random rays

add command line options
make a formal api
2024-11-22 21:25:22 -08:00 · 2024-11-22 21:25:22 -08:00 · 2024-11-22 21:25:22 -08:00 · 2024-11-22 21:25:22 -08:00 · 2024-11-22 21:25:22 -08:00 · 2024-11-22 21:25:22 -08:00
128 changed files with 1633 additions and 1722 deletions
@@ -525,14 +525,11 @@ else()
    option(FILAMENT_BUILD_FILAMAT "Build filamat and JNI buildings" OFF)
 endif()

-# By default, link in matdbg/fgviewer for Desktop + Debug only since it pulls in filamat and a web server.
+# By default, link in matdbg for Desktop + Debug only since it pulls in filamat and a web server.
 if (CMAKE_BUILD_TYPE STREQUAL "Debug" AND IS_HOST_PLATFORM)
    option(FILAMENT_ENABLE_MATDBG "Enable the material debugger" ON)
-    # TODO: Uncomment below when fgviewer is ready
-    # option(FILAMENT_ENABLE_FGVIEWER "Enable the frame graph viewer" ON)
 else()
    option(FILAMENT_ENABLE_MATDBG "Enable the material debugger" OFF)
-    option(FILAMENT_ENABLE_FGVIEWER "Enable the frame graph viewer" OFF)
 endif()

 # Only optimize materials in Release mode (so error message lines match the source code)
@@ -781,11 +778,6 @@ if (FILAMENT_BUILD_FILAMAT OR IS_HOST_PLATFORM)
    if (FILAMENT_ENABLE_MATDBG OR IS_HOST_PLATFORM)
        add_subdirectory(${LIBRARIES}/matdbg)
    endif()
-
-    # TODO: Uncomment below when fgviewer is ready
-    # if (FILAMENT_ENABLE_FGVIEWER OR IS_HOST_PLATFORM)
-    #     add_subdirectory(${LIBRARIES}/fgviewer)
-    # endif()
 endif()

 if (FILAMENT_SUPPORTS_VULKAN)
@@ -830,6 +822,7 @@ if (IS_HOST_PLATFORM)
    endif()
    add_subdirectory(${TOOLS}/mipgen)
    add_subdirectory(${TOOLS}/normal-blending)
+    add_subdirectory(${TOOLS}/rainbowgen)
    add_subdirectory(${TOOLS}/resgen)
    add_subdirectory(${TOOLS}/rgb-to-lmsr)
    add_subdirectory(${TOOLS}/roughness-prefilter)
@@ -31,7 +31,7 @@ repositories {
 }

 dependencies {
-    implementation 'com.google.android.filament:filament-android:1.56.6'
+    implementation 'com.google.android.filament:filament-android:1.56.1'
 }
 ```

@@ -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', '~> 1.56.6'
+pod 'Filament', '~> 1.56.1'
 ```

 ## Documentation
@@ -7,19 +7,6 @@ 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).

-## v1.56.6
-
-fix crash: the 'target_node' of Animation Channel may be nullpointer.
-
-## v1.56.5
-
-
-## v1.56.4
-
-
-## v1.56.3
-
-
 ## v1.56.2

 - vk: fix stage pool gc logic
@@ -11,9 +11,6 @@
 // com.google.android.filament.exclude-vulkan
 //     When set, support for Vulkan will be excluded.
 //
-// com.google.android.filament.fgviewer
-//     When set, enables fgviewer
-//
 // com.google.android.filament.matdbg
 //     When set, enables matdbg
 //
@@ -64,11 +61,6 @@ buildscript {
            .gradleProperty("com.google.android.filament.exclude-vulkan")
            .isPresent()

-    // TODO: Uncomment below when fgviewer is ready
-    // def fgviewer = providers
-    //         .gradleProperty("com.google.android.filament.fgviewer")
-    //         .isPresent()
-
    def matdbg = providers
            .gradleProperty("com.google.android.filament.matdbg")
            .isPresent()
@@ -123,8 +115,6 @@ buildscript {
        "-DANDROID_STL=c++_static",
        "-DFILAMENT_DIST_DIR=${filamentPath}".toString(),
        "-DFILAMENT_SUPPORTS_VULKAN=${excludeVulkan ? 'OFF' : 'ON'}".toString(),
-        // TODO: Uncomment below when fgviewer is ready
-        // "-DFILAMENT_ENABLE_FGVIEWER=${fgviewer ? 'ON' : 'OFF'}".toString(),
        "-DFILAMENT_ENABLE_MATDBG=${matdbg ? 'ON' : 'OFF'}".toString(),
        "-DFILAMENT_DISABLE_MATOPT=${matnopt ? 'ON' : 'OFF'}".toString()
    ]
@@ -2,8 +2,6 @@ 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_MATDBG "Enables Material debugger" OFF)
 option(FILAMENT_DISABLE_MATOPT "Disables material optimizations" OFF)

@@ -53,13 +51,6 @@ 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_MATDBG)
    add_library(matdbg STATIC IMPORTED)
    set_target_properties(matdbg PROPERTIES IMPORTED_LOCATION
@@ -126,8 +117,6 @@ 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_MATDBG},ON>:matdbg>
    $<$<STREQUAL:${FILAMENT_ENABLE_MATDBG},ON>:filamat>
    $<$<STREQUAL:${FILAMENT_SUPPORTS_VULKAN},ON>:bluevk>
@@ -38,7 +38,7 @@ Java_com_google_android_filament_EntityManager_nCreateArray(JNIEnv* env, jclass,
    // (which it is), but still.
    em->create((size_t) n, reinterpret_cast<Entity *>(entities));

-    env->ReleaseIntArrayElements(entities_, entities, 0);
+    env->ReleaseIntArrayElements(entities_, entities, JNI_ABORT);
 }

 extern "C" JNIEXPORT jint JNICALL
@@ -167,7 +167,7 @@ Java_com_google_android_filament_IndirectLight_nGetDirectionEstimateStatic(JNIEn
    jfloat *outDirection = env->GetFloatArrayElements(outDirection_, NULL);
    *reinterpret_cast<filament::math::float3*>(outDirection) = IndirectLight::getDirectionEstimate((filament::math::float3*)sh);
    env->ReleaseFloatArrayElements(outDirection_, outDirection, 0);
-    env->ReleaseFloatArrayElements(sh_, sh, 0);
+    env->ReleaseFloatArrayElements(sh_, sh, JNI_ABORT);
 }

 extern "C" JNIEXPORT void JNICALL
@@ -178,5 +178,5 @@ Java_com_google_android_filament_IndirectLight_nGetColorEstimateStatic(JNIEnv *e
    *reinterpret_cast<filament::math::float4*>(outColor) =
            IndirectLight::getColorEstimate((filament::math::float3*)sh, math::float3{x, y, z});
    env->ReleaseFloatArrayElements(outColor_, outColor, 0);
-    env->ReleaseFloatArrayElements(sh_, sh, 0);
+    env->ReleaseFloatArrayElements(sh_, sh, JNI_ABORT);
 }
@@ -25,14 +25,13 @@ using namespace filament;

 extern "C" JNIEXPORT jlong JNICALL
 Java_com_google_android_filament_Material_nBuilderBuild(JNIEnv *env, jclass,
-        jlong nativeEngine, jobject buffer_, jint size, jint shBandCount, jint shadowQuality) {
+        jlong nativeEngine, jobject buffer_, jint size, jint shBandCount) {
    Engine* engine = (Engine*) nativeEngine;
    AutoBuffer buffer(env, buffer_, size);
    auto builder = Material::Builder();
    if (shBandCount) {
        builder.sphericalHarmonicsBandCount(shBandCount);
    }
-    builder.shadowSamplingQuality((Material::Builder::ShadowSamplingQuality)shadowQuality);
    Material* material = builder
            .package(buffer.getData(), buffer.getSize())
            .build(*engine);
@@ -138,13 +138,6 @@ Java_com_google_android_filament_Texture_nBuilderImportTexture(JNIEnv*, jclass,
    builder->import((intptr_t)id);
 }

-extern "C"
-JNIEXPORT void JNICALL
-Java_com_google_android_filament_Texture_nBuilderExternal(JNIEnv*, jclass, jlong nativeBuilder) {
-    Texture::Builder *builder = (Texture::Builder *) nativeBuilder;
-    builder->external();
-}
-
 extern "C" JNIEXPORT jlong JNICALL
 Java_com_google_android_filament_Texture_nBuilderBuild(JNIEnv*, jclass,
        jlong nativeBuilder, jlong nativeEngine) {
@@ -344,18 +344,9 @@ public class Material {
    }

    public static class Builder {
-        public enum ShadowSamplingQuality {
-            /** 2x2 PCF */
-            HARD,
-            /** 3x3 gaussian filter */
-            LOW,
-        }
-
        private Buffer mBuffer;
        private int mSize;
        private int mShBandCount = 0;
-        private ShadowSamplingQuality mShadowSamplingQuality = ShadowSamplingQuality.LOW;
-

        /**
         * Specifies the material data. The material data is a binary blob produced by
@@ -387,18 +378,6 @@ public class Material {
            return this;
        }

-        /**
-         * Set the quality of shadow sampling. This is only taken into account
-         * if this material is lit and in the surface domain.
-         * @param quality
-         * @return Reference to this Builder for chaining calls.
-         */
-        @NonNull
-        public Builder shadowSamplingQuality(ShadowSamplingQuality quality) {
-            mShadowSamplingQuality = quality;
-            return this;
-        }
-
        /**
         * Creates and returns the Material object.
         *
@@ -411,7 +390,7 @@ public class Material {
        @NonNull
        public Material build(@NonNull Engine engine) {
            long nativeMaterial = nBuilderBuild(engine.getNativeObject(),
-                mBuffer, mSize, mShBandCount, mShadowSamplingQuality.ordinal());
+                mBuffer, mSize, mShBandCount);
            if (nativeMaterial == 0) throw new IllegalStateException("Couldn't create Material");
            return new Material(nativeMaterial);
        }
@@ -1062,7 +1041,7 @@ public class Material {
        mNativeObject = 0;
    }

-    private static native long nBuilderBuild(long nativeEngine, @NonNull Buffer buffer, int size, int shBandCount, int shadowQuality);
+    private static native long nBuilderBuild(long nativeEngine, @NonNull Buffer buffer, int size, int shBandCount);
    private static native long nCreateInstance(long nativeMaterial);
    private static native long nCreateInstanceWithName(long nativeMaterial, @NonNull String name);
    private static native long nGetDefaultInstance(long nativeMaterial);
@@ -800,19 +800,6 @@ public class Texture {
            return this;
        }

-        /**
-         * Creates an external texture. The content must be set using setExternalImage().
-         * The sampler can be SAMPLER_EXTERNAL or SAMPLER_2D depending on the format. Generally
-         * YUV formats must use SAMPLER_EXTERNAL. This depends on the backend features and is not
-         * validated.
-         * @return This Builder, for chaining calls.
-         */
-        @NonNull
-        public Builder external() {
-            nBuilderExternal(mNativeBuilder);
-            return this;
-        }
-
        /**
         * Creates a new <code>Texture</code> instance.
         * @param engine The {@link Engine} to associate this <code>Texture</code> with.
@@ -1274,7 +1261,6 @@ public class Texture {
    private static native void nBuilderUsage(long nativeBuilder, int flags);
    private static native void nBuilderSwizzle(long nativeBuilder, int r, int g, int b, int a);
    private static native void nBuilderImportTexture(long nativeBuilder, long id);
-    private static native void nBuilderExternal(long nativeBuilder);
    private static native long nBuilderBuild(long nativeBuilder, long nativeEngine);

    private static native int nGetWidth(long nativeTexture, int level);
@@ -1,5 +1,5 @@
 GROUP=com.google.android.filament
-VERSION_NAME=1.56.6
+VERSION_NAME=1.56.1

 POM_DESCRIPTION=Real-time physically based rendering engine for Android.

@@ -23,8 +23,6 @@ function print_help {
    echo "        This is sometimes needed instead of -c (which still misses some clean steps)."
    echo "    -d"
    echo "        Enable matdbg."
-    echo "    -t"
-    echo "        Enable fgviewer."
    echo "    -f"
    echo "        Always invoke CMake before incremental builds."
    echo "    -g"
@@ -127,27 +125,6 @@ function print_matdbg_help {
    echo ""
 }

-function print_fgviewer_help {
-    echo "fgviewer is enabled in the build, but some extra steps are needed."
-    echo ""
-    echo "FOR DESKTOP BUILDS:"
-    echo ""
-    echo "Please set the port environment variable before launching. e.g., on macOS do:"
-    echo "   export FILAMENT_FGVIEWER_PORT=8085"
-    echo ""
-    echo "FOR ANDROID BUILDS:"
-    echo ""
-    echo "1) For Android Studio builds, make sure to set:"
-    echo "       -Pcom.google.android.filament.fgviewer"
-    echo "   option in Preferences > Build > Compiler > Command line options."
-    echo ""
-    echo "2) The port number is hardcoded to 8085 so you will need to do:"
-    echo "       adb forward tcp:8085 tcp:8085"
-    echo ""
-    echo "3) Be sure to enable INTERNET permission in your app's manifest file."
-    echo ""
-}
-
 # Unless explicitly specified, NDK version will be selected as highest available version within same major release chain
 FILAMENT_NDK_VERSION=${FILAMENT_NDK_VERSION:-$(cat `dirname $0`/build/android/ndk.version | cut -f 1 -d ".")}

@@ -195,7 +172,6 @@ VULKAN_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=""

 MATOPT_OPTION=""
@@ -264,7 +240,6 @@ function build_desktop_target {
            -DCMAKE_BUILD_TYPE="$1" \
            -DCMAKE_INSTALL_PREFIX="../${lc_target}/filament" \
            ${EGL_ON_LINUX_OPTION} \
-            ${FGVIEWER_OPTION} \
            ${MATDBG_OPTION} \
            ${MATOPT_OPTION} \
            ${ASAN_UBSAN_OPTION} \
@@ -401,7 +376,6 @@ function build_android_target {
            -DFILAMENT_NDK_VERSION="${FILAMENT_NDK_VERSION}" \
            -DCMAKE_INSTALL_PREFIX="../android-${lc_target}/filament" \
            -DCMAKE_TOOLCHAIN_FILE="../../build/toolchain-${arch}-linux-android.cmake" \
-            ${FGVIEWER_OPTION} \
            ${MATDBG_OPTION} \
            ${MATOPT_OPTION} \
            ${VULKAN_ANDROID_OPTION} \
@@ -639,7 +613,6 @@ function build_ios_target {
            -DPLATFORM_NAME="${platform}" \
            -DIOS=1 \
            -DCMAKE_TOOLCHAIN_FILE=../../third_party/clang/iOS.cmake \
-            ${FGVIEWER_OPTION} \
            ${MATDBG_OPTION} \
            ${MATOPT_OPTION} \
            ${STEREOSCOPIC_OPTION} \
@@ -829,7 +802,7 @@ function check_debug_release_build {

 pushd "$(dirname "$0")" > /dev/null

-while getopts ":hacCfgijmp:q:uvslwedtk:bx:S:X:" opt; do
+while getopts ":hacCfgijmp:q:uvslwedk:bx:S:X:" opt; do
    case ${opt} in
        h)
            print_help
@@ -850,12 +823,6 @@ while getopts ":hacCfgijmp:q:uvslwedtk:bx:S:X:" opt; do
            MATDBG_OPTION="-DFILAMENT_ENABLE_MATDBG=ON, -DFILAMENT_BUILD_FILAMAT=ON"
            MATDBG_GRADLE_OPTION="-Pcom.google.android.filament.matdbg"
            ;;
-        t)
-            # TODO: Uncomment below when fgviewer is ready
-            # PRINT_FGVIEWER_HELP=true
-            # FGVIEWER_OPTION="-DFILAMENT_ENABLE_FGVIEWER=ON"
-            #FGVIEWER_GRADLE_OPTION="-Pcom.google.android.filament.fgviewer"
-            ;;
        f)
            ISSUE_CMAKE_ALWAYS=true
            ;;
@@ -1060,7 +1027,3 @@ fi
 if [[ "${PRINT_MATDBG_HELP}" == "true" ]]; then
    print_matdbg_help
 fi
-
-if [[ "${PRINT_FGVIEWER_HELP}" == "true" ]]; then
-    print_fgviewer_help
-fi
@@ -1787,12 +1787,7 @@ non-shader data.
      declare a variable called <code>eyeDirection</code> you can access it in the fragment shader using
      <code>variable_eyeDirection</code>. In the vertex shader, the interpolant name is simply a member of
      the <code>MaterialVertexInputs</code> structure (<code>material.eyeDirection</code> in your example). Each
-      interpolant is of type <code>float4</code> (<code>vec4</code>) in the shaders. By default the precision of the
-      interpolant is <code>highp</code> in <em class="asterisk">both</em> the vertex and fragment shaders.
-      An alternate syntax can be used to specify both the name and precision of the interpolant.
-      In this case the specified precision is used as-is in both fragment and vertex stages, in
-      particular if <code>default</code> is specified the default precision is used is the fragment shader
-      (<code>mediump</code>) and in the vertex shader (<code>highp</code>).
+      interpolant is of type <code>float4</code> (<code>vec4</code>) in the shaders.

