Stash old version of rebind logic to still be able to play with it.

Co-authored-by: Bill Napier <napier@google.com>
Co-authored-by: Powei Feng <powei@google.com>

.clang-format

@@ -0,0 +1,68 @@
+BasedOnStyle: LLVM
+AccessModifierOffset: -4
+AlignAfterOpenBracket: DontAlign
+AlignConsecutiveAssignments: None
+AlignOperands: Align
+AllowAllArgumentsOnNextLine: false
+AllowAllConstructorInitializersOnNextLine: false
+AllowAllParametersOfDeclarationOnNextLine: true
+AllowShortBlocksOnASingleLine: false
+AllowShortCaseLabelsOnASingleLine: false
+AllowShortFunctionsOnASingleLine: All
+AllowShortIfStatementsOnASingleLine: Always
+AllowShortLambdasOnASingleLine: All
+AllowShortLoopsOnASingleLine: true
+AlwaysBreakAfterReturnType: None
+AlwaysBreakTemplateDeclarations: Yes
+BreakBeforeBraces: Custom
+BracedInitializerIndentWidth: 4
+BraceWrapping:
+  AfterCaseLabel: false
+  AfterClass: false
+  AfterControlStatement: Never
+  AfterEnum: false
+  AfterFunction: false
+  AfterNamespace: false
+  AfterUnion: false
+  BeforeCatch: false
+  BeforeElse: false
+  IndentBraces: false
+  SplitEmptyFunction: false
+  SplitEmptyRecord: true
+BreakBeforeBinaryOperators: None
+BreakBeforeTernaryOperators: true
+BreakConstructorInitializers: BeforeColon
+BreakInheritanceList: BeforeColon
+ColumnLimit: 100
+CompactNamespaces: false
+ContinuationIndentWidth: 8
+Cpp11BracedListStyle: false
+IndentCaseLabels: true
+IndentPPDirectives: None
+IndentWidth: 4
+KeepEmptyLinesAtTheStartOfBlocks: true
+MaxEmptyLinesToKeep: 2
+NamespaceIndentation: None
+ObjCSpaceAfterProperty: false
+ObjCSpaceBeforeProtocolList: true
+PointerAlignment: Left
+ReflowComments: true
+SpaceAfterCStyleCast: true
+SpaceAfterLogicalNot: false
+SpaceAfterTemplateKeyword: false
+SpaceBeforeAssignmentOperators: true
+SpaceBeforeCpp11BracedList: false
+SpaceBeforeCtorInitializerColon: true
+SpaceBeforeInheritanceColon: true
+SpaceBeforeParens: ControlStatements
+SpaceBeforeRangeBasedForLoopColon: false
+SpaceInEmptyParentheses: false
+SpacesBeforeTrailingComments: 0
+SpacesInAngles: false
+SpacesInCStyleCastParentheses: false
+SpacesInContainerLiterals: false
+SpacesInParentheses: false
+SpacesInSquareBrackets: false
+TabWidth: 4
+UseTab: Never
+PackConstructorInitializers: Never

.github/workflows/android-continuous.yml

@@ -12,7 +12,7 @@ jobs:
     name: build-android
     # We intentially use a larger runner here to enable larger disk space
     # (standard linux runner will fail on disk space and faster build time).
-    runs-on: ubuntu-22.04-32core
+    runs-on: 'ubuntu-24.04-32core'
 
     steps:
       - uses: actions/checkout@v4.1.6

.github/workflows/linux-continuous.yml

@@ -10,7 +10,7 @@ on:
 jobs:
   build-linux:
     name: build-linux
-    runs-on: ubuntu-22.04-16core
+    runs-on: 'ubuntu-24.04-16core'
 
     steps:
       - uses: actions/checkout@v4.1.6

.github/workflows/presubmit.yml

@@ -40,7 +40,7 @@ jobs:
 
   build-android:
     name: build-android
-    runs-on: ubuntu-22.04-16core
+    runs-on: 'ubuntu-24.04-16core'
 
     steps:
       - uses: actions/checkout@v4.1.6
@@ -69,7 +69,7 @@ jobs:
 
   build-web:
     name: build-web
-    runs-on: ubuntu-22.04-16core
+    runs-on: 'ubuntu-24.04-16core'
 
     steps:
       - uses: actions/checkout@v4.1.6
@@ -79,7 +79,7 @@ jobs:
 
   validate-docs:
     name: validate-docs
-    runs-on: ubuntu-22.04-4core
+    runs-on: 'ubuntu-24.04-4core'
     steps:
       - uses: actions/checkout@v4.1.6
         with:
@@ -92,15 +92,33 @@ jobs:
 
   test-renderdiff:
     name: test-renderdiff
-    runs-on: ubuntu-22.04-16core
-
+    runs-on: macos-14-xlarge
     steps:
       - uses: actions/checkout@v4.1.6
-      - uses: ./.github/actions/ubuntu-apt-add-src
+      - name: Set up Homebrew
+        id: set-up-homebrew
+        uses: Homebrew/actions/setup-homebrew@master
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.x'
+      - name: Install python prereqs
+        run: pip install mako setuptools pyyaml
       - name: Run script
         run: |
-          source ./build/linux/ci-common.sh && bash test/renderdiff/test.sh
+          bash test/renderdiff/test.sh
       - uses: actions/upload-artifact@v4
         with:
           name: presubmit-renderdiff-result
           path: ./out/renderdiff_tests
+
+  validate-wgsl-webgpu:
+    name: validate-wgsl-webgpu
+    runs-on: 'ubuntu-24.04-8core'
+
+    steps:
+      - uses: actions/checkout@v4.1.6
+      - name: Run build script
+        run: source ./build/linux/ci-common.sh && ./build.sh -W debug test_filamat filament
+      - name: Run test
+        run: ./out/cmake-debug/libs/filamat/test_filamat --gtest_filter=MaterialCompiler.Wgsl*

.github/workflows/release.yml

@@ -9,6 +9,11 @@ name: Release
 env:
   RELEASE_TAG: ${{ github.event.inputs.release_tag }}
 
+# The default GITHUB_TOKEN does not have write permissions, which is needed to upload release
+# assets.
+permissions:
+  contents: write
+
 on:
   workflow_dispatch:
     inputs:
@@ -65,7 +70,7 @@ jobs:
 
   build-web:
     name: build-web
-    runs-on: ubuntu-22.04-16core
+    runs-on: 'ubuntu-24.04-16core'
     if: github.event_name == 'release' || github.event.inputs.platform == 'web'
 
     steps:
@@ -98,7 +103,7 @@ jobs:
 
   build-android:
     name: build-android
-    runs-on: ubuntu-22.04-16core
+    runs-on: 'ubuntu-24.04-16core'
     if: github.event_name == 'release' || github.event.inputs.platform == 'android'
 
     steps:

.github/workflows/web-continuous.yml

@@ -10,7 +10,7 @@ on:
 jobs:
   build-web:
     name: build-web
-    runs-on: ubuntu-22.04-16core
+    runs-on: 'ubuntu-24.04-16core'
 
     steps:
       - uses: actions/checkout@v4.1.6

CMakeLists.txt

@@ -139,14 +139,14 @@ else()
     set(LINUX FALSE)
 endif()
 
-if (LINUX)
-    if (NOT FILAMENT_OSMESA_PATH STREQUAL "")
-        if (NOT EXISTS ${FILAMENT_OSMESA_PATH}/)
-            message(FATAL_ERROR "Cannot find specified OSMesa build directory: ${FILAMENT_OSMESA_PATH}")
-        endif()
-        set(FILAMENT_SUPPORTS_OSMESA TRUE)
+if (NOT FILAMENT_OSMESA_PATH STREQUAL "")
+    if (NOT EXISTS ${FILAMENT_OSMESA_PATH}/)
+        message(FATAL_ERROR "Cannot find specified OSMesa build directory: ${FILAMENT_OSMESA_PATH}")
     endif()
+    set(FILAMENT_SUPPORTS_OSMESA TRUE)
+endif()
 
+if (LINUX)
     if (FILAMENT_SUPPORTS_WAYLAND)
         add_definitions(-DFILAMENT_SUPPORTS_WAYLAND)
         set(FILAMENT_SUPPORTS_X11 FALSE)
@@ -184,6 +184,12 @@ if (NOT ANDROID AND NOT WEBGL AND NOT IOS AND NOT FILAMENT_LINUX_IS_MOBILE)
     set(IS_HOST_PLATFORM TRUE)
 endif()
 
+if (APPLE)
+    if (FILAMENT_SUPPORTS_OSMESA)
+        add_definitions(-DFILAMENT_SUPPORTS_OSMESA)
+    endif()
+endif()
+
 if (WIN32)
     # Link statically against c/c++ lib to avoid missing redistriburable such as
     # "VCRUNTIME140.dll not found. Try reinstalling the app.", but give users
@@ -601,6 +607,14 @@ if (FILAMENT_SUPPORTS_METAL)
     set(MATC_API_FLAGS ${MATC_API_FLAGS} -a metal)
 endif()
 
+# With WebGPU, push constants are not supported. Skinning uses them.
+# WebGPU has a proposal to add push constants at https://github.com/gpuweb/gpuweb/blob/main/proposals/push-constants.md
+# With WebGPU, Tint does not support ClipDistance which is used in Stereo.  Mentioned in comment
+# https://github.com/google/dawn/blob/855d17b08abdf02f9142bf5a8f14d0ea088810a4/src/tint/lang/spirv/reader/ast_parser/function.cc#L4434
+if (FILAMENT_SUPPORTS_WEBGPU)
+    set(MATC_API_FLAGS ${MATC_API_FLAGS} -a webgpu --variant-filter=skinning,stereo)
+endif()
+
 # Disable ESSL 1.0 code generation.
 if (NOT FILAMENT_ENABLE_FEATURE_LEVEL_0)
     set(MATC_API_FLAGS ${MATC_API_FLAGS} -1)

NEW_RELEASE_NOTES.md

@@ -7,4 +7,3 @@ for next branch cut* header.
 appropriate header in [RELEASE_NOTES.md](./RELEASE_NOTES.md).
 
 ## Release notes for next branch cut
-- materials: add an optional `transformName` field to sampler material parameters. [⚠️ **New Material Version**]

README.md

@@ -31,7 +31,7 @@ repositories {
 }
 
 dependencies {
-    implementation 'com.google.android.filament:filament-android:1.58.0'
+    implementation 'com.google.android.filament:filament-android:1.58.2'
 }
 ```
 
@@ -51,7 +51,7 @@ Here are all the libraries available in the group `com.google.android.filament`:
 iOS projects can use CocoaPods to install the latest release:
 
