-03 18:25:15 +01:00
xbzk	6065e9aa09	[cmake] Reduce spam on builds for MSVC (#3454 ) Fix for disable-warning-as-error flags set for MSVC not CLANG dripping into externals and causing override warnings on Ninja (cl) Signed-off-by: xbzk <xbzk@eden-emu.dev> Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3454 Reviewed-by: Lizzie <lizzie@eden-emu.dev> Reviewed-by: Maufeat <sahyno1996@gmail.com> Co-authored-by: xbzk <xbzk@eden-emu.dev> Co-committed-by: xbzk <xbzk@eden-emu.dev>	2026-02-03 13:53:07 +01:00