 </p></dd></dl><p></p><pre class="listing tilde"><code><span class="line">material {</span>
 <span class="line">    name : Skybox,</span>
@@ -1803,11 +1798,7 @@ non-shader data.
 <span class="line">        }</span>
 <span class="line">    ],</span>
 <span class="line">    variables : [</span>
-<span class="line">         eyeDirection,</span>
-<span class="line">         {</span>
-<span class="line">            name : eyeColor,</span>
-<span class="line">            precision : medium</span>
-<span class="line">         }</span>
+<span class="line">         eyeDirection</span>
 <span class="line">    ],</span>
 <span class="line">    vertexDomain : device,</span>
 <span class="line">    depthWrite : <span class="hljs-literal">false</span>,</span>
@@ -561,13 +561,6 @@ target_link_libraries(${TARGET} PUBLIC filaflat)
 target_link_libraries(${TARGET} PUBLIC filabridge)
 target_link_libraries(${TARGET} PUBLIC ibl-lite)

-if (FILAMENT_ENABLE_FGVIEWER)
-    target_link_libraries(${TARGET} PUBLIC fgviewer)
-    add_definitions(-DFILAMENT_ENABLE_FGVIEWER=1)
-else()
-    add_definitions(-DFILAMENT_ENABLE_FGVIEWER=0)
-endif()
-
 if (FILAMENT_ENABLE_MATDBG)
    target_link_libraries(${TARGET} PUBLIC matdbg)
    add_definitions(-DFILAMENT_ENABLE_MATDBG=1)
@@ -226,12 +226,10 @@ if (FILAMENT_SUPPORTS_VULKAN)
            src/vulkan/VulkanUtility.cpp
            src/vulkan/VulkanUtility.h
    )
-    if (LINUX OR WIN32)
-        list(APPEND SRCS src/vulkan/platform/VulkanPlatformLinuxWindows.cpp)
+    if (ANDROID OR LINUX OR WIN32)
+        list(APPEND SRCS src/vulkan/platform/VulkanPlatformAndroidLinuxWindows.cpp)
    elseif (APPLE OR IOS)
        list(APPEND SRCS src/vulkan/platform/VulkanPlatformApple.mm)
-    elseif (ANDROID)
-        list(APPEND SRCS src/vulkan/platform/VulkanPlatformAndroid.cpp)
    endif()
 endif()

@@ -292,64 +292,8 @@ public:
     */
    VkQueue getProtectedGraphicsQueue() const noexcept;

-    struct ExternalImageMetadata {
-        /**
-         * The width of the external image
-         */
-        uint32_t width;
-
-        /**
-         * The height of the external image
-         */
-        uint32_t height;
-
-        /**
-         * The layer count of the external image
-         */
-        uint32_t layers;
-
-        /**
-         * The format of the external image
-         */
-        VkFormat format;
-
-        /**
-         * An external buffer can be protected. This tells you if it is.
-         */
-        bool isProtected;
-
-        /**
-         * The type of external format (opaque int) if used.
-         */
-        uint64_t externalFormat;
-
-        /**
-         * Image usage
-         */
-        VkImageUsageFlags usage;
-
-        /**
-         * Allocation size
-         */
-        VkDeviceSize allocationSize;
-
-        /**
-         * Heap information
-         */
-        uint32_t memoryTypeBits;
-    };
-    virtual ExternalImageMetadata getExternalImageMetadata(void* externalImage);
-
-    using ImageData = std::pair<VkImage, VkDeviceMemory>;
-    virtual ImageData createExternalImage(void* externalImage,
-            const ExternalImageMetadata& metadata);
-
 private:
    static ExtensionSet getSwapchainInstanceExtensions();
-    static ExternalImageMetadata getExternalImageMetadataImpl(void* externalImage,
-            VkDevice device);
-    static ImageData createExternalImageImpl(void* externalImage, VkDevice device,
-            const VkAllocationCallbacks* allocator, const ExternalImageMetadata& metadata);

    // Platform dependent helper methods
    using SurfaceBundle = std::tuple<VkSurfaceKHR, VkExtent2D>;
@@ -214,7 +214,6 @@ DECL_DRIVER_API_R_N(backend::TextureHandle, createTextureViewSwizzle,
        backend::TextureSwizzle, a)

 DECL_DRIVER_API_R_N(backend::TextureHandle, createTextureExternalImage,
-        backend::SamplerType, target,
        backend::TextureFormat, format,
        uint32_t, width,
        uint32_t, height,
@@ -400,6 +399,17 @@ DECL_DRIVER_API_N(update3DImage,
 DECL_DRIVER_API_N(generateMipmaps,
        backend::TextureHandle, th)

+// Deprecated
+DECL_DRIVER_API_N(setExternalImage,
+        backend::TextureHandle, th,
+        void*, image)
+
+// Deprecated
+DECL_DRIVER_API_N(setExternalImagePlane,
+        backend::TextureHandle, th,
+        void*, image,
+        uint32_t, plane)
+
 DECL_DRIVER_API_N(setExternalStream,
        backend::TextureHandle, th,
        backend::StreamHandle, sh)
@@ -24,7 +24,7 @@
    #if defined(FILAMENT_SUPPORTS_OPENGL) && !defined(FILAMENT_USE_EXTERNAL_GLES3)
        #include "backend/platforms/PlatformEGLAndroid.h"
    #endif
-#elif defined(FILAMENT_IOS)
+#elif defined(IOS)
    #if defined(FILAMENT_SUPPORTS_OPENGL) && !defined(FILAMENT_USE_EXTERNAL_GLES3)
        #include "backend/platforms/PlatformCocoaTouchGL.h"
    #endif
@@ -88,7 +88,7 @@ Platform* PlatformFactory::create(Backend* backend) noexcept {
        *backend = Backend::OPENGL;
 #elif defined(__ANDROID__)
        *backend = Backend::OPENGL;
-#elif defined(FILAMENT_IOS) || defined(__APPLE__)
+#elif defined(IOS) || defined(__APPLE__)
        *backend = Backend::METAL;
 #elif defined(FILAMENT_DRIVER_SUPPORTS_VULKAN)
        *backend = Backend::VULKAN;
@@ -119,7 +119,7 @@ Platform* PlatformFactory::create(Backend* backend) noexcept {
            return nullptr;
        #elif defined(__ANDROID__)
            return new PlatformEGLAndroid();
-        #elif defined(FILAMENT_IOS)
+        #elif defined(IOS)
            return new PlatformCocoaTouchGL();
        #elif defined(__APPLE__)
            return new PlatformCocoaGL();
@@ -250,7 +250,7 @@ public:
    // In practice, MetalRingBuffer is used for argument buffers, which are kept in the constant
    // address space. Constant buffers have specific alignment requirements when specifying an
    // offset.
-#if defined(FILAMENT_IOS)
+#if defined(IOS)
 #if TARGET_OS_SIMULATOR
    // The iOS simulator has differing alignment requirements.
    static constexpr auto METAL_CONSTANT_BUFFER_OFFSET_ALIGNMENT = 256;
@@ -135,7 +135,6 @@ struct MetalContext {
    bool supportsTextureSwizzling = false;
    bool supportsAutoDepthResolve = false;
    bool supportsMemorylessRenderTargets = false;
-    bool supportsDepthClamp = false;
    uint8_t maxColorRenderTargets = 4;
    struct {
        uint8_t common;
@@ -155,7 +154,6 @@ struct MetalContext {

    RenderPassFlags currentRenderPassFlags;
    MetalRenderTarget* currentRenderTarget = nullptr;
-    bool validPipelineBound = false;

    // State trackers.
    PipelineStateTracker pipelineState;
@@ -151,11 +151,6 @@ MetalDriver::MetalDriver(
            mContext->highestSupportedGpuFamily.mac   >= 2;     // newer macOS GPUs
    }

-    mContext->supportsDepthClamp = false;
-    if (@available(macOS 10.11, iOS 11.0, *)) {
-        mContext->supportsDepthClamp = true;
-    }
-
    // In order to support resolve store action on depth attachment, the GPU needs to support it.
    // Note that support for depth resolve implies support for stencil resolve using .sample0 resolve filter.
    // (Other resolve filters are supported starting .apple5 and .mac2 families).
@@ -482,9 +477,7 @@ void MetalDriver::createTextureViewSwizzleR(Handle<HwTexture> th, Handle<HwTextu
    mContext->textures.insert(construct_handle<MetalTexture>(th, *mContext, src, r, g, b, a));
 }

-void MetalDriver::createTextureExternalImageR(Handle<HwTexture> th,
-        backend::SamplerType target,
-        backend::TextureFormat format,
+void MetalDriver::createTextureExternalImageR(Handle<HwTexture> th, backend::TextureFormat format,
        uint32_t width, uint32_t height, backend::TextureUsage usage, void* image) {
    mContext->textures.insert(construct_handle<MetalTexture>(
            th, *mContext, format, width, height, usage, (CVPixelBufferRef)image));
@@ -909,7 +902,7 @@ void MetalDriver::terminate() {
 }

 ShaderModel MetalDriver::getShaderModel() const noexcept {
-#if defined(FILAMENT_IOS)
+#if defined(IOS)
    return ShaderModel::MOBILE;
 #else
    return ShaderModel::DESKTOP;
@@ -983,7 +976,7 @@ bool MetalDriver::isTextureFormatMipmappable(TextureFormat format) {
        case MTLPixelFormatRGBA16Float:
            return true;

-#if !defined(FILAMENT_IOS)
+#if !defined(IOS)
        // Mipmappable only on desktop:
        case MTLPixelFormatR32Float:
        case MTLPixelFormatRG32Float:
@@ -991,7 +984,7 @@ bool MetalDriver::isTextureFormatMipmappable(TextureFormat format) {
            return true;
 #endif

-#if defined(FILAMENT_IOS)
+#if defined(IOS)
        // Mipmappable only on iOS:
        case MTLPixelFormatRGB9E5Float:
            return true;
@@ -1070,7 +1063,7 @@ bool MetalDriver::isProtectedTexturesSupported() {
 }

 bool MetalDriver::isDepthClampSupported() {
-    return mContext->supportsDepthClamp;
+    return true;
 }

 bool MetalDriver::isWorkaroundNeeded(Workaround workaround) {
@@ -1193,6 +1186,10 @@ void MetalDriver::setupExternalImage(void* image) {
    CVPixelBufferRetain(pixelBuffer);
 }

+void MetalDriver::setExternalImage(Handle<HwTexture> th, void* image) {}
+
+void MetalDriver::setExternalImagePlane(Handle<HwTexture> th, void* image, uint32_t plane) {}
+
 void MetalDriver::setExternalStream(Handle<HwTexture> th, Handle<HwStream> sh) {
 }

@@ -1261,7 +1258,6 @@ void MetalDriver::beginRenderPass(Handle<HwRenderTarget> rth,
    mContext->cullModeState.invalidate();
    mContext->windingState.invalidate();
    mContext->scissorRectState.invalidate();
-    mContext->depthClampState.invalidate();
    mContext->currentPolygonOffset = {0.0f, 0.0f};

    mContext->finalizedDescriptorSets.clear();
@@ -1364,7 +1360,7 @@ void MetalDriver::startCapture(int) {
    if (@available(iOS 13, *)) {
        MTLCaptureDescriptor* descriptor = [MTLCaptureDescriptor new];
        descriptor.captureObject = mContext->device;
-#if defined(FILAMENT_IOS)
+#if defined(IOS)
        descriptor.destination = MTLCaptureDestinationDeveloperTools;
 #else
        descriptor.destination = MTLCaptureDestinationGPUTraceDocument;
@@ -1428,7 +1424,7 @@ void MetalDriver::readPixels(Handle<HwRenderTarget> src, uint32_t x, uint32_t y,
                                                               width:srcTextureSize.width
                                                              height:srcTextureSize.height
                                                           mipmapped:NO];
-#if defined(FILAMENT_IOS)
+#if defined(IOS)
    textureDescriptor.storageMode = MTLStorageModeShared;
 #else
    textureDescriptor.storageMode = MTLStorageModeManaged;
@@ -1447,7 +1443,7 @@ void MetalDriver::readPixels(Handle<HwRenderTarget> src, uint32_t x, uint32_t y,

    mContext->blitter->blit(getPendingCommandBuffer(mContext), args, "readPixels blit");

-#if !defined(FILAMENT_IOS)
+#if !defined(IOS)
    // Managed textures on macOS require explicit synchronization between GPU / CPU.
    id <MTLBlitCommandEncoder> blitEncoder = [getPendingCommandBuffer(mContext) blitCommandEncoder];
    [blitEncoder synchronizeResource:readPixelsTexture];
@@ -1655,7 +1651,6 @@ void MetalDriver::bindPipeline(PipelineState const& ps) {
    // during the draw call when the program is invalid. The shader compile error has already been
    // dumped to the console at this point, so it's fine to simply return early.
    if (FILAMENT_ENABLE_MATDBG && UTILS_UNLIKELY(!functions)) {
-        mContext->validPipelineBound = false;
        return;
    }

@@ -1734,12 +1729,10 @@ void MetalDriver::bindPipeline(PipelineState const& ps) {
    }

    // depth clip mode
-    if (mContext->supportsDepthClamp) {
-        MTLDepthClipMode depthClipMode = rs.depthClamp ? MTLDepthClipModeClamp : MTLDepthClipModeClip;
-        mContext->depthClampState.updateState(depthClipMode);
-        if (mContext->depthClampState.stateChanged()) {
-            [mContext->currentRenderPassEncoder setDepthClipMode:depthClipMode];
-        }
+    MTLDepthClipMode depthClipMode = rs.depthClamp ? MTLDepthClipModeClamp : MTLDepthClipModeClip;
+    mContext->depthClampState.updateState(depthClipMode);
+    if (mContext->depthClampState.stateChanged()) {
+        [mContext->currentRenderPassEncoder setDepthClipMode:depthClipMode];
    }

    // Set the depth-stencil state, if a state change is needed.
@@ -1788,8 +1781,6 @@ void MetalDriver::bindPipeline(PipelineState const& ps) {
                                                   clamp:0.0];
        mContext->currentPolygonOffset = ps.polygonOffset;
    }
-
-    mContext->validPipelineBound = true;
 }

 void MetalDriver::bindRenderPrimitive(Handle<HwRenderPrimitive> rph) {
@@ -1887,10 +1878,6 @@ void MetalDriver::draw2(uint32_t indexOffset, uint32_t indexCount, uint32_t inst
            << "draw() without a valid command encoder.";
    DEBUG_LOG("draw2(...)\n");

-    if (FILAMENT_ENABLE_MATDBG && UTILS_UNLIKELY(!mContext->validPipelineBound)) {
-        return;
-    }
-
    // Bind the offset data.
    if (mContext->dynamicOffsets.isDirty()) {
        const auto [size, data] = mContext->dynamicOffsets.getOffsets();
@@ -196,7 +196,7 @@ inline MTLPixelFormat getMetalFormatLinear(MTLPixelFormat format) {
        case MTLPixelFormatRG8Unorm_sRGB: return MTLPixelFormatRG8Unorm;
        case MTLPixelFormatRGBA8Unorm_sRGB: return MTLPixelFormatRGBA8Unorm;
        case MTLPixelFormatBGRA8Unorm_sRGB: return MTLPixelFormatBGRA8Unorm;
-#if !defined(FILAMENT_IOS)
+#if !defined(IOS)
        case MTLPixelFormatBC1_RGBA_sRGB: return MTLPixelFormatBC1_RGBA;
        case MTLPixelFormatBC2_RGBA_sRGB: return MTLPixelFormatBC2_RGBA;
        case MTLPixelFormatBC3_RGBA_sRGB: return MTLPixelFormatBC3_RGBA;
@@ -268,7 +268,7 @@ constexpr inline bool isMetalFormatStencil(MTLPixelFormat format) {
        case MTLPixelFormatStencil8:
        case MTLPixelFormatDepth32Float_Stencil8:
        case MTLPixelFormatX32_Stencil8:
-#if !defined(FILAMENT_IOS)
+#if !defined(IOS)
        case MTLPixelFormatDepth24Unorm_Stencil8:
        case MTLPixelFormatX24_Stencil8:
 #endif
@@ -157,7 +157,7 @@ id<MTLTexture> MetalSwapChain::acquireDrawable() {
        textureDescriptor.height = headlessHeight;
        // Specify MTLTextureUsageShaderRead so the headless surface can be blitted from.
        textureDescriptor.usage = MTLTextureUsageRenderTarget | MTLTextureUsageShaderRead;
-#if defined(FILAMENT_IOS)
+#if defined(IOS)
        textureDescriptor.storageMode = MTLStorageModeShared;
 #else
        textureDescriptor.storageMode = MTLStorageModeManaged;
@@ -520,7 +520,7 @@ MetalTexture::MetalTexture(MetalContext& context, SamplerType target, uint8_t le
    const BOOL mipmapped = levels > 1;
    const BOOL multisampled = samples > 1;

-#if defined(FILAMENT_IOS)
+#if defined(IOS)
    const BOOL textureArray = target == SamplerType::SAMPLER_2D_ARRAY;
    FILAMENT_CHECK_PRECONDITION(!textureArray || !multisampled)
            << "iOS does not support multisampled texture arrays.";
@@ -533,7 +533,7 @@ MetalTexture::MetalTexture(MetalContext& context, SamplerType target, uint8_t le
        switch (target) {
            case SamplerType::SAMPLER_2D:
                return MTLTextureType2DMultisample;
-#if !defined(FILAMENT_IOS)
+#if !defined(IOS)
            case SamplerType::SAMPLER_2D_ARRAY:
                return MTLTextureType2DMultisampleArray;
 #endif
@@ -856,7 +856,7 @@ void MetalTexture::loadWithBlit(uint32_t level, uint32_t slice, MTLRegion region
    descriptor.height = region.size.height;
    descriptor.depth = region.size.depth;