 ```shell
-pod 'Filament', '~> 1.58.0'
+pod 'Filament', '~> 1.58.2'
 ```
 
 ## Documentation

RELEASE_NOTES.md

@@ -7,6 +7,21 @@ A new header is inserted each time a *tag* is created.
 Instead, if you are authoring a PR for the main branch, add your release note to
 [NEW_RELEASE_NOTES.md](./NEW_RELEASE_NOTES.md).
 
+## v1.59.0
+
+- materials: five custom variables (varyings) are now available on the condition that the `color` attribute is not requested (b/404930099). [⚠️ **New Material Version**]
+
+## v1.58.2
+
+- engine: Generate 1D instead of 3D LUTs for color grading whenever possible.
+
+## v1.58.1
+
+
+## v1.58.0
+
+- materials: add an optional `transformName` field to sampler material parameters. [⚠️ **New Material Version**]
+
 ## v1.57.3
 
 - android: breaking changes to API KTX1Loader::createIndirectLight and KTX1Loader::createSkybox

android/build.gradle

@@ -71,10 +71,9 @@ buildscript {
             .gradleProperty("com.google.android.filament.include-webgpu")
             .isPresent()
 
-    // TODO: Uncomment below when fgviewer is ready
-    // def fgviewer = providers
-    //         .gradleProperty("com.google.android.filament.fgviewer")
-    //         .isPresent()
+    def fgviewer = providers
+                 .gradleProperty("com.google.android.filament.fgviewer")
+                 .isPresent()
 
     def matdbg = providers
             .gradleProperty("com.google.android.filament.matdbg")
@@ -131,8 +130,7 @@ buildscript {
         "-DFILAMENT_DIST_DIR=${filamentPath}".toString(),
         "-DFILAMENT_SUPPORTS_VULKAN=${excludeVulkan ? 'OFF' : 'ON'}".toString(),
         "-DFILAMENT_SUPPORTS_WEBGPU=${includeWebGPU ? 'ON' : 'OFF'}".toString(),
-        // TODO: Uncomment below when fgviewer is ready
-        // "-DFILAMENT_ENABLE_FGVIEWER=${fgviewer ? 'ON' : 'OFF'}".toString(),
+        "-DFILAMENT_ENABLE_FGVIEWER=${fgviewer ? 'ON' : 'OFF'}".toString(),
         "-DFILAMENT_ENABLE_MATDBG=${matdbg ? 'ON' : 'OFF'}".toString(),
         "-DFILAMENT_DISABLE_MATOPT=${matnopt ? 'ON' : 'OFF'}".toString()
     ]

android/filamat-android/CMakeLists.txt

@@ -11,6 +11,13 @@ if (FILAMENT_SUPPORTS_VULKAN)
     message("Library filamat ignores Vulkan settings")
 endif()
 
+if (FILAMENT_SUPPORTS_WEBGPU)
+    # See third_party/dawn/tnt/CMakeLists.txt and libs/filamat/CMakeLists.txt
+    add_library(tint STATIC IMPORTED)
+    set_target_properties(tint PROPERTIES IMPORTED_LOCATION
+            ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libtint_combined.a)
+endif()
+
 add_library(${FILAMAT_FLAVOR} STATIC IMPORTED)
 set_target_properties(${FILAMAT_FLAVOR} PROPERTIES IMPORTED_LOCATION
         ${FILAMENT_DIR}/lib/${ANDROID_ABI}/lib${FILAMAT_FLAVOR}.a)
@@ -50,4 +57,5 @@ target_link_libraries(filamat-jni
         utils
         log
         smol-v
+        $<$<STREQUAL:${FILAMENT_SUPPORTS_WEBGPU},ON>:tint>
 )

android/filament-android/CMakeLists.txt

@@ -2,8 +2,7 @@ cmake_minimum_required(VERSION 3.19)
 project(filament-android)
 
 option(FILAMENT_SUPPORTS_VULKAN "Enables Vulkan on Android" OFF)
-# TODO: Uncomment below when fgviewer is ready
-# option(FILAMENT_ENABLE_FGVIEWER "Enables Frame Graph Viewer" OFF)
+option(FILAMENT_ENABLE_FGVIEWER "Enables Frame Graph Viewer" OFF)
 option(FILAMENT_ENABLE_MATDBG "Enables Material debugger" OFF)
 option(FILAMENT_DISABLE_MATOPT "Disables material optimizations" OFF)
 option(FILAMENT_SUPPORTS_WEBGPU "Enables WebGPU on Android" OFF)
@@ -30,10 +29,6 @@ add_library(filaflat STATIC IMPORTED)
 set_target_properties(filaflat PROPERTIES IMPORTED_LOCATION
         ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libfilaflat.a)
 
-add_library(filamat STATIC IMPORTED)
-set_target_properties(filamat PROPERTIES IMPORTED_LOCATION
-        ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libfilamat.a)
-
 add_library(geometry STATIC IMPORTED)
 set_target_properties(geometry PROPERTIES IMPORTED_LOCATION
         ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libgeometry.a)
@@ -60,12 +55,11 @@ add_library(smol-v STATIC IMPORTED)
 set_target_properties(smol-v PROPERTIES IMPORTED_LOCATION
         ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libsmol-v.a)
 
-# TODO: Uncomment below when fgviewer is ready
-# if (FILAMENT_ENABLE_FGVIEWER)
-#     add_library(fgviewer STATIC IMPORTED)
-#     set_target_properties(fgviewer PROPERTIES IMPORTED_LOCATION
-#             ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libfgviewer.a)
-# endif()
+if (FILAMENT_ENABLE_FGVIEWER)
+    add_library(fgviewer STATIC IMPORTED)
+    set_target_properties(fgviewer PROPERTIES IMPORTED_LOCATION
+            ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libfgviewer.a)
+endif()
 
 if (FILAMENT_ENABLE_MATDBG)
     add_library(matdbg STATIC IMPORTED)
@@ -133,8 +127,7 @@ target_link_libraries(filament-jni
     # libgeometry is PUBLIC because gltfio uses it.
     PUBLIC geometry
 
-    # TODO: Uncomment below when fgviewer is ready
-    # $<$<STREQUAL:${FILAMENT_ENABLE_FGVIEWER},ON>:fgviewer>
+    $<$<STREQUAL:${FILAMENT_ENABLE_FGVIEWER},ON>:fgviewer>
     $<$<STREQUAL:${FILAMENT_ENABLE_MATDBG},ON>:matdbg>
     $<$<STREQUAL:${FILAMENT_ENABLE_MATDBG},ON>:filamat>
     $<$<STREQUAL:${FILAMENT_SUPPORTS_VULKAN},ON>:bluevk>

android/filament-android/src/main/cpp/Texture.cpp

@@ -68,6 +68,21 @@ Java_com_google_android_filament_Texture_nIsTextureSwizzleSupported(JNIEnv*, jcl
     return (jboolean) Texture::isTextureSwizzleSupported(*engine);
 }
 
+
+extern "C" JNIEXPORT jint JNICALL
+Java_com_google_android_filament_Texture_nGetMaxTextureSize(JNIEnv *, jclass,
+        jlong nativeEngine, jint sampler) {
+    Engine *engine = (Engine *) nativeEngine;
+    return Texture::getMaxTextureSize(*engine, (Texture::Sampler)sampler);
+}
+
+extern "C" JNIEXPORT jint JNICALL
+Java_com_google_android_filament_Texture_nGetMaxArrayTextureLayers(JNIEnv *, jclass,
+        jlong nativeEngine) {
+    Engine *engine = (Engine *) nativeEngine;
+    return Texture::getMaxArrayTextureLayers(*engine);
+}
+
 extern "C" JNIEXPORT jboolean JNICALL
 Java_com_google_android_filament_Texture_nValidatePixelFormatAndType(JNIEnv*, jclass,
         jint internalFormat, jint pixelDataFormat, jint pixelDataType) {

android/filament-android/src/main/java/com/google/android/filament/ColorGrading.java

@@ -46,9 +46,8 @@ import static com.google.android.filament.Asserts.assertFloat4In;
  *
  * <h1>Performance</h1>
  *
- * Creating a new ColorGrading object may be more expensive than other Filament objects as a
- * 3D LUT may need to be generated. The generation of a 3D LUT, if necessary, may happen on
- * the CPU.
+ * Creating a new ColorGrading object may be more expensive than other Filament objects as a LUT may
+ * need to be generated. The generation of this LUT, if necessary, may happen on the CPU.
  *
  * <h1>Ordering</h1>
  *
@@ -160,6 +159,10 @@ public class ColorGrading {
          * quality level will use a 32x32x32 10 bit LUT, a high quality will use a 32x32x32 16 bit
          * LUT, and a ultra quality will use a 64x64x64 16 bit LUT.
          *
+         * This setting has no effect if generating a 1D LUT.
+         *
+         * This overrides the values set by format() and dimensions().
+         *
          * The default quality is {@link QualityLevel#MEDIUM}.
          *
          * @param qualityLevel The desired quality of the color grading process
@@ -175,6 +178,8 @@ public class ColorGrading {
          * When color grading is implemented using a 3D LUT, this sets the texture format of
          * of the LUT. This overrides the value set by quality().
          *
+         * This setting has no effect if generating a 1D LUT.
+         *
          * The default is INTEGER
          *
          * @param format The desired format of the 3D LUT.
@@ -190,6 +195,8 @@ public class ColorGrading {
          * When color grading is implemented using a 3D LUT, this sets the dimension of the LUT.
          * This overrides the value set by quality().
          *
+         * This setting has no effect if generating a 1D LUT.
+         *
          * The default is 32
          *
          * @param dim The desired dimension of the LUT. Between 16 and 64.
@@ -616,4 +623,3 @@ public class ColorGrading {
 
     private static native long nBuilderBuild(long nativeBuilder, long nativeEngine);
 }
-

android/filament-android/src/main/java/com/google/android/filament/Texture.java

@@ -691,6 +691,24 @@ public class Texture {
                 pixelDataType.ordinal());
     }
 
+    /**
+     * @param engine {@link Engine}
+     * @param type Texture sampler type
+     * @return The maximum size in texels of a texture of type \p type. At least 2048 for
+     *         2D textures, 256 for 3D textures
+     */
+    public static int getMaxTextureSize(@NonNull Engine engine, Sampler type) {
+        return nGetMaxTextureSize(engine.getNativeObject(), type.ordinal());
+    }
+
+    /**
+     * @param engine {@link Engine}
+     * @return The maximum number of layers supported by texture arrays. At least 256.
+     */
+    public static int getMaxArrayTextureLayers(@NonNull Engine engine) {
+        return nGetMaxArrayTextureLayers(engine.getNativeObject());
+    }
+
     /**
      * Use <code>Builder</code> to construct a <code>Texture</code> object instance.
      */
@@ -1289,6 +1307,8 @@ public class Texture {
     private static native boolean nIsTextureFormatSupported(long nativeEngine, int internalFormat);
     private static native boolean nIsTextureFormatMipmappable(long nativeEngine, int internalFormat);
     private static native boolean nIsTextureSwizzleSupported(long nativeEngine);
+    private static native int nGetMaxTextureSize(long nativeObject, int ordinal);
+    private static native int nGetMaxArrayTextureLayers(long nativeObject);
 
     private static native boolean nValidatePixelFormatAndType(int internalFormat, int pixelDataFormat,
             int pixelDataType);

android/filament-android/src/main/java/com/google/android/filament/ToneMapper.java

@@ -31,8 +31,6 @@ package com.google.android.filament;
  *   <li>DisplayRangeToneMapper</li>
  * </ul>
  * </ul>
- *
- * You can create custom tone mapping operators by subclassing ToneMapper.
  */
 public class ToneMapper {
     private final long mNativeObject;

android/gradle.properties

@@ -1,5 +1,5 @@
 GROUP=com.google.android.filament
-VERSION_NAME=1.58.0
+VERSION_NAME=1.58.2
 
 POM_DESCRIPTION=Real-time physically based rendering engine for Android.
 
@@ -30,4 +30,5 @@ com.google.android.filament.abis=all
 #com.google.android.filament.include-webgpu
 
 #com.google.android.filament.matdbg
+#com.google.android.filament.fgviewer
 #com.google.android.filament.matnopt

build.sh

@@ -200,8 +200,9 @@ WEBGPU_ANDROID_GRADLE_OPTION=""
 EGL_ON_LINUX_OPTION="-DFILAMENT_SUPPORTS_EGL_ON_LINUX=OFF"
 
 MATDBG_OPTION="-DFILAMENT_ENABLE_MATDBG=OFF"
-FGVIEWER_OPTION="-DFILAMENT_ENABLE_FGVIEWER=OFF"
 MATDBG_GRADLE_OPTION=""
+FGVIEWER_OPTION="-DFILAMENT_ENABLE_FGVIEWER=OFF"
+FGVIEWER_GRADLE_OPTION=""
 
 MATOPT_OPTION=""
 MATOPT_GRADLE_OPTION=""
@@ -533,6 +534,7 @@ function build_android {
             ${VULKAN_ANDROID_GRADLE_OPTION} \
             ${WEBGPU_ANDROID_GRADLE_OPTION} \
             ${MATDBG_GRADLE_OPTION} \
+            ${FGVIEWER_GRADLE_OPTION} \
             ${MATOPT_GRADLE_OPTION} \
             :filament-android:assembleDebug \
             :gltfio-android:assembleDebug \
@@ -541,6 +543,7 @@ function build_android {
         ./gradlew \
             -Pcom.google.android.filament.dist-dir=../out/android-debug/filament \
             -Pcom.google.android.filament.abis=${ABI_GRADLE_OPTION} \
+            ${WEBGPU_ANDROID_GRADLE_OPTION} \
             :filamat-android:assembleDebug
 
         if [[ "${BUILD_ANDROID_SAMPLES}" == "true" ]]; then
@@ -583,6 +586,7 @@ function build_android {
             ${VULKAN_ANDROID_GRADLE_OPTION} \
             ${WEBGPU_ANDROID_GRADLE_OPTION} \
             ${MATDBG_GRADLE_OPTION} \
+            ${FGVIEWER_GRADLE_OPTION} \
             ${MATOPT_GRADLE_OPTION} \
             :filament-android:assembleRelease \
             :gltfio-android:assembleRelease \
@@ -591,6 +595,7 @@ function build_android {
         ./gradlew \
             -Pcom.google.android.filament.dist-dir=../out/android-release/filament \
             -Pcom.google.android.filament.abis=${ABI_GRADLE_OPTION} \
+            ${WEBGPU_ANDROID_GRADLE_OPTION} \
             :filamat-android:assembleRelease
 
         if [[ "${BUILD_ANDROID_SAMPLES}" == "true" ]]; then

build/linux/ci-common.sh

@@ -30,6 +30,9 @@ if [[ "$GITHUB_WORKFLOW" ]]; then
     sudo apt-get install clang-$GITHUB_CLANG_VERSION libc++-$GITHUB_CLANG_VERSION-dev libc++abi-$GITHUB_CLANG_VERSION-dev
     sudo apt-get install mesa-common-dev libxi-dev libxxf86vm-dev
 
+    # For dawn
+    sudo apt-get install libxrandr-dev libxinerama-dev libxcursor-dev libxi-dev libx11-xcb-dev
+
     sudo update-alternatives --install /usr/bin/cc cc /usr/bin/clang-${GITHUB_CLANG_VERSION} 100
     sudo update-alternatives --install /usr/bin/c++ c++ /usr/bin/clang++-${GITHUB_CLANG_VERSION} 100
 fi

docs/Materials.md.html

@@ -13,7 +13,7 @@ This document is part of the [Filament project](https://github.com/google/filame
 ## Authors
 
 - [Romain Guy](https://github.com/romainguy), [@romainguy](https://twitter.com/romainguy)
-- [Mathias Agopian](https://github.com/pixelflinger), [@darthmoosious](https://twitter.com/darthmoosious)
+- [Mathias Agopian](https://github.com/pixelflinger), [@pixelflinger](https://bsky.app/profile/pixelflinger.bsky.social)
 
 # Overview
 
@@ -78,26 +78,27 @@ in table [standardProperties].
         Property        |      Definition
 -----------------------:|:---------------------
 **baseColor**           | Diffuse albedo for non-metallic surfaces, and specular color for metallic surfaces
-**metallic**            | Whether a surface appears to be dielectric (0.0) or conductor (1.0). Often used as a binary value (0 or 1)
 **roughness**           | Perceived smoothness (1.0) or roughness (0.0) of a surface. Smooth surfaces exhibit sharp reflections
+**metallic**            | Whether a surface appears to be dielectric (0.0) or conductor (1.0). Often used as a binary value (0 or 1)
 **reflectance**         | Fresnel reflectance at normal incidence for dielectric surfaces. This directly controls the strength of the reflections
-**sheenColor**          | Strength of the sheen layer
-**sheenRoughness**      | Perceived smoothness or roughness of the sheen layer
+**ambientOcclusion**    | Defines how much of the ambient light is accessible to a surface point. It is a per-pixel shadowing factor between 0.0 and 1.0
 **clearCoat**           | Strength of the clear coat layer
 **clearCoatRoughness**  | Perceived smoothness or roughness of the clear coat layer
+**clearCoatNormal**     | A detail normal used to perturb the clear coat layer using _bump mapping_ (_normal mapping_)
 **anisotropy**          | Amount of anisotropy in either the tangent or bitangent direction
 **anisotropyDirection** | Local surface direction in tangent space
-**ambientOcclusion**    | Defines how much of the ambient light is accessible to a surface point. It is a per-pixel shadowing factor between 0.0 and 1.0
-**normal**              | A detail normal used to perturb the surface using _bump mapping_ (_normal mapping_)
-**bentNormal**          | A normal pointing in the average unoccluded direction. Can be used to improve indirect lighting quality
-**clearCoatNormal**     | A detail normal used to perturb the clear coat layer using _bump mapping_ (_normal mapping_)
-**emissive**            | Additional diffuse albedo to simulate emissive surfaces (such as neons, etc.) This property is mostly useful in an HDR pipeline with a bloom pass
-**postLightingColor**   | Additional color that can be blended with the result of the lighting computations. See `postLightingBlending`
-**ior**                 | Index of refraction, either for refractive objects or as an alternative to reflectance
-**transmission**        | Defines how much of the diffuse light of a dielectric is transmitted through the object, in other words this defines how transparent an object is
-**absorption**          | Absorption factor for refractive objects
-**microThickness**      | Thickness of the thin layer of refractive objects
 **thickness**           | Thickness of the solid volume of refractive objects
+**sheenColor**          | Strength of the sheen layer
+**sheenRoughness**      | Perceived smoothness or roughness of the sheen layer
+**emissive**            | Additional diffuse albedo to simulate emissive surfaces (such as neons, etc.) This property is mostly useful in an HDR pipeline with a bloom pass
+**normal**              | A detail normal used to perturb the surface using _bump mapping_ (_normal mapping_)
+**postLightingColor**   | Additional color that can be blended with the result of the lighting computations. See `postLightingBlending`
+**absorption**          | Absorption factor for refractive objects
+**transmission**        | Defines how much of the diffuse light of a dielectric is transmitted through the object, in other words this defines how transparent an object is
+**ior**                 | Index of refraction, either for refractive objects or as an alternative to reflectance
+**microThickness**      | Thickness of the thin layer of refractive objects
+**bentNormal**          | A normal pointing in the average unoccluded direction. Can be used to improve indirect lighting quality
+**shadowStrength**      | Strength factor between 0 and 1 for all shadows received by this material
 [Table [standardProperties]: Properties of the standard model]
 
 The type and range of each property is described in table [standardPropertiesTypes].
@@ -1272,6 +1273,9 @@ Description
       when selecting any shading model that is not `unlit`. See the shader sections of this document
       for more information on how to access these attributes from the shaders.
 
+!!! Note: Interaction with custom variables
+    When the `color` attribute is specified, only four custom variables are available instead of five.
+
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ JSON
 material {
     parameters : [
@@ -1302,7 +1306,7 @@ Type
 :    array of `string`
 
 Value
-:     Up to 4 strings, each must be a valid GLSL identifier.
+:     Up to 5 strings, each must be a valid GLSL identifier.
 
 Description
 :     Defines custom interpolants (or variables) that are output by the material's vertex shader.
@@ -1318,6 +1322,10 @@ Description
       particular if `default` is specified the default precision is used is the fragment shader
       (`mediump`) and in the vertex shader (`highp`).
 
+!!! Warning: Interaction with required attributes
+    If the `color` attribute is specified in the `required` list, then only four variables can be used
+    instead of five.
+
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ JSON
 material {
     name : Skybox,

filament/backend/CMakeLists.txt

@@ -108,8 +108,12 @@ if (FILAMENT_SUPPORTS_OPENGL AND NOT FILAMENT_USE_EXTERNAL_GLES3)
         list(APPEND SRCS src/opengl/platforms/PlatformCocoaTouchGL.mm)
         list(APPEND SRCS src/opengl/platforms/CocoaTouchExternalImage.mm)
     elseif (APPLE)
-        list(APPEND SRCS src/opengl/platforms/PlatformCocoaGL.mm)
-        list(APPEND SRCS src/opengl/platforms/CocoaExternalImage.mm)
+        if (FILAMENT_SUPPORTS_OSMESA)
+            list(APPEND SRCS src/opengl/platforms/PlatformOSMesa.cpp)
+        else()
+            list(APPEND SRCS src/opengl/platforms/PlatformCocoaGL.mm)
+            list(APPEND SRCS src/opengl/platforms/CocoaExternalImage.mm)
+        endif()
     elseif (WEBGL)
         list(APPEND SRCS src/opengl/platforms/PlatformWebGL.cpp)
     elseif (LINUX)
@@ -173,10 +177,12 @@ endif()
 if (FILAMENT_SUPPORTS_VULKAN)
     list(APPEND SRCS
             include/backend/platforms/VulkanPlatform.h
-            src/vulkan/caching/VulkanDescriptorSetManager.cpp
-            src/vulkan/caching/VulkanDescriptorSetManager.h
-            src/vulkan/caching/VulkanPipelineLayoutCache.cpp
-            src/vulkan/caching/VulkanPipelineLayoutCache.h
+            src/vulkan/VulkanDescriptorSetCache.cpp
+            src/vulkan/VulkanDescriptorSetCache.h
+            src/vulkan/VulkanDescriptorSetLayoutCache.cpp
+            src/vulkan/VulkanDescriptorSetLayoutCache.h
+            src/vulkan/VulkanPipelineLayoutCache.cpp
+            src/vulkan/VulkanPipelineLayoutCache.h
             src/vulkan/memory/ResourceManager.cpp
             src/vulkan/memory/ResourceManager.h
             src/vulkan/memory/ResourcePointer.h
@@ -233,17 +239,44 @@ if (FILAMENT_SUPPORTS_VULKAN)
     elseif (APPLE OR IOS)
         list(APPEND SRCS src/vulkan/platform/VulkanPlatformApple.mm)
     elseif (ANDROID)
-        list(APPEND SRCS src/vulkan/platform/VulkanPlatformAndroid.cpp)
+        list(APPEND SRCS
+            include/backend/platforms/VulkanPlatformAndroid.h
+            src/vulkan/platform/VulkanPlatformAndroid.cpp
+        )
     endif()
 endif()
 
 if (FILAMENT_SUPPORTS_WEBGPU)
     list(APPEND SRCS
+            include/backend/platforms/WebGPUPlatform.h
+            src/webgpu/platform/WebGPUPlatform.cpp
+            src/webgpu/WebGPUConstants.h
             src/webgpu/WebGPUDriver.cpp
             src/webgpu/WebGPUDriver.h
-            src/webgpu/WebGPUPlatform.cpp
-            src/webgpu/WebGPUPlatform.h
     )
+    if (WIN32)
+        list(APPEND SRCS src/webgpu/platform/WebGPUPlatformWindows.cpp)
+    elseif (LINUX)
+        list(APPEND SRCS src/webgpu/platform/WebGPUPlatformLinux.cpp)
+    elseif (APPLE OR IOS)
+        list(APPEND SRCS src/webgpu/platform/WebGPUPlatformApple.mm)
+    elseif (ANDROID)
+        list(APPEND SRCS src/webgpu/platform/WebGPUPlatformAndroid.cpp)
+    endif()
+
+    if (TNT_DEV)
+        set(FILAMENT_WEBGPU_IMMEDIATE_ERROR_HANDLING_DEFAULT ON)
+    else()
+        set(FILAMENT_WEBGPU_IMMEDIATE_ERROR_HANDLING_DEFAULT OFF)
+    endif()
+
+    option(FILAMENT_WEBGPU_IMMEDIATE_ERROR_HANDLING
+            "Enable immediate_error_handling for the WebGPU backend Dawn implementation"
+            ${FILAMENT_WEBGPU_IMMEDIATE_ERROR_HANDLING_DEFAULT})
+endif()
+
+if (ANDROID)
+    list(APPEND SRCS src/BackendUtilsAndroid.cpp)
 endif()
 
 # ==================================================================================================
@@ -389,9 +422,12 @@ set(LINUX_LINKER_OPTIMIZATION_FLAGS
         -Wl,--exclude-libs,bluegl
 )
 
-if (LINUX AND FILAMENT_SUPPORTS_OSMESA)
-    set(OSMESA_COMPILE_FLAGS
-        -I${FILAMENT_OSMESA_PATH}/include/GL)
+if (FILAMENT_SUPPORTS_OSMESA)
+    if (LINUX)
+        set(OSMESA_COMPILE_FLAGS -I${FILAMENT_OSMESA_PATH}/include/GL)
+    elseif (APPLE)
+        set(OSMESA_COMPILE_FLAGS -I${FILAMENT_OSMESA_PATH}/include)
+    endif()
 endif()
 
 if (MSVC)
@@ -426,6 +462,10 @@ if (FILAMENT_SUPPORTS_METAL)
     target_compile_definitions(${TARGET} PRIVATE $<$<BOOL:${FILAMENT_METAL_PROFILING}>:FILAMENT_METAL_PROFILING>)
 endif()
 
+if (FILAMENT_SUPPORTS_WEBGPU)
+    target_compile_definitions(${TARGET} PRIVATE $<$<BOOL:${FILAMENT_WEBGPU_IMMEDIATE_ERROR_HANDLING}>:FILAMENT_WEBGPU_IMMEDIATE_ERROR_HANDLING>)
+endif()
+
 target_link_libraries(${TARGET} PRIVATE
         ${OSMESA_LINKER_FLAGS}
         $<$<AND:$<PLATFORM_ID:Linux>,$<CONFIG:Release>>:${LINUX_LINKER_OPTIMIZATION_FLAGS}>
@@ -454,6 +494,9 @@ if (APPLE OR LINUX)
         test/ShaderGenerator.cpp
         test/TrianglePrimitive.cpp
         test/Arguments.cpp
+        test/ImageExpectations.cpp
+        test/Lifetimes.cpp
+        test/Shader.cpp
         test/test_FeedbackLoops.cpp
         test/test_Blit.cpp
         test/test_MissingRequiredAttributes.cpp
@@ -469,6 +512,7 @@ if (APPLE OR LINUX)
         test/test_Handles.cpp
     )
     set(BACKEND_TEST_LIBS
+        absl::str_format
         backend
         getopt
         gtest
@@ -484,7 +528,7 @@ if (APPLE AND NOT IOS)
     list(APPEND BACKEND_TEST_SRC
          test/test_RenderExternalImage.cpp)
     add_library(backend_test STATIC ${BACKEND_TEST_SRC})
-    target_link_libraries(backend_test PRIVATE ${BACKEND_TEST_LIBS})
+    target_link_libraries(backend_test PUBLIC ${BACKEND_TEST_LIBS})
 
     set(BACKEND_TEST_DEPS
             OSDependent

filament/backend/include/backend/DriverEnums.h

@@ -126,6 +126,10 @@ static_assert(MAX_VERTEX_BUFFER_COUNT <= MAX_VERTEX_ATTRIBUTE_COUNT,
 static constexpr size_t CONFIG_UNIFORM_BINDING_COUNT = 9;   // This is guaranteed by OpenGL ES.
 static constexpr size_t CONFIG_SAMPLER_BINDING_COUNT = 4;   // This is guaranteed by OpenGL ES.
 
+static constexpr uint8_t EXTERNAL_SAMPLER_DATA_INDEX_UNUSED =
+        uint8_t(-1);// Case where the descriptor set binding isnt using any external sampler state
+                     // and therefore doesn't have a valid entry.
+
 /**
  * Defines the backend's feature levels.
  */
@@ -182,6 +186,7 @@ enum class ShaderLanguage {
     SPIRV = 2,
     MSL = 3,
     METAL_LIBRARY = 4,
+    WGSL = 5,
 };
 
 static constexpr const char* shaderLanguageToString(ShaderLanguage shaderLanguage) {
@@ -196,6 +201,8 @@ static constexpr const char* shaderLanguageToString(ShaderLanguage shaderLanguag
             return "MSL";
         case ShaderLanguage::METAL_LIBRARY:
             return "Metal precompiled library";
+        case ShaderLanguage::WGSL:
+            return "WGSL";
     }
 }
 
@@ -249,21 +256,21 @@ struct DescriptorSetLayoutBinding {
     DescriptorFlags flags = DescriptorFlags::NONE;
     uint16_t count = 0;
 
-    friend inline bool operator==(
-            DescriptorSetLayoutBinding const& lhs,
+    //  TODO: uncomment when needed.  Note that this class is used as hash key.  We need to ensure
+    //  no uninitialized padding bytes.
+    //    uint8_t externalSamplerDataIndex = EXTERNAL_SAMPLER_DATA_INDEX_UNUSED;
+
+    friend inline bool operator==(DescriptorSetLayoutBinding const& lhs,
             DescriptorSetLayoutBinding const& rhs) noexcept {
         return lhs.type == rhs.type &&
                lhs.flags == rhs.flags &&
                lhs.count == rhs.count &&
                lhs.stageFlags == rhs.stageFlags;
+//               lhs.stageFlags == rhs.stageFlags &&
+//               lhs.externalSamplerDataIndex == rhs.externalSamplerDataIndex;
     }
 };
 
-struct DescriptorSetLayout {
-    utils::FixedCapacityVector<DescriptorSetLayoutBinding> bindings;
-};
-
-
 /**
  * Bitmask for selecting render buffers
  */
@@ -807,6 +814,33 @@ static constexpr bool isStencilFormat(TextureFormat format) noexcept {
     }
 }
 
+inline constexpr bool isColorFormat(TextureFormat format) noexcept {
+    switch (format) {
+        // Standard color formats
+        case TextureFormat::R8:
+        case TextureFormat::RG8:
+        case TextureFormat::RGBA8:
+        case TextureFormat::R16F:
+        case TextureFormat::RG16F:
+        case TextureFormat::RGBA16F:
+        case TextureFormat::R32F:
+        case TextureFormat::RG32F:
+        case TextureFormat::RGBA32F:
+        case TextureFormat::RGB10_A2:
+        case TextureFormat::R11F_G11F_B10F:
+        case TextureFormat::SRGB8:
+        case TextureFormat::SRGB8_A8:
+        case TextureFormat::RGB8:
+        case TextureFormat::RGB565:
+        case TextureFormat::RGB5_A1:
+        case TextureFormat::RGBA4:
+            return true;
+        default:
+            break;
+    }
+    return false;
+}
+
 static constexpr bool isUnsignedIntFormat(TextureFormat format) {
     switch (format) {
         case TextureFormat::R8UI:
@@ -938,8 +972,28 @@ enum class SamplerCompareFunc : uint8_t {
     N           //!< Never. The depth / stencil test always fails.
 };
 
+//! this API is copied from (and only applies to) the Vulkan spec.
+//! These specify YUV to RGB conversions.
+enum class SamplerYcbcrModelConversion : uint8_t {
+    RGB_IDENTITY = 0,
+    YCBCR_IDENTITY = 1,
+    YCBCR_709 = 2,
+    YCBCR_601 = 3,
+    YCBCR_2020 = 4,
+};
+
+enum class SamplerYcbcrRange : uint8_t {
+    ITU_FULL = 0,
+    ITU_NARROW = 1,
+};
+
+enum class ChromaLocation : uint8_t {
+    COSITED_EVEN = 0,
+    MIDPOINT = 1,
+};
+
 //! Sampler parameters
-struct SamplerParams { // NOLINT
+struct SamplerParams {             // NOLINT
     SamplerMagFilter filterMag      : 1;    //!< magnification filter (NEAREST)
     SamplerMinFilter filterMin      : 3;    //!< minification filter  (NEAREST)
     SamplerWrapMode wrapS           : 2;    //!< s-coordinate wrap mode (CLAMP_TO_EDGE)
@@ -994,6 +1048,7 @@ private:
         return SamplerParams::LessThan{}(lhs, rhs);
     }
 };
+
 static_assert(sizeof(SamplerParams) == 4);
 
 // The limitation to 64-bits max comes from how we store a SamplerParams in our JNI code
@@ -1001,6 +1056,95 @@ static_assert(sizeof(SamplerParams) == 4);
 static_assert(sizeof(SamplerParams) <= sizeof(uint64_t),
         "SamplerParams must be no more than 64 bits");
 
+//! Sampler parameters
+struct SamplerYcbcrConversion {// NOLINT
+    SamplerYcbcrModelConversion ycbcrModel : 4;
+    TextureSwizzle r : 4;
+    TextureSwizzle g : 4;
+    TextureSwizzle b : 4;
+    TextureSwizzle a : 4;
+    SamplerYcbcrRange ycbcrRange : 1;
+    ChromaLocation xChromaOffset : 1;
+    ChromaLocation yChromaOffset : 1;
+    SamplerMagFilter chromaFilter : 1;
+    uint8_t padding;
+
+    struct Hasher {
+        size_t operator()(const SamplerYcbcrConversion p) const noexcept {
+            // we don't use std::hash<> here, so we don't have to include <functional>
+            return *reinterpret_cast<uint32_t const*>(reinterpret_cast<char const*>(&p));
+        }
+    };
+
+    struct EqualTo {
+        bool operator()(SamplerYcbcrConversion lhs, SamplerYcbcrConversion rhs) const noexcept {
+            assert_invariant(lhs.padding == 0);
+            auto* pLhs = reinterpret_cast<uint32_t const*>(reinterpret_cast<char const*>(&lhs));
+            auto* pRhs = reinterpret_cast<uint32_t const*>(reinterpret_cast<char const*>(&rhs));
+            return *pLhs == *pRhs;
+        }
+    };
+
+    struct LessThan {
+        bool operator()(SamplerYcbcrConversion lhs, SamplerYcbcrConversion rhs) const noexcept {
+            assert_invariant(lhs.padding == 0);
+            auto* pLhs = reinterpret_cast<uint32_t const*>(reinterpret_cast<char const*>(&lhs));
+            auto* pRhs = reinterpret_cast<uint32_t const*>(reinterpret_cast<char const*>(&rhs));
+            return *pLhs < *pRhs;
+        }
+    };
+
+private:
+    friend inline bool operator == (SamplerYcbcrConversion lhs, SamplerYcbcrConversion rhs)
+            noexcept {
+        return SamplerYcbcrConversion::EqualTo{}(lhs, rhs);
+    }
+    friend inline bool operator != (SamplerYcbcrConversion lhs, SamplerYcbcrConversion rhs)
+            noexcept {
+        return  !SamplerYcbcrConversion::EqualTo{}(lhs, rhs);
+    }
+    friend inline bool operator < (SamplerYcbcrConversion lhs, SamplerYcbcrConversion rhs)
+            noexcept {
+        return SamplerYcbcrConversion::LessThan{}(lhs, rhs);
+    }
+};
+
+static_assert(sizeof(SamplerYcbcrConversion) == 4);
+
+static_assert(sizeof(SamplerYcbcrConversion) <= sizeof(uint64_t),
+        "SamplerYcbcrConversion must be no more than 64 bits");
+
+struct ExternalSamplerDatum {
+    ExternalSamplerDatum(SamplerYcbcrConversion ycbcr, SamplerParams spm, uint32_t extFmt)
+        : YcbcrConversion(ycbcr),
+          samplerParams(spm),
+          externalFormat(extFmt) {}
+    bool operator==(ExternalSamplerDatum const& rhs) const {
+        return (YcbcrConversion == rhs.YcbcrConversion && samplerParams == rhs.samplerParams &&
+                externalFormat == rhs.externalFormat);
+    }
+    struct EqualTo {
+        bool operator()(const ExternalSamplerDatum& lhs,
+                const ExternalSamplerDatum& rhs) const noexcept {
+            return (lhs.YcbcrConversion == rhs.YcbcrConversion &&
+                lhs.samplerParams == rhs.samplerParams &&
+                lhs.externalFormat == rhs.externalFormat);
+        }
+    };
+    SamplerYcbcrConversion YcbcrConversion;
+    SamplerParams samplerParams;
+    uint32_t externalFormat;
+};
+// No implicit padding allowed due to it being a hash key.
+static_assert(sizeof(ExternalSamplerDatum) == 12);
+
+struct DescriptorSetLayout {
+    utils::FixedCapacityVector<DescriptorSetLayoutBinding> bindings;
+
+//  TODO: uncomment when needed
+//    utils::FixedCapacityVector<ExternalSamplerDatum> externalSamplerData;
+};
+
 //! blending equation function
 enum class BlendEquation : uint8_t {
     ADD,                    //!< the fragment is added to the color buffer

filament/backend/include/backend/platforms/PlatformCocoaGL.h

@@ -33,7 +33,7 @@ public:
     PlatformCocoaGL();
     ~PlatformCocoaGL() noexcept override;
 
-    ExternalImageHandle createExternalImage(void* cvPixelBuffer) noexcept;
+    ExternalImageHandle UTILS_PUBLIC createExternalImage(void* cvPixelBuffer) noexcept;
 
 protected:
     // --------------------------------------------------------------------------------------------

filament/backend/include/backend/platforms/PlatformCocoaTouchGL.h

@@ -32,7 +32,7 @@ public:
     PlatformCocoaTouchGL();
     ~PlatformCocoaTouchGL() noexcept override;
 
-    ExternalImageHandle createExternalImage(void* cvPixelBuffer) noexcept;
+    ExternalImageHandle UTILS_PUBLIC createExternalImage(void* cvPixelBuffer) noexcept;
 
     // --------------------------------------------------------------------------------------------
     // Platform Interface

filament/backend/include/backend/platforms/PlatformEGL.h

@@ -50,7 +50,7 @@ public:
     /**
      * Creates an ExternalImage from a EGLImageKHR
      */
-    ExternalImageHandle createExternalImage(EGLImageKHR eglImage) noexcept;
+    ExternalImageHandle UTILS_PUBLIC createExternalImage(EGLImageKHR eglImage) noexcept;
 
 protected:
     // --------------------------------------------------------------------------------------------

filament/backend/include/backend/platforms/PlatformEGLAndroid.h

@@ -23,6 +23,7 @@
 #include <backend/platforms/PlatformEGL.h>
 
 #include <utils/android/PerformanceHintManager.h>
+#include <utils/compiler.h>
 
 #include <chrono>
 
@@ -43,8 +44,21 @@ public:
     PlatformEGLAndroid() noexcept;
     ~PlatformEGLAndroid() noexcept override;
 
-protected:
+    /**
+     * Creates an ExternalImage from a EGLImageKHR
+     */
+    ExternalImageHandle UTILS_PUBLIC createExternalImage(AHardwareBuffer const* buffer, bool sRGB) noexcept;
 
+    struct UTILS_PUBLIC ExternalImageDescAndroid {
+        uint32_t width;      // Texture width
+        uint32_t height;     // Texture height
+        TextureFormat format;// Texture format
+        TextureUsage usage;  // Texture usage flags
+    };
+
+    ExternalImageDescAndroid UTILS_PUBLIC getExternalImageDesc(ExternalImageHandle externalImage) noexcept;
+
+protected:
     // --------------------------------------------------------------------------------------------
     // Platform Interface
 
@@ -57,11 +71,6 @@ protected:
     Driver* createDriver(void* sharedContext,
             const Platform::DriverConfig& driverConfig) noexcept override;
 
-    /**
-     * Creates an ExternalImage from a EGLImageKHR
-     */
-    ExternalImageHandle createExternalImage(AHardwareBuffer const *buffer, bool sRGB) noexcept;
-
     // --------------------------------------------------------------------------------------------
     // OpenGLPlatform Interface
 
@@ -95,15 +104,26 @@ protected:
      */
     AcquiredImage transformAcquiredImage(AcquiredImage source) noexcept override;
 
-    bool setExternalImage(ExternalImageHandleRef externalImage, ExternalTexture* texture) noexcept override;
+    OpenGLPlatform::ExternalTexture* createExternalImageTexture() noexcept override;
+    void destroyExternalImageTexture(ExternalTexture* texture) noexcept override;
 
     struct ExternalImageEGLAndroid : public ExternalImageEGL {
         AHardwareBuffer* aHardwareBuffer = nullptr;
+        uint32_t width;      // Texture width
+        uint32_t height;     // Texture height
+        TextureFormat format;// Texture format
+        TextureUsage usage;  // Texture usage flags
         bool sRGB = false;
+
     protected:
         ~ExternalImageEGLAndroid() override;
     };
 
+    bool setExternalImage(ExternalImageHandleRef externalImage,
+            ExternalTexture* texture) noexcept override;
+    bool setImage(ExternalImageEGLAndroid const* eglExternalImage,
+            ExternalTexture* texture) noexcept;
+
 protected:
     bool makeCurrent(ContextType type,
             SwapChain* drawSwapChain,

filament/backend/include/backend/platforms/PlatformOSMesa.h

@@ -21,7 +21,12 @@
 
 #include "bluegl/BlueGL.h"
 
+#if defined(__linux__)
 #include <osmesa.h>
+#elif defined(__APPLE__)
+#undef GLAPI
+#include <GL/osmesa.h>
+#endif
 
 #include <backend/platforms/OpenGLPlatform.h>
 #include <backend/DriverEnums.h>

filament/backend/include/backend/platforms/VulkanPlatform.h

@@ -18,6 +18,7 @@
 #define TNT_FILAMENT_BACKEND_PLATFORMS_VULKANPLATFORM_H
 
 #include <backend/Platform.h>
+#include <backend/DriverEnums.h>
 
 #include <bluevk/BlueVK.h>
 
@@ -317,6 +318,11 @@ public:
          */
         uint32_t layers;
 
+        /**
+         * The numbers of samples per texel
+         */
+        VkSampleCountFlagBits samples;
+
         /**
          * The format of the external image
          */
@@ -347,22 +353,48 @@ public:
          */
         uint32_t memoryTypeBits;
     };
-    virtual ExternalImageMetadata getExternalImageMetadata(void* externalImage);
+    virtual ExternalImageMetadata getExternalImageMetadata(ExternalImageHandleRef externalImage);
 
     using ImageData = std::pair<VkImage, VkDeviceMemory>;
-    virtual ImageData createExternalImage(void* externalImage,
-            const ExternalImageMetadata& metadata);
+    virtual ImageData createExternalImageData(ExternalImageHandleRef externalImage,
+            const ExternalImageMetadata& metadata, uint32_t memoryTypeIndex,
+            VkImageUsageFlags usage);
+
+    virtual VkSampler createExternalSampler(SamplerYcbcrConversion chroma,
+            SamplerParams sampler,
+            uint32_t internalFormat);
+
+    virtual VkImageView createExternalImageView(SamplerYcbcrConversion chroma,
+            uint32_t internalFormat, VkImage image, VkImageSubresourceRange range,
+            VkImageViewType viewType, VkComponentMapping swizzle);
+
+protected:
+    virtual ExtensionSet getSwapchainInstanceExtensions() const;
+
+    using SurfaceBundle = std::tuple<VkSurfaceKHR, VkExtent2D>;
+    virtual SurfaceBundle createVkSurfaceKHR(void* nativeWindow, VkInstance instance,
+            uint64_t flags) const noexcept;
 
 private:
-    static ExtensionSet getSwapchainInstanceExtensions();
-    static ExternalImageMetadata getExternalImageMetadataImpl(void* externalImage,
+    // Platform dependent helper methods
+    static ExtensionSet getSwapchainInstanceExtensionsImpl();
+
+    static ExternalImageMetadata getExternalImageMetadataImpl(ExternalImageHandleRef externalImage,
             VkDevice device);
-    static ImageData createExternalImageImpl(void* externalImage, VkDevice device,
-            const VkAllocationCallbacks* allocator, const ExternalImageMetadata& metadata);
+
+    static ImageData createExternalImageDataImpl(ExternalImageHandleRef externalImage,
+            VkDevice device, const ExternalImageMetadata& metadata, uint32_t memoryTypeIndex,
+            VkImageUsageFlags usage);
+    static VkSampler createExternalSamplerImpl(VkDevice device,
+            SamplerYcbcrConversion chroma, SamplerParams sampler,
+            uint32_t internalFormat);
+    static VkImageView createExternalImageViewImpl(VkDevice device,
+            SamplerYcbcrConversion chroma, uint32_t internalFormat, VkImage image,
+            VkImageSubresourceRange range, VkImageViewType viewType,
+            VkComponentMapping swizzle);
 
     // Platform dependent helper methods
-    using SurfaceBundle = std::tuple<VkSurfaceKHR, VkExtent2D>;
-    static SurfaceBundle createVkSurfaceKHR(void* nativeWindow, VkInstance instance,
+    static SurfaceBundle createVkSurfaceKHRImpl(void* nativeWindow, VkInstance instance,
             uint64_t flags) noexcept;
 
     friend struct VulkanPlatformPrivate;

filament/backend/include/backend/platforms/VulkanPlatformAndroid.h

@@ -0,0 +1,71 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_BACKEND_PLATFORMS_VULKAN_PLATFORM_ANDROID_H
+#define TNT_FILAMENT_BACKEND_PLATFORMS_VULKAN_PLATFORM_ANDROID_H
+
+#include <backend/DriverEnums.h>
+#include <backend/platforms/VulkanPlatform.h>
+
+#include <android/hardware_buffer.h>
+
+namespace filament::backend {
+
+class VulkanPlatformAndroid : public VulkanPlatform {
+public:
+    Platform::ExternalImageHandle UTILS_PUBLIC createExternalImage(AHardwareBuffer const* buffer,
+            bool sRGB) noexcept;
+
+    struct UTILS_PUBLIC ExternalImageDescAndroid {
+        uint32_t width;      // Texture width
+        uint32_t height;     // Texture height
+        TextureFormat format;// Texture format
+        TextureUsage usage;  // Texture usage flags
+    };
+
+    ExternalImageDescAndroid UTILS_PUBLIC getExternalImageDesc(
+            ExternalImageHandleRef externalImage) const noexcept;
+
+protected:
+    struct ExternalImageVulkanAndroid : public Platform::ExternalImage {
+        AHardwareBuffer* aHardwareBuffer = nullptr;
+        bool sRGB = false;
+        unsigned int width;  // Texture width
+        unsigned int height; // Texture height
+        TextureFormat format;// Texture format
+        TextureUsage usage;  // Texture usage flags
+
+    protected:
+        ~ExternalImageVulkanAndroid() override;
+    };
+
+    virtual ExternalImageMetadata getExternalImageMetadata(ExternalImageHandleRef externalImage);
+
+    using ImageData = VulkanPlatform::ImageData;
+    virtual ImageData createExternalImageData(ExternalImageHandleRef externalImage,
+            const ExternalImageMetadata& metadata, uint32_t memoryTypeIndex,
+            VkImageUsageFlags usage);
+
+    virtual ExtensionSet getSwapchainInstanceExtensions() const;
+
+    using SurfaceBundle = VulkanPlatform::SurfaceBundle;
+    virtual SurfaceBundle createVkSurfaceKHR(void* nativeWindow, VkInstance instance,
+            uint64_t flags) const noexcept;
+};
+
+}// namespace filament::backend
+
+#endif// TNT_FILAMENT_BACKEND_PLATFORMS_VULKAN_PLATFORM_ANDROID_H

filament/backend/include/backend/platforms/WebGPUPlatform.h

@@ -0,0 +1,59 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_BACKEND_PLATFORMS_WEBGPUPLATFORM_H
+#define TNT_FILAMENT_BACKEND_PLATFORMS_WEBGPUPLATFORM_H
+
+#include <backend/Platform.h>
+
+#include <webgpu/webgpu_cpp.h>
+
+#include <cstdint>
+
+namespace filament::backend {
+
+/**
+ * A Platform interface, handling the environment-specific concerns, e.g. OS, for creating a WebGPU
+ * driver (backend).
+ */
+class WebGPUPlatform final : public Platform {
+public:
+    WebGPUPlatform();
+    ~WebGPUPlatform() override = default;
+
+    [[nodiscard]] int getOSVersion() const noexcept final { return 0; }
+
+    [[nodiscard]] wgpu::Instance& getInstance() noexcept { return mInstance; }
+
+    // either returns a valid surface or panics
+    [[nodiscard]] wgpu::Surface createSurface(void* nativeWindow, uint64_t flags);
+    // either returns a valid adapter or panics
+    [[nodiscard]] wgpu::Adapter requestAdapter(wgpu::Surface const& surface);
+    // either returns a valid device or panics
+    [[nodiscard]] wgpu::Device requestDevice(wgpu::Adapter const& adapter);
+
+protected:
+    [[nodiscard]] Driver* createDriver(void* sharedContext,
+            const Platform::DriverConfig& driverConfig) noexcept override;
+
+private:
+    // we may consider having the driver own this in the future
+    wgpu::Instance mInstance;
+};
+
+}// namespace filament::backend
+
+#endif// TNT_FILAMENT_BACKEND_PLATFORMS_WEBGPUPLATFORM_H

filament/backend/include/private/backend/BackendUtilsAndroid.h

@@ -14,29 +14,23 @@
  * limitations under the License.
  */
 
-#ifndef TNT_FILAMENT_BACKEND_WEBGPUPLATFORM_H
-#define TNT_FILAMENT_BACKEND_WEBGPUPLATFORM_H
+#ifndef TNT_FILAMENT_BACKEND_PRIVATE_BACKENDUTILSANDROID_H
+#define TNT_FILAMENT_BACKEND_PRIVATE_BACKENDUTILSANDROID_H
 
 #include <backend/DriverEnums.h>
-#include <backend/Platform.h>
-
-#include "webgpu/webgpu_cpp.h"
 
 namespace filament::backend {
 
-class WebGPUPlatform
-    final : public Platform {
-public:
+/**
+ * Maps AHardwareBuffer format to TextureFormat.
+ */
+TextureFormat mapToFilamentFormat(unsigned int format, bool isSrgbTransfer) noexcept;
 
-    int getOSVersion() const noexcept final { return 0; }
-
-    ~WebGPUPlatform() noexcept override = default;
-
-protected:
-
-    Driver* createDriver(void* sharedContext, const Platform::DriverConfig& driverConfig) noexcept override;
-};
+/**
+ * Maps AHardwareBuffer usage to TextureUsage.
+ */
+TextureUsage mapToFilamentUsage(unsigned int usage, TextureFormat format) noexcept;
 
 } // namespace filament
 
-#endif // TNT_FILAMENT_BACKEND_WEBGPUPLATFORM_H
+#endif // TNT_FILAMENT_BACKEND_PRIVATE_BACKENDUTILSANDROID_H

filament/backend/include/private/backend/DriverAPI.inc

@@ -362,6 +362,8 @@ DECL_DRIVER_API_SYNCHRONOUS_0(bool, isProtectedTexturesSupported)
 DECL_DRIVER_API_SYNCHRONOUS_0(bool, isDepthClampSupported)
 DECL_DRIVER_API_SYNCHRONOUS_0(uint8_t, getMaxDrawBuffers)
 DECL_DRIVER_API_SYNCHRONOUS_0(size_t, getMaxUniformBufferSize)
+DECL_DRIVER_API_SYNCHRONOUS_N(size_t, getMaxTextureSize, backend::SamplerType, target)
+DECL_DRIVER_API_SYNCHRONOUS_0(size_t, getMaxArrayTextureLayers)
 DECL_DRIVER_API_SYNCHRONOUS_0(math::float2, getClipSpaceParams)
 DECL_DRIVER_API_SYNCHRONOUS_N(void, setupExternalImage2, backend::Platform::ExternalImageHandleRef, image)
 DECL_DRIVER_API_SYNCHRONOUS_N(void, setupExternalImage, void*, image)

filament/backend/src/BackendUtilsAndroid.cpp

@@ -0,0 +1,98 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "private/backend/BackendUtilsAndroid.h"
+
+#include <android/hardware_buffer.h>
+
+namespace filament::backend {
+
+TextureFormat mapToFilamentFormat(unsigned int format, bool isSrgbTransfer) noexcept {
+    if (isSrgbTransfer) {
+        switch (format) {
+            case AHARDWAREBUFFER_FORMAT_R8G8B8_UNORM:
+                return TextureFormat::SRGB8;
+            default:
+                return TextureFormat::SRGB8_A8;
+        }
+    }
+    switch (format) {
+        case AHARDWAREBUFFER_FORMAT_R8G8B8A8_UNORM:
+        case AHARDWAREBUFFER_FORMAT_R8G8B8X8_UNORM:
+            return TextureFormat::RGBA8;
+        case AHARDWAREBUFFER_FORMAT_R8G8B8_UNORM:
+        case AHARDWAREBUFFER_FORMAT_Y8Cb8Cr8_420:
+            return TextureFormat::RGB8;
+        case AHARDWAREBUFFER_FORMAT_R5G6B5_UNORM:
+            return TextureFormat::RGB565;
+        case AHARDWAREBUFFER_FORMAT_R16G16B16A16_FLOAT:
+            return TextureFormat::RGBA16F;
+        case AHARDWAREBUFFER_FORMAT_R10G10B10A2_UNORM:
+            return TextureFormat::RGB10_A2;
+        case AHARDWAREBUFFER_FORMAT_D16_UNORM:
+            return TextureFormat::DEPTH16;
+        case AHARDWAREBUFFER_FORMAT_D24_UNORM:
+            return TextureFormat::DEPTH24;
+        case AHARDWAREBUFFER_FORMAT_D24_UNORM_S8_UINT:
+            return TextureFormat::DEPTH24_STENCIL8;
+        case AHARDWAREBUFFER_FORMAT_D32_FLOAT:
+            return TextureFormat::DEPTH32F;
+        case AHARDWAREBUFFER_FORMAT_D32_FLOAT_S8_UINT:
+            return TextureFormat::DEPTH32F_STENCIL8;
+        case AHARDWAREBUFFER_FORMAT_S8_UINT:
+            return TextureFormat::STENCIL8;
+        case AHARDWAREBUFFER_FORMAT_R8_UNORM:
+            return TextureFormat::R8;
+        default:
+            return TextureFormat::UNUSED;
+    }
+}
+
+TextureUsage mapToFilamentUsage(unsigned int usage, TextureFormat format) noexcept {
+    TextureUsage usageFlags = TextureUsage::NONE;
+
+    if (usage & AHARDWAREBUFFER_USAGE_GPU_SAMPLED_IMAGE) {
+        usageFlags |= TextureUsage::SAMPLEABLE;
+    }
+
+    if (usage & AHARDWAREBUFFER_USAGE_GPU_FRAMEBUFFER) {
+        if (isDepthFormat(format)) {
+            usageFlags |= TextureUsage::DEPTH_ATTACHMENT;
+        }
+        if (isStencilFormat(format)) {
+            usageFlags |= TextureUsage::STENCIL_ATTACHMENT;
+        }
+        if (isColorFormat(format)) {
+            usageFlags |= TextureUsage::COLOR_ATTACHMENT;
+        }
+    }
+
+    if (usage & AHARDWAREBUFFER_USAGE_GPU_DATA_BUFFER) {
+        usageFlags |= TextureUsage::UPLOADABLE;
+    }
+
+    if (usage & AHARDWAREBUFFER_USAGE_PROTECTED_CONTENT) {
+        usageFlags |= TextureUsage::PROTECTED;
+    }
+
+    if (usageFlags == TextureUsage::NONE) {
+        usageFlags = TextureUsage::COLOR_ATTACHMENT | TextureUsage::SAMPLEABLE;
+    }
+
+    return usageFlags;
+}
+
+} // namespace backend::filament

filament/backend/src/PlatformFactory.cpp

@@ -19,6 +19,11 @@
 #include <utils/Systrace.h>
 #include <utils/debug.h>
 
+// We need to keep this up top for the linux (X11) name collisions.
+#if defined(FILAMENT_SUPPORTS_WEBGPU)
+    #include "backend/platforms/WebGPUPlatform.h"
+#endif
+
 #if defined(__ANDROID__)
     #include <sys/system_properties.h>
     #if defined(FILAMENT_SUPPORTS_OPENGL) && !defined(FILAMENT_USE_EXTERNAL_GLES3)
@@ -30,7 +35,11 @@
     #endif
 #elif defined(__APPLE__)
     #if defined(FILAMENT_SUPPORTS_OPENGL) && !defined(FILAMENT_USE_EXTERNAL_GLES3)
-        #include <backend/platforms/PlatformCocoaGL.h>
+        #if defined(FILAMENT_SUPPORTS_OSMESA)
+            #include <backend/platforms/PlatformOSMesa.h>
+        #else
+            #include <backend/platforms/PlatformCocoaGL.h>
+        #endif
     #endif
 #elif defined(__linux__)
     #if defined(FILAMENT_SUPPORTS_X11)
@@ -55,7 +64,11 @@
 #endif
 
 #if defined(FILAMENT_DRIVER_SUPPORTS_VULKAN)
-    #include "backend/platforms/VulkanPlatform.h"
+    #if defined(__ANDROID__)
+        #include "backend/platforms/VulkanPlatformAndroid.h"
+    #else
+        #include "backend/platforms/VulkanPlatform.h"
+    #endif
 #endif
 
 #if defined (FILAMENT_SUPPORTS_METAL)
@@ -65,9 +78,6 @@ filament::backend::Platform* createDefaultMetalPlatform();
 #endif
 
 #include "noop/PlatformNoop.h"
-#if defined(FILAMENT_SUPPORTS_WEBGPU)
-    #include "webgpu/WebGPUPlatform.h"
-#endif
 
 namespace filament::backend {
 
@@ -104,7 +114,11 @@ Platform* PlatformFactory::create(Backend* backend) noexcept {
     }
     if (*backend == Backend::VULKAN) {
         #if defined(FILAMENT_DRIVER_SUPPORTS_VULKAN)
-            return new VulkanPlatform();
+            #if defined(__ANDROID__)
+                return new VulkanPlatformAndroid();
+            #else
+                return new VulkanPlatform();
+            #endif
         #else
             return nullptr;
         #endif
@@ -132,7 +146,11 @@ Platform* PlatformFactory::create(Backend* backend) noexcept {
         #elif defined(FILAMENT_IOS)
             return new PlatformCocoaTouchGL();
         #elif defined(__APPLE__)
-            return new PlatformCocoaGL();
+            #if defined(FILAMENT_SUPPORTS_OSMESA)
+                return new PlatformOSMesa();
+            #else
+                return new PlatformCocoaGL();
+            #endif
         #elif defined(__linux__)
             #if defined(FILAMENT_SUPPORTS_X11)
                 return new PlatformGLX();

filament/backend/src/metal/MetalDriver.mm

@@ -1177,6 +1177,24 @@ size_t MetalDriver::getMaxUniformBufferSize() {
     return 256 * 1024 * 1024;   // TODO: return the actual size instead of hardcoding the minspec
 }
 
+size_t MetalDriver::getMaxTextureSize(SamplerType type) {
+    if (type == SamplerType::SAMPLER_3D) {
+        return 2048;
+    }
+    if (mContext->highestSupportedGpuFamily.apple >= 3 ||
+            mContext->highestSupportedGpuFamily.mac >= 2) {
+        return 16384;
+    } else if (mContext->highestSupportedGpuFamily.apple >= 2) {
+        return 8192;
+    }
+    return 4096;
+}
+
+size_t MetalDriver::getMaxArrayTextureLayers() {
+    // TODO: return the actual size instead of hardcoding the minspec
+    return 256;
+}
+
 void MetalDriver::updateIndexBuffer(Handle<HwIndexBuffer> ibh, BufferDescriptor&& data,
         uint32_t byteOffset) {
     FILAMENT_CHECK_PRECONDITION(data.buffer)

filament/backend/src/metal/MetalShaderCompiler.mm

@@ -138,6 +138,7 @@ bool MetalShaderCompiler::isParallelShaderCompileSupported() const noexcept {
             case ShaderLanguage::ESSL1:
             case ShaderLanguage::ESSL3:
             case ShaderLanguage::SPIRV:
+            case ShaderLanguage::WGSL:
                 break;
         }
 

filament/backend/src/noop/NoopDriver.cpp

@@ -240,6 +240,16 @@ size_t NoopDriver::getMaxUniformBufferSize() {
     return 16384u;
 }
 
+size_t NoopDriver::getMaxTextureSize(SamplerType target) {
+    // NoopDriver is being actively used for other purposes.  This needs to be resolved before we
+    // can change it to 2048. b/406832484
+    return 16384u;
+}
+
+size_t NoopDriver::getMaxArrayTextureLayers() {
+    return 256u;
+}
+
 void NoopDriver::updateIndexBuffer(Handle<HwIndexBuffer> ibh, BufferDescriptor&& p,
         uint32_t byteOffset) {
     scheduleDestroy(std::move(p));

filament/backend/src/noop/PlatformNoop.cpp

@@ -20,7 +20,7 @@
 
 namespace filament::backend {
 
-Driver* PlatformNoop::createDriver(void* const sharedGLContext, const Platform::DriverConfig& driverConfig) noexcept {
+Driver* PlatformNoop::createDriver(void* sharedGLContext, const Platform::DriverConfig& driverConfig) noexcept {
     return NoopDriver::create();
 }
 

filament/backend/src/opengl/OpenGLContext.cpp

@@ -94,18 +94,22 @@ OpenGLContext::OpenGLContext(OpenGLPlatform& platform,
     }
     #endif
 
-    OpenGLContext::initExtensions(&ext, state.major, state.minor);
+    initExtensions(&ext, state.major, state.minor);
 
-    OpenGLContext::initProcs(&procs, ext, state.major, state.minor);
+    initProcs(&procs, ext, state.major, state.minor);
 
-    OpenGLContext::initBugs(&bugs, ext, state.major, state.minor,
+    initBugs(&bugs, ext, state.major, state.minor,
             state.vendor, state.renderer, state.version, state.shader);
 
     glGetIntegerv(GL_MAX_RENDERBUFFER_SIZE,             &gets.max_renderbuffer_size);
     glGetIntegerv(GL_MAX_TEXTURE_IMAGE_UNITS,           &gets.max_texture_image_units);
     glGetIntegerv(GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS,  &gets.max_combined_texture_image_units);
+    glGetIntegerv(GL_MAX_TEXTURE_SIZE,                  &gets.max_texture_size);
+    glGetIntegerv(GL_MAX_CUBE_MAP_TEXTURE_SIZE,         &gets.max_cubemap_texture_size);
+    glGetIntegerv(GL_MAX_3D_TEXTURE_SIZE,               &gets.max_3d_texture_size);
+    glGetIntegerv(GL_MAX_ARRAY_TEXTURE_LAYERS,          &gets.max_array_texture_layers);
 
-    mFeatureLevel = OpenGLContext::resolveFeatureLevel(state.major, state.minor, ext, gets, bugs);
+    mFeatureLevel = resolveFeatureLevel(state.major, state.minor, ext, gets, bugs);
 
 #ifdef BACKEND_OPENGL_VERSION_GLES
     mShaderModel = ShaderModel::MOBILE;
@@ -177,6 +181,14 @@ OpenGLContext::OpenGLContext(OpenGLPlatform& platform,
                     << gets.max_anisotropy << '\n'
             << "GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS = "
                     << gets.max_combined_texture_image_units << '\n'
+            << "GL_MAX_TEXTURE_SIZE = "
+                    << gets.max_texture_size << '\n'
+            << "GL_MAX_CUBE_MAP_TEXTURE_SIZE = "
+                    << gets.max_cubemap_texture_size << '\n'
+            << "GL_MAX_3D_TEXTURE_SIZE = "
+                    << gets.max_3d_texture_size << '\n'
+            << "GL_MAX_ARRAY_TEXTURE_LAYERS = "
+                    << gets.max_array_texture_layers << '\n'
             << "GL_MAX_DRAW_BUFFERS = "
                     << gets.max_draw_buffers << '\n'
             << "GL_MAX_RENDERBUFFER_SIZE = "
@@ -554,7 +566,11 @@ void OpenGLContext::initBugs(Bugs* bugs, Extensions const& exts,
         } else if (strstr(renderer, "AMD") ||
                    strstr(renderer, "ATI")) {
             // AMD/ATI GPU
-        } else if (strstr(vendor, "Mesa")) {
+        } else if (strstr(renderer, "Mozilla")) {
+            bugs->disable_invalidate_framebuffer = true;
+        }
+
+        if (strstr(vendor, "Mesa")) {
             // Seen on
             //  [Mesa],
             //  [llvmpipe (LLVM 17.0.6, 256 bits)],
@@ -562,8 +578,6 @@ void OpenGLContext::initBugs(Bugs* bugs, Extensions const& exts,
             //  [4.50]
             // not known which version are affected
             bugs->rebind_buffer_after_deletion = true;
-        } else if (strstr(renderer, "Mozilla")) {
-            bugs->disable_invalidate_framebuffer = true;
         }
     } else {
         // When running under ANGLE, it's a different set of workaround that we need.
@@ -575,6 +589,19 @@ void OpenGLContext::initBugs(Bugs* bugs, Extensions const& exts,
         }
     }
 
+    if (strstr(vendor, "Mozilla")) {
+        // Seen on
+        //  [Mozilla],
+        //  [GeForce GTX 980, or similar]
+        //    or [ANGLE (NVIDIA, NVIDIA GeForce GTX 980 Direct3D11 vs_5_0 ps_5_0), or similar]
+        //    or anything else,
+        //  [OpenGL ES 3.0 (WebGL 2.0)],
+        //  [OpenGL ES GLSL ES 3.00 (WebGL GLSL ES 3.00)]
+        // For Mozilla, the issue appears to be observed regardless of whether the renderer is
+        // ANGLE or not. (b/376125497)
+        bugs->rebind_buffer_after_deletion = true;
+    }
+
 #ifdef BACKEND_OPENGL_VERSION_GLES
 #   ifndef FILAMENT_IOS // FILAMENT_IOS is guaranteed to have ES3.x
     if (UTILS_UNLIKELY(major == 2)) {

filament/backend/src/opengl/OpenGLContext.h

@@ -203,6 +203,10 @@ public:
         GLint max_renderbuffer_size;
         GLint max_samples;
         GLint max_texture_image_units;
+        GLint max_texture_size;
+        GLint max_cubemap_texture_size;
+        GLint max_3d_texture_size;
+        GLint max_array_texture_layers;
         GLint max_transform_feedback_separate_attribs;
         GLint max_uniform_block_size;
         GLint max_uniform_buffer_bindings;
@@ -316,8 +320,8 @@ public:
         // a glFinish. So we must delay the destruction until we know the GPU is finished.
         bool delay_fbo_destruction;
 
-        // Mesa sometimes clears the generic buffer binding when *another* buffer is destroyed,
-        // if that other buffer is bound on an *indexed* buffer binding.
+        // Mesa and Mozilla(web) sometimes clear the generic buffer binding when *another* buffer
+        // is destroyed, if that other buffer is bound on an *indexed* buffer binding.
         bool rebind_buffer_after_deletion;
 
         // Force feature level 0. Typically used for low end ES3 devices with significant driver

filament/backend/src/opengl/OpenGLDriver.cpp

@@ -148,8 +148,8 @@ using namespace utils;
 namespace filament::backend {
 
 Driver* OpenGLDriverFactory::create(
-        OpenGLPlatform* const platform,
-        void* const sharedGLContext,
+        OpenGLPlatform* platform,
+        void* sharedGLContext,
         const Platform::DriverConfig& driverConfig) noexcept {
     return OpenGLDriver::create(platform, sharedGLContext, driverConfig);
 }
@@ -159,10 +159,10 @@ using namespace GLUtils;
 // ------------------------------------------------------------------------------------------------
 
 UTILS_NOINLINE
-OpenGLDriver* OpenGLDriver::create(OpenGLPlatform* const platform,
-        void* const /*sharedGLContext*/, const Platform::DriverConfig& driverConfig) noexcept {
+OpenGLDriver* OpenGLDriver::create(OpenGLPlatform* platform,
+        void* /*sharedGLContext*/, const Platform::DriverConfig& driverConfig) noexcept {
     assert_invariant(platform);
-    OpenGLPlatform* const ec = platform;
+    OpenGLPlatform* ec = platform;
 
 #if 0
     // this is useful for development, but too verbose even for debug builds
@@ -230,7 +230,7 @@ OpenGLDriver* OpenGLDriver::create(OpenGLPlatform* const platform,
     constexpr size_t defaultSize = FILAMENT_OPENGL_HANDLE_ARENA_SIZE_IN_MB * 1024U * 1024U;
     Platform::DriverConfig validConfig{ driverConfig };
     validConfig.handleArenaSize = std::max(driverConfig.handleArenaSize, defaultSize);
-    OpenGLDriver* const driver = new(std::nothrow) OpenGLDriver(ec, validConfig);
+    OpenGLDriver* driver = new(std::nothrow) OpenGLDriver(ec, validConfig);
     return driver;
 }
 
@@ -2460,6 +2460,26 @@ size_t OpenGLDriver::getMaxUniformBufferSize() {
     return mContext.gets.max_uniform_block_size;
 }
 
+size_t OpenGLDriver::getMaxTextureSize(SamplerType target) {
+    switch (target) {
+        case SamplerType::SAMPLER_2D:
+        case SamplerType::SAMPLER_2D_ARRAY:
+        case SamplerType::SAMPLER_EXTERNAL:
+            return mContext.gets.max_texture_size;
+        case SamplerType::SAMPLER_CUBEMAP:
+            return mContext.gets.max_cubemap_texture_size;
+        case SamplerType::SAMPLER_3D:
+            return mContext.gets.max_3d_texture_size;
+        case SamplerType::SAMPLER_CUBEMAP_ARRAY:
+            return mContext.gets.max_cubemap_texture_size;
+    }
+    return 0;
+}
+
+size_t OpenGLDriver::getMaxArrayTextureLayers() {
+    return mContext.gets.max_array_texture_layers;
+}
+
 // ------------------------------------------------------------------------------------------------
 // Swap chains
 // ------------------------------------------------------------------------------------------------

filament/backend/src/opengl/platforms/PlatformCocoaTouchGL.mm

@@ -64,7 +64,7 @@ PlatformCocoaTouchGL::~PlatformCocoaTouchGL() noexcept {
     delete pImpl;
 }
 
-Driver* PlatformCocoaTouchGL::createDriver(void* const sharedGLContext, const Platform::DriverConfig& driverConfig) noexcept {
+Driver* PlatformCocoaTouchGL::createDriver(void* sharedGLContext, const Platform::DriverConfig& driverConfig) noexcept {
     EAGLSharegroup* sharegroup = (__bridge EAGLSharegroup*) sharedGLContext;
 
     EAGLContext *context = [[EAGLContext alloc] initWithAPI:kEAGLRenderingAPIOpenGLES3 sharegroup:sharegroup];

filament/backend/src/opengl/platforms/PlatformEGLAndroid.cpp

@@ -19,6 +19,7 @@
 #include <backend/platforms/PlatformEGL.h>
 #include <backend/platforms/PlatformEGLAndroid.h>
 
+#include <private/backend/BackendUtilsAndroid.h>
 #include <private/backend/VirtualMachineEnv.h>
 
 #include "opengl/GLUtils.h"
@@ -34,6 +35,8 @@
 #include <utils/ostream.h>
 #include <utils/Panic.h>
 #include <utils/Log.h>
+#include <utils/compiler.h>
+#include <utils/ostream.h>
 
 #include <EGL/egl.h>
 #include <EGL/eglext.h>
@@ -231,26 +234,133 @@ Driver* PlatformEGLAndroid::createDriver(void* sharedContext,
     return driver;
 }
 
-PlatformEGLAndroid::ExternalImageEGLAndroid::~ExternalImageEGLAndroid() = default;
+PlatformEGLAndroid::ExternalImageEGLAndroid::~ExternalImageEGLAndroid() {
+    if (__builtin_available(android 26, *)) {
+        if (aHardwareBuffer) {
+            AHardwareBuffer_release(aHardwareBuffer);
+        }
+    }
+}
 
-Platform::ExternalImageHandle PlatformEGLAndroid::createExternalImage(AHardwareBuffer const* buffer, bool sRGB) noexcept {
-    auto* const p = new(std::nothrow) ExternalImageEGLAndroid;
-    p->aHardwareBuffer = const_cast<AHardwareBuffer*>(buffer);
-    p->sRGB = sRGB;
-    return ExternalImageHandle{ p };
+Platform::ExternalImageHandle PlatformEGLAndroid::createExternalImage(AHardwareBuffer const* buffer,
+        bool sRGB) noexcept {
+    if (__builtin_available(android 26, *)) {
+        auto* const p = new (std::nothrow) ExternalImageEGLAndroid;
+        auto hardwareBuffer = const_cast<AHardwareBuffer*>(buffer);
+        AHardwareBuffer_acquire(hardwareBuffer);
+        p->aHardwareBuffer = hardwareBuffer;
+        p->sRGB = sRGB;
+        AHardwareBuffer_Desc hardwareBufferDescription = {};
+        AHardwareBuffer_describe(hardwareBuffer, &hardwareBufferDescription);
+        p->height = hardwareBufferDescription.height;
+        p->width = hardwareBufferDescription.width;
+        auto textureFormat = mapToFilamentFormat(hardwareBufferDescription.format, sRGB);
+        p->usage = mapToFilamentUsage(hardwareBufferDescription.usage, textureFormat);
+        return ExternalImageHandle{ p };
+    }
+
+    return Platform::ExternalImageHandle{};
+}
+
+PlatformEGLAndroid::ExternalImageDescAndroid PlatformEGLAndroid::getExternalImageDesc(
+        ExternalImageHandle externalImage) noexcept {
+    auto const* const eglExternalImage =
+            static_cast<ExternalImageEGLAndroid const*>(externalImage.get());
+    ExternalImageDescAndroid metadata = {};
+    if (!eglExternalImage) {
+        return metadata;
+    }
+    metadata.height = eglExternalImage->height;
+    metadata.width = eglExternalImage->width;
+    metadata.format = eglExternalImage->format;
+    metadata.usage = eglExternalImage->usage;
+    return metadata;
 }
 
 bool PlatformEGLAndroid::setExternalImage(ExternalImageHandleRef externalImage,
         UTILS_UNUSED_IN_RELEASE ExternalTexture* texture) noexcept {
-    auto const* const eglExternalImage = static_cast<ExternalImageEGLAndroid const*>(externalImage.get());
+    auto const* const eglExternalImage =
+            static_cast<ExternalImageEGLAndroid const*>(externalImage.get());
     if (eglExternalImage->aHardwareBuffer) {
-        // TODO: implement PlatformEGLAndroid::setExternalImage w/ AHardwareBuffer
-        return true;
+        return PlatformEGLAndroid::setImage(eglExternalImage, texture);
     }
     // not a AHardwareBuffer, fallback to the inherited version
     return PlatformEGL::setExternalImage(externalImage, texture);
 }
 
+OpenGLPlatform::ExternalTexture* PlatformEGLAndroid::createExternalImageTexture() noexcept {
+    ExternalTexture* outTexture = new (std::nothrow) ExternalTexture{};
+    glGenTextures(1, &outTexture->id);
+    return outTexture;
+}
+
+void PlatformEGLAndroid::destroyExternalImageTexture(ExternalTexture* texture) noexcept {
+    glDeleteTextures(1, &texture->id);
+    delete texture;
+}
+
+bool PlatformEGLAndroid::setImage(ExternalImageEGLAndroid const* eglExternalImage,
+        UTILS_UNUSED_IN_RELEASE ExternalTexture* texture) noexcept {
+    AHardwareBuffer* hardwareBuffer = eglExternalImage->aHardwareBuffer;
+
+    // Get the EGL client buffer from AHardwareBuffer
+    EGLClientBuffer clientBuffer = eglGetNativeClientBufferANDROID(hardwareBuffer);
+    EGLint imageAttrs[] = {
+        EGL_IMAGE_PRESERVED_KHR, EGL_TRUE,
+        EGL_NONE, EGL_NONE,  // Reserve space
+        EGL_NONE, EGL_NONE,  // Reserve space
+        EGL_NONE             // Ensure the list always ends with EGL_NONE
+    };
+    int attrIndex = 2;
+    if (eglExternalImage->sRGB) {
+        imageAttrs[attrIndex++] = EGL_GL_COLORSPACE;
+        imageAttrs[attrIndex++] = EGL_GL_COLORSPACE_SRGB;
+    }
+
+    if (static_cast<bool>(eglExternalImage->usage & TextureUsage::PROTECTED)) {
+        imageAttrs[attrIndex++] = EGL_PROTECTED_CONTENT_EXT;
+        imageAttrs[attrIndex++] = EGL_TRUE;
+    }
+    // Create an EGLImage from the client buffer
+    EGLImageKHR eglImage = eglCreateImageKHR(eglGetCurrentDisplay(), EGL_NO_CONTEXT,
+            EGL_NATIVE_BUFFER_ANDROID, clientBuffer, imageAttrs);
+    if (eglImage == EGL_NO_IMAGE_KHR) {
+        // Handle error
+        slog.e << "Failed to create EGL image" << io::endl;
+        glDeleteTextures(1, &texture->id);
+        return false;
+    }
+
+    // Create and bind the OpenGL texture
+    GLint prevActiveTexture, prevTexture;
+    glGetIntegerv(GL_ACTIVE_TEXTURE, &prevActiveTexture);
+    glGetIntegerv(GL_TEXTURE_BINDING_2D, &prevTexture);
+    glActiveTexture(GL_TEXTURE0);
+    glBindTexture(texture->target, texture->id);
+    GLenum error = glGetError();
+    if (UTILS_UNLIKELY(error != GL_NO_ERROR)) {
+        slog.e << "Error after glBindTexture: " << error << io::endl;
+        glDeleteTextures(1, &texture->id);
+        eglDestroyImageKHR(eglGetCurrentDisplay(), eglImage);
+        glActiveTexture(prevActiveTexture);
+        glBindTexture(GL_TEXTURE_2D, prevTexture);
+        return false;
+    }
+    glEGLImageTargetTexture2DOES(texture->target, static_cast<GLeglImageOES>(eglImage));
+    error = glGetError();
+    if (UTILS_UNLIKELY(error != GL_NO_ERROR)) {
+        slog.e << "Error after glEGLImageTargetTexture2DOES: " << error << io::endl;
+        glDeleteTextures(1, &texture->id);
+        eglDestroyImageKHR(eglGetCurrentDisplay(), eglImage);
+        glActiveTexture(prevActiveTexture);
+        glBindTexture(GL_TEXTURE_2D, prevTexture);
+        return false;
+    }
+    glActiveTexture(prevActiveTexture);
+    glBindTexture(GL_TEXTURE_2D, prevTexture);
+    return true;
+}
+
 void PlatformEGLAndroid::setPresentationTime(int64_t presentationTimeInNanosecond) noexcept {
     EGLSurface currentDrawSurface = eglGetCurrentSurface(EGL_DRAW);
     if (currentDrawSurface != EGL_NO_SURFACE) {

filament/backend/src/opengl/platforms/PlatformGLX.cpp

@@ -128,7 +128,7 @@ namespace filament::backend {
 
 using namespace backend;
 
-Driver* PlatformGLX::createDriver(void* const sharedGLContext,
+Driver* PlatformGLX::createDriver(void* sharedGLContext,
         const DriverConfig& driverConfig) noexcept {
     loadLibraries();
     // Get the display device

filament/backend/src/opengl/platforms/PlatformOSMesa.cpp

@@ -22,6 +22,12 @@
 #include <dlfcn.h>
 #include <memory>
 
+#if defined(__linux__)
+// This is to ensure that linking during compilation will not fail even if
+// OSMesaGetProcAddress is not linked.
+__attribute__((weak)) OSMESAproc OSMesaGetProcAddress(char const*);
+#endif
+
 namespace filament::backend {
 
 using namespace backend;
@@ -44,20 +50,27 @@ struct OSMesaSwapchain {
 
 struct OSMesaAPI {
 private:
-    using CreateContextFunc = OSMesaContext (*)(GLenum format, OSMesaContext);
+    using CreateContextAttribsFunc = OSMesaContext (*)(const int *, OSMesaContext);
     using DestroyContextFunc = GLboolean (*)(OSMesaContext);
     using MakeCurrentFunc = GLboolean (*)(OSMesaContext ctx, void* buffer, GLenum type,
             GLsizei width, GLsizei height);
     using GetProcAddressFunc = OSMESAproc (*)(const char* funcName);
 
 public:
-    CreateContextFunc OSMesaCreateContext;
-    DestroyContextFunc OSMesaDestroyContext;
-    MakeCurrentFunc OSMesaMakeCurrent;
-    GetProcAddressFunc OSMesaGetProcAddress;
+    CreateContextAttribsFunc fOSMesaCreateContextAttribs;
+    DestroyContextFunc fOSMesaDestroyContext;
+    MakeCurrentFunc fOSMesaMakeCurrent;
+    GetProcAddressFunc fOSMesaGetProcAddress;
 
     OSMesaAPI() {
-        constexpr char const* libraryNames[] = {"libOSMesa.so", "libosmesa.so"};
+        static constexpr char const* libraryNames[] = {
+#if defined(__linux__)
+            "libOSMesa.so",
+            "libosmesa.so",
+#elif defined(__APPLE__)
+            "libOSMesa.dylib",
+#endif
+        };
         for (char const* libName: libraryNames) {
             mLib = dlopen(libName, RTLD_GLOBAL | RTLD_NOW);
             if (mLib) {
@@ -65,18 +78,27 @@ public:
             }
         }
         if (mLib) {
-            OSMesaGetProcAddress = (GetProcAddressFunc) dlsym(mLib, "OSMesaGetProcAddress");
-        } else {
-            OSMesaGetProcAddress = (GetProcAddressFunc) dlsym(RTLD_LOCAL, "OSMesaGetProcAddress");
+            // Loading from a libosmesa.os
+            fOSMesaGetProcAddress = (GetProcAddressFunc) dlsym(mLib, "OSMesaGetProcAddress");
         }
+#if defined(__linux__)
+        else {
+            // Filament is built into a .so
+            fOSMesaGetProcAddress = (GetProcAddressFunc) dlsym(RTLD_LOCAL, "OSMesaGetProcAddress");
+        }
+        if (!fOSMesaGetProcAddress) {
+            // Statically linking osmesa
+            fOSMesaGetProcAddress = OSMesaGetProcAddress;
+        }
+#endif // __linux__
 
-        FILAMENT_CHECK_PRECONDITION(OSMesaGetProcAddress)
-                << "Unable to against libOSMesa to create a software GL context";
+        FILAMENT_CHECK_PRECONDITION(fOSMesaGetProcAddress)
+                << "Unable to link against libOSMesa to create a software GL context";
 
-        OSMesaCreateContext = (CreateContextFunc) OSMesaGetProcAddress("OSMesaCreateContext");
-        OSMesaDestroyContext =
-                (DestroyContextFunc) OSMesaGetProcAddress("OSMesaDestroyContext");
-        OSMesaMakeCurrent = (MakeCurrentFunc) OSMesaGetProcAddress("OSMesaMakeCurrent");
+        fOSMesaCreateContextAttribs =
+                (CreateContextAttribsFunc) fOSMesaGetProcAddress("OSMesaCreateContextAttribs");
+        fOSMesaDestroyContext = (DestroyContextFunc) fOSMesaGetProcAddress("OSMesaDestroyContext");
+        fOSMesaMakeCurrent = (MakeCurrentFunc) fOSMesaGetProcAddress("OSMesaMakeCurrent");
     }
 
     ~OSMesaAPI() {
@@ -90,14 +112,24 @@ private:
 
 }// anonymous namespace
 
-Driver* PlatformOSMesa::createDriver(void* const sharedGLContext,
+Driver* PlatformOSMesa::createDriver(void* sharedGLContext,
         const DriverConfig& driverConfig) noexcept {
+
     OSMesaAPI* api = new OSMesaAPI();
     mOsMesaApi = api;
 
+    static constexpr int attribs[] = {
+        OSMESA_FORMAT, GL_RGBA,
+        OSMESA_DEPTH_BITS, 24,
+        OSMESA_STENCIL_BITS, 8,
+        OSMESA_ACCUM_BITS, 0,
+        OSMESA_PROFILE, OSMESA_CORE_PROFILE,
+        0,
+    };
+
     FILAMENT_CHECK_PRECONDITION(sharedGLContext == nullptr)
             << "shared GL context is not supported with PlatformOSMesa";
-    mContext = api->OSMesaCreateContext(GL_RGBA, NULL);
+    mContext = api->fOSMesaCreateContextAttribs(attribs, NULL);
 
     // We need to do a no-op makecurrent here so that the context will be in a correct state before
     // any GL calls.
@@ -113,7 +145,7 @@ Driver* PlatformOSMesa::createDriver(void* const sharedGLContext,
 
 void PlatformOSMesa::terminate() noexcept {
     OSMesaAPI* api = (OSMesaAPI*) mOsMesaApi;
-    api->OSMesaDestroyContext(mContext);
+    api->fOSMesaDestroyContext(mContext);
     delete api;
     mOsMesaApi = nullptr;
 
@@ -141,7 +173,7 @@ bool PlatformOSMesa::makeCurrent(ContextType type, SwapChain* drawSwapChain,
     OSMesaAPI* api = (OSMesaAPI*) mOsMesaApi;
     OSMesaSwapchain* impl = (OSMesaSwapchain*) drawSwapChain;
 
-    auto result = api->OSMesaMakeCurrent(mContext, (BackingType*) impl->buffer.get(),
+    auto result = api->fOSMesaMakeCurrent(mContext, (BackingType*) impl->buffer.get(),
             BACKING_GL_TYPE, impl->width, impl->height);
     FILAMENT_CHECK_POSTCONDITION(result == GL_TRUE) << "OSMesaMakeCurrent failed!";
 

filament/backend/src/opengl/platforms/PlatformWGL.cpp

@@ -75,7 +75,7 @@ struct WGLSwapChain {
 
 static PFNWGLCREATECONTEXTATTRIBSARBPROC wglCreateContextAttribs = nullptr;
 
-Driver* PlatformWGL::createDriver(void* const sharedGLContext,
+Driver* PlatformWGL::createDriver(void* sharedGLContext,
         const Platform::DriverConfig& driverConfig) noexcept {
     int result = 0;
     int pixelFormat = 0;

filament/backend/src/opengl/platforms/PlatformWebGL.cpp

@@ -20,7 +20,7 @@ namespace filament::backend {
 
 using namespace backend;
 
-Driver* PlatformWebGL::createDriver(void* const sharedGLContext,
+Driver* PlatformWebGL::createDriver(void* sharedGLContext,
         const Platform::DriverConfig& driverConfig) noexcept {
     return OpenGLPlatform::createDefaultDriver(this, sharedGLContext, driverConfig);
 }

filament/backend/src/vulkan/VulkanDescriptorSetCache.cpp

@@ -14,35 +14,25 @@
  * limitations under the License.
  */
 
-#include "VulkanDescriptorSetManager.h"
+#include "VulkanDescriptorSetCache.h"
 
-#include "vulkan/VulkanCommands.h"
-#include "vulkan/VulkanHandles.h"
-#include "vulkan/VulkanConstants.h"
+#include "VulkanCommands.h"
+#include "VulkanHandles.h"
+#include "VulkanConstants.h"
 
 #include <utils/FixedCapacityVector.h>
 #include <utils/Panic.h>
 
-#include <math.h>
-
 #include <algorithm>
 #include <memory>
-#include <type_traits>
 #include <vector>
 
 namespace filament::backend {
 
 namespace {
 
-using BitmaskGroup = VulkanDescriptorSetLayout::Bitmask;
 using DescriptorCount = VulkanDescriptorSetLayout::Count;
-using DescriptorSetLayoutArray = VulkanDescriptorSetManager::DescriptorSetLayoutArray;
-using BitmaskGroupHashFn = utils::hash::MurmurHashFn<BitmaskGroup>;
-struct BitmaskGroupEqual {
-    bool operator()(BitmaskGroup const& k1, BitmaskGroup const& k2) const {
-        return k1 == k2;
-    }
-};
+using DescriptorSetLayoutArray = VulkanDescriptorSetCache::DescriptorSetLayoutArray;
 
 // We create a pool for each layout as defined by the number of descriptors of each type. For
 // example, a layout of
@@ -199,72 +189,12 @@ struct Equal {
     }
 };
 
-template<typename Bitmask>
-uint32_t createBindings(VkDescriptorSetLayoutBinding* toBind, uint32_t count, VkDescriptorType type,
-        Bitmask const& mask) {
-    Bitmask alreadySeen;
-    mask.forEachSetBit([&](size_t index) {
-        VkShaderStageFlags stages = 0;
-        uint32_t binding = 0;
-        if (index < fvkutils::getFragmentStageShift<Bitmask>()) {
-            binding = (uint32_t) index;
-            stages |= VK_SHADER_STAGE_VERTEX_BIT;
-            auto fragIndex = index + fvkutils::getFragmentStageShift<Bitmask>();
-            if (mask.test(fragIndex)) {
-                stages |= VK_SHADER_STAGE_FRAGMENT_BIT;
-                alreadySeen.set(fragIndex);
-            }
-        } else if (!alreadySeen.test(index)) {
-            // We are in fragment stage bits
-            binding = (uint32_t) (index - fvkutils::getFragmentStageShift<Bitmask>());
-            stages |= VK_SHADER_STAGE_FRAGMENT_BIT;
-        }
-
-        if (stages) {
-            toBind[count++] = {
-                .binding = binding,
-                .descriptorType = type,
-                .descriptorCount = 1,
-                .stageFlags = stages,
-            };
-        }
-    });
-    return count;
-}
-
-inline VkDescriptorSetLayout createLayout(VkDevice device, BitmaskGroup const& bitmaskGroup) {
-    // Note that the following *needs* to be static so that VkDescriptorSetLayoutCreateInfo will not
-    // refer to stack memory.
-    VkDescriptorSetLayoutBinding toBind[VulkanDescriptorSetLayout::MAX_BINDINGS];
-    uint32_t count = 0;
-
-    count = createBindings(toBind, count, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC,
-            bitmaskGroup.dynamicUbo);
-    count = createBindings(toBind, count, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, bitmaskGroup.ubo);
-    count = createBindings(toBind, count, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
-            bitmaskGroup.sampler);
-    count = createBindings(toBind, count, VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT,
-            bitmaskGroup.inputAttachment);
-
-    assert_invariant(count != 0 && "Need at least one binding for descriptor set layout.");
-    VkDescriptorSetLayoutCreateInfo dlinfo = {
-            .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
-            .pNext = nullptr,
-            .bindingCount = count,
-            .pBindings = toBind,
-    };
-
-    VkDescriptorSetLayout layout;
-    vkCreateDescriptorSetLayout(device, &dlinfo, VKALLOC, &layout);
-    return layout;
-}
-
 } // anonymous namespace
 
 // This is an ever-expanding pool of sets where it
 //    1. Keeps a list of smaller pools of different layout-dimensions.
 //    2. Will add a pool if existing pool are not compatible with the requested layout o runs out.
-class VulkanDescriptorSetManager::DescriptorInfinitePool {
+class VulkanDescriptorSetCache::DescriptorInfinitePool {
 private:
     static constexpr uint16_t EXPECTED_SET_COUNT = 10;
     static constexpr float SET_COUNT_GROWTH_FACTOR = 1.5;
@@ -319,61 +249,34 @@ private:
     std::vector<std::unique_ptr<DescriptorPool>> mPools;
 };
 
-class VulkanDescriptorSetManager::DescriptorSetLayoutManager {
-public:
-    DescriptorSetLayoutManager(VkDevice device)
-        : mDevice(device) {}
 
-    VkDescriptorSetLayout getVkLayout(VulkanDescriptorSetLayout::Bitmask const& bitmasks) {
-        if (auto itr = mVkLayouts.find(bitmasks); itr != mVkLayouts.end()) {
-            return itr->second;
-        }
-        auto vklayout = createLayout(mDevice, bitmasks);
-        mVkLayouts[bitmasks] = vklayout;
-        return vklayout;
-    }
-
-    ~DescriptorSetLayoutManager() {
-        for (auto& itr: mVkLayouts) {
-            vkDestroyDescriptorSetLayout(mDevice, itr.second, VKALLOC);
-        }
-    }
-
-private:
-    VkDevice mDevice;
-    tsl::robin_map<BitmaskGroup, VkDescriptorSetLayout, BitmaskGroupHashFn, BitmaskGroupEqual>
-            mVkLayouts;
-};
-
-VulkanDescriptorSetManager::VulkanDescriptorSetManager(VkDevice device,
+VulkanDescriptorSetCache::VulkanDescriptorSetCache(VkDevice device,
         fvkmemory::ResourceManager* resourceManager)
     : mDevice(device),
       mResourceManager(resourceManager),
-      mLayoutManager(std::make_unique<DescriptorSetLayoutManager>(device)),
       mDescriptorPool(std::make_unique<DescriptorInfinitePool>(device)) {}
 
-VulkanDescriptorSetManager::~VulkanDescriptorSetManager() = default;
+VulkanDescriptorSetCache::~VulkanDescriptorSetCache() = default;
 
-void VulkanDescriptorSetManager::terminate() noexcept{
-    mLayoutManager.reset();
+void VulkanDescriptorSetCache::terminate() noexcept{
     mDescriptorPool.reset();
     clearHistory();
 }
 
 // bind() is not really binding the set but just stashing until we have all the info
 // (pipelinelayout).
-void VulkanDescriptorSetManager::bind(uint8_t setIndex,
+void VulkanDescriptorSetCache::bind(uint8_t setIndex,
         fvkmemory::resource_ptr<VulkanDescriptorSet> set,
         backend::DescriptorSetOffsetArray&& offsets) {
     set->setOffsets(std::move(offsets));
     mStashedSets[setIndex] = set;
 }
 
-void VulkanDescriptorSetManager::unbind(uint8_t setIndex) {
+void VulkanDescriptorSetCache::unbind(uint8_t setIndex) {
     mStashedSets[setIndex] = {};
 }
 
-void VulkanDescriptorSetManager::commit(VulkanCommandBuffer* commands,
+void VulkanDescriptorSetCache::commit(VulkanCommandBuffer* commands,
         VkPipelineLayout pipelineLayout, fvkutils::DescriptorSetMask const& setMask) {
     // setMask indicates the set of descriptor sets the driver wants to bind, curMask is the
     // actual set of sets that *needs* to be bound.
@@ -412,7 +315,7 @@ void VulkanDescriptorSetManager::commit(VulkanCommandBuffer* commands,
     };
 }
 
-void VulkanDescriptorSetManager::updateBuffer(fvkmemory::resource_ptr<VulkanDescriptorSet> set,
+void VulkanDescriptorSetCache::updateBuffer(fvkmemory::resource_ptr<VulkanDescriptorSet> set,
         uint8_t binding, fvkmemory::resource_ptr<VulkanBufferObject> bufferObject,
         VkDeviceSize offset, VkDeviceSize size) noexcept {
     VkDescriptorBufferInfo const info = {
@@ -438,7 +341,7 @@ void VulkanDescriptorSetManager::updateBuffer(fvkmemory::resource_ptr<VulkanDesc
     set->acquire(bufferObject);
 }
 
-void VulkanDescriptorSetManager::updateSampler(fvkmemory::resource_ptr<VulkanDescriptorSet> set,
+void VulkanDescriptorSetCache::updateSampler(fvkmemory::resource_ptr<VulkanDescriptorSet> set,
         uint8_t binding, fvkmemory::resource_ptr<VulkanTexture> texture,
         VkSampler sampler) noexcept {
     VkDescriptorImageInfo info{
@@ -470,13 +373,13 @@ void VulkanDescriptorSetManager::updateSampler(fvkmemory::resource_ptr<VulkanDes
     set->acquire(texture);
 }
 
-void VulkanDescriptorSetManager::updateInputAttachment(
+void VulkanDescriptorSetCache::updateInputAttachment(
         fvkmemory::resource_ptr<VulkanDescriptorSet> set,
         VulkanAttachment const& attachment) noexcept {
-    // TOOD: fill-in this region
+    // TOOD: fill this in.
 }
 
-fvkmemory::resource_ptr<VulkanDescriptorSet> VulkanDescriptorSetManager::createSet(
+fvkmemory::resource_ptr<VulkanDescriptorSet> VulkanDescriptorSetCache::createSet(
         Handle<HwDescriptorSet> handle, fvkmemory::resource_ptr<VulkanDescriptorSetLayout> layout) {
     auto const vkSet = mDescriptorPool->obtainSet(layout);
     auto const& count = layout->count;
@@ -492,12 +395,7 @@ fvkmemory::resource_ptr<VulkanDescriptorSet> VulkanDescriptorSetManager::createS
             });
 }
 
-void VulkanDescriptorSetManager::initVkLayout(
-        fvkmemory::resource_ptr<VulkanDescriptorSetLayout> layout) {
-    layout->setVkLayout(mLayoutManager->getVkLayout(layout->bitmask));
-}
-
-void VulkanDescriptorSetManager::clearHistory() {
+void VulkanDescriptorSetCache::clearHistory() {
     mStashedSets = {};
 }
 

filament/backend/src/vulkan/VulkanDescriptorSetCache.h

@@ -14,10 +14,10 @@
  * limitations under the License.
  */
 
-#ifndef TNT_FILAMENT_BACKEND_CACHING_VULKANDESCRIPTORSETMANAGER_H
-#define TNT_FILAMENT_BACKEND_CACHING_VULKANDESCRIPTORSETMANAGER_H
+#ifndef TNT_FILAMENT_BACKEND_CACHING_VULKANDESCRIPTORSETCACHE_H
+#define TNT_FILAMENT_BACKEND_CACHING_VULKANDESCRIPTORSETCACHE_H
 
-#include "vulkan/VulkanHandles.h"
+#include "VulkanHandles.h"
 #include "vulkan/memory/ResourcePointer.h"
 #include "vulkan/utils/Definitions.h"  // For DescriptorSetMask
 
@@ -34,20 +34,16 @@
 
 namespace filament::backend {
 
-// [GDSR]: Great-Descriptor-Set-Refactor: As of 03/20/24, the Filament frontend is planning to
-// introduce descriptor set. This PR will arrive before that change is complete. As such, some of
-// the methods introduced here will be obsolete, and certain logic will be generalized.
-
-// Abstraction over the pool and the layout cache.
-class VulkanDescriptorSetManager {
+// Abstraction over the descriptor set pool.
+class VulkanDescriptorSetCache {
 public:
     static constexpr uint8_t UNIQUE_DESCRIPTOR_SET_COUNT =
             VulkanDescriptorSetLayout::UNIQUE_DESCRIPTOR_SET_COUNT;
 
     using DescriptorSetLayoutArray = VulkanDescriptorSetLayout::DescriptorSetLayoutArray;
 
-    VulkanDescriptorSetManager(VkDevice device, fvkmemory::ResourceManager* resourceManager);
-    ~VulkanDescriptorSetManager();
+    VulkanDescriptorSetCache(VkDevice device, fvkmemory::ResourceManager* resourceManager);
+    ~VulkanDescriptorSetCache();
 
     void terminate() noexcept;
 
@@ -72,12 +68,9 @@ public:
     fvkmemory::resource_ptr<VulkanDescriptorSet> createSet(Handle<HwDescriptorSet> handle,
             fvkmemory::resource_ptr<VulkanDescriptorSetLayout> layout);
 
-    void initVkLayout(fvkmemory::resource_ptr<VulkanDescriptorSetLayout> layout);
-
     void clearHistory();
 
 private:
-    class DescriptorSetLayoutManager;
     class DescriptorInfinitePool;
 
     using DescriptorSetArray =
@@ -85,7 +78,6 @@ private:
 
     VkDevice mDevice;
     fvkmemory::ResourceManager* mResourceManager;
-    std::unique_ptr<DescriptorSetLayoutManager> mLayoutManager;
     std::unique_ptr<DescriptorInfinitePool> mDescriptorPool;
     std::pair<VulkanAttachment, VkDescriptorImageInfo> mInputAttachment;
     DescriptorSetArray mStashedSets = {};
@@ -99,4 +91,4 @@ private:
 
 }// namespace filament::backend
 
-#endif// TNT_FILAMENT_BACKEND_CACHING_VULKANDESCRIPTORSETMANAGER_H
+#endif// TNT_FILAMENT_BACKEND_CACHING_VULKANDESCRIPTORSETCACHE_H

filament/backend/src/vulkan/VulkanDescriptorSetLayoutCache.cpp

@@ -0,0 +1,110 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "VulkanDescriptorSetLayoutCache.h"
+
+#include "VulkanHandles.h"
+
+namespace filament::backend {
+
+namespace {
+
+using BitmaskGroup = VulkanDescriptorSetLayout::Bitmask;
+
+template<typename Bitmask>
+uint32_t appendBindings(VkDescriptorSetLayoutBinding* toBind, VkDescriptorType type,
+        Bitmask const& mask) {
+    uint32_t count = 0;
+    Bitmask alreadySeen;
+    mask.forEachSetBit([&](size_t index) {
+        VkShaderStageFlags stages = 0;
+        uint32_t binding = 0;
+        if (index < fvkutils::getFragmentStageShift<Bitmask>()) {
+            binding = (uint32_t) index;
+            stages |= VK_SHADER_STAGE_VERTEX_BIT;
+            auto fragIndex = index + fvkutils::getFragmentStageShift<Bitmask>();
+            if (mask.test(fragIndex)) {
+                stages |= VK_SHADER_STAGE_FRAGMENT_BIT;
+                alreadySeen.set(fragIndex);
+            }
+        } else if (!alreadySeen.test(index)) {
+            // We are in fragment stage bits
+            binding = (uint32_t) (index - fvkutils::getFragmentStageShift<Bitmask>());
+            stages |= VK_SHADER_STAGE_FRAGMENT_BIT;
+        }
+
+        if (stages) {
+            toBind[count++] = {
+                .binding = binding,
+                .descriptorType = type,
+                .descriptorCount = 1,
+                .stageFlags = stages,
+            };
+        }
+    });
+    return count;
+}
+
+} // anonymous namespace
+
+VulkanDescriptorSetLayoutCache::VulkanDescriptorSetLayoutCache(VkDevice device,
+        fvkmemory::ResourceManager* resourceManager)
+    : mDevice(device),
+      mResourceManager(resourceManager) {}
+
+VulkanDescriptorSetLayoutCache::~VulkanDescriptorSetLayoutCache() = default;
+
+void VulkanDescriptorSetLayoutCache::terminate() noexcept {
+    for (auto& itr: mVkLayouts) {
+        vkDestroyDescriptorSetLayout(mDevice, itr.second, VKALLOC);
+    }
+}
+
+fvkmemory::resource_ptr<VulkanDescriptorSetLayout> VulkanDescriptorSetLayoutCache::createLayout(
+        Handle<HwDescriptorSetLayout> handle, backend::DescriptorSetLayout&& info) {
+    auto layout = fvkmemory::resource_ptr<VulkanDescriptorSetLayout>::make(mResourceManager, handle,
+            info);
+    VkDescriptorSetLayout vklayout = VK_NULL_HANDLE;
+    auto const& bitmasks = layout->bitmask;
+    if (auto itr = mVkLayouts.find(bitmasks); itr != mVkLayouts.end()) {
+        vklayout = itr->second;
+    } else {
+        VkDescriptorSetLayoutBinding toBind[VulkanDescriptorSetLayout::MAX_BINDINGS];
+        uint32_t count = 0;
+        count += appendBindings(&toBind[count], VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC,
+                bitmasks.dynamicUbo);
+        count += appendBindings(&toBind[count], VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, bitmasks.ubo);
+        count += appendBindings(&toBind[count], VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+                bitmasks.sampler);
+        count += appendBindings(&toBind[count], VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT,
+                bitmasks.inputAttachment);
+
+        assert_invariant(count != 0 && "Need at least one binding for descriptor set layout.");
+        VkDescriptorSetLayoutCreateInfo dlinfo = {
+            .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
+            .pNext = nullptr,
+            .bindingCount = count,
+            .pBindings = toBind,
+        };
+
+        vkCreateDescriptorSetLayout(mDevice, &dlinfo, VKALLOC, &vklayout);
+        mVkLayouts[bitmasks] = vklayout;
+    }
+    layout->setVkLayout(vklayout);
+    return layout;
+}
+
+} // namespace filament::backend

filament/backend/src/vulkan/VulkanDescriptorSetLayoutCache.h

@@ -0,0 +1,63 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_BACKEND_CACHING_VULKANDESCRIPTORSETLAYOUTCACHE_H
+#define TNT_FILAMENT_BACKEND_CACHING_VULKANDESCRIPTORSETLAYOUTCACHE_H
+
+#include "VulkanHandles.h"
+
+#include "vulkan/memory/ResourcePointer.h"
+
+#include <backend/DriverEnums.h>
+#include <backend/Program.h>
+#include <backend/TargetBufferInfo.h>
+
+#include <utils/bitset.h>
+
+#include <bluevk/BlueVK.h>
+#include <tsl/robin_map.h>
+
+#include <memory>
+
+namespace filament::backend {
+
+class VulkanDescriptorSetLayoutCache {
+public:
+    VulkanDescriptorSetLayoutCache(VkDevice device, fvkmemory::ResourceManager* resourceManager);
+    ~VulkanDescriptorSetLayoutCache();
+
+    void terminate() noexcept;
+
+    fvkmemory::resource_ptr<VulkanDescriptorSetLayout> createLayout(
+            Handle<HwDescriptorSetLayout> handle, backend::DescriptorSetLayout&& info);
+
+private:
+    VkDevice mDevice;
+    fvkmemory::ResourceManager* mResourceManager;
+
+    using BitmaskGroup = VulkanDescriptorSetLayout::Bitmask;
+    using BitmaskGroupHashFn = utils::hash::MurmurHashFn<BitmaskGroup>;
+    struct BitmaskGroupEqual {
+        bool operator()(BitmaskGroup const& k1, BitmaskGroup const& k2) const { return k1 == k2; }
+    };
+
+    tsl::robin_map<BitmaskGroup, VkDescriptorSetLayout, BitmaskGroupHashFn, BitmaskGroupEqual>
+            mVkLayouts;
+};
+
+} // namespace filament::backend
+
+#endif// TNT_FILAMENT_BACKEND_CACHING_VULKANDESCRIPTORSETLAYOUTCACHE_H

filament/backend/src/vulkan/VulkanDriver.cpp

@@ -17,7 +17,6 @@
 #include "VulkanDriver.h"
 
 #include "CommandStreamDispatcher.h"
-#include "DataReshaper.h"
 #include "SystraceProfile.h"
 #include "VulkanAsyncHandles.h"
 #include "VulkanBuffer.h"
@@ -35,7 +34,6 @@
 #include <backend/platforms/VulkanPlatform.h>
 
 #include <utils/CString.h>
-#include <utils/FixedCapacityVector.h>
 #include <utils/Panic.h>
 
 #ifndef NDEBUG
@@ -44,8 +42,6 @@
 
 using namespace bluevk;
 
-using utils::FixedCapacityVector;
-
 #if defined(__clang__)
 // Vulkan functions often immediately dereference pointers, so it's fine to pass in a pointer