-#if defined(FILAMENT_IOS)
+#if defined(IOS)
    descriptor.storageMode = MTLStorageModeShared;
 #else
    descriptor.storageMode = MTLStorageModeManaged;
@@ -37,7 +37,7 @@ Driver* MetalPlatform::createDriver(void* /*sharedContext*/, const Platform::Dri
 id<MTLDevice> MetalPlatform::createDevice() noexcept {
    id<MTLDevice> result;

-#if !defined(FILAMENT_IOS)
+#if !defined(IOS)
    const bool forceIntegrated =
            NSProcessInfo.processInfo.environment[@"FILAMENT_FORCE_INTEGRATED_GPU"] != nil;
    if (forceIntegrated) {
@@ -226,11 +226,9 @@ MetalShaderCompiler::program_token_t MetalShaderCompiler::createProgram(
            CompilerPriorityQueue const priorityQueue = program.getPriorityQueue();
            mCompilerThreadPool.queue(priorityQueue, token,
                    [this, name, device = mDevice, program = std::move(program), token]() {
-                        @autoreleasepool {
-                            MetalFunctionBundle compiledProgram = compileProgram(program, device);
-                            token->set(compiledProgram);
-                            mCallbackManager.put(token->handle);
-                        }
+                        MetalFunctionBundle compiledProgram = compileProgram(program, device);
+                        token->set(compiledProgram);
+                        mCallbackManager.put(token->handle);
                    });

            break;
@@ -150,7 +150,7 @@ id<MTLSamplerState> SamplerStateCreator::operator()(id<MTLDevice> device,
                MTLCompareFunctionNever : getCompareFunction(params.compareFunc);
    samplerDescriptor.supportArgumentBuffers = YES;

-#if defined(FILAMENT_IOS)
+#if defined(IOS)
    // Older Apple devices (and the simulator) don't support setting a comparison function in
    // MTLSamplerDescriptor.
    // In practice, this means shadows are not supported when running in the simulator.
@@ -32,7 +32,7 @@ Dispatcher NoopDriver::getDispatcher() const noexcept {
 }

 ShaderModel NoopDriver::getShaderModel() const noexcept {
-#if defined(__ANDROID__) || defined(FILAMENT_IOS) || defined(__EMSCRIPTEN__)
+#if defined(__ANDROID__) || defined(IOS) || defined(__EMSCRIPTEN__)
    return ShaderModel::MOBILE;
 #else
    return ShaderModel::DESKTOP;
@@ -265,6 +265,12 @@ TimerQueryResult NoopDriver::getTimerQueryValue(Handle<HwTimerQuery> tqh, uint64
    return TimerQueryResult::ERROR;
 }

+void NoopDriver::setExternalImage(Handle<HwTexture> th, void* image) {
+}
+
+void NoopDriver::setExternalImagePlane(Handle<HwTexture> th, void* image, uint32_t plane) {
+}
+
 void NoopDriver::setExternalStream(Handle<HwTexture> th, Handle<HwStream> sh) {
 }

@@ -173,7 +173,6 @@ void GLDescriptorSet::update(OpenGLContext& gl,

            arg.target = t ? t->gl.target : 0;
            arg.id = t ? t->gl.id : 0;
-            arg.external = t ? t->gl.external :  false;
            if constexpr (std::is_same_v<T, Sampler> ||
                          std::is_same_v<T, SamplerWithAnisotropyWorkaround>) {
                if constexpr (std::is_same_v<T, SamplerWithAnisotropyWorkaround>) {
@@ -286,7 +285,7 @@ void GLDescriptorSet::bind(
            } else if constexpr (std::is_same_v<T, Sampler>) {
                GLuint const unit = p.getTextureUnit(set, binding);
                if (arg.target) {
-                    gl.bindTexture(unit, arg.target, arg.id, arg.external);
+                    gl.bindTexture(unit, arg.target, arg.id);
                    gl.bindSampler(unit, arg.sampler);
                    if (UTILS_UNLIKELY(arg.ref)) {
                        updateTextureView(gl, handleAllocator, unit, arg);
@@ -297,7 +296,7 @@ void GLDescriptorSet::bind(
            } else if constexpr (std::is_same_v<T, SamplerWithAnisotropyWorkaround>) {
                GLuint const unit = p.getTextureUnit(set, binding);
                if (arg.target) {
-                    gl.bindTexture(unit, arg.target, arg.id, arg.external);
+                    gl.bindTexture(unit, arg.target, arg.id);
                    gl.bindSampler(unit, arg.sampler);
                    if (UTILS_UNLIKELY(arg.ref)) {
                        updateTextureView(gl, handleAllocator, unit, arg);
@@ -315,7 +314,7 @@ void GLDescriptorSet::bind(
                // in ES2 the sampler parameters need to be set on the texture itself
                GLuint const unit = p.getTextureUnit(set, binding);
                if (arg.target) {
-                    gl.bindTexture(unit, arg.target, arg.id, arg.external);
+                    gl.bindTexture(unit, arg.target, arg.id);
                    SamplerParams const params = arg.params;
                    glTexParameteri(arg.target, GL_TEXTURE_MIN_FILTER,
                            (GLint)GLUtils::getTextureFilter(params.filterMin));
@@ -111,9 +111,7 @@ private:

    // A sampler descriptor
    struct Sampler {
-        uint16_t target;                        // 2 (GLenum)
-        bool external = false;                  // 1
-        bool reserved = false;                  // 1
+        GLenum target = 0;                      // 4
        GLuint id = 0;                          // 4
        GLuint sampler = 0;                     // 4
        Handle<GLTextureRef> ref;               // 4
@@ -128,9 +126,7 @@ private:
    };

    struct SamplerWithAnisotropyWorkaround {
-        uint16_t target;                        // 2 (GLenum)
-        bool external = false;                  // 1
-        bool reserved = false;                  // 1
+        GLenum target = 0;                      // 4
        GLuint id = 0;                          // 4
        GLuint sampler = 0;                     // 4
        Handle<GLTextureRef> ref;               // 4
@@ -147,9 +143,7 @@ private:

    // A sampler descriptor for ES2
    struct SamplerGLES2 {
-        uint16_t target;                        // 2 (GLenum)
-        bool external = false;                  // 1
-        bool reserved = false;                  // 1
+        GLenum target = 0;                      // 4
        GLuint id = 0;                          // 4
        SamplerParams params{};                 // 4
        float anisotropy = 1.0f;                // 4
@@ -35,7 +35,7 @@ struct GLTextureRef {
    GLTextureRef() = default;
    // view reference counter
    uint16_t count = 1;
-    // Current per-view values of the texture (in GL we can only have a single View active at
+    // current per-view values of the texture (in GL we can only have a single View active at
    // a time, and this tracks that state). It's used to avoid unnecessarily change state.
    int8_t baseLevel = 127;
    int8_t maxLevel = -1;
@@ -50,7 +50,7 @@ struct GLTextureRef {
 struct GLTexture : public HwTexture {
    using HwTexture::HwTexture;
    struct GL {
-        GL() noexcept : imported(false), external(false), sidecarSamples(1), reserved1(0) {}
+        GL() noexcept : imported(false), sidecarSamples(1), reserved1(0) {}
        GLuint id = 0;          // texture or renderbuffer id
        GLenum target = 0;
        GLenum internalFormat = 0;
@@ -62,8 +62,7 @@ struct GLTexture : public HwTexture {
        int8_t maxLevel = -1;
        uint8_t reserved0 = 0;
        bool imported           : 1;
-        bool external           : 1;
-        uint8_t sidecarSamples  : 3;
+        uint8_t sidecarSamples  : 4;
        uint8_t reserved1       : 3;
        std::array<TextureSwizzle, 4> swizzle{
                TextureSwizzle::CHANNEL_0,
@@ -408,7 +408,7 @@ void OpenGLContext::initProcs(Procs* procs,
    procs->maxShaderCompilerThreadsKHR = +[](GLuint) {};

 #ifdef BACKEND_OPENGL_VERSION_GLES
-#   ifndef FILAMENT_IOS // FILAMENT_IOS is guaranteed to have ES3.x
+#   ifndef IOS // IOS is guaranteed to have ES3.x
 #       ifndef __EMSCRIPTEN__
    if (UTILS_UNLIKELY(major == 2)) {
        // Runtime OpenGL version is ES 2.x
@@ -438,7 +438,7 @@ void OpenGLContext::initProcs(Procs* procs,
        procs->maxShaderCompilerThreadsKHR = glMaxShaderCompilerThreadsKHR;
    }
 #       endif // __EMSCRIPTEN__
-#   endif // FILAMENT_IOS
+#   endif // IOS
 #else
    procs->maxShaderCompilerThreadsKHR = glMaxShaderCompilerThreadsARB;
 #endif
@@ -500,13 +500,11 @@ void OpenGLContext::initBugs(Bugs* bugs, Extensions const& exts,
            // qualcomm seems to have no problem with this (which is good for us)
            bugs->allow_read_only_ancillary_feedback_loop = true;

-#ifndef __EMSCRIPTEN__
            // Older Adreno devices that support ES3.0 only tend to be extremely buggy, so we
            // fall back to ES2.0.
            if (major == 3 && minor == 0) {
                bugs->force_feature_level0 = true;
            }
-#endif
        } else if (strstr(renderer, "Mali")) {
            // ARM GPU
            bugs->vao_doesnt_store_element_array_buffer_binding = true;
@@ -576,14 +574,14 @@ void OpenGLContext::initBugs(Bugs* bugs, Extensions const& exts,
    }

 #ifdef BACKEND_OPENGL_VERSION_GLES
-#   ifndef FILAMENT_IOS // FILAMENT_IOS is guaranteed to have ES3.x
+#   ifndef IOS // IOS is guaranteed to have ES3.x
    if (UTILS_UNLIKELY(major == 2)) {
        if (UTILS_UNLIKELY(!exts.OES_vertex_array_object)) {
            // we activate this workaround path, which does the reset of array buffer
            bugs->vao_doesnt_store_element_array_buffer_binding = true;
        }
    }
-#   endif // FILAMENT_IOS
+#   endif // IOS
 #else
    // feedback loops are allowed on GL desktop as long as writes are disabled
    bugs->allow_read_only_ancillary_feedback_loop = true;
@@ -622,7 +620,7 @@ FeatureLevel OpenGLContext::resolveFeatureLevel(GLint major, GLint minor,
            }
        }
    }
-#   ifndef FILAMENT_IOS // FILAMENT_IOS is guaranteed to have ES3.x
+#   ifndef IOS // IOS is guaranteed to have ES3.x
    else if (UTILS_UNLIKELY(major == 2)) {
        // Runtime OpenGL version is ES 2.x
        // note: mandatory extensions (all supported by Mali-400 and Adreno 304)
@@ -634,7 +632,7 @@ FeatureLevel OpenGLContext::resolveFeatureLevel(GLint major, GLint minor,
        //      OES_texture_npot
        featureLevel = FeatureLevel::FEATURE_LEVEL_0;
    }
-#   endif // FILAMENT_IOS
+#   endif // IOS
 #else
    assert_invariant(gets.max_texture_image_units >= 16);
    assert_invariant(gets.max_combined_texture_image_units >= 32);
@@ -131,7 +131,7 @@ public:
    }

    inline bool isES2() const noexcept {
-#if defined(BACKEND_OPENGL_VERSION_GLES) && !defined(FILAMENT_IOS)
+#if defined(BACKEND_OPENGL_VERSION_GLES) && !defined(IOS)
 #   ifndef BACKEND_OPENGL_LEVEL_GLES30
            return true;
 #   else
@@ -153,7 +153,7 @@ public:

          void pixelStore(GLenum, GLint) noexcept;
    inline void activeTexture(GLuint unit) noexcept;
-    inline void bindTexture(GLuint unit, GLuint target, GLuint texId, bool external) noexcept;
+    inline void bindTexture(GLuint unit, GLuint target, GLuint texId) noexcept;

           void unbindTexture(GLenum target, GLuint id) noexcept;
           void unbindTextureUnit(GLuint unit) noexcept;
@@ -761,9 +761,7 @@ void OpenGLContext::bindBufferRange(GLenum target, GLuint index, GLuint buffer,
 #endif
 }

-void OpenGLContext::bindTexture(GLuint unit, GLuint target, GLuint texId, bool external) noexcept {
-    //  another texture is bound to the same unit with a different target,
-    //  unbind the texture from the current target
+void OpenGLContext::bindTexture(GLuint unit, GLuint target, GLuint texId) noexcept {
    update_state(state.textures.units[unit].target, target, [&]() {
        activeTexture(unit);
        glBindTexture(state.textures.units[unit].target, 0);
@@ -771,7 +769,7 @@ void OpenGLContext::bindTexture(GLuint unit, GLuint target, GLuint texId, bool e
    update_state(state.textures.units[unit].id, texId, [&]() {
        activeTexture(unit);
        glBindTexture(target, texId);
-    }, external);
+    }, target == GL_TEXTURE_EXTERNAL_OES);
 }

 void OpenGLContext::useProgram(GLuint program) noexcept {
@@ -402,7 +402,7 @@ void OpenGLDriver::setPushConstant(backend::ShaderStage stage, uint8_t index,

 void OpenGLDriver::bindTexture(GLuint unit, GLTexture const* t) noexcept {
    assert_invariant(t != nullptr);
-    mContext.bindTexture(unit, t->gl.target, t->gl.id, t->gl.external);
+    mContext.bindTexture(unit, t->gl.target, t->gl.id);
 }

 bool OpenGLDriver::useProgram(OpenGLProgram* p) noexcept {
@@ -1052,49 +1052,17 @@ void OpenGLDriver::createTextureViewSwizzleR(Handle<HwTexture> th, Handle<HwText
    CHECK_GL_ERROR(utils::slog.e)
 }

-void OpenGLDriver::createTextureExternalImageR(Handle<HwTexture> th, SamplerType target,
-    TextureFormat format, uint32_t width, uint32_t height, TextureUsage usage, void* image) {
-    DEBUG_MARKER()
-
-    usage |= TextureUsage::SAMPLEABLE;
-    usage &= ~TextureUsage::UPLOADABLE;
-
-    auto& gl = mContext;
-    GLenum internalFormat = getInternalFormat(format);
-    if (UTILS_UNLIKELY(gl.isES2())) {
-        // on ES2, format and internal format must match
-        // FIXME: handle compressed texture format
-        internalFormat = textureFormatToFormatAndType(format).first;
-    }
-    assert_invariant(internalFormat);
-
-    GLTexture* const t = construct<GLTexture>(th, target, 1, 1, width, height, 1, format, usage);
-    assert_invariant(t);
-
-    t->externalTexture = mPlatform.createExternalImageTexture();
-    if (t->externalTexture) {
-        t->gl.target = t->externalTexture->target;
-        t->gl.id = t->externalTexture->id;
-        // internalFormat actually depends on the external image, but it doesn't matter
-        // because it's not used anywhere for anything important.
-        t->gl.internalFormat = internalFormat;
-        t->gl.baseLevel = 0;
-        t->gl.maxLevel = 0;
-        t->gl.external = true; // forces bindTexture() call (they're never cached)
-    }
-
-    bindTexture(OpenGLContext::DUMMY_TEXTURE_BINDING, t);
-    if (mPlatform.setExternalImage(image, t->externalTexture)) {
-        // the target and id can be reset each time
-        t->gl.target = t->externalTexture->target;
-        t->gl.id = t->externalTexture->id;
-    }
+void OpenGLDriver::createTextureExternalImageR(Handle<HwTexture> th, backend::TextureFormat format,
+        uint32_t width, uint32_t height, backend::TextureUsage usage, void* image) {
+    createTextureR(th, SamplerType::SAMPLER_EXTERNAL, 1, format, 1, width, height, 1, usage);
+    setExternalImage(th, image);
 }

 void OpenGLDriver::createTextureExternalImagePlaneR(Handle<HwTexture> th,
        backend::TextureFormat format, uint32_t width, uint32_t height, backend::TextureUsage usage,
        void* image, uint32_t plane) {
-    // not relevant for the OpenGL backend
+    createTextureR(th, SamplerType::SAMPLER_EXTERNAL, 1, format, 1, width, height, 1, usage);
+    setExternalImagePlane(th, image, plane);
 }

 void OpenGLDriver::importTextureR(Handle<HwTexture> th, intptr_t id,
@@ -1114,7 +1082,6 @@ void OpenGLDriver::importTextureR(Handle<HwTexture> th, intptr_t id,
    switch (target) {
        case SamplerType::SAMPLER_EXTERNAL:
            t->gl.target = GL_TEXTURE_EXTERNAL_OES;
-            t->gl.external = true; // forces bindTexture() call (they're never cached)
            break;
        case SamplerType::SAMPLER_2D:
            t->gl.target = GL_TEXTURE_2D;
@@ -1381,13 +1348,13 @@ void OpenGLDriver::framebufferTexture(TargetBufferInfo const& binfo,
 #ifndef FILAMENT_SILENCE_NOT_SUPPORTED_BY_ES2

                // TODO: support multiview for iOS and WebGL
-#if !defined(__EMSCRIPTEN__) && !defined(FILAMENT_IOS)
+#if !defined(__EMSCRIPTEN__) && !defined(IOS)
                if (layerCount > 1) {
                    // if layerCount > 1, it means we use the multiview extension.
                    glFramebufferTextureMultiviewOVR(GL_FRAMEBUFFER, attachment,
                        t->gl.id, 0, binfo.layer, layerCount);
                } else
-#endif // !defined(__EMSCRIPTEN__) && !defined(FILAMENT_IOS)
+#endif // !defined(__EMSCRIPTEN__) && !defined(IOS)
                {
                    // GL_TEXTURE_2D_MULTISAMPLE_ARRAY is not supported in GLES
                    glFramebufferTextureLayer(GL_FRAMEBUFFER, attachment,
@@ -2807,6 +2774,25 @@ void OpenGLDriver::setupExternalImage(void* image) {
    mPlatform.retainExternalImage(image);
 }

+void OpenGLDriver::setExternalImage(Handle<HwTexture> th, void* image) {
+    DEBUG_MARKER()
+    GLTexture* t = handle_cast<GLTexture*>(th);
+    assert_invariant(t);
+    assert_invariant(t->target == SamplerType::SAMPLER_EXTERNAL);
+
+    bindTexture(OpenGLContext::DUMMY_TEXTURE_BINDING, t);
+    if (mPlatform.setExternalImage(image, t->externalTexture)) {
+        // the target and id can be reset each time
+        t->gl.target = t->externalTexture->target;
+        t->gl.id = t->externalTexture->id;
+        bindTexture(OpenGLContext::DUMMY_TEXTURE_BINDING, t);
+    }
+}
+
+void OpenGLDriver::setExternalImagePlane(Handle<HwTexture> th, void* image, uint32_t plane) {
+    DEBUG_MARKER()
+}
+
 void OpenGLDriver::setExternalStream(Handle<HwTexture> th, Handle<HwStream> sh) {
    auto& gl = mContext;
    if (gl.ext.OES_EGL_image_external_essl3) {
@@ -3526,7 +3512,7 @@ void OpenGLDriver::endFrame(UTILS_UNUSED uint32_t frameId) {
    auto& gl = mContext;
    gl.bindVertexArray(nullptr);
    for (int unit = OpenGLContext::DUMMY_TEXTURE_BINDING; unit >= 0; unit--) {
-        gl.bindTexture(unit, GL_TEXTURE_2D, 0, false);
+        gl.bindTexture(unit, GL_TEXTURE_2D, 0);
    }
    gl.disable(GL_CULL_FACE);
    gl.depthFunc(GL_LESS);
@@ -56,7 +56,7 @@
    #   define FILAMENT_SILENCE_NOT_SUPPORTED_BY_ES2
    #endif

-#elif defined(FILAMENT_IOS)
+#elif defined(IOS)

    #define GLES_SILENCE_DEPRECATION

@@ -81,7 +81,7 @@
 #if defined(GL_VERSION_4_5)
 #elif defined(GL_ES_VERSION_3_1)
 #elif defined(GL_ES_VERSION_3_0)
-#   if !defined(FILAMENT_IOS) && !defined(__EMSCRIPTEN__)
+#   if !defined(IOS) && !defined(__EMSCRIPTEN__)
 #       error "GLES 3.0 headers only supported on iOS and WebGL2"
 #   endif
 #elif defined(GL_ES_VERSION_2_0)
@@ -98,7 +98,7 @@

 #if defined(GL_ES_VERSION_2_0)  // this basically means all versions of GLES

-#if defined(FILAMENT_IOS)
+#if defined(IOS)

 // iOS headers only provide prototypes, nothing to do.

@@ -271,7 +271,7 @@ void glGetBufferSubData(GLenum target, GLintptr offset, GLsizeiptr size, void *d
 #endif

 #ifdef GL_ES_VERSION_2_0
-#   ifndef FILAMENT_IOS
+#   ifndef IOS
 #      ifndef GL_OES_vertex_array_object
 #          error "Headers with GL_OES_vertex_array_object are mandatory unless on iOS"
 #      endif
@@ -715,16 +715,14 @@ void PlatformEGL::destroyExternalImage(ExternalTexture* texture) noexcept {

 bool PlatformEGL::setExternalImage(void* externalImage,
        UTILS_UNUSED_IN_RELEASE ExternalTexture* texture) noexcept {
-
-    // OES_EGL_image_external_essl3 must be present if the target is TEXTURE_EXTERNAL_OES
-    // GL_OES_EGL_image must be present if TEXTURE_2D is used
-
-#if defined(GL_OES_EGL_image) || defined(GL_OES_EGL_image_external_essl3)
+    if (UTILS_LIKELY(ext.gl.OES_EGL_image_external_essl3)) {
+        assert_invariant(texture->target == GL_TEXTURE_EXTERNAL_OES);
        // the texture is guaranteed to be bound here.
-        glEGLImageTargetTexture2DOES(texture->target,
+#ifdef GL_OES_EGL_image
+        glEGLImageTargetTexture2DOES(GL_TEXTURE_EXTERNAL_OES,
                static_cast<GLeglImageOES>(externalImage));
 #endif
-
+    }
    return true;
 }

@@ -64,14 +64,10 @@ public:
                break;
            }
        }
-        if (mLib) {
-            OSMesaGetProcAddress = (GetProcAddressFunc) dlsym(mLib, "OSMesaGetProcAddress");
-        } else {
-            OSMesaGetProcAddress = (GetProcAddressFunc) dlsym(RTLD_LOCAL, "OSMesaGetProcAddress");
-        }
+        FILAMENT_CHECK_PRECONDITION(mLib)
+                << "Unable to dlopen libOSMesa to create a software GL context";

-        FILAMENT_CHECK_PRECONDITION(OSMesaGetProcAddress)
-                << "Unable to against libOSMesa to create a software GL context";
+        OSMesaGetProcAddress = (GetProcAddressFunc) dlsym(mLib, "OSMesaGetProcAddress");

        OSMesaCreateContext = (CreateContextFunc) OSMesaGetProcAddress("OSMesaCreateContext");
        OSMesaDestroyContext =
@@ -80,9 +76,7 @@ public:
    }

    ~OSMesaAPI() {
-        if (mLib) {
-            dlclose(mLib);
-        }
+        dlclose(mLib);
    }
 private:
    void* mLib = nullptr;
@@ -288,7 +288,7 @@ Driver* VulkanDriver::create(VulkanPlatform* platform, VulkanContext const& cont
 }

 ShaderModel VulkanDriver::getShaderModel() const noexcept {
-#if defined(__ANDROID__) || defined(FILAMENT_IOS)
+#if defined(__ANDROID__) || defined(IOS)
    return ShaderModel::MOBILE;
 #else
    return ShaderModel::DESKTOP;
@@ -543,27 +543,8 @@ void VulkanDriver::createTextureViewSwizzleR(Handle<HwTexture> th, Handle<HwText
   texture.inc();
 }

-void VulkanDriver::createTextureExternalImageR(Handle<HwTexture> th,
-        backend::SamplerType target,  backend::TextureFormat format,
-        uint32_t width, uint32_t height, backend::TextureUsage usage, 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(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, usage, mStagePool);
-
-    texture.inc();
+void VulkanDriver::createTextureExternalImageR(Handle<HwTexture> th, backend::TextureFormat format,
+        uint32_t width, uint32_t height, backend::TextureUsage usage, void* image) {
 }

 void VulkanDriver::createTextureExternalImagePlaneR(Handle<HwTexture> th,
@@ -1190,6 +1171,12 @@ TimerQueryResult VulkanDriver::getTimerQueryValue(Handle<HwTimerQuery> tqh, uint
    return TimerQueryResult::AVAILABLE;
 }

+void VulkanDriver::setExternalImage(Handle<HwTexture> th, void* image) {
+}
+
+void VulkanDriver::setExternalImagePlane(Handle<HwTexture> th, void* image, uint32_t plane) {
+}
+
 void VulkanDriver::setExternalStream(Handle<HwTexture> th, Handle<HwStream> sh) {
 }

@@ -75,16 +75,15 @@ void VulkanSwapChain::update() {
    }
    for (auto const color: bundle.colors) {
        auto colorTexture = fvkmemory::resource_ptr<VulkanTexture>::construct(mResourceManager,
-                device, mAllocator, mResourceManager, mCommands, color, VK_NULL_HANDLE,
-                bundle.colorFormat, 1, bundle.extent.width, bundle.extent.height, colorUsage,
+                device, mAllocator, mResourceManager, mCommands, color, bundle.colorFormat, 1,
+                bundle.extent.width, bundle.extent.height, TextureUsage::COLOR_ATTACHMENT,
                mStagePool);
        mColors.push_back(colorTexture);
    }

-    mDepth = fvkmemory::resource_ptr<VulkanTexture>::construct(mResourceManager, device,
-        mAllocator, mResourceManager, mCommands, bundle.depth, VK_NULL_HANDLE,
-        bundle.depthFormat, 1, bundle.extent.width, bundle.extent.height, depthUsage,
-        mStagePool);
+    mDepth = fvkmemory::resource_ptr<VulkanTexture>::construct(mResourceManager, device, mAllocator,
+            mResourceManager, mCommands, bundle.depth, bundle.depthFormat, 1, bundle.extent.width,
+            bundle.extent.height, TextureUsage::DEPTH_ATTACHMENT, mStagePool);

    mExtent = bundle.extent;
 }
@@ -164,15 +164,14 @@ VulkanTextureState::VulkanTextureState(VkDevice device, VmaAllocator allocator,
 // Constructor for internally passed VkImage
 VulkanTexture::VulkanTexture(VkDevice device, VmaAllocator allocator,
        fvkmemory::ResourceManager* resourceManager, VulkanCommands* commands, VkImage image,
-        VkDeviceMemory memory, VkFormat format, uint8_t samples, uint32_t width,
-        uint32_t height, TextureUsage tusage, VulkanStagePool& stagePool)
+        VkFormat format, uint8_t samples, uint32_t width, uint32_t height, TextureUsage tusage,
+        VulkanStagePool& stagePool)
    : HwTexture(SamplerType::SAMPLER_2D, 1, samples, width, height, 1, TextureFormat::UNUSED,
              tusage),
      mState(fvkmemory::resource_ptr<VulkanTextureState>::construct(resourceManager, device,
              allocator, commands, stagePool, format, imgutil::getViewType(SamplerType::SAMPLER_2D),
              1, 1, getDefaultLayoutImpl(tusage), any(usage & TextureUsage::PROTECTED))) {
    mState->mTextureImage = image;
-    mState->mTextureImageMemory = memory;
    mPrimaryViewRange = mState->mFullViewRange;
 }

@@ -94,8 +94,8 @@ struct VulkanTexture : public HwTexture, fvkmemory::Resource {
    // The texture will never destroy the given VkImage, but it does manages its subresources.
    VulkanTexture(VkDevice device, VmaAllocator allocator,
            fvkmemory::ResourceManager* resourceManager, VulkanCommands* commands, VkImage image,
-            VkDeviceMemory memory, VkFormat format, uint8_t samples, uint32_t width, uint32_t height,
-            TextureUsage tusage, VulkanStagePool& stagePool);
+            VkFormat format, uint8_t samples, uint32_t width, uint32_t height, TextureUsage tusage,
+            VulkanStagePool& stagePool);

    // Constructor for creating a texture view for wrt specific mip range
    VulkanTexture(VkDevice device, VkPhysicalDevice physicalDevice, VulkanContext const& context,
@@ -66,8 +66,7 @@ struct Resource {
        : resManager(nullptr),
          id(HandleBase::nullid),
          mCount(0),
-          restype(ResourceType::UNDEFINED_TYPE),
-          mHandleConsideredDestroyed(false) {}
+          restype(ResourceType::UNDEFINED_TYPE) {}

 private:
    inline void inc() noexcept {
@@ -81,16 +80,6 @@ private:
        }
    }

-    // To be able to detect use-after-free, we need a bit to signify if the handle should be
-    // consider destroyed (from Filament's perspective).
-    inline void setHandleConsiderDestroyed() noexcept {
-        mHandleConsideredDestroyed = true;
-    }
-
-    inline bool isHandleConsideredDestroyed() const {
-        return mHandleConsideredDestroyed;
-    }
-
    template <typename T>
    inline void init(HandleId id, ResourceManager* resManager) {
        this->id = id;
@@ -103,9 +92,7 @@ private:
    ResourceManager* resManager; // 8
    HandleId id;                 // 4
    uint32_t mCount      : 24;
-    ResourceType restype : 7;
-    bool mHandleConsideredDestroyed : 1;  // restype + mCount + mHandleConsideredDestroyed
-                                          //   is 4 bytes.
+    ResourceType restype : 6;    // restype + mCount is 4 bytes.

    friend class ResourceManager;

@@ -118,8 +105,7 @@ struct ThreadSafeResource {
        : resManager(nullptr),
          id(HandleBase::nullid),
          mCount(0),
-          restype(ResourceType::UNDEFINED_TYPE),
-          mHandleConsideredDestroyed(false) {}
+          restype(ResourceType::UNDEFINED_TYPE) {}

 private:
    inline void inc() noexcept {
@@ -132,16 +118,6 @@ private:
        }
    }

-    // To be able to detect use-after-free, we need a bit to signify if the handle should be
-    // consider destroyed (from Filament's perspective).
-    inline void setHandleConsiderDestroyed() noexcept {
-        mHandleConsideredDestroyed = true;
-    }
-
-    inline bool isHandleConsideredDestroyed() const {
-        return mHandleConsideredDestroyed;
-    }
-
    template <typename T>
    inline void init(HandleId id, ResourceManager* resManager) {
        this->id = id;
@@ -154,8 +130,7 @@ private:
    ResourceManager* resManager;  // 8
    HandleId id;                  // 4
    std::atomic<uint32_t> mCount; // 4
-    ResourceType restype : 7;
-    bool mHandleConsideredDestroyed : 1;  // restype + mHandleConsideredDestroyed is 1 byte
+    ResourceType restype;         // 1

    friend class ResourceManager;

@@ -21,7 +21,6 @@
 #include "vulkan/memory/ResourceManager.h"

 #include <backend/Handle.h>
-#include <utils/compiler.h>

 #include <utility>

@@ -60,10 +59,6 @@ public:
    static enabled_resource_ptr<B> cast(ResourceManager* resManager,
            Handle<B> const& handle) noexcept {
        D* ptr = resManager->handle_cast<D*, B>(handle);
-        FILAMENT_CHECK_PRECONDITION(!ptr->isHandleConsideredDestroyed())
-                << "Handle id=" << ptr->id << " (" << getTypeStr(ptr->restype)
-                << ") is being used after it has been freed";
-
        return {ptr};
    }

@@ -161,8 +156,6 @@ private:
    // only be used from VulkanDriver.
    inline void dec() {
        assert_invariant(mRef);
-        assert_invariant(!mRef->isHandleConsideredDestroyed());
-        mRef->setHandleConsiderDestroyed();
        mRef->dec();
    }

@@ -194,15 +194,6 @@ ExtensionSet getDeviceExtensions(VkPhysicalDevice device) {
 #if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
            VK_EXT_DEBUG_MARKER_EXTENSION_NAME,
 #endif
-    // We only support external image for Android for now, but nothing bars us from 
-    // supporting other platforms.
-#if defined(__ANDROID__)
-            VK_KHR_EXTERNAL_MEMORY_EXTENSION_NAME,
-            VK_KHR_DEDICATED_ALLOCATION_EXTENSION_NAME,
-            VK_KHR_SAMPLER_YCBCR_CONVERSION_EXTENSION_NAME,
-            VK_ANDROID_EXTERNAL_MEMORY_ANDROID_HARDWARE_BUFFER_EXTENSION_NAME,
- #endif
-
            VK_KHR_PORTABILITY_SUBSET_EXTENSION_NAME,
            VK_KHR_MAINTENANCE1_EXTENSION_NAME,
            VK_KHR_MAINTENANCE2_EXTENSION_NAME,
@@ -954,16 +945,6 @@ VkQueue VulkanPlatform::getProtectedGraphicsQueue() const noexcept {
    return mImpl->mProtectedGraphicsQueue;
 }

-VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadata(
-        void* externalImage) {
-    return getExternalImageMetadataImpl(externalImage, mImpl->mDevice);
-}
-
-VulkanPlatform::ImageData VulkanPlatform::createExternalImage(void* externalImage,
-            const ExternalImageMetadata& metadata) {
-    return createExternalImageImpl(externalImage, mImpl->mDevice, nullptr, metadata);
-}
-
 #undef SWAPCHAIN_RET_FUNC