 // to a stack-allocated variable.
@@ -219,7 +215,8 @@ VulkanDriver::VulkanDriver(VulkanPlatform* platform, VulkanContext const& contex
       mSamplerCache(mPlatform->getDevice()),
       mBlitter(mPlatform->getPhysicalDevice(), &mCommands),
       mReadPixels(mPlatform->getDevice()),
-      mDescriptorSetManager(mPlatform->getDevice(), &mResourceManager),
+      mDescriptorSetLayoutCache(mPlatform->getDevice(), &mResourceManager),
+      mDescriptorSetCache(mPlatform->getDevice(), &mResourceManager),
       mQueryManager(mPlatform->getDevice()),
       mIsSRGBSwapChainSupported(mPlatform->getCustomization().isSRGBSwapChainSupported),
       mStereoscopicType(driverConfig.stereoscopicType) {
@@ -329,7 +326,8 @@ void VulkanDriver::terminate() {
     mPipelineCache.terminate();
     mFramebufferCache.reset();
     mSamplerCache.terminate();
-    mDescriptorSetManager.terminate();
+    mDescriptorSetLayoutCache.terminate();
+    mDescriptorSetCache.terminate();
     mPipelineLayoutCache.terminate();
 
     // Before terminating ResourceManager, we must make sure all of the resource_ptrs have been unset.
@@ -365,7 +363,7 @@ void VulkanDriver::collectGarbage() {
     FVK_SYSTRACE_SCOPE();
     // Command buffers need to be submitted and completed before other resources can be gc'd.
     mCommands.gc();
-    mDescriptorSetManager.clearHistory();
+    mDescriptorSetCache.clearHistory();
     mStagePool.gc();
     mFramebufferCache.gc();
     mPipelineCache.gc();
@@ -408,7 +406,7 @@ void VulkanDriver::updateDescriptorSetBuffer(
     FVK_SYSTRACE_SCOPE();
     auto set = resource_ptr<VulkanDescriptorSet>::cast(&mResourceManager, dsh);
     auto buffer = resource_ptr<VulkanBufferObject>::cast(&mResourceManager, boh);
-    mDescriptorSetManager.updateBuffer(set, binding, buffer, offset, size);
+    mDescriptorSetCache.updateBuffer(set, binding, buffer, offset, size);
 }
 
 void VulkanDriver::updateDescriptorSetTexture(
@@ -421,7 +419,7 @@ void VulkanDriver::updateDescriptorSetTexture(
     auto texture = resource_ptr<VulkanTexture>::cast(&mResourceManager, th);
 
     VkSampler const vksampler = mSamplerCache.getSampler(params);
-    mDescriptorSetManager.updateSampler(set, binding, texture, vksampler);
+    mDescriptorSetCache.updateSampler(set, binding, texture, vksampler);
 }
 
 void VulkanDriver::flush(int) {
@@ -554,13 +552,45 @@ void VulkanDriver::createTextureViewSwizzleR(Handle<HwTexture> th, Handle<HwText
    texture.inc();
 }
 
-void VulkanDriver::createTextureExternalImage2R(Handle<HwTexture> th,
-        backend::SamplerType target,  backend::TextureFormat format,
-        uint32_t width, uint32_t height, backend::TextureUsage usage,
+void VulkanDriver::createTextureExternalImage2R(Handle<HwTexture> th, backend::SamplerType target,
+        backend::TextureFormat format, uint32_t width, uint32_t height, backend::TextureUsage usage,
         Platform::ExternalImageHandleRef externalImage) {
     FVK_SYSTRACE_SCOPE();
+    const auto& metadata = mPlatform->getExternalImageMetadata(externalImage);
+    if (metadata.isProtected) {
+        usage |= backend::TextureUsage::PROTECTED;
+    }
 
-    // FIXME: implement createTextureExternalImage2R
+    VkImageUsageFlags vkUsage = metadata.usage;
+    if (any(usage & TextureUsage::BLIT_SRC)) {
+        vkUsage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
+    }
+
+    if (any(usage & (TextureUsage::BLIT_DST & TextureUsage::UPLOADABLE))) {
+        vkUsage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT;
+    }
+
+    assert_invariant(width == metadata.width);
+    assert_invariant(height == metadata.height);
+    assert_invariant(fvkutils::getVkFormat(format) == metadata.format);
+
+    VkMemoryPropertyFlags const requiredMemoryFlags = any(usage & TextureUsage::UPLOADABLE)
+                                                              ? VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT
+                                                              : VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT;
+    uint32_t const memoryTypeIndex =
+            mContext.selectMemoryType(metadata.memoryTypeBits, requiredMemoryFlags);
+    FILAMENT_CHECK_POSTCONDITION(memoryTypeIndex != VK_MAX_MEMORY_TYPES)
+            << "failed to find a valid memory type for external image memory.";
+
+    const auto& data =
+            mPlatform->createExternalImageData(externalImage, metadata, memoryTypeIndex, vkUsage);
+
+    auto texture = resource_ptr<VulkanTexture>::make(&mResourceManager, th, mPlatform->getDevice(),
+            mAllocator, &mResourceManager, &mCommands, data.first, data.second, metadata.format,
+            metadata.samples, metadata.width, metadata.height, metadata.layerCount, usage,
+            mStagePool);
+
+    texture.inc();
 }
 
 void VulkanDriver::createTextureExternalImageR(Handle<HwTexture> th, backend::SamplerType target,
@@ -568,22 +598,7 @@ void VulkanDriver::createTextureExternalImageR(Handle<HwTexture> th, backend::Sa
         void* externalImage) {
     FVK_SYSTRACE_SCOPE();
 
-    const auto& metadata = mPlatform->getExternalImageMetadata(externalImage);
-    if (metadata.isProtected) {
-        usage |= backend::TextureUsage::PROTECTED;
-    }
-
-    assert_invariant(width == metadata.width);
-    assert_invariant(height == metadata.height);
-    assert_invariant(fvkutils::getVkFormat(format) == metadata.format);
-
-    const auto& data = mPlatform->createExternalImage(externalImage, metadata);
-
-    auto texture = resource_ptr<VulkanTexture>::make(&mResourceManager, th, mPlatform->getDevice(),
-        mAllocator, &mResourceManager, &mCommands, data.first, data.second, metadata.format,
-        1, metadata.width, metadata.height, /*depth=*/1, usage, mStagePool);
-
-    texture.inc();
+    // not supported in this backend
 }
 
 void VulkanDriver::createTextureExternalImagePlaneR(Handle<HwTexture> th,
@@ -761,8 +776,7 @@ void VulkanDriver::createTimerQueryR(Handle<HwTimerQuery> tqh, int) {
 
 void VulkanDriver::createDescriptorSetLayoutR(Handle<HwDescriptorSetLayout> dslh,
         backend::DescriptorSetLayout&& info) {
-    auto layout = resource_ptr<VulkanDescriptorSetLayout>::make(&mResourceManager, dslh, info);
-    mDescriptorSetManager.initVkLayout(layout);
+    auto layout = mDescriptorSetLayoutCache.createLayout(dslh, std::move(info));
     layout.inc();
 }
 
@@ -771,7 +785,7 @@ void VulkanDriver::createDescriptorSetR(Handle<HwDescriptorSet> dsh,
     FVK_SYSTRACE_SCOPE();
     fvkmemory::resource_ptr<VulkanDescriptorSetLayout> layout =
             fvkmemory::resource_ptr<VulkanDescriptorSetLayout>::cast(&mResourceManager, dslh);
-    auto set = mDescriptorSetManager.createSet(dsh, layout);
+    auto set = mDescriptorSetCache.createSet(dsh, layout);
     set.inc();
 }
 
@@ -1043,7 +1057,7 @@ bool VulkanDriver::isDepthStencilBlitSupported(TextureFormat format) {
            formats.end();
 }
 
-bool VulkanDriver::isProtectedTexturesSupported() { return false; }
+bool VulkanDriver::isProtectedTexturesSupported() { return isProtectedContentSupported(); }
 
 bool VulkanDriver::isDepthClampSupported() {
     return mContext.isDepthClampSupported();
@@ -1120,14 +1134,28 @@ math::float2 VulkanDriver::getClipSpaceParams() {
 }
 
 uint8_t VulkanDriver::getMaxDrawBuffers() {
-    return MRT::MIN_SUPPORTED_RENDER_TARGET_COUNT; // TODO: query real value
+    return mContext.getPhysicalDeviceLimits().maxColorAttachments;
 }
 
 size_t VulkanDriver::getMaxUniformBufferSize() {
-    // TODO: return the actual size instead of hardcoded value
-    // TODO: devices that return less than 32768 should be rejected. This represents only 3%
-    //       of android devices.
-    return 32768;
+    return mContext.getPhysicalDeviceLimits().maxUniformBufferRange;
+}
+
+size_t VulkanDriver::getMaxTextureSize(SamplerType type) {
+    switch (type) {
+        case SamplerType::SAMPLER_2D:
+            return mContext.getPhysicalDeviceLimits().maxImageDimension2D;
+        case SamplerType::SAMPLER_3D:
+            return mContext.getPhysicalDeviceLimits().maxImageDimension3D;
+        case SamplerType::SAMPLER_CUBEMAP:
+            return mContext.getPhysicalDeviceLimits().maxImageDimensionCube;
+        default:
+            return mContext.getPhysicalDeviceLimits().maxImageDimension1D;
+    }
+}
+
+size_t VulkanDriver::getMaxArrayTextureLayers() {
+    return mContext.getPhysicalDeviceLimits().maxImageArrayLayers;
 }
 
 void VulkanDriver::setVertexBufferObject(Handle<HwVertexBuffer> vbh, uint32_t index,
@@ -1266,10 +1294,9 @@ void VulkanDriver::generateMipmaps(Handle<HwTexture> th) {
             mBlitter.blit(VK_FILTER_LINEAR, dst, dstOffsets, src, srcOffsets);
         }
 
-        level++;
         srcw = dstw;
         srch = dsth;
-    } while ((srcw > 1 || srch > 1) && level < t->levels);
+    } while ((srcw > 1 || srch > 1) && ++level < t->levels - 1);
 }
 
 void VulkanDriver::compilePrograms(CompilerPriorityQueue priority,
@@ -1467,7 +1494,7 @@ void VulkanDriver::nextSubpass(int) {
 
     if (mCurrentRenderPass.params.subpassMask & 0x1) {
         VulkanAttachment& subpassInput = renderTarget->getColor0();
-        mDescriptorSetManager.updateInputAttachment({}, subpassInput);
+        mDescriptorSetCache.updateInputAttachment({}, subpassInput);
     }
 }
 
@@ -1794,9 +1821,9 @@ void VulkanDriver::bindDescriptorSet(
         backend::DescriptorSetOffsetArray&& offsets) {
     if (dsh) {
         auto set = resource_ptr<VulkanDescriptorSet>::cast(&mResourceManager, dsh);
-        mDescriptorSetManager.bind(setIndex, set, std::move(offsets));
+        mDescriptorSetCache.bind(setIndex, set, std::move(offsets));
     } else {
-        mDescriptorSetManager.unbind(setIndex);
+        mDescriptorSetCache.unbind(setIndex);
     }
 }
 
@@ -1804,7 +1831,7 @@ void VulkanDriver::draw2(uint32_t indexOffset, uint32_t indexCount, uint32_t ins
     FVK_SYSTRACE_SCOPE();
     VkCommandBuffer cmdbuffer = mCurrentRenderPass.commandBuffer->buffer();
 
-    mDescriptorSetManager.commit(mCurrentRenderPass.commandBuffer,
+    mDescriptorSetCache.commit(mCurrentRenderPass.commandBuffer,
             mBoundPipeline.pipelineLayout,
             mBoundPipeline.descriptorSetMask);
 

filament/backend/src/vulkan/VulkanDriver.h

@@ -27,8 +27,9 @@
 #include "VulkanSamplerCache.h"
 #include "VulkanStagePool.h"
 #include "VulkanQueryManager.h"
-#include "vulkan/caching/VulkanDescriptorSetManager.h"
-#include "vulkan/caching/VulkanPipelineLayoutCache.h"
+#include "vulkan/VulkanDescriptorSetCache.h"
+#include "vulkan/VulkanDescriptorSetLayoutCache.h"
+#include "vulkan/VulkanPipelineLayoutCache.h"
 #include "vulkan/memory/ResourceManager.h"
 #include "vulkan/memory/ResourcePointer.h"
 #include "vulkan/utils/Definitions.h"
@@ -137,7 +138,8 @@ private:
     VulkanSamplerCache mSamplerCache;
     VulkanBlitter mBlitter;
     VulkanReadPixels mReadPixels;
-    VulkanDescriptorSetManager mDescriptorSetManager;
+    VulkanDescriptorSetLayoutCache mDescriptorSetLayoutCache;
+    VulkanDescriptorSetCache mDescriptorSetCache;
     VulkanQueryManager mQueryManager;
 
     // This is necessary for us to write to push constants after binding a pipeline.

filament/backend/src/vulkan/VulkanPipelineCache.cpp

@@ -15,15 +15,12 @@
  */
 
 #include "VulkanPipelineCache.h"
-#include "VulkanMemory.h"
-#include "caching/VulkanDescriptorSetManager.h"
 
 #include <utils/Log.h>
 #include <utils/Panic.h>
 
 #include "VulkanConstants.h"
 #include "VulkanHandles.h"
-#include "VulkanTexture.h"
 #include "vulkan/utils/Conversion.h"
 
 #if defined(__clang__)

filament/backend/src/vulkan/VulkanPipelineLayoutCache.h

@@ -17,7 +17,8 @@
 #ifndef TNT_FILAMENT_BACKEND_VULKANPIPELINELAYOUTCACHE_H
 #define TNT_FILAMENT_BACKEND_VULKANPIPELINELAYOUTCACHE_H
 
-#include <vulkan/VulkanHandles.h>
+#include "VulkanHandles.h"
+
 #include <bluevk/BlueVK.h>
 
 #include <utils/Hash.h>

filament/backend/src/vulkan/VulkanSamplerCache.cpp

@@ -23,79 +23,6 @@ using namespace bluevk;
 
 namespace filament::backend {
 
-constexpr inline VkSamplerAddressMode getWrapMode(SamplerWrapMode mode) noexcept {
-    switch (mode) {
-        case SamplerWrapMode::REPEAT:
-            return VK_SAMPLER_ADDRESS_MODE_REPEAT;
-        case SamplerWrapMode::CLAMP_TO_EDGE:
-            return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE;
-        case SamplerWrapMode::MIRRORED_REPEAT:
-            return VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT;
-    }
-}
-
-constexpr inline VkFilter getFilter(SamplerMinFilter filter) noexcept {
-    switch (filter) {
-        case SamplerMinFilter::NEAREST:
-            return VK_FILTER_NEAREST;
-        case SamplerMinFilter::LINEAR:
-            return VK_FILTER_LINEAR;
-        case SamplerMinFilter::NEAREST_MIPMAP_NEAREST:
-            return VK_FILTER_NEAREST;
-        case SamplerMinFilter::LINEAR_MIPMAP_NEAREST:
-            return VK_FILTER_LINEAR;
-        case SamplerMinFilter::NEAREST_MIPMAP_LINEAR:
-            return VK_FILTER_NEAREST;
-        case SamplerMinFilter::LINEAR_MIPMAP_LINEAR:
-            return VK_FILTER_LINEAR;
-    }
-}
-
-constexpr inline VkFilter getFilter(SamplerMagFilter filter) noexcept {
-    switch (filter) {
-        case SamplerMagFilter::NEAREST:
-            return VK_FILTER_NEAREST;
-        case SamplerMagFilter::LINEAR:
-            return VK_FILTER_LINEAR;
-    }
-}
-
-constexpr inline VkSamplerMipmapMode getMipmapMode(SamplerMinFilter filter) noexcept {
-    switch (filter) {
-        case SamplerMinFilter::NEAREST:
-            return VK_SAMPLER_MIPMAP_MODE_NEAREST;
-        case SamplerMinFilter::LINEAR:
-            return VK_SAMPLER_MIPMAP_MODE_NEAREST;
-        case SamplerMinFilter::NEAREST_MIPMAP_NEAREST:
-            return VK_SAMPLER_MIPMAP_MODE_NEAREST;
-        case SamplerMinFilter::LINEAR_MIPMAP_NEAREST:
-            return VK_SAMPLER_MIPMAP_MODE_NEAREST;
-        case SamplerMinFilter::NEAREST_MIPMAP_LINEAR:
-            return VK_SAMPLER_MIPMAP_MODE_LINEAR;
-        case SamplerMinFilter::LINEAR_MIPMAP_LINEAR:
-            return VK_SAMPLER_MIPMAP_MODE_LINEAR;
-    }
-}
-
-constexpr inline float getMaxLod(SamplerMinFilter filter) noexcept {
-    switch (filter) {
-        case SamplerMinFilter::NEAREST:
-        case SamplerMinFilter::LINEAR:
-            // The Vulkan spec recommends a max LOD of 0.25 to "disable" mipmapping.
-            // See "Mapping of OpenGL to Vulkan filter modes" in the VK Spec.
-            return 0.25f;
-        case SamplerMinFilter::NEAREST_MIPMAP_NEAREST:
-        case SamplerMinFilter::LINEAR_MIPMAP_NEAREST:
-        case SamplerMinFilter::NEAREST_MIPMAP_LINEAR:
-        case SamplerMinFilter::LINEAR_MIPMAP_LINEAR:
-            return VK_LOD_CLAMP_NONE;
-    }
-}
-
-constexpr inline VkBool32 getCompareEnable(SamplerCompareMode mode) noexcept {
-    return mode == SamplerCompareMode::NONE ? VK_FALSE : VK_TRUE;
-}
-
 VulkanSamplerCache::VulkanSamplerCache(VkDevice device)
     : mDevice(device) {}
 
@@ -106,18 +33,18 @@ VkSampler VulkanSamplerCache::getSampler(SamplerParams params) noexcept {
     }
     VkSamplerCreateInfo samplerInfo {
         .sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO,
-        .magFilter = getFilter(params.filterMag),
-        .minFilter = getFilter(params.filterMin),
-        .mipmapMode = getMipmapMode(params.filterMin),
-        .addressModeU = getWrapMode(params.wrapS),
-        .addressModeV = getWrapMode(params.wrapT),
-        .addressModeW = getWrapMode(params.wrapR),
-        .anisotropyEnable = params.anisotropyLog2 == 0 ? 0u : 1u,
+        .magFilter = fvkutils::getFilter(params.filterMag),
+        .minFilter = fvkutils::getFilter(params.filterMin),
+        .mipmapMode = fvkutils::getMipmapMode(params.filterMin),
+        .addressModeU = fvkutils::getWrapMode(params.wrapS),
+        .addressModeV = fvkutils::getWrapMode(params.wrapT),
+        .addressModeW = fvkutils::getWrapMode(params.wrapR),
+        .anisotropyEnable = params.anisotropyLog2 == 0 ? VK_FALSE : VK_TRUE,
         .maxAnisotropy = (float)(1u << params.anisotropyLog2),
-        .compareEnable = getCompareEnable(params.compareMode),
+        .compareEnable = fvkutils::getCompareEnable(params.compareMode),
         .compareOp = fvkutils::getCompareOp(params.compareFunc),
         .minLod = 0.0f,
-        .maxLod = getMaxLod(params.filterMin),
+        .maxLod = fvkutils::getMaxLod(params.filterMin),
         .borderColor = VK_BORDER_COLOR_INT_OPAQUE_BLACK,
         .unnormalizedCoordinates = VK_FALSE
     };

filament/backend/src/vulkan/VulkanTexture.cpp

@@ -55,63 +55,61 @@ VkComponentMapping composeSwizzle(VkComponentMapping const& prev, VkComponentMap
         VK_COMPONENT_SWIZZLE_A,
     };
 
-    auto const compose = [](VkComponentSwizzle out, VkComponentMapping const& prev,
-                                 uint8_t channelIndex) {
-        // We need to first change all identities to its equivalent channel.
-        if (out == VK_COMPONENT_SWIZZLE_IDENTITY) {
-            out = IDENTITY[channelIndex];
+    auto const compose = [](VkComponentMapping const& prev,
+                                 VkComponentMapping const& next) -> VkComponentMapping {
+        VkComponentSwizzle vals[4] = { next.r, next.g, next.b, next.a };
+        for (auto& out: vals) {
+            switch (out) {
+                case VK_COMPONENT_SWIZZLE_R:
+                    out = prev.r;
+                    break;
+                case VK_COMPONENT_SWIZZLE_G:
+                    out = prev.g;
+                    break;
+                case VK_COMPONENT_SWIZZLE_B:
+                    out = prev.b;
+                    break;
+                case VK_COMPONENT_SWIZZLE_A:
+                    out = prev.a;
+                    break;
+                // Do not modify the value
+                case VK_COMPONENT_SWIZZLE_IDENTITY:
+                case VK_COMPONENT_SWIZZLE_ZERO:
+                case VK_COMPONENT_SWIZZLE_ONE:
+                // Below is not exposed in Vulkan's API, but needs to be there for compilation.
+                case VK_COMPONENT_SWIZZLE_MAX_ENUM:
+                    break;
+            }
         }
-        switch (out) {
-            case VK_COMPONENT_SWIZZLE_R:
-                out = prev.r;
-                break;
-            case VK_COMPONENT_SWIZZLE_G:
-                out = prev.g;
-                break;
-            case VK_COMPONENT_SWIZZLE_B:
-                out = prev.b;
-                break;
-            case VK_COMPONENT_SWIZZLE_A:
-                out = prev.a;
-                break;
-            case VK_COMPONENT_SWIZZLE_IDENTITY:
-            case VK_COMPONENT_SWIZZLE_ZERO:
-            case VK_COMPONENT_SWIZZLE_ONE:
-                return out;
-            // Below is not exposed in Vulkan's API, but needs to be there for compilation.
-            case VK_COMPONENT_SWIZZLE_MAX_ENUM:
-                break;
-        }
-        // If the result correctly corresponds to the identity, just return identity.
-        if (IDENTITY[channelIndex] == out) {
-            return VK_COMPONENT_SWIZZLE_IDENTITY;
-        }
-        return out;
+        return { vals[0], vals[1], vals[2], vals[3] };
     };
 
-    auto const identityToChannel = [](VkComponentSwizzle val, uint8_t channelIndex) {
-        if (val != VK_COMPONENT_SWIZZLE_IDENTITY) {
-            return val;
+    auto const identityToChannel = [](VkComponentMapping const& mapping) -> VkComponentMapping {
+        VkComponentSwizzle vals[4] = { mapping.r, mapping.g, mapping.b, mapping.a };
+        for (uint8_t i = 0; i < 4; i++) {
+            if (vals[i] != VK_COMPONENT_SWIZZLE_IDENTITY) {
+                continue;
+            }
+            vals[i] = IDENTITY[i];
         }
-        return IDENTITY[channelIndex];
+        return { vals[0], vals[1], vals[2], vals[3] };
+    };
+    auto const channelToIdentity = [](VkComponentMapping const& mapping) -> VkComponentMapping {
+        VkComponentSwizzle vals[4] = { mapping.r, mapping.g, mapping.b, mapping.a };
+        for (uint8_t i = 0; i < 4; i++) {
+            if (IDENTITY[i] != vals[i]) {
+                continue;
+            }
+            vals[i] = VK_COMPONENT_SWIZZLE_IDENTITY;
+        }
+        return { vals[0], vals[1], vals[2], vals[3] };
     };
 
     // We make sure all all identities are mapped into respective channels so that actual channel
     // mapping will be passed onto the output.
-    VkComponentMapping const prevExplicit = {
-            identityToChannel(prev.r, 0),
-            identityToChannel(prev.g, 1),
-            identityToChannel(prev.b, 2),
-            identityToChannel(prev.a, 3),
-    };
-
-    // Note that the channel index corresponds to the VkComponentMapping struct layout.
-    return {
-        compose(next.r, prevExplicit, 0),
-        compose(next.g, prevExplicit, 1),
-        compose(next.b, prevExplicit, 2),
-        compose(next.a, prevExplicit, 3),
-    };
+    VkComponentMapping const prevExplicit = identityToChannel(prev);
+    VkComponentMapping const nextExplicit = identityToChannel(next);
+    return channelToIdentity(compose(prevExplicit, nextExplicit));
 }
 
 inline VulkanLayout getDefaultLayoutImpl(TextureUsage usage) {
@@ -153,7 +151,6 @@ uint8_t getLayerCountFromDepth(uint32_t const depth) {
     return getLayerCount(getSamplerTypeFromDepth(depth), depth);
 }
 
-
 } // anonymous namespace
 
 VulkanTextureState::VulkanTextureState(VkDevice device, VmaAllocator allocator,
@@ -249,12 +246,19 @@ VulkanTexture::VulkanTexture(VkDevice device, VkPhysicalDevice physicalDevice,
 
     // Determine if we can use the transient usage flag combined with lazily allocated memory.
     const bool useTransientAttachment =
-        // Lazily allocated memory is available.
-        context.isLazilyAllocatedMemorySupported() &&
-        // Usage consists of attachment flags only.
-        none(tusage & ~TextureUsage::ALL_ATTACHMENTS) &&
-        // Usage contains at least one attachment flag.
-        any(tusage & TextureUsage::ALL_ATTACHMENTS);
+            // Lazily allocated memory is available.
+            context.isLazilyAllocatedMemorySupported() &&
+            // Usage consists of attachment flags only.
+            none(tusage & ~TextureUsage::ALL_ATTACHMENTS) &&
+            // Usage contains at least one attachment flag.
+            any(tusage & TextureUsage::ALL_ATTACHMENTS) &&
+            // Depth resolve cannot use transient attachment because it uses a custom shader.
+            // TODO: see VulkanDriver::isDepthStencilResolveSupported() to know when to remove this
+            // restriction.
+            // Note that the custom shader does not resolve stencil. We do need to move to vk 1.2
+            // and above to be able to support stencil resolve (along with depth).
+            !(any(usage & TextureUsage::DEPTH_ATTACHMENT) && samples > 1);
+
     mState->mIsTransientAttachment = useTransientAttachment;
 
     const VkImageUsageFlags transientFlag =

filament/backend/src/vulkan/platform/VulkanPlatform.cpp

@@ -736,10 +736,6 @@ Driver* VulkanPlatform::createDriver(void* sharedContext,
     chainStruct(&context.mPhysicalDeviceFeatures, &queryProtectedMemoryFeatures);
     chainStruct(&context.mPhysicalDeviceProperties, &protectedMemoryProperties);
 
-    // We know we need to allocate the protected version of the VK objects
-    context.mProtectedMemorySupported =
-            static_cast<bool>(queryProtectedMemoryFeatures.protectedMemory);
-
     // Initialize the following fields: physicalDeviceProperties, memoryProperties,
     // physicalDeviceFeatures, graphicsQueueFamilyIndex.
     vkGetPhysicalDeviceProperties2(mImpl->mPhysicalDevice, &context.mPhysicalDeviceProperties);
@@ -759,6 +755,10 @@ Driver* VulkanPlatform::createDriver(void* sharedContext,
         mImpl->mGraphicsQueueIndex = 0;
     }
 
+    // We know we need to allocate the protected version of the VK objects
+    context.mProtectedMemorySupported =
+            static_cast<bool>(queryProtectedMemoryFeatures.protectedMemory);
+
     if (context.mProtectedMemorySupported) {
         mImpl->mProtectedGraphicsQueueFamilyIndex = identifyGraphicsQueueFamilyIndex(
                 mImpl->mPhysicalDevice, (VK_QUEUE_GRAPHICS_BIT | VK_QUEUE_PROTECTED_BIT));
@@ -978,15 +978,37 @@ VkQueue VulkanPlatform::getProtectedGraphicsQueue() const noexcept {
 }
 
 VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadata(
-        void* externalImage) {
+        ExternalImageHandleRef externalImage) {
     return getExternalImageMetadataImpl(externalImage, mImpl->mDevice);
 }
 
-VulkanPlatform::ImageData VulkanPlatform::createExternalImage(void* externalImage,
-            const ExternalImageMetadata& metadata) {
-    return createExternalImageImpl(externalImage, mImpl->mDevice, nullptr, metadata);
+VulkanPlatform::ImageData VulkanPlatform::createExternalImageData(
+        ExternalImageHandleRef externalImage, const ExternalImageMetadata& metadata,
+        uint32_t memoryTypeIndex, VkImageUsageFlags usage) {
+    return createExternalImageDataImpl(externalImage, mImpl->mDevice, metadata, memoryTypeIndex,
+            usage);
 }
 
+VkSampler VulkanPlatform::createExternalSampler(SamplerYcbcrConversion chroma,
+        SamplerParams sampler, uint32_t internalFormat) {
+    return createExternalSamplerImpl(mImpl->mDevice, chroma, sampler, internalFormat);
+}
+
+VkImageView VulkanPlatform::createExternalImageView(SamplerYcbcrConversion chroma,
+        uint32_t internalFormat, VkImage image, VkImageSubresourceRange range,
+        VkImageViewType viewType, VkComponentMapping swizzle) {
+    return createExternalImageViewImpl(mImpl->mDevice, chroma, internalFormat, image, range,
+            viewType, swizzle);
+}
+
+ExtensionSet VulkanPlatform::getSwapchainInstanceExtensions() const {
+    return getSwapchainInstanceExtensionsImpl();
+}
+
+VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHR(void* nativeWindow,
+        VkInstance instance, uint64_t flags) const noexcept {
+    return createVkSurfaceKHRImpl(nativeWindow, instance, flags);
+}
 #undef SWAPCHAIN_RET_FUNC
 
 }// namespace filament::backend

filament/backend/src/vulkan/platform/VulkanPlatformAndroid.cpp

@@ -13,28 +13,60 @@
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-#include <backend/platforms/VulkanPlatform.h>
+#include <backend/platforms/VulkanPlatformAndroid.h>
+
+#include <backend/DriverEnums.h>
+#include <private/backend/BackendUtilsAndroid.h>
 
 #include "vulkan/VulkanConstants.h"
 
 #include <utils/Panic.h>
+#include "vulkan/utils/Image.h"
+#include "vulkan/utils/Conversion.h"
 
 #include <bluevk/BlueVK.h>
 
 #include <android/hardware_buffer.h>
 #include <android/native_window.h>
 
+#include <utility>
+
 using namespace bluevk;
 
 namespace filament::backend {
 
 namespace {
-void getVKFormatAndUsage(const AHardwareBuffer_Desc& desc, VkFormat& format,
-        VkImageUsageFlags& usage, bool& isProtected) {
+
+VkFormat transformVkFormat(VkFormat format, bool sRGB) {
+    if (!sRGB) {
+        return format;
+    }
+
+    switch (format) {
+        case VK_FORMAT_R8G8B8A8_UNORM:
+            format = VK_FORMAT_R8G8B8A8_SRGB;
+            break;
+        case VK_FORMAT_R8G8B8_UNORM:
+            format = VK_FORMAT_R8G8B8_SRGB;
+            break;
+        default:
+            break;
+    }
+
+    return format;
+}
+
+bool isProtectedFromUsage(uint64_t usage) {
+    return (usage & AHARDWAREBUFFER_USAGE_PROTECTED_CONTENT) ? true : false;
+}
+
+std::pair<VkFormat, VkImageUsageFlags> getVKFormatAndUsage(const AHardwareBuffer_Desc& desc,
+        bool sRGB) {
+    VkFormat format = VK_FORMAT_UNDEFINED;
+    VkImageUsageFlags usage = 0;
     // Refer to "11.2.17. External Memory Handle Types" in the spec, and
     // Tables 13/14 for how the following derivation works.
     bool isDepthFormat = false;
-    isProtected = false;
     switch (desc.format) {
         case AHARDWAREBUFFER_FORMAT_R8G8B8A8_UNORM:
             format = VK_FORMAT_R8G8B8A8_UNORM;
@@ -82,6 +114,8 @@ void getVKFormatAndUsage(const AHardwareBuffer_Desc& desc, VkFormat& format,
             format = VK_FORMAT_UNDEFINED;
     }
 
+    format = transformVkFormat(format, sRGB);
+
     // The following only concern usage flags derived from Table 14.
     usage = 0;
     if (desc.usage & AHARDWAREBUFFER_USAGE_GPU_SAMPLED_IMAGE) {
@@ -98,16 +132,14 @@ void getVKFormatAndUsage(const AHardwareBuffer_Desc& desc, VkFormat& format,
     if (desc.usage & AHARDWAREBUFFER_USAGE_GPU_DATA_BUFFER) {
         usage = VK_IMAGE_USAGE_STORAGE_BIT;
     }
-    if (desc.usage & AHARDWAREBUFFER_USAGE_PROTECTED_CONTENT) {
-        isProtected = true;
-    }
+
+    return { format, usage };
 }
 
-VulkanPlatform::ImageData allocateExternalImage(void* externalBuffer,
-        VkDevice device, const VkAllocationCallbacks* allocator,
-        VulkanPlatform::ExternalImageMetadata const& metadata) {
+VulkanPlatform::ImageData allocateExternalImage(AHardwareBuffer* buffer, VkDevice device,
+        VulkanPlatform::ExternalImageMetadata const& metadata, uint32_t memoryTypeIndex,
+        VkImageUsageFlags usage) {
     VulkanPlatform::ImageData data;
-    AHardwareBuffer* buffer = static_cast<AHardwareBuffer*>(externalBuffer);
 
     // if external format we need to specifiy it in the allocation
     const bool useExternalFormat = metadata.format == VK_FORMAT_UNDEFINED;
@@ -115,8 +147,8 @@ VulkanPlatform::ImageData allocateExternalImage(void* externalBuffer,
     const VkExternalFormatANDROID externalFormat = {
         .sType = VK_STRUCTURE_TYPE_EXTERNAL_FORMAT_ANDROID,
         .pNext = nullptr,
-        .externalFormat = metadata
-                .externalFormat,// pass down the format (external means we don't have it VK defined)
+        // pass down the format (external means we don't have it VK defined)
+        .externalFormat = metadata.externalFormat,
     };
     const VkExternalMemoryImageCreateInfo externalCreateInfo = {
         .sType = VK_STRUCTURE_TYPE_EXTERNAL_MEMORY_IMAGE_CREATE_INFO,
@@ -124,7 +156,7 @@ VulkanPlatform::ImageData allocateExternalImage(void* externalBuffer,
         .handleTypes = VK_EXTERNAL_MEMORY_HANDLE_TYPE_ANDROID_HARDWARE_BUFFER_BIT_ANDROID,
     };
 
-    VkImageCreateInfo imageInfo{.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO};
+    VkImageCreateInfo imageInfo{ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO };
     imageInfo.pNext = &externalCreateInfo;
     imageInfo.format = metadata.format;
     imageInfo.imageType = VK_IMAGE_TYPE_2D;
@@ -135,12 +167,10 @@ VulkanPlatform::ImageData allocateExternalImage(void* externalBuffer,
     };
     imageInfo.mipLevels = 1;
     imageInfo.arrayLayers = metadata.layers;
-    imageInfo.usage = metadata.usage;
-    // In the unprotected case add R/W capabilities
-    if (metadata.isProtected == false)
-        imageInfo.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
+    imageInfo.samples = metadata.samples;
+    imageInfo.usage = usage;
 
-    VkResult result = vkCreateImage(device, &imageInfo, allocator, &data.first);
+    VkResult result = vkCreateImage(device, &imageInfo, VKALLOC, &data.first);
     FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
             << "vkCreateImage failed with error=" << static_cast<int32_t>(result);
 
@@ -156,11 +186,13 @@ VulkanPlatform::ImageData allocateExternalImage(void* externalBuffer,
         .image = data.first,
         .buffer = VK_NULL_HANDLE,
     };
-    VkMemoryAllocateInfo allocInfo = {.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
+    VkMemoryAllocateInfo allocInfo = {
+        .sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
         .pNext = &memoryDedicatedAllocateInfo,
         .allocationSize = metadata.allocationSize,
-        .memoryTypeIndex = metadata.memoryTypeBits};
-    result = vkAllocateMemory(device, &allocInfo, allocator, &data.second);
+        .memoryTypeIndex = memoryTypeIndex,
+    };
+    result = vkAllocateMemory(device, &allocInfo, VKALLOC, &data.second);
     FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
             << "vkAllocateMemory failed with error=" << static_cast<int32_t>(result);
 
@@ -169,23 +201,60 @@ VulkanPlatform::ImageData allocateExternalImage(void* externalBuffer,
 
 }// namespace
 
-VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadataImpl(void* externalImage,
-        VkDevice device) {
+VulkanPlatformAndroid::ExternalImageVulkanAndroid::~ExternalImageVulkanAndroid() = default;
+
+Platform::ExternalImageHandle VulkanPlatformAndroid::createExternalImage(
+        AHardwareBuffer const* buffer, bool sRGB) noexcept {
+    if (__builtin_available(android 26, *)) {
+        AHardwareBuffer_Desc hardwareBufferDescription = {};
+        AHardwareBuffer_describe(buffer, &hardwareBufferDescription);
+
+        auto* const p = new (std::nothrow) ExternalImageVulkanAndroid;
+        p->aHardwareBuffer = const_cast<AHardwareBuffer*>(buffer);
+        p->sRGB = sRGB;
+        p->height = hardwareBufferDescription.height;
+        p->width = hardwareBufferDescription.width;
+        TextureFormat textureFormat = mapToFilamentFormat(hardwareBufferDescription.format, sRGB);
+        p->format = textureFormat;
+        p->usage = mapToFilamentUsage(hardwareBufferDescription.usage, textureFormat);
+        return Platform::ExternalImageHandle{ p };
+    }
+
+    return Platform::ExternalImageHandle{};
+}
+
+VulkanPlatformAndroid::ExternalImageDescAndroid VulkanPlatformAndroid::getExternalImageDesc(
+        ExternalImageHandleRef externalImage) const noexcept {
+    auto const* fvkExternalImage =
+            static_cast<ExternalImageVulkanAndroid const*>(externalImage.get());
+
+    return {
+        .width = fvkExternalImage->width,
+        .height = fvkExternalImage->height,
+        .format = fvkExternalImage->format,
+        .usage = fvkExternalImage->usage,
+    };
+}
+
+VulkanPlatform::ExternalImageMetadata VulkanPlatformAndroid::getExternalImageMetadata(
+        ExternalImageHandleRef externalImage) {
+    auto const* fvkExternalImage =
+            static_cast<ExternalImageVulkanAndroid const*>(externalImage.get());
+
     ExternalImageMetadata metadata;
-    AHardwareBuffer* buffer = static_cast<AHardwareBuffer*>(externalImage);
+    AHardwareBuffer* buffer = fvkExternalImage->aHardwareBuffer;
     if (__builtin_available(android 26, *)) {
         AHardwareBuffer_Desc bufferDesc;
         AHardwareBuffer_describe(buffer, &bufferDesc);
         metadata.width = bufferDesc.width;
         metadata.height = bufferDesc.height;
         metadata.layers = bufferDesc.layers;
+        metadata.isProtected = isProtectedFromUsage(bufferDesc.usage);
+        std::tie(metadata.format, metadata.usage) =
+                getVKFormatAndUsage(bufferDesc, fvkExternalImage->sRGB);
+    }
 
-        getVKFormatAndUsage(bufferDesc, metadata.format, metadata.usage, metadata.isProtected);
-    }
-    // In the unprotected case add R/W capabilities
-    if (metadata.isProtected == false) {
-        metadata.usage |= VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_TRANSFER_SRC_BIT;
-    }
+    metadata.samples = VK_SAMPLE_COUNT_1_BIT;
 
     VkAndroidHardwareBufferFormatPropertiesANDROID formatInfo = {
         .sType = VK_STRUCTURE_TYPE_ANDROID_HARDWARE_BUFFER_FORMAT_PROPERTIES_ANDROID,
@@ -195,36 +264,155 @@ VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadataIm
         .sType = VK_STRUCTURE_TYPE_ANDROID_HARDWARE_BUFFER_PROPERTIES_ANDROID,
         .pNext = &formatInfo,
     };
-    VkResult result = vkGetAndroidHardwareBufferPropertiesANDROID(device, buffer, &properties);
+    VkResult result = vkGetAndroidHardwareBufferPropertiesANDROID(getDevice(), buffer, &properties);
     FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
             << "vkGetAndroidHardwareBufferProperties failed with error="
             << static_cast<int32_t>(result);
-    FILAMENT_CHECK_POSTCONDITION(metadata.format == formatInfo.format)
-            << "mismatched image format for external image (AHB)";
+
+    VkFormat bufferPropertiesFormat = transformVkFormat(formatInfo.format, fvkExternalImage->sRGB);
+    FILAMENT_CHECK_POSTCONDITION(metadata.format == bufferPropertiesFormat)
+            << "mismatched image format( " << metadata.format << ") and queried format("
+            << bufferPropertiesFormat << ") for external image (AHB)";
     metadata.externalFormat = formatInfo.externalFormat;
     metadata.allocationSize = properties.allocationSize;
     metadata.memoryTypeBits = properties.memoryTypeBits;
     return metadata;
 }
 
-VulkanPlatform::ImageData VulkanPlatform::createExternalImageImpl(void* externalImage,
-        VkDevice device, const VkAllocationCallbacks* allocator,
-        const ExternalImageMetadata& metadata) {
-    ImageData data = allocateExternalImage(externalImage, device, allocator, metadata);
-    VkResult result = vkBindImageMemory(device, data.first, data.second, 0);
+VulkanPlatformAndroid::ImageData VulkanPlatformAndroid::createExternalImageData(
+        ExternalImageHandleRef externalImage, const ExternalImageMetadata& metadata,
+        uint32_t memoryTypeIndex, VkImageUsageFlags usage) {
+    auto const* fvkExternalImage =
+            static_cast<ExternalImageVulkanAndroid const*>(externalImage.get());
+    ImageData data = allocateExternalImage(fvkExternalImage->aHardwareBuffer, getDevice(), metadata,
+            memoryTypeIndex, usage);
+    VkResult result = vkBindImageMemory(getDevice(), data.first, data.second, 0);
     FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
-            << "vkBindImageMemory error=" << static_cast<int32_t>(result);
+        << "vkBindImageMemory error=" << static_cast<int32_t>(result);
     return data;
 }
 
-VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensions() {
+VkImageView VulkanPlatform::createExternalImageViewImpl(VkDevice device, SamplerYcbcrConversion chroma,
+            uint32_t internalFormat, VkImage image, VkImageSubresourceRange range,
+            VkImageViewType viewType, VkComponentMapping swizzle){
+        VkExternalFormatANDROID externalFormat = {
+        .sType = VK_STRUCTURE_TYPE_EXTERNAL_FORMAT_ANDROID,
+        .pNext = nullptr,
+        .externalFormat = internalFormat,
+    };
+
+    TextureSwizzle const swizzleArray[] = {chroma.r, chroma.g, chroma.b, chroma.a};
+    VkSamplerYcbcrConversionCreateInfo conversionInfo = {
+        .sType = VK_STRUCTURE_TYPE_SAMPLER_YCBCR_CONVERSION_CREATE_INFO,
+        .pNext = &externalFormat,
+        .format = VK_FORMAT_UNDEFINED,
+        .ycbcrModel = fvkutils::getYcbcrModelConversion(chroma.ycbcrModel),
+        .ycbcrRange = fvkutils::getYcbcrRange(chroma.ycbcrRange),
+        .components = fvkutils::getSwizzleMap(swizzleArray),
+        .xChromaOffset = fvkutils::getChromaLocation(chroma.xChromaOffset),
+        .yChromaOffset = fvkutils::getChromaLocation(chroma.yChromaOffset),
+        .chromaFilter = fvkutils::getFilter(chroma.chromaFilter),
+    };
+    VkSamplerYcbcrConversion conversion = VK_NULL_HANDLE;
+    VkResult result = vkCreateSamplerYcbcrConversion(device, &conversionInfo,
+                                                     nullptr, &conversion);
+    FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
+        << "Unable to create Ycbcr Conversion."
+        << " error=" << static_cast<int32_t>(result);
+
+    VkSamplerYcbcrConversionInfo samplerYcbcrConversionInfo = {
+        .sType = VK_STRUCTURE_TYPE_SAMPLER_YCBCR_CONVERSION_INFO,
+        .pNext = nullptr,
+        .conversion = conversion,
+    };
+
+    VkImageViewCreateInfo viewInfo = {
+        .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+        .pNext = &samplerYcbcrConversionInfo,
+        .flags = 0,
+        .image = image,
+        .viewType = viewType,
+        .format = VK_FORMAT_UNDEFINED,
+        .components = swizzle,
+        .subresourceRange = range,
+    };
+    VkImageView imageView;
+    result = vkCreateImageView(device, &viewInfo, VKALLOC, &imageView);
+    FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
+        << "Unable to create VkImageView."
+        << " error=" << static_cast<int32_t>(result);
+
+    return imageView;
+}
+
+VkSampler VulkanPlatform::createExternalSamplerImpl(
+        VkDevice device, SamplerYcbcrConversion chroma, SamplerParams params,
+        uint32_t internalFormat) {
+    VkExternalFormatANDROID externalFormat = {
+        .sType = VK_STRUCTURE_TYPE_EXTERNAL_FORMAT_ANDROID,
+        .pNext = nullptr,
+        .externalFormat = internalFormat,
+    };
+
+    TextureSwizzle const swizzleArray[] = {chroma.r, chroma.g, chroma.b, chroma.a};
+    VkSamplerYcbcrConversionCreateInfo conversionInfo = {
+        .sType = VK_STRUCTURE_TYPE_SAMPLER_YCBCR_CONVERSION_CREATE_INFO,
+        .pNext = &externalFormat,
+        .format = VK_FORMAT_UNDEFINED,
+        .ycbcrModel = fvkutils::getYcbcrModelConversion(chroma.ycbcrModel),
+        .ycbcrRange = fvkutils::getYcbcrRange(chroma.ycbcrRange),
+        .components = fvkutils::getSwizzleMap(swizzleArray),
+        .xChromaOffset = fvkutils::getChromaLocation(chroma.xChromaOffset),
+        .yChromaOffset = fvkutils::getChromaLocation(chroma.yChromaOffset),
+        .chromaFilter = fvkutils::getFilter(chroma.chromaFilter),
+    };
+    VkSamplerYcbcrConversion conversion = VK_NULL_HANDLE;
+    VkResult result = vkCreateSamplerYcbcrConversion(device, &conversionInfo,
+                                                     nullptr, &conversion);
+    FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
+        << "Unable to create Ycbcr Conversion."
+        << " error=" << static_cast<int32_t>(result);
+
+    VkSamplerYcbcrConversionInfo samplerYcbcrConversionInfo = {
+        .sType = VK_STRUCTURE_TYPE_SAMPLER_YCBCR_CONVERSION_INFO,
+        .pNext = nullptr,
+        .conversion = conversion,
+    };
+
+    VkSamplerCreateInfo samplerInfo = {
+        .sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO,
+        .pNext = &samplerYcbcrConversionInfo,
+        .magFilter = fvkutils::getFilter(params.filterMag),
+        .minFilter = fvkutils::getFilter(params.filterMin),
+        .mipmapMode = fvkutils::getMipmapMode(params.filterMin),
+        .addressModeU = fvkutils::getWrapMode(params.wrapS),
+        .addressModeV = fvkutils::getWrapMode(params.wrapT),
+        .addressModeW = fvkutils::getWrapMode(params.wrapR),
+        .anisotropyEnable = params.anisotropyLog2 == 0 ? VK_FALSE : VK_TRUE,
+        .maxAnisotropy = (float)(1u << params.anisotropyLog2),
+        .compareEnable = fvkutils::getCompareEnable(params.compareMode),
+        .compareOp = fvkutils::getCompareOp(params.compareFunc),
+        .minLod = 0.0f,
+        .maxLod = fvkutils::getMaxLod(params.filterMin),
+        .borderColor = VK_BORDER_COLOR_INT_OPAQUE_BLACK,
+        .unnormalizedCoordinates = VK_FALSE,
+    };
+    VkSampler sampler;
+    result = vkCreateSampler(device, &samplerInfo, VKALLOC, &sampler);
+    FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
+        << "Unable to create sampler."
+        << " error=" << static_cast<int32_t>(result);
+    return sampler;
+}
+
+VulkanPlatform::ExtensionSet VulkanPlatformAndroid::getSwapchainInstanceExtensions() const {
     return {
         VK_KHR_ANDROID_SURFACE_EXTENSION_NAME,
     };
 }
 
-VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHR(void* nativeWindow,
-    VkInstance instance, uint64_t flags) noexcept {
+VulkanPlatform::SurfaceBundle VulkanPlatformAndroid::createVkSurfaceKHR(void* nativeWindow,
+        VkInstance instance, uint64_t flags) const noexcept {
     VkSurfaceKHR surface;
     VkExtent2D extent;
 
@@ -236,6 +424,26 @@ VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHR(void* nativeWin
             vkCreateAndroidSurfaceKHR(instance, &createInfo, VKALLOC, (VkSurfaceKHR*) &surface);
     FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
             << "vkCreateAndroidSurfaceKHR with error=" << static_cast<int32_t>(result);
-    return {surface, extent};
+    return { surface, extent };
 }
+
+// Deprecated platform dependent helper methods
+VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensionsImpl() { return {}; }
+
+VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadataImpl(
+        ExternalImageHandleRef externalImage, VkDevice device) {
+    return ExternalImageMetadata{};
 }
+
+VulkanPlatform::ImageData VulkanPlatform::createExternalImageDataImpl(
+        ExternalImageHandleRef externalImage, VkDevice device,
+        const ExternalImageMetadata& metadata, uint32_t memoryTypeIndex, VkImageUsageFlags usage) {
+    return ImageData{};
+}
+
+VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHRImpl(void* nativeWindow,
+        VkInstance instance, uint64_t flags) noexcept {
+    return SurfaceBundle{};
+}
+
+}// namespace filament::backend

filament/backend/src/vulkan/platform/VulkanPlatformApple.mm

@@ -52,7 +52,7 @@ using namespace bluevk;
 
 namespace filament::backend {
 
-VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensions() {
+VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensionsImpl() {
     ExtensionSet const ret = {
     #if defined(__APPLE__)
         VK_MVK_MACOS_SURFACE_EXTENSION_NAME,  // TODO: replace with VK_EXT_metal_surface
@@ -64,17 +64,30 @@ VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensions() {
 }
 
 VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadataImpl(
-        void* externalImage, VkDevice device) {
+        ExternalImageHandleRef externalImage, VkDevice device) {
     return {};
 }
 
-VulkanPlatform::ImageData VulkanPlatform::createExternalImageImpl(void* externalImage,
-        VkDevice device, const VkAllocationCallbacks* allocator,
-        const ExternalImageMetadata& metadata) {
+VulkanPlatform::ImageData VulkanPlatform::createExternalImageDataImpl(
+        ExternalImageHandleRef externalImage, VkDevice device,
+        const ExternalImageMetadata& metadata, uint32_t memoryTypeIndex, VkImageUsageFlags usage) {
     return {};
 }
 
-VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHR(void* nativeWindow,
+VkSampler VulkanPlatform::createExternalSamplerImpl(VkDevice device,
+    SamplerYcbcrConversion chroma,
+    SamplerParams sampler,
+    uint32_t internalFormat) {
+    return VK_NULL_HANDLE;
+}
+
+VkImageView VulkanPlatform::createExternalImageViewImpl(VkDevice device,
+        SamplerYcbcrConversion chroma, uint32_t internalFormat, VkImage image,
+        VkImageSubresourceRange range, VkImageViewType viewType, VkComponentMapping swizzle) {