 }// namespace filament::backend
@@ -1,240 +0,0 @@
-/*
- * Copyright (C) 2024 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/VulkanPlatform.h>
-
-#include "vulkan/VulkanConstants.h"
-
-#include <utils/Panic.h>
-
-#include <bluevk/BlueVK.h>
-
-#include <android/hardware_buffer.h>
-#include <android/native_window.h>
-
-using namespace bluevk;
-
-namespace filament::backend {
-
-namespace {
-void getVKFormatAndUsage(const AHardwareBuffer_Desc& desc, VkFormat& format,
-        VkImageUsageFlags& usage, bool& isProtected) {
-    // 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;
-            break;
-        case AHARDWAREBUFFER_FORMAT_R8G8B8X8_UNORM:
-            format = VK_FORMAT_R8G8B8A8_UNORM;
-            break;
-        case AHARDWAREBUFFER_FORMAT_R8G8B8_UNORM:
-            format = VK_FORMAT_R8G8B8_UNORM;
-            break;
-        case AHARDWAREBUFFER_FORMAT_R5G6B5_UNORM:
-            format = VK_FORMAT_R5G6B5_UNORM_PACK16;
-            break;
-        case AHARDWAREBUFFER_FORMAT_R16G16B16A16_FLOAT:
-            format = VK_FORMAT_R16G16B16A16_SFLOAT;
-            break;
-        case AHARDWAREBUFFER_FORMAT_R10G10B10A2_UNORM:
-            format = VK_FORMAT_A2B10G10R10_UNORM_PACK32;
-            break;
-        case AHARDWAREBUFFER_FORMAT_D16_UNORM:
-            isDepthFormat = true;
-            format = VK_FORMAT_D16_UNORM;
-            break;
-        case AHARDWAREBUFFER_FORMAT_D24_UNORM:
-            isDepthFormat = true;
-            format = VK_FORMAT_X8_D24_UNORM_PACK32;
-            break;
-        case AHARDWAREBUFFER_FORMAT_D24_UNORM_S8_UINT:
-            isDepthFormat = true;
-            format = VK_FORMAT_D24_UNORM_S8_UINT;
-            break;
-        case AHARDWAREBUFFER_FORMAT_D32_FLOAT:
-            isDepthFormat = true;
-            format = VK_FORMAT_D32_SFLOAT;
-            break;
-        case AHARDWAREBUFFER_FORMAT_D32_FLOAT_S8_UINT:
-            isDepthFormat = true;
-            format = VK_FORMAT_D32_SFLOAT_S8_UINT;
-            break;
-        case AHARDWAREBUFFER_FORMAT_S8_UINT:
-            isDepthFormat = true;
-            format = VK_FORMAT_S8_UINT;
-            break;
-        default:
-            format = VK_FORMAT_UNDEFINED;
-    }
-
-    // The following only concern usage flags derived from Table 14.
-    usage = 0;
-    if (desc.usage & AHARDWAREBUFFER_USAGE_GPU_SAMPLED_IMAGE) {
-        usage |= VK_IMAGE_USAGE_SAMPLED_BIT;
-        usage |= VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT;
-    }
-    if (desc.usage & AHARDWAREBUFFER_USAGE_GPU_FRAMEBUFFER) {
-        if (isDepthFormat) {
-            usage |= VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
-        } else {
-            usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
-        }
-    }
-    if (desc.usage & AHARDWAREBUFFER_USAGE_GPU_DATA_BUFFER) {
-        usage = VK_IMAGE_USAGE_STORAGE_BIT;
-    }
-    if (desc.usage & AHARDWAREBUFFER_USAGE_PROTECTED_CONTENT) {
-        isProtected = true;
-    }
-}
-
-VulkanPlatform::ImageData allocateExternalImage(void* externalBuffer,
-        VkDevice device, const VkAllocationCallbacks* allocator,
-        VulkanPlatform::ExternalImageMetadata const& metadata) {
-    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;
-
-    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)
-    };
-    const VkExternalMemoryImageCreateInfo externalCreateInfo = {
-        .sType = VK_STRUCTURE_TYPE_EXTERNAL_MEMORY_IMAGE_CREATE_INFO,
-        .pNext = useExternalFormat ? &externalFormat : nullptr,
-        .handleTypes = VK_EXTERNAL_MEMORY_HANDLE_TYPE_ANDROID_HARDWARE_BUFFER_BIT_ANDROID,
-    };
-
-    VkImageCreateInfo imageInfo{.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO};
-    imageInfo.pNext = &externalCreateInfo;
-    imageInfo.format = metadata.format;
-    imageInfo.imageType = VK_IMAGE_TYPE_2D;
-    imageInfo.extent = {
-        metadata.width,
-        metadata.height,
-        1u,
-    };
-    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;
-
-    VkResult result = vkCreateImage(device, &imageInfo, allocator, &data.first);
-    FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
-            << "vkCreateImage failed with error=" << result;
-
-    // Allocate the memory
-    VkImportAndroidHardwareBufferInfoANDROID androidHardwareBufferInfo = {
-        .sType = VK_STRUCTURE_TYPE_IMPORT_ANDROID_HARDWARE_BUFFER_INFO_ANDROID,
-        .pNext = nullptr,
-        .buffer = buffer,
-    };
-    VkMemoryDedicatedAllocateInfo memoryDedicatedAllocateInfo = {
-        .sType = VK_STRUCTURE_TYPE_MEMORY_DEDICATED_ALLOCATE_INFO,
-        .pNext = &androidHardwareBufferInfo,
-        .image = data.first,
-        .buffer = VK_NULL_HANDLE,
-    };
-    VkMemoryAllocateInfo allocInfo = {.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
-        .pNext = &memoryDedicatedAllocateInfo,
-        .allocationSize = metadata.allocationSize,
-        .memoryTypeIndex = metadata.memoryTypeBits};
-    result = vkAllocateMemory(device, &allocInfo, allocator, &data.second);
-    FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
-            << "vkAllocateMemory failed with error=" << result;
-
-    return data;
-}
-
-}// namespace
-
-VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadataImpl(void* externalImage,
-        VkDevice device) {
-    ExternalImageMetadata metadata;
-    AHardwareBuffer* buffer = static_cast<AHardwareBuffer*>(externalImage);
-    if (__builtin_available(android 26, *)) {
-        AHardwareBuffer_Desc bufferDesc;
-        AHardwareBuffer_describe(buffer, &bufferDesc);
-        metadata.width = bufferDesc.width;
-        metadata.height = bufferDesc.height;
-        metadata.layers = bufferDesc.layers;
-
-        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;
-    }
-
-    VkAndroidHardwareBufferFormatPropertiesANDROID formatInfo = {
-        .sType = VK_STRUCTURE_TYPE_ANDROID_HARDWARE_BUFFER_FORMAT_PROPERTIES_ANDROID,
-        .pNext = nullptr,
-    };
-    VkAndroidHardwareBufferPropertiesANDROID properties = {
-        .sType = VK_STRUCTURE_TYPE_ANDROID_HARDWARE_BUFFER_PROPERTIES_ANDROID,
-        .pNext = &formatInfo,
-    };
-    VkResult result = vkGetAndroidHardwareBufferPropertiesANDROID(device, buffer, &properties);
-    FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
-            << "vkGetAndroidHardwareBufferProperties failed with error=" << result;
-
-    FILAMENT_CHECK_POSTCONDITION(metadata.format == formatInfo.format);
-    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);
-    FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
-            << "vkBindImageMemory error=" << result << ".";
-    return data;
-}
-
-VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensions() {
-    return {
-        VK_KHR_ANDROID_SURFACE_EXTENSION_NAME,
-    };
-}
-
-VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHR(void* nativeWindow,
-    VkInstance instance, uint64_t flags) noexcept {
-    VkSurfaceKHR surface;
-    VkExtent2D extent;
-
-    VkAndroidSurfaceCreateInfoKHR const createInfo{
-        .sType = VK_STRUCTURE_TYPE_ANDROID_SURFACE_CREATE_INFO_KHR,
-        .window = (ANativeWindow*) nativeWindow,
-    };
-    VkResult const result =
-            vkCreateAndroidSurfaceKHR(instance, &createInfo, VKALLOC, (VkSurfaceKHR*) &surface);
-    FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
-            << "vkCreateAndroidSurfaceKHR with error=" << result;
-    return {surface, extent};
-}
-}
@@ -34,7 +34,9 @@
 #endif

 // Platform specific includes and defines
-#if defined(__linux__) && defined(FILAMENT_SUPPORTS_WAYLAND)
+#if defined(__ANDROID__)
+    #include <android/native_window.h>
+#elif defined(__linux__) && defined(FILAMENT_SUPPORTS_WAYLAND)
    #include <dlfcn.h>
    namespace {
        typedef struct _wl {
@@ -84,20 +86,11 @@ using namespace bluevk;

 namespace filament::backend {

-VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadataImpl(
-        void* externalImage, VkDevice device) {
-    return {};
-}
-
-VulkanPlatform::ImageData VulkanPlatform::createExternalImageImpl(void* externalImage,
-        VkDevice device, const VkAllocationCallbacks* allocator,
-        const ExternalImageMetadata& metadata) {
-    return {};
-}
-
 VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensions() {
    VulkanPlatform::ExtensionSet const ret = {
-#if defined(__linux__) && defined(FILAMENT_SUPPORTS_WAYLAND)
+#if defined(__ANDROID__)
+        VK_KHR_ANDROID_SURFACE_EXTENSION_NAME,
+#elif defined(__linux__) && defined(FILAMENT_SUPPORTS_WAYLAND)
        VK_KHR_WAYLAND_SURFACE_EXTENSION_NAME,
 #elif defined(LINUX_OR_FREEBSD) && defined(FILAMENT_SUPPORTS_X11)
    #if defined(FILAMENT_SUPPORTS_XCB)
@@ -123,7 +116,15 @@ VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHR(void* nativeWin
    // swap chain extent.
    VkExtent2D extent;

-    #if defined(__linux__) && defined(FILAMENT_SUPPORTS_WAYLAND)
+    #if defined(__ANDROID__)
+        VkAndroidSurfaceCreateInfoKHR const createInfo{
+                .sType = VK_STRUCTURE_TYPE_ANDROID_SURFACE_CREATE_INFO_KHR,
+                .window = (ANativeWindow*) nativeWindow,
+        };
+        VkResult const result = vkCreateAndroidSurfaceKHR(instance, &createInfo, VKALLOC,
+                (VkSurfaceKHR*) &surface);
+        FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS) << "vkCreateAndroidSurfaceKHR error.";
+    #elif defined(__linux__) && defined(FILAMENT_SUPPORTS_WAYLAND)
        wl* ptrval = reinterpret_cast<wl*>(nativeWindow);
        extent.width = ptrval->width;
        extent.height = ptrval->height;
@@ -32,7 +32,7 @@
    #ifndef VK_MVK_macos_surface
         #error VK_MVK_macos_surface is not defined
    #endif
-#elif defined(FILAMENT_IOS)
+#elif defined(IOS)
    // Metal is not available when building for the iOS simulator on Desktop.
    #define METAL_AVAILABLE __has_include(<QuartzCore/CAMetalLayer.h>)
    #if METAL_AVAILABLE
@@ -56,24 +56,13 @@ VulkanPlatform::ExtensionSet VulkanPlatform::getSwapchainInstanceExtensions() {
    ExtensionSet const ret = {
    #if defined(__APPLE__)
        VK_MVK_MACOS_SURFACE_EXTENSION_NAME,  // TODO: replace with VK_EXT_metal_surface
-    #elif defined(FILAMENT_IOS) && defined(METAL_AVAILABLE)
+    #elif defined(IOS) && defined(METAL_AVAILABLE)
        VK_MVK_IOS_SURFACE_EXTENSION_NAME,
    #endif
    };
    return ret;
 }

-VulkanPlatform::ExternalImageMetadata VulkanPlatform::getExternalImageMetadataImpl(
-        void* externalImage, VkDevice device) {
-    return {};
-}
-
-VulkanPlatform::ImageData VulkanPlatform::createExternalImageImpl(void* externalImage,
-        VkDevice device, const VkAllocationCallbacks* allocator,
-        const ExternalImageMetadata& metadata) {
-    return {};
-}
-
 VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHR(void* nativeWindow,
        VkInstance instance, uint64_t flags) noexcept {
    VkSurfaceKHR surface;
@@ -90,7 +79,7 @@ VulkanPlatform::SurfaceBundle VulkanPlatform::createVkSurfaceKHR(void* nativeWin
        VkResult result = vkCreateMacOSSurfaceMVK((VkInstance) instance, &createInfo, VKALLOC,
                (VkSurfaceKHR*) &surface);
        FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS) << "vkCreateMacOSSurfaceMVK error.";
-    #elif defined(FILAMENT_IOS) && defined(METAL_AVAILABLE)
+    #elif defined(IOS) && defined(METAL_AVAILABLE)
        CAMetalLayer* metalLayer = (CAMetalLayer*) nativeWindow;
        // Create the VkSurface.
        FILAMENT_CHECK_POSTCONDITION(vkCreateIOSSurfaceMVK)
@@ -33,7 +33,7 @@ static constexpr size_t CONFIG_COMMAND_BUFFERS_SIZE     = 3 * CONFIG_MIN_COMMAND
 using namespace filament;
 using namespace filament::backend;

-#ifndef FILAMENT_IOS
+#ifndef IOS
 #include <imageio/ImageEncoder.h>
 #include <image/ColorTransform.h>

@@ -177,7 +177,7 @@ void BackendTest::readPixelsAndAssertHash(const char* testName, size_t width, si

                // Export a screenshot, if requested.
                if (c->exportScreenshot) {
-#ifndef FILAMENT_IOS
+#ifndef IOS
                    LinearImage image(c->width, c->height, 4);
                    image = toLinearWithAlpha<uint8_t>(c->width, c->height, c->width * 4,
                            (uint8_t*) buffer);
@@ -24,7 +24,7 @@

 #include <fstream>

-#ifndef FILAMENT_IOS
+#ifndef IOS
 #include <imageio/ImageEncoder.h>
 #include <image/ColorTransform.h>
 #endif
@@ -67,7 +67,7 @@ struct ScreenshotParams {
    uint32_t pixelHashResult;
 };

-#ifdef FILAMENT_IOS
+#ifdef IOS
 static void dumpScreenshot(DriverApi& dapi, Handle<HwRenderTarget> rt, ScreenshotParams* params) {}
 #else
 static void dumpScreenshot(DriverApi& dapi, Handle<HwRenderTarget> rt, ScreenshotParams* params) {
@@ -31,7 +31,7 @@
 #include <stddef.h>
 #include <stdint.h>

-#ifndef FILAMENT_IOS
+#ifndef IOS
 #include <imageio/ImageEncoder.h>
 #include <image/ColorTransform.h>

@@ -112,7 +112,7 @@ static void dumpScreenshot(DriverApi& dapi, Handle<HwRenderTarget> rt) {
        int w = kTexWidth, h = kTexHeight;
        const uint32_t* texels = (uint32_t*) buffer;
        sPixelHashResult = utils::hash::murmur3(texels, size / 4, 0);
-#ifndef FILAMENT_IOS
+#ifndef IOS
        LinearImage image(w, h, 4);
        image = toLinearWithAlpha<uint8_t>(w, h, w * 4, (uint8_t*) buffer);
        std::ofstream pngstrm("feedback.png", std::ios::binary | std::ios::trunc);
@@ -28,7 +28,7 @@
 using namespace filament;
 using namespace filament::backend;

-#ifndef FILAMENT_IOS
+#ifndef IOS
 #include <imageio/ImageEncoder.h>
 #include <image/ColorTransform.h>

@@ -137,7 +137,7 @@ TEST_F(ReadPixelsTest, ReadPixels) {
        }

        void exportScreenshot(void* pixelData) const {
-            #ifndef FILAMENT_IOS
+            #ifndef IOS
            const size_t width = readRect.width, height = readRect.height;
            LinearImage image(width, height, 4);
            if (format == PixelDataFormat::RGBA && type == PixelDataType::UBYTE) {
@@ -227,8 +227,7 @@ TEST_F(BackendTest, RenderExternalImage) {

    api.setupExternalImage(pixBuffer);
    backend::Handle<HwTexture> texture =
-            api.createTextureExternalImage(SamplerType::SAMPLER_EXTERNAL,
-                    TextureFormat::RGBA8, 1024, 1024, usage, pixBuffer);
+            api.createTextureExternalImage(TextureFormat::RGBA8, 1024, 1024, usage, pixBuffer);