+    return VK_NULL_HANDLE;
+}
+
+VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHRImpl(void* nativeWindow,
         VkInstance instance, uint64_t flags) noexcept {
     VkSurfaceKHR surface;
     #if defined(__APPLE__)

filament/backend/src/vulkan/platform/VulkanPlatformLinuxWindows.cpp

@@ -85,17 +85,30 @@ using namespace bluevk;
 namespace filament::backend {
 
 VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadataImpl(
-        void* externalImage, VkDevice device) {
+        ExternalImageHandleRef externalImage, VkDevice device) {
     return {};
 }
 
-VulkanPlatform::ImageData VulkanPlatform::createExternalImageImpl(void* externalImage,
-        VkDevice device, const VkAllocationCallbacks* allocator,
-        const ExternalImageMetadata& metadata) {
+VulkanPlatform::ImageData VulkanPlatform::createExternalImageDataImpl(
+        ExternalImageHandleRef externalImage, VkDevice device,
+        const ExternalImageMetadata& metadata, uint32_t memoryTypeIndex, VkImageUsageFlags usage) {
     return {};
 }
 
-VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensions() {
+VkSampler VulkanPlatform::createExternalSamplerImpl(VkDevice device,
+    SamplerYcbcrConversion chroma,
+    SamplerParams sampler,
+    uint32_t internalFormat) {
+    return VK_NULL_HANDLE;
+}
+
+VkImageView VulkanPlatform::createExternalImageViewImpl(VkDevice device,
+        SamplerYcbcrConversion chroma, uint32_t internalFormat, VkImage image,
+        VkImageSubresourceRange range, VkImageViewType viewType, VkComponentMapping swizzle) {
+    return VK_NULL_HANDLE;
+}
+
+VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensionsImpl() {
     VulkanPlatform::ExtensionSet const ret = {
 #if defined(__linux__) && defined(FILAMENT_SUPPORTS_WAYLAND)
         VK_KHR_WAYLAND_SURFACE_EXTENSION_NAME,
@@ -113,7 +126,7 @@ VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensions() {
     return ret;
 }
 
-VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHR(void* nativeWindow,
+VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHRImpl(void* nativeWindow,
         VkInstance instance, uint64_t flags) noexcept {
     VkSurfaceKHR surface;
 

filament/backend/src/vulkan/utils/Conversion.cpp

@@ -588,6 +588,79 @@ VkComponentMapping getSwizzleMap(TextureSwizzle const swizzle[4]) {
     return map;
 }
 
+VkFilter getFilter(SamplerMinFilter filter) {
+    switch (filter) {
+    case SamplerMinFilter::NEAREST:
+        return VK_FILTER_NEAREST;
+    case SamplerMinFilter::LINEAR:
+        return VK_FILTER_LINEAR;
+    case SamplerMinFilter::NEAREST_MIPMAP_NEAREST:
+        return VK_FILTER_NEAREST;
+    case SamplerMinFilter::LINEAR_MIPMAP_NEAREST:
+        return VK_FILTER_LINEAR;
+    case SamplerMinFilter::NEAREST_MIPMAP_LINEAR:
+        return VK_FILTER_NEAREST;
+    case SamplerMinFilter::LINEAR_MIPMAP_LINEAR:
+        return VK_FILTER_LINEAR;
+    }
+}
+
+VkFilter getFilter(SamplerMagFilter filter) {
+    switch (filter) {
+    case SamplerMagFilter::NEAREST:
+        return VK_FILTER_NEAREST;
+    case SamplerMagFilter::LINEAR:
+        return VK_FILTER_LINEAR;
+    }
+}
+
+VkSamplerMipmapMode getMipmapMode(SamplerMinFilter filter) {
+    switch (filter) {
+    case SamplerMinFilter::NEAREST:
+        return VK_SAMPLER_MIPMAP_MODE_NEAREST;
+    case SamplerMinFilter::LINEAR:
+        return VK_SAMPLER_MIPMAP_MODE_NEAREST;
+    case SamplerMinFilter::NEAREST_MIPMAP_NEAREST:
+        return VK_SAMPLER_MIPMAP_MODE_NEAREST;
+    case SamplerMinFilter::LINEAR_MIPMAP_NEAREST:
+        return VK_SAMPLER_MIPMAP_MODE_NEAREST;
+    case SamplerMinFilter::NEAREST_MIPMAP_LINEAR:
+        return VK_SAMPLER_MIPMAP_MODE_LINEAR;
+    case SamplerMinFilter::LINEAR_MIPMAP_LINEAR:
+        return VK_SAMPLER_MIPMAP_MODE_LINEAR;
+    }
+}
+
+VkSamplerAddressMode getWrapMode(SamplerWrapMode mode) {
+    switch (mode) {
+    case SamplerWrapMode::REPEAT:
+        return VK_SAMPLER_ADDRESS_MODE_REPEAT;
+    case SamplerWrapMode::CLAMP_TO_EDGE:
+        return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE;
+    case SamplerWrapMode::MIRRORED_REPEAT:
+        return VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT;
+    }
+}
+
+VkBool32 getCompareEnable(SamplerCompareMode mode) {
+    return mode == SamplerCompareMode::NONE ? VK_FALSE : VK_TRUE;
+}
+
+float getMaxLod(SamplerMinFilter filter) {
+    switch (filter) {
+    case SamplerMinFilter::NEAREST:
+    case SamplerMinFilter::LINEAR:
+        // The Vulkan spec recommends a max LOD of 0.25 to "disable" mipmapping.
+        // See "Mapping of OpenGL to Vulkan filter modes" in the VK Spec.
+        return 0.25f;
+    case SamplerMinFilter::NEAREST_MIPMAP_NEAREST:
+    case SamplerMinFilter::LINEAR_MIPMAP_NEAREST:
+    case SamplerMinFilter::NEAREST_MIPMAP_LINEAR:
+    case SamplerMinFilter::LINEAR_MIPMAP_LINEAR:
+        return VK_LOD_CLAMP_NONE;
+    }
+}
+
 VkShaderStageFlags getShaderStageFlags(ShaderStageFlags stageFlags) {
     VkShaderStageFlags flags = 0x0;
     if (any(stageFlags & ShaderStageFlags::VERTEX))     flags |= VK_SHADER_STAGE_VERTEX_BIT;

filament/backend/src/vulkan/utils/Conversion.h

@@ -56,6 +56,19 @@ uint32_t getComponentCount(VkFormat format);
 VkComponentMapping getSwizzleMap(TextureSwizzle const swizzle[4]);
 VkShaderStageFlags getShaderStageFlags(ShaderStageFlags stageFlags);
 
+// Needed by the Platform for external sampler creation
+VkFilter getFilter(SamplerMinFilter filter);
+VkFilter getFilter(SamplerMagFilter filter);
+VkSamplerMipmapMode getMipmapMode(SamplerMinFilter filter);
+VkSamplerAddressMode getWrapMode(SamplerWrapMode mode);
+VkBool32 getCompareEnable(SamplerCompareMode mode);
+float getMaxLod(SamplerMinFilter filter);
+
+// Ycbcr related functions
+VkSamplerYcbcrModelConversion getYcbcrModelConversion(SamplerYcbcrModelConversion model);
+VkSamplerYcbcrRange getYcbcrRange(SamplerYcbcrRange range);
+VkChromaLocation getChromaLocation(ChromaLocation loc);
+
 inline VkImageViewType getViewType(SamplerType target) {
     switch (target) {
         case SamplerType::SAMPLER_CUBEMAP:

filament/backend/src/vulkan/utils/Image.cpp

@@ -193,6 +193,49 @@ uint8_t reduceSampleCount(uint8_t sampleCount, VkSampleCountFlags mask) {
     return mostSignificantBit((sampleCount - 1) & mask);
 }
 
+VkSamplerYcbcrModelConversion getYcbcrModelConversion(
+    SamplerYcbcrModelConversion model) {
+    switch (model) {
+    case SamplerYcbcrModelConversion::RGB_IDENTITY:
+        return VK_SAMPLER_YCBCR_MODEL_CONVERSION_RGB_IDENTITY;
+    case SamplerYcbcrModelConversion::YCBCR_IDENTITY:
+        return VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_IDENTITY;
+    case SamplerYcbcrModelConversion::YCBCR_709:
+        return VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_709;
+    case SamplerYcbcrModelConversion::YCBCR_601:
+        return VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_601;
+    case SamplerYcbcrModelConversion::YCBCR_2020:
+        return VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_2020;
+    default:
+        assert_invariant(false &&
+            "Unknown data type, conversion is not supported.");
+    }
+}
+
+VkSamplerYcbcrRange getYcbcrRange(SamplerYcbcrRange range) {
+    switch (range) {
+    case SamplerYcbcrRange::ITU_FULL:
+        return VK_SAMPLER_YCBCR_RANGE_ITU_FULL;
+    case SamplerYcbcrRange::ITU_NARROW:
+        return VK_SAMPLER_YCBCR_RANGE_ITU_NARROW;
+    default:
+        assert_invariant(false &&
+            "Unknown data type, conversion is not supported.");
+    }
+}
+
+VkChromaLocation getChromaLocation(ChromaLocation loc) {
+    switch (loc) {
+    case ChromaLocation::COSITED_EVEN:
+        return VK_CHROMA_LOCATION_COSITED_EVEN;
+    case ChromaLocation::MIDPOINT:
+        return VK_CHROMA_LOCATION_MIDPOINT;
+    default:
+        assert_invariant(false &&
+            "Unknown data type, conversion is not supported.");
+    }
+}
+
 } // namespace filament::backend::fvkutils
 
 bool operator<(const VkImageSubresourceRange& a, const VkImageSubresourceRange& b) {

filament/backend/src/webgpu/WebGPUConstants.h

@@ -0,0 +1,66 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_BACKEND_WEBGPUCONSTANTS_H
+#define TNT_FILAMENT_BACKEND_WEBGPUCONSTANTS_H
+
+#include <utils/Log.h>
+
+#include <cstdint>
+
+// FWGPU is short for Filament WebGPU
+
+// turn on runtime validation, namely for debugging, that would normally not run (for release)
+#define FWGPU_DEBUG_VALIDATION     0x00000001
+// print/log system component details to the console, e.g. about the
+// instance, surface, adapter, device, etc.
+#define FWGPU_PRINT_SYSTEM         0x00000002
+
+// Set this to enable logging "only" to one output stream. This is useful in the case where we want
+// to debug with print statements and want ordered logging (e.g slog.i and slog.e will not appear in
+// order of calls).
+#define FWGPU_DEBUG_FORCE_LOG_TO_I 0x00000004
+
+// Useful default combinations
+#define FWGPU_DEBUG_EVERYTHING     0xFFFFFFFF
+
+#if defined(FILAMENT_BACKEND_DEBUG_FLAG)
+    #define FWGPU_DEBUG_FORWARDED_FLAG (FILAMENT_BACKEND_DEBUG_FLAG & FWGPU_DEBUG_EVERYTHING)
+#else
+    #define FWGPU_DEBUG_FORWARDED_FLAG 0
+#endif
+
+#ifndef NDEBUG
+    #define FWGPU_DEBUG_FLAGS FWGPU_DEBUG_FORWARDED_FLAG
+#else
+    #define FWGPU_DEBUG_FLAGS 0
+#endif
+
+#define FWGPU_ENABLED(flags) (((FWGPU_DEBUG_FLAGS) & (flags)) == (flags))
+
+#if FWGPU_ENABLED(FWGPU_DEBUG_FORCE_LOG_TO_I)
+    #define FWGPU_LOGI (utils::slog.i)
+    #define FWGPU_LOGD FWGPU_LOGI
+    #define FWGPU_LOGE FWGPU_LOGI
+    #define FWGPU_LOGW FWGPU_LOGI
+#else
+    #define FWGPU_LOGE (utils::slog.e)
+    #define FWGPU_LOGW (utils::slog.w)
+    #define FWGPU_LOGD (utils::slog.d)
+    #define FWGPU_LOGI (utils::slog.i)
+#endif
+
+#endif// TNT_FILAMENT_BACKEND_WEBGPUCONSTANTS_H

filament/backend/src/webgpu/WebGPUDriver.cpp

@@ -14,21 +14,254 @@
  * limitations under the License.
  */
 
+#include "webgpu/WebGPUDriver.h"
+
+#include "webgpu/WebGPUConstants.h"
+#include <backend/platforms/WebGPUPlatform.h>
+
+#include "CommandStreamDispatcher.h"
+#include "DriverBase.h"
+#include "private/backend/Dispatcher.h"
 #include <backend/DriverEnums.h>
 #include <backend/Handle.h>
 
-#include "webgpu/WebGPUDriver.h"
-#include "CommandStreamDispatcher.h"
+#include <math/mat3.h>
+#include <utils/CString.h>
+#include <utils/ostream.h>
 
-#include <stdint.h>
+#include <dawn/webgpu_cpp_print.h>
+#include <webgpu/webgpu_cpp.h>
+
+#include <algorithm>
+#include <cstddef>
+#include <cstdint>
+#include <sstream>
+#include <string_view>
+#include <utility>
+#include <variant>
 
 namespace filament::backend {
 
-Driver* WebGPUDriver::create() {
-    return new WebGPUDriver();
+namespace {
+
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+void printInstanceDetails(wgpu::Instance const& instance) {
+    wgpu::SupportedWGSLLanguageFeatures supportedWGSLLanguageFeatures{};
+    if (!instance.GetWGSLLanguageFeatures(&supportedWGSLLanguageFeatures)) {
+        FWGPU_LOGW << "Failed to get WebGPU instance supported WGSL language features"
+                   << utils::io::endl;
+    } else {
+        FWGPU_LOGI << "WebGPU instance supported WGSL language features ("
+                   << supportedWGSLLanguageFeatures.featureCount << "):" << utils::io::endl;
+        if (supportedWGSLLanguageFeatures.featureCount > 0 &&
+                supportedWGSLLanguageFeatures.features != nullptr) {
+            std::for_each(supportedWGSLLanguageFeatures.features,
+                    supportedWGSLLanguageFeatures.features +
+                            supportedWGSLLanguageFeatures.featureCount,
+                    [](wgpu::WGSLLanguageFeatureName const featureName) {
+                        std::stringstream nameStream{};
+                        nameStream << featureName;
+                        FWGPU_LOGI << "  " << nameStream.str() << utils::io::endl;
+                    });
+        }
+    }
+}
+#endif
+
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+void printLimit(std::string_view name, const std::variant<uint32_t, uint64_t> value) {
+    FWGPU_LOGI << "  " << name.data() << ": ";
+    bool undefined = true;
+    if (std::holds_alternative<uint32_t>(value)) {
+        if (std::get<uint32_t>(value) != WGPU_LIMIT_U32_UNDEFINED) {
+            undefined = false;
+            FWGPU_LOGI << std::get<uint32_t>(value);
+        }
+    } else if (std::holds_alternative<uint64_t>(value)) {
+        if (std::get<uint64_t>(value) != WGPU_LIMIT_U64_UNDEFINED) {
+            undefined = false;
+            FWGPU_LOGI << std::get<uint64_t>(value);
+        }
+    }
+    if (undefined) {
+        FWGPU_LOGI << "UNDEFINED";
+    }
+    FWGPU_LOGI << utils::io::endl;
+}
+#endif
+
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+void printLimits(wgpu::Limits const& limits) {
+    printLimit("maxTextureDimension1D", limits.maxTextureDimension1D);
+    printLimit("maxTextureDimension2D", limits.maxTextureDimension2D);
+    printLimit("maxTextureDimension3D", limits.maxTextureDimension3D);
+    printLimit("maxTextureArrayLayers", limits.maxTextureArrayLayers);
+    printLimit("maxBindGroups", limits.maxBindGroups);
+    printLimit("maxBindGroupsPlusVertexBuffers", limits.maxBindGroupsPlusVertexBuffers);
+    printLimit("maxBindingsPerBindGroup", limits.maxBindingsPerBindGroup);
+    printLimit("maxDynamicUniformBuffersPerPipelineLayout",
+            limits.maxDynamicUniformBuffersPerPipelineLayout);
+    printLimit("maxDynamicStorageBuffersPerPipelineLayout",
+            limits.maxDynamicStorageBuffersPerPipelineLayout);
+    printLimit("maxSampledTexturesPerShaderStage", limits.maxSampledTexturesPerShaderStage);
+    printLimit("maxSamplersPerShaderStage", limits.maxSamplersPerShaderStage);
+    printLimit("maxStorageBuffersPerShaderStage", limits.maxStorageBuffersPerShaderStage);
+    printLimit("maxStorageTexturesPerShaderStage", limits.maxStorageTexturesPerShaderStage);
+    printLimit("maxUniformBuffersPerShaderStage", limits.maxUniformBuffersPerShaderStage);
+    printLimit("maxUniformBufferBindingSize", limits.maxUniformBufferBindingSize);
+    printLimit("maxStorageBufferBindingSize", limits.maxStorageBufferBindingSize);
+    printLimit("minUniformBufferOffsetAlignment", limits.minUniformBufferOffsetAlignment);
+    printLimit("minStorageBufferOffsetAlignment", limits.minStorageBufferOffsetAlignment);
+    printLimit("maxVertexBuffers", limits.maxVertexBuffers);
+    printLimit("maxBufferSize", limits.maxBufferSize);
+    printLimit("maxVertexAttributes", limits.maxVertexAttributes);
+    printLimit("maxVertexBufferArrayStride", limits.maxVertexBufferArrayStride);
+    printLimit("maxInterStageShaderComponents", limits.maxInterStageShaderComponents);
+    printLimit("maxInterStageShaderVariables", limits.maxInterStageShaderVariables);
+    printLimit("maxColorAttachments", limits.maxColorAttachments);
+    printLimit("maxColorAttachmentBytesPerSample", limits.maxColorAttachmentBytesPerSample);
+    printLimit("maxComputeWorkgroupStorageSize", limits.maxComputeWorkgroupStorageSize);
+    printLimit("maxComputeInvocationsPerWorkgroup", limits.maxComputeInvocationsPerWorkgroup);
+    printLimit("maxComputeWorkgroupSizeX", limits.maxComputeWorkgroupSizeX);
+    printLimit("maxComputeWorkgroupSizeY", limits.maxComputeWorkgroupSizeY);
+    printLimit("maxComputeWorkgroupSizeZ", limits.maxComputeWorkgroupSizeZ);
+    printLimit("maxComputeWorkgroupsPerDimension", limits.maxComputeWorkgroupsPerDimension);
+    printLimit("maxStorageBuffersInVertexStage", limits.maxStorageBuffersInVertexStage);
+    printLimit("maxStorageTexturesInVertexStage", limits.maxStorageTexturesInVertexStage);
+    printLimit("maxStorageBuffersInFragmentStage", limits.maxStorageBuffersInFragmentStage);
+    printLimit("maxStorageTexturesInFragmentStage", limits.maxStorageTexturesInFragmentStage);
+}
+#endif
+
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+void printAdapterDetails(wgpu::Adapter const& adapter) {
+    wgpu::DawnAdapterPropertiesPowerPreference powerPreferenceProperties{};
+    wgpu::AdapterInfo adapterInfo{};
+    adapterInfo.nextInChain = &powerPreferenceProperties;
+    if (!adapter.GetInfo(&adapterInfo)) {
+        FWGPU_LOGW << "Failed to get WebGPU adapter info" << utils::io::endl;
+    } else {
+        std::stringstream backendTypeStream{};
+        backendTypeStream << adapterInfo.backendType;
+        std::stringstream adapterTypeStream{};
+        adapterTypeStream << adapterInfo.adapterType;
+        std::stringstream powerPreferenceStream{};
+        powerPreferenceStream << powerPreferenceProperties.powerPreference;
+        FWGPU_LOGI << "WebGPU adapter info:" << utils::io::endl;
+        FWGPU_LOGI << "  vendor: " << adapterInfo.vendor.data << utils::io::endl;
+        FWGPU_LOGI << "  architecture: " << adapterInfo.architecture.data << utils::io::endl;
+        FWGPU_LOGI << "  device: " << adapterInfo.device.data << utils::io::endl;
+        FWGPU_LOGI << "  description: " << adapterInfo.description.data << utils::io::endl;
+        FWGPU_LOGI << "  backend type: " << backendTypeStream.str().data() << utils::io::endl;
+        FWGPU_LOGI << "  adapter type: " << adapterTypeStream.str().data() << utils::io::endl;
+        FWGPU_LOGI << "  device ID: " << adapterInfo.deviceID << utils::io::endl;
+        FWGPU_LOGI << "  vendor ID: " << adapterInfo.vendorID << utils::io::endl;
+        FWGPU_LOGI << "  subgroup min size: " << adapterInfo.subgroupMinSize << utils::io::endl;
+        FWGPU_LOGI << "  subgroup max size: " << adapterInfo.subgroupMaxSize << utils::io::endl;
+        FWGPU_LOGI << "  compatibility mode: " << bool(adapterInfo.compatibilityMode)
+                   << utils::io::endl;
+        FWGPU_LOGI << "  power preference: " << powerPreferenceStream.str() << utils::io::endl;
+    }
+    wgpu::SupportedFeatures supportedFeatures{};
+    adapter.GetFeatures(&supportedFeatures);
+    FWGPU_LOGI << "WebGPU adapter supported features (" << supportedFeatures.featureCount
+               << "):" << utils::io::endl;
+    if (supportedFeatures.featureCount > 0 && supportedFeatures.features != nullptr) {
+        std::for_each(supportedFeatures.features,
+                supportedFeatures.features + supportedFeatures.featureCount,
+                [](wgpu::FeatureName const name) {
+                    std::stringstream nameStream{};
+                    nameStream << name;
+                    FWGPU_LOGI << "  " << nameStream.str().data() << utils::io::endl;
+                });
+    }
+    wgpu::SupportedLimits supportedLimits{};
+    if (!adapter.GetLimits(&supportedLimits)) {
+        FWGPU_LOGW << "Failed to get WebGPU adapter supported limits" << utils::io::endl;
+    } else {
+        FWGPU_LOGI << "WebGPU adapter supported limits:" << utils::io::endl;
+        printLimits(supportedLimits.limits);
+    }
+}
+#endif
+
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+void printSurfaceCapabilitiesDetails(wgpu::SurfaceCapabilities const& capabilities) {
+    std::stringstream usages_stream{};
+    usages_stream << capabilities.usages;
+    FWGPU_LOGI << "WebGPU surface capabilities:" << utils::io::endl;
+    FWGPU_LOGI << "  surface usages: " << usages_stream.str().data() << utils::io::endl;
+    FWGPU_LOGI << "  surface formats (" << capabilities.formatCount << "):" << utils::io::endl;
+    if (capabilities.formatCount > 0 && capabilities.formats != nullptr) {
+        std::for_each(capabilities.formats, capabilities.formats + capabilities.formatCount,
+                [](wgpu::TextureFormat const format) {
+                    std::stringstream format_stream{};
+                    format_stream << format;
+                    FWGPU_LOGI << "    " << format_stream.str().data() << utils::io::endl;
+                });
+    }
+    FWGPU_LOGI << "  surface present modes (" << capabilities.presentModeCount
+               << "):" << utils::io::endl;
+    if (capabilities.presentModeCount > 0 && capabilities.presentModes != nullptr) {
+        std::for_each(capabilities.presentModes,
+                capabilities.presentModes + capabilities.presentModeCount,
+                [](wgpu::PresentMode const presentMode) {
+                    std::stringstream present_mode_stream{};
+                    present_mode_stream << presentMode;
+                    FWGPU_LOGI << "    " << present_mode_stream.str().data() << utils::io::endl;
+                });
+    }
+    FWGPU_LOGI << "  surface alpha modes (" << capabilities.alphaModeCount
+               << "):" << utils::io::endl;
+    if (capabilities.alphaModeCount > 0 && capabilities.alphaModes != nullptr) {
+        std::for_each(capabilities.alphaModes,
+                capabilities.alphaModes + capabilities.alphaModeCount,
+                [](wgpu::CompositeAlphaMode const alphaMode) {
+                    std::stringstream alpha_mode_stream{};
+                    alpha_mode_stream << alphaMode;
+                    FWGPU_LOGI << "    " << alpha_mode_stream.str().data() << utils::io::endl;
+                });
+    }
+}
+#endif
+
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+void printDeviceDetails(wgpu::Device const& device) {
+    wgpu::SupportedFeatures supportedFeatures{};
+    device.GetFeatures(&supportedFeatures);
+    FWGPU_LOGI << "WebGPU device supported features (" << supportedFeatures.featureCount
+               << "):" << utils::io::endl;
+    if (supportedFeatures.featureCount > 0 && supportedFeatures.features != nullptr) {
+        std::for_each(supportedFeatures.features,
+                supportedFeatures.features + supportedFeatures.featureCount,
+                [](wgpu::FeatureName const name) {
+                    std::stringstream nameStream{};
+                    nameStream << name;
+                    FWGPU_LOGI << "  " << nameStream.str().data() << utils::io::endl;
+                });
+    }
+    wgpu::SupportedLimits supportedLimits{};
+    if (!device.GetLimits(&supportedLimits)) {
+        FWGPU_LOGW << "Failed to get WebGPU supported device limits" << utils::io::endl;
+    } else {
+        FWGPU_LOGI << "WebGPU device supported limits:" << utils::io::endl;
+        printLimits(supportedLimits.limits);
+    }
+}
+#endif
+
+}// namespace
+
+Driver* WebGPUDriver::create(WebGPUPlatform& platform) noexcept {
+    return new WebGPUDriver(platform);
 }
 
-WebGPUDriver::WebGPUDriver() noexcept = default;
+WebGPUDriver::WebGPUDriver(WebGPUPlatform& platform) noexcept
+    : mPlatform(platform) {
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+    printInstanceDetails(mPlatform.getInstance());
+#endif
+}
 
 WebGPUDriver::~WebGPUDriver() noexcept = default;
 
@@ -45,7 +278,7 @@ ShaderModel WebGPUDriver::getShaderModel() const noexcept {
 }
 
 ShaderLanguage WebGPUDriver::getShaderLanguage() const noexcept {
-    return ShaderLanguage::ESSL3;
+    return ShaderLanguage::WGSL;
 }
 
 // explicit instantiation of the Dispatcher
@@ -123,6 +356,188 @@ void WebGPUDriver::destroyDescriptorSetLayout(Handle<HwDescriptorSetLayout> tqh)
 void WebGPUDriver::destroyDescriptorSet(Handle<HwDescriptorSet> tqh) {
 }
 
+Handle<HwSwapChain> WebGPUDriver::createSwapChainS() noexcept {
+    return Handle<HwSwapChain>((Handle<HwSwapChain>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwSwapChain> WebGPUDriver::createSwapChainHeadlessS() noexcept {
+    return Handle<HwSwapChain>((Handle<HwSwapChain>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwTexture> WebGPUDriver::createTextureS() noexcept {
+    return Handle<HwTexture>((Handle<HwTexture>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwTexture> WebGPUDriver::importTextureS() noexcept {
+    return Handle<HwTexture>((Handle<HwTexture>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwProgram> WebGPUDriver::createProgramS() noexcept {
+    return Handle<HwProgram>((Handle<HwProgram>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwFence> WebGPUDriver::createFenceS() noexcept {
+    return Handle<HwFence>((Handle<HwFence>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwTimerQuery> WebGPUDriver::createTimerQueryS() noexcept {
+    return Handle<HwTimerQuery>((Handle<HwTimerQuery>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwIndexBuffer> WebGPUDriver::createIndexBufferS() noexcept {
+    return Handle<HwIndexBuffer>((Handle<HwIndexBuffer>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwTexture> WebGPUDriver::createTextureViewS() noexcept {
+    return Handle<HwTexture>((Handle<HwTexture>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwBufferObject> WebGPUDriver::createBufferObjectS() noexcept {
+    return Handle<HwBufferObject>((Handle<HwBufferObject>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwRenderTarget> WebGPUDriver::createRenderTargetS() noexcept {
+    return Handle<HwRenderTarget>((Handle<HwRenderTarget>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwVertexBuffer> WebGPUDriver::createVertexBufferS() noexcept {
+    return Handle<HwVertexBuffer>((Handle<HwVertexBuffer>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwDescriptorSet> WebGPUDriver::createDescriptorSetS() noexcept {
+    return Handle<HwDescriptorSet>((Handle<HwDescriptorSet>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwRenderPrimitive> WebGPUDriver::createRenderPrimitiveS() noexcept {
+    return Handle<HwRenderPrimitive>((Handle<HwRenderPrimitive>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwVertexBufferInfo> WebGPUDriver::createVertexBufferInfoS() noexcept {
+    return Handle<HwVertexBufferInfo>((Handle<HwVertexBufferInfo>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwTexture> WebGPUDriver::createTextureViewSwizzleS() noexcept {
+    return Handle<HwTexture>((Handle<HwTexture>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwRenderTarget> WebGPUDriver::createDefaultRenderTargetS() noexcept {
+    return Handle<HwRenderTarget>((Handle<HwRenderTarget>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwDescriptorSetLayout> WebGPUDriver::createDescriptorSetLayoutS() noexcept {
+    return Handle<HwDescriptorSetLayout>(
+            (Handle<HwDescriptorSetLayout>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwTexture> WebGPUDriver::createTextureExternalImageS() noexcept {
+    return Handle<HwTexture>((Handle<HwTexture>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwTexture> WebGPUDriver::createTextureExternalImage2S() noexcept {
+    return Handle<HwTexture>((Handle<HwTexture>::HandleId) mNextFakeHandle++);
+}
+
+Handle<HwTexture> WebGPUDriver::createTextureExternalImagePlaneS() noexcept {
+    return Handle<HwTexture>((Handle<HwTexture>::HandleId) mNextFakeHandle++);
+}
+
+void WebGPUDriver::createSwapChainR(Handle<HwSwapChain> sch, void* nativeWindow, uint64_t flags) {
+    mSurface= mPlatform.createSurface(nativeWindow, flags);
+    mAdapter = mPlatform.requestAdapter(mSurface);
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+    printAdapterDetails(mAdapter);
+    wgpu::SurfaceCapabilities surfaceCapabilities{};
+    if (!mSurface.GetCapabilities(mAdapter, &surfaceCapabilities)) {
+        FWGPU_LOGW << "Failed to get WebGPU surface capabilities" << utils::io::endl;
+    } else {
+        printSurfaceCapabilitiesDetails(surfaceCapabilities);
+    }
+#endif
+    mDevice = mPlatform.requestDevice(mAdapter);
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+    printDeviceDetails(mDevice);
+#endif
+    mQueue = mDevice.GetQueue();
+    // TODO configure the surface (maybe before or after creating the swapchain?
+    //                             how do we get the surface extent?)
+    // TODO actually create the swapchain
+    FWGPU_LOGW << "WebGPU support is still essentially a no-op at this point in development (only "
+                  "background components have been instantiated/selected, such as surface/screen, "
+                  "graphics device/GPU, etc.), thus nothing is being drawn to the screen."
+               << utils::io::endl;
+#if !FWGPU_ENABLED(FWGPU_PRINT_SYSTEM) && !defined(NDEBUG)
+    FWGPU_LOGI << "If the FILAMENT_BACKEND_DEBUG_FLAG variable were set with the " << utils::io::hex
+               << FWGPU_PRINT_SYSTEM << utils::io::dec
+               << " bit flag on during build time the application would print system details "
+                  "about the selected graphics device, surface, etc. To see this try "
+                  "rebuilding Filament with that flag, e.g. ./build.sh -x "
+               << FWGPU_PRINT_SYSTEM << " ..." << utils::io::endl;
+#endif
+}
+
+void WebGPUDriver::createSwapChainHeadlessR(Handle<HwSwapChain> sch, uint32_t width,
+        uint32_t height, uint64_t flags) {}
+
+void WebGPUDriver::createVertexBufferInfoR(Handle<HwVertexBufferInfo> vbih, uint8_t bufferCount,
+        uint8_t attributeCount, AttributeArray attributes) {}
+
+void WebGPUDriver::createVertexBufferR(Handle<HwVertexBuffer> vbh, uint32_t vertexCount,
+        Handle<HwVertexBufferInfo> vbih) {}
+
+void WebGPUDriver::createIndexBufferR(Handle<HwIndexBuffer> ibh, ElementType elementType,
+        uint32_t indexCount, BufferUsage usage) {}
+
+void WebGPUDriver::createBufferObjectR(Handle<HwBufferObject> boh, uint32_t byteCount,
+        BufferObjectBinding bindingType, BufferUsage usage) {}
+
+void WebGPUDriver::createTextureR(Handle<HwTexture> th, SamplerType target, uint8_t levels,
+        TextureFormat format, uint8_t samples, uint32_t w, uint32_t h, uint32_t depth,
+        TextureUsage usage) {}
+
+void WebGPUDriver::createTextureViewR(Handle<HwTexture> th, Handle<HwTexture> srch,
+        uint8_t baseLevel, uint8_t levelCount) {}
+
+void WebGPUDriver::createTextureViewSwizzleR(Handle<HwTexture> th, Handle<HwTexture> srch,
+        backend::TextureSwizzle r, backend::TextureSwizzle g, backend::TextureSwizzle b,
+        backend::TextureSwizzle a) {}
+
+void WebGPUDriver::createTextureExternalImage2R(Handle<HwTexture> th, backend::SamplerType target,
+        backend::TextureFormat format, uint32_t width, uint32_t height, backend::TextureUsage usage,
+        Platform::ExternalImageHandleRef externalImage) {}
+
+void WebGPUDriver::createTextureExternalImageR(Handle<HwTexture> th, backend::SamplerType target,
+        backend::TextureFormat format, uint32_t width, uint32_t height, backend::TextureUsage usage,
+        void* externalImage) {}
+
+void WebGPUDriver::createTextureExternalImagePlaneR(Handle<HwTexture> th,
+        backend::TextureFormat format, uint32_t width, uint32_t height, backend::TextureUsage usage,
+        void* image, uint32_t plane) {}
+
+void WebGPUDriver::importTextureR(Handle<HwTexture> th, intptr_t id, SamplerType target,
+        uint8_t levels, TextureFormat format, uint8_t samples, uint32_t w, uint32_t h,
+        uint32_t depth, TextureUsage usage) {}
+
+void WebGPUDriver::createRenderPrimitiveR(Handle<HwRenderPrimitive> rph, Handle<HwVertexBuffer> vbh,
+        Handle<HwIndexBuffer> ibh, PrimitiveType pt) {}
+
+void WebGPUDriver::createProgramR(Handle<HwProgram> ph, Program&& program) {}
+
+void WebGPUDriver::createDefaultRenderTargetR(Handle<HwRenderTarget> rth, int) {}
+
+void WebGPUDriver::createRenderTargetR(Handle<HwRenderTarget> rth, TargetBufferFlags targets,
+        uint32_t width, uint32_t height, uint8_t samples, uint8_t layerCount, MRT color,
+        TargetBufferInfo depth, TargetBufferInfo stencil) {}
+
+void WebGPUDriver::createFenceR(Handle<HwFence> fh, int) {}
+
+void WebGPUDriver::createTimerQueryR(Handle<HwTimerQuery> tqh, int) {}
+
+void WebGPUDriver::createDescriptorSetLayoutR(Handle<HwDescriptorSetLayout> dslh,
+        backend::DescriptorSetLayout&& info) {}
+
+void WebGPUDriver::createDescriptorSetR(Handle<HwDescriptorSet> dsh,
+        Handle<HwDescriptorSetLayout> dslh) {}
+
 Handle<HwStream> WebGPUDriver::createStreamNative(void* nativeStream) {
     return {};
 }
@@ -131,9 +546,11 @@ Handle<HwStream> WebGPUDriver::createStreamAcquired() {
     return {};
 }
 
-void WebGPUDriver::setAcquiredImage(Handle<HwStream> sh, void* image,
-        CallbackHandler* handler, StreamCallback cb, void* userData) {
-}
+void WebGPUDriver::setAcquiredImage(Handle<HwStream> sh, void* image, const math::mat3f& transform,
+        CallbackHandler* handler, StreamCallback cb, void* userData) {}
+
+void WebGPUDriver::registerBufferObjectStreams(Handle<HwBufferObject> boh,
+        BufferObjectStreamDescriptor&& streams) {}
 
 void WebGPUDriver::setStreamDimensions(Handle<HwStream> sh, uint32_t width, uint32_t height) {
 }
@@ -238,6 +655,14 @@ size_t WebGPUDriver::getMaxUniformBufferSize() {
     return 16384u;
 }
 
+size_t WebGPUDriver::getMaxTextureSize(SamplerType target) {
+    return 2048u;
+}
+
+size_t WebGPUDriver::getMaxArrayTextureLayers() {
+    return 256u;
+}
+
 void WebGPUDriver::updateIndexBuffer(Handle<HwIndexBuffer> ibh, BufferDescriptor&& p,
         uint32_t byteOffset) {
     scheduleDestroy(std::move(p));

filament/backend/src/webgpu/WebGPUDriver.h

@@ -17,26 +17,44 @@
 #ifndef TNT_FILAMENT_BACKEND_WEBGPUDRIVER_H
 #define TNT_FILAMENT_BACKEND_WEBGPUDRIVER_H
 
-#include "private/backend/Driver.h"
+#include <backend/platforms/WebGPUPlatform.h>
+
 #include "DriverBase.h"
+#include "private/backend/Dispatcher.h"
+#include "private/backend/Driver.h"
+#include <backend/DriverEnums.h>
 
 #include <utils/compiler.h>
 
+#include <webgpu/webgpu_cpp.h>
+
+#include <cstdint>
+
 namespace filament::backend {
 
+/**
+ * WebGPU backend (driver) implementation
+ */
 class WebGPUDriver final : public DriverBase {
-    WebGPUDriver() noexcept;
-    ~WebGPUDriver() noexcept override;
-    Dispatcher getDispatcher() const noexcept final;
-
 public:
-    static Driver* create();
+    ~WebGPUDriver() noexcept override;
+
+    [[nodiscard]] Dispatcher getDispatcher() const noexcept final;
+    [[nodiscard]] static Driver* create(WebGPUPlatform& platform) noexcept;
 
 private:
-    ShaderModel getShaderModel() const noexcept final;
-    ShaderLanguage getShaderLanguage() const noexcept final;
+    explicit WebGPUDriver(WebGPUPlatform& platform) noexcept;
+    [[nodiscard]] ShaderModel getShaderModel() const noexcept final;
+    [[nodiscard]] ShaderLanguage getShaderLanguage() const noexcept final;
 
-    uint64_t nextFakeHandle = 1;
+    // the platform (e.g. OS) specific aspects of the WebGPU backend are strictly only
+    // handled in the WebGPUPlatform
+    WebGPUPlatform& mPlatform;
+    wgpu::Surface mSurface = nullptr;
+    wgpu::Adapter mAdapter = nullptr;
+    wgpu::Device mDevice = nullptr;
+    wgpu::Queue mQueue = nullptr;
+    uint64_t mNextFakeHandle = 1;
 
     /*
      * Driver interface
@@ -45,20 +63,19 @@ private:
     template<typename T>
     friend class ConcreteDispatcher;
 
-#define DECL_DRIVER_API(methodName, paramsDecl, params) \
+#define DECL_DRIVER_API(methodName, paramsDecl, params)                                            \
     UTILS_ALWAYS_INLINE inline void methodName(paramsDecl);
 
-#define DECL_DRIVER_API_SYNCHRONOUS(RetType, methodName, paramsDecl, params) \
+#define DECL_DRIVER_API_SYNCHRONOUS(RetType, methodName, paramsDecl, params)                       \
     RetType methodName(paramsDecl) override;
 
-#define DECL_DRIVER_API_RETURN(RetType, methodName, paramsDecl, params) \
-    RetType methodName##S() noexcept override { \
-        return RetType((RetType::HandleId)nextFakeHandle++); } \
-    UTILS_ALWAYS_INLINE inline void methodName##R(RetType, paramsDecl) { }
+#define DECL_DRIVER_API_RETURN(RetType, methodName, paramsDecl, params)                            \
+    RetType methodName##S() noexcept override;                                                     \
+    UTILS_ALWAYS_INLINE inline void methodName##R(RetType, paramsDecl);
 
 #include "private/backend/DriverAPI.inc"
 };
 
-} // namespace filament
+}// namespace filament::backend
 
-#endif // TNT_FILAMENT_BACKEND_WEBGPUDRIVER_H
+#endif// TNT_FILAMENT_BACKEND_WEBGPUDRIVER_H

filament/backend/src/webgpu/WebGPUPlatform.cpp

@@ -1,36 +0,0 @@
-/*
- * Copyright (C) 2025 The Android Open Source Project
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- */
-
-#include "webgpu/WebGPUPlatform.h"
-#include "webgpu/WebGPUDriver.h"
-
-#include "utils/Log.h"
-
-namespace filament::backend {
-
-Driver* WebGPUPlatform::createDriver(void* const sharedContext , const Platform::DriverConfig& driverConfig) noexcept {
-    wgpu::InstanceDescriptor instance_descriptor {};
-    wgpu::Instance instance = wgpu::CreateInstance(&instance_descriptor);
-    if (instance) {
-        utils::slog.i << " WebGPU instance created\n";
-    } else {
-        utils::slog.e << " WebGPU instance failed to create\n";
-        return nullptr;
-    }
-    return WebGPUDriver::create();
-}
-
-} // namespace filament

filament/backend/src/webgpu/platform/WebGPUPlatform.cpp

@@ -0,0 +1,159 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <backend/platforms/WebGPUPlatform.h>
+
+#include "webgpu/WebGPUConstants.h"
+#include "webgpu/WebGPUDriver.h"
+
+#include <backend/Platform.h>
+#include <utils/Panic.h>
+#include <utils/ostream.h>
+
+#include <dawn/webgpu_cpp_print.h>
+#include <webgpu/webgpu_cpp.h>
+
+#include <cstdint>
+#include <sstream>
+
+/**
+ * WebGPU Backend implementation common across platforms or operating systems (at least for now).
+ * Some of these functions may likely be refactored to platform/OS-specific implementations
+ * over time as needed. The caller of the WebGPUPlatform doesn't need to care which is the case.
+ */
+
+namespace filament::backend {
+
+namespace {
+
+//either returns a valid instance or panics
+[[nodiscard]] wgpu::Instance createInstance() {
+    wgpu::DawnTogglesDescriptor dawnTogglesDescriptor{};
+#if defined(FILAMENT_WEBGPU_IMMEDIATE_ERROR_HANDLING)
+#if FWGPU_ENABLED(FWGPU_PRINT_SYSTEM)
+    FWGPU_LOGI << "setting on toggle enable_immediate_error_handling" << utils::io::endl;
+#endif
+    /**
+     * Have the un-captured error callback invoked immediately when an error occurs, rather than
+     * waiting for the next Tick. This enables using the stack trace in which the un-captured error
+     * occurred when breaking into the un-captured error callback.
+     * https://crbug.com/dawn/1789
+     */
+    static const char* toggleName = "enable_immediate_error_handling";
+    dawnTogglesDescriptor.enabledToggleCount = 1;
+    dawnTogglesDescriptor.enabledToggles = &toggleName;
+#endif
+    wgpu::InstanceDescriptor instanceDescriptor{
+        .nextInChain = &dawnTogglesDescriptor,
+        .capabilities = {
+            .timedWaitAnyEnable = true// TODO consider using pure async instead
+        }
+    };
+    wgpu::Instance instance = wgpu::CreateInstance(&instanceDescriptor);
+    FILAMENT_CHECK_POSTCONDITION(instance != nullptr) << "Unable to create webgpu instance.";
+    return instance;
+}
+
+}// namespace
+
+wgpu::Adapter WebGPUPlatform::requestAdapter(wgpu::Surface const& surface) {
+    // TODO consider power preference etc. (can be custom preferences passed to the platform or
+    //                                      based on whether this is a Mobile or Desktop system,
+    //                                      etc...)
+    wgpu::RequestAdapterOptions adaptorOptions{ .compatibleSurface = surface };
+    // note this just gets the first adapter
+    wgpu::Adapter adapter = nullptr;
+    wgpu::WaitStatus status = mInstance.WaitAny(
+            mInstance.RequestAdapter(&adaptorOptions, wgpu::CallbackMode::WaitAnyOnly,
+                    [&adapter](wgpu::RequestAdapterStatus const status,
+                            wgpu::Adapter const& readyAdapter, wgpu::StringView const message) {
+                        // TODO consider more robust error handling
+                        FILAMENT_CHECK_POSTCONDITION(status == wgpu::RequestAdapterStatus::Success)
+                                << "Unable to request a WebGPU adapter. Status "
+                                << static_cast<uint32_t>(status)
+                                << " with message: " << message.data;
+                        adapter = readyAdapter;
+                    }),
+            UINT16_MAX);// TODO define reasonable timeout (or do this asynchronously)
+    FILAMENT_CHECK_POSTCONDITION(status == wgpu::WaitStatus::Success)
+            << "Non-successful wait status requesting a WebGPU adapter "
+            << static_cast<uint32_t>(status);
+    FILAMENT_CHECK_POSTCONDITION(adapter != nullptr)
+            << "Failed to get a WebGPU adapter for the platform";
+    // TODO consider validating adapter has required features and/or limits
+    return adapter;
+}
+
+wgpu::Device WebGPUPlatform::requestDevice(wgpu::Adapter const& adapter) {
+    // TODO consider passing required features and/or limits
+    wgpu::DeviceDescriptor deviceDescriptor{};
+    deviceDescriptor.label = "graphics_device";
+    deviceDescriptor.defaultQueue.label = "default_queue";
+    deviceDescriptor.SetDeviceLostCallback(wgpu::CallbackMode::AllowSpontaneous,
+            [](wgpu::Device const&, wgpu::DeviceLostReason const& reason,
+                    wgpu::StringView message) {
+                if (reason == wgpu::DeviceLostReason::Destroyed) {
+                    FWGPU_LOGD << "WebGPU device lost due to being destroyed (expected)"
+                               << utils::io::endl;
+                    return;
+                }
+                // TODO try recreating the device instead of just panicking
+                std::stringstream reasonStream{};
+                reasonStream << reason;
+                FILAMENT_CHECK_POSTCONDITION(reason != wgpu::DeviceLostReason::Destroyed)
+                        << "WebGPU device lost: " << reasonStream.str() << " " << message.data;
+            });
+    deviceDescriptor.SetUncapturedErrorCallback(
+            [](wgpu::Device const&, wgpu::ErrorType errorType, wgpu::StringView message) {
+                std::stringstream typeStream{};
+                typeStream << errorType;
+                FWGPU_LOGE << "WebGPU device error: " << typeStream.str() << " " << message.data
+                           << utils::io::endl;
+            });
+    wgpu::Device device = nullptr;
+    wgpu::WaitStatus status = mInstance.WaitAny(
+            adapter.RequestDevice(&deviceDescriptor, wgpu::CallbackMode::WaitAnyOnly,
+                    [&device](wgpu::RequestDeviceStatus const status,
+                            wgpu::Device const& readyDevice, wgpu::StringView const message) {
+                        FILAMENT_CHECK_POSTCONDITION(status == wgpu::RequestDeviceStatus::Success)
+                                << "Unable to request a WebGPU device. Status: "
+                                << static_cast<uint32_t>(status)
+                                << " with message: " << message.data;
+                        device = readyDevice;
+                    }),
+            UINT64_MAX);// TODO define reasonable timeout (or do this asynchronously)
+    FILAMENT_CHECK_POSTCONDITION(status == wgpu::WaitStatus::Success)
+            << "Non-successful wait status requesting a WebGPU device "
+            << static_cast<uint32_t>(status);
+    FILAMENT_CHECK_POSTCONDITION(device != nullptr)
+            << "Failed to get a WebGPU device for the platform.";
+    return device;
+}
+
+Driver* WebGPUPlatform::createDriver(void* sharedContext,
+        const Platform::DriverConfig& /*driverConfig*/) noexcept {
+    if (sharedContext) {
+        FWGPU_LOGW << "sharedContext is ignored/unused in the WebGPU backend. A non-null "
+                      "sharedContext was provided, but it will be ignored."
+                   << utils::io::endl;
+    }
+    return WebGPUDriver::create(*this);
+}
+
+WebGPUPlatform::WebGPUPlatform()
+    : mInstance(createInstance()) {}
+
+}// namespace filament::backend

filament/backend/src/webgpu/platform/WebGPUPlatformAndroid.cpp

@@ -0,0 +1,43 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <backend/platforms/WebGPUPlatform.h>
+
+#include <utils/Panic.h>
+
+#include <webgpu/webgpu_cpp.h>
+
+#include <cstdint>
+
+/**
+ * Android OS specific implementation aspects of the WebGPU backend
+ */
+
+namespace filament::backend {
+
+wgpu::Surface WebGPUPlatform::createSurface(void* nativeWindow, uint64_t /*flags*/) {
+    wgpu::SurfaceSourceAndroidNativeWindow surfaceSourceAndroidWindow{};
+    surfaceSourceAndroidWindow.window = nativeWindow;
+    wgpu::SurfaceDescriptor surfaceDescriptor{
+        .nextInChain = &surfaceSourceAndroidWindow,
+        .label = "android_surface"
+    };
+    wgpu::Surface surface = mInstance.CreateSurface(&surfaceDescriptor);
+    FILAMENT_CHECK_POSTCONDITION(surface != nullptr) << "Unable to create Android-backed surface.";
+    return surface;
+}
+
+}// namespace filament::backend

filament/backend/src/webgpu/platform/WebGPUPlatformApple.mm

@@ -0,0 +1,81 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <backend/platforms/WebGPUPlatform.h>
+
+#include <utils/Panic.h>
+#include <utils/ostream.h>
+
+#include <webgpu/webgpu_cpp.h>
+
+#include <cstdint>
+
+// Platform specific includes and defines
+#if defined(__APPLE__)
+    #include <Cocoa/Cocoa.h>
+    #import <QuartzCore/CAMetalLayer.h>
+#elif defined(FILAMENT_IOS)
+    // Metal is not available when building for the iOS simulator on Desktop.
+    #define METAL_AVAILABLE __has_include(<QuartzCore/CAMetalLayer.h>)
+    #if METAL_AVAILABLE
+        #import <Metal/Metal.h>
+        #import <QuartzCore/CAMetalLayer.h>
+    #endif
+    // is this needed?
+    #define METALVIEW_TAG 255
+#else
+    #error Not a supported Apple + WebGPU platform
+#endif
+
+/**
+ * Apple (Mac OS and IOS) specific implementation aspects of the WebGPU backend
+ */
+
+namespace filament::backend {
+
+wgpu::Surface WebGPUPlatform::createSurface(void* nativeWindow, uint64_t /*flags*/) {
+    wgpu::Surface surface = nullptr;
+#if defined(__APPLE__)
+    auto nsView = (__bridge NSView*) nativeWindow;
+    FILAMENT_CHECK_POSTCONDITION(nsView) << "Unable to obtain Metal-backed NSView.";
+    [nsView setWantsLayer:YES];
+    id metalLayer = [CAMetalLayer layer];
+    [nsView setLayer:metalLayer];
+    wgpu::SurfaceSourceMetalLayer surfaceSourceMetalLayer{};
+    surfaceSourceMetalLayer.layer = (__bridge void*) metalLayer;
+    wgpu::SurfaceDescriptor surfaceDescriptor = {
+        .nextInChain = &surfaceSourceMetalLayer,
+        .label = "metal_surface",
+    };
+    surface = mInstance.CreateSurface(&surfaceDescriptor);
+    FILAMENT_CHECK_POSTCONDITION(surface != nullptr) << "Unable to create Metal-backed surface.";
+#elif defined(FILAMENT_IOS)
+    CAMetalLayer* metalLayer = (CAMetalLayer*) nativeWindow;
+    wgpu::SurfaceSourceMetalLayer surfaceSourceMetalLayer{};
+    surfaceSourceMetalLayer.layer = (__bridge void*) metalLayer;
+    wgpu::SurfaceDescriptor surfaceDescriptor = {
+        .nextInChain = &surfaceSourceMetalLayer,
+        .label = "metal_surface",
+    };
+    surface = mInstance.CreateSurface(&surfaceDescriptor);
+    FILAMENT_CHECK_POSTCONDITION(surface != nullptr) << "Unable to create Metal-backed surface.";
+#else
+    #error Not a supported Apple + WebGPU platform
+#endif
+    return surface;
+}
+
+}// namespace filament::backend

filament/backend/src/webgpu/platform/WebGPUPlatformLinux.cpp

@@ -0,0 +1,163 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include <backend/platforms/WebGPUPlatform.h>
+
+#include <backend/DriverEnums.h>
+
+#include <utils/Panic.h>
+
+#include <webgpu/webgpu_cpp.h>
+
+#include <cstddef>
+#include <cstdint>
+
+#if defined(__linux__) || defined(__FreeBSD__)
+    #define LINUX_OR_FREEBSD 1
+#endif
+
+// Platform specific includes and defines
+#if defined(__linux__) && defined(FILAMENT_SUPPORTS_WAYLAND)
+    #include <dlfcn.h>
+    namespace {
+    typedef struct _wl {
+        struct wl_display* display;
+        struct wl_surface* surface;
+        uint32_t width;
+        uint32_t height;
+    } wl;
+    }// namespace
+#elif defined(LINUX_OR_FREEBSD) && defined(FILAMENT_SUPPORTS_X11)
+    // TODO: we should allow for headless on Linux explicitly. Right now this is the headless path
+    // (with no FILAMENT_SUPPORTS_XCB or FILAMENT_SUPPORTS_XLIB).
+    #include <dlfcn.h>
+    #if defined(FILAMENT_SUPPORTS_XCB)
+        #include <xcb/xcb.h>
+        namespace {
+        typedef xcb_connection_t* (*XCB_CONNECT)(const char* displayname, int* screenp);
+        }// namespace
+    #endif
+    #if defined(FILAMENT_SUPPORTS_XLIB)
+        #include <X11/Xlib.h>
+        namespace {
+        typedef Display* (*X11_OPEN_DISPLAY)(const char*);
+        }// namespace
+    #endif
+    static constexpr const char* LIBRARY_X11 = "libX11.so.6";
+    namespace {
+    struct XEnv {
+    #if defined(FILAMENT_SUPPORTS_XCB)
+        XCB_CONNECT xcbConnect;
+        xcb_connection_t* connection;
+    #endif
+    #if defined(FILAMENT_SUPPORTS_XLIB)
+        X11_OPEN_DISPLAY openDisplay;
+        Display* display;
+    #endif
+        void* library = nullptr;
+    } g_x11;
+    }// namespace
+#else
+    #error Not a supported Linux or FeeBSD + WebGPU platform
+#endif
+
+/**
+ * Linux OS specific implementation aspects of the WebGPU Backend
+ */
+
+namespace filament::backend {
+
+wgpu::Surface WebGPUPlatform::createSurface(void* nativeWindow, uint64_t flags) {
+    wgpu::Surface surface = nullptr;
+#if defined(__linux__) && defined(FILAMENT_SUPPORTS_WAYLAND)
+    wl* ptrval = reinterpret_cast<wl*>(nativeWindow);
+    wgpu::SurfaceSourceWaylandSurface surfaceSourceWayland{};
+    surfaceSourceWayland.display = ptrval->display;
+    surfaceSourceWayland.surface = ptrval->surface;
+    wgpu::SurfaceDescriptor surfaceDescriptor{
+        .nextInChain = &surfaceSourceWayland,
+        .label = "linux_wayland_surface"
+    };
+    surface = mInstance.CreateSurface(&surfaceDescriptor);
+    FILAMENT_CHECK_POSTCONDITION(surface != nullptr)
+            << "Unable to create Linux Wayland-backed surface.";
+#elif defined(LINUX_OR_FREEBSD) && defined(FILAMENT_SUPPORTS_X11)
+    if (g_x11.library == nullptr) {
+        g_x11.library = dlopen(LIBRARY_X11, RTLD_LOCAL | RTLD_NOW);
+        FILAMENT_CHECK_PRECONDITION(g_x11.library) << "Unable to open X11 library.";
+        #if defined(FILAMENT_SUPPORTS_XCB)
+            g_x11.xcbConnect = (XCB_CONNECT) dlsym(g_x11.library, "xcb_connect");
+            int screen = 0;
+            g_x11.connection = g_x11.xcbConnect(nullptr, &screen);
+        #endif
+        #if defined(FILAMENT_SUPPORTS_XLIB)
+            g_x11.openDisplay = (X11_OPEN_DISPLAY) dlsym(g_x11.library, "XOpenDisplay");
+            g_x11.display = g_x11.openDisplay(NULL);
+            FILAMENT_CHECK_PRECONDITION(g_x11.display) << "Unable to open X11 display.";
+        #endif
+    }
+    #if defined(FILAMENT_SUPPORTS_XCB) || defined(FILAMENT_SUPPORTS_XLIB)
+        bool useXcb = false;
+    #endif
+    #if defined(FILAMENT_SUPPORTS_XCB)
+        #if defined(FILAMENT_SUPPORTS_XLIB)
+            useXcb = (flags & SWAP_CHAIN_CONFIG_ENABLE_XCB) != 0;
+        #else
+            useXcb = true;
+        #endif
+        if (useXcb) {
+            wgpu::SurfaceSourceXCBWindow surfaceSourceXcb{};
+            surfaceSourceXcb.connection = g_x11.connection;
+
+            // TODO: this looks really wrong, please fix!!
+            surfaceSourceXcb.window = *((uint32_t*) nativeWindow);
+            wgpu::SurfaceDescriptor surfaceDescriptor{
+                .nextInChain = &surfaceSourceXcb,
+                .label = "linux_xcb_surface"
+            };
+            surface = mInstance.CreateSurface(&surfaceDescriptor);
+            FILAMENT_CHECK_POSTCONDITION(surface != nullptr)
+                    << "Unable to create Linux (or FreeBSD) XCB-backed surface.";
+        }
+    #endif
+    #if defined(FILAMENT_SUPPORTS_XLIB)
+        if (!useXcb) {
+            wgpu::SurfaceSourceXlibWindow surfaceSourceXlib{};
+            surfaceSourceXlib.display = g_x11.display;
+            surfaceSourceXlib.window = reinterpret_cast<uint64_t>(nativeWindow);
+            wgpu::SurfaceDescriptor surfaceDescriptor{
+                .nextInChain = &surfaceSourceXlib,
+                .label = "linux_xlib_surface"
+            };
+            surface = mInstance.CreateSurface(&surfaceDescriptor);
+            FILAMENT_CHECK_POSTCONDITION(surface != nullptr)
+                    << "Unable to create Linux (or FreeBSD) XLib-backed surface.";
+        }
+    #endif
+    FILAMENT_CHECK_POSTCONDITION(surface != nullptr)
+            << "Cannot create WebGPU X11 surface for Linux (or FreeBSD) OS "
+               "(not built with support for XCB or XLIB?)";
+#elif defined(__linux__)
+    FILAMENT_CHECK_POSTCONDITION(surface != nullptr)
+            << "Cannot create WebGPU surface for Linux (or FreeBSD) OS "
+               "(not built with support for Wayland or X11?)";
+#else
+    FILAMENT_CHECK_POSTCONDITION(surface != nullptr)
+            << "Not a supported (Linux) OS + WebGPU platform";
+#endif
+    return surface;
+}
+
+}// namespace filament::backend

filament/backend/src/webgpu/platform/WebGPUPlatformWindows.cpp

@@ -0,0 +1,53 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <backend/platforms/WebGPUPlatform.h>
+
+#include <utils/Panic.h>
+
+#include <webgpu/webgpu_cpp.h>
+
+#include <cstdint>
+
+/**
+ * Windows OS specific implementation aspects of the WebGPU backend
+ */
+
+namespace filament::backend {
+
+wgpu::Surface WebGPUPlatform::createSurface(void* nativeWindow, uint64_t /*flags*/) {
+    // TODO verify this is necessary for Dawn implementation as well:
+    // On (at least) NVIDIA drivers, the Vulkan implementation (specifically the call to
+    // vkGetPhysicalDeviceSurfaceCapabilitiesKHR()) does not correctly handle the fact that
+    // each native window has its own DPI_AWARENESS_CONTEXT, and erroneously uses the context
+    // of the calling thread. As a workaround, we set the current thread's DPI_AWARENESS_CONTEXT
+    // to that of the native window we've been given. This isn't a perfect solution, because an
+    // application could create swap chains on multiple native windows with varying DPI-awareness,
+    // but even then, at least one of the windows would be guaranteed to work correctly.
+    SetThreadDpiAwarenessContext(GetWindowDpiAwarenessContext((HWND) nativeWindow));
+    wgpu::SurfaceSourceWindowsHWND surfaceSourceWin{};
+    surfaceSourceWin.hinstance = GetModuleHandle(nullptr);
+    surfaceSourceWin.hwnd = nativeWindow;
+    wgpu::SurfaceDescriptor surfaceDescriptor{
+        .nextInChain = &surfaceSourceWin,
+        .label = "windows_surface"
+    };
+    wgpu::Surface surface = mInstance.CreateSurface(&surfaceDescriptor);
+    FILAMENT_CHECK_POSTCONDITION(surface != nullptr) << "Unable to create Windows-backed surface.";
+    return surface;
+}
+
+}// namespace filament::backend

filament/backend/test/ImageExpectations.cpp

@@ -0,0 +1,172 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "ImageExpectations.h"
+
+#include "gtest/gtest.h"
+#include "gmock/gmock.h"
+#include "absl/strings/str_format.h"
+#include "utils/Hash.h"
+#include <fstream>
+
+#include "backend/PixelBufferDescriptor.h"
+#include "private/backend/DriverApi.h"
+
+#ifndef FILAMENT_IOS
+
+#include <imageio/ImageEncoder.h>
+#include <image/ColorTransform.h>
+
+#endif
+
+ScreenshotParams::ScreenshotParams(int width, int height, std::string fileName,
+        uint32_t expectedPixelHash)
+        : mWidth(width), mHeight(height), mFileName(std::move(fileName)),
+          mExpectedPixelHash(expectedPixelHash) {}
+
+int ScreenshotParams::width() const {
+    return mWidth;
+}
+
+int ScreenshotParams::height() const {
+    return mHeight;
+}
+
+uint32_t ScreenshotParams::expectedHash() const {
+    return mExpectedPixelHash;
+}
+
+std::string ScreenshotParams::outputDirectoryPath() const {
+    return ".";
+}
+
+std::string ScreenshotParams::generatedActualFileName() const {
+    return absl::StrFormat("%s_actual.png", mFileName);
+}
+
+std::string ScreenshotParams::generatedActualFilePath() const {
+    return absl::StrFormat("%s/%s", outputDirectoryPath(), generatedActualFileName());
+}
+
+std::string ScreenshotParams::goldenFileName() const {
+    return absl::StrFormat("%s_golden.png", mFileName);
+}
+
+std::string ScreenshotParams::goldenFilePath() const {
+    return absl::StrFormat("%s/%s", outputDirectoryPath(), goldenFileName());
+}
+
+ImageExpectation::ImageExpectation(const char* fileName, int lineNumber,
+        filament::backend::DriverApi& api, ScreenshotParams params,
+        filament::backend::RenderTargetHandle renderTarget)
+        : mFileName(fileName), mLineNumber(lineNumber), mParams(std::move(params)),
+          mResult(api, renderTarget, mParams) {}
+
+void ImageExpectation::evaluate() {
+    // Ensure this is only evaluated once.
+    if (mEvaluated) {
+        return;
+    }
+    mEvaluated = true;
+
+    // Do the actual image comparison inside a scoped trace with the stored file and line.
+    {
+        testing::ScopedTrace trace(mFileName, mLineNumber, "");
+        compareImage();
+    }
+}
+
+void ImageExpectation::compareImage() const {
+    bool bytesFilled = mResult.bytesFilled();
+    EXPECT_THAT(bytesFilled, testing::IsTrue())
+                        << "Render target wasn't copied to the buffer for " << mFileName;
+    if (bytesFilled) {
+        uint32_t actualHash = mResult.hash();
+        EXPECT_THAT(actualHash, testing::Eq(mParams.expectedHash()));
+    }
+}
+
+ImageExpectations::ImageExpectations(filament::backend::DriverApi& api) : mApi(api) {}
+
+ImageExpectations::~ImageExpectations() {
+    // Guarantee that all expectations are evaluated when this leaves scope even if the caller
+    // forgot to manually evaluate them.
+    evaluate();
+}
+
+void ImageExpectations::addExpectation(const char* fileName, int lineNumber,
+        filament::backend::RenderTargetHandle renderTarget, ScreenshotParams params) {
+    mExpectations.emplace_back(fileName, lineNumber, mApi, std::move(params), renderTarget);
+}
+
+void ImageExpectations::evaluate() {
+    for (auto& expectation: mExpectations) {
+        expectation.evaluate();
+    }
+    mExpectations.clear();
+}
+
+RenderTargetDump::RenderTargetDump(filament::backend::DriverApi& api,
+        filament::backend::RenderTargetHandle renderTarget, const ScreenshotParams& params)
+        : mInternal(std::make_unique<RenderTargetDump::Internal>(params)) {
+#ifdef FILAMENT_IOS
+    bytesFilled_ = true;
+    bytes_.resize(size);
+    std::fill(bytes_.begin(), bytes_.end(), 0);
+#else
+    const size_t size = mInternal->params.width() * mInternal->params.height() * 4;
+    mInternal->bytes.resize(size);
+
+    auto cb = [](void* buffer, size_t size, void* user) {
+        auto* internal = static_cast<RenderTargetDump::Internal*>(user);
+        image::LinearImage image(internal->params.width(), internal->params.width(), 4);
+        image = image::toLinearWithAlpha<uint8_t>(internal->params.width(),
+                internal->params.height(),
+                internal->params.width() * 4, (uint8_t*)buffer);
+        std::string filePath = internal->params.generatedActualFilePath();
+        std::ofstream pngStream(filePath, std::ios::binary | std::ios::trunc);
+        image::ImageEncoder::encode(pngStream, image::ImageEncoder::Format::PNG, image, "",
+                filePath);
+        internal->bytesFilled = true;
+    };
+    filament::backend::PixelBufferDescriptor pb(mInternal->bytes.data(), size,
+            filament::backend::PixelDataFormat::RGBA, filament::backend::PixelDataType::UBYTE, cb,
+            (void*)mInternal.get());
+    api.readPixels(renderTarget, 0, 0, mInternal->params.width(), mInternal->params.height(),
+            std::move(pb));
+#endif
+}
+
+RenderTargetDump::~RenderTargetDump() {
+    // If the GPU callback hasn't been made yet then there's a callback elsewhere that has a copy of
+    // the internal pointer. But there's no guarantee that the callback will be ever made if the GPU
+    // pipeline wasn't run for some reason. So it is necessary to leak the memory.
+    // It would be possible to try to coordinate with the callback to have it clean up the memory,
+    // but if this condition happens there's already an issue with the test case so there's no need.
+    if (!bytesFilled()) {
+        mInternal.release();
+    }
+}
+
+uint32_t RenderTargetDump::hash() const {
+    return utils::hash::murmur3((uint32_t*)mInternal->bytes.data(), mInternal->bytes.size() / 4, 0);
+}
+
+bool RenderTargetDump::bytesFilled() const {
+    return mInternal->bytesFilled;
+}
+
+RenderTargetDump::Internal::Internal(const ScreenshotParams& params) : params(params) {}

filament/backend/test/ImageExpectations.h

@@ -0,0 +1,136 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_IMAGE_EXPECTATIONS_H
+#define TNT_IMAGE_EXPECTATIONS_H
+
+#include <vector>
+
+#include "gtest/gtest.h"
+
+#include "backend/Handle.h"
+#include "backend/DriverApiForward.h"
+
+// Arguments are (RenderTargetHandle renderTarget, ImageExpectations& expectations,
+// ScreenshotParams screenshotParams)
+#define EXPECT_IMAGE(renderTarget, expectations, screenshotParams)  \
+do {                                                                \
+    expectations.addExpectation(                                    \
+        __FILE__,                                                   \
+        __LINE__,                                                   \
+        renderTarget,                                               \
+        screenshotParams);                                          \
+} while (0)
+
+/**
+ * Stores user-provided configuration values for an image expectation
+ */
+class ScreenshotParams {
+public:
+    ScreenshotParams(int width, int height, std::string fileName, uint32_t expectedPixelHash);
+
+    int width() const;
+    int height() const;
+    uint32_t expectedHash() const;
+
+    std::string outputDirectoryPath() const;
+    std::string generatedActualFileName() const;
+    std::string generatedActualFilePath() const;
+    std::string goldenFileName() const;
+    std::string goldenFilePath() const;
+
+private:
+    int mWidth;
+    int mHeight;
+    uint32_t mExpectedPixelHash;
+    std::string mFileName;
+};
+
+/**
+ * When created adds a command to the GPU pipeline to copy the render target into a buffer and
+ * stores the result.
+ * If this object is destroyed before the GPU pipeline is flushed it will leak memory in order to
+ * avoid the GPU pipeline callback being a use-after-free.
+ */
+class RenderTargetDump {
+public:
+    RenderTargetDump(filament::backend::DriverApi& api,
+            filament::backend::RenderTargetHandle renderTarget, const ScreenshotParams& params);
+    RenderTargetDump(RenderTargetDump&& other) = default;
+    RenderTargetDump& operator=(RenderTargetDump&& other) = default;
+    ~RenderTargetDump();
+
+    /**
+     * Should only bue used if BytesFilled returns true.
+     * @return The hash of the stored bytes.
+     */
+    uint32_t hash() const;
+    /**
+     * Thread safe as this is backed by an atomic.
+     * Once this returns true it will never return false.
+     * @return Whether the bytes have actually been copied from the GPU to the buffer.
+     */
+    bool bytesFilled() const;
+
+private:
+    struct Internal {
+        explicit Internal(const ScreenshotParams& params);
+        ScreenshotParams params;
+        std::atomic<bool> bytesFilled = false;
+        std::vector<unsigned char> bytes;
+    };
+
+    // We need a memory location that won't be invalidated to pass to GPU callbacks as they can't
+    // be canceled during the destructor.
+    std::unique_ptr<Internal> mInternal;
+};
+
+class ImageExpectation {
+public:
+    ImageExpectation(const char* fileName, int lineNumber, filament::backend::DriverApi& api,
+            ScreenshotParams params, filament::backend::RenderTargetHandle renderTarget);
+
+    void evaluate();
+
+private:
+    void compareImage() const;
+
+    bool mEvaluated = false;
+    const char* mFileName;
+    int mLineNumber;
+    ScreenshotParams mParams;
+    RenderTargetDump mResult;
+};
+
+class ImageExpectations {
+public:
+    explicit ImageExpectations(filament::backend::DriverApi& api);
+    ~ImageExpectations();
+
+    /**
+     * Not meant to be called directly, use EXPECT_IMAGE to get the file name and line number
+     */
+    void addExpectation(const char* fileName, int lineNumber,
+            filament::backend::RenderTargetHandle renderTarget, ScreenshotParams params);
+
+    void evaluate();
+
+private:
+    filament::backend::DriverApi& mApi;
+    std::vector<ImageExpectation> mExpectations;
+};
+
+#endif //TNT_IMAGE_EXPECTATIONS_H

filament/backend/test/Lifetimes.cpp

@@ -0,0 +1,92 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "Lifetimes.h"
+
+#include "backend/Platform.h"
+#include "private/backend/DriverApi.h"
+
+RenderFrame::RenderFrame(filament::backend::DriverApi& api): mApi(api) {
+    mApi.beginFrame(0, 0, 0);
+}
+
+RenderFrame::~RenderFrame() {
+    mApi.endFrame(0);
+}
+
+Cleanup::Cleanup(filament::backend::DriverApi& driverApi) : mDriverApi(driverApi) {}
+
+Cleanup::~Cleanup() {
+    for (const auto swapChain: mSwapChains) {
+        mDriverApi.destroySwapChain(swapChain);
+    }
+    for (const auto descriptorSet : mDescriptorSets) {
+        mDriverApi.destroyDescriptorSet(descriptorSet);
+    }
+    for (const auto descriptorSetLayout : mDescriptorSetLayouts) {
+        mDriverApi.destroyDescriptorSetLayout(descriptorSetLayout);
+    }
+    for (const auto bufferObject : mBufferObjects) {
+        mDriverApi.destroyBufferObject(bufferObject);
+    }
+    for (const auto program : mPrograms) {
+        mDriverApi.destroyProgram(program);
+    }
+    for (const auto texture : mTextures) {
+        mDriverApi.destroyTexture(texture);
+    }
+    for (const auto renderTarget : mRenderTargets) {
+        mDriverApi.destroyRenderTarget(renderTarget);
+    }
+
+    for (const auto& callback : mCallbacks) {
+        callback();
+    }
+}
+
+template <>
+void Cleanup::addInternal(filament::backend::SwapChainHandle handle) {
+    mSwapChains.push_back(handle);
+}
+template <>
+void Cleanup::addInternal(filament::backend::DescriptorSetHandle handle) {
+    mDescriptorSets.push_back(handle);
+}
+template <>
+void Cleanup::addInternal(
+        filament::backend::DescriptorSetLayoutHandle handle) {
+    mDescriptorSetLayouts.push_back(handle);
+}
+template <>
+void Cleanup::addInternal(filament::backend::BufferObjectHandle handle) {
+    mBufferObjects.push_back(handle);
+}
+template <>
+void Cleanup::addInternal(filament::backend::ProgramHandle handle) {
+    mPrograms.push_back(handle);
+}
+template <>
+void Cleanup::addInternal(filament::backend::TextureHandle handle) {
+    mTextures.push_back(handle);
+}
+template <>
+void Cleanup::addInternal(filament::backend::RenderTargetHandle handle) {
+    mRenderTargets.push_back(handle);
+}
+
+void Cleanup::addPostCall(std::function<void()> callback){
+    mCallbacks.emplace_back(std::move(callback));
+}

filament/backend/test/Lifetimes.h

@@ -0,0 +1,78 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_LIFETIMES_H
+#define TNT_LIFETIMES_H
+
+#include <vector>
+#include <functional>
+#include <optional>
+
+#include "backend/Handle.h"
+#include "backend/DriverApiForward.h"
+#include "BackendTest.h"
+
+class RenderFrame {
+public:
+    explicit RenderFrame(filament::backend::DriverApi& api);
+    ~RenderFrame();
+
+    RenderFrame(RenderFrame const& rhs) noexcept = delete;
+    RenderFrame& operator=(RenderFrame const& rhs) noexcept = delete;
+    RenderFrame(RenderFrame const&& rhs) noexcept = delete;
+    RenderFrame& operator=(RenderFrame const&& rhs) noexcept = delete;
+
+private:
+    filament::backend::DriverApi& mApi;
+};
+
+class Cleanup {
+public:
+    /**
+     * @param driverApi Must be valid for the entire lifetime of the Cleanup object.
+     */
+    explicit Cleanup(filament::backend::DriverApi& driverApi);
+    ~Cleanup();
+
+    template <typename HandleType>
+    filament::backend::Handle<HandleType> add(filament::backend::Handle<HandleType> handle);
+
+    void addPostCall(std::function<void()> callback);
+
+private:
+    template <typename HandleType>
+    void addInternal(filament::backend::Handle<HandleType> handle);
+
+    filament::backend::DriverApi & mDriverApi;
+
+    std::vector<filament::backend::SwapChainHandle> mSwapChains;
+    std::vector<filament::backend::DescriptorSetHandle> mDescriptorSets;
+    std::vector<filament::backend::DescriptorSetLayoutHandle> mDescriptorSetLayouts;
+    std::vector<filament::backend::BufferObjectHandle> mBufferObjects;
+    std::vector<filament::backend::ProgramHandle> mPrograms;
+    std::vector<filament::backend::TextureHandle> mTextures;
+    std::vector<filament::backend::RenderTargetHandle> mRenderTargets;
+
+    std::vector<std::function<void()>> mCallbacks;
+};
+
+template <typename HandleType>
+filament::backend::Handle<HandleType> Cleanup::add(filament::backend::Handle<HandleType> handle) {
+    addInternal(handle);
+    return handle;
+}
+
+#endif //TNT_LIFETIMES_H

filament/backend/test/PlatformRunner.h

@@ -30,7 +30,8 @@ enum class Backend : uint8_t {
     OPENGL = 1,
     VULKAN = 2,
     METAL = 3,
-    NOOP = 4,
+    WEBGPU = 4,
+    NOOP = 5,
 };
 
 struct NativeView {

filament/backend/test/Shader.cpp

@@ -0,0 +1,64 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "Shader.h"
+
+#include "BackendTest.h"
+#include "ShaderGenerator.h"
+
+namespace test {
+
+using namespace filament::backend;
+
+Shader::Shader(DriverApi& api, Cleanup& cleanup, ShaderConfig config) {
+    utils::FixedCapacityVector<DescriptorSetLayoutBinding> kLayouts(config.uniformNames.size());
+    for (unsigned char i = 0; i < config.uniformNames.size(); ++i) {
+        kLayouts[i] =
+                { DescriptorType::UNIFORM_BUFFER, ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, i };
+    };
+
+    filamat::DescriptorSets descriptors;
+    for (unsigned char i = 0; i < config.uniformNames.size(); ++i) {
+        descriptors[i + 1] = {{ config.uniformNames[i], kLayouts[i], {}}};
+    }
+    ShaderGenerator shaderGen(
+            std::move(config.vertexShader), std::move(config.fragmentShader), BackendTest::sBackend,
+            BackendTest::sIsMobilePlatform, std::move(descriptors));
+    Program prog = shaderGen.getProgram(api);
+    for (unsigned char i = 0; i < config.uniformNames.size(); ++i) {
+        prog.descriptorBindings(1, {{ config.uniformNames[i], DescriptorType::UNIFORM_BUFFER, i }});
+    }
+    mProgram = cleanup.add(api.createProgram(std::move(prog)));
+
+    mDescriptorSetLayout = cleanup.add(
+            api.createDescriptorSetLayout(DescriptorSetLayout{ kLayouts }));
+
+    mDescriptorSet = cleanup.add(api.createDescriptorSet(mDescriptorSetLayout));
+}
+
+filament::backend::ProgramHandle Shader::getProgram() const {
+    return mProgram;
+}
+
+filament::backend::DescriptorSetLayoutHandle Shader::getDescriptorSetLayout() const {
+    return mDescriptorSetLayout;
+}
+
+filament::backend::DescriptorSetHandle Shader::getDescriptorSet() const {
+    return mDescriptorSet;
+}
+
+} // namespace test

filament/backend/test/Shader.h

@@ -0,0 +1,143 @@
+/*
+ * Copyright (C) 2025 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_SHADER_H
+#define TNT_SHADER_H
+
+#include "Lifetimes.h"
+
+namespace test {
+
+// All describing a shader that should be created.
+struct ShaderConfig {
+    std::string vertexShader;
+    std::string fragmentShader;
+    std::vector<utils::CString> uniformNames;
+};
+
+// All values describing a uniform.
+struct ResolvedUniformBindingConfig {
+    size_t dataSize;
+    size_t bufferSize;
+    uint32_t byteOffset;
+    filament::backend::descriptor_set_t set;
+    filament::backend::descriptor_binding_t binding;
+};
+
+// An equivalent to ResolvedUniformBindingConfig with all fields optional.
+// resolve can be called to create a ResolvedBindingConfig with default values
+// use in place of nullopt values.
+struct UniformBindingConfig {
+    std::optional<size_t> dataSize;
+    std::optional<size_t> bufferSize;
+    std::optional<uint32_t> byteOffset;
+    std::optional<filament::backend::descriptor_set_t> set;
+    std::optional<filament::backend::descriptor_binding_t> binding;
+
+    template<typename UniformType>
+    ResolvedUniformBindingConfig resolve();
+};
+
+class Shader {
+public:
+    // All graphics resources have their lifetime controlled by the Cleanup and not this object.
+    // Shader must not outlive either the DriverApi or the Cleanup.
+    Shader(filament::backend::DriverApi& api, Cleanup& cleanup, ShaderConfig config);
+
+    // Uploads a uniform with the default config.
+    template<typename UniformType>
+    void uploadUniform(filament::backend::DriverApi& api,
+            filament::backend::Handle<filament::backend::HwBufferObject> hwBuffer,
+            UniformType uniforms) const;
+    // Binds a uniform with the default config.
+    template<typename UniformType>
+    void bindUniform(filament::backend::DriverApi& api,
+            filament::backend::Handle<filament::backend::HwBufferObject> hwBuffer) const;
+    // Uploads a uniform, with UniformBindingConfig providing the arguments.
+    template<typename UniformType>
+    void uploadUniform(filament::backend::DriverApi& api,
+            filament::backend::Handle<filament::backend::HwBufferObject> hwBuffer,
+            UniformBindingConfig config, UniformType uniforms) const;
+    // Binds a uniform, with UniformBindingConfig providing the arguments.
+    template<typename UniformType>
+    void bindUniform(filament::backend::DriverApi& api,
+            filament::backend::Handle<filament::backend::HwBufferObject> hwBuffer,
+            UniformBindingConfig config) const;
+
+    filament::backend::ProgramHandle getProgram() const;
+    filament::backend::DescriptorSetLayoutHandle getDescriptorSetLayout() const;
+    filament::backend::DescriptorSetHandle getDescriptorSet() const;
+
+protected:
+    filament::backend::ProgramHandle mProgram;
+    filament::backend::DescriptorSetLayoutHandle mDescriptorSetLayout;
+    filament::backend::DescriptorSetHandle mDescriptorSet;
+};
+
+template<typename UniformType>
+ResolvedUniformBindingConfig UniformBindingConfig::resolve() {
+    auto resolvedDataSize = dataSize.value_or(sizeof(UniformType));
+    return ResolvedUniformBindingConfig{
+            .dataSize = resolvedDataSize,
+            .bufferSize = bufferSize.value_or(resolvedDataSize),
+            .byteOffset = byteOffset.value_or(0),
+            .set = set.value_or(1),
+            .binding = binding.value_or(0)
+    };
+}
+
+template<typename UniformType>
+void Shader::uploadUniform(filament::backend::DriverApi& api,
+        filament::backend::Handle<filament::backend::HwBufferObject> hwBuffer,
+        UniformBindingConfig config, UniformType uniforms) const {
+    auto resolvedConfig = config.resolve<UniformType>();
+
+    UniformType* tmp = new UniformType(uniforms);
+    auto cb = [](void* buffer, size_t size, void* user) {
+        UniformType* sp = (UniformType*)buffer;
+        delete sp;
+    };
+    filament::backend::BufferDescriptor bd(tmp, resolvedConfig.dataSize, cb);
+    api.updateBufferObject(hwBuffer, std::move(bd), resolvedConfig.byteOffset);
+}
+
+template<typename UniformType>
+void Shader::bindUniform(filament::backend::DriverApi& api,
+        filament::backend::Handle<filament::backend::HwBufferObject> hwBuffer,
+        UniformBindingConfig config) const {
+    auto resolvedConfig = config.resolve<UniformType>();
+
+    api.updateDescriptorSetBuffer(getDescriptorSet(), resolvedConfig.binding, hwBuffer, 0,
+            resolvedConfig.bufferSize);
+    api.bindDescriptorSet(getDescriptorSet(), resolvedConfig.set, {});
+}
+
+template<typename UniformType>
+void Shader::uploadUniform(filament::backend::DriverApi& api,
+        filament::backend::Handle<filament::backend::HwBufferObject> hwBuffer,
+        UniformType uniforms) const {
+    uploadUniform(api, hwBuffer, UniformBindingConfig{}, uniforms);
+}
+
+template<typename UniformType>
+void Shader::bindUniform(filament::backend::DriverApi& api,
+        filament::backend::Handle<filament::backend::HwBufferObject> hwBuffer) const {
+    bindUniform<UniformType>(api, hwBuffer, UniformBindingConfig{});
+}
+
+} // namespace test
+
+#endif //TNT_SHADER_H

filament/backend/test/ShaderGenerator.cpp

@@ -90,6 +90,9 @@ ShaderGenerator::ShaderGenerator(std::string vertex, std::string fragment, Backe
         case Backend::METAL:
             mShaderLanguage = filament::backend::ShaderLanguage::MSL;
             break;
+        case Backend::WEBGPU:
+            mShaderLanguage = filament::backend::ShaderLanguage::WGSL;
+            break;
         case Backend::NOOP:
             mShaderLanguage = filament::backend::ShaderLanguage::ESSL3;
             break;
@@ -115,6 +118,8 @@ ShaderGenerator::Blob ShaderGenerator::transpileShader(ShaderStage stage, std::s
         shader.insert(pos, "#define TARGET_METAL_ENVIRONMENT\n");
     } else if (backend == Backend::VULKAN) {
         shader.insert(pos, "#define TARGET_VULKAN_ENVIRONMENT\n");
+    } else if (backend == Backend::WEBGPU) {
+        shader.insert(pos, "#define TARGET_VULKAN_ENVIRONMENT\n");
     }
 
     const char* shaderCString = shader.c_str();
@@ -150,7 +155,8 @@ ShaderGenerator::Blob ShaderGenerator::transpileShader(ShaderStage stage, std::s
 
     assert_invariant(backend == Backend::OPENGL ||
            backend == Backend::METAL  ||
-           backend == Backend::VULKAN);
+           backend == Backend::VULKAN ||
+           backend == Backend::WEBGPU);
 
     if (backend == Backend::OPENGL) {
         if (isMobile) {
@@ -166,6 +172,8 @@ ShaderGenerator::Blob ShaderGenerator::transpileShader(ShaderStage stage, std::s
         return { result.c_str(), result.c_str() + result.length() + 1 };
     } else if (backend == Backend::VULKAN) {
         return { (uint8_t*)spirv.data(), (uint8_t*)(spirv.data() + spirv.size()) };
+    } else if (backend == Backend::WEBGPU){
+        filamat::GLSLPostProcessor::spirvToWgsl(&spirv, &result);
     }
 
     return {};

filament/backend/test/test_Blit.cpp

@@ -16,7 +16,9 @@
 
 #include "BackendTest.h"
 
-#include "ShaderGenerator.h"
+#include "ImageExpectations.h"
+#include "Lifetimes.h"
+#include "Shader.h"
 #include "TrianglePrimitive.h"
 
 #include <utils/Hash.h>
@@ -25,8 +27,10 @@
 #include <fstream>
 
 #ifndef FILAMENT_IOS
+
 #include <imageio/ImageEncoder.h>
 #include <image/ColorTransform.h>
+
 #endif
 
 namespace test {
@@ -36,6 +40,21 @@ using namespace filament::backend;
 using namespace filament::math;
 using namespace utils;
 
+struct MaterialParams {
+    float4 color;
+    float4 scale;
+};
+
+class BlitTest : public BackendTest {
+public:
+    BlitTest() : mCleanup(getDriverApi()) {}
+
+protected:
+    Shader createShader();
+
+    Cleanup mCleanup;
+};
+
 static const char* const triangleVs = R"(#version 450 core
 layout(location = 0) in vec4 mesh_position;
 layout(binding = 0, set = 1) uniform Params { highp vec4 color; highp vec4 scale; } params;
@@ -55,49 +74,12 @@ void main() {
     fragColor = params.color;
 })";
 
-struct MaterialParams {
-    float4 color;
-    float4 scale;
-};
-
-struct ScreenshotParams {
-    int width;
-    int height;
-    const char* filename;
-    uint32_t pixelHashResult;
-};
-
-#ifdef FILAMENT_IOS
-static void dumpScreenshot(DriverApi& dapi, Handle<HwRenderTarget> rt, ScreenshotParams* params) {}
-#else
-static void dumpScreenshot(DriverApi& dapi, Handle<HwRenderTarget> rt, ScreenshotParams* params) {
-    using namespace image;
-    const size_t size = params->width * params->height * 4;
-    void* buffer = calloc(1, size);
-    auto cb = [](void* buffer, size_t size, void* user) {
-        ScreenshotParams* params = (ScreenshotParams*) user;
-        int w = params->width, h = params->height;
-        const uint32_t* texels = (uint32_t*) buffer;
-        params->pixelHashResult = utils::hash::murmur3(texels, size / 4, 0);
-        LinearImage image(w, h, 4);
-        image = toLinearWithAlpha<uint8_t>(w, h, w * 4, (uint8_t*) buffer);
-        std::ofstream pngstrm(params->filename, std::ios::binary | std::ios::trunc);
-        ImageEncoder::encode(pngstrm, ImageEncoder::Format::PNG, image, "", params->filename);
-    };
-    PixelBufferDescriptor pb(buffer, size, PixelDataFormat::RGBA, PixelDataType::UBYTE, cb,
-            (void*) params);
-    dapi.readPixels(rt, 0, 0, params->width, params->height, std::move(pb));
-}
-#endif
-
-static void uploadUniforms(DriverApi& dapi, Handle<HwBufferObject> ubh, MaterialParams params) {
-    MaterialParams* tmp = new MaterialParams(params);
-    auto cb = [](void* buffer, size_t size, void* user) {
-        MaterialParams* sp = (MaterialParams*) buffer;
-        delete sp;
-    };
-    BufferDescriptor bd(tmp, sizeof(MaterialParams), cb);
-    dapi.updateBufferObject(ubh, std::move(bd), 0);
+Shader BlitTest::createShader() {
+    return Shader(getDriverApi(), mCleanup, ShaderConfig{
+            .vertexShader = triangleVs,
+            .fragmentShader = triangleFs,
+            .uniformNames = { "Params" },
+    });
 }
 
 static uint32_t toUintColor(float4 color) {
@@ -115,7 +97,7 @@ static void createBitmap(DriverApi& dapi, Handle<HwTexture> texture, int baseWid
     const int width = baseWidth >> level;
     const int height = baseHeight >> level;
     const size_t size0 = height * width * 4;
-    uint8_t* buffer0 = (uint8_t*) calloc(size0, 1);
+    uint8_t* buffer0 = (uint8_t*)calloc(size0, 1);
     PixelBufferDescriptor pb(buffer0, size0, PixelDataFormat::RGBA, PixelDataType::UBYTE, cb);
 
     const float3 foreground = color;
@@ -139,7 +121,7 @@ static void createBitmap(DriverApi& dapi, Handle<HwTexture> texture, int baseWid
     // |   ........
     // low addresses
     // This is because OpenGL automatically flips image data when uploading into the texture.
-    uint32_t* texels = (uint32_t*) buffer0;
+    uint32_t* texels = (uint32_t*)buffer0;
     for (int row = 0; row < height; row++) {
         for (int col = 0; col < width; col++) {
             float2 uv = { (col - width / 2.0f) / width, (row - height / 2.0f) / height };
@@ -167,7 +149,7 @@ static void createFaces(DriverApi& dapi, Handle<HwTexture> texture, int baseWidt
     const int width = baseWidth >> level;
     const int height = baseHeight >> level;
     const size_t size0 = height * width * 4 * 6;
-    uint8_t* buffer0 = (uint8_t*) calloc(size0, 1);
+    uint8_t* buffer0 = (uint8_t*)calloc(size0, 1);
     PixelBufferDescriptor pb(buffer0, size0, PixelDataFormat::RGBA, PixelDataType::UBYTE, cb);
 
     const float3 foreground = color;
@@ -175,7 +157,7 @@ static void createFaces(DriverApi& dapi, Handle<HwTexture> texture, int baseWidt
     const float radius = 0.25f;
 
     // Draw a circle on a yellow background.
-    uint32_t* texels = (uint32_t*) buffer0;
+    uint32_t* texels = (uint32_t*)buffer0;
     for (int face = 0; face < 6; face++) {
         for (int row = 0; row < height; row++) {
             for (int col = 0; col < width; col++) {
@@ -193,8 +175,9 @@ static void createFaces(DriverApi& dapi, Handle<HwTexture> texture, int baseWidt
     dapi.update3DImage(texture, level, 0, 0, 0, width, height, 6, std::move(pb));
 }
 
-TEST_F(BackendTest, ColorMagnify) {
+TEST_F(BlitTest, ColorMagnify) {
     auto& api = getDriverApi();
+    mCleanup.addPostCall([&]() { executeCommands(); });
 
     constexpr int kSrcTexWidth = 256;
     constexpr int kSrcTexHeight = 256;
@@ -205,71 +188,64 @@ TEST_F(BackendTest, ColorMagnify) {
     constexpr int kNumLevels = 3;
 
     // Create a SwapChain and make it current. We don't really use it so the res doesn't matter.
-    auto swapChain = api.createSwapChainHeadless(256, 256, 0);
+    auto swapChain = mCleanup.add(api.createSwapChainHeadless(256, 256, 0));
     api.makeCurrent(swapChain, swapChain);
 
     // Create a source texture.
-    Handle<HwTexture> const srcTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> const srcTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
+            TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT));
     const bool flipY = sBackend == Backend::OPENGL;
     createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 0, float3(0.5, 0, 0), flipY);
     createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 1, float3(0, 0, 0.5), flipY);
 
     // Create a destination texture.
-    Handle<HwTexture> const dstTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kDstTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> const dstTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, kNumLevels, kDstTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
+            TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT));
 
     // Create a RenderTarget for each texture's miplevel.
     Handle<HwRenderTarget> srcRenderTargets[kNumLevels];
     Handle<HwRenderTarget> dstRenderTargets[kNumLevels];
     for (uint8_t level = 0; level < kNumLevels; level++) {
-        srcRenderTargets[level] = api.createRenderTarget( TargetBufferFlags::COLOR,
-                kSrcTexWidth >> level, kSrcTexHeight >> level, 1, 0, { srcTexture, level, 0 }, {}, {});
-        dstRenderTargets[level] = api.createRenderTarget( TargetBufferFlags::COLOR,
-                kDstTexWidth >> level, kDstTexHeight >> level, 1, 0, { dstTexture, level, 0 }, {}, {});
+        srcRenderTargets[level] = mCleanup.add(api.createRenderTarget(TargetBufferFlags::COLOR,
+                kSrcTexWidth >> level, kSrcTexHeight >> level, 1, 0, { srcTexture, level, 0 }, {},
+                {}));
+        dstRenderTargets[level] = mCleanup.add(api.createRenderTarget(TargetBufferFlags::COLOR,
+                kDstTexWidth >> level, kDstTexHeight >> level, 1, 0, { dstTexture, level, 0 }, {},
+                {}));
     }
 
     // Do a "magnify" blit from level 1 of the source RT to the level 0 of the destination RT.
     const int srcLevel = 1;
     api.blitDEPRECATED(TargetBufferFlags::COLOR0, dstRenderTargets[0],
-            {0, 0, kDstTexWidth, kDstTexHeight}, srcRenderTargets[srcLevel],
-            {0, 0, kSrcTexWidth >> srcLevel, kSrcTexHeight >> srcLevel}, SamplerMagFilter::LINEAR);
+            { 0, 0, kDstTexWidth, kDstTexHeight }, srcRenderTargets[srcLevel],
+            { 0, 0, kSrcTexWidth >> srcLevel, kSrcTexHeight >> srcLevel },
+            SamplerMagFilter::LINEAR);
 
     // Push through an empty frame to allow the texture to upload and the blit to execute.
-    api.beginFrame(0, 0, 0);
-    api.commit(swapChain);
-    api.endFrame(0);
+    {
+        RenderFrame frame(api);
+        api.commit(swapChain);
+    }
 
-    // Grab a screenshot.
-    ScreenshotParams params { kDstTexWidth, kDstTexHeight, "ColorMagnify.png" };
-    api.beginFrame(0, 0, 0);
-    dumpScreenshot(api, dstRenderTargets[0], &params);
-    api.commit(swapChain);
-    api.endFrame(0);
+    {
+        ImageExpectations expectations(api);
 
-    // Wait for the ReadPixels result to come back.
-    api.finish();
-    executeCommands();
-    getDriver().purge();
+        {
+            RenderFrame frame(api);
+            EXPECT_IMAGE(dstRenderTargets[0], expectations,
+                    ScreenshotParams(kDstTexWidth, kDstTexHeight, "ColorMagnify", 0x410bdd31));
+            api.commit(swapChain);
+        }
 
-    // Check if the image matches perfectly to our golden run.
-    const uint32_t expected = 0x410bdd31;
-    printf("Computed hash is 0x%8.8x, Expected 0x%8.8x\n", params.pixelHashResult, expected);
-    EXPECT_TRUE(params.pixelHashResult == expected);
-
-    // Cleanup.
-    api.destroyTexture(srcTexture);
-    api.destroyTexture(dstTexture);
-    api.destroySwapChain(swapChain);
-    for (auto rt : srcRenderTargets)  api.destroyRenderTarget(rt);
-    for (auto rt : dstRenderTargets)  api.destroyRenderTarget(rt);
-    executeCommands();
+        flushAndWait();
+    }
 }
 
-TEST_F(BackendTest, ColorMinify) {
+TEST_F(BlitTest, ColorMinify) {
     auto& api = getDriverApi();
+    mCleanup.addPostCall([&]() { executeCommands(); });
 
     constexpr int kSrcTexWidth = 1024;
     constexpr int kSrcTexHeight = 1024;
@@ -280,62 +256,54 @@ TEST_F(BackendTest, ColorMinify) {
     constexpr int kNumLevels = 3;
 
     // Create a SwapChain and make it current. We don't really use it so the res doesn't matter.
-    auto swapChain = api.createSwapChainHeadless(256, 256, 0);
+    auto swapChain = mCleanup.add(api.createSwapChainHeadless(256, 256, 0));
     api.makeCurrent(swapChain, swapChain);
 
     // Create a source texture.
-    Handle<HwTexture> const srcTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> const srcTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
+            TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT));
     const bool flipY = sBackend == Backend::OPENGL;
     createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 0, float3(0.5, 0, 0), flipY);
     createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 1, float3(0, 0, 0.5), flipY);
 
     // Create a destination texture.
-    Handle<HwTexture> const dstTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kDstTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> const dstTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, kNumLevels, kDstTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
+            TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT));
 
     // Create a RenderTarget for each texture's miplevel.
     Handle<HwRenderTarget> srcRenderTargets[kNumLevels];
     Handle<HwRenderTarget> dstRenderTargets[kNumLevels];
     for (uint8_t level = 0; level < kNumLevels; level++) {
-        srcRenderTargets[level] = api.createRenderTarget( TargetBufferFlags::COLOR,
-                kSrcTexWidth >> level, kSrcTexHeight >> level, 1, 0, { srcTexture, level, 0 }, {}, {});
-        dstRenderTargets[level] = api.createRenderTarget( TargetBufferFlags::COLOR,
-                kDstTexWidth >> level, kDstTexHeight >> level, 1, 0, { dstTexture, level, 0 }, {}, {});
+        srcRenderTargets[level] = mCleanup.add(api.createRenderTarget(TargetBufferFlags::COLOR,
+                kSrcTexWidth >> level, kSrcTexHeight >> level, 1, 0, { srcTexture, level, 0 }, {},
+                {}));
+        dstRenderTargets[level] = mCleanup.add(api.createRenderTarget(TargetBufferFlags::COLOR,
+                kDstTexWidth >> level, kDstTexHeight >> level, 1, 0, { dstTexture, level, 0 }, {},
+                {}));
     }
 
     // Do a "minify" blit from level 1 of the source RT to the level 0 of the destination RT.
     const int srcLevel = 1;
 
     api.blitDEPRECATED(TargetBufferFlags::COLOR0,
-            dstRenderTargets[0], {0, 0, kDstTexWidth, kDstTexHeight},
-            srcRenderTargets[srcLevel], {0, 0, kSrcTexWidth >> srcLevel, kSrcTexHeight >> srcLevel},
+            dstRenderTargets[0], { 0, 0, kDstTexWidth, kDstTexHeight },
+            srcRenderTargets[srcLevel],
+            { 0, 0, kSrcTexWidth >> srcLevel, kSrcTexHeight >> srcLevel },
             SamplerMagFilter::LINEAR);
 
-    // Grab a screenshot.
-    ScreenshotParams params { kDstTexWidth, kDstTexHeight, "ColorMinify.png" };
-    dumpScreenshot(api, dstRenderTargets[0], &params);
+    {
+        ImageExpectations expectations(api);
 
-    // Wait for the ReadPixels result to come back.
-    flushAndWait();
+        EXPECT_IMAGE(dstRenderTargets[0], expectations,
+                ScreenshotParams(kDstTexWidth, kDstTexHeight, "ColorMinify", 0xf3d9c53f));
 
-    // Check if the image matches perfectly to our golden run.
-    const uint32_t expected = 0xf3d9c53f;
-    printf("Computed hash is 0x%8.8x, Expected 0x%8.8x\n", params.pixelHashResult, expected);
-    EXPECT_TRUE(params.pixelHashResult == expected);
-
-    // Cleanup.
-    api.destroyTexture(srcTexture);
-    api.destroyTexture(dstTexture);
-    api.destroySwapChain(swapChain);
-    for (auto rt : srcRenderTargets)  api.destroyRenderTarget(rt);
-    for (auto rt : dstRenderTargets)  api.destroyRenderTarget(rt);
-    executeCommands();
+        flushAndWait();
+    }
 }
 
-TEST_F(BackendTest, ColorResolve) {
+TEST_F(BlitTest, ColorResolve) {
     auto& api = getDriverApi();
 
     constexpr int kSrcTexWidth = 256;
@@ -345,51 +313,31 @@ TEST_F(BackendTest, ColorResolve) {
     constexpr auto kColorTexFormat = TextureFormat::RGBA8;
     constexpr int kSampleCount = 4;
 
-    // Create a program.
-    ProgramHandle program;
-    {
-        filamat::DescriptorSets descriptors;
-        descriptors[1] = { { "Params",
-                { DescriptorType::UNIFORM_BUFFER, ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 0 },
-                {} } };
-        ShaderGenerator shaderGen(
-                triangleVs, triangleFs, sBackend, sIsMobilePlatform, std::move(descriptors));
-        Program prog = shaderGen.getProgram(api);
-        prog.descriptorBindings(1, {{ "Params", DescriptorType::UNIFORM_BUFFER, 0 }});
-        program = api.createProgram(std::move(prog));
-    }
-