    // We're now free to release the buffer.
    CVBufferRelease(pixBuffer);
@@ -61,7 +61,7 @@ namespace filament {
 *  filament::Camera* myCamera = engine->createCamera(myCameraEntity);
 *  myCamera->setProjection(45, 16.0/9.0, 0.1, 1.0);
 *  myCamera->lookAt({0, 1.60, 1}, {0, 0, 0});
- *  engine->destroyCameraComponent(myCameraEntity);
+ *  engine->destroyCameraComponent(myCamera);
 * ~~~~~~~~~~~
 *
 *
@@ -103,11 +103,6 @@ public:
        Builder& operator=(Builder const& rhs) noexcept;
        Builder& operator=(Builder&& rhs) noexcept;

-        enum class ShadowSamplingQuality : uint8_t {
-            HARD,   // 2x2 PCF
-            LOW     // 3x3 gaussian filter
-        };
-
        /**
         * Specifies the material data. The material data is a binary blob produced by
         * libfilamat or by matc.
@@ -157,14 +152,6 @@ public:
         */
        Builder& sphericalHarmonicsBandCount(size_t shBandCount) noexcept;

-        /**
-         * Set the quality of shadow sampling. This is only taken into account
-         * if this material is lit and in the surface domain.
-         * @param quality
-         * @return
-         */
-        Builder& shadowSamplingQuality(ShadowSamplingQuality quality) noexcept;
-
        /**
         * Creates the Material object and returns a pointer to it.
         *
@@ -438,7 +438,7 @@ struct MultiSampleAntiAliasingOptions {
 * @see setTemporalAntiAliasingOptions()
 */
 struct TemporalAntiAliasingOptions {
-    float filterWidth = 1.0f;   //!< reconstruction filter width typically between 1 (sharper) and 2 (smoother)
+    float filterWidth = 1.0f;   //!< reconstruction filter width typically between 0.2 (sharper, aliased) and 1.5 (smoother)
    float feedback = 0.12f;     //!< history feedback, between 0 (maximum temporal AA) and 1 (no temporal AA).
    float lodBias = -1.0f;      //!< texturing lod bias (typically -1 or -2)
    float sharpness = 0.0f;     //!< post-TAA sharpen, especially useful when upscaling is true.
@@ -261,18 +261,6 @@ public:
         */
        Builder& import(intptr_t id) noexcept;

-        /**
-         * Creates an external texture. The content must be set using setExternalImage().
-         * The sampler can be SAMPLER_EXTERNAL or SAMPLER_2D depending on the format. Generally
-         * YUV formats must use SAMPLER_EXTERNAL. This depends on the backend features and is not
-         * validated.
-         *
-         * If the Sampler is set to SAMPLER_EXTERNAL, external() is implied.
-         *
-         * @return
-         */
-        Builder& external() noexcept;
-
    private:
        friend class FTexture;
    };
@@ -407,7 +395,7 @@ public:


    /**
-     * Specify the external image to associate with this Texture. Typically, the external
+     * Specify the external image to associate with this Texture. Typically the external
     * image is OS specific, and can be a video or camera frame.
     * There are many restrictions when using an external image as a texture, such as:
     *   - only the level of detail (lod) 0 can be specified
@@ -432,7 +420,7 @@ public:
    void setExternalImage(Engine& engine, void* UTILS_NONNULL image) noexcept;

    /**
-     * Specify the external image and plane to associate with this Texture. Typically, the external
+     * Specify the external image and plane to associate with this Texture. Typically the external
     * image is OS specific, and can be a video or camera frame. When using this method, the
     * external image must be a planar type (such as a YUV camera frame). The plane parameter
     * selects which image plane is bound to this texture.
@@ -2622,7 +2622,7 @@ void PostProcessManager::TaaJitterCamera(
    current.projection = inoutCameraInfo->projection * inoutCameraInfo->getUserViewMatrix();
    current.frameId = previous.frameId + 1;

-    auto jitterPosition = [pattern = taaOptions.jitterPattern](size_t frameIndex) -> float2 {
+    auto jitterPosition = [pattern = taaOptions.jitterPattern](size_t frameIndex){
        using JitterPattern = TemporalAntiAliasingOptions::JitterPattern;
        switch (pattern) {
            case JitterPattern::RGSS_X4:
@@ -2636,7 +2636,6 @@ void PostProcessManager::TaaJitterCamera(
            case JitterPattern::HALTON_23_X32:
                return sHaltonSamples(frameIndex);
        }
-        return { 0.0f, 0.0f };
    };

    // sample position within a pixel [-0.5, 0.5]
@@ -2760,31 +2759,15 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::taa(FrameGraph& fg,
                }};

                constexpr float2 sampleOffsets[9] = {
-                        { -1.0f, -1.0f }, {  0.0f, -1.0f }, {  1.0f, -1.0f }, { -1.0f,  0.0f },
-                        {  0.0f,  0.0f },
-                        {  1.0f,  0.0f }, { -1.0f,  1.0f }, {  0.0f,  1.0f }, {  1.0f,  1.0f },
+                        { -1.0f, -1.0f }, {  0.0f, -1.0f }, {  1.0f, -1.0f },
+                        { -1.0f,  0.0f }, {  0.0f,  0.0f }, {  1.0f,  0.0f },
+                        { -1.0f,  1.0f }, {  0.0f,  1.0f }, {  1.0f,  1.0f },
                };

                constexpr float2 subSampleOffsets[4] = {
-                        { -0.25f,  0.25f },
-                        {  0.25f,  0.25f },
-                        {  0.25f, -0.25f },
-                        { -0.25f, -0.25f }
+                        { -0.25f, 0.25f }, {  0.25f, 0.25f }, { 0.25f, -0.25f }, { -0.25f, -0.25f }
                };

-                UTILS_UNUSED
-                auto const lanczos = [](float x, float a) -> float {
-                    if (x <= std::numeric_limits<float>::epsilon()) {
-                        return 1.0f;
-                    }
-                    if (std::abs(x) <= a) {
-                        return (a * std::sin(f::PI * x) * std::sin(f::PI * x / a))
-                               / ((f::PI * f::PI) * (x * x));
-                    }
-                    return 0.0f;
-                };
-
-                float const filterWidth = std::clamp(taaOptions.filterWidth, 1.0f, 2.0f);
                float4 sum = 0.0;
                float4 weights[9];

@@ -2794,9 +2777,11 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::taa(FrameGraph& fg,
                for (size_t i = 0; i < 9; i++) {
                    float2 const o = sampleOffsets[i];
                    for (size_t j = 0; j < 4; j++) {
-                        float2 const subPixelOffset = taaOptions.upscaling ? subSampleOffsets[j] : float2{ 0 };
-                        float2 const d = (o - (current.jitter - subPixelOffset)) / filterWidth;
-                        weights[i][j] = lanczos(length(d), filterWidth);
+                        float2 const s = taaOptions.upscaling ? subSampleOffsets[j] : float2{ 0 };
+                        float2 const d = (o - current.jitter - s) / taaOptions.filterWidth;
+                        // This is a gaussian fit of a 3.3-wide Blackman-Harris window
+                        // see: "High Quality Temporal Supersampling" by Brian Karis
+                        weights[i][j] = std::exp(-2.29f * (d.x * d.x + d.y * d.y));
                    }
                    sum += weights[i];
                }
@@ -428,7 +428,7 @@ private:

    template<size_t SIZE>
    struct JitterSequence {
-        math::float2 operator()(size_t i) const noexcept { return positions[i % SIZE] - 0.5f; }
+        auto operator()(size_t i) const noexcept { return positions[i % SIZE] - 0.5f; }
        const std::array<math::float2, SIZE> positions;
    };

@@ -680,6 +680,8 @@ RenderPass::Command* RenderPass::generateCommandsImpl(RenderPass::CommandTypeFla
        cmd.info.hasMorphing = (bool)morphing.handle;
        cmd.info.hasSkinning = (bool)skinning.handle;

+        assert_invariant(cmd.info.hasHybridInstancing || cmd.info.instanceCount <= 1);
+
        // soaInstanceInfo[i].count is the number of instances the user has requested, either for
        // manual or hybrid instancing. Instanced stereo multiplies the number of instances by the
        // eye count.
@@ -862,44 +864,21 @@ backend::Viewport RenderPass::Executor::applyScissorViewport(
        backend::Viewport const& scissorViewport, backend::Viewport const& scissor) noexcept {
    // scissor is set, we need to apply the offset/clip
    // clang vectorizes this!
-    constexpr int64_t maxvali = std::numeric_limits<int32_t>::max();
-
-    // we do all the offsetting/clipping math in 64 bits below to avoid overflows
-    struct {
-        int64_t l;
-        int64_t b;
-        int64_t r;
-        int64_t t;
-    } svp, s;
-
-    // convert scissorViewport to {left,bottom,right,top} format
-    svp.l = scissorViewport.left;
-    svp.b = scissorViewport.bottom;
-    svp.r = svp.l + scissorViewport.width;
-    svp.t = svp.b + scissorViewport.height;
-
-    // convert scissor to {left,bottom,right,top} format and offset by scissorViewport's left,bottom
-    s.l = svp.l + scissor.left;
-    s.b = svp.b + scissor.bottom;
-    s.r = s.l + scissor.width;
-    s.t = s.b + scissor.height;
-
-    // clip to the scissorViewport
-    s.l = std::max(s.l, svp.l);
-    s.b = std::max(s.b, svp.b);
-    s.r = std::min(s.r, svp.r);
-    s.t = std::min(s.t, svp.t);
-
-    // clip to positive int32_t
-    s.l = std::max(s.l, int64_t(0));
-    s.b = std::max(s.b, int64_t(0));
-    s.r = std::min(s.r, maxvali);
-    s.t = std::min(s.t, maxvali);
-
-    assert_invariant(s.r >= s.l && s.t >= s.b);
-
-    // convert back to Viewport format
-    return { int32_t(s.l), int32_t(s.b), uint32_t(s.r - s.l), uint32_t(s.t - s.b) };
+    constexpr int32_t maxvali = std::numeric_limits<int32_t>::max();
+    // compute new left/bottom, assume no overflow
+    int32_t const l = scissor.left + scissorViewport.left;
+    int32_t const b = scissor.bottom + scissorViewport.bottom;
+    // compute right/top without overflowing, scissor.width/height guaranteed
+    // to convert to int32
+    int32_t r = (l > maxvali - int32_t(scissor.width)) ? maxvali : l + int32_t(scissor.width);
+    int32_t t = (b > maxvali - int32_t(scissor.height)) ? maxvali : b + int32_t(scissor.height);
+    // clip to the viewport
+    assert_invariant(l == std::max(l, scissorViewport.left));
+    assert_invariant(b == std::max(b, scissorViewport.bottom));
+    r = std::min(r, scissorViewport.left + int32_t(scissorViewport.width));
+    t = std::min(t, scissorViewport.bottom + int32_t(scissorViewport.height));
+    assert_invariant(r >= l && t >= b);
+    return { l, b, uint32_t(r - l), uint32_t(t - b) };
 }

 UTILS_NOINLINE // no need to be inlined
@@ -427,7 +427,7 @@ float FLightManager::getSpotLightInnerCone(Instance i) const noexcept {

 void LightManager::ShadowCascades::computeUniformSplits(float splitPositions[3], uint8_t cascades) {
    size_t s = 0;
-    cascades = min(cascades, (uint8_t) 4u);
+    cascades = max(cascades, (uint8_t) 4u);
    for (size_t c = 1; c < cascades; c++) {
        splitPositions[s++] = float(c) / float(cascades);
    }
@@ -436,7 +436,7 @@ void LightManager::ShadowCascades::computeUniformSplits(float splitPositions[3],
 void LightManager::ShadowCascades::computeLogSplits(float splitPositions[3], uint8_t cascades,
        float near, float far) {
    size_t s = 0;
-    cascades = min(cascades, (uint8_t) 4u);
+    cascades = max(cascades, (uint8_t) 4u);
    for (size_t c = 1; c < cascades; c++) {
        splitPositions[s++] =
            (near * std::pow(far / near, float(c) / float(cascades)) - near) / (far - near);
@@ -447,7 +447,7 @@ void LightManager::ShadowCascades::computePracticalSplits(float splitPositions[3
        float near, float far, float lambda) {
    float uniformSplits[3];
    float logSplits[3];
-    cascades = min(cascades, (uint8_t) 4u);
+    cascades = max(cascades, (uint8_t) 4u);
    computeUniformSplits(uniformSplits, cascades);
    computeLogSplits(logSplits, cascades, near, far);
    size_t s = 0;
@@ -1159,13 +1159,8 @@ bool FEngine::destroy(const FMaterial* ptr) {
            if (!ASSERT_PRECONDITION_NON_FATAL(pos->second.empty(),
                    "destroying material \"%s\" but %u instances still alive",
                    ptr->getName().c_str(), (*pos).second.size())) {
-                // We return true here, because the material has been destroyed. However, this
-                // leaks this material's ResourceList<FMaterialInstance> until the engine
-                // is shut down. Note that it's still possible for the MaterialInstances to be
-                // destroyed manually.
-                return true;
+                return false;
            }
-            mMaterialInstances.erase(pos);
        }
    }
    return success;
@@ -123,7 +123,6 @@ struct Material::BuilderDetails {
    size_t mSize = 0;
    bool mDefaultMaterial = false;
    int32_t mShBandsCount = 3;
-    Builder::ShadowSamplingQuality mShadowSamplingQuality = Builder::ShadowSamplingQuality::LOW;
    std::unordered_map<
        utils::CString,
        std::variant<int32_t, float, bool>,
@@ -153,11 +152,6 @@ Material::Builder& Material::Builder::sphericalHarmonicsBandCount(size_t shBandC
    return *this;
 }

-Material::Builder& Material::Builder::shadowSamplingQuality(ShadowSamplingQuality quality) noexcept {
-    mImpl->mShadowSamplingQuality = quality;
-    return *this;
-}
-
 template<typename T, typename>
 Material::Builder& Material::Builder::constant(const char* name, size_t nameLength, T value) {
    FILAMENT_CHECK_PRECONDITION(name != nullptr) << "name cannot be null";
@@ -556,7 +550,7 @@ Program FMaterial::getProgramWithVariants(
            << "The material '" << mName.c_str()
            << "' has not been compiled to include the required GLSL or SPIR-V chunks for the "
               "vertex shader (variant="
-            << +variant.key << ", filtered=" << +vertexVariant.key << ").";
+            << variant.key << ", filtered=" << vertexVariant.key << ").";

    /*
     * Fragment shader
@@ -571,7 +565,7 @@ Program FMaterial::getProgramWithVariants(
            << "The material '" << mName.c_str()
            << "' has not been compiled to include the required GLSL or SPIR-V chunks for the "
               "fragment shader (variant="
-            << +variant.key << ", filtered=" << +fragmentVariant.key << ").";
+            << variant.key << ", filtered=" << ").";

    Program program;
    program.shader(ShaderStage::VERTEX, vsBuilder.data(), vsBuilder.size())
@@ -611,10 +605,10 @@ void FMaterial::createAndCacheProgram(Program&& p, Variant variant) const noexce
    FEngine const& engine = mEngine;
    DriverApi& driverApi = mEngine.getDriverApi();

-    bool const isShared = isSharedVariant(variant);
-
    // Check if the default material has this program cached
-    if (isShared) {
+    if (mMaterialDomain == MaterialDomain::SURFACE &&
+            !mIsDefaultMaterial && !mHasCustomDepthShader &&
+            Variant::isValidDepthVariant(variant)) {
        FMaterial const* const pDefaultMaterial = engine.getDefaultMaterial();
        if (pDefaultMaterial) {
            auto program = pDefaultMaterial->mCachedPrograms[variant.key];
@@ -633,7 +627,9 @@ void FMaterial::createAndCacheProgram(Program&& p, Variant variant) const noexce
    // If the default material doesn't already have this program cached, and all caching conditions
    // are met (Surface Domain and no custom depth shader), cache it now.
    // New Materials will inherit these program automatically.
-    if (isShared) {
+    if (mMaterialDomain == MaterialDomain::SURFACE &&
+            !mIsDefaultMaterial && !mHasCustomDepthShader &&
+            Variant::isValidDepthVariant(variant)) {
        FMaterial const* const pDefaultMaterial = engine.getDefaultMaterial();
        if (pDefaultMaterial && !pDefaultMaterial->mCachedPrograms[variant.key]) {
            // set the tag to the default material name
@@ -1002,11 +998,7 @@ void FMaterial::processSpecializationConstants(FEngine& engine, Material::Builde
            +ReservedSpecializationConstants::CONFIG_STEREO_EYE_COUNT,
            (int)engine.getConfig().stereoscopicEyeCount });
    mSpecializationConstants.push_back({
-            +ReservedSpecializationConstants::CONFIG_SH_BANDS_COUNT,
-            builder->mShBandsCount });
-    mSpecializationConstants.push_back({
-            +ReservedSpecializationConstants::CONFIG_SHADOW_SAMPLING_METHOD,
-            (int32_t)builder->mShadowSamplingQuality });