-    DescriptorSetLayoutHandle descriptorSetLayout = api.createDescriptorSetLayout({
-            {{
-                     DescriptorType::UNIFORM_BUFFER,
-                     ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 0,
-                     DescriptorFlags::NONE, 0
-             }}});
-
-    DescriptorSetHandle descriptorSet = api.createDescriptorSet(descriptorSetLayout);
+    Shader shader = createShader();
 
     // Create a VertexBuffer, IndexBuffer, and RenderPrimitive.
     TrianglePrimitive const triangle(api);
 
     // Create 4-sample texture.
-    Handle<HwTexture> const srcColorTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, 1, kColorTexFormat, kSampleCount, kSrcTexWidth, kSrcTexHeight, 1,
-        TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> const srcColorTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, 1, kColorTexFormat, kSampleCount, kSrcTexWidth, kSrcTexHeight,
+            1,
+            TextureUsage::COLOR_ATTACHMENT));
 
     // Create 1-sample texture.
-    Handle<HwTexture> const dstColorTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, 1, kColorTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> const dstColorTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, 1, kColorTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
+            TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT));
 
     // Create a 4-sample render target with the 4-sample texture.
-    Handle<HwRenderTarget> const srcRenderTarget = api.createRenderTarget(
+    Handle<HwRenderTarget> const srcRenderTarget = mCleanup.add(api.createRenderTarget(
             TargetBufferFlags::COLOR, kSrcTexWidth, kSrcTexHeight, kSampleCount, 0,
-            {{ srcColorTexture }}, {}, {});
+            {{ srcColorTexture }}, {}, {}));
 
     // Create a 1-sample render target with the 1-sample texture.
-    Handle<HwRenderTarget> const dstRenderTarget = api.createRenderTarget(
+    Handle<HwRenderTarget> const dstRenderTarget = mCleanup.add(api.createRenderTarget(
             TargetBufferFlags::COLOR, kDstTexWidth, kDstTexHeight, 1, 0,
-            {{ dstColorTexture }}, {}, {});
+            {{ dstColorTexture }}, {}, {}));
 
     // Prep for rendering.
     RenderPassParams params = {};
@@ -401,65 +349,52 @@ TEST_F(BackendTest, ColorResolve) {
     params.viewport.height = kSrcTexHeight;
 
     PipelineState state = {};
-    state.program = program;
-    state.pipelineLayout.setLayout[1] = { descriptorSetLayout };
+    state.program = shader.getProgram();
+    state.pipelineLayout.setLayout[1] = { shader.getDescriptorSetLayout() };
     state.rasterState.colorWrite = true;
     state.rasterState.depthWrite = false;
     state.rasterState.depthFunc = RasterState::DepthFunc::A;
     state.rasterState.culling = CullingMode::NONE;
 
-    auto ubuffer = api.createBufferObject(sizeof(MaterialParams),
-            BufferObjectBinding::UNIFORM, BufferUsage::STATIC);
+    auto ubuffer = mCleanup.add(api.createBufferObject(sizeof(MaterialParams),
+            BufferObjectBinding::UNIFORM, BufferUsage::STATIC));
     // Draw red triangle into srcRenderTarget.
-    uploadUniforms(api, ubuffer, {
-        .color = float4(1, 0, 0, 1),
-        .scale = float4(1, 1, 0.5, 0),
+    shader.uploadUniform(api, ubuffer, MaterialParams{
+            .color = float4(1, 0, 0, 1),
+            .scale = float4(1, 1, 0.5, 0),
     });
-
-    api.updateDescriptorSetBuffer(descriptorSet, 0, ubuffer, 0, sizeof(MaterialParams));
-    api.bindDescriptorSet(descriptorSet, 1, {});
+    shader.bindUniform<MaterialParams>(api, ubuffer);
 
     // FIXME: on Metal this triangle is not drawn. Can't understand why.
-    api.beginFrame(0, 0, 0);
+    {
+        RenderFrame frame(api);
         api.beginRenderPass(srcRenderTarget, params);
-            api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
         api.endRenderPass();
-    api.endFrame(0);
+    }
 
     // Resolve the MSAA render target into the single-sample render target.
     api.blitDEPRECATED(TargetBufferFlags::COLOR,
-            dstRenderTarget, {0, 0, kDstTexWidth, kDstTexHeight},
-            srcRenderTarget, {0, 0, kSrcTexWidth, kSrcTexHeight},
+            dstRenderTarget, { 0, 0, kDstTexWidth, kDstTexHeight },
+            srcRenderTarget, { 0, 0, kSrcTexWidth, kSrcTexHeight },
             SamplerMagFilter::NEAREST);
 
-    // Grab a screenshot.
-    ScreenshotParams sparams{ kDstTexWidth, kDstTexHeight, "ColorResolve.png" };
-    dumpScreenshot(api, dstRenderTarget, &sparams);
+    {
+        ImageExpectations expectations(api);
 
-    // Wait for the ReadPixels result to come back.
-    flushAndWait();
+        EXPECT_IMAGE(dstRenderTarget, expectations,
+                ScreenshotParams(kDstTexWidth, kDstTexHeight, "ColorResolve", 0xebfac2ef));
 
-    // Check if the image matches perfectly to our golden run.
-    const uint32_t expected = 0xebfac2ef;
-    printf("Computed hash is 0x%8.8x, Expected 0x%8.8x\n", sparams.pixelHashResult, expected);
-    EXPECT_TRUE(sparams.pixelHashResult == expected);
-
-    // Cleanup.
-    api.destroyDescriptorSet(descriptorSet);
-    api.destroyDescriptorSetLayout(descriptorSetLayout);
-    api.destroyBufferObject(ubuffer);
-    api.destroyProgram(program);
-    api.destroyTexture(srcColorTexture);
-    api.destroyTexture(dstColorTexture);
-    api.destroyRenderTarget(srcRenderTarget);
-    api.destroyRenderTarget(dstRenderTarget);
-    flushAndWait();
+        flushAndWait();
+    }
 }
 
-TEST_F(BackendTest, Blit2DTextureArray) {
+TEST_F(BlitTest, Blit2DTextureArray) {
     auto& api = getDriverApi();
+    mCleanup.addPostCall([&]() { executeCommands(); });
 
     api.startCapture(0);
+    mCleanup.addPostCall([&]() { api.stopCapture(0); });
 
     constexpr int kSrcTexWidth = 256;
     constexpr int kSrcTexHeight = 256;
@@ -474,69 +409,65 @@ TEST_F(BackendTest, Blit2DTextureArray) {
     constexpr int kDstTexLayer = 0;
 
     // Create a SwapChain and make it current. We don't really use it so the res doesn't matter.
-    auto swapChain = api.createSwapChainHeadless(256, 256, 0);
+    auto swapChain = mCleanup.add(api.createSwapChainHeadless(256, 256, 0));
     api.makeCurrent(swapChain, swapChain);
 
     // Create a source texture.
-    Handle<HwTexture> srcTexture = api.createTexture(
-            SamplerType::SAMPLER_2D_ARRAY, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, kSrcTexDepth,
-            TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> srcTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D_ARRAY, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth,
+            kSrcTexHeight, kSrcTexDepth,
+            TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT));
     const bool flipY = sBackend == Backend::OPENGL;
-    createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 0, kSrcTexLayer, float3(0.5, 0, 0), flipY);
+    createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 0, kSrcTexLayer, float3(0.5, 0, 0),
+            flipY);
 
     // Create a destination texture.
-    Handle<HwTexture> dstTexture = api.createTexture(
+    Handle<HwTexture> dstTexture = mCleanup.add(api.createTexture(
             SamplerType::SAMPLER_2D, kNumLevels, kDstTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
-            TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT);
+            TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT));
 
     // Create two RenderTargets.
     const int level = 0;
-    Handle<HwRenderTarget> srcRenderTarget = api.createRenderTarget( TargetBufferFlags::COLOR,
-            kSrcTexWidth >> level, kSrcTexHeight >> level, 1, 0, { srcTexture, level, kSrcTexLayer }, {}, {});
-    Handle<HwRenderTarget> dstRenderTarget = api.createRenderTarget( TargetBufferFlags::COLOR,
-            kDstTexWidth >> level, kDstTexHeight >> level, 1, 0, { dstTexture, level, kDstTexLayer }, {}, {});
+    Handle<HwRenderTarget> srcRenderTarget = mCleanup.add(
+            api.createRenderTarget(TargetBufferFlags::COLOR,
+                    kSrcTexWidth >> level, kSrcTexHeight >> level, 1, 0,
+                    { srcTexture, level, kSrcTexLayer }, {}, {}));
+    Handle<HwRenderTarget> dstRenderTarget = mCleanup.add(
+            api.createRenderTarget(TargetBufferFlags::COLOR,
+                    kDstTexWidth >> level, kDstTexHeight >> level, 1, 0,
+                    { dstTexture, level, kDstTexLayer }, {}, {}));
 
     // Do a blit from kSrcTexLayer of the source RT to kDstTexLayer of the destination RT.
     const int srcLevel = 0;
     api.blitDEPRECATED(TargetBufferFlags::COLOR0, dstRenderTarget,
-            {0, 0, kDstTexWidth, kDstTexHeight}, srcRenderTarget,
-            {0, 0, kSrcTexWidth >> srcLevel, kSrcTexHeight >> srcLevel}, SamplerMagFilter::LINEAR);
+            { 0, 0, kDstTexWidth, kDstTexHeight }, srcRenderTarget,
+            { 0, 0, kSrcTexWidth >> srcLevel, kSrcTexHeight >> srcLevel },
+            SamplerMagFilter::LINEAR);
 
     // Push through an empty frame to allow the texture to upload and the blit to execute.
-    api.beginFrame(0, 0, 0);
-    api.commit(swapChain);
-    api.endFrame(0);
+    {
+        RenderFrame frame(api);
+        api.commit(swapChain);
+    }
 
-    // Grab a screenshot.
-    ScreenshotParams params { kDstTexWidth, kDstTexHeight, "Blit2DTextureArray.png" };
-    api.beginFrame(0, 0, 0);
-    dumpScreenshot(api, dstRenderTarget, &params);
-    api.commit(swapChain);
-    api.endFrame(0);
+    {
+        ImageExpectations expectations(api);
 
-    // Wait for the ReadPixels result to come back.
-    api.finish();
-    executeCommands();
-    getDriver().purge();
+        {
+            RenderFrame frame(api);
+            EXPECT_IMAGE(dstRenderTarget, expectations,
+                    ScreenshotParams(kDstTexWidth, kDstTexHeight, "Blit2DTextureArray",
+                            0x8de7d55b));
+            api.commit(swapChain);
+        }
 
-    // Check if the image matches perfectly to our golden run.
-    const uint32_t expected = 0x8de7d55b;
-    printf("Computed hash is 0x%8.8x, Expected 0x%8.8x\n", params.pixelHashResult, expected);
-    EXPECT_TRUE(params.pixelHashResult == expected);
-
-    // Cleanup.
-    api.destroyTexture(srcTexture);
-    api.destroyTexture(dstTexture);
-    api.destroySwapChain(swapChain);
-    api.destroyRenderTarget(srcRenderTarget);
-    api.destroyRenderTarget(dstRenderTarget);
-    executeCommands();
-
-    api.stopCapture(0);
+        flushAndWait();
+    }
 }
 
-TEST_F(BackendTest, BlitRegion) {
+TEST_F(BlitTest, BlitRegion) {
     auto& api = getDriverApi();
+    mCleanup.addPostCall([&]() { executeCommands(); });
 
     constexpr int kSrcTexWidth = 1024;
     constexpr int kSrcTexHeight = 1024;
@@ -549,88 +480,75 @@ TEST_F(BackendTest, BlitRegion) {
     constexpr int kDstLevel = 0;
 
     // Create a SwapChain and make it current. We don't really use it so the res doesn't matter.
-    auto swapChain = api.createSwapChainHeadless(256, 256, 0);
+    auto swapChain = mCleanup.add(api.createSwapChainHeadless(256, 256, 0));
     api.makeCurrent(swapChain, swapChain);
 
     // Create a source texture.
-    Handle<HwTexture> srcTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> srcTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
+            TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT));
     const bool flipY = sBackend == Backend::OPENGL;
     createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 0, float3(0.5, 0, 0), flipY);
     createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 1, float3(0, 0, 0.5), flipY);
 
     // Create a destination texture.
-    Handle<HwTexture> dstTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kDstTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> dstTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, kNumLevels, kDstTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
+            TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT));
 
     // Blit one-quarter of src level 1 to dst level 0.
     Viewport srcRect = {
-        .left = 0,
-        .bottom = 0,
-        .width = (kSrcTexWidth >> kSrcLevel) / 2,
-        .height = (kSrcTexHeight >> kSrcLevel) / 2,
+            .left = 0,
+            .bottom = 0,
+            .width = (kSrcTexWidth >> kSrcLevel) / 2,
+            .height = (kSrcTexHeight >> kSrcLevel) / 2,
     };
     Viewport dstRect = {
-        .left = 10,
-        .bottom = 10,
-        .width = kDstTexWidth - 10,
-        .height = kDstTexHeight - 10,
+            .left = 10,
+            .bottom = 10,
+            .width = kDstTexWidth - 10,
+            .height = kDstTexHeight - 10,
     };
 
     // Create a RenderTarget for each texture's miplevel.
     // We purposely set the render target width and height to smaller than the texture, to check
     // that this case is handled correctly.
     Handle<HwRenderTarget> srcRenderTarget =
-            api.createRenderTarget(TargetBufferFlags::COLOR, srcRect.width,
-                    srcRect.height, 1, 0, {srcTexture, kSrcLevel, 0}, {}, {});
+            mCleanup.add(api.createRenderTarget(TargetBufferFlags::COLOR, srcRect.width,
+                    srcRect.height, 1, 0, { srcTexture, kSrcLevel, 0 }, {}, {}));
     Handle<HwRenderTarget> dstRenderTarget =
-            api.createRenderTarget(TargetBufferFlags::COLOR, kDstTexWidth >> kDstLevel,
-                    kDstTexHeight >> kDstLevel, 1, 0, {dstTexture, kDstLevel, 0}, {}, {});
+            mCleanup.add(api.createRenderTarget(TargetBufferFlags::COLOR, kDstTexWidth >> kDstLevel,
+                    kDstTexHeight >> kDstLevel, 1, 0, { dstTexture, kDstLevel, 0 }, {}, {}));
 
-    api.blitDEPRECATED(TargetBufferFlags::COLOR0, dstRenderTarget, dstRect, srcRenderTarget, srcRect,
+    api.blitDEPRECATED(TargetBufferFlags::COLOR0, dstRenderTarget, dstRect, srcRenderTarget,
+            srcRect,
             SamplerMagFilter::LINEAR);
 
     // Push through an empty frame to allow the texture to upload and the blit to execute.
-    api.beginFrame(0, 0, 0);
-    api.commit(swapChain);
-    api.endFrame(0);
+    {
+        RenderFrame frame(api);
+        api.commit(swapChain);
+    }
 
-    // Grab a screenshot.
-    ScreenshotParams params { kDstTexWidth, kDstTexHeight, "BlitRegion.png" };
-    api.beginFrame(0, 0, 0);
-    dumpScreenshot(api, dstRenderTarget, &params);
-    api.commit(swapChain);
-    api.endFrame(0);
+    {
+        ImageExpectations expectations(api);
 
-    // Wait for the ReadPixels result to come back.
-    api.finish();
-    executeCommands();
-    getDriver().purge();
+        {
+            RenderFrame frame(api);
+            // TODO: for some reason, this test has very, very slight (as in one pixel) differences
+            // between OpenGL and Metal. So disable golden checking for now.
+            // EXPECT_IMAGE(dstRenderTarget, expectations, ScreenshotParams(kDstTexWidth,
+            //         kDstTexHeight, "BlitRegion", 0x74fa34ed));
+            api.commit(swapChain);
+        }
 
-    // Check if the image matches perfectly to our golden run.
-    //
-    // TODO: for some reason, this test has very, very slight (as in one pixel) differences between
-    // OpenGL and Metal. So disable golden checking for now.
-    // Use the compare tool from ImageMagick to see visual differences:
-    // compare -verbose -metric mae BlitRegion_Metal.png BlitRegion_OpenGL.png difference.png
-    //
-    // const uint32_t expected = 0x74fa34ed;
-    // printf("Computed hash is 0x%8.8x, Expected 0x%8.8x\n", params.pixelHashResult, expected);
-    // EXPECT_TRUE(params.pixelHashResult == expected);
-
-    // Cleanup.
-    api.destroyTexture(srcTexture);
-    api.destroyTexture(dstTexture);
-    api.destroySwapChain(swapChain);
-    api.destroyRenderTarget(srcRenderTarget);
-    api.destroyRenderTarget(dstRenderTarget);
-    executeCommands();
+        flushAndWait();
+    }
 }
 
-TEST_F(BackendTest, BlitRegionToSwapChain) {
+TEST_F(BlitTest, BlitRegionToSwapChain) {
     auto& api = getDriverApi();
+    mCleanup.addPostCall([&]() { executeCommands(); });
 
     constexpr int kSrcTexWidth = 1024;
     constexpr int kSrcTexHeight = 1024;
@@ -640,14 +558,15 @@ TEST_F(BackendTest, BlitRegionToSwapChain) {
     constexpr int kNumLevels = 3;
 
     // Create a SwapChain and make it current.
-    auto swapChain = createSwapChain();
-    Handle<HwRenderTarget> dstRenderTarget = api.createDefaultRenderTarget();
+    auto swapChain = mCleanup.add(createSwapChain());
     api.makeCurrent(swapChain, swapChain);
 
+    Handle<HwRenderTarget> dstRenderTarget = mCleanup.add(api.createDefaultRenderTarget());
+
     // Create a source texture.
-    Handle<HwTexture> srcTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT);
+    Handle<HwTexture> srcTexture = mCleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
+            TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT));
     const bool flipY = sBackend == Backend::OPENGL;
     createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 0, float3(0.5, 0, 0), flipY);
     createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 1, float3(0, 0, 0.5), flipY);
@@ -655,48 +574,46 @@ TEST_F(BackendTest, BlitRegionToSwapChain) {
     // Create a RenderTarget for each texture's miplevel.
     Handle<HwRenderTarget> srcRenderTargets[kNumLevels];
     for (uint8_t level = 0; level < kNumLevels; level++) {
-        srcRenderTargets[level] = api.createRenderTarget( TargetBufferFlags::COLOR,
-                kSrcTexWidth >> level, kSrcTexHeight >> level, 1, 0, { srcTexture, level, 0 }, {}, {});
+        srcRenderTargets[level] = mCleanup.add(api.createRenderTarget(TargetBufferFlags::COLOR,
+                kSrcTexWidth >> level, kSrcTexHeight >> level, 1, 0, { srcTexture, level, 0 }, {},
+                {}));
     }
 
     // Blit one-quarter of src level 1 to dst level 0.
     const int srcLevel = 1;
     Viewport srcRect = {
-        .left = (kSrcTexWidth >> srcLevel) / 2,
-        .bottom = (kSrcTexHeight >> srcLevel) / 2,
-        .width = (kSrcTexWidth >> srcLevel) / 2,
-        .height = (kSrcTexHeight >> srcLevel) / 2,
+            .left = (kSrcTexWidth >> srcLevel) / 2,
+            .bottom = (kSrcTexHeight >> srcLevel) / 2,
+            .width = (kSrcTexWidth >> srcLevel) / 2,
+            .height = (kSrcTexHeight >> srcLevel) / 2,
     };
     Viewport dstRect = {
-        .left = 10,
-        .bottom = 10,
-        .width = kDstTexWidth - 10,
-        .height = kDstTexHeight - 10,
+            .left = 10,
+            .bottom = 10,
+            .width = kDstTexWidth - 10,
+            .height = kDstTexHeight - 10,
     };
 
-    api.beginFrame(0, 0, 0);
+    {
+        ImageExpectations expectations(api);
+        {
+            {
+                RenderFrame frame(api);
 
-    api.blitDEPRECATED(TargetBufferFlags::COLOR0, dstRenderTarget,
-            dstRect, srcRenderTargets[srcLevel],
-            srcRect, SamplerMagFilter::LINEAR);
+                api.blitDEPRECATED(TargetBufferFlags::COLOR0, dstRenderTarget,
+                        dstRect, srcRenderTargets[srcLevel],
+                        srcRect, SamplerMagFilter::LINEAR);
 
-    ScreenshotParams params { kDstTexWidth, kDstTexHeight, "BlitRegionToSwapChain.png" };
-    dumpScreenshot(api, dstRenderTarget, &params);
+                api.commit(swapChain);
+            }
 
-    api.commit(swapChain);
-
-    api.endFrame(0);
-
-    // Wait for the ReadPixels result to come back.
-    api.finish();
-    executeCommands();
-    getDriver().purge();
-
-    // Cleanup.
-    api.destroyTexture(srcTexture);
-    api.destroySwapChain(swapChain);
-    for (auto rt : srcRenderTargets)  api.destroyRenderTarget(rt);
-    executeCommands();
+            // TODO: for some reason, this test has been disabled. It needs to be tested on all
+            // machines.
+            // EXPECT_IMAGE(dstRenderTarget, expectations,
+            //         ScreenshotParams(kDstTexWidth, kDstTexHeight, "BlitRegionToSwapChain", 0x0));
+        }
+        flushAndWait();
+    }
 }
 
 } // namespace test

filament/backend/test/test_BufferUpdates.cpp

@@ -16,6 +16,8 @@
 
 #include "BackendTest.h"
 
+#include "Lifetimes.h"
+#include "Shader.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
 
@@ -78,8 +80,28 @@ struct MaterialParams {
     math::float4 offset;
 };
 static_assert(sizeof(MaterialParams) == 8 * sizeof(float));
+// Uniform config for writing MaterialParams to the shader uniform with 64 bytes of padding.
+const UniformBindingConfig kBindingConfig = {
+        .dataSize = sizeof(MaterialParams),
+        .bufferSize = sizeof(MaterialParams) + 64,
+        .byteOffset = 64
+};
 
-TEST_F(BackendTest, VertexBufferUpdate) {
+class BufferUpdatesTest : public BackendTest {
+public:
+    BufferUpdatesTest() : mCleanup(getDriverApi()) {}
+
+protected:
+    Shader createShader() {
+        return Shader(getDriverApi(), mCleanup, ShaderConfig{
+           vertex, fragment, {"Params"}
+        });
+    }
+
+    Cleanup mCleanup;
+};
+
+TEST_F(BufferUpdatesTest, VertexBufferUpdate) {
     const bool largeBuffers = false;
 
     // If updateIndices is true, then even-numbered triangles will have their indices set to
@@ -90,31 +112,15 @@ TEST_F(BackendTest, VertexBufferUpdate) {
     // executeCommands().
     {
         auto& api = getDriverApi();
+        Cleanup cleanup(api);
 
         // Create a platform-specific SwapChain and make it current.
-        auto swapChain = createSwapChain();
+        auto swapChain = cleanup.add(createSwapChain());
         api.makeCurrent(swapChain, swapChain);
 
-        // Create a program.
-        filamat::DescriptorSets descriptors;
-        descriptors[1] = { { "Params",
-                { DescriptorType::UNIFORM_BUFFER, ShaderStageFlags::FRAGMENT, 0 }, {} } };
-        ShaderGenerator shaderGen(
-                vertex, fragment, sBackend, sIsMobilePlatform, std::move(descriptors));
-        Program p = shaderGen.getProgram(api);
-        p.descriptorBindings(1, {{ "Params", DescriptorType::UNIFORM_BUFFER, 0 }});
-        auto program = api.createProgram(std::move(p));
+        Shader shader = createShader();
 
-        DescriptorSetLayoutHandle descriptorSetLayout = api.createDescriptorSetLayout({
-                {{
-                         DescriptorType::UNIFORM_BUFFER,
-                         ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 0,
-                         DescriptorFlags::NONE, 0
-                 }}});
-
-        DescriptorSetHandle descriptorSet = api.createDescriptorSet(descriptorSetLayout);
-
-        auto defaultRenderTarget = api.createDefaultRenderTarget(0);
+        auto defaultRenderTarget = cleanup.add(api.createDefaultRenderTarget(0));
 
         // To test large buffers (which exercise a different code path) create an extra large
         // buffer. Only the first 3 vertices will be used.
@@ -128,38 +134,29 @@ TEST_F(BackendTest, VertexBufferUpdate) {
         params.flags.discardEnd = TargetBufferFlags::NONE;
 
         PipelineState state;
-        state.program = program;
-        state.pipelineLayout.setLayout[1] = { descriptorSetLayout };
+        state.program = shader.getProgram();
+        state.pipelineLayout.setLayout[1] = { shader.getDescriptorSetLayout() };
         state.rasterState.colorWrite = true;
         state.rasterState.depthWrite = false;
         state.rasterState.depthFunc = RasterState::DepthFunc::A;
         state.rasterState.culling = CullingMode::NONE;
 
         // Create a uniform buffer.
-        // We use STATIC here, even though the buffer is updated, to force the Metal backend to use a
-        // GPU buffer, which is more interesting to test.
-        auto ubuffer = api.createBufferObject(sizeof(MaterialParams) + 64,
-                BufferObjectBinding::UNIFORM, BufferUsage::STATIC);
+        // We use STATIC here, even though the buffer is updated, to force the Metal backend to use
+        // a GPU buffer, which is more interesting to test.
+        auto ubuffer = cleanup.add(api.createBufferObject(sizeof(MaterialParams) + 64,
+                BufferObjectBinding::UNIFORM, BufferUsage::STATIC));
 
-        api.updateDescriptorSetBuffer(descriptorSet, 0, ubuffer, 0, sizeof(MaterialParams) + 64);
-        api.bindDescriptorSet(descriptorSet, 1, {});
+        shader.bindUniform<MaterialParams>(api, ubuffer, kBindingConfig);
 
         api.startCapture(0);
 
-        // Upload uniforms.
-        {
-            MaterialParams params {
-                    .color = { 1.0f, 1.0f, 1.0f, 1.0f },
-                    .offset = { 0.0f, 0.0f, 0.0f, 0.0f }
-            };
-            auto* tmp = new MaterialParams(params);
-            auto cb = [](void* buffer, size_t size, void* user) {
-                auto* sp = (MaterialParams*) buffer;
-                delete sp;
-            };
-            BufferDescriptor bd(tmp, sizeof(MaterialParams), cb);
-            api.updateBufferObject(ubuffer, std::move(bd), 64);
-        }
+        // Upload the uniform, but with an offset to accommodate the padding in the shader's
+        // uniform definition.
+        shader.uploadUniform(api, ubuffer, kBindingConfig, MaterialParams{
+                .color = { 1.0f, 1.0f, 1.0f, 1.0f },
+                .offset = { 0.0f, 0.0f, 0.0f, 0.0f }
+        });
 
         api.makeCurrent(swapChain, swapChain);
         api.beginFrame(0, 0, 0);
@@ -202,13 +199,6 @@ TEST_F(BackendTest, VertexBufferUpdate) {
         api.endFrame(0);
 
         api.stopCapture(0);
-
-        api.destroyProgram(program);
-        api.destroySwapChain(swapChain);
-        api.destroyDescriptorSet(descriptorSet);
-        api.destroyDescriptorSetLayout(descriptorSetLayout);
-        api.destroyBufferObject(ubuffer);
-        api.destroyRenderTarget(defaultRenderTarget);
     }
 
     executeCommands();
@@ -216,62 +206,38 @@ TEST_F(BackendTest, VertexBufferUpdate) {
 
 // This test renders two triangles in two separate draw calls. Between the draw calls, a uniform
 // buffer object is partially updated.
-TEST_F(BackendTest, BufferObjectUpdateWithOffset) {
+TEST_F(BufferUpdatesTest, BufferObjectUpdateWithOffset) {
     auto& api = getDriverApi();
+    Cleanup cleanup(api);
 
     // Create a platform-specific SwapChain and make it current.
-    auto swapChain = createSwapChain();
+    auto swapChain = cleanup.add(createSwapChain());
     api.makeCurrent(swapChain, swapChain);
 
     // Create a program.
-    filamat::DescriptorSets descriptors;
-    descriptors[1] = { { "Params",
-            { DescriptorType::UNIFORM_BUFFER, ShaderStageFlags::FRAGMENT, 0 }, {} } };
-    ShaderGenerator shaderGen(
-            vertex, fragment, sBackend, sIsMobilePlatform, std::move(descriptors));
-    Program p = shaderGen.getProgram(api);
-    p.descriptorBindings(1, {{ "Params", DescriptorType::UNIFORM_BUFFER, 0 }});
-    auto program = api.createProgram(std::move(p));
-
-    DescriptorSetLayoutHandle descriptorSetLayout = api.createDescriptorSetLayout({
-            {{
-                     DescriptorType::UNIFORM_BUFFER,
-                     ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 0,
-                     DescriptorFlags::NONE, 0
-             }}});
-
-    DescriptorSetHandle descriptorSet = api.createDescriptorSet(descriptorSetLayout);
-
+    Shader shader = createShader();
 
     // Create a uniform buffer.
     // We use STATIC here, even though the buffer is updated, to force the Metal backend to use a
     // GPU buffer, which is more interesting to test.
-    auto ubuffer = api.createBufferObject(sizeof(MaterialParams) + 64,
-            BufferObjectBinding::UNIFORM, BufferUsage::STATIC);
+    auto ubuffer = cleanup.add(api.createBufferObject(sizeof(MaterialParams) + 64,
+            BufferObjectBinding::UNIFORM, BufferUsage::STATIC));
 
-    api.updateDescriptorSetBuffer(descriptorSet, 0, ubuffer, 0, sizeof(MaterialParams) + 64);
-    api.bindDescriptorSet(descriptorSet, 1, {});
+    shader.bindUniform<MaterialParams>(api, ubuffer, kBindingConfig);
 
     // Create a render target.
-    auto colorTexture = api.createTexture(SamplerType::SAMPLER_2D, 1,
-            TextureFormat::RGBA8, 1, 512, 512, 1, TextureUsage::COLOR_ATTACHMENT);
-    auto renderTarget = api.createRenderTarget(
-            TargetBufferFlags::COLOR0, 512, 512, 1, 0, {{colorTexture}}, {}, {});
+    auto colorTexture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
+            TextureFormat::RGBA8, 1, 512, 512, 1, TextureUsage::COLOR_ATTACHMENT));
+    auto renderTarget = cleanup.add(api.createRenderTarget(
+            TargetBufferFlags::COLOR0, 512, 512, 1, 0, {{colorTexture}}, {}, {}));
 
     // Upload uniforms for the first triangle.
-    {
-        MaterialParams params {
+    // Upload the uniform, but with an offset to accommodate the padding in the shader's
+    // uniform definition.
+    shader.uploadUniform(api, ubuffer, kBindingConfig, MaterialParams{
             .color = { 1.0f, 0.0f, 0.5f, 1.0f },
             .offset = { 0.0f, 0.0f, 0.0f, 0.0f }
-        };
-        auto* tmp = new MaterialParams(params);
-        auto cb = [](void* buffer, size_t size, void* user) {
-            auto* sp = (MaterialParams*) buffer;
-            delete sp;
-        };
-        BufferDescriptor bd(tmp, sizeof(MaterialParams), cb);
-        api.updateBufferObject(ubuffer, std::move(bd), 64);
-    }
+    });
 
     RenderPassParams params = {};
     params.flags.clear = TargetBufferFlags::COLOR;
@@ -280,29 +246,27 @@ TEST_F(BackendTest, BufferObjectUpdateWithOffset) {
     params.flags.discardEnd = TargetBufferFlags::NONE;
     params.viewport.height = 512;
     params.viewport.width = 512;
-    renderTriangle({{ DescriptorSetLayoutHandle{}, descriptorSetLayout }},
-            renderTarget, swapChain, program, params);
+    renderTriangle({{ DescriptorSetLayoutHandle{}, shader.getDescriptorSetLayout() }},
+            renderTarget, swapChain, shader.getProgram(), params);
 
     // Upload uniforms for the second triangle. To test partial buffer updates, we'll only update
     // color.b, color.a, offset.x, and offset.y.
-    {
-        MaterialParams params {
-                .color = { 1.0f, 0.0f, 1.0f, 1.0f },
-                .offset = { 0.5f, 0.5f, 0.0f, 0.0f }
-        };
-        auto* tmp = new MaterialParams(params);
-        auto cb = [](void* buffer, size_t size, void* user) {
-            auto* sp = (MaterialParams*) ((char*)buffer - offsetof(MaterialParams, color.b));
-            delete sp;
-        };
-        BufferDescriptor bd((char*)tmp + offsetof(MaterialParams, color.b), sizeof(float) * 4, cb);
-        api.updateBufferObject(ubuffer, std::move(bd), 64 + offsetof(MaterialParams, color.b));
-    }
+    shader.uploadUniform(api, ubuffer, UniformBindingConfig{
+                    .dataSize = sizeof(std::array<float, 4>),
+                    .bufferSize = kBindingConfig.bufferSize,
+                    .byteOffset = *kBindingConfig.byteOffset + offsetof(MaterialParams, color.b),
+            },
+            std::array<float, 4>{
+                    // color.b, color.a
+                    1.0f, 1.0f,
+                    // offset.x, offset.y
+                    0.5f, 0.5f }
+    );
 
     params.flags.clear = TargetBufferFlags::NONE;
     params.flags.discardStart = TargetBufferFlags::NONE;
-    renderTriangle({{ DescriptorSetLayoutHandle{}, descriptorSetLayout }},
-            renderTarget, swapChain, program, params);
+    renderTriangle({{ DescriptorSetLayoutHandle{}, shader.getDescriptorSetLayout() }},
+            renderTarget, swapChain, shader.getProgram(), params);
 
     static const uint32_t expectedHash = 91322442;
     readPixelsAndAssertHash(
@@ -312,14 +276,6 @@ TEST_F(BackendTest, BufferObjectUpdateWithOffset) {
     api.commit(swapChain);
     api.endFrame(0);
 
-    api.destroyProgram(program);
-    api.destroySwapChain(swapChain);
-    api.destroyDescriptorSet(descriptorSet);
-    api.destroyDescriptorSetLayout(descriptorSetLayout);
-    api.destroyBufferObject(ubuffer);
-    api.destroyRenderTarget(renderTarget);
-    api.destroyTexture(colorTexture);
-
     // This ensures all driver commands have finished before exiting the test.
     api.finish();
 

filament/backend/test/test_Callbacks.cpp

@@ -16,6 +16,8 @@
 
 #include "BackendTest.h"
 
+#include "Lifetimes.h"
+
 using namespace filament;
 using namespace filament::backend;
 
@@ -23,13 +25,14 @@ namespace test {
 
 TEST_F(BackendTest, FrameScheduledCallback) {
     auto& api = getDriverApi();
+    Cleanup cleanup(api);
 
     // Create a SwapChain.
     // In order for the frameScheduledCallback to be called, this must be a real SwapChain (not
     // headless) so we obtain a drawable.
-    auto swapChain = createSwapChain();
+    auto swapChain = cleanup.add(createSwapChain());
 
-    Handle<HwRenderTarget> renderTarget = api.createDefaultRenderTarget();
+    Handle<HwRenderTarget> renderTarget = cleanup.add(api.createDefaultRenderTarget());
 
     int callbackCountA = 0;
     api.setFrameScheduledCallback(swapChain, nullptr, [&callbackCountA](PresentCallable callable) {
@@ -79,9 +82,10 @@ TEST_F(BackendTest, FrameScheduledCallback) {
 
 TEST_F(BackendTest, FrameCompletedCallback) {
     auto& api = getDriverApi();
+    Cleanup cleanup(api);
 
     // Create a SwapChain.
-    auto swapChain = api.createSwapChainHeadless(256, 256, 0);
+    auto swapChain = cleanup.add(api.createSwapChainHeadless(256, 256, 0));
 
     int callbackCountA = 0;
     api.setFrameCompletedCallback(swapChain, nullptr,

filament/backend/test/test_FeedbackLoops.cpp

@@ -16,6 +16,7 @@
 
 #include "BackendTest.h"
 
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
 
@@ -129,12 +130,13 @@ static void dumpScreenshot(DriverApi& dapi, Handle<HwRenderTarget> rt) {
 // backend's readPixels does not work correctly with textures that have image data uploaded.
 TEST_F(BackendTest, FeedbackLoops) {
     auto& api = getDriverApi();
+    Cleanup cleanup(api);
 
     // The test is executed within this block scope to force destructors to run before
     // executeCommands().
     {
         // Create a platform-specific SwapChain and make it current.
-        auto swapChain = createSwapChain();
+        auto swapChain = cleanup.add(createSwapChain());
         api.makeCurrent(swapChain, swapChain);
 
         // Create a program.
@@ -155,10 +157,10 @@ TEST_F(BackendTest, FeedbackLoops) {
                     { "test_tex", DescriptorType::SAMPLER, 0 },
                     { "Params", DescriptorType::UNIFORM_BUFFER, 1 }
             });
-            program = api.createProgram(std::move(prog));
+            program = cleanup.add(api.createProgram(std::move(prog)));
         }
 
-        DescriptorSetLayoutHandle descriptorSetLayout = api.createDescriptorSetLayout({
+        DescriptorSetLayoutHandle descriptorSetLayout = cleanup.add(api.createDescriptorSetLayout({
                 {{
                          DescriptorType::SAMPLER,
                          ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 0,
@@ -168,27 +170,27 @@ TEST_F(BackendTest, FeedbackLoops) {
                          DescriptorType::UNIFORM_BUFFER,
                          ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 1,
                          DescriptorFlags::NONE, 0
-                 }}});
+                 }}}));
 
 
         TrianglePrimitive const triangle(getDriverApi());
 
         // Create a texture.
         auto usage = TextureUsage::COLOR_ATTACHMENT | TextureUsage::SAMPLEABLE;
-        Handle<HwTexture> const texture = api.createTexture(
-            SamplerType::SAMPLER_2D, kNumLevels, kTexFormat, 1, kTexWidth, kTexHeight, 1, usage);
+        Handle<HwTexture> const texture = cleanup.add(api.createTexture(
+            SamplerType::SAMPLER_2D, kNumLevels, kTexFormat, 1, kTexWidth, kTexHeight, 1, usage));
 
         // Create ubo
-        auto ubuffer = api.createBufferObject(sizeof(MaterialParams),
-                BufferObjectBinding::UNIFORM, BufferUsage::STATIC);
+        auto ubuffer = cleanup.add(api.createBufferObject(sizeof(MaterialParams),
+                BufferObjectBinding::UNIFORM, BufferUsage::STATIC));
 
         // Create a RenderTarget for each miplevel.
         Handle<HwRenderTarget> renderTargets[kNumLevels];
         for (uint8_t level = 0; level < kNumLevels; level++) {
             slog.i << "Level " << int(level) << ": " <<
                     (kTexWidth >> level) << "x" << (kTexHeight >> level) << io::endl;
-            renderTargets[level] = api.createRenderTarget( TargetBufferFlags::COLOR,
-                    kTexWidth >> level, kTexHeight >> level, 1, 0, { texture, level, 0 }, {}, {});
+            renderTargets[level] = cleanup.add(api.createRenderTarget( TargetBufferFlags::COLOR,
+                    kTexWidth >> level, kTexHeight >> level, 1, 0, { texture, level, 0 }, {}, {}));
         }
 
         // Fill the base level of the texture with interesting colors.
@@ -226,12 +228,13 @@ TEST_F(BackendTest, FeedbackLoops) {
             params.flags.discardStart = TargetBufferFlags::ALL;
             state.rasterState.disableBlending();
             for (int targetLevel = 1; targetLevel < kNumLevels; targetLevel++) {
+                Cleanup passCleanup(api);
                 const uint32_t sourceLevel = targetLevel - 1;
                 params.viewport.width = kTexWidth >> targetLevel;
                 params.viewport.height = kTexHeight >> targetLevel;
 
-                auto textureView = api.createTextureView(texture, sourceLevel, 1);
-                DescriptorSetHandle descriptorSet = api.createDescriptorSet(descriptorSetLayout);
+                auto textureView = passCleanup.add(api.createTextureView(texture, sourceLevel, 1));
+                DescriptorSetHandle descriptorSet = passCleanup.add(api.createDescriptorSet(descriptorSetLayout));
                 api.updateDescriptorSetTexture(descriptorSet, 0, textureView, {
                         .filterMag = SamplerMagFilter::LINEAR,
                         .filterMin = SamplerMinFilter::LINEAR_MIPMAP_NEAREST
@@ -247,8 +250,6 @@ TEST_F(BackendTest, FeedbackLoops) {
                 api.beginRenderPass(renderTargets[targetLevel], params);
                 api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
                 api.endRenderPass();
-                api.destroyTexture(textureView);
-                api.destroyDescriptorSet(descriptorSet);
             }
 
             // Upsample passes
@@ -256,12 +257,13 @@ TEST_F(BackendTest, FeedbackLoops) {
             state.rasterState.blendFunctionSrcRGB = BlendFunction::ONE;
             state.rasterState.blendFunctionDstRGB = BlendFunction::ONE;
             for (int targetLevel = kNumLevels - 2; targetLevel >= 0; targetLevel--) {
+                Cleanup passCleanup(api);
                 const uint32_t sourceLevel = targetLevel + 1;
                 params.viewport.width = kTexWidth >> targetLevel;
                 params.viewport.height = kTexHeight >> targetLevel;
 
-                auto textureView = api.createTextureView(texture, sourceLevel, 1);
-                DescriptorSetHandle descriptorSet = api.createDescriptorSet(descriptorSetLayout);
+                auto textureView = passCleanup.add(api.createTextureView(texture, sourceLevel, 1));
+                DescriptorSetHandle descriptorSet = passCleanup.add(api.createDescriptorSet(descriptorSetLayout));
                 api.updateDescriptorSetTexture(descriptorSet, 0, textureView, {
                         .filterMag = SamplerMagFilter::LINEAR,
                         .filterMin = SamplerMinFilter::LINEAR_MIPMAP_NEAREST
@@ -277,8 +279,6 @@ TEST_F(BackendTest, FeedbackLoops) {
                 api.beginRenderPass(renderTargets[targetLevel], params);
                 api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
                 api.endRenderPass();
-                api.destroyTexture(textureView);
-                api.destroyDescriptorSet(descriptorSet);
             }
 
             // Read back the render target corresponding to the base level.
@@ -296,15 +296,6 @@ TEST_F(BackendTest, FeedbackLoops) {
             executeCommands();
             getDriver().purge();
         }
-
-        api.destroyDescriptorSetLayout(descriptorSetLayout);
-        api.destroyProgram(program);
-        api.destroySwapChain(swapChain);
-        api.destroyTexture(texture);
-        api.destroyBufferObject(ubuffer);
-        for (auto rt : renderTargets)  {
-            api.destroyRenderTarget(rt);
-        }
     }
 
     const uint32_t expected = 0x70695aa1;

filament/backend/test/test_LoadImage.cpp

@@ -16,9 +16,10 @@
 
 #include "BackendTest.h"
 
+#include "BackendTestUtils.h"
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
-#include "BackendTestUtils.h"
 
 #include <backend/DriverEnums.h>
 #include <backend/Handle.h>
@@ -295,10 +296,12 @@ TEST_F(BackendTest, UpdateImage2D) {
     // The test is executed within this block scope to force destructors to run before
     // executeCommands().
     for (const auto& t : testCases) {
+        Cleanup cleanup(api);
+
         // Create a platform-specific SwapChain and make it current.
-        auto swapChain = createSwapChain();
+        auto swapChain = cleanup.add(createSwapChain());
         api.makeCurrent(swapChain, swapChain);
-        auto defaultRenderTarget = api.createDefaultRenderTarget(0);
+        auto defaultRenderTarget = cleanup.add(api.createDefaultRenderTarget(0));
 
         // Create a program.
         filament::SamplerInterfaceBlock::SamplerInfo samplerInfo { "test", "tex", 0,
@@ -315,21 +318,21 @@ TEST_F(BackendTest, UpdateImage2D) {
         Program prog = shaderGen.getProgram(api);
         prog.descriptorBindings(1, {{"test_tex", DescriptorType::SAMPLER, 0}});
 
-        ProgramHandle const program = api.createProgram(std::move(prog));
+        ProgramHandle const program = cleanup.add(api.createProgram(std::move(prog)));
 
-        DescriptorSetLayoutHandle descriptorSetLayout = api.createDescriptorSetLayout({
+        DescriptorSetLayoutHandle descriptorSetLayout = cleanup.add(api.createDescriptorSetLayout({
                 {{
                          DescriptorType::SAMPLER,
                          ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 0,
                          DescriptorFlags::NONE, 0
-                 }}});
+                 }}}));
 
-        DescriptorSetHandle descriptorSet = api.createDescriptorSet(descriptorSetLayout);
+        DescriptorSetHandle descriptorSet = cleanup.add(api.createDescriptorSet(descriptorSetLayout));
 
         // Create a Texture.
         auto usage = TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE;
-        Handle<HwTexture> const texture = api.createTexture(SamplerType::SAMPLER_2D, 1,
-                t.textureFormat, 1, 512, 512, 1u, usage);
+        Handle<HwTexture> const texture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
+                t.textureFormat, 1, 512, 512, 1u, usage));
 
         // Upload some pixel data.
         if (t.uploadSubregions) {
@@ -359,13 +362,6 @@ TEST_F(BackendTest, UpdateImage2D) {
 
         api.commit(swapChain);
         api.endFrame(0);
-
-        api.destroyDescriptorSet(descriptorSet);
-        api.destroyDescriptorSetLayout(descriptorSetLayout);
-        api.destroyProgram(program);
-        api.destroySwapChain(swapChain);
-        api.destroyRenderTarget(defaultRenderTarget);
-        api.destroyTexture(texture);
     }
 
     api.finish();

filament/backend/test/test_MRT.cpp

@@ -16,6 +16,7 @@
 
 #include "BackendTest.h"
 
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
 
@@ -60,25 +61,28 @@ using namespace filament;
 using namespace filament::backend;
 
 TEST_F(BackendTest, MRT) {
+    DriverApi& api = getDriverApi();
+    Cleanup cleanup(api);
+
     // The test is executed within this block scope to force destructors to run before
     // executeCommands().
     {
         // Create a platform-specific SwapChain and make it current.
-        auto swapChain = createSwapChain();
-        getDriverApi().makeCurrent(swapChain, swapChain);
+        auto swapChain = cleanup.add(createSwapChain());
+        api.makeCurrent(swapChain, swapChain);
 
         // Create a program.
         ShaderGenerator shaderGen(vertex, fragment, sBackend, sIsMobilePlatform);
-        Program p = shaderGen.getProgram(getDriverApi());
-        auto program = getDriverApi().createProgram(std::move(p));
+        Program p = shaderGen.getProgram(api);
+        auto program = cleanup.add(api.createProgram(std::move(p)));
 
-        TrianglePrimitive triangle(getDriverApi());
+        TrianglePrimitive triangle(api);
 
-        auto defaultRenderTarget = getDriverApi().createDefaultRenderTarget(0);
+        auto defaultRenderTarget = cleanup.add(api.createDefaultRenderTarget(0));
 
         // Create two Textures.
         auto usage = TextureUsage::COLOR_ATTACHMENT | TextureUsage::SAMPLEABLE;
-        Handle<HwTexture> textureA = getDriverApi().createTexture(
+        Handle<HwTexture> textureA = cleanup.add(api.createTexture(
                     SamplerType::SAMPLER_2D,            // target
                     1,                                  // levels
                     TextureFormat::RGBA8,               // format
@@ -86,8 +90,8 @@ TEST_F(BackendTest, MRT) {
                     512,                                // width
                     512,                                // height
                     1,                                  // depth
-                    usage);                             // usage
-        Handle<HwTexture> textureB = getDriverApi().createTexture(
+                    usage));                             // usage
+        Handle<HwTexture> textureB = cleanup.add(api.createTexture(
                     SamplerType::SAMPLER_2D,            // target
                     1,                                  // levels
                     TextureFormat::RGBA8,               // format
@@ -95,10 +99,10 @@ TEST_F(BackendTest, MRT) {
                     512,                                // width
                     512,                                // height
                     1,                                  // depth
-                    usage);                             // usage
+                    usage));                             // usage
 
         // Create a RenderTarget with two attachments.
-        Handle<HwRenderTarget> renderTarget = getDriverApi().createRenderTarget(