+            +ReservedSpecializationConstants::CONFIG_SH_BANDS_COUNT, builder->mShBandsCount });
    if (UTILS_UNLIKELY(parser->getShaderLanguage() == ShaderLanguage::ESSL1)) {
        // The actual value of this spec-constant is set in the OpenGLDriver backend.
        mSpecializationConstants.push_back({
@@ -1084,21 +1076,6 @@ void FMaterial::processPushConstants(FEngine& engine, MaterialParser const* pars
 }

 void FMaterial::precacheDepthVariants(FEngine& engine) {
-    // pre-cache all depth variants inside the default material. Note that this should be
-    // entirely optional; if we remove this pre-caching, these variants will be populated
-    // later, when/if needed by createAndCacheProgram(). Doing it now potentially uses more
-    // memory and increases init time, but reduces hiccups during the first frame.
-    if (UTILS_UNLIKELY(mIsDefaultMaterial)) {
-        auto const allDepthVariants = VariantUtils::getDepthVariants();
-        for (auto const variant: allDepthVariants) {
-            assert_invariant(Variant::isValidDepthVariant(variant));
-            if (hasVariant(variant)) {
-                prepareProgram(variant);
-            }
-        }
-        return;
-    }
-
    // if possible pre-cache all depth variants from the default material
    if (mMaterialDomain == MaterialDomain::SURFACE &&
            !mIsDefaultMaterial &&
@@ -156,13 +156,6 @@ public:
        std::unique_lock<utils::Mutex> lock(mActiveProgramsLock);
        mActivePrograms.set(variant.key);
        lock.unlock();
-
-        if (isSharedVariant(variant)) {
-            FMaterial const* const pDefaultMaterial = mEngine.getDefaultMaterial();
-            if (pDefaultMaterial && pDefaultMaterial->mCachedPrograms[variant.key]) {
-                return pDefaultMaterial->getProgram(variant);
-            }
-        }
 #endif
        assert_invariant(mCachedPrograms[variant.key]);
        return mCachedPrograms[variant.key];
@@ -297,11 +290,6 @@ private:

    void createAndCacheProgram(backend::Program&& p, Variant variant) const noexcept;

-    inline bool isSharedVariant(Variant variant) const {
-        return (mMaterialDomain == MaterialDomain::SURFACE) && !mIsDefaultMaterial &&
-               !mHasCustomDepthShader && Variant::isValidDepthVariant(variant);
-    }
-
    // try to order by frequency of use
    mutable std::array<backend::Handle<backend::HwProgram>, VARIANT_COUNT> mCachedPrograms;
    DescriptorSetLayout mPerViewDescriptorSetLayout;
@@ -963,11 +963,6 @@ void FRenderer::renderJob(RootArenaScope& rootArenaScope, FView& view) {
                            uint32_t(float(xvp.width ) * aoOptions.resolution),
                            uint32_t(float(xvp.height) * aoOptions.resolution)});

-                // this needs to reset the sampler that are only set in RendererUtils::colorPass(), because
-                // this descriptor-set is also used for ssr/picking/structure and these could be stale
-                // it would be better to use a separate desriptor-set for those two cases so that we don't
-                // have to do this
-                view.unbindSamplers(driver);
                view.commitUniformsAndSamplers(driver);
            });

@@ -81,7 +81,6 @@ struct Texture::BuilderDetails {
    Usage mUsage = Usage::NONE;
    bool mHasBlitSrc = false;
    bool mTextureIsSwizzled = false;
-    bool mExternal = false;
    std::array<Swizzle, 4> mSwizzle = {
           Swizzle::CHANNEL_0, Swizzle::CHANNEL_1,
           Swizzle::CHANNEL_2, Swizzle::CHANNEL_3 };
@@ -137,11 +136,6 @@ Texture::Builder& Texture::Builder::import(intptr_t id) noexcept {
    return *this;
 }

-Texture::Builder& Texture::Builder::external() noexcept {
-    mImpl->mExternal = true;
-    return *this;
-}
-
 Texture::Builder& Texture::Builder::swizzle(Swizzle r, Swizzle g, Swizzle b, Swizzle a) noexcept {
    mImpl->mTextureIsSwizzled = true;
    mImpl->mSwizzle = { r, g, b, a };
@@ -160,11 +154,6 @@ Texture* Texture::Builder::build(Engine& engine) {
            (isProtectedTexturesSupported && useProtectedMemory) || !useProtectedMemory)
            << "Texture is PROTECTED but protected textures are not supported";

-    // SAMPLER_EXTERNAL implies imported.
-    if (mImpl->mTarget == SamplerType::SAMPLER_EXTERNAL) {
-        mImpl->mExternal = true;
-    }
-
    uint8_t maxLevelCount;
    switch (mImpl->mTarget) {
        case SamplerType::SAMPLER_2D:
@@ -185,7 +174,7 @@ Texture* Texture::Builder::build(Engine& engine) {
        mImpl->mUsage = TextureUsage::DEFAULT;
        if (mImpl->mLevels > 1 &&
            (mImpl->mWidth > 1 || mImpl->mHeight > 1) &&
-            !mImpl->mExternal) {
+            mImpl->mTarget != SamplerType::SAMPLER_EXTERNAL) {
            const bool formatMipmappable =
                    downcast(engine).getDriverApi().isTextureFormatMipmappable(mImpl->mFormat);
            if (formatMipmappable) {
@@ -254,10 +243,9 @@ FTexture::FTexture(FEngine& engine, const Builder& builder) {
    mSwizzle = builder->mSwizzle;
    mTextureIsSwizzled = builder->mTextureIsSwizzled;
    mHasBlitSrc = builder->mHasBlitSrc;
-    mExternal = builder->mExternal;

    bool const isImported = builder->mImportedId != 0;
-    if (mExternal && !isImported) {
+    if (mTarget == SamplerType::SAMPLER_EXTERNAL && !isImported) {
        // mHandle and mHandleForSampling will be created in setExternalImage()
        // If this Texture is used for sampling before setExternalImage() is called,
        // we'll lazily create a 1x1 placeholder texture.
@@ -330,8 +318,8 @@ void FTexture::setImage(FEngine& engine, size_t level,
            << "level=" << unsigned(level) << " is >= to levelCount=" << unsigned(mLevelCount)
            << ".";

-    FILAMENT_CHECK_PRECONDITION(!mExternal)
-            << "External Texture not supported for this operation.";
+    FILAMENT_CHECK_PRECONDITION(mTarget != SamplerType::SAMPLER_EXTERNAL)
+            << "Texture SamplerType::SAMPLER_EXTERNAL not supported for this operation.";

    FILAMENT_CHECK_PRECONDITION(mSampleCount <= 1) << "Operation not supported with multisample ("
                                                   << unsigned(mSampleCount) << ") texture.";
@@ -466,14 +454,15 @@ void FTexture::setImage(FEngine& engine, size_t level,
 }

 void FTexture::setExternalImage(FEngine& engine, void* image) noexcept {
-    FILAMENT_CHECK_PRECONDITION(mExternal) << "The texture must be external.";
-
+    if (mTarget != Sampler::SAMPLER_EXTERNAL) {
+        return;
+    }
    // The call to setupExternalImage is synchronous, and allows the driver to take ownership of the
    // external image on this thread, if necessary.
    auto& api = engine.getDriverApi();
    api.setupExternalImage(image);

-    auto texture = api.createTextureExternalImage(mTarget, mFormat, mWidth, mHeight, mUsage, image);
+    auto texture = api.createTextureExternalImage(mFormat, mWidth, mHeight, mUsage, image);

    if (mTextureIsSwizzled) {
        auto const& s = mSwizzle;
@@ -486,8 +475,9 @@ void FTexture::setExternalImage(FEngine& engine, void* image) noexcept {
 }

 void FTexture::setExternalImage(FEngine& engine, void* image, size_t plane) noexcept {
-    FILAMENT_CHECK_PRECONDITION(mExternal) << "The texture must be external.";
-
+    if (mTarget != Sampler::SAMPLER_EXTERNAL) {
+        return;
+    }
    // The call to setupExternalImage is synchronous, and allows the driver to take ownership of
    // the external image on this thread, if necessary.
    auto& api = engine.getDriverApi();
@@ -507,7 +497,9 @@ void FTexture::setExternalImage(FEngine& engine, void* image, size_t plane) noex
 }

 void FTexture::setExternalStream(FEngine& engine, FStream* stream) noexcept {
-    FILAMENT_CHECK_PRECONDITION(mExternal) << "The texture must be external.";
+    if (mTarget != Sampler::SAMPLER_EXTERNAL) {
+        return;
+    }

    auto& api = engine.getDriverApi();
    auto texture = api.createTexture(
@@ -532,7 +524,7 @@ void FTexture::setExternalStream(FEngine& engine, FStream* stream) noexcept {
 }

 void FTexture::generateMipmaps(FEngine& engine) const noexcept {
-    FILAMENT_CHECK_PRECONDITION(!mExternal)
+    FILAMENT_CHECK_PRECONDITION(mTarget != SamplerType::SAMPLER_EXTERNAL)
            << "External Textures are not mipmappable.";

    FILAMENT_CHECK_PRECONDITION(mTarget != SamplerType::SAMPLER_3D)
@@ -552,11 +544,12 @@ void FTexture::generateMipmaps(FEngine& engine) const noexcept {
 }

 bool FTexture::textureHandleCanMutate() const noexcept {
-    return (any(mUsage & Usage::SAMPLEABLE) && mLevelCount > 1) || mExternal;
+    return (any(mUsage & Usage::SAMPLEABLE) && mLevelCount > 1) ||
+            mTarget == SamplerType::SAMPLER_EXTERNAL;
 }

 void FTexture::updateLodRange(uint8_t baseLevel, uint8_t levelCount) noexcept {
-    assert_invariant(!mExternal);
+    assert_invariant(mTarget != SamplerType::SAMPLER_EXTERNAL);
    if (any(mUsage & Usage::SAMPLEABLE) && mLevelCount > 1) {
        auto& range = mLodRange;
        uint8_t const last = int8_t(baseLevel + levelCount);
@@ -608,7 +601,7 @@ backend::Handle<backend::HwTexture> FTexture::createPlaceholderTexture(
 }

 backend::Handle<backend::HwTexture> FTexture::getHwHandleForSampling() const noexcept {
-    if (UTILS_UNLIKELY(mExternal && !mHandleForSampling)) {
+    if (UTILS_UNLIKELY(mTarget == SamplerType::SAMPLER_EXTERNAL && !mHandleForSampling)) {
        return setHandleForSampling(createPlaceholderTexture(*mDriver));
    }
    auto const& range = mLodRange;
@@ -617,10 +610,9 @@ backend::Handle<backend::HwTexture> FTexture::getHwHandleForSampling() const noe
    if (UTILS_UNLIKELY(lodRangeChanged)) {
        activeRange = range;
        if (range.empty() || hasAllLods(range)) {
-            std::ignore = setHandleForSampling(mHandle);
+            setHandleForSampling(mHandle);
        } else {
-            std::ignore = setHandleForSampling(mDriver->createTextureView(
-                mHandle, range.first, range.size()));
+            setHandleForSampling(mDriver->createTextureView(mHandle, range.first, range.size()));
        }
    }
    return mHandleForSampling;
@@ -173,8 +173,7 @@ private:
    // Indicates whether the user has set the TextureUsage::BLIT_SRC usage. This will be used to
    // temporarily validate whether this texture can be used for readPixels.
    bool mHasBlitSrc = false;
-    bool mExternal = false;
-    // there is 4 bytes of padding here
+    // there is 5 bytes of padding here

    FStream* mStream = nullptr; // only needed for streaming textures
 };
@@ -912,10 +912,6 @@ void FView::commitUniformsAndSamplers(DriverApi& driver) const noexcept {
    mColorPassDescriptorSet.commit(driver);
 }

-void FView::unbindSamplers(DriverApi& driver) noexcept {
-    mColorPassDescriptorSet.unbindSamplers(driver);
-}
-
 void FView::commitFroxels(DriverApi& driverApi) const noexcept {
    if (mHasDynamicLighting) {
        mFroxelizer.commit(driverApi);
@@ -174,7 +174,6 @@ public:

    void commitFroxels(backend::DriverApi& driverApi) const noexcept;
    void commitUniformsAndSamplers(backend::DriverApi& driver) const noexcept;
-    void unbindSamplers(backend::DriverApi& driver) noexcept;

    utils::JobSystem::Job* getFroxelizerSync() const noexcept { return mFroxelizerSync; }
    void setFroxelizerSync(utils::JobSystem::Job* sync) noexcept { mFroxelizerSync = sync; }
@@ -541,17 +541,6 @@ void ColorPassDescriptorSet::commit(backend::DriverApi& driver) noexcept {
    }
 }

-void ColorPassDescriptorSet::unbindSamplers(DriverApi&) noexcept {
-    // this needs to reset the sampler that are only set in RendererUtils::colorPass(), because
-    // this descriptor-set is also used for ssr/picking/structure and these could be stale
-    // it would be better to use a separate descriptor-set for those two cases so that we don't
-    // have to do this
-    setSampler(+PerViewBindingPoints::STRUCTURE, {}, {});
-    setSampler(+PerViewBindingPoints::SHADOW_MAP, {}, {});
-    setSampler(+PerViewBindingPoints::SSAO, {}, {});
-    setSampler(+PerViewBindingPoints::SSR, {}, {});
-}
-
 void ColorPassDescriptorSet::setSampler(backend::descriptor_binding_t binding,
        TextureHandle th, SamplerParams params) noexcept {
    for (size_t i = 0; i < DESCRIPTOR_LAYOUT_COUNT; i++) {
@@ -158,8 +158,6 @@ public:
    // update local data into GPU UBO
    void commit(backend::DriverApi& driver) noexcept;

-    void unbindSamplers(backend::DriverApi& driver) noexcept;
-
    // bind this UBO
    void bind(backend::DriverApi& driver, uint8_t index) const noexcept {
        mDescriptorSet[index].bind(driver, DescriptorSetBindingPoints::PER_VIEW);
@@ -149,7 +149,7 @@ void DescriptorSet::setSampler(

 DescriptorSet DescriptorSet::duplicate(DescriptorSetLayout const& layout) const noexcept {
    DescriptorSet set{layout};
-    set.mDescriptors = mDescriptors; // Use the vector's assignment operator
+    memcpy(set.mDescriptors.data(), mDescriptors.data(), mDescriptors.size() * sizeof(Desc));
    set.mDirty = mDirty;
    set.mValid = mValid;
    return set;
@@ -127,8 +127,7 @@ float lumaYCoCg(const vec3 c) {
 }

 float luma(const vec3 c) {
-    return (materialConstants_useYCoCg && materialConstants_boxClipping != BOX_CLIPPING_NONE) ?
-            lumaYCoCg(c) : lumaRGB(c);
+    return materialConstants_useYCoCg ? lumaYCoCg(c) : lumaRGB(c);
 }

 vec3 tonemap(const vec3 c) {
@@ -279,6 +278,10 @@ void postProcess(inout PostProcessInputs postProcess) {
        history = textureLod(materialParams_history, uv.zw, 0.0);
    }

+    if (materialConstants_useYCoCg) {
+        history.rgb = RGB_YCoCg(history.rgb);
+    }
+
    highp vec2 size = vec2(textureSize(materialParams_color, 0));
    highp vec2 p = (floor(uv.xy * size) + 0.5) / size;
    vec4 filtered = textureLod(materialParams_color, p, 0.0);
@@ -294,46 +297,47 @@ void postProcess(inout PostProcessInputs postProcess) {
        s[6] = textureLodOffset(materialParams_color, p, 0.0, ivec2(-1,  1)).rgb;
        s[7] = textureLodOffset(materialParams_color, p, 0.0, ivec2( 0,  1)).rgb;
        s[8] = textureLodOffset(materialParams_color, p, 0.0, ivec2( 1,  1)).rgb;
-    }
-
-    int j = 0;
-    float confidence = 1.0;
-    if (materialConstants_upscaling) {
-        highp vec2 subPixelOffset = (p - uv.xy) * size;  // +/- [0.25, 0.25]
-
-        // we reduce the contribution of a sample based on the distance
-        // to the high resolution pixel center
-        const float cutoff = 0.5;
-        highp float l = length(materialParams.jitter - subPixelOffset) / cutoff;
-        confidence = saturate(1.0 - l * l);
-
-        if (materialConstants_filterInput) {
-            int jxp = subPixelOffset.y > 0.0 ? 1 : 2;
-            int jxn = subPixelOffset.y > 0.0 ? 0 : 3;
-            j = subPixelOffset.x > 0.0 ? jxp : jxn;
-        }
-    }
-
-    if (materialConstants_filterInput) {
-        // unjitter/filter input
-        filtered = vec4(vec3(0), filtered.a);
-        for (int i = 0; i < 9; i++) {
-            float w = materialParams.filterWeights[i][j];
-            filtered.rgb += s[i] * w;
-        }
-        filtered.rgb = max(filtered.rgb, vec3(0));
-    }
-
-    // build the history clamping box
-    if (materialConstants_boxClipping != BOX_CLIPPING_NONE) {
        if (materialConstants_useYCoCg) {
-            history.rgb = RGB_YCoCg(history.rgb);
-            filtered.rgb = RGB_YCoCg(filtered.rgb);
            for (int i = 0; i < 9; i++) {
                s[i] = RGB_YCoCg(s[i]);
            }
        }
+    }

+    vec2 subPixelOffset = p - uv.xy;  // +/- [0.25, 0.25]
+    float confidence = materialConstants_upscaling ? 0.0 : 1.0;
+
+    if (materialConstants_filterInput) {
+        // unjitter/filter input
+        // figure out which set of coeficients to use
+        filtered = vec4(0, 0, 0, filtered.a);
+        if (materialConstants_upscaling) {
+            int jxp = subPixelOffset.y > 0.0 ? 3 : 0;
+            int jxn = subPixelOffset.y > 0.0 ? 2 : 1;
+            int j   = subPixelOffset.x > 0.0 ? jxp : jxn;
+            for (int i = 0; i < 9; i++) {
+                float w = materialParams.filterWeights[i][j];
+                filtered.rgb += s[i] * w;
+                confidence = max(confidence, w);
+            }
+        } else {
+            for (int i = 0; i < 9; i++) {
+                float w = materialParams.filterWeights[i][0];