+        Handle<HwRenderTarget> renderTarget = cleanup.add(api.createRenderTarget(
                 TargetBufferFlags::COLOR0 | TargetBufferFlags::COLOR1,
                 // The width and height must match the width and height of the respective mip
                 // level (at least for OpenGL).
@@ -108,7 +112,7 @@ TEST_F(BackendTest, MRT) {
                 0,                                         // layerCount
                 {{textureA },{textureB }},                 // color
                 {},                                        // depth
-                {});                                       // stencil
+                {}));                                       // stencil
 
         RenderPassParams params = {};
         fullViewport(params);
@@ -124,25 +128,21 @@ TEST_F(BackendTest, MRT) {
         state.rasterState.depthFunc = RasterState::DepthFunc::A;
         state.rasterState.culling = CullingMode::NONE;
 
-        getDriverApi().startCapture(0);
+        api.startCapture(0);
 
-        getDriverApi().makeCurrent(swapChain, swapChain);
-        getDriverApi().beginFrame(0, 0, 0);
+        api.makeCurrent(swapChain, swapChain);
+        api.beginFrame(0, 0, 0);
 
         // Draw a triangle.
-        getDriverApi().beginRenderPass(renderTarget, params);
-        getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
-        getDriverApi().endRenderPass();
+        api.beginRenderPass(renderTarget, params);
+        api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        api.endRenderPass();
 
-        getDriverApi().flush();
-        getDriverApi().commit(swapChain);
-        getDriverApi().endFrame(0);
+        api.flush();
+        api.commit(swapChain);
+        api.endFrame(0);
 
-        getDriverApi().stopCapture(0);
-
-        getDriverApi().destroyProgram(program);
-        getDriverApi().destroySwapChain(swapChain);
-        getDriverApi().destroyRenderTarget(defaultRenderTarget);
+        api.stopCapture(0);
     }
 
     executeCommands();

filament/backend/test/test_MipLevels.cpp

@@ -17,9 +17,10 @@
 
 #include "BackendTest.h"
 
+#include "BackendTestUtils.h"
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
-#include "BackendTestUtils.h"
 
 #include <backend/DriverEnums.h>
 #include <backend/Handle.h>
@@ -78,12 +79,13 @@ using namespace filament::backend;
 TEST_F(BackendTest, TextureViewLod) {
     auto& api = getDriverApi();
     api.startCapture(0);
+    Cleanup cleanup(api);
 
     // The test is executed within this block scope to force destructors to run before
     // executeCommands().
     {
         // Create a SwapChain and make it current.
-        auto swapChain = createSwapChain();
+        auto swapChain = cleanup.add(createSwapChain());
         api.makeCurrent(swapChain, swapChain);
 
         // Create a program that draws only white.
@@ -92,7 +94,7 @@ TEST_F(BackendTest, TextureViewLod) {
             ShaderGenerator shaderGen(vertex, whiteFragment, sBackend, sIsMobilePlatform);
             Program p = shaderGen.getProgram(api);
             p.descriptorBindings(0, {{"backend_test_sib_tex", DescriptorType::SAMPLER, 0}});
-            whiteProgram = api.createProgram(std::move(p));
+            whiteProgram = cleanup.add(api.createProgram(std::move(p)));
         }
 
         // Create a program that samples a texture.
@@ -106,19 +108,19 @@ TEST_F(BackendTest, TextureViewLod) {
             ShaderGenerator shaderGen(vertex, fragment, sBackend, sIsMobilePlatform, std::move(descriptors));
             Program p = shaderGen.getProgram(api);
             p.descriptorBindings(0, {{"backend_test_sib_tex", DescriptorType::SAMPLER, 0}});
-            textureProgram = api.createProgram(std::move(p));
+            textureProgram = cleanup.add(api.createProgram(std::move(p)));
         }
 
-        DescriptorSetLayoutHandle descriptorSetLayout = api.createDescriptorSetLayout({
+        DescriptorSetLayoutHandle descriptorSetLayout = cleanup.add(api.createDescriptorSetLayout({
                 {{
                          DescriptorType::SAMPLER,
                          ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 0,
                          DescriptorFlags::NONE, 0
-                 }}});
+                 }}}));
 
         DescriptorSetHandle descriptorSet[2];
-        descriptorSet[0] = api.createDescriptorSet(descriptorSetLayout);
-        descriptorSet[1] = api.createDescriptorSet(descriptorSetLayout);
+        descriptorSet[0] = cleanup.add(api.createDescriptorSet(descriptorSetLayout));
+        descriptorSet[1] = cleanup.add(api.createDescriptorSet(descriptorSetLayout));
 
         // Create a texture that has 4 mip levels. Each level is a different color.
         // Level 0: 128x128 (red)
@@ -127,9 +129,9 @@ TEST_F(BackendTest, TextureViewLod) {
         // Level 3:   16x16 (yellow)
         const size_t kTextureSize = 128;
         const size_t kMipLevels = 4;
-        Handle<HwTexture> texture = api.createTexture(SamplerType::SAMPLER_2D, kMipLevels,
+        Handle<HwTexture> texture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, kMipLevels,
                 TextureFormat::RGBA8, 1, kTextureSize, kTextureSize, 1,
-                TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT | TextureUsage::UPLOADABLE);
+                TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT | TextureUsage::UPLOADABLE));
 
         // Create image data.
         auto pixelFormat = PixelDataFormat::RGBA;
@@ -163,13 +165,13 @@ TEST_F(BackendTest, TextureViewLod) {
         // Level 1:   64x64 (green)             <-- base
         // Level 2:   32x32 (blue)              <--- white triangle rendered
         // Level 3:   16x16 (yellow)            <-- max
-        auto texture13 = api.createTextureView(texture, 1, 3);
+        auto texture13 = cleanup.add(api.createTextureView(texture, 1, 3));
 
         // Render a white triangle into level 2.
         // We specify mip level 2, because minMaxLevels has no effect when rendering into a texture.
-        Handle<HwRenderTarget> renderTarget = api.createRenderTarget(
+        Handle<HwRenderTarget> renderTarget = cleanup.add(api.createRenderTarget(
                 TargetBufferFlags::COLOR, 32, 32, 1, 0,
-                {texture, 2 /* level */, 0 /* layer */}, {}, {});
+                {texture, 2 /* level */, 0 /* layer */}, {}, {}));
         {
             RenderPassParams params = {};
             fullViewport(params);
@@ -185,7 +187,7 @@ TEST_F(BackendTest, TextureViewLod) {
             api.endRenderPass();
         }
 
-        backend::Handle<HwRenderTarget> defaultRenderTarget = api.createDefaultRenderTarget(0);
+        backend::Handle<HwRenderTarget> defaultRenderTarget = cleanup.add(api.createDefaultRenderTarget(0));
 
         RenderPassParams params = {};
         fullViewport(params);
@@ -217,7 +219,7 @@ TEST_F(BackendTest, TextureViewLod) {
         api.endRenderPass();
 
         // Adjust the base mip to 2.
-        auto texture22 = api.createTextureView(texture, 2, 2);
+        auto texture22 = cleanup.add(api.createTextureView(texture, 2, 2));
 
         api.updateDescriptorSetTexture(descriptorSet[1], 0, texture22, {
                 .filterMag = SamplerMagFilter::NEAREST,
@@ -244,18 +246,6 @@ TEST_F(BackendTest, TextureViewLod) {
         api.endFrame(0);
 
         api.stopCapture(0);
-
-        // Cleanup.
-        api.destroySwapChain(swapChain);
-        api.destroyRenderTarget(renderTarget);
-        api.destroyDescriptorSet(descriptorSet[0]);
-        api.destroyDescriptorSet(descriptorSet[1]);
-        api.destroyDescriptorSetLayout(descriptorSetLayout);
-        api.destroyTexture(texture);
-        api.destroyTexture(texture13);
-        api.destroyTexture(texture22);
-        api.destroyProgram(whiteProgram);
-        api.destroyProgram(textureProgram);
     }
 
     api.finish();

filament/backend/test/test_MissingRequiredAttributes.cpp

@@ -16,6 +16,7 @@
 
 #include "BackendTest.h"
 
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
 
@@ -70,18 +71,20 @@ TEST_F(BackendTest, MissingRequiredAttributes) {
     // The test is executed within this block scope to force destructors to run before
     // executeCommands().
     {
+        DriverApi& api = getDriverApi();
+        Cleanup cleanup(api);
         // Create a platform-specific SwapChain and make it current.
-        auto swapChain = createSwapChain();
-        getDriverApi().makeCurrent(swapChain, swapChain);
+        auto swapChain = cleanup.add(createSwapChain());
+        api.makeCurrent(swapChain, swapChain);
 
         // Create a program.
         ShaderGenerator shaderGen(vertex, fragment, sBackend, sIsMobilePlatform);
-        Program p = shaderGen.getProgram(getDriverApi());
-        auto program = getDriverApi().createProgram(std::move(p));
+        Program p = shaderGen.getProgram(api);
+        auto program = cleanup.add(api.createProgram(std::move(p)));
 
-        auto defaultRenderTarget = getDriverApi().createDefaultRenderTarget(0);
+        auto defaultRenderTarget = cleanup.add(api.createDefaultRenderTarget(0));
 
-        TrianglePrimitive triangle(getDriverApi());
+        TrianglePrimitive triangle(api);
 
         RenderPassParams params = {};
         fullViewport(params);
@@ -97,25 +100,21 @@ TEST_F(BackendTest, MissingRequiredAttributes) {
         state.rasterState.depthFunc = RasterState::DepthFunc::A;
         state.rasterState.culling = CullingMode::NONE;
 
-        getDriverApi().startCapture(0);
+        api.startCapture(0);
 
-        getDriverApi().makeCurrent(swapChain, swapChain);
-        getDriverApi().beginFrame(0, 0, 0);
+        api.makeCurrent(swapChain, swapChain);
+        api.beginFrame(0, 0, 0);
 
         // Render a triangle.
-        getDriverApi().beginRenderPass(defaultRenderTarget, params);
-        getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
-        getDriverApi().endRenderPass();
+        api.beginRenderPass(defaultRenderTarget, params);
+        api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        api.endRenderPass();
 
-        getDriverApi().flush();
-        getDriverApi().commit(swapChain);
-        getDriverApi().endFrame(0);
+        api.flush();
+        api.commit(swapChain);
+        api.endFrame(0);
 
-        getDriverApi().stopCapture(0);
-
-        getDriverApi().destroyProgram(program);
-        getDriverApi().destroySwapChain(swapChain);
-        getDriverApi().destroyRenderTarget(defaultRenderTarget);
+        api.stopCapture(0);
     }
 
     executeCommands();

filament/backend/test/test_PushConstants.cpp

@@ -16,6 +16,7 @@
 
 #include "BackendTest.h"
 
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
 
@@ -80,20 +81,21 @@ TEST_F(BackendTest, PushConstants) {
     auto& api = getDriverApi();
 
     api.startCapture(0);
+    Cleanup cleanup(api);
 
     // The test is executed within this block scope to force destructors to run before
     // executeCommands().
     {
         // Create a SwapChain and make it current.
-        auto swapChain = createSwapChain();
+        auto swapChain = cleanup.add(createSwapChain());
         api.makeCurrent(swapChain, swapChain);
 
         // Create a program.
         ShaderGenerator shaderGen(triangleVs, triangleFs, sBackend, sIsMobilePlatform);
         Program p = shaderGen.getProgram(api);
-        ProgramHandle program = api.createProgram(std::move(p));
+        ProgramHandle program = cleanup.add(api.createProgram(std::move(p)));
 
-        Handle<HwRenderTarget> renderTarget = api.createDefaultRenderTarget();
+        Handle<HwRenderTarget> renderTarget = cleanup.add(api.createDefaultRenderTarget());
 
         TrianglePrimitive triangle(api);
 
@@ -150,10 +152,6 @@ TEST_F(BackendTest, PushConstants) {
 
         api.commit(swapChain);
         api.endFrame(0);
-
-        // Cleanup.
-        api.destroySwapChain(swapChain);
-        api.destroyRenderTarget(renderTarget);
     }
 
     api.stopCapture(0);

filament/backend/test/test_ReadPixels.cpp

@@ -16,9 +16,10 @@
 
 #include "BackendTest.h"
 
+#include "BackendTestUtils.h"
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
-#include "BackendTestUtils.h"
 
 #include <utils/Hash.h>
 
@@ -236,17 +237,20 @@ TEST_F(ReadPixelsTest, ReadPixels) {
 
     TestCase const testCases[] = { t0, t2, t3, t4, t5, t6, t7, t8 };
 
+    DriverApi& api = getDriverApi();
+    Cleanup cleanup(api);
+
     // Create programs.
     Handle<HwProgram> programFloat, programUint;
     {
         ShaderGenerator shaderGen(vertex, fragmentFloat, sBackend, sIsMobilePlatform);
-        Program p = shaderGen.getProgram(getDriverApi());
-        programFloat = getDriverApi().createProgram(std::move(p));
+        Program p = shaderGen.getProgram(api);
+        programFloat = cleanup.add(api.createProgram(std::move(p)));
     }
     {
         ShaderGenerator shaderGen(vertex, fragmentUint, sBackend, sIsMobilePlatform);
-        Program p = shaderGen.getProgram(getDriverApi());
-        programUint = getDriverApi().createProgram(std::move(p));
+        Program p = shaderGen.getProgram(api);
+        programUint = cleanup.add(api.createProgram(std::move(p)));
     }
 
 
@@ -255,35 +259,35 @@ TEST_F(ReadPixelsTest, ReadPixels) {
         // Create a platform-specific SwapChain and make it current.
         Handle<HwSwapChain> swapChain;
         if (t.useDefaultRT) {
-            swapChain = createSwapChain();
+            swapChain = cleanup.add(createSwapChain());
         } else {
-            swapChain = getDriverApi().createSwapChainHeadless(t.getRenderTargetSize(),
-                    t.getRenderTargetSize(), 0);
+            swapChain = cleanup.add(api.createSwapChainHeadless(t.getRenderTargetSize(),
+                    t.getRenderTargetSize(), 0));
         }
 
-        getDriverApi().makeCurrent(swapChain, swapChain);
+        api.makeCurrent(swapChain, swapChain);
 
         // Create a Texture and RenderTarget to render into.
         auto usage = TextureUsage::COLOR_ATTACHMENT | TextureUsage::SAMPLEABLE;
-        Handle<HwTexture> const texture = getDriverApi().createTexture(SamplerType::SAMPLER_2D,
+        Handle<HwTexture> const texture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D,
                 t.mipLevels, t.textureFormat, 1, renderTargetBaseSize, renderTargetBaseSize, 1,
-                usage);
+                usage));
 
         Handle<HwRenderTarget> renderTarget;
         if (t.useDefaultRT) {
             // The width and height must match the width and height of the respective mip
             // level (at least for OpenGL).
-            renderTarget = getDriverApi().createDefaultRenderTarget();
+            renderTarget = cleanup.add(api.createDefaultRenderTarget());
         } else {
             // The width and height must match the width and height of the respective mip
             // level (at least for OpenGL).
-            renderTarget = getDriverApi().createRenderTarget(
+            renderTarget = cleanup.add(api.createRenderTarget(
                     TargetBufferFlags::COLOR, t.getRenderTargetSize(),
                     t.getRenderTargetSize(), t.samples, 0, {{ texture, uint8_t(t.mipLevel) }}, {},
-                    {});
+                    {}));
         }
 
-        TrianglePrimitive const triangle(getDriverApi());
+        TrianglePrimitive const triangle(api);
 
         RenderPassParams params = {};
         fullViewport(params);
@@ -294,11 +298,11 @@ TEST_F(ReadPixelsTest, ReadPixels) {
         params.viewport.height = t.getRenderTargetSize();
         params.viewport.width = t.getRenderTargetSize();
 
-        getDriverApi().makeCurrent(swapChain, swapChain);
-        getDriverApi().beginFrame(0, 0, 0);
+        api.makeCurrent(swapChain, swapChain);
+        api.beginFrame(0, 0, 0);
 
         // Render a white triangle over blue.
-        getDriverApi().beginRenderPass(renderTarget, params);
+        api.beginRenderPass(renderTarget, params);
 
         PipelineState state;
         state.program = programFloat;
@@ -309,21 +313,21 @@ TEST_F(ReadPixelsTest, ReadPixels) {
         state.rasterState.depthWrite = false;
         state.rasterState.depthFunc = RasterState::DepthFunc::A;
         state.rasterState.culling = CullingMode::NONE;
-        getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
 
-        getDriverApi().endRenderPass();
+        api.endRenderPass();
 
         if (t.mipLevel > 0) {
+            Cleanup localCleanup(api);
             // Render red to the first mip level to check that the backend is actually reading the
             // correct mip.
             RenderPassParams p = params;
-            Handle<HwRenderTarget> mipLevelOneRT = getDriverApi().createRenderTarget(
+            Handle<HwRenderTarget> mipLevelOneRT = localCleanup.add(api.createRenderTarget(
                     TargetBufferFlags::COLOR, renderTargetBaseSize, renderTargetBaseSize, 1, 0,
-                    {{ texture }}, {}, {});
+                    {{ texture }}, {}, {}));
             p.clearColor = {1.f, 0.f, 0.f, 1.f};
-            getDriverApi().beginRenderPass(mipLevelOneRT, p);
-            getDriverApi().endRenderPass();
-            getDriverApi().destroyRenderTarget(mipLevelOneRT);
+            api.beginRenderPass(mipLevelOneRT, p);
+            api.endRenderPass();
         }
 
         // Read pixels.
@@ -347,26 +351,19 @@ TEST_F(ReadPixelsTest, ReadPixels) {
                     free(buffer);
                 }, (void*) &t);
 
-        getDriverApi().readPixels(renderTarget, t.readRect.x, t.readRect.y, t.readRect.width,
+        api.readPixels(renderTarget, t.readRect.x, t.readRect.y, t.readRect.width,
                 t.readRect.height, std::move(descriptor));
 
         // Now render red over what was just rendered. This ensures that readPixels captures the
         // state of rendering between render passes.
         params.clearColor = {1.f, 0.f, 0.f, 1.f};
-        getDriverApi().beginRenderPass(renderTarget, params);
-        getDriverApi().endRenderPass();
+        api.beginRenderPass(renderTarget, params);
+        api.endRenderPass();
 
-        getDriverApi().commit(swapChain);
-        getDriverApi().endFrame(0);
-
-        getDriverApi().destroySwapChain(swapChain);
-        getDriverApi().destroyRenderTarget(renderTarget);
-        getDriverApi().destroyTexture(texture);
+        api.commit(swapChain);
+        api.endFrame(0);
     }
 
-    getDriverApi().destroyProgram(programFloat);
-    getDriverApi().destroyProgram(programUint);
-
     // This ensures all driver commands have finished before exiting the test.
     flushAndWait();
 }
@@ -375,18 +372,21 @@ TEST_F(ReadPixelsTest, ReadPixelsPerformance) {
     const size_t renderTargetSize = 2000;
     const int iterationCount = 100;
 
+    DriverApi& api = getDriverApi();
+    Cleanup cleanup(api);
+
     // Create a platform-specific SwapChain and make it current.
-    auto swapChain = getDriverApi().createSwapChainHeadless(renderTargetSize, renderTargetSize, 0);
-    getDriverApi().makeCurrent(swapChain, swapChain);
+    auto swapChain = cleanup.add(api.createSwapChainHeadless(renderTargetSize, renderTargetSize, 0));
+    api.makeCurrent(swapChain, swapChain);
 
     // Create a program.
     ShaderGenerator shaderGen(vertex, fragmentFloat, sBackend, sIsMobilePlatform);
-    Program p = shaderGen.getProgram(getDriverApi());
-    auto program = getDriverApi().createProgram(std::move(p));
+    Program p = shaderGen.getProgram(api);
+    auto program = cleanup.add(api.createProgram(std::move(p)));
 
     // Create a Texture and RenderTarget to render into.
     auto usage = TextureUsage::COLOR_ATTACHMENT | TextureUsage::SAMPLEABLE;
-    Handle<HwTexture> texture = getDriverApi().createTexture(
+    Handle<HwTexture> texture = cleanup.add(api.createTexture(
                 SamplerType::SAMPLER_2D,            // target
                 1,                                  // levels
                 TextureFormat::RGBA8,               // format
@@ -394,9 +394,9 @@ TEST_F(ReadPixelsTest, ReadPixelsPerformance) {
                 renderTargetSize,                   // width
                 renderTargetSize,                   // height
                 1,                                  // depth
-                usage);                             // usage
+                usage));                             // usage
 
-    Handle<HwRenderTarget> renderTarget = getDriverApi().createRenderTarget(
+    Handle<HwRenderTarget> renderTarget = cleanup.add(api.createRenderTarget(
             TargetBufferFlags::COLOR,
             renderTargetSize,                          // width
             renderTargetSize,                          // height
@@ -404,9 +404,9 @@ TEST_F(ReadPixelsTest, ReadPixelsPerformance) {
             0,                                         // layerCount
             {{ texture }},                             // color
             {},                                        // depth
-            {});                                       // stencil
+            {}));                                       // stencil
 
-    TrianglePrimitive triangle(getDriverApi());
+    TrianglePrimitive triangle(api);
 
     RenderPassParams params = {};
     fullViewport(params);
@@ -433,13 +433,13 @@ TEST_F(ReadPixelsTest, ReadPixelsPerformance) {
             printf("Executing test %d / %d\n", iteration, iterationCount);
         }
 
-        getDriverApi().makeCurrent(swapChain, swapChain);
-        getDriverApi().beginFrame(0, 0, 0);
+        api.makeCurrent(swapChain, swapChain);
+        api.beginFrame(0, 0, 0);
 
         // Render some content, just so we don't read back uninitialized data.
-        getDriverApi().beginRenderPass(renderTarget, params);
-        getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
-        getDriverApi().endRenderPass();
+        api.beginRenderPass(renderTarget, params);
+        api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        api.endRenderPass();
 
         PixelBufferDescriptor descriptor(buffer, renderTargetSize * renderTargetSize * 4,
                 PixelDataFormat::RGBA, PixelDataType::UBYTE, 1, 0, 0, renderTargetSize,
@@ -448,9 +448,9 @@ TEST_F(ReadPixelsTest, ReadPixelsPerformance) {
                     test->readPixelsFinished = true;
                 }, this);
 
-        getDriverApi().readPixels(renderTarget, 0, 0, renderTargetSize, renderTargetSize, std::move(descriptor));
-        getDriverApi().commit(swapChain);
-        getDriverApi().endFrame(0);
+        api.readPixels(renderTarget, 0, 0, renderTargetSize, renderTargetSize, std::move(descriptor));
+        api.commit(swapChain);
+        api.endFrame(0);
 
         flushAndWait();
         getDriver().purge();
@@ -460,11 +460,7 @@ TEST_F(ReadPixelsTest, ReadPixelsPerformance) {
 
     free(buffer);
 
-    getDriverApi().destroyProgram(program);
-    getDriverApi().destroySwapChain(swapChain);
-    getDriverApi().destroyRenderTarget(renderTarget);
-    getDriverApi().destroyTexture(texture);
-    getDriverApi().finish();
+    api.finish();
     executeCommands();
 }
 

filament/backend/test/test_RenderExternalImage.cpp

@@ -16,6 +16,7 @@
 
 #include "BackendTest.h"
 
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
 
@@ -66,10 +67,11 @@ using namespace filament::backend;
 // Rendering an external image without setting any data should not crash.
 TEST_F(BackendTest, RenderExternalImageWithoutSet) {
     auto& api = getDriverApi();
+    Cleanup cleanup(api);
 
     TrianglePrimitive triangle(api);
 
-    auto swapChain = createSwapChain();
+    auto swapChain = cleanup.add(createSwapChain());
 
     filament::SamplerInterfaceBlock::SamplerInfo samplerInfo { "test", "tex", 0,
         SamplerType::SAMPLER_EXTERNAL, SamplerFormat::FLOAT, Precision::HIGH, false };
@@ -82,22 +84,22 @@ TEST_F(BackendTest, RenderExternalImageWithoutSet) {
     // Create a program that samples a texture.
     Program p = shaderGen.getProgram(api);
     p.descriptorBindings(1, {{"test_tex", DescriptorType::SAMPLER, 0}});
-    backend::Handle<HwProgram> program = api.createProgram(std::move(p));
-    DescriptorSetLayoutHandle descriptorSetLayout = api.createDescriptorSetLayout({
+    backend::Handle<HwProgram> program = cleanup.add(api.createProgram(std::move(p)));
+    DescriptorSetLayoutHandle descriptorSetLayout = cleanup.add(api.createDescriptorSetLayout({
             {{
                      DescriptorType::SAMPLER,
                      ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 0,
                      DescriptorFlags::NONE, 0
-             }}});
+             }}}));
 
-    DescriptorSetHandle descriptorSet = api.createDescriptorSet(descriptorSetLayout);
+    DescriptorSetHandle descriptorSet = cleanup.add(api.createDescriptorSet(descriptorSetLayout));
 
-    backend::Handle<HwRenderTarget> defaultRenderTarget = api.createDefaultRenderTarget(0);
+    backend::Handle<HwRenderTarget> defaultRenderTarget = cleanup.add(api.createDefaultRenderTarget(0));
 
     // Create a texture that will be backed by an external image.
     auto usage = TextureUsage::COLOR_ATTACHMENT | TextureUsage::SAMPLEABLE;
     const NativeView& view = getNativeView();
-    backend::Handle<HwTexture> texture = api.createTexture(
+    backend::Handle<HwTexture> texture = cleanup.add(api.createTexture(
                 SamplerType::SAMPLER_EXTERNAL,      // target
                 1,                                  // levels
                 TextureFormat::RGBA8,               // format
@@ -105,7 +107,7 @@ TEST_F(BackendTest, RenderExternalImageWithoutSet) {
                 view.width,                         // width
                 view.height,                        // height
                 1,                                  // depth
-                usage);                             // usage
+                usage));                             // usage
 
     RenderPassParams params = {};
     fullViewport(params);
@@ -140,15 +142,6 @@ TEST_F(BackendTest, RenderExternalImageWithoutSet) {
 
     api.stopCapture(0);
 
-    // Delete our resources.
-    api.destroyDescriptorSet(descriptorSet);
-    api.destroyDescriptorSetLayout(descriptorSetLayout);
-    api.destroyTexture(texture);
-
-    // Destroy frame resources.
-    api.destroyProgram(program);
-    api.destroyRenderTarget(defaultRenderTarget);
-
     api.finish();
 
     executeCommands();
@@ -156,10 +149,11 @@ TEST_F(BackendTest, RenderExternalImageWithoutSet) {
 
 TEST_F(BackendTest, RenderExternalImage) {
     auto& api = getDriverApi();
+    Cleanup cleanup(api);
 
     TrianglePrimitive triangle(api);
 
-    auto swapChain = createSwapChain();
+    auto swapChain = cleanup.add(createSwapChain());
 
     filament::SamplerInterfaceBlock::SamplerInfo samplerInfo { "test", "tex", 0,
         SamplerType::SAMPLER_EXTERNAL, SamplerFormat::FLOAT, Precision::HIGH, false };
@@ -172,18 +166,18 @@ TEST_F(BackendTest, RenderExternalImage) {
     // Create a program that samples a texture.
     Program p = shaderGen.getProgram(api);
     p.descriptorBindings(1, {{"test_tex", DescriptorType::SAMPLER, 0}});
-    auto program = api.createProgram(std::move(p));
+    auto program = cleanup.add(api.createProgram(std::move(p)));
 
-    DescriptorSetLayoutHandle descriptorSetLayout = api.createDescriptorSetLayout({
+    DescriptorSetLayoutHandle descriptorSetLayout = cleanup.add(api.createDescriptorSetLayout({
             {{
                      DescriptorType::SAMPLER,
                      ShaderStageFlags::ALL_SHADER_STAGE_FLAGS, 0,
                      DescriptorFlags::NONE, 0
-             }}});
+             }}}));
 
-    DescriptorSetHandle descriptorSet = api.createDescriptorSet(descriptorSetLayout);
+    DescriptorSetHandle descriptorSet = cleanup.add(api.createDescriptorSet(descriptorSetLayout));
 
-    backend::Handle<HwRenderTarget> defaultRenderTarget = api.createDefaultRenderTarget(0);
+    backend::Handle<HwRenderTarget> defaultRenderTarget = cleanup.add(api.createDefaultRenderTarget(0));
 
     // require users to create two Filament textures and have two material parameters
     // add a "plane" parameter to setExternalImage
@@ -227,8 +221,8 @@ TEST_F(BackendTest, RenderExternalImage) {
 
     api.setupExternalImage(pixBuffer);
     backend::Handle<HwTexture> texture =
-            api.createTextureExternalImage(SamplerType::SAMPLER_EXTERNAL,
-                    TextureFormat::RGBA8, 1024, 1024, usage, pixBuffer);
+            cleanup.add(api.createTextureExternalImage(SamplerType::SAMPLER_EXTERNAL,
+                    TextureFormat::RGBA8, 1024, 1024, usage, pixBuffer));
 
     // We're now free to release the buffer.
     CVBufferRelease(pixBuffer);
@@ -269,15 +263,6 @@ TEST_F(BackendTest, RenderExternalImage) {
 
     api.stopCapture(0);
 
-    // Delete our resources.
-    api.destroyDescriptorSet(descriptorSet);
-    api.destroyDescriptorSetLayout(descriptorSetLayout);
-    api.destroyTexture(texture);
-
-    // Destroy frame resources.
-    api.destroyProgram(program);
-    api.destroyRenderTarget(defaultRenderTarget);
-
     api.finish();
 
     executeCommands();

filament/backend/test/test_Scissor.cpp

@@ -16,6 +16,7 @@
 
 #include "BackendTest.h"
 
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
 
@@ -55,6 +56,7 @@ TEST_F(BackendTest, ScissorViewportRegion) {
     constexpr int kSrcRtHeight = 384;
 
     api.startCapture(0);
+    Cleanup cleanup(api);
 
     //    color texture (mip level 1) 512x512           depth texture (mip level 0) 512x512
     // +----------------------------------------+   +------------------------------------------+
@@ -80,21 +82,21 @@ TEST_F(BackendTest, ScissorViewportRegion) {
     // executeCommands().
     {
         // Create a SwapChain and make it current. We don't really use it so the res doesn't matter.
-        auto swapChain = api.createSwapChainHeadless(256, 256, 0);
+        auto swapChain = cleanup.add(api.createSwapChainHeadless(256, 256, 0));
         api.makeCurrent(swapChain, swapChain);
 
         // Create a program.
         ShaderGenerator shaderGen(triangleVs, triangleFs, sBackend, sIsMobilePlatform);
         Program p = shaderGen.getProgram(api);
-        ProgramHandle program = api.createProgram(std::move(p));
+        ProgramHandle program = cleanup.add(api.createProgram(std::move(p)));
 
         // Create source color and depth textures.
-        Handle<HwTexture> srcTexture = api.createTexture(SamplerType::SAMPLER_2D, kNumLevels,
+        Handle<HwTexture> srcTexture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, kNumLevels,
                 kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
-                TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT);
-        Handle<HwTexture> depthTexture = api.createTexture(SamplerType::SAMPLER_2D, 1,
+                TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT));
+        Handle<HwTexture> depthTexture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
                 TextureFormat::DEPTH16, 1, 512, 512, 1,
-                TextureUsage::DEPTH_ATTACHMENT);
+                TextureUsage::DEPTH_ATTACHMENT));
 
         // Render into the bottom-left quarter of the texture.
         Viewport srcRect = {
@@ -112,13 +114,13 @@ TEST_F(BackendTest, ScissorViewportRegion) {
 
         // We purposely set the render target width and height to smaller than the texture, to check
         // that this case is handled correctly.
-        Handle<HwRenderTarget> srcRenderTarget = api.createRenderTarget(
+        Handle<HwRenderTarget> srcRenderTarget = cleanup.add(api.createRenderTarget(
                 TargetBufferFlags::COLOR | TargetBufferFlags::DEPTH, kSrcRtHeight, kSrcRtHeight, 1, 0,
-                {srcTexture, kSrcLevel, 0}, {depthTexture, 0, 0}, {});
+                {srcTexture, kSrcLevel, 0}, {depthTexture, 0, 0}, {}));
 
-        Handle<HwRenderTarget> fullRenderTarget = api.createRenderTarget(TargetBufferFlags::COLOR,
+        Handle<HwRenderTarget> fullRenderTarget = cleanup.add(api.createRenderTarget(TargetBufferFlags::COLOR,
                 kSrcTexHeight >> kSrcLevel, kSrcTexWidth >> kSrcLevel, 1, 0,
-                {srcTexture, kSrcLevel, 0}, {}, {});
+                {srcTexture, kSrcLevel, 0}, {}, {}));
 
         TrianglePrimitive triangle(api);
 
@@ -150,11 +152,6 @@ TEST_F(BackendTest, ScissorViewportRegion) {
         api.endFrame(0);
 
         api.stopCapture(0);
-
-        // Cleanup.
-        api.destroyTexture(srcTexture);
-        api.destroySwapChain(swapChain);
-        api.destroyRenderTarget(srcRenderTarget);
     }
 
     // Wait for the ReadPixels result to come back.
@@ -169,23 +166,24 @@ TEST_F(BackendTest, ScissorViewportEdgeCases) {
     auto& api = getDriverApi();
 
     api.startCapture(0);
+    Cleanup cleanup(api);
 
     // The test is executed within this block scope to force destructors to run before
     // executeCommands().
     {
         // Create a SwapChain and make it current. We don't really use it so the res doesn't matter.
-        auto swapChain = api.createSwapChainHeadless(256, 256, 0);
+        auto swapChain = cleanup.add(api.createSwapChainHeadless(256, 256, 0));
         api.makeCurrent(swapChain, swapChain);
 
         // Create a program.
         ShaderGenerator shaderGen(triangleVs, triangleFs, sBackend, sIsMobilePlatform);
         Program p = shaderGen.getProgram(api);
-        ProgramHandle program = api.createProgram(std::move(p));
+        ProgramHandle program = cleanup.add(api.createProgram(std::move(p)));
 
         // Create a source color textures.
-        Handle<HwTexture> srcTexture = api.createTexture(SamplerType::SAMPLER_2D, 1,
+        Handle<HwTexture> srcTexture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
                 TextureFormat::RGBA8, 1, 512, 512, 1,
-                TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT);
+                TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT));
 
         // Render into the bottom-left quarter of the texture, checking 3 special cases.
         // 1. negative viewport left/bottom
@@ -207,9 +205,9 @@ TEST_F(BackendTest, ScissorViewportEdgeCases) {
         Viewport scissor = {0, 0, (uint32_t)std::numeric_limits<int32_t>::max(),
                 (uint32_t)std::numeric_limits<int32_t>::max()};
 
-        Handle<HwRenderTarget> renderTarget = api.createRenderTarget(
+        Handle<HwRenderTarget> renderTarget = cleanup.add(api.createRenderTarget(
                 TargetBufferFlags::COLOR, 512, 512, 1, 0,
-                {srcTexture, 0, 0}, {}, {});
+                {srcTexture, 0, 0}, {}, {}));
 
         TrianglePrimitive triangle(api);
 
@@ -249,11 +247,6 @@ TEST_F(BackendTest, ScissorViewportEdgeCases) {
         api.endFrame(0);
 
         api.stopCapture(0);
-
-        // Cleanup.
-        api.destroyTexture(srcTexture);
-        api.destroySwapChain(swapChain);
-        api.destroyRenderTarget(renderTarget);
     }
 
     // Wait for the ReadPixels result to come back.

filament/backend/test/test_StencilBuffer.cpp

@@ -16,6 +16,7 @@
 
 #include "BackendTest.h"
 
+#include "Lifetimes.h"
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
 
@@ -63,20 +64,23 @@ namespace test {
 class BasicStencilBufferTest : public BackendTest {
 public:
 
-    Handle<HwSwapChain> swapChain;
-    ProgramHandle program;
+    Handle<HwSwapChain> mSwapChain;
+    ProgramHandle mProgram;
+    Cleanup mCleanup;
+
+    BasicStencilBufferTest() : mCleanup(getDriverApi()) {}
 
     void SetUp() override {
         auto& api = getDriverApi();
 
         // Create a platform-specific SwapChain and make it current.
-        swapChain = createSwapChain();
-        api.makeCurrent(swapChain, swapChain);
+        mSwapChain = mCleanup.add(createSwapChain());
+        api.makeCurrent(mSwapChain, mSwapChain);
 
         // Create a program.
         ShaderGenerator shaderGen(vertex, fragment, sBackend, sIsMobilePlatform);
         Program p = shaderGen.getProgram(api);
-        program = api.createProgram(std::move(p));
+        mProgram = mCleanup.add(api.createProgram(std::move(p)));
     }
 
     void RunTest(Handle<HwRenderTarget> renderTarget) {
@@ -103,13 +107,13 @@ public:
         params.flags.discardEnd = TargetBufferFlags::NONE;
 
         PipelineState ps = {};
-        ps.program = program;
+        ps.program = mProgram;
         ps.rasterState.colorWrite = false;
         ps.rasterState.depthWrite = false;
         ps.stencilState.stencilWrite = true;
         ps.stencilState.front.stencilOpDepthStencilPass = StencilOperation::INCR;
 
-        api.makeCurrent(swapChain, swapChain);
+        api.makeCurrent(mSwapChain, mSwapChain);
         api.beginFrame(0, 0, 0);
 
         api.beginRenderPass(renderTarget, params);
@@ -129,29 +133,27 @@ public:
         api.draw(ps, triangle.getRenderPrimitive(), 0, 3, 1);
         api.endRenderPass();
 
-        api.commit(swapChain);
+        api.commit(mSwapChain);
         api.endFrame(0);
     }
 
     void TearDown() override {
-        auto& api = getDriverApi();
-        api.destroyProgram(program);
-        api.destroySwapChain(swapChain);
         flushAndWait();
     }
 };
 
 TEST_F(BasicStencilBufferTest, StencilBuffer) {
     auto& api = getDriverApi();
+    Cleanup cleanup(api);
 
     // Create two textures: a color and a stencil, and an associated RenderTarget.
-    auto colorTexture = api.createTexture(SamplerType::SAMPLER_2D, 1,
-            TextureFormat::RGBA8, 1, 512, 512, 1, TextureUsage::COLOR_ATTACHMENT);
-    auto stencilTexture = api.createTexture(SamplerType::SAMPLER_2D, 1,
-            TextureFormat::STENCIL8, 1, 512, 512, 1, TextureUsage::STENCIL_ATTACHMENT);
-    auto renderTarget = getDriverApi().createRenderTarget(
+    auto colorTexture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
+            TextureFormat::RGBA8, 1, 512, 512, 1, TextureUsage::COLOR_ATTACHMENT));
+    auto stencilTexture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
+            TextureFormat::STENCIL8, 1, 512, 512, 1, TextureUsage::STENCIL_ATTACHMENT));
+    auto renderTarget = cleanup.add(api.createRenderTarget(
             TargetBufferFlags::COLOR0 | TargetBufferFlags::STENCIL, 512, 512, 1, 0,
-            {{colorTexture}}, {}, {{stencilTexture}});
+            {{colorTexture}}, {}, {{stencilTexture}}));
 
     RunTest(renderTarget);
 
@@ -159,23 +161,20 @@ TEST_F(BasicStencilBufferTest, StencilBuffer) {
 
     flushAndWait();
     getDriver().purge();
-
-    api.destroyTexture(colorTexture);
-    api.destroyTexture(stencilTexture);
-    api.destroyRenderTarget(renderTarget);
 }
 
 TEST_F(BasicStencilBufferTest, DepthAndStencilBuffer) {
     auto& api = getDriverApi();
+    Cleanup cleanup(api);
 
     // Create two textures: a color and a stencil, and an associated RenderTarget.
-    auto colorTexture = api.createTexture(SamplerType::SAMPLER_2D, 1,
-            TextureFormat::RGBA8, 1, 512, 512, 1, TextureUsage::COLOR_ATTACHMENT);
-    auto depthStencilTexture = api.createTexture(SamplerType::SAMPLER_2D, 1,
-            TextureFormat::DEPTH24_STENCIL8, 1, 512, 512, 1, TextureUsage::STENCIL_ATTACHMENT | TextureUsage::DEPTH_ATTACHMENT);
-    auto renderTarget = getDriverApi().createRenderTarget(
+    auto colorTexture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
+            TextureFormat::RGBA8, 1, 512, 512, 1, TextureUsage::COLOR_ATTACHMENT));
+    auto depthStencilTexture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
+            TextureFormat::DEPTH24_STENCIL8, 1, 512, 512, 1, TextureUsage::STENCIL_ATTACHMENT | TextureUsage::DEPTH_ATTACHMENT));
+    auto renderTarget = cleanup.add(api.createRenderTarget(
             TargetBufferFlags::COLOR0 | TargetBufferFlags::STENCIL, 512, 512, 1, 0,
-            {{colorTexture}}, {depthStencilTexture}, {{depthStencilTexture}});
+            {{colorTexture}}, {depthStencilTexture}, {{depthStencilTexture}}));
 
     RunTest(renderTarget);
 
@@ -183,30 +182,27 @@ TEST_F(BasicStencilBufferTest, DepthAndStencilBuffer) {
 
     flushAndWait();
     getDriver().purge();
-
-    api.destroyTexture(colorTexture);
-    api.destroyTexture(depthStencilTexture);
-    api.destroyRenderTarget(renderTarget);
 }
 
 TEST_F(BasicStencilBufferTest, StencilBufferMSAA) {
     auto& api = getDriverApi();
+    Cleanup cleanup(api);
 
     // Create two textures: a single-sampled color and a MSAA stencil texture.
     // We also create two RenderTargets, one for each pass:
     // Pass 0: Render a triangle only into the MSAA stencil buffer.
     // Pass 1: Render a triangle into (an auto-created) MSAA color buffer using the stencil test.
     //         Performs an auto-resolve on the color.
-    auto colorTexture = api.createTexture(SamplerType::SAMPLER_2D, 1,
-            TextureFormat::RGBA8, 1, 512, 512, 1, TextureUsage::COLOR_ATTACHMENT | TextureUsage::SAMPLEABLE);
-    auto depthStencilTextureMSAA = api.createTexture(SamplerType::SAMPLER_2D, 1,
-            TextureFormat::DEPTH24_STENCIL8, 4, 512, 512, 1, TextureUsage::STENCIL_ATTACHMENT | TextureUsage::DEPTH_ATTACHMENT);
-    auto renderTarget0 = getDriverApi().createRenderTarget(
+    auto colorTexture = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
+            TextureFormat::RGBA8, 1, 512, 512, 1, TextureUsage::COLOR_ATTACHMENT | TextureUsage::SAMPLEABLE));
+    auto depthStencilTextureMSAA = cleanup.add(api.createTexture(SamplerType::SAMPLER_2D, 1,
+            TextureFormat::DEPTH24_STENCIL8, 4, 512, 512, 1, TextureUsage::STENCIL_ATTACHMENT | TextureUsage::DEPTH_ATTACHMENT));
+    auto renderTarget0 = cleanup.add(api.createRenderTarget(
             TargetBufferFlags::DEPTH_AND_STENCIL, 512, 512, 4, 0,
-            {{}}, {depthStencilTextureMSAA}, {depthStencilTextureMSAA});
-    auto renderTarget1 = getDriverApi().createRenderTarget(
+            {{}}, {depthStencilTextureMSAA}, {depthStencilTextureMSAA}));
+    auto renderTarget1 = cleanup.add(api.createRenderTarget(
             TargetBufferFlags::COLOR0 | TargetBufferFlags::DEPTH_AND_STENCIL, 512, 512, 4, 0,
-            {{colorTexture}}, {depthStencilTextureMSAA}, {depthStencilTextureMSAA});
+            {{colorTexture}}, {depthStencilTextureMSAA}, {depthStencilTextureMSAA}));
 
     api.startCapture(0);
 
@@ -230,13 +226,13 @@ TEST_F(BasicStencilBufferTest, StencilBufferMSAA) {
     params.flags.discardEnd = TargetBufferFlags::NONE;
 
     PipelineState ps = {};
-    ps.program = program;
+    ps.program = mProgram;
     ps.rasterState.colorWrite = false;
     ps.rasterState.depthWrite = false;
     ps.stencilState.stencilWrite = true;
     ps.stencilState.front.stencilOpDepthStencilPass = StencilOperation::INCR;
 
-    api.makeCurrent(swapChain, swapChain);
+    api.makeCurrent(mSwapChain, mSwapChain);
     api.beginFrame(0, 0, 0);
 
     api.beginRenderPass(renderTarget0, params);
@@ -258,7 +254,7 @@ TEST_F(BasicStencilBufferTest, StencilBufferMSAA) {
     api.draw(ps, triangle.getRenderPrimitive(), 0, 3, 1);
     api.endRenderPass();
 
-    api.commit(swapChain);
+    api.commit(mSwapChain);
     api.stopCapture(0);
     api.endFrame(0);
 
@@ -266,11 +262,6 @@ TEST_F(BasicStencilBufferTest, StencilBufferMSAA) {
 
     flushAndWait();
     getDriver().purge();
-
-    api.destroyTexture(colorTexture);
-    api.destroyTexture(depthStencilTextureMSAA);
-    api.destroyRenderTarget(renderTarget0);
-    api.destroyRenderTarget(renderTarget1);
 }
 
 } // namespace test

filament/include/filament/BufferObject.h

@@ -25,6 +25,7 @@
 #include <backend/BufferDescriptor.h>
 
 #include <utils/compiler.h>
+#include <utils/StaticString.h>
 
 #include <stdint.h>
 #include <stddef.h>
@@ -90,9 +91,21 @@ public:
          * @param name A string to identify this BufferObject
          * @param len Length of name, should be less than or equal to 128
          * @return This Builder, for chaining calls.
+         * @deprecated Use name(utils::StaticString const&) instead.
          */
+        UTILS_DEPRECATED
         Builder& name(const char* UTILS_NONNULL name, size_t len) noexcept;
 
+        /**
+         * Associate an optional name with this BufferObject for debugging purposes.
+         *
+         * name will show in error messages and should be kept as short as possible.
+         *
+         * @param name A string literal to identify this BufferObject
+         * @return This Builder, for chaining calls.
+         */
+        Builder& name(utils::StaticString const& name) noexcept;
+
         /**
          * Creates the BufferObject and returns a pointer to it. After creation, the buffer
          * object is uninitialized. Use BufferObject::setBuffer() to initialize it.

filament/include/filament/ColorGrading.h

@@ -64,9 +64,8 @@ class ColorSpace;
  * Performance
  * ===========
  *
- * Creating a new ColorGrading object may be more expensive than other Filament objects as a
- * 3D LUT may need to be generated. The generation of a 3D LUT, if necessary, may happen on
- * the CPU.
+ * Creating a new ColorGrading object may be more expensive than other Filament objects as a LUT may
+ * need to be generated. The generation of this LUT, if necessary, may happen on the CPU.
  *
  * Ordering
  * ========
@@ -155,6 +154,9 @@ public:
          * 3D texture. For instance, a low quality level will use a 16x16x16 10 bit LUT, a medium
          * quality level will use a 32x32x32 10 bit LUT, a high quality will use a 32x32x32 16 bit
          * LUT, and a ultra quality will use a 64x64x64 16 bit LUT.
+         *
+         * This setting has no effect if generating a 1D LUT.
+         *
          * This overrides the values set by format() and dimensions().
          *
          * The default quality is medium.
@@ -169,6 +171,8 @@ public:
          * When color grading is implemented using a 3D LUT, this sets the texture format of
          * of the LUT. This overrides the value set by quality().
          *
+         * This setting has no effect if generating a 1D LUT.
+         *
          * The default is INTEGER
          *
          * @param format The desired format of the 3D LUT.
@@ -181,6 +185,8 @@ public:
          * When color grading is implemented using a 3D LUT, this sets the dimension of the LUT.
          * This overrides the value set by quality().
          *
+         * This setting has no effect if generating a 1D LUT.
+         *
          * The default is 32
          *
          * @param dim The desired dimension of the LUT. Between 16 and 64.