+                filtered.rgb += s[i] * w;
+            }
+        }
+    } else {
+        if (materialConstants_useYCoCg) {
+            filtered.rgb = RGB_YCoCg(filtered.rgb);
+        }
+        if (materialConstants_upscaling) {
+            confidence = float(materialParams.jitter.x * subPixelOffset.x > 0.0 &&
+                               materialParams.jitter.y * subPixelOffset.y > 0.0);
+        }
+    }
+
+    // build the history clamping box
+    if (materialConstants_boxClipping != BOX_CLIPPING_NONE) {
        vec3 boxmin;
        vec3 boxmax;
        if (materialConstants_boxType == BOX_TYPE_AABB ||
@@ -342,7 +346,7 @@ void postProcess(inout PostProcessInputs postProcess) {
            boxmax = max(s[4], max(max(s[1], s[3]), max(s[5], s[7])));
            vec3 box9min = min(boxmin, min(min(s[0], s[2]), min(s[6], s[8])));
            vec3 box9max = max(boxmax, max(max(s[0], s[2]), max(s[6], s[8])));
-            // round the corners of the 3x3 box, giving less importance to the corner samples
+            // round the corners of the 3x3 box
            boxmin = (boxmin + box9min) * 0.5;
            boxmax = (boxmax + box9max) * 0.5;
        }
@@ -384,11 +388,9 @@ void postProcess(inout PostProcessInputs postProcess) {
    }

    // go back to RGB space before tonemapping
-    if (materialConstants_boxClipping != BOX_CLIPPING_NONE) {
-        if (materialConstants_useYCoCg) {
-            filtered.rgb = YCoCg_RGB(filtered.rgb);
-            history.rgb = YCoCg_RGB(history.rgb);
-        }
+    if (materialConstants_useYCoCg) {
+        filtered.rgb = YCoCg_RGB(filtered.rgb);
+        history.rgb = YCoCg_RGB(history.rgb);
    }

    // tonemap before mixing
@@ -1,12 +1,12 @@
 Pod::Spec.new do |spec|
  spec.name = "Filament"
-  spec.version = "1.56.6"
+  spec.version = "1.56.1"
  spec.license = { :type => "Apache 2.0", :file => "LICENSE" }
  spec.homepage = "https://google.github.io/filament"
  spec.authors = "Google LLC."
  spec.summary = "Filament is a real-time physically based rendering engine for Android, iOS, Windows, Linux, macOS, and WASM/WebGL."
  spec.platform = :ios, "11.0"
-  spec.source = { :http => "https://github.com/google/filament/releases/download/v1.56.6/filament-v1.56.6-ios.tgz" }
+  spec.source = { :http => "https://github.com/google/filament/releases/download/v1.56.1/filament-v1.56.1-ios.tgz" }

  # Fix linking error with Xcode 12; we do not yet support the simulator on Apple silicon.
  spec.pod_target_xcconfig = {
@@ -38,13 +38,12 @@ bool initBinder() {
        }
    }
    if (!g_driver.library) {
-        // The library has been linked explicitly during compile.
-        g_driver.OSMesaGetProcAddress = (ProcAddressFunc) dlsym(RTLD_LOCAL, "OSMesaGetProcAddress");
-    } else {
-        g_driver.OSMesaGetProcAddress =
-                (ProcAddressFunc) dlsym(g_driver.library, "OSMesaGetProcAddress");
+        return false;
    }

+    g_driver.OSMesaGetProcAddress = (ProcAddressFunc)
+            dlsym(g_driver.library, "OSMesaGetProcAddress");
+
    return g_driver.OSMesaGetProcAddress;
 }

@@ -53,9 +52,7 @@ void* loadFunction(const char* name) {
 }

 void shutdownBinder() {
-    if (g_driver.library) {
-        dlclose(g_driver.library);
-    }
+    dlclose(g_driver.library);
    memset(&g_driver, 0, sizeof(g_driver));
 }

@@ -40,7 +40,7 @@

    #if defined(__ANDROID__)
    #define VK_USE_PLATFORM_ANDROID_KHR 1
-    #elif defined(FILAMENT_IOS)
+    #elif defined(IOS)
    #define VK_USE_PLATFORM_IOS_MVK 1
    #elif defined(__linux__)
    #if defined(FILAMENT_SUPPORTS_XCB)
@@ -1,71 +0,0 @@
-cmake_minimum_required(VERSION 3.19)
-project(fgviewer C ASM)
-
-set(TARGET fgviewer)
-set(PUBLIC_HDR_DIR include)
-
-if (CMAKE_CROSSCOMPILING)
-    include(${IMPORT_EXECUTABLES})
-endif()
-
-# ==================================================================================================
-# Sources and headers
-# ==================================================================================================
-
-set(PUBLIC_HDRS
-    include/fgviewer/DebugServer.h
-    include/fgviewer/JsonWriter.h
-)
-
-set(SRCS
-    src/ApiHandler.cpp
-    src/ApiHandler.h
-    src/DebugServer.cpp
-)
-
-# ==================================================================================================
-# Include and target definitions
-# ==================================================================================================
-
-include_directories(${PUBLIC_HDR_DIR})
-
-add_library(${TARGET} STATIC ${PUBLIC_HDRS} ${SRCS})
-
-target_link_libraries(${TARGET} PUBLIC
-        civetweb
-        utils
-)
-
-target_include_directories(${TARGET} PRIVATE ${filamat_SOURCE_DIR}/src)
-
-target_include_directories(${TARGET} PUBLIC ${PUBLIC_HDR_DIR})
-set_target_properties(${TARGET} PROPERTIES FOLDER Libs)
-
-# ==================================================================================================
-# Compiler flags
-# ==================================================================================================
-
-if (MSVC)
-else()
-    target_compile_options(${TARGET} PRIVATE -Wno-deprecated-register)
-endif()
-
-# ==================================================================================================
-# Installation
-# ==================================================================================================
-
-# matdbg has dependencies on non-installed libraries. Here we bundle them all together into a single
-# library that gets copied into the installation folder so users are only required to link against
-# matdbg.
-set(FGVIEWER_DEPS
-    fgviewer
-    civetweb
-    )
-
-set(FGVIEWER_COMBINED_OUTPUT "${CMAKE_CURRENT_BINARY_DIR}/libfgviewer_combined.a")
-combine_static_libs(fgviewer "${FGVIEWER_COMBINED_OUTPUT}" "${FGVIEWER_DEPS}")
-
-set(FGVIEWER_LIB_NAME ${CMAKE_STATIC_LIBRARY_PREFIX}fgviewer${CMAKE_STATIC_LIBRARY_SUFFIX})
-install(FILES "${FGVIEWER_COMBINED_OUTPUT}" DESTINATION lib/${DIST_DIR} RENAME ${FGVIEWER_LIB_NAME})
-# We do not need fgviewer headers in the install directory
-# install(DIRECTORY ${PUBLIC_HDR_DIR}/fgviewer DESTINATION include)
@@ -1,88 +0,0 @@
-/*
- * Copyright (C) 2024 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 FGVIEWER_DEBUGSERVER_H
-#define FGVIEWER_DEBUGSERVER_H
-
-#include <utils/CString.h>
-#include <utils/Mutex.h>
-
-#include <unordered_map>
-#include <vector>
-
-class CivetServer;
-
-namespace filament::fgviewer {
-
-using FrameGraphInfoKey = uint32_t;
-
-struct FrameGraphPassInfo {
-    utils::CString pass_name;
-    // TODO: Add struct detail properties
-};
-
-struct FrameGraphInfo {
-    utils::CString view_name;
-    std::vector<FrameGraphPassInfo> passes;
-};
-
-/**
- * Server-side frame graph debugger.
- *
- * This class manages an HTTP server. It receives frame graph packages from the Filament C++ engine or
- * from a standalone tool such as fginfo.
- */
-class DebugServer {
-public:
-    static std::string_view const kSuccessHeader;
-    static std::string_view const kErrorHeader;
-
-    DebugServer(int port);
-    ~DebugServer();
-
-    /**
-     * Notifies the debugger that a new view has been added.
-     */
-    void addView(const utils::CString& name, FrameGraphInfo info);
-
-    /**
-     * Notifies the debugger that the given view has been deleted.
-     */
-    void removeView(const utils::CString& name);
-
-    /**
-     * Updates the information for a given view.
-     */
-    void updateView(const utils::CString& name, FrameGraphInfo info);
-
-    bool isReady() const { return mServer; }
-
-private:
-    CivetServer* mServer;
-
-    std::unordered_map<FrameGraphInfoKey, FrameGraphInfo> mViews;
-    mutable utils::Mutex mViewsMutex;
-
-    class FileRequestHandler* mFileHandler = nullptr;
-    class ApiHandler* mApiHandler = nullptr;
-
-    friend class FileRequestHandler;
-    friend class ApiHandler;
-};
-
-} // namespace filament::fgviewer
-
-#endif  // FGVIEWER_DEBUGSERVER_H
@@ -1,45 +0,0 @@
-/*
- * Copyright (C) 2024 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 FGVIEWER_JSONWRITER_H
-#define FGVIEWER_JSONWRITER_H
-
-#include <utils/CString.h>
-
-namespace filament::fgviewer {
-
-struct FrameGraphInfo;
-
-// This class generates portions of JSON messages that are sent to the web client.
-// Note that some portions of these JSON strings are generated by directly in DebugServer,
-// as well as CommonWriter.
-class JsonWriter {
-public:
-
-    // Retrieves the most recently generated string.
-    const char* getJsonString() const;
-    size_t getJsonSize() const;
-
-    // Generates a JSON string describing the given FrameGraphInfo.
-    bool writeFrameGraphInfo(const FrameGraphInfo& frameGraph);
-
-private:
-    utils::CString mJsonString;
-};
-
-} // namespace filament::fgviewer
-
-#endif  // FGVIEWER_JSONWRITER_H
@@ -1,69 +0,0 @@
-/*
- * Copyright (C) 2024 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 "ApiHandler.h"
-
-#include <fgviewer/DebugServer.h>
-
-#include <utils/FixedCapacityVector.h>
-#include <utils/Log.h>
-
-#include <CivetServer.h>
-
-namespace filament::fgviewer {
-
-using namespace std::chrono_literals;
-
-namespace {
-
-auto const& kSuccessHeader = DebugServer::kSuccessHeader;
-auto const& kErrorHeader = DebugServer::kErrorHeader;
-
-auto const error = [](int line, std::string const& uri) {
-    utils::slog.e << "[fgviewer] DebugServer: 404 at line " << line << ": " << uri << utils::io::endl;
-    return false;
-};
-
-} // anonymous
-
-bool ApiHandler::handleGetApiFgInfo(struct mg_connection* conn,
-        struct mg_request_info const* request) {
-    auto const softError = [conn, request](char const* msg) {
-        utils::slog.e << "[fgviewer] DebugServer: " << msg << ": " << request->query_string << utils::io::endl;
-        mg_printf(conn, kErrorHeader.data(), "application/txt");
-        mg_write(conn, msg, strlen(msg));
-        return true;
-    };
-
-    // TODO: Implement the method
-    return true;
-}
-
-void ApiHandler::addFrameGraph(FrameGraphInfo const* framegraph) {
-    // TODO: Implement the method
-}
-
-
-bool ApiHandler::handleGet(CivetServer* server, struct mg_connection* conn) {
-    struct mg_request_info const* request = mg_get_request_info(conn);
-    std::string const& uri = request->local_uri;
-
-    // TODO: Implement the method
-    return true;
-}
-
-} // filament::fgviewer
@@ -1,49 +0,0 @@
-/*
- * Copyright (C) 2024 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 FGVIEWER_APIHANDLER_H
-#define FGVIEWER_APIHANDLER_H
-
-#include <CivetServer.h>
-
-namespace filament::fgviewer {
-
-class DebugServer;
-struct FrameGraphInfo;
-
-// Handles the following REST requests, where {id} is an 8-digit hex string.
-//
-//    GET /api/framegraphs
-//    GET /api/framegraph?fg={viewname}
-//
-class ApiHandler : public CivetHandler {
-public:
-    explicit ApiHandler(DebugServer* server)
-        : mServer(server) {}
-    ~ApiHandler() = default;
-
-    bool handleGet(CivetServer* server, struct mg_connection* conn);
-    void addFrameGraph(FrameGraphInfo const* frameGraph);
-
-private:
-    bool handleGetApiFgInfo(struct mg_connection* conn, struct mg_request_info const* request);
-
-    DebugServer* mServer;
-};
-
-} // filament::fgviewer
-
-#endif // FGVIEWER_APIHANDLER_H
@@ -1,131 +0,0 @@
-/*
- * Copyright (C) 2024 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 <fgviewer/DebugServer.h>
-
-#include "ApiHandler.h"
-
-#include <CivetServer.h>
-
-#include <utils/FixedCapacityVector.h>
-#include <utils/Hash.h>
-#include <utils/Log.h>
-
-#include <string>
-#include <string_view>
-
-namespace filament::fgviewer {
-
-namespace {
-std::string const BASE_URL = "libs/fgviewer/web";
-
-FrameGraphInfoKey getKeybyString(const utils::CString &input,
-                                              uint32_t seed) {
-    return utils::hash::murmurSlow(reinterpret_cast<uint8_t const*>(
-        input.c_str()), input.size(), 0);
-}
-} // anonymous
-
-using namespace utils;
-
-std::string_view const DebugServer::kSuccessHeader =
-        "HTTP/1.1 200 OK\r\nContent-Type: %s\r\n"
-        "Connection: close\r\n\r\n";
-
-std::string_view const DebugServer::kErrorHeader =
-        "HTTP/1.1 404 Not Found\r\nContent-Type: %s\r\n"
-        "Connection: close\r\n\r\n";
-
-class FileRequestHandler : public CivetHandler {
-public:
-    FileRequestHandler(DebugServer* server) : mServer(server) {}
-    bool handleGet(CivetServer *server, struct mg_connection *conn) {
-        auto const& kSuccessHeader = DebugServer::kSuccessHeader;
-        struct mg_request_info const* request = mg_get_request_info(conn);
-        std::string uri(request->request_uri);
-        if (uri == "/") {
-            uri = "/index.html";
-        }
-
-        if (uri == "/index.html" || uri == "/app.js" || uri == "/api.js") {
-            mg_send_file(conn, (BASE_URL + uri).c_str());
-            return true;
-        }
-        slog.e << "[fgviewer] DebugServer: bad request at line " <<  __LINE__ << ": " << uri << io::endl;
-        return false;
-    }
-private:
-    DebugServer* mServer;
-};
-
-DebugServer::DebugServer(int port) {
-    // By default the server spawns 50 threads so we override this to 10. According to the civetweb
-    // documentation, "it is recommended to use num_threads of at least 5, since browsers often
-    /// establish multiple connections to load a single web page, including all linked documents
-    // (CSS, JavaScript, images, ...)."  If this count is too small, the web app basically hangs.
-    const char* kServerOptions[] = {
-        "listening_ports", "8085",
-        "num_threads", "10",
-        "error_log_file", "civetweb.txt",
-        nullptr
-    };
-    std::string portString = std::to_string(port);
-    kServerOptions[1] = portString.c_str();
-
-    mServer = new CivetServer(kServerOptions);
-    if (!mServer->getContext()) {
-        delete mServer;
-        mServer = nullptr;
-        slog.e << "[fgviewer] Unable to start DebugServer, see civetweb.txt for details." << io::endl;
-        return;
-    }
-
-    mFileHandler = new FileRequestHandler(this);
-    mApiHandler = new ApiHandler(this);
-
-    mServer->addHandler("/api", mApiHandler);
-    mServer->addHandler("", mFileHandler);
-
-    slog.i << "[fgviewer] DebugServer listening at http://localhost:" << port << io::endl;
-}
-
-DebugServer::~DebugServer() {
-    mServer->close();
-
-    delete mFileHandler;
-    delete mApiHandler;
-    delete mServer;
-}
-
-void DebugServer::addView(const utils::CString &name, FrameGraphInfo info) {
-    std::unique_lock<utils::Mutex> lock(mViewsMutex);
-    const FrameGraphInfoKey key = getKeybyString(name, 0);
-    mViews.insert({key, info});
-}
-
-void DebugServer::removeView(const utils::CString& name) {
-    std::unique_lock<utils::Mutex> lock(mViewsMutex);
-    const FrameGraphInfoKey key = getKeybyString(name, 0);
-    mViews.erase(key);
-}
-
-void DebugServer::updateView(const utils::CString& name, FrameGraphInfo info) {
-    std::unique_lock<utils::Mutex> lock(mViewsMutex);
-    const FrameGraphInfoKey key = getKeybyString(name, 0);
-    mViews[key] = info;
-}
-
-} // namespace filament::fgviewer
@@ -81,9 +81,7 @@ enum class ReservedSpecializationConstants : uint8_t {
    CONFIG_DEBUG_DIRECTIONAL_SHADOWMAP = 6,
    CONFIG_DEBUG_FROXEL_VISUALIZATION = 7,
    CONFIG_STEREO_EYE_COUNT = 8, // don't change (hardcoded in ShaderCompilerService.cpp)
-    CONFIG_SH_BANDS_COUNT = 9,
-    CONFIG_SHADOW_SAMPLING_METHOD = 10
-    // check CONFIG_MAX_RESERVED_SPEC_CONSTANTS below
+    CONFIG_SH_BANDS_COUNT = 9
 };

 enum class PushConstantIds : uint8_t  {
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Mathias Agopian	107f25ca75	random rays	2024-11-22 21:25:22 -08:00
Mathias Agopian	46806735e4	add command line options	2024-11-22 21:25:22 -08:00
Mathias Agopian	c410acb69f	make a formal api	2024-11-22 21:25:22 -08:00
Mathias Agopian	cb9eba3b3e	more	2024-11-22 21:25:22 -08:00
Mathias Agopian	dcf51ba20e	more	2024-11-22 21:25:22 -08:00
Mathias Agopian	73e6ccad96	secondary wip	2024-11-22 21:25:22 -08:00
Mathias Agopian	ad5c409317	more better	2024-11-22 21:25:22 -08:00
Mathias Agopian	db597af272	raytracing	2024-11-22 21:25:22 -08:00
Mathias Agopian	4a46955bfb	more	2024-11-22 21:25:22 -08:00
Mathias Agopian	864bc3c1c1	rainbowgen	2024-11-22 21:25:22 -08:00
Mathias Agopian	2da395caf8	rainbowgen	2024-11-22 21:25:22 -08